Effects on Populations of Exposure to Low Levels of Ionizing Radiation (1972)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

CHAPTER VII SOMATIC EFFECTS OF IONIZING RADIATION Summary and Conclusions I: Introduction 85 II. Some Principles Underlying Induction of Somatic Effects 85 III. Cancer Incidence and Radiation Dose 86 FV. Probability of Cancer Induction at Low Doses and Low Dose Rates 87 V. Relative Biological Effectiveness 88 VI. The Linear Hypothesis 89 VII. Risk Estimation 89 VIII. Summary 91 References 91 Appendices to Chapter VII I. Review of Scientific Bases of Evaluation of Risks of Low- Level Radiation 93 A. Types of Effects 93 B. Evidence of the Causal Nature of Associations 93 C. Dose-Effect Models 95 D. Ways of Expressing Risk 99 II. Estimates of Risk from Human and Animal Data 100 A. Cancers 100 1. Exposure during Childhood or Adult Life 100 a. Leukemia 100 b. Radiation-induced Thyroid Cancer 120 c. Radiation-induced Bone Cancer 125 d. Skin 132 e. Breast 136 f. Lung 145 g. Other Neoplasms of Specific Types 157 h. All Cancers Other than Leukemia 160 2. Cancer following Irradiation before Conception or dur- ing Intrauterine Life 160 3. Total Cancer Risk 167 B. Mortality from Causes other than Cancer 174 1. Adult 174 2. Infant Mortality and Ionizing Radiation 177 C. Morbidity from Causes other than Cancer 179 1. Cataracts 179 2. Central Nervous System Effects 179 3. Impairment of Fertility 179 83

III. Analysis of Viewpoints on Record 181 IV. Calculation of Confidence Limits for Risk Estimation 189 V. Radiation Dosimetry of Heavily Irradiated Sites in Patients Treated for Ankylosing Spondylitis 190 VI. Definitions and Notes to Accompany Reference Tables Summarizing Quantitative Data on Carcinogenic Effects of Ionizing Radiation 195 84

CHAPTER VII SOMATIC EFFECTS OF IONIZING RADIATION Summary and Conclusions I. Introduction Consideration has been given in this portion of the report to those effects of ionizing radia- tion that are manifested in exposed individuals themselves (i.e., somatic effects) as contrasted to effects that are manifested in subsequent generations (i.e., genetic, or inherited, effects). Effects of radiation on prenatal and juvenile development are dealt with in Chapter VI. In reviewing existing knowledge of the ef- fects of interest, the Subcommittee has had access to: (1) previous evaluations by other committees of the National Academy of Sci- ences of the United States, the National Coun- cil on Radiation Protection and Measurements, the International Commission on Radiological Protection, the United Nations Scientific Com- mittee on the Effects of Atomic Radiation, and other panels of experts,i (2) published reports available in the world scientific literature; (3) technical documents, including hitherto una- vailable data, provided by the Atomic Bomb Casualty Commission, the United States Public Health Service, The Atomic Energy Commis- sion, and other agencies; and (4) private com- munications from individual scientists, some of which were supplied in response to a wide- spread appeal to the scientific community for information on the effects of low-level radia- tion. In general, the Committee has not considered acute effects of irradiation, since these are al- ready well documented and occur only at dose levels well above those of interest in the setting of protection standards. With few exceptions, the somatic effects we have considered mani- fest themselves only after an interval of years or decades following irradiation and are indis- Federal Radiation Council, 1964; Unitcil Vsitions ---••ieii- tific Committee on the Effects of Atomic R:" ;ition. 1962, 1964, 1966, 1969; International Commission o;i Radiologi- cal Protection, 1966, 1969; National Council on Radiologi- cal Protection and Measurements, 1971 (see references). tinguishable from lesions that occur naturally in nonirradiated populations; thus, their rela- tionship to radiation is detectable only in a statistical sensed Thus in any given individual a particular effect cannot be attributed conclu- sively to radiation, as opposed to some other cause, and the smaller the dose of radiation, the less the likelihood that radiation was in fact a prime cause. II. Some Principles Underlying Induction of Somatic Effects To specify numerically the risks of radiation effects under conditions of low-level exposure requires better knowledge than is now availa- ble of the mechanisms involved in the produc- tion of such effects, of their dose-response rela- tionships, and of the susceptibility of human populations at risk. For none of the effects can the dose-response relationship be defined pre- cisely over a wide range of dose and dose rate, and only for induction of certain types of ef- fects, such as cataracts and impairment of fer- tility, are the mechanisms of induction known with some precision. For the most part, there- fore, estimation of the risks of effects at low dose levels involves extrapolation from obser- vations at higher dose levels, based on assump- tions about the nature of the dose-response relationship, the mechanisms involved, the susceptibility of the population at risk, and other factors.3 For induction of cataract of the lens and impairment of fertility there is radiobiological and clinical evidence of a nonlinear relation- ship between effect and dose, these effects pre- sumably depending in large measure on the kill- ing of cells in the lens and gonads, respective- ly.4 These considerations imply that there is 2See Appendix I A 3See Appendix I B See Appendix II C 1 (cataracts) and IIC 3 (impairment of fertility). 85

little or no risk of inducing such effects at doses and dose rates approaching natural back- ground radiation levels. For induction of some tumors, on the other hand, a linear non-thresh- old dose-effect relationship cannot be excluded from existing data, nor can the possibility that such effects might result from sublethal injury in only one or a few cells of the body.5 Moreo- ver, evidence available at this time indicates that the most important effect of radiation on the mortality of human populations results from carcinogenesis, including leukemogene- sis.6 Whether in human populations there are radiation effects on mortality from causes oth- er than cancer is still uncertain.1? In regard to the induction of cancer by ioniz- ing radiation, the following observations are pertinent: (1) the cancers induced by radiation are indistinguishable individually from those occurring naturally, and hence their existence can be inferred only in terms of an excess above the natural incidence; (2) the natural incidence of cancer varies over several orders of magni- tude, depending on the type and site of the neo- plastic growth, age, sex, and other factors; (3) cancer of any one type occurs with sufficiently low incidence in man that few irradiated popu- lations are large enough to provide convincing quantitative data on the incidence of tumors of any one type or site; (4) the time elapsing between irradiation and the appearance of a clinically detectable neoplasm is characteristi- cally long; i.e., years or even decades; (5) this long induction time complicates the prospec- tive follow-up of irradiated populations for observation of possible tumor development, and it also complicates the retrospective evalu- ation of cancer patients for possible history of relevant radiation exposure; (6) many of the existing human and animal data on radiation- induced tumors come from populations exposed to internally deposited radionuclides, in which the dose-incidence relation is obscured by marked nonuniformities in the temporal and spatial distribution of radiation; (7) in several other instances, data have been derived from studies of therapeutically irradiated patients, in whom the effects of radiation may have been complicated by effects of the underlying dis- ease itself or of treatments other than radia- tion; and (8) some of the data concern mortality whereas others concern disease incidence; in the case of cancer it is relevant to differentiate those radiation-induced malignancies that do not greatly alter the death rate (e.g., thyroid carcinoma) from others that in the present state of knowledge are generally fatal (e.g., leukemia). III. Cancer Incidence and Radiation Dose Despite the difficulties mentioned, a clear-cut increase in incidence with increasing radiation dose has been documented for several types of cancer in human populations, as well as for many types of neoplasms in experimental ani- mals. Although, with few exceptions, the ob- served dose-incidence data pertain to relatively high doses (above 50 rem8 and high dose rates (above 1 rem per minute), he findings for any given neoplasm are reasont b y consistent from one irradiated human population to another, suggesting that they may be applicable within limits to the general population for purposes of risk evaluation.9 In atomic bomb survivors of Hiroshima and Nagasaki, and in British patients treated with intensive spinal irradiation for ankylosing spondylitis, an increased incidence of all forms of leukemia except the chronic lymphocytic type has been observed. The relationship be- tween the excess in incidence and the radiation dose, as observed at relatively high doses and high dose rates, is consistent with a linear dose-incidence function. For purposes of esti- mating risk at low doses, it is necessary to ex- trapolate this linear relationship through the origin at zero dose, on the assumption that the incidence at zero dose is a point on the curve. 5See Appendix IC 6See Appendix IIA ?See Appendix II B The dose equivalent in rem in strict usage is the product of the absorbed dose in rads and an assigned quality factor, and other necessary modifying factors, and is reserved for use in radiation protection. However, the term "rem" has also been used in the radiological literature to indicate the product of the absorbed dose in rads and RBE, and will sometimes be so used in this report for simplicity. This use of rem is not in strict conformity with definitions of the International Commission on Radiological Units (ICRU). Also, the term "dose" is used broadly to apply not only to the absorbed dose in rads, but to the dose equivalent in rem as well. "See Appendix II A 3 86

The slope of the fitted straight-line corre- sponds to a risk of about 1 case of leukemia per 106 exposed persons, per year, per rem.io The excess in incidence, which was evident within 3- 4 years after irradiation, declined within 15 years but still persists at a diminished level in atomic bomb survivors, now 25 years after exposure. Data for other irradiated popula- tions are less quantitative but imply, for high doses, a comparable excess of leukemia per unit of average dose to the bone marrow, despite wide differences in the conditions of exposure. The evidence suggests that susceptibility to induction of leukemia is several times higher in those irradiated in utero or during childhood.il as well as in those irradiated late in adult life.io than in individuals of intermediate ages. Tumors of the thyroid gland also have been found to show a systematic increase in inci- dence with increasing dose in irradiated popula- tions.12 The dose-effect relationship as ob- served at relatively high doses and high dose rates, like that for leukemia, can be represent- ed by a linear, non-threshold function, corre- sponding to a risk of 2.5-9.3 cases (not deaths) of cancer per 106 exposed children, per year, per rem to the thyroid gland, averaged over the fifth to twenty-fifth years after exposure. Sus- ceptibility to induction of these tumors seems to be several times higher in children than in adults. For tumors of other types and sites, the ex- isting dose-response data are more limited, and the estimates of risk correspondingly less reli- able. For cancer of the Iung,i3 the mortality at high doses has been observed to approximate one death per 106 exposed persons per year, per rem. For cancer of the breast, 14 the mortality at high doses has been observed to approxi- mate three deaths per 106 exposed women per year, per rem. For cancer of the skeleton.is the mortality at high doses has been observed to approximate two deaths per 10? exposed per- sons per year, per rem. For cancer of the GI tract including the stomach,i6 the mortality at high doses has been observed to approximate one death per 106 persons per year, per rem. "See Appendix II A 1 a "See Appendix II A 2 i-See Appendix II A 1 b l:tSee Appendix II A 1 f uSee Appendix II A 1 e lftSee Appendix II A 1 c i6See Appendix IIAlg Cancer at other sitesi' may contribute a fur- ther one death per 106 persons per year, per rem. These rates have been derived from the period after irradiation during which an excess in the incidence of these tumors has been evi- dent. Although cancers of other types have been observed to occur in heavily irradiated tissues (for example, cancer of the skin), there are no quantitative dose-incidence data for such growths comparable to those cited above. The findings imply either that susceptibility to such malignancies is low by comparison with suscep- tibility to the specific types of cancer men- tioned earlierie or that the distribution of la- tent periods for such malignancies extends well beyond the upper limit of 25 years of follow-up achieved for the major long-term studies thus far. In fact, the overall excess mortality from cancer, including leukemia, in irradiated popu- lations can be accounted for largely by the spe- cific types of tumors mentioned above. In the Hiroshima and Nagasaki population, this ex- cess at high doses and dose rates amounts to about 2.5 deaths/106/rem/year, averaged over the periodi8 in which the excess was observed.!* Some studies suggest that after prenatal irra- diation the overall juvenile cancer mortality may be increased by about 50 cases/ 106/rem/year, averaged over the first 10 years of life; however, the possibility remains that the excess observed in these studies may be dependent on factors other than radiation.il The observed variations in susceptibility to induction of different types of cancer by irra- diation, which are apparently unrelated to the marked variations in the natural incidence of the diverse types, make it clear that the con- cept of a uniform doubling dose of radiation for induction of all types of cancer is invalid. IV. Probability of Cancer Induction at Low Doses and Low Dose Rates The dose-mortality figures cited above, which pertain chiefly to human populations exposed at high doses and high dose rates, may be used i-See Appendix II A3 lsFollow-up observations on the survivors of the A- bombs dropped on Hiroshima and Nagasaki in 1945 gener- ally began 1 October 1950. Thus, data through 1970 are for the 20-year period five-to-25 years, not O-to-25 or O-to-20. Conversion to a rem dose has been done, in this instance, on the assumption that the RBE for neutrons is 5. i9See Appendix III. 87

to estimate the probability of cancer in other populations exposed at lower doses and lower dose rates if it is assumed that the relationship between mortality and dose remains the same irrespective of changes in dose, dose rate, and population at risk. There are cogent radiobio- logical reasons for doubting that the dose-in- cidence relationship for cancer induction in man does, in fact, remain constant in the face of such changes, one of which is the wide- spread occurrence of repair of most other types of radiation injury at low doses and low dose rates, particularly in the case of low-LET ra- diations, i? The dose rate characteristic for background radiation (approximately 0.1 rem/ year) is one-hundred-million to one-billion times lower than the dose rate at which effects have been observed in most irradiated study populations. At background radiation levels, ionizing events in individual mammalian cell nuclei occur at a rate of much less than one per day, whereas at the higher dose rates men- tioned, ionization events occur in cells at a fre- quency of the order of 2600 per second. This enormous difference may have important impli- cations with respect to the production of radia- tion damage within cells and its repair at the molecular level. On the basis of the likelihood of such repair, the risk of cancer induction at low doses and low dose rates might be expected to be appreciably smaller per unit dose than at high doses and high dose rates, as has been observed to be the case in certain radiation- induced tumors of experimental animals. 5,10,19 Hence, expectations based on linear extrapola- tion from the known effects in man of larger doses delivered at high dose rates in the range of rising dose-incidence relationship may well overestimate the risks of low-LET radiation at low dose rates and may, therefore, be regarded as upper limits of risk for low-level low-LET irradiation. The lower limit, depending on the shape of the dose-incidence curve for low-LET radiation and the efficiency of repair processes in counteracting carcinogenic effects, could be appreciably smaller (the possibility of zero is not excluded by the data). On the other hand, because there is greater killing of susceptible cells at high doses and high dose rates, extrapo- lation based on effects observed under these exposure conditions may be postulated to un- derestimate the risks of irradiation at low dos- es and low dose rates. V. Relative Biological Effectiveness Another factor complicating extrapolation from the available human data is wide varia- tion in relative biological effectiveness among different types of ionizing radiations. This var- iation, which depends on differences in the mi- crodistribution of radiation energy, or linear energy transfer (LET), may cause the same total dose to differ in its effects by a factor of 10 or more, depending on the radiation in ques- tion. This problem pertains directly to the in- terpretation of data from several of the princi- pal available sources; namely, atomic bomb survivors of Hiroshima, underground miners exposed to radon gas and its radioactive decay products, and a number of populations with high body burdens of alpha-emitting radionu- clides. In the case of Hiroshima, the numbers of survivors are larger, and the statistics corre- spondingly better, than in the case of Naga- saki; but since the radiations at Hiroshima in- cluded an appreciable component of fast neu- trons, it is necessary to estimate the relative biological effectiveness (RBE) of this compo- nent in order that dose-effect data for the two cities can be appropriately compared. The best estimate of the RBE, at high doses and dose rates, derived from intercomparison of the Hiroshima and Nagasaki data for leukemia, is between 1 and 5,10 a range of values which is consistent with findings in experimental ani- mals. The value of the RBE, which denotes the ratio between the doses of high-LET and of low- LET radiations for equivalent effects, rises with increase in the spatial concentration of the radiation energy delivered during a given exposure, i.e., with increasing LET. Also, many radiobiological data indicate that the risk-per-rad of low-LET radiations, such as x rays and gamma rays, decreases to a greater degree with decrease in the dose and dose rate than does the effectiveness of high- LET radiations, which may decrease little if at all. Hence the RBE of high-LET radiations can be expected to increase with decrease in the dose and dose rate. The RBE value of 1-5 for leukemia induction, cited above, may thus be considerably smaller than the RBE value appli- cable to low doses and dose rates. In this re- port, values of 1 and 5 have been used for the RBE of neutrons in the Hiroshima experience for the purpose of calculating risk per rem. The 88

data available on human populations exposed to alpha emitters (underground miners, thoro- trast-or radium-treated patients, and radium dial painters) indicate that for cancer produc- tion alpha particles delivered at relatively low dose rates are 5-10 times more effective per rad average tissue dose than x rays or gamma rays delivered at high dose rates. VI. The Linear Hypothesis Although experimental evidence indicates that the dose-effect relationship for x rays and gamma rays may not be a linear function that is invariant with dose and dose rate, the use of a non-linear hypothesis for estimating risks in support of public policy on radiation protec- tion would be impractical in the present state of knowledge, since it would require considera- tion of individual variations in temporal and spatial distribution of tissue dose, as well as allowance for other variables which cannot be analyzed at this time. The possible significance of the experimental data is not the only element of uncertainty in interpreting the human data. It is the whole population from birth to death that is to be protected, and no body of human observations provides dose-specific risk estimates for longer than about 25 years. Further, the human fetus may be especially susceptible to radiation leu- kemogenesis, possibly to carcinogensis gener- ally, but the various studies are not in agree- ment on the size and nature of the effects of radiation, and no study provides more than 15 years of follow-up. The lifetime cost (to human health) of a particular radiation protection gxiide may therefore be highly sensitive to the effects of fetal irradiation. Thus, there is no certainty, and in a situation that calls for a careful weighing of costs and benefits it has seemed prudent to present numerical risk esti- mates for man on the basis of exclusively hu- man data with linear interpolation into the re- gion of low dose, merely indicating at which points the experimental data, or further human observations, might modify such estimates in the future. At this time, then, the linear hypothesis, which allows the mean tissue dose to be used as the appropriate measure of radiation exposure, provides the only workable approach to numer- ical estimation of the risk in a population. Fur- ther, since there is no means at present of de- termining the value of the dose-effect slope in the low-dose region of interest, use of the linear extrapolation from data obtained at high doses and dose rates may be justified on pragmatic grounds as a basis for risk estimation. VII. Risk Estimation To estimate the actual risk of cancer attrib- utable to a particular increase in the level of exposure of the general population to ionizing radiation would require systematic informa- tion on the effect of life-long, low-dose irradiation that is simply not available. How- ever, an approximate calculation at the level of mortality can be made on the basis of the 25- year follow-up studies on A-bomb survivors and on patients treated with intensive spinal irradiation for ankylosing spondylitis. In the Japanese, this excess mortality from all forms of cancer, including leukemia, corresponds to roughly 50 to 78 deaths per 106 exposed persons per rem over the 20-year period from 1950- 1970, i.e., five to 25 years after exposure. In the spondylitics, the excess mortality corresponds to a cumulative total of roughly 92-165 deaths from cancer per 106 persons per rem during the first 27 years after irradiation. If such rates, extrapolated to low-dose levels without allow- ance for the possible dependence of the effect of dose and dose rate, are assumed to apply gener- ally, then exposure of the U.S. population of about 200 million persons to an additional 0.1 rem during one year (approximately equivalent to a doubling of irradiation from background sources), for example, could be expected to cause 1350-3300 deaths from cancer during the 25 years following irradiation, or about 50 to 130 deaths per year. Continual exposure of the population to the additional 0.1 rem per year could be expected ultimately to cause 1350 to 3300 deaths annually, provided that the effect of a given increment of dose did not persist beyond 25 years after exposure. However, use of a factor, if known for man, to take into ac- count the influence of dose and dose rate on the dose-effect relationship might reduce these es- timates appreciably. When consideration is given to the full cumu- lative experience of an entire population, more specific attention should be paid to age at expo- sure, duration of latency, and to the size and 89

duration of the effect; and calculations should be made on the basis of the actual age distribu- tion of the population and the presently-ob- served age-specific mortality from leukemia and other forms of cancer. Since virtually all human data derive from much higher doses and dose rates than those of present interest, and do not extend beyond 25 years of systematic follow-up, the Subcommittee has considered it advisable to illustrate the uncertainty that must necessarily, at this time, characterize es- timates of the effect of a particular level of chronic low-dose irradiation on the entire popu- lation by choosing, for both leukemia and all other cancers combined, a range of values for each parameter entering into such estimates. The estimation process, which is fully de- scribed in the report,9 yields figures for the annual number of cancer deaths, without al- lowance for the influence of dose rate. These figures range from roughly 2,000 to 9,000, de- pending on the values selected for the parame- ters in question and on the choice of model used. The Subcommittee considers the most likely estimate from this type of model to be approxi- mately 3000-4000, which is equivalent to rough- ly 1% of the spontaneous cancer deaths per year. (Since 0.1 rem per year approximates the average value of the natural background ra- diation level in the U.S., these figures represent the number of cancer deaths attributable to irradiation from natural sources). Because a linear extrapolation model has been used in the calculations, the number of cancer deaths at- tributable to any dose other than 0.1 rem/year can be estimated by simple multiplication; however, it must be borne in mind that the fore- going estimates of mortality from radiation exposure may be too high, or too low, for a vari- ety of reasons: (1) the carcinogenic effects of a given dose of low-LET radiation may be lower at low dose rates than at the high dose rates on which these estimates have been based; (2) con- versely, the carcinogenic effects per unit dose may be higher at low doses and low dose rates, owing to less killing of cells susceptible to can- cer induction; (3) insofar as high dose data have provided a basis for the estimates given here, the risks may have been overestimated, owing to side effects at high dose levels which can enhance the carcinogenic action of radia- tion under certain conditions; (4) longer periods of follow-up may lead to estimates of risk that differ in magnitude from those above; (5) none of the estimates of risk used by the Committee derives from a sufficiently large experience to be free of sampling variation; i.e., the data on most radiation-induced tumors are too scanty to allow construction of dose-incidence curves adequate for extrapolation; (6) uncertainty attaches to the RBE values which must pres- ently be used for alpha and neutron radiations; and (7) further uncertainty attaches to the rel- evant organ or tissue dose, owing to attenua- tion of the radiation with depth in the body and to other sources of nonuniformity in the spa- tial distribution of the dose. The figures presented in the foregoing are not to be taken as precise estimates of risk since they are based on the incomplete evidence presently available. Moreover, the values are based on mortality data and do not, therefore, represent the number of individuals affected. If expressed in terms of incidence, including non- fatal cancers, estimates of risk could be higher by a factor of roughly 2. Follow-up studies are just now attaining sufficient scope to provide information on the magnitude and duration of the overall cancer risk in irradiated popula- tions. Nevertheless, these estimates illustrate the gravity of the problem facing those who must set radiation protection guides or stand- ards. It is essential not only that the mean dose of radiation from all manmade sources that is received by the population be as low as is prac- ticable, but that the dose to the individual also be mini mi zed. Whether there are other somatic effects that deserve to be considered in the same category with cancer in evaluating the risks of low-level irradiation remains to be determined. For those effects that may be conceived to fall into this category, however—induction of cata- racts,4 life-shortening from causes other than cancer,? and impairment of fertility4 — exist- ing dose-effect data suggest these are not likely to occur at dose levels compatible with present radiation protection guides. Hence, it seems reasonable to limit consideration to cancer alone for the purpose of this evaluation. Despite the incompleteness of the data, the evaluation of risks based on the approach summarized above affords a rational means of appraising the adequacy of radiation protec- tion standards in perspective with other fac- tors to be considered in the relevant cost-bene- 90

fit analysis. It is essential, however, that these problems remain under constant review, to observe, record, and evaluate all relevant new data, in order to insure that the estimation of risk from radiation in exposed populations be as precise as possible. VIII. Summary Cancer induction is considered to be the only source of somatic risk that needs to be taken into account in setting radiation protection standards for the general population. Despite many uncertainties, an approximate estimate of overall cancer mortality can be made on the basis of follow-up studies on Japanese atomic bomb survivors and patients treated with ra- diation for diseases other than cancer. In these populations, the excess mortality from all forms of cancer corresponds to roughly 50-165 deaths per 106 persons per rem during the first 25-27 years after irradiation. By extrapola- tion, it can be estimated that the number of deaths per 0.17 rem per year in the entire U.S. population might range roughly from 3,000 to 15,000 with the most likely value falling in the range of 5,000 to 7,000 (or 3,500 per 0.1 rem per year). It is emphasized that the risk estimates lack precision but do indicate that the mean dose both to the population and to each individual must be kept as low as practicable. REFERENCES (1) Federal Radiation Council, Implications to Man of Irradiation by Internally Deposited Strontium-89, Strontium-90, and Cesium-137, a report of an advisory committee from the Division of Medical Sciences: Na- tional Academy of Sciences-National Research Council, Washington, D.C., Federal Radiation Council, 1964. (2) International Commission on Radiological Protection. The Evaluation of Risks from Radiation. Health Phys- ics, 12:239-302, 1966. (X) International Commission on Radiological Protection. Radiosensitivity and Spatial Distribution of Dose. ICRP Publication 14. Oxford, Pergamon Press, 1969. (4) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. General Assembly, Official Records: 17th Session, Supplement No. 16 (A/5216), New York 1962. (5) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. General Assembly, Official Records: 19th Session, Supplement No. 14 (A/5814), New York, 1964. (6) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. General Assembly, Official Records: 21st Session Supplement No. 14 (A/6314), New York 1966. (7) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. General Assembly, Official Records; 24th Session, Supplement No. 13 (A/7613), New York, 1969. (lf) National Council on Radiological Protection and Mea- surements. Basic Radiation Protection Criteria. NCRP Report No. 39. National Council on Radiological Protec- tion and Measurements, Washington, D.C., 1971. 91

APPENDICES TO CHAPTER VII Appendix I. Review of Scientific Bases of Eval- uation of Risks of Low-Level Radiation A. Types of Effects Aside from cytologic (1, 2) and cytogenetic (3) abnormalities, the pathologic significance of which is unknown, no radiation injuries have been documented in man or other mammals un- der exposure conditions compatible with exist- ing radiation protection guides. Nevertheless, the possibility of certain somatic effects of low- level irradiation cannot be excluded. These in- clude neoplasms, opacities of the lens of the eye, impairment of fertility, defective develop- ment of the fetus, and life shortening (4, 5). Of these, cancer is the chief concern, because it usually involves greater detriment to an af- fected individual than do any of the others and because the risk of cancer may conceivably be increased by smaller amounts of radiation than are required to cause any of the other effects in question (5). In this report, therefore, primary emphasis is placed on evaluating the possible risks of cancer associated with low-level irra- diation. Complete evaluation of the risk of cancer, or of any other somatic effect, requires knowledge of the dose-effect relation. Because, however, dose-effect data for all the effects of interest are fragmentary, particularly under condi- tions of low-level irradiation, the evaluation must be based largely on extrapolation from data that have become available through stud- ies of the effects of larger amounts of radiation on human and animal populations. Analysis of these data entails assessment of the nature of the association between exposure to radiation and the effects in question, interpreted in the light of existing knowledge of the possible mechanisms of production of such effects. REFERENCES (1) Dobson, R. L. In "Immediate and Low Level Effects of Ionizing Radiations", (A. A. Buzzati-Traverso, editor), London, Taylor & Francis, pp. 247-252,1960. (2) Ingram, M. ibid, pp. 232-246. (3) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. General Assembly. Official Records: 24th Session, Supplement No. 13 (A/7614), New York, pp. 98-155,1969. (4) International Commission on Radiological Protection. The Evaluation of Risks from Radiation. Health Phys- ics, 12:239-302, 1966. (5) International Commission on Radiological Protection. Publication 14. Radiosensitivity and Spatial Distribu- tion of Dose. Oxford, Pergamon Press, 1969. B. Evidence of the Causal Nature of Associa- tions In the field of human radiobiology there are many situations in which the question at issue is not whether an association exists between radiation exposure and a particular disease or other manifestation, but whether the associa- tion is a causal one. That is to say, whether or not reduction or elimination of the radiation exposure per se would be followed by a reduc- tion in frequency of the disease. Contrary to widespread belief, evaluation of the causal nature of an observed association is a statistical problem, and does not involve the concept of "proof" in any definitive sense. The evaluation requires the assembly of informa- tion and concepts from many different sources and their integration into an overall estimate of the likelihood that the association is or is not causal. Where controversy exists, atten- tion should primarily be focused not on wheth- er the association is "proven", but on the limi- tations, the adequacy of the data, the lack of design and corresponding artifacts, and finally the choice of the appropriate "control" for the 93

situation. In this process of assembly and inte- gration of data there may come a point at which it becomes - for the purpose of making decisions or taking actions of practical import - more prudent to act as though the association were causal, than to continue to regard it as non-causal. Where controversy exists, it should be focussed on whether or not currently availa- ble data lead us to this point, rather than on the unanswerable question of whether the cau- sal nature of the association is or is not "prov- en". It is important to recognize that both the evaluation of individual pieces of evidence and the relative weights assigned to evidence of different kinds contain substantial elements of subjectivity, and that there are few biologic issues on which belief in the strength of the evidence of causality does not vary widely among experts in the field. The kinds of evidence pertinent to evaluation of causality have been the subject of philosoph- ical discourse for many centuries (recent re- views are in 1 and 2). When it exists, the strong- est evidence is usually derived from observa- tion of the sequelae of deliberately and ran- domly assigning individuals to different levels of the exposure. As is so obvious in the field of radiobiology, the limitation of this method is in the very few problems to which it can be ap- plied. In the absence of experimental evidence, three types of consideration are useful: 1. Time sequence. For radiation to be consi- dered a cause of a particular disease, it is clear that the radiation must precede the appearance of the disease. 2. Strength of the association. The "strong- er" the association (that is to say, for example, the higher the ratio of the inci- dence of the disease following a given dose of radiation to the incidence of the disease at lower doses or in the absence of prior radiation), the more likely is the associa- tion to be causal. 3. Consonance with existing knowledge. Be- lief in the causal nature of an association is supported by knowledge of a cellular or subcellular mechanism that makes a cau- sal relationship reasonable in the light of existing knowledge in relevant sciences, by analogy from experimental work in other species, and by evidence that the distribution of the disease in populations follows the distribution of the supposed cause. Evidence obtained through exclu- sion may also be pertinent - the more ex- tensive the efforts that have been made to identify non-causal explanations of an association, the more one is likely to be- lieve, if these efforts have been unsuccess- ful, that the association is causal. In most situations in which the existence of a causal association has become widely accepted on the basis of non-experimental evidence (e.g., fecal contamination of water and cholera, ciga- rette smoking and lung cancer, high doses of radiation and leukemia) the second type of con- sideration - strength of the association - has played a major role, for all these associations have involved "relative risks" (ratios of risk in exposed to risk in non-exposed) of the order of 10 or more. This poses a particular problem in regard to the effects of very low doses of radia- tion, for it is clear that if, for example, cancer were in fact increased in persons exposed to very small doses, the overall rate in such per- sons would probably be only slightly higher than that in non-exposed persons. Indeed, even the demonstration that such an association existed - much less evaluation of its causal im- plications - may be beyond the realm of feasi- bility, although if it did exist, such an increase in rates would be important because of the very large number of people exposed to low doses of radiation. It is extremely difficult to exclude non-causal explanations of relatively small increases in disease rates, and in this particu- lar situation considerable weight must be placed on the types of consideration in the third category, namely, consonance with other knowledge. At the present time, evaluation of effects in man of low doses of radiation must depend heavily on consideration of possible mechanisms and on extrapolation across doses and, sometimes, species. Consideration of radiation carcinogenesis in human populations exposed to low doses of ion- izing radiation is complicated by noteworthy difficulties: (1) such large populations must be studied to obtain precise data on the incidence of neoplasms at any given site that few irra- diated populations are large enough to yield quantitative dose-incidence data for any one type of neoplasm; (2) the long latent period in- tervening between irradiation and the appear- ance of many neoplasms hampers the follow-up of exposed individuals in prospective studies 94

and hampers evaluation of the exposure histo- ry of individuals in retrospective studies; (3) because of the long duration of the latent peri- od and the limited follow-up of irradiated popu- lations to date, it is not yet possible to estimate the risks of radiation-induced cancer for the entire life span; (4) many of the existing data are based on patients who were exposed to ra- diation for medical purposes and whose risk of cancer may have been influenced by other treatments or by underlying disease itself, complicating the applicability of such data to the general population; (5) because the natural incidence of cancer varies widely from one or- gan to another and is influenced by genetic background, age, sex, geographic location, diet, socio-economic factors, and other variables, the action of which is not fully understood, dose-incidence data derived from one popula- tion may not be directly applicable to another. REFERENCES (1) Blalock, H.M., Jr.: Causal Inferences in Nonexperi- mental Research. Univ. North Carolina Press, Chapel Hill, N.C., 1961. (2) MacMahon, B. and Pujjh, T.F.: Epidemiology. Princi- ples and Methods. Little, Brown, Boston, Mass., 1970. C. Dose-Effect Models The following considerations pertain primarily to the induction of cancer, but some of the general principles are also pertinent to the induction of non-neoplastic lesions. 1. General Aspects of the Causation of Cancer Although the mechanisms of carcinogenesis, or of radiation carcinogenesis in particular, are not fully known, available information implies that most, if not all, types of cancer develop as a result of the combined effects of multiple factors. These causative factors may include: prezygotic (inherited) mutations of chromosomal aberrations, which can spread during development to many kinds of cells; somatic cell mutations or chromosomal aberra- tions, which can be acquired at any time after conception; changes resulting from the action of viruses; and changes in systemic growth factors (e.g., depressed immune competence, hormonal imbalance) and in local tissue regula- tion (disorganization, damage), such as may result from diseases other than cancer or from advancing age (1). Although point mutations, chromosomal aberrations, and other changes at the cellular and molecular level may require only small doses, tissue disorganization and gross dis- turbances in physiology are unlikely without larger doses (2). Of the many types of changes which radia- tion can cause in cells or tissues, none is consi- dered to be unique for radiation. Many, if not all, such changes can presumably result from a variety of other agents. REFERENCES (1) Cole, L.J. and Newell, P.C.: Radiation Carcinogenesis: The Sequence of Events. Science 150:1782-6, 1965. (2) Rubin, P. and Casarett, G.W.: Clinical Radiation Path- ology, Vols, I and II., Philadelphia, Saunders, 1968. 2. The Question of a Threshold Dose for Carcinogenesis The amount of radiation required to induce cancer in an individual (individual dose thresh- old) can be assumed to be a variable, dependent upon the extent to which causes other than the radiation contribute to the total carcinogenic process. It is the distribution of individual dose thresholds for any particular population that will determine the dose-effect function for that population. The term "threshold dose" for car- cinogenesis is a heuristic concept used to de- note the radiation dose required to cause can- cer in the most susceptible (to radiation) indi- vidual in the population. It is taken to be an absolute value below which cancer will not be induced by radiation, and at or above which it will be induced in susceptible individuals. The concept is an attractive one because there are well-confirmed non-linear responses to radiation in the experimental literature. In radiation carcinogenesis the concept may have value in the design and interpretation of expe- riments, but its applicability to the setting of guidelines for human exposure is highly ques- tionable. There is no sufficient theory of radia- tion carcinogenesis from which the concept may be deduced, and an empirical demonstra- tion has not been made. Most human data apply 95

to high doses and high dose rates, and to utilize these data in developing guidelines for human exposure at low doses and low-dose rates re- quires interpolation between the region of high dose and zero dose. The significance of the threshold-dose concept enters at the point where one asks: How is the interpolation to be done? Linear models are easy to apply and give clear-cut estimates - but provide estimates of non-zero risk even at the lowest portion of the dose-scale. Some human populations are so large that even very small linear estimates of risk, in the region of dose prescribed by current guidelines, yield finite estimates of induced cancers, i.e., deaths. These estimates of risk are beyond empirical demonstration. It is unlikely that the presence or absence of a true thresh- old for cancer in human populations can be proved. If the intent of authorities is to mini- mize the loss of life that radiation exposure may entail, they must, indeed, be guided by such estimates, and will not rely on notions of a threshold. 3. Dose Bate and Relevant Dose in Carcino- genesis Reduction of the dose rate by protraction or fractionation of exposure over extended peri- ods of time generally permits substantial re- covery of cells and tissues from radiation dam- age, at least in the case of low-LET radiations under experimental conditions. Whether the induction of cancer is correspondingly affected may be expected to depend on the effectiveness of the total dose when delivered at a high dose rate and on any change in susceptibility that might be associated with increase in the age at irradiation resulting from the prolongation of exposure. Reduction of the dose rate in deliver- ing a dose at a very high level has been ob- served to increase its oncogenic effectiveness in experimental animals (1), presumably by reduc- ing lethal damage of cells and tissues, whereas reduction of the dose rate in delivering a small- er, suboptimal dose has been observed experi- mentally to reduce its carcinogenic effective- ness (2). When irradiation is prolonged, by protrac- tion or fractionation, the portion of the total accumulated dose which is relevant (effective) for the induction of cancer is uncertain. If irra- diation is prolonged beyond the point at which the cancer is induced, the corresponding part of the remaining accumulated dose will be irrele- vant as far as the induction of cancer is con- cerned, although it might conceivably influence the length of the latent period (2). Nonuniformity in the spatial distribution of dose within tissue (microdistribution) and non- uniformity in the distribution of dose among individuals within a population are other fac- tors which may complicate determination of the dose that is relevant for an observed induc- tion of cancer. For example, the existence of "hot spots" in the distribution of radioactive isotopes in tissue raises the question of wheth- er the relevant carcinogenic dose is the dose in or around the "hot spots", the dose in regions of more diffuse distribution of the isotopes, the average dose for the whole organ; or the corre- sponding dose to those particular anatomical types of cells within the organ that are consi- dered to be the source of the cancer (3). REFERENCES (1) Upton, A.C.: Comparative Observations on Radiation Carcinogenesis in Man and Animals. In: Carcinog-enesis: A Broad Critique. Williams and Wilkins Company. Balti- more, pp. 631-675,1967. (2) Upton, A.C., Randolph, M.L., and Conklin, J.W.: Late Effects of Fast Neutrons and Gamma Rays in Mice as Influenced by the Dose Rate of Irradiation: Induction of Neoplasia. Radiation Res. 41:467-491,1970. (3) International Commission on Radiological Protection. Publication 14. Radiosensitivity and Spatial Distribu- tion of Dose. Oxford, Pergamon Press, 1969. 4. Dose-Effect Relationship For cancers induced experimentally by irra- diation, there is generally an increase in inci- dence and a decrease in latent period as the dose is increased within a certain dose range (1). Above this range, the probability of cancer induction tends to reach a maximum and then decrease ("turn-down") with further increase in the dose. The turning-down of the dose-inci- dence curve at high dose levels has been attrib- uted to excessive cell killing, tissue destruction, and shortening of life span from causes other than the cancer in question. Experimental data for many dose intervals over a wide range of- ten give rise to a sigmoid curve, with a rising concave portion followed by a rising, fairly lin- ear portion, followed by a plateau, and then a falling portion in the region of highest dose.

When only parts or combined fragments of the total dose-incidence curve are observed and used to extrapolate to the low-dose range for which there are no adequate, concrete data, as the basis for estimating the risk of low-level irradiation for man, assumptions must be made concerning the shape of the curve in the low- dose region. In concept, this will depend on the particular carcinogenic mechanisms, the influ- ence of dose rate, the distribution of individual thresholds within the population, and the la- tent periods for carcinogenic effects as influ- enced by dose and dose rate. In practice, these factors are little known. In regard to cellular neoplastic initiation in the carcinogenic mechanisms, the shapes of dose-response curves for genetic effects of ra- diation may be pertinent to the consideration of dose-response curves for leukemia and other neoplasia. The induction of neoplasia by radia- tion may well involve direct injury to individual cells or groups of cells, and possibly to the ge- netic apparatus of such cells. Thus, the lesion produced in somatic cells may be analogous to that produced in germ cells, i.e., a "somatic mutation" of some type may be involved in the induction of neoplasia. In the female mouse, a reduction in dose rate has been observed to reduce the number of mutations to that seen in nonexposed controls. In the male mouse, a less marked dose-rate effect was seen, the slope of the dose-effect curve at low dose rates being one-third to one-fourth of that seen at high dose rates (See Chapter V, p. 52). Figure 1 is a schematic representation of observed (solid parts of lines) and unobserved (dashed parts of lines) portions of various in- complete experimental dose-response curves for both leukemia and genetic effects. Curve "a" (solid line portion) represents observations made at high doses and high dose-rates and is in itself insufficient to determine the functional form of the relationship, particularly in the region of low dose. The observed (solid portion) straight line "a" is consistent with the data, and so is the observed (solid portion) straight line "b" but "a" extrapolates linearly to zero dose and zero effect,, while "b" is curvilinear and does not. With reduced dose rate, the ob- servations may be represented in some systems by "c" (e.g., genetic effects in the male mouse) or by "d" (e.g., genetic effects in the female mouse). The available data on radiation-induced can- cer in man are relatively scanty, the conditions of exposure nonuniform and uncertain, the ir- radiated samples highly heterogeneous, the controls uncertainly or crudely matched, the observations confined to limited (high) or ill- defined ranges of dose, dose rate, and fraction of total possible post-exposure risk time, and the effects of variables other than radiation incompletely known. In view of the gaps in our understanding of radiation carcinogenesis in man, and in view of its more conservative implications, the linear, nonthreshold hypothesis warrants use in deter- mining public policy on radiation protection; however, explicit explanation and qualification of the assumptions and procedures involved in such risk estimates are called for to prevent their acceptance as scientific dogma. Further- more, the linear hypothesis is the only one which permits the selection of the mean accu- mulated tissue dose to characterize the radia- tion exposure of a group under conditions of nonuniform exposure and exposure rate. The mean accumulated tissue dose is the only prac- tical quantity that can be used to estimate the risk of cancer in such populations until the influence of the many interacting variables can be better specified (2). There is also some theoretical and logical basis for use of the linear hypothesis at low dose levels. If the dose and dose rate are small (e.g., at maximal permissible levels of low-LET radiation), the spatial and temporal separation of ionizations is sufficiently large so that one would expect effects to be caused principally by "single track" radiation, and that interactions of radiation tracks within cells would be so improbable as to be negligible. This argument implies a linear dose-effect relationship for molecular and cellular effects at low dose lev- els, even though larger doses, which may cause "multi-track" interactions at the cellular level or at the tissue level, may be associated with a nonlinear relationship. In the attempt to derive risk estimates with which to set dose limits for protection against radiation carcinogenesis, it is necessary to consider the range of dose and dose rate over which dose-effect data can provide a reasona- bly valid slope for extrapolation to low dose levels. If the slope (rate of increase in incidence with increasing dose) to be used for linear ex- 9-797 O • n - * 97

Dose —•- Fig. 1. Hypothetical dose-response curves for leukemia and genetic effects. Solid lines = observed; dashed extension of solid lines = unobserved. Lines "a" and "b": possible dose-response curves at high doses and dose rates Line "c" = genetic effects in the male mouse Line "d" = genetic effects in the female mouse Parallel dashed lines - rough limits of error for lines a and b. 98

trapolation to low dose levels is obtained by interpolation between effects observed at zero dose and those observed in the most rapidly ris- ing: segment of the curve, the estimated risk per unit dose at the low dose levels is likely to over- estimate the real risk at low doses. However, if the slope for linear extrapolation is obtained by interpolation between effects observed at zero dose and those observed within the high dose range of the dose-effect curve, where the doses exceed the maximum effective induction dose, the risk per unit dose at low dose levels may be underestimated. Estimates of risk are, of course, not scientifically reliable except in the range of observations from which they are derived and under corresponding conditions of exposure. REFERENCES (1) Upton, A. C.: Comparative Observations on .Radiation Carcinog-enesis in Man and Animals. In: Carcinogenesis: A Broad Critique. Williams & Wilkins Company, Balti- more, 1967, pp. 631-675. (2) International Commission on Radiological Protection. Publication 14. Radiosensitivity and Spatial Distribu- tion of Dose. Oxford, Pergamon Press, 1969. D. Ways of Expressing Risk Estimates of risk may be expressed in abso- lute terms or in relative (comparative) terms. The absolute risk is the excess of risk due to irradiation. In practice, this is the difference between the risk in the irradiated population and the risk in the nonirradiated population. For example, under the linear dose-incidence model, the absolute risk may be expressed as the number of excess (radiation-related) cases of cancer per unit of time in an exposed popula- tion of given size per unit of dose (e.g., 1 case/ 106 exposed persons/year/rad). The relative risk is the ratio between the risk in the irradiated population and the risk in the nonirradiated population. It is usually stated a? a multiple of the natural risk. The doubling dose, i.e., the dose that will double the standard (natural) risk, is a special example of a calcula- tion of relative risk. Absolute risk estimates are generally more useful for purposes of radiation protection than are relative risk estimates, because they specify directly the number of persons affected (1). On the other hand, if the risk due to radia- tion were found to increase in proportion to the natural risk, then the relative risk would pro- vide the more appropriate estimate. Since the existing knowledge of radiation carcinogenesis is not always sufficient to indicate which type of estimate applies best in a given situation (2), both the absolute risk (where possible) and the relative risk are given in this report. Either type of estimate when applied to dose levels for which human data are lacking involves as- sumptions concerning the dose-effect relation- ship. Either type of estimate also involves as- sumptions concerning the distribution of risk with time after irradiation — namely, the time elapsing before the risk becomes elevated ("la- tent-period") — since pertinent information for most types of cancer in human populations is fragmentary as yet. As suggested by the ICRP (3), the expression of risk estimates in absolute terms — for exam- ple, 2 cases per million exposed people per year per rem — might be misinterpreted as implying considerably greater accuracy than the facts justify. For this reason, estimates are some- times expressed in terms of "orders of risk", e.g., 1 to 10 cases/106/year/rad is a 6th order risk. In the tables prepared by the Committee to summarize the human data, use is made of 80 percent confidence intervals on absolute risk estimates to facilitate comparison of the data from different studies and not to imply greater accuracy than is warranted. In order to minimize their misinterpretation and misuse, numerical risk estimates should regularly be accompanied by the following qualifying information: (a) Range of doses, dose rates, and exposure conditions on which the risk estimate is based and for which it is, therefore, scien- tifically valid. (b) Any biomedical conditions or indications for irradiation affecting the population for which the risk estimate is scientifical- ly valid. (c) Age range (at irradiation) for which the risk estimate is scientifically valid. (d) Sex for which the risk estimate is scien- tifically valid. (e) Years of observation or person-years at risk for which the estimated average annual risk or total risk are valid. The expression of risk on an annual basis averaged over a short period may give an 99

underestimation of the lifetime risk for cancers with a longer modal latent peri- od, or may give an overestimation of lifetime risk for cancers with a short modal latent period. One of the most seri- ous problems is the fact that existing knowledge of cancer induction in man is based on a limited number of years after exposure, and information is lacking about risk during later years when the natural cancer incidence increases great- ly- REFERENCES (1) International Commission on Radiological Protection Publication 14. Radiosensitivity and Spatial Distribu- tion of Dose. Oxford, Pergamon Press, 1969. (2) Mole, R. H.: Radiation Effects in Man: Current Views and Prospects. Health Physics 20:485-490,1971. (3) International Commission on Radiological Protection. The Evaluation of Risks from Radiation. Health Physics 12:239-302, 1966. Appendix II. Estimates of Risk from Human and Animal Data A. Cancers 1. Expos ure during Childhood or Adult Life a. Leukemia Introduction Although the mechanism of leukemia induc- tion remains unknown, radiation leukemogene- sis has been empirically established by experi- mental studies and by numerous epidemiolog- ic surveys on man. Previous estimates of the risk of radiation-induced leukemia in man have been made principally from data on Japanese survivors of the atomic bombings of Hiroshima and Nagasaki, and on patients in Great Britain who were given therapeutic x-irradiation of the spine for ankylosing spondylitis (1, 2). The dose-effect curves for both of these experiences have been taken to be consistent with a linear increase in incidence with dose, from which the excess risk from radiation at high doses and high dose rates has been inferred to be of the order of 20/106/rad over a period of 20 to 25 years following exposure to 100 rem or more, or 1-2 cases/106/rad/year (1, 2). This estimate was considered to represent average risk for expo- sure at high doses and dose rates; linear ex- trapolation of the curve downward (or interpo- lation if zero effect at zero dose is assumed to represent a point on the curve) was considered to represent an upper limit of risk at low doses and dose rates. The fact that an appreciable part of the total dose in Hiroshima was due to neutrons was not taken into account in the esti- mate cited, nor was any numerical factor in- troduced to adjust the risk for exposure at low dose rates. The A-bomb Survivors The data from Hiroshima and Nagasaki have since been updated and reanalyzed, at the level of incidence for the period October 1950 to Sep- tember 1966 (3), and at the level of mortality of October 1950 to December 1970 (4), on the basis of improved (T-65) estimates of both neutron and gamma doses in the two cities1. These new appraisals, from the Atomic Bomb Casualty Some of the most useful data available for the evalua- tion of the late effects of radiation on man come from the Atomic Bomb Casualty Commission: The population of A- bomb survivors is large and received doses ranging from the trivial to the near lethal; estimated doses are available for almost all; the population is relatively unselected (ex- cept for mortality at the upper end of the dose range), in contrast to series arising from therapy administered for pre-existing disease or from occupational exposure. Although the ABCC data are probably among the best available, they are subject to certain constraints and qual- ifications. Most important is the fact that, as in all large human radiation studies, since there was no randomization of survivors to dose, there is no guarantee that high- and zero-dose survivors are actually alike in all respects other than the radiation doses received. The radiation dose de- pended on the place where a survivor was located at the instant the bomb was detonated; for women, this was usually the place of residence, for men often the place of employment. Survivors at different distances are not exact- ly alike with respect to patterns of employment, age, or sex. Although these factors can be allowed for explicitly in sta- tistical analysis, there remains the possibility that other factors, not measured, also distinguish survivors in differ- ent dose classes and are responsible for some of the ob- served differences in subsequent experience. The radiation experienced by the survivors was, of course, received at a very high dose rate. The problems involved in extrapolating data which pertain to large doses at a high dose rate to small doses at a low rate are consi- dered elsewhere in this report. Similarly, the fact that in Hiroshima there was a considerable admixture of neutrons in the radiation which emanated from the bomb raises a question about the relative biological effectiveness of high- ly energetic neutrons as compared with gamma radiation. 100

Commission (ABCC) in Japan, provide the first truly dose-specific analysis of the leukemia experience of the A-bomb survivors. Even sec- ond-generation dose estimates and 25 years of surveillance of the A-bomb survivors do not, however, provide definite answers to all the important questions relating to the influence of ionizing radiation on the empirically deter- mined risk of leukemia, entirely apart from questions of etiologic mechanisms. This is be- cause the variables capable of influencing that risk are numerous and are not all represented in the experience of A-bomb survivors (e.g., dose rate), and because, for the region of low dose, even the population of A-bomb survivors is not large enough to permit accurate assess- ment of the presumably low risks in question. Certainly by 1950, and perhaps as early as 1948, an excess of leukemia was apparent in A- bomb survivors (6) so that the period of latency of radiation-induced leukemia can safely be regarded as less than 5 years following whole- body, single-dose exposure over a wide range of dose, for both Nagasaki (essentially gamma radiation alone) and for Hiroshima (mixed gamma rays and neutrons). In 1960, Heyssel et al. (7) showed that the incidence of leukemia in A-bomb survivors had already exceeded their lifetime expectation at the rates then prevail- ing in Japan. In 1965-1970, 20-25 years after exposure, the excess was still apparent (4), but greatly reduced from the peak incidence in 1951; acute forms of leukemia were more clear- ly in excess than chronic, however, in the period of 1960-1966(3). Although the Leukemia Registry maintained by ABCC in both cities with the aid of the medi- cal schools, hospitals, and physicians of the community, included 1,360 cases on 30 June 1967, only 430 were A-bomb survivors, and only 117 were definite or probable cases within the ABCC Master Sample of 117,000 survivors on 1 October 1950, with onset between 1 October 1950 and 30 September 1966, and with known dose (Table a-1). The current (T-65) dosimetry (8, 9) like that used previously (6,7), leads to All doses of radiation received by the exposed Japanese have been corrected for shielding, and refer to the dose "free in air" at a point corresponding to the center of the body (technically, the kerma in air at a location corre- sponding to the center of the body). Depth-dose curves have been obtained for gamma and neutron radiations from atomic weapons detonated in field studies (5). The dose of gamma radiation, referenced to the dose "free in air", falls off from 100 percent on the side of the body trunk nearest the weapon to about 65 percent on the opposite side of the body. The neutron dose, also referenced to the "air dose", is about 75 percent at the body surface nearest the weapon, of the order of 15 percent or 20 percent at the midline and approximately 40 percent on the side most distant from the weapon. These considerations should in principle be taken into account in specifying the absorbed dose to the bone marrow and other tissues in the body. However, depth-dose relationships differ appreciably depending upon the precise geometry of exposure. Since this is highly variable and in fact unknown in many instances, it was deemed impractical at present to give values for absorbed dose. Therefore, all doses are given only in terms of the "air dose". An important feature of the Hiroshima and Nagasaki experience is that, in fact, the largest part of the energy released by the bombs was in the form of heat and blast. Many of the survivors were burned, either by radiant heat from the fire-ball or by the fires that engulfed the cities di- rectly after the bombings. Homes were destroyed; food was short; living patterns were profoundly disrupted. The in- fluence of this concatenation of disasters upon the subse- quent health of the survivors is unknowable. The issue is further complicated by the possibility that, at least to some degree, the less hardy members of the population had higher mortality during the first few weeks after the bomb- ings, from either disease or radiation effects. Thus the sur- vivors may on the one hand be selected as among the most fit of the bombed population but on the other hand have suffered a variety of experiences all combining to reduce their future fitness. Two classes of "controls" are employed in the ABCC data: persons who came into the cities after the bombings (so called Not-in-City or NIC group) and survivors who were in the cities, but at such great distances or protected by so much shielding that their estimated doses were small. Nei- ther group is ideal as a control, the NIC group because the immigrants are clearly distinguishable from survivors according to many socioeconomic and perhaps health fac- tors which undoubtedly affect subsequent morbidity and mortality; the low-dose survivors because they may include some small fraction who actually did receive large doses that some relatively small errors in dosimetry are poten- tially less disturbing than the known large differences that mark the NIC group. respect to occupation, social class, and perhaps other fac- tors as well. The Subcommittee judges that the more suitable compar- ison group consists of the low-dose survivors, those who received doses estimated to be less than ten rads, believing that some relatively small contamination on the side of do- simetry is potentially less disturbing than the known large differences that mark the NIC group. The data coming from the ABCC Master Sample are con- cerned with morbidity and mortality that occurred only after October 1950, that is, about five years after the bomb- ings. This deficiency is of little real importance for the study of solid tumors in relation to radiation, since the ev- idence is good that for such tumors latent periods exceed five years. Leukemia, however, has a shorter latent period, and for the complete study of radiation leukemogenesis, recourse must be had to the whole data of the Hiroshima and Nagasaki Leukemia Registries, not merely that part which concerns the Master Sample. 101

Table a-1 Incidence of Leukemia in A-bomb Survivors, by T-65 Dose and by City 1950-1966. ABCC Master Sample (from Ref. 3). T 65 dose (rads) Median values No. of subjects Thousands of person-years Leukemia Range Gamma Neutron Total Cases Rate Hiroshima 300+ 323 112 427 825 12.1 17 140.5 200-299 185 49 241 606 9.0 5 55.6 100-199 105 27 131 1,652 24.1 10 41.5 50-99 56 13 68 2,611 38.3 7 18.3 20-49 26 5 30 4,555 67.0 14 20.9 5-19 8 2 10 10,541 156.0 8 5.1 Under 5 0 0 0 62,515 915.1 27 3.0 TOTAL - - - 83,305 1,222.7 88 7.2 Nagasaki 300+ 417 7 427 566 8.4 6 71.4 200-299 238 3 240 693 10.4 6 57.7 100-199 145 2 146 1,174 17.7 3 16.9 50-99 69 0 69 1,173 17.6 0 0 20-49 31 0 31 1,354 20.0 0 0 5-19 10 0 10 4,501 66.3 2 3.0 Under 5 0 0 0 20,403 297.2 12 4.0 TOTAL 29,864 437.6 29 6.6 102

different dose-response curves for the two cit- ies, but the curve for Hiroshima now appears steeper than that for Nagasaki (Figure a-1). Apart from questions of possible error in the dosimetry which cannot be entirely excluded (10), and of genetic and other differences be- tween the populations of the two cities, the primary factor differentiating the two cities is the quality of the radiation received, with neu- trons playing an important role in Hiroshima (Table a-1) and almost none in Nagasaki. The dose-response curve for Hiroshima can be rep- resented by a linear function, although other relationships cannot be excluded. If gamma rays and neutrons are assumed to be equal in relative biological effectiveness (RBE) for in- duction of leukemia, and the dose-response curve for Hiroshima is assumed to be linear, the excess of leukemia corresponds to about 3 cases/106/year/rad of whole-body external ra- diation (3). A variety of curves, including a straight line, can be fitted to the smaller Naga- saki experience. Hence one cannot conclude, from these data alone, that the dose-response effect of gamma radiation in man is or is not linear. On the linear assumption, and with ex- cess cases at all dose-levels pooled and divided by person-year rads, the data of Ishimaru et al. suggest that excess leukemia cases in Naga- saki amount to about one per 106/year/rad. In Hiroshima a significant excess of leukemia was seen at and above the dose range of 20-49 rads. In Nagasaki, on the other hand, the excess was not significant below 100 rads. The paucity of cases in the interval 20-99 rads may well result from a combination of sampling varia- tion, on the one hand, and the restraints im- posed by this particular analysis, especially its requirements that (1) the T-65 dose be calcula- ble (for two cases, otherwise eligible, it was not), (2) the subject be in the ABCC Master Sample defined on the basis of the October 1950 Census, and (S) onset be in the interval October 1950- September 1966. For example, Brill et al. (61) list several cases in the zone 1.5 - 2.0 km from the Nagasaki hypocenter (air dose about 18 to 120 rads), with onset after 1950, that do not appear in the list of Ishimaru et al., pre- sumably because they were not in the Master Sample. Tomonaga et al. also list six cases ex- posed at these distances whose date of onset was unknown. In the most recent analysis of death certificate data, Jablon and Kato (4) show two deaths in the period October 1950 December 1964 for those with T-65 doses rang- ing from 10-49 rads, none in the interval 50-99 rads. The difference between the Hiroshima and Nagasaki data is not only a matter of dose-re- sponse and quality of radiation, but also of type of leukemia (Table a-2). Hiroshima and Nagasaki survivors differ markedly in the ra- tio of acute: chronic types of leukemia (chronic lymphocytic leukemia, rare in Japan, was not associated with radiation exposure in either city). This difference exists despite the diagnos- tic standardization that has been continually exercised at ABCC under competent hematolog- ic leadership (12). In Hiroshima survivors the ratio is 58:30, and in Nagasaki 24:5. If the data of Ishimaru et al. on the type of leukemia are examined in relation to dose, it appears that most of the difference between the two cities lies in the number of chronic cases. For acute cases the curves of incidence are much more similar. Since a possible cause of the difference in in- cidence between the two cities is a difference in the quality of radiation, a number of RBE con- stants have been fitted (3, 10) to the total leuke- mia incidence in the two cities, and an RBE of 5 has been found to give the best fit. However, the authors (10) state that "...The Nagasaki curve is quite erratic, betraying the relatively small number of cases upon which it is based, and by the same token, reminds us that we can regard any conclusions that we reach as being, at best, crude approximations to the truth."ilf the dose in rem2 at each point is determined by weighting the dose of neutrons by five and the dose of gamma rays by one, the incidence plot can still be represented by a straight line (Fig- ure a-2), and the excess leukemia incidence is approximately 1.7/106/year/rem in Hiroshima survivors. There are, then, two estimates ob- tainable from this report (3), on the linear hy- pothesis, differing by the RBE used for neu- trons, for all forms of leukemia combined, for Hiroshima: RBE = 1 3.1 RBE = 5 1.7 Lewis, in a personal communication (12) points out that a weighted linear regression analysis shows that the leuke- mia data for Hiroshima and Nagasaki are consistent with a linear model for all values of the RBE from 1.0 to 5.0. 2 See footnote 8, page 86 on usage of the term "rem". 103

Rate 140 120 100 80 8 Of u •z. o u ^. 60 20 NAGASAKI 100 200 300 400 TENTATIVE 65 TOTAL DOSE 500 iod Figure a-t. Annual incidence rate of definite and probable leukemia (all forms) per 100,000 population of A-bomb survi- vors in the ABCC Master Sample by tentative 65 total dose and city: Oct. 1950-Sept. 1966. (From Ref. 3) 100 r- 80 uj 60 U z UJ a D 40 20 HIROSHIMA Y = .182X-3.13 NAGASAKI Y=.152X+5.50 100 200 300 400 500 T65 RBE DOSE (RBE:5) (X) 600 rod Figure a-2. Definite and probable leukemia in the ABCC Master Sample by T65 RBE dose, 1950-66. (From Ref. 2) 104

Table a-2 Incidence of Leukemia in A-bomb Survivors by T-65 Dose, City, and Type of Leukemia, 1950-1966, ABCC Master Sample (from Ref. 3) T-65 total dose (rads) Person-years at risk Leukemia cases Acute Chronic Leukemia cases/100,000/yr. Acute Chronic Hiroshima 300+ 20Q-299 100-199 50-99 20-49 5-19 Under 5 TOTAL 12.1 9.0 24.1 38.3 67.0 156.0 915.1 1,222.7 12 2 8 3 6 4 23 58 5 3 2 4 8 4 4 30 99.17 22.22 33.20 7.83 8.96 56 51 4.74 41.32 33.33 8.30 10.44 11.94 2.56 .44 2.45 300+ 200-299 100-199 50-99 20-49 5-19 Under 5 TOTAL 8.4 10. 17, 17.6 20.0 66. 297, .4 ,7 ,3 ,2 437.6 Nagasaki 5 5 3 0 0 1 10 24 1 1 0 0 0 1 2 59.52 48.08 16.95 0 0 1. 3. 51 36 5.48 11.90 9.62 0 0 0 1.51 .67 1.14 105

The excess rate of leukemia for Nagasaki re- mains about 1.0/106/year/rem under either as- sumption concerning RBE, since neutrons did not contribute significantly to the dose in that city. These estimated excesses can be factored into acute and chronic cases and when this is done (Table a-3) the factors for calculating excess cases/106/year/rem become reasonably similar for acute leukemia in the two cities (0.9 and 0.8), but remain appreciably different for chronic leukemia. Additional observations on the Nagasaki survivors are not likely to estab- lish the shape of the dose-response curve in view of the small numbers to be expected. The estimate of 5 for the RBE of neutrons in inducing leukemia depends heavily on data from the high dose range in Hiroshima. In mammalian radiobiology the shapes of the dose-response curves for neutrons are general- ly different from those for x rays or gamma- rays, causing the RBE of neutrons to depend on the dose level (or on the magnitude of biological effect). The RBE generally increases with de- creasing dose (or effect). From the Hiroshima- Nagasaki comparison, no definite conclusion can be drawn concerning the variation of RBE with dose, although the data are consistent with an increase in the RBE with decreasing dose (10). It is considered likely, therefore, that the RBE of neutrons for induction of leukemia at low doses and low dose rates may be higher than the above range of values derived from the Hiroshima and Nagasaki data; however, we emphasize that there were differences between the Hiroshima and Nagasaki experiences other than those of radiation quality, and estimates of the RBE of neutrons derived by comparison of data from the two cities must be regarded as highly tentative. The published report of Ishimaru et al. in- cludes no information on age at exposure, al- though the leukemogenic effect is known to be greater in those exposed as children (3). In its attempt to provide the scientific basis for practical guidelines for protecting the general population the Subcommittee has preferred to summarize risk estimates with attention to age at exposure, distinguishing exposure in utero, during childhood (usually under 10 years of age), and later (usually age 10 + ). For this rea- son, and because their report presents data for other forms of cancer in A-bomb survivors through December 1970, the latest mortality data of Jablon et al. (4) have been used in pref- erence to the 1950-1966 incidence data of Ishi- maru et al. (3) in the summary tables at the end of the chapter. Table a-4 provides a summary of the 1950-1970 mortality data for all ages. When the mortality data for 1950-1970 are compared with the incidence data for 1950-1966 it will be seen that (a) the comparable dose-spe- cific rates are lower for the mortality rates, largely because the numbers of leukemia cases change little while the person-years accumu- late; (b) the baseline rates are defined different- ly, those for incidence being based on persons exposed to less than 5 rads and those for mor- tality on less than 10, and are slightly lower for incidence; (c) the use of the lower dose for cal- culating baseline incidence yields a lower aver- age dose to which excess cases are referred than is the case in the mortality calculation (54 vs. 72 in Hiroshima, 80 vs. 105 in Nagasaki). In combination these differences reduce the esti- mates of leukemia deaths/106/years/rad below those of Ishimaru et al. The calculations at the level of mortality are shown in Table a-5 and yield estimates of 2.2 for Hiroshima and 0.88 for Nagasaki, vs. the incidence estimates of 3.1 and 1.0 respectively for the shorter period. Table a-5 also provides estimates of excess leu- kemia mortality by age ATB and by city. Both absolute and relative risks, relative even more than absolute, are higher for those irradiated in childhood. X-ray Therapy For Ankylosing Spondylitis The series of British patients treated with spinal irradiation for ankylosing spondylitis (13, 14) has been repeatedly analyzed, to yield, on the assumption of linearity in dose response, about one leukemia death per 106 persons per year per rad, averaged over a follow-up period ranging up to 25 years. In the latest report (13) on this series of 14,554 adult patients treated with therapeutic doses of x rays, ranging (when added over all courses of treatment) from approximately 375 to more than 2,750 rads (mean spinal dose) per patient, deaths from leukemia number 60 observed through 1960, vs. 5.48 expected at British age-, sex-, and time-specific death rates in 141,496 person- years of follow-up. These data are shown in detail in Table a-6. The difference, 46.52 deaths, represents an average annual risk of 329 per 106 person-years of observation. If the average 106

Table a-3 Approximate Factors for Estimating Excess Cases of Leukemia, Based on the Experience of the A-bomb Survivors, 1950-1966 (From Ref. 3) Type of Leukemia City of exposure Total Acute Chronic (excess cases leukemia/ro°/year/rem) REE for neutrons = 1 Hiroshima 3.1 1.6 1.5 Nagasaki 1.0 0.8 0.2 RBE for neutrons_ = 5 Hiroshima 1.7 0.9 0.8 Nagasaki 1.0 0.8 0.2 107

Table a-4 Mortality from Leukemia in A-bomb Survivors, by T-65 Dose and by City, 1950-1970, JN1H-ABCC Mortality Sample (Extended) (From Ref. 4) Range T-65 Dose Gamma Mean Neutron Total No. of Thousands of Leukemia Deaths subj ects person-years Number Rate* 200+* 100-199 50-99 10-49 1-9 0 Unknown TOTAL ,3 .5 269. 108. 56.9 17.6 2.9 0 93.9 30.1 13.3 4.3 0.8 0 363.2 138.6 70.2 21.9 3.7 0 Hiroshima 1,460 1,677 2,665 10,707 13,787 29,943 1,670 61,909 26 30 48 195 251 543 29.8 71 9/ 1,126.2 27 10 7 17 34 5 100 81.6 33.1 14.5 8.7 4.3 16.8 8.9 200+* 100-199 50-99 10-49 1-9 0 Unknown TOTAL 329. 144. 70, 21. 4.0 0 .1 ,3 .3 ,3 5.6 1.4 0.2 0.0 0.0 0 ,7 .7 .5 334. 145. 70. 21.0 4.0 0 Nagasaki 1,310 1,229 1,231 3,700 6,705 4,699 1,461 20,335 24.3 23.0 22.9 67.6 123.0! 86.9: 27.0 374.8 15 3 0 2 11 3 34 61.6 13.0 0 3.0 5.2 11.1 9.1 *The T-65 dose estimates for some survivors are so high as to raise serious questions of error. There are 335 persons (Hiroshima 217 and Nagasaki 138) with T-65 doses estimated at more than 600 rads. For these individuals, 600 rads has been substituted for the calculated dose. 108

Table a-5 Calculation of Excess Deaths from Leukemia. 1950-1970. by ARC ATB and City (From Ref. 4) Age ATB Total 0-9 10+ Factor H N n+rr H N Person years x 10"3 300.7 137.9 90.03 243.1 105.5 Base-line rate 4.27 5.24 3.25 4.76 5.04 Leukemia deaths: obs. (0) 61 20 19 50 12 Exp. (E) 12.85 7.23 2.93 11.7 5.31 0-E 48.15 12.77 16.07 38.3 6.69 Mean dose, rads 72 105 69 74 113 Excess deaths/106/year/rad 2.2 0.88 2.6 2.1 0.56 109

Table a-6 Observed and Expected Deaths from Leukemia, 1935-1960 Among 14.554 Patients with Ankylosing Spondylitis Treated by X ray, 1935-1955 Leukemia deaths Followup Of observation Observed (0) Expected (E) 0/E 0-2 35,453 7 1.10 6.4 3-5 40,746 19 1.49 12.8 6-8 31,906 14 1.32 10.6 9-11 19,247 6 0.86 7.0 12-14 9,558 5 0.45 11.1 15-24 4,886 1 0.27 3.7 TOTAL 141,496 52 5.48 9.5 110

dose to the entire marrow, given a spinal mar- row dose ranging from 525 to 894 R (14), is esti- mated (see p. 197) to be 372 rads, an estimate of 0.88 deaths from leukemia/106/year/rad to the bone marrow is obtained. Although the spondylitic series provides one of the largest bodies of data on leukemia in man following x or gamma radiation, certain limitations should be borne in mind. Adminis- tered in the therapy for a chronic disease, the irradiation was partial-body and varied in pat- tern of dose, dose rate, and dose-distribution. The derivation of estimates for equivalent whole-body radiation dose thus involves several assumptions. Of perhaps equal import- ance is the fact that the excess leukemia deaths can presently be estimated only on the assump- tion that such patients have, apart from their x-ray therapy, the same expectation of leuke- mia as the general population. The study lacks an intrinsic control consisting of patients with the same disease who did not receive x-ray ther- apy and whose treatment was otherwise the same. A possible contribution of carcinogenic drugs to the tumor incidence in these patients has been suggested but no specific mechanism put forth. Other Human Experience Determination of the possible leukemogenic effect of protracted irradiation was sought in a cooperative study of 36,000 patients in 26 medi- cal centers in the U.S. (15), who were treated for hyperthyroidism either by radioiodine-131, which has become the most widely used treat- ment for this disease, or by surgery. Patients treated by surgery, the major alternative form of treatment, were used as controls, although many of them were treated earlier and hence were followed up longer (average 10.6 years vs. 6.5 years). Mean bone marrow doses from 131I were of the order of 7-15 rads received at a comparatively low dose rate. Comparison of the two treatment groups, in accordance with the design of the survey, failed to detect any leukemogenic effects of i3iI at this low dose, either in the total sample or when the compari- sons were repeated for follow-up periods of fixed length. However, the observed 39 deaths from leukemia were found to be 1.5 times expec- tation based on U.S. vital statistics, from which it was concluded that patients with hy- perthyroidism have an enhanced risk of leuke- mia, whether treated surgically or with 131I (12). The study did not have the power to detect an increase in acute leukemia of 1-2 cases per 106 per rad, independent of underlying risk. Lewis (16) has since pointed out that the ob- served rate of mortality from leukemia in the 131I patients was significantly in excess of that seen in the general U.S. population, especially the rate of mortality from acute leukemia in patients of older ages, whereas the excess in the surgically treated group alone was not sta- tistically significant. Nevertheless, because the excess in the i31I group did not differ signif- icantly from that in the surgery group, the data fail to establish the existence of an excess attributable to the radioiodine treatment per se. No excess in leukemia, or in other cancer, has been documented in populations exposed at dose rates within present occupation exposure limits, but workers exposed in the days before current safety practices have shown an excess in leukemia (1). Unfortunately, however, it is not possible to make dose estimates of any real precision for such workers, and the major stud- ies provide no basis for quantitative estimates of the leukemogenic effects of occupational radiation (17-21). Two of these studies (20-21) are negative but include too few person-years of follow-up to exclude the existence of leuke- mogenenic effects of radiation. The best data are those of Seltser and Sartwell (17) who com- pared members of the American Academy of Ophthalmology and Otolaryngology (AAOO) as to incidence of death from leukemia in the peri- od of 1935-1958. Their observations follow, by age at risk, and with expectations based on the experience of the AAOO: Age Obs. Exp. Ratio 35-49 2 50-64 8 65-79 9 Total 19 1.9 1.1 4/7 7.7 1.0 7.3 1.9 2.5 Although the experience is not large, it is much larger, in terms of leukemia mortality, than the British radiologists, experience reported by Court Brown and Doll (20), and this fact, or a difference in dose, may explain the difference in findings. I11

The Seltser-Sartwell data on U.S. radiolo- gists are also of interest for their possible re- levance to the question of a dose-rate effect for x radiation in man. The total dose received by radiologists practicing in the 1935-1958 period of the survey cannot be quantified in any real sense, but Braestrup (22) attempted an approx- imate estimate for the years 1930-1954 in re- sponse to Warren,s 1956 report on age at death of radiologists vs. other specialists. Averaging highly variable readings on 1933-1937 installa- tions, Braestrup estimated that the radiologist at that time may have received 100 R per year in contrast to about 1 R in 1957, and suggested 2,000 R as a possible life-time (40-year) occupa- tional dose for radiologists dying from 1930- 1954. Marinelli (23) has recently taken this es- timate as equivalent to about 600 rads to the marrow, and derived from the Seltser-Sartwell report on radiologists an estimated absolute risk of 0.4 deaths from leukemia per 106 man- year-rads. The lower 90 percent and upper 90 percent confidence limits on this estimate are 0.20 and 0.56, apart from the uncertainties in the dosimetry, and are below the range of most estimates in the table on pp. 117-118. For sever- al reasons, however, Marinelli,s estimate pro- vides no upper bound on the probable risk; (a) the Seltser-Sartwell cohorts had not all com- pleted their professional lives; (b) the exposure period (1945-1958) is later than that to which Braestrup had reference; and (c) Braestrup,s 1933-1937 estimate of 400 mR/day is four times the NCRP maximum permissible dose in effect from 1934 to 1949. If the NCRP maxima are taken as the average dose year by year from 1935-1958, a mean dose would be more nearly 800 R on a 40-year basis, and appreciably less for the partial life-time actually observed by Seltser and Sartwell. If Marinelli,s calculation had been done on the basis of 800 R the ob- served excess would have been about one death/ 106/year/rad. Another occupational study is that of Miller and Jablon (21), who compared 6,560 World War II Army-trained x-ray technologists with 6,826 pharmacy or medical laboratory technol- ogists as to mortality through 1963, without finding any clear evidence of radiation-induced leukemia over the 18-year period. Failure to observe a significant increase in leukemia mor- tality in this instance may well reflect the shorter duration of radiation exposure and lower cumulative doses received by the technol- ogists in comparison with the radiologists; also, a two- or three-fold increase in risk could by chance have gone undetected in a survey of this size. In three independent studies from Denmark (24), England and Wales (25) and the U.S. (26), an association exists between leukemia in adults and exposure to diagnostic x rays. In each study the association is demonstrable for the myeloid forms of leukemia but not for the chronic lymphatic form. The type of leukemia reported is the same as that seen in persons exposed to heavy doses of radiation. In each study the association is most pronounced for exposures to regions of the body containing active marrow in the adult, i.e., the trunk, as opposed to peripheral regions. In the U.S. study the association is reported only for males; none was shown for the smaller number of females. This greater sensitivity of males is also seen among A-bomb survivors, as well as in the natural incidence of the disease. In the U.S. study, which involved 1,414 leukemia cases occurring in three states during the years 1959-1962, Gibson and colleagues (25) estimate the relative risks associated with specific num- bers of x-ray films. For example, they find that the risk to males of chronic myeloid leukemia following exposure to 11 or more films to the trunk is 2.2 times the risk from exposure to less than 11 such films. The average dose to the to- tal red marrow from such exposures is expected to be less than 0.5 rad per film. These data are consistent, therefore, with the possibility that repeated small doses of radiation given to males may result in a relative risk of leukemia per rad which is comparable to that incurred from a single large dose. Cogent observations come from studies of patients subject to pelvic irradiation. Four such studies concern radiation used to castrate female patients, for relief of certain benign gynecologic conditions (27-30); three studies concern radiotherapy for cervical cancer (31- 34); and one study concerns both uses of irra- diation (26). The reports of Doll and Smith (27) and Alderson and Jackson (29), agree in the finding of leukemia excess after radiation cas- tration, which is greatest in the interval five to nine years following irradiation. The study of Brinkley and Haybittle (28) is much smaller and finds no cases of leukemia, as compared 112

with an expected number of 0.24 determined from general population incidence data on can- cer registration. Waggoner (35) found a leuke- mia excess in patients exposed to radiation for castration, the excess appearing at all inter- vals five years or more following exposure. Randall et al. (30) found an increased overall cancer mortality following radiation for cas- tration, but their data do not give a breakdown of the cancer into leukemia or other specific sites. Thus, four of the studies find a leukemia excess, one study finds a cancer excess but does not investigate leukemia specifically, and one study is too small to identify an excess of the magnitude report by the others. Relative risks of death from leukemia following radiation- induced castration in three of the positive stud- ies are given as 4.58 (27), 2.27 (29) and 2.32 (35). The dose is given in two of these papers (20, 21) in terms of dose to the ovary or to some other intrapelvic position as in the range of 300-1500 rads. Two other papers (18, 26) report the mean dose to bone marrow, which range from 40 to 300 rads. Randall et al. give the ex- posure as 1200 milligram hours intracavitary radium or greater. In contrast to the studies on radiation cas- tration, none of the reported studies on irradia- tion for cervical cancer (31-35) shows a leuke- mia excess. The mean marrow dose for the pa- tients in these studies is estimated to be be- tween 300 and 1500 rads, or usually well above the range of dose received in radiation castra- tion. Only a small portion (about 15 percent) of the patients given radiotherapy for carcinoma of the cervix received intracavitary radium alone, in some of whom the radiation dose was in the same range as that received by patients given radiation for castration. The great ma- jority of the cervical cancer patients received external radiotherapy, alone or in addition to intracavitary radium. The finding of no leuke- mia excess at the higher doses associated with radiotherapy for cancer is not adequately ex- plained. It is conceivable that some sufficiently high dose of radiation is lethal to a high pro- portion of leukocyte stem cells and that this lethal effect outweighs any leukemogenic effect at these doses (36). No model has been present- ed, however, to describe these two effects quan- titatively in the patients of these series. It is also conceivable that the risk of leukemia decreases disproportionately with diminution in the fraction of marrow that is irradiated, as has been observed to be the case in the induc- tion of certain lymphoid leukemias of the mouse (37), such that when the irradiated frac- tion falls below a certain "critical" level (as might be true in patients irradiated for cervi- cal cancer) the risk per rem becomes drastical- ly smaller, if not negligible. Again, however, the existing human data are insufficient to re- solve this issue. An excess of leukemia has been observed in each of three series (38-40) of patients investigated after being injected for diagnostic purposes with thorotrast, a colloidal suspen- sion of thorium dioxide. In these series, which together included more than 2,000 patients, the leukemias appeared after an average of 17 years following injection of the radioactive col- loid, during which time the marrow was irra- diated continually with alpha particles emitted by the thorium, at a rate estimated to average 8-13 rads per year mean tissue dose. Interpre- tation of this dose in terms of a dose-effect rela- tionship is greatly complicated, however, by nonunif ormity in the spatial distribution of the dose within the marrow, the radiation tending to be concentrated in microscopic "hot spots", and by the high LET of the alpha radiations responsible for the dose. In view of the greater sensitivity of children to the leukemogenic effect of radiation, the fol- low-up studies of children treated by x ray for thymic enlargement and for tinea capitis are of special interest despite their small size. Among the 1,451 subjects irradiated as infants for thymic enlargement, with an average follow-up of 18 years, Hemplemann et al. found six cases of leukemia in comparison with 0.96 expected on the basis of upstate New York incidence, a relative risk of 6.2 which translates into an absolute risk of 3.0/106/year/rem when the mean marrow dose of 65 rads is taken into ac- count (41). A similar estimate is derived from the report of Albert and Omran on 2,043 child- ren treated by x ray (30 rads estimated marrow dose) for tinea capitis and observed for an av- erage of about 15 years (42). They found four cases vs. 0.9 expected, from which the absolute risk has been estimated at 3.4/106/year/rem. Other human series have been investigated for evidence of a possible relationship between leukemia incidence and radiation exposure, the results of which have been reviewed elsewhere 4OT-W O - It - 9 113

(1). These include groups exposed medically to radiation from internal and external sources other than those mentioned above, as well as groups residing at different environmental background radiation levels. The studies are insufficient themselves to establish a dose-ef- fect relationship. Hence, the evidence from such studies is not reviewed again in this re- port. Radiobiological Considerations Pertaining to Temporal and Spatial Distribution of Dose. From experiments with animals it may be inferred that the leukemogenic effect of ioniz- ing radiation varies with the quality of radia- tion, LET, dose rate, total accumulated dose, and perhaps other physical factors as well (37). The induction of myelocytic leukemia in mice has been studied following exposure to both gamma rays and neutrons (43), and the results are shown in Figure a-3. Although there was no dose-rate dependence with neutron exposure, a striking dependence was seen with gamma ra- diation, the effect of a single dose being per- haps five or more times that of the same total dose accumulated through daily exposures. The lack of dose-rate dependence for neutrons (45) counters the argument (46) that the dose-rate effect seen with gamma rays was only apparent and resulted from a difference in age-sensitivi- ty to leukemia induction. Even if that argu- ment were correct, the fact remains that a giv- en dose of gamma radiation delivered at low dose rates over a sizeable part of an animals, life span yielded fewer leukemias than did the same dose delivered in a single brief exposure at an early age, as has been observed repeated- ly (37). The dose-rate dependence in leukemogenic effectiveness of gamma rays and the contrast- ing lack of dose rate dependence in leukemogenic effectiveness of fast neutrons typify the influ- ence of dose rate on the effectiveness of low- LET and high-LET radiations as observed for most effects in mammalian radiobiology (43). In general, the reduced effectiveness of low-LET radiations at low dose rates is attributed to repair of incipient stages of injury at low dose levels, associated with which there is usually a corresponding departure of -the dose-effect curve from linearity even at high dose rates (Fig. 1, p. 98 curve "b"); in contrast to this situation, the curves for high-LET radiations are typically relatively insensitive to dose rate and are linear (Fig. 1, p. 98 curve "a") (43). The induction of neoplasia by radiation may well involve direct injury to individual cells or groups of cells, and possibly to the genetic apparatus of such cells. A definite dose-rate dependence of effect has been seen with respect to genetic effects in the mouse. The data have bearing on a possible dose-rate effect for leuke- mogenesis in man only if the assumption is made that a "somatic mutation" represents the underlying mechanism. The data from Hiroshima (large neutron dose component) are consistent with linearity; the Nagasaki curve (essentially no neutron compo- nent) suggests curvilinearity with increas- ing slope as the dose increases. However, be- cause of the small number of cases involved, particularly in the Nagasaki data, it is not possible to conclude that the curves are in fact different. Nevertheless, the apparent differ- ences in the curves do correspond to what has been observed widely in radiobiology, both in tissue culture and for late effects such as leuke- mia induction (31, 32) and lens opacification in the mouse (47, 48). Because of the different shapes of the curves for high and low-LET ra- diation, the RBE increases with decreasing dose. Furthermore, there is generally a differ- ential dose-rate dependence, with little or no dose-rate dependence observed following expo- sure with high-LET radiation and a definite dose-rate dependence following exposure to low-LET radiation. In summary there are no data on the human being that allow one to conclude that a dose- rate effect is or is not operative in radiation- induced leukemia in man. A dose-rate effect has been demonstrated for the induction of myelo- cytic leukemia in the mouse following exposure to low-LET radiation. Radiobiological consid- erations are cited which are consistent with the dose-rate effect seen in the mouse. Mechanisms of Leukemogenesis The pathogenesis of leukemia, like that of other forms of cancer, remains largely un- known. The differences in age distribution of the various forms of the disease in non-irra- diated populations (49-50) and the absence of 114

5CH 300 500 DOSE (iods) 700 Figure a-3. Myeloid leukemia in male mice. 0 single exposure; Q daily exposures. (Open symbols denote results with gamma rays and x rays; solid symbols, neutrons.) (from Ref. 44) 115

an increase of the chronic lymphocytic form in irradiated populations imply that pathogenetic mechanisms may vary appreciably from one form of leukemia to another. The presence of the Philadelphia chromo- some in the great majority of patients with chronic granulocytic leukemia and the high susceptibility of leukemia associated with con- genitial diseases involving chromosomal abnor- malities (e.g., Down,s syndrome, Bloom,s syn- drome, and Fanconi,s syndrome) suggest that the leukemogenic effects of radiation may pos- sibly be related to radiation -induced chromoso- mal or genetic damage (51). Studies on experi- mental animals have also implicated leukemo- genic viruses, at least in the induction of cer- tain forms of leukemia in irradiated mice (37). The experimental induction of leukemias of diverse types has been studied extensively in irradiated mice and, to a lesser extent, in irra- diated rats, guinea pigs, dogs, and monkeys (37). At dose rates less than 0.01 rad/min., bea- gles (52), swine (53), and mice (54) have shown significant increases in the rate of leukemia. Furthermore, reduction of the dose rate below this level by a factor of 3 in the beagles and by a factor of approximately 10 in the swine, failed to abolish the significant increase in the frequency of leukemia and related diseases. In general, the observations in these species are insufficient to establish dose-effect relation- ships at low dose levels, or to characterize the process of leukemogenesis in detail. Hence, for purposes of estimating risks in human popula- tions, reliance cannot be placed on knowledge of mechanisms but must be placed chiefly on empirical observations derived from study of the occurrence of leukemia in man himself. Summary of Empirically Determined Risks for Man Both absolute and relative risk estimates obtained from the major reports cited above are summarized in Table a-7 to provide a basis for arriving at a "best" current estimate for administrative and scientific purposes. This is the first of the summary tables of standard format more fully described in Appendix VI. Only studies generating information directly useful in quantifying risk have been included. The table lists, first, estimates for individuals exposed at age 10 or more; next those exposed as children; and finally, those exposed in utero. The in utero are shown here merely for compar- ison; they are discussed specifically in Appen- dix IIA 2. In addition to the standard definitions and explanations given in Appendix VI, there are special footnotes to Table a-7 relating to specific studies. For each study listed, Table a-7 gives infor- mation on conditions of radiation exposure (type of radiation, duration of exposure, range of external dose, mean tissue dose), size of expe- rience, demographic character of subjects, source of baseline (control) observations, and measures of risk. The risk estimates are of three forms: 1. Relative risk (col. 16) - For subjects ex- posed to a given (mean) dose, the ratio of the observed number of leukemia cases (or deaths) to the expected number. 2. Percentage increase in relative risk per rem (col. 18) 3. Absolute risk (cols 19-20) — The leukemo- genic effect is expressed in terms of the excess number of cases (or deaths) per mil- lion person-years of observation per rem. Appendix VI should be consulted for more spe- cific definitions and methods of calculation. In view of the fact that the estimates for A- bomb survivors pertain to external dose, for which no adjustment to marrow dose has seemed possible, the tabulated studies on indi- viduals exposed as adults are in good general agreement, as are those on subjects exposed as children. The various measures of risk may be summarized as follows: Percent increase in relative risk/rem Absolute risk (cases/ 106/year/rad) Adults Children 2 to 3 5 to 10 1 to 2 2 to 3 These estimates are intended to apply to large population groups; particular subgroups may be at higher, or lower, risk than these average values. 116

2 J - 1 10 8a30H300J /sa8B3 JO i 33a3 SJ 1ad ui asn.uDui ju .3,i,,C, 3SH (3/0) XaS (SJBdA) jo siaafqns 20 (SJC.T)A) (SJBaX) dn-«0[[" JO uoT3BJ UBdW JO JO • « • • • • a • I | u M 1 ii m o 0 Mortality; SAX male 4J C -4 a r4 C ~* b • Morbidity V 14 1-i « b * « O o • iH £ • £ • b 4J > M^ m* i; Ji. & 3« 3* 4J • U 5^3 ao 04 -"JD m ao --t iTs ov p- l-i ^' (N •H i-4 CN1 iH * « £ O> C^ O O 0 t-4 m r*. C3 r*. in ov *£> ^ S CN r4 r-l — * in uS in " ^ CM iH •-i i-4 0 -* cn r-4 i-4 o* C*> • * r^ (N m CN •* CN i—i — J M N •-i U1 r4 iTl - - H - I i*~ 9 i*i - ". « in CN • • m I ••* •4 100 i*! • «Q t-< i=)": •* (N sO CN iN ^ co a- in « - in *-i r-' • r> i i-i m • m m CN • 0> *D ?! Os T3 « -0 W C « Syotland &1 • "4 0 i- i 2 H £ M ^ X * fc. Z J) Bk X ifl b. X £ 4 h, b. m cn iO so g m in m <n « >* | 4 4 j>* i* iH r-i m s CN m S s in CN P*. NO H = n H r4 m i o iS R°. 0 0 i~l Q s§ O O SS- £ *H 5 CN CN M H I S in m CN 1 *n m o^ g r^ CN 1 n i-l •• ifi lH 00 CN 3 iN CN O CN r*- oo S K § 1 CO M? •'» N m *» O oT r^ m rH CN i-H £ x: 5-S O CN m *» CN 1 O 4J 4J £ X CN sO in 4-1 4J i N sO lA M vO fi Of O U"i in m m m r*. vO CN CN CN o 01 cn iO N in m CN i m m CN iN CN 1 ma r-4 b b V O 5,o 2 COB iH i-* •3" : « |f_l v v X ** + f- + c X -* •^ •* JJ CO Ol 1 CN l-s 1 „ •rt • Menor- rhagla Patients 1940-60 i-i w -j ^S O m x cm •Q n « i B -H *H m &*j 4J cn a o* ca a. " •?Bs .0 z: -J 1 CT-- < ^ * < -> , -O O IS s e II 22 c v * *• O io -• 91 O 4) T3 " . •a a O w E-£ .2 o l o as io *) O cp O •!-) T3 O •S c- 5.2 117

SUUimooo JO /SOSEO JO Uf O-,-C..JDii ' ji.".-J • (3/0) jo "S II 35 (8p»a) «»8u»a 30 •. i. -I -|» J" paeeq aJ> 3J«V M»TM« ao -JM.JJi jo n" i"" :".i (8JBa'C) d JO .i'.(Hi- i aJnsodxd UOTJCIPCJ JO uoT3BJlXI jo * • a Morbidity. See footnote (b) Morbidity. See footnote (y) a ) M « 0 c m r-i u • £ * w^ S^" *• in r- 1** C* ae •-i « •1 O ^4 so sO 0 -» fJ ^* r> cn o •-i ao m ^4 GO M t-i m P4 ^4 I • sr fi m ax 9i i4> Oi CN •tf o* -* •-i • v0 o « M O • • - C 4J • V « >i •O 4J *H • • O • w -a • •o 1 io O ^ axc ::l O M X •« fc. SB •• * x * fc. m in o i>• •i i m O 1 M o C^ 1-1 s U1 8 o 8 iH + sO A « if V 3 ao m M •4 3 8 o o S p i*- H cn s m g ** ** M £ J3 in m i 4J 4J 1 C4 *3 o O CN in CO m M m i m 1 CN M iN in r> « b C B H V 1 >- + a X X d- CN .=r -^ 1Z- •?+» 9i « ^ PN. in • • 4J ON -Ha-fr « 4J « 1 « -H *H O i Si •< » ^ >•i * CN j= 1 CTi H X ^ c a t •* Cm m o* o o.•-i C i- £ e •i T3 — c 'c ss '/) i4-l O « O X 4-* e o < Li v O..C -H -i tt m ai 4) 4) T3 .0 cfl C — J! * O 4) 3 cn o j= -a 300 >• Li a) u 7 ;< - « X B V 4) — co o oi cu CO C id •) a * B 4-1 Li 2 s S S C L, S S. 4-. X E C 4-* V V £ JhC E OC 3 4-i •** 4) Q U *- i U Li C 41 i -^ J- 33a) 4) u J Z CO O Li u *o - CO 4 ) B W » 4J 4) 4) O 4) 4J 4J O 4-1 41 T3 CX 4-1 4) 4 CO n U m 4) c 01 L, -^ CO 4J 4J 4J O CO 4J • jJ j= 4^ m -*-i : C0 4J 3 u -o 1-4 4J 4 O CO C1) 3 iN Li 4J 4) CO •* n •-• r TJ ~* •)>,- io JD ^ CO £1 C -H •a c ^3 Li 118

REFERENCES (1) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. General Assembly, 19th Ses- sion. Supplement No. 14 (A/5814), p. 85,1964. (2) International Commission on Radiological Protection. Publication 14. Radiosensitivity and Spatial Distribu- tion of Dose, Oxford, Pergamon Press, 1969. (3) Ishimaru, Toranosuke, et al.: Leukemia in Atomic Bomb Survivors, Hiroshima and Nagasaki, 1 October 1950 - 30 September 1966, Radiation Res. 45:216-233, 1971, and Atomic Bomb Casualty Commission Technical Report #25-69. (4) Jablon, S., and Kato, H.: JNIH-ABCC Life Span Study Report 6, Radiation Dose and Mortality of A-bomb survi- vors - 1950-1970, Atomic Bomb Casualty Commission Technical Report #10-71. (5) Bond, V.P., Imirie, G. W., Cronkite, E.P., and Stickley, E.E.: Distribution in Tissue of Dose from Penetrating Gamma and Neutron Radiation. In: Radioactivity in Man, George R. Meneely, M.D., Editor. Charles Thomas, Publisher, Springfield, Illinois, 1961. (S) Brill, A.B., et al.: Leukemia in Man following Exposure to Ionizing Radiation, Summary of Findings in Hiroshi- ma and Nagasaki, and Comparison with Other Human Experience, Ann. Intern. Med. 56:590-609,1962. (7) Heyssel, R.M., et al.: Leukemia in Hiroshima Atomic Bomb Survivors, Blood 15:313-331,1960. (8) Auxier, J. A., et al.: Free-field Radiation-dose Distru- tions from Hiroshima and Nagasaki Bombings, Health Physics 12425-429,1966. (9) Milton, R. C. and Shohoji, T.: Tentative 1965 Dose Esti- mation for A-bomb Survivors, Hiroshima and Nagasaki, Atomic Bomb Casualty Commission Technical Report/1- 68, Hiroshima. (10) Jablon, S., Fujita, S., Fukushima, K., Ishimaru, T., and Auxier, J.A.: RBE of Neutrons in Japanese Survi- vors. Proceedings of the Symposium on Neutrons in Ra- diobiology, November 11-14, 1969, Oak Ridge, Tennessee, CONF-691106, Biology and Medicine (TID-4500). (11) Finch, S.C., et al.: Detection of Leukemia and Related Disorders, Hiroshima and Nagasaki, Atomic Bomb Cas- ualty Commission Technical Report #5-65, Hiroshima. (12) Letter to Dr. A.C. Upton, Chairman, Subcommittee on Somatic Effects, from Dr. E.B. Lewis, California Insti- tute of Technology, 16 Feb 1972. (13) Court Brown, W. M., and Doll, R.: Mortality from Cancer and Other Causes after Radiotherapy for Anky- losing - Spondylitis. Brit. Med. J. 5474:1327-1332,1965. (14) Court Brown, W. M. and Doll, R.: Leukemia and Aplastic Anaemia in Patients Irradiated for Ankylosing Spondylitis, London, Her Majesty,s Stationery Office, 1957. (15) Saenger, E. L., Thoma, C. £., and Tompkins, E. A.: Incidence of Leukemia Following Treatment of Hyper- thyroidism - Preliminary Report of the Cooperative Thyrotoxicosis Therapy Follow-up Study, JAMA 205:855-862,1968. (16) Lewis, E. B.: Leukemia, Radiation, and Hyperthyroid- ism. Science 174:454, 1971. (17) Seltser, R., and Sartwell, P.: The Influence of Occupa- tional Exposure to Radiation on the Mortality of Ameri- can Radiologists and other Medical Specialists, Amer. J. Epid. 81:2-22, 1965. (18) Dublin, L. I., and Spiegelman, M.: Mortality of Medi- cal Specialists, 1938-1942, JAMA 137:1519-1524,1948. (19) Warren, S.: Longevity and Causes of Death from Ir- radiation in Physicians, JAMA 162:464-468,1956. (20) Court Brown, W.M., and Doll, R.: Expectation of Life and Mortality from Cancer Among British Radiologists, Brit. Med. J. 2:181-187, 1958. (21) Miller, R. W., and Jablon, S.: A Search for Late Ra- diation Effects among Men Who Served as X-ray technol- ogists in the V.S. Army During World War II, Radiology 96:269-274,1960. (22) Braestrup, C. B.: Past and Present Radiation Expo- sure from the Point of View of Life Expectancy. Amer. J. Roent. 78:988-992,1957. (23) Marinelli, L.D.: Estimates of the Radiation-induced Leukemic Risk in Man, Argonne National Laboratory Report #7760, Part II, June 1970. (24) Faber, M., Andreasen, E., and Uhrbrand, H.: Further Studies on Irradiation-Induced Leukemia in Denmark. In: Transactions of the Sixth Congress of the European Society of Haematology, A. Videback, Ed., Basel, Swit- zerland, and New York: S. Karger.Part 2, pp. S211-S213, 1958. (25) Gibson, R., Graham, S., Lilienfeld, A., Schuman, L., Dowd, J.E., and Levin, M.L.: Irradiation in the Epidemiol- ogy of Leukemia Among Adults, J. Nat. Cancer Inst. 48:301-311,1972. (26) Stewart, A., Pennybacker, W., and Barber, R.: Adult Leukaemias and Diagnostic X-rays, Brit. Med. J. 2:882- 890,1962. (27) Doll, R. and Smith, P.G.: The Long-Term Effects of X- irradiation in Patients Treated for Metropathia Hae- morrhagica, Brit. J. Radiol. 41:362-368,1968. (28) Brinkley, D., and Haybittle, J.L.: The Late Effects of Artificial Menopause by X-irradiation, Brit. J. Radiol. 42:519-521,1969. (29) Alderson, M.R., and Jackson, S.M.: Long Term Foi- low-up of Patients with Menorrhag-ia Treated by Irra- diation, Brit. J. Radiol. 44:295-298,1971. (30) Randall, C. L., Paloucek, R. P., Graham, J. R. and Graham S.: Causes of Death in Cases of Preclimateric Mennorhagia, Amer. J. Obst. Gyn. 88:880-887,1964. (31) Zippin, C., Bailar, J.C., Kohn, H. I., Lum, D. and Eisenberg, H.: Radiation Therapy for Cervical Cancer: Late Effect on Life Span and on Leukemia Incidence, Cancer 28:937-942,1971. (32) Kohn, H. I., Bailar, J. C., and Zippin, C.: Radiation Therapy for Cancer of the Cervix: Its late Effect on the Life span as a Function of Regional Dose, J. Nat. Cancer Inst. 34:345-361,1965. (33) Simon, M., Brucer, M. and Hayes, R.: Radiation and Leukemia in Carcinoma of the Cervix, Radiol. 74:905- 911,1960. (34) Hutchison, G. B.: Leukemia in Patients with Cancer of the Cervix Uteri Treated with Radiation. A Report Covering the First Five Years of an International Stud- y, J. Nat. Cancer Inst. 40:951-982,1968. (35) Waggoner, J.K.: Leukemia and Other Malignancies Following Radiation Therapy for Benign Gynecological Disorders. Unpublished. Paper presented at American Public Health Association Annual Meeting, November 12,1969. (36) Gray, L.H.: Radiation Biology and Cancer. In: Cellu- lar Radiation Biology, 18th M.D. Anderson Hospital and Tumor Institute Symposium on Fundamental Cancer Research, 7-25, Baltimore, Maryland, Williams and Wilk- ins, 1965. (37) Upton, A.C., and Cosgrove, G.E., Jr.: Radiation Leu- kemia. In: Experimental Leukemia, edited by M. Rich, New York, Appleton-Century-Crofts, pp. 131-158, 1968. (38) Blomberg, R. L., Larsson, E., Lindell, B., and Lindren, E.: Late Effects of Thorotrast in Cerebral Angiography, Acta Radiol. (Diagnosis) 1:955-1006,1963. 119

(39) de Silva Horta, J., Abbat, J.D., Cayolla da Motta, L.A.R.C., and Roriz, M.L.: Malignancy and Other Late Effects Following Administration of Thorotrast, Lancet 2:201,1965. (40) Faber, M.: Epidemiological Experience with Thoro- trast in Denmark. The Dosimetry and Toxicity of Thoro- trast. International Atomic Energy Technical Report IAEA-106, Vienna, 1968, pp. 139-146. (41)Hempelmann, L.H., Pifer, J.W., Burke, G.J., Terry, R., and Ames, W.R.: Neoplasms in Persons Treated with X- rays in Infancy for Thymic Enlargement. A report of third follow-up survey, J. Nat. Cancer Inst. 38:317-341, 1967. (42) Albert, E.R., and Omran, A.R.: Follow-up Study of Patients Treated by X-ray Epilation for Tinea Capitis, Arch. Environ. Health 17:899-950,1968. (43) Upton, A.C.: The Influence of Dose Rate in Mammali- an Radiation Biology: Quality Effects. In: Dose Rate in Mammalian Radiation Biology, edited by D.G. Brown, R.G. Cragle, and T.R. Noonan, Division of Technical In- formation, U.S. Atomic Energy Commission, CONF- 680410, Oak Ridge, Tennessee, 1968, pp. 22.1-22.18. (44) Upton, A.C., Randolph, M.L., and Conklin, J.W.: Late Effects of Fast Neutrons and Gamma Rays in Mice as Influenced by the Dose Rate of Irradiation; Induction of Neoplasia, Rad Res. 41:467-491,1970. (45) Russell, W.L.: Studies in Mammalian Radiation Ge- netics, Nucleonics 23, No. 1,1965. (46) Gofman, J.W., and Tamplin, A.R.: The Mechanism of Radiation Carcinogenesis. GT-109-70. Joint Committee on Atomic Energy, 1970. (47) Bond, V.P: Lens Opacification in the Mouse; Implica- tions for RBEand QF. Presented at the Second Panel on the Biophysical Aspects of Radiation Quality, Vienna, April 10-14,1967. (48) Bateman, J.L., Rossi, H. H., Kellerer, A.M., Robinson, C.N., and Bond, V.P.: Dose-Dependence of Fast Neutron RBEfor Lens Opacification in Mice, Radiation Research, in press. (49) Court Brown, W.M., and Doll, R.: Adult Leukemia. Trends in Mortality in Relation to Aetiology, Brit. Med. J. 1:1063-1069, 1959. . (50) Court Brown, W.M., and Doll, R.: Leukemia in Child- hood and Young Adult Life. Trends in Mortality in Rela- tion to Aetiology, Brit. Med. J. 1:981-988,1961. (51) Miller, R.W.: Persons with Exceptionally High Risk of Leukemia, Cancer Res. 21:2420-2423,1967. (52) Dungworth, D.L., Goldman, M., McKelvie, D.M., et al.: Development of a Form of Myelogenous Leukemia in Beagles Exposed Continuously to Strontium-90. In: Mye- loproliferative Disorders to Animals and Man, W. J. Clarke, E.B. Howard, and P.L. Hackett, Eds., USAEC, Symposium Series #19, pp. 272-289,1970. (53) Howard, E.B., and Clarke, W.J.: Induction of Hema- topoietic Neoplasms in Minature Swine by Chronic Feed- ing of Strontium-90, J. Nat. Cancer Inst. 44:21-38, 1970. (54) Warren, Shields and Gates Olive: The Induction of Leukemia and Life-shortening in Mice by Continuous Low-level External Gamma Radiation Rad. Res. 47:480- 490,1971. (55) Stewart, A., and Kneale, G.W.: Changes in the Cancer Risk Associated with Obstetric Radiography, Lancet 1:104-107, 1968. (56) MacMahon, B.: Prenatal X-ray Exposure and Child- hood Cancer, J. Nat. Cancer Inst. 28:1173,1962. (57) Jablon, S., and Kato, H.: Childhood Cancer in Rela- tion to Prenatal Exposure to Atomic-Bomb Radiation, Lancet 2:1000,1970. (58) Ferber, B., Handy, V.H., Gerhardt, P.R., and Solo- mon, M.: Cancer in New York State, Exclusive of New York City, 1941-1960, Albany, Bureau of Cancer Con- trol, New York State Department of Health, 1962. (59) Woolf, B.: On Estimating the Relation Between Blood Group and Disease, Ann. Hum. Genet. 19:251-253, 1954. (60) Haldane, J.B.: The Estimation and Significance of the Logarithm of a Ratio of Frequencies, Ann. Hum. Genet. 20:309-311,1955. (61) Brill, A.B., Tomonaga, M. and Heyssel, R.M.: Leuke- mia in Humans Following Exposure to Ionizing Radia- tion, ABCC Technical Report 15-59,1959. b. Radiation-induced Thyroid Cancer Although the thyroid gland is comparatively resistant to destruction by radiation, studies in man and animals during the past decade have demonstrated it to be relatively susceptible to the induction of neoplastic lesions (1). The ra- diation doses associated with the observed neoplastic changes, although lower for x rays than for 131I beta rays in rats (2), appear to be of the same order of magnitude in man and sev- eral species of experimental animals (3). It should be pointed out that in the case of the human population the radiation exposure, al- though considered to be partial body, involved the entire thyroid gland. Dose Response For external x radiation, there are reasona- bly good dose-incidence o. ta for thyroid neo- plasms in man (4, 5, 10) ar.u in rats (6, 7). The shape of the dose-response cu/ve has not been clearly denned, but with moderately high doses (over 1000 rads) the induction of neoplasms (mainly benign) approaches 100 percent in per- sons exposed during childhood (Fig. 6-1) (4, 8). One study of clinically palpable nodules in three groups of persons irradiated in childhood suggests a linear response (perhaps curvili- near at the higher dose range), with linearity down to relatively low doses (above 20 rads) (5) (Fig. b-2). Also, in the Japanese who were under 20 years of age when exposed to the atom bomb explosions, a distinct dose incidence correlation has been reported (9,10). Dose Rate There is no good information about dose rate effects for human thyroid cells; however, on the basis of the physical dose absorbed in the 120

5.0- 40- o t o I 1.0- Lfe- 100 200 300 400 500 THYROID DOSE (RADS-l 600 Figure b-1—Incidence of thyroid neoplasms versus thyroid dose (rads): O refers to tumor incidence in the AP-treated individuals of the oldest cohort (born be- fore 1940), + refers to all persons in oldest cohort, and X to those in the two oldest cohorts combined (born before 1950) (Ref. 4). MARSHALL ESE CHILDREN JW 200 900 400 500 600 700 800 900 1000 1100 1200 Estimated cumulative thyroid dose (rad) Figure b-2. Incidence of thyroid modularity in relation to estimated cumulative dose to the thyroid gland. The points represent values based on real (or assumed) incidence of nodularity and estimated mean cumulative physical doses to the thyroid gland. The horizontal dashed line and arrow repre- sent the direction in which the mean dose of the Marshallese should be adjusted to take into account the fact that beta-rays from Iodine-131 are probably less effective than x-rays in inducing thyroid neoplasms. Animal experiments indicate that beta radiation is one-tenth to one-fifteenth as effective as x irradiation (13); this obviously could not be the case here (unless the estimated dose is too small), because a correction of this magnitude would reduce the dose to the Marshallese thyroids to less than that of the Subgroup C children. The horizontal dashed line and arrow for Subgroup C takes into consideration the possibility that the fractionated doses might be less effective than sin- gle exposures—a presumption not substantiated by the evidence at hand (Ref. 5). 121

gland, there is some circumstantial evidence that beta rays from a mixture of internally deposited radioiodine isotopes are as effective as x rays in initiating tumor formation in child- ren (5). The observations in the Marshallese irradiated as children primarily with iodine isotopes from an H-bomb explosion in 1954 are also consistent with those noted after x ray exposure (8), although the number of cases is small (one case of thyroid cancer has been found). In this instance, however, the shorter- lived radioiodine isotopes, which are 10 to 20 times more biologically effective than 131I (11), were responsible for much of the tissue dam- age. Hence, the dose rate was not as much lower than that associated with x ray exposures as it would have been if 131I had been the only source of radiation.i In the cooperative thyrotoxicosis study of patients treated with large doses of 131I, no clear-cut increase in the number of cases of thyroid cancer was noted above that found in hyperthyroid patients not given 131I (12). Al- though the failure of radiation to induce cancer in these patients might conceivably have re- sulted from resistance of their hyperactive thyroid cells to malignant transformation (13), it seems more likely to have been due to the possibility that their doses were in excess of an optimal dose for tumor induction, since their thyroids received many thousands of rads. The presumed explanation for a decrease in the oncogenicity of radiation at high levels, which has been observed in animals as well as in man, is that the radiation damage to thyroid cells at these levels is so severe as to kill the cells or render them incapable of sustained prolifera- tion. Recent evidence has shown upwards of a 50 percent incidence of hypothyroidism five to ten years after treatment of hyperthyroidism with 131I, suggesting that the relevant doses of radiation are in fact sufficient to cause exten- sive death of thyroid cells (22). Studies of direct chromosomal damage in man have shown that pre-operative tracer dos- es of 131I which delivered 50-100 rads to the •Approximately seven-eighths of the total dose due to radioiodine came from decay of i3iI and i:)5I, which irradiat- ed the gland at initial dose rates of 0.28 and 0.6 rads per minute, respectively. thyroid produced no increase in chromosome aberrations demonstrable in cells cultured from thyroid tissue excised surgically soon af- ter irradiation (14, 21). This is in contrast to the large increase in aberrations (up to 33%) noted in cells cultured from patients who had received 400-780 rads of x rays to the thyroid gland in infancy 31-37 years previously (14). In rats, the carcinogenic effectiveness of per rad absorbed dose, is approximately one- tenth of that of x rays (2). Studies show that 131I is also about one-tenth as efficient in kill- ing thyroid cells in rats as in the same dose of xrays (15). Similar but less well quantitated experiments in sheep also indicate that 131I is much less effective in cell killing than is x irra- diation (16). In producing chromosome aberra- tions in the hamster thyroid, on the other hand. 131I has been reported to be as efficient per rad as xrays (17, 18). In evaluating such observa- tions for their implications concerning dose rate, one must remember that the radiations from 131I and from xrays differ in energy dis- tribution within the gland as well as in dose rate. Host Factors Little is known of the influence of host fac- tors in man; however, the action of thyroid stimulating hormone (TSH) is required for in- duction of thyroid cancer in animals after car- cinogenic stimuli, including radiation exposure (19). Furthermore, after irradiation, increasing amounts of TSH are associated with increasing cancer incidence. Cell proliferation kinetics (rapid during adolescence, slower during child- hood, and almost static during adulthood) pos- sibly explain the fact that the thyroid cells seem to be more sensitive to the carcinogenic action of radiation in human beings exposed as juveniles than in those exposed as adults (4, 9). The well-differentiated forms of thyroid can- cer tend to run a relatively benign course in young adulthood and middle age, as is perhaps reflected also by the high incidence of benign appearing "occult" thyroid cancers observed at autopsy in the Japanese (20). In older age groups, "spontaneous" thyroid cancers have a more malignant course, but whether this is also the case with radiation-induced malignancies in these age groups is not known. 122

Mechanisms In animals and in human beings, there is ev- idence that the pathogenesis of thyroid cancer is a multistage process, involving a primary event causing lasting damage (possibly chro- mosomal) to thyroid cells, followed by second- ary events which promote cell division, thereby allowing the neoplastic potential of the altered cells to be expressed (19). Visible damage to chromosomes (aberrations) has been demon- strated in a substantial proportion of cultured cells from irradiated thyroid glands in humans (14,21) and Chinese hamsters (17, 18); in one of the human cases, exposure to x rays had oc- curred 45 years before (21). Of the secondary factors which promote ex- pression of the malignant potential of dam- aged, yet viable, thyroid cells, stimulation of proliferation by thyroid-stimulating hormone (TSH) is clearly important. In rats, the neo- plastic process after thyroid irradiation pro- gresses through a spectrum of stages begin- ning with cellular hyperplasia, followed by benign neoplasia, and ultimately by malignant transformation (19). Risk Estimates On the assumption of linearity in dose re- sponse, even in the low dose range, the risk of thyroid cancer appearing in adolescence or young adulthood (from birth to 25-30 years) after irradiation in childhood may be estimated (Table b-1) to be of the order of 1.6 to 9.3 cases per year per million children exposed per rem (4, 5, 10, 22, 23). Since the time of development of the radiation-induced tumors is age-depend- ent, the actual risk of tumor induction during childhood is lower than this, and during adoles- cence it is higher. There is a suggestion that cancer induction may decline as the irradiated population enters the third decade, implying a decrease in risk at later ages (4). In the Japanese A-bomb survivors, the rela- tive risk at high dose levels is not clearly in- creased for persons 20 years of age or older at the time of exposure, but is definitely increased for those under 20 at exposure (9, 10). Another study of subjects treated as adults (age>20 years) for an average of 22 y^ars previously who had received an average dose of 2100 rads of external x radiation showed no cases of thy- roid cancer (24), although if the absolute risk of cancer induction were 2 cases/rad/year per 106 subjects, 20 cases would have been expect- ed. These studies confirm the lesser susceptibil- ity of the adult thyroid to radiation-induced carcinogenesis, as compared with the thyroid of the infant or child. It should be emphasized that the values given apply only to exposure at high dose rates. Lit- tle is known about the risk of tumor induction at low dose rates (<0.1 rad/hour). Several studies have not been utilized in making the risk estimates tabulated, for a va- riety of reasons. Some of the studies, which appeared to show an association between ra- diation and an increased cancer incidence, have not been used because they did not lend them- selves well to quantitative treatment. Other studies did not show such an association and could not be used to estimate a risk other than zero. All reports should, of course, be consulted for a proper understanding of the complexity of the problem of estimating the risks in ques- tion. REFERENCES (1) United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). Report. Supplement 14 (A/5814), New York, United Nations, 1964. (2) Doniach, I.: Effects Including Carcinogenesis of P3i and X-rays on the Thyroid of Experimental Animals. Health Phys. 9:1357-1362,1963. (3) United Nations Scientific Committee on the Effects of Atomic Radiation. Report on Current Problems in Exper- imental Radiation Carcinogenesis, (in preparation). (4) Hempelmann, L. H., Pifer, J. W., Burke, G. W., Terry, R., and Ames, W. R.: Neoplasms in Persons Treated with X-rays in Infancy for Thymic Enlargement. A Report of Third Follow-up Survey. J. Nat. Cancer Inst. 38:317-341, 1967. (5) Hempelmann, L. H.: Risk of Thyroid Neoplasms after Irradiation in Childhood. Science 160:169-163,1968. (6) Lindsay, S., Potter, G. D., and Chaikoff, I. L.: Thyroid Neoplasms in the Rat. A Comparison of Naturally Occur- ring- and Ii3i-Induced Tumors. Cancer Res., 17:183-189, 1963. (7) Vasilenki, IIA. and Klassovskii, IaA.: Tumor Effects of Radioiodine Isotopes, Vest. Akad. Med. Nauk USSR, 22(12):30-33,1967. (8) Conard, R. A., Dobyns, B. M. and Sutow, W. W.: Thy- roid Neoplasia as Late Effects of Exposure to Radioac- tive Iodine in Fallout, J.A.M.A. 214 (No. 2): 316-324, 1972. (9) Wood, J. W., Tamagaki, H., Neriishi, S., Sato, T., Shel- don, W. F., Archer, P. G., Hamilton, H. B., and Johnson, K. G.: Thyroid Carcinoma in Atomic Bomb Survivors, Hiroshima and Nagasaki, Amer. J. Epidemiol. 89:4-14, 1969. (10) Jablon, S., Belsky, J. L., Tachikawa, K., and Steer, A.: Cancer in Japanese Exposed as Children to Atomic Bombs, Lancet 1:927-932,1971. 123

OnatnDOt) loq3o Jo 8d30l13004 § Morbidity; se footnote (b) Morbidity; se footnote (b) RBE for n taken as S, gamma as L Js Morbidity 9 • 5B OS o CO d ^ . 00 CsJ 2 g i>1STJ a^njosqy J 5 ° ie .saa § in CN o CN r+ m.>J .. id -4-iij a A; iv[.-J ;u,oi.,j 00 ON CN uT aseaJouT r-i U1 - ass r+- - - - (a/0) 1«W aATae-[an § o ii i-. |O co in o "* d Upper N.Y. State Inyidenye « «i 4-1 U jo § CO C CI) -r4 D. >• o o. . c D 2 M O M xaS xf *:** r ia u. s: 08 h (sJeaK) "••« 2 O o . — CN in O o i7* 1 O 0 3 i UBaH ON rH CN CN ,-M •— O § o o 1 0 o - o o 1 00 •-* O ON O »— S 8 m CN JO 00 CD 00 CN cn 00 (SJBaA) paBBq ,1.1 B O uo -piiJ.i i ,iL' i rt- po^Jdj c n m i °.l (.aBaX) UBaH ^-* 2 CN CN JO UOI 1C.HH] a.UBH s CO CN CO CM 10 1 r*. aJnsodxa UO-HBTPBJ 2 • Jl CO O C O 1) C O C) — 1 4-i J jjij V B 3 uoT3BlpeJ jo adXi c X X t- + c a3UaJajaH CN - - o m TK" « m 3 >* 1 i i III = J* ? s! r ao n H X rH - £ rc 4) u 3 4J O ,,S rs fi m ns E^ .-i -a D « « •r-l >* u c i) '"3 ? a 124

(11) Klassovskii, IuA.: Dependency of Irradiation Effects on the Determination of Dose in Thyroid Histological Structures. Akademiia Meditsinskipk., 40-42, 1967. (12) Saenger, E.L., and Tompkins, E. A.: Personal Commu- nication (13) Sloan, L.W., and Frantz, V.K.: Thyroid Cancer: Clini- cal Aspects. In: The Thyroid, Werner, S. C., ed., 2nd ed., New York, Harper and Row, 1962, Chapt. 26, pp. 445-468. (14) Doida, Y., Hoke, C., and Hempelmann, L.H.: Damage in Thyroid Cells of Adults Irradiated with X-rays in In- fancy. Rad. Res., 45: 645,1971. (15) Greijj, W.R., Smith, J.P.B., Orr, J.S. and Foster, C.J.: Comparative Survivals of Rat Thyroid Cells in Vivo af- ter P3i, Ii35 and X-irradiation, Brit.J. Radiol 43:542- 548,1970. (1ff) McClellan, R. O., Clarke, W. J., Ragen, H. A., Wood, D. R, and Bustad, L.K.: Comparative Effects of Ii-" and X- irradiation on Sheep Thyroid, Health Phys. 9(12): 1363- 1368.1963. (17) Moore, W., Jr., and Colvin, M.: Chromosomal Changes in the Chinese Hamster Thyroid Following X-irradiation in Vivo, Int. J. Rad. Biol., 14:161-167.1968. (18) Moore, W., Jr., and Colvin, M.: The Effect of I73i on the Aberration Rate of Chromosomes from Chinese Hamster Thyroids, Int. J. Rad. Biol. 10:391-401,1966. (19) Furth, J.: Radiation Neoplasia and Endocrine Sys- tems. In: Radiation and Biology and Cancer, Univ. Texas Press, Austin, Texas, 1959. (20) Sampson, R. J., Oka, H., Key, C. R., Buncher, C. R., and lijima, S.: Metastasis from Occult Thyroid Carci- noma. An Autopsy Study from Hiroshima and Nagasaki, Japan, Cancer 25:803-811,1970. (21) Socolow, E.L., Engle, E., Mantooth, L., and Stanbury, J.B.: Chromosomes in Human Thyroid Tumors, Cytoge- netics 3:394-415,1964. (22) Green, M., and Wilson, G.M.: Thyrotoxicosis Treated by Surgery or Iodine-131. With Special Reference to Development of Hypothyroidism, Brit. Med. J 1:1005- 1010.1964. (23) Beach, S.A., and Dolphin, G. W.: A Study of the Rela- tionship of X-ray Dose Delivered to the Thyroids of Children and Subsequent Development of Malignant Tumors, Phys. Med. Biol., 6:583,1962. (24) DeLawter, D. S., and Winship, T.: Follow-up Study of Adults Treated With Roentgen Rays for Thyroid Dis- ease, Cancer 16:1028-1031,1963. c. Radiation-Induced Bone Cancer Bone cancer is a relatively rare form of can- cer in man, and environmental factors which contribute to production of bone cancer are not well known. The first environmental factor identified was radioactivity introduced into workers from exposure to radium dial paint in the early part of the century. Other irradiated groups that have been studied include patients receiving radioactivity or radiation exposure in the therapy of various diseases; e.g., injec- tion of radium isotopes or external x radiation to the skeleton for ankylosing spondylitis or other bone diseases. x Skeletal tumors developing at the site of previous therapeutic external irradiation have been reported in a few dozen instances. The neoplasms, which are of cartilage as well as of bone, include malignant and benign types. The osteosarcomas have arisen after doses general- ly varying from 3000 rads to more than 15,000 rads, with a latent period averaging about nine years. The ages of affected patients have ranged from less than ten years to more than 60 years (1). The benign tumors, which devel- oped after exposure to much smaller doses Ca- sually less than 500 R in air), are chiefly osteo- chondromas. Among a group of such tumors developing after radiotherapy to the medias- tinum in infancy, the interval between irradia- tion and histologic diagnosis averaged roughly 11 years (2). The lower doses associated with the osteochondromas, as compared with the osteosarcomas, imply that susceptibility to induction of bone tumor may be higher in in- fants than in adults, and that the skeletal tu- mors induced by irradiation in infancy tend to be predominantly benign, whereas those in- duced by irradiation in adult life tend to be predominantly malignant (3). Relation of Cancer Rate to Radiation Exposure Except for the cases of bone cancer associat- ed with x ray therapy, most cases have been associated with deposition of radionuclides in the skeleton. Recent data from the study of Hiroshima and Nagasaki survivors indicate that some bone cancers are beginning to ap- pear in the highest dose group, but the cases are too few for analysis. Spondylitis patients given x ray therapy to the spine represent a group of approximately 14,000 individuals, 84% males, in whom four cases of bone cancer have been observed, where- as only 0.63 case was expected (4, 5). One cancer case also received 224Ra therapy (11). This is the largest group in which quantitative risk estimates can be derived, and it has the further advantage that marked local variations in the dose to bone cells is not a major problem, as it is for internally deposited radionuclides. The follow-up period has been short, however, and the number of observed cases, four, is small; in addition, these patients already had a disorder of musculoskeletal tissues at the time of irra- diation. For these reasons interpretation of the findings in relation to estimation of the risk to the general population is difficult. 125

Another important group of subjects in which the relationship between cancer inci- dence and radiation exposure has been investi- gated includes about 770 exposed individuals, mainly dial painters, observed in studies at M.I.T. (6) and Argonne National Laboratory (7). These subjects, studied originally in two series, are now being followed as a single group by the Argonne National Laboratory Center for the Study of Human Radiobiology. In this group of individuals, most of whom are still alive, there have been 51 bone sarcomas and 21 carcinomas of the head and paranasal sinuses. Four of the carcinoma cases have occurred in individuals who have also developed bone sar- comas. The dose-response data up to the pre- sent have been recently summarized by Row- land and associates (8, 9). No sarcomas or car- cinomas have been seen below a total accumu- lated mean bone dose of more than 500 rads, but the incidence rises sharply above this point, particularly in the case of the sarcomas. Rowland and coworkers have suggested that an empirical equation of the type I=KD2 exp (D/D') provides the best fit for the sarcoma data, where I is cancer incidence, D is cumula- tive mean bone dose, and K and D, are con- stants. The exponential term is introduced to account for an apparent fall in incidence in the highest dose range. At lower values of D, an equation of this type reduces to a proportional- ity between incidence and the square of the cumulative dose. This particular relationship in effect implies a very small probability of radiation-induced bone cancer at low cumula- tive doses, with a sharply rising value for the higher ranges, certainly a better fit to the ob- served incidence rates in this group than a strictly linear fit to the data. It should be noted that the majority of the patients in this rather small series have cumulative radiation doses in the lower ranges. A group of dial painters in Czechoslovakia, Poland, and Switzerland is currently being studied, after having absorbed strontium-90 and some radium-226 during dial painting with- in the past few decades (10). The estimated dos- es are only a few rads to the bone in these indi- viduals, however, and the likelihood of much pertinent information on cancer risk being available from them within the next decade or two is slight. Another major group of subjects includes approximately 900 patients given radium-224 for therapy of bone tuberculosis or ankylosing spondylitis in Germany within the last two decades. The results of a follow-up averaging 20 years have recently been reported by Spiess and Mays (11). In this group 53 patients have developed bone sarcomas, with periods from initial injection to the time of appearance of the disease ranging from 4 to 20 years. It is significant that within this group of patients with intravenous radium-224 exposure there have been no carcinomas of the cranial or na- sal sinuses. The majority of the bone sarcomas occurred in younger patients, with 35 cases observed in 217 individuals less than 20 years of age at the time radium was injected, and 12 in 708 individuals greater than 20 years of age at injection (6 cases have been omitted because the injected dose is not known). A plot of the incidence of bone sarcoma ver- sus the accumulated skeletal dose appears to be approximately linear for adult as well as juvenile patients, but again, as in so many in- stances, the lowest dose range has not shown any sarcoma cases, a circumstance explicable by chance, since the numbers of subjects are small. The lowest cumulative dose at which a bone sarcoma has occurred is 90 rads, and the tumor occurred in one of the adults. According to Spiess and Mays there are at least another 1,000 to 2,000 individuals in Germany who have been given radium-224 therapeutically, and a follow-up study of these patients would be help- ful in defining the dose-response curve in the lower ranges. Since this form of treatment of ankylosing spondylitis is still being used in Europe, additional patients may be expected also to become available for study. With regard to the dosimetry from ingested radium-226, radium-228, radium-224, or other members of the decay chain of the major urani- um series, the most detailed treatment of this problem has been given by Marshall in a pre- liminary report (13). In his analysis he at- tempts to separate the dose to bone surfaces from that delivered to the interior portion of the noncellular calcified bone crystals. For the alpha-emitting radionuclides that are bone seekers, the calculation of the dose to the sur- face cells is believed to be the most relevant parameter, since these cells are not efficiently 126

irradiated from radioactivity deposited in min- eral volume. On the basis of this model, the apparent discrepancy between the results ob- tained with radium-224 and those with radium- 226 and radium-228 is clarified. Spiess and Mays have calculated the effective dose to the bone surface with a model such as has been described by Marshall and Rowland, and ac- cording to their calculations the average skele- tal dose, which has been given above in relation to the cancer incidence, is not the relevant ex- posure criterion; it is the dose to cells on the bone surface that is relevant (11). When the mean local dose to the soft tissue layer within 10 microns from the bone surface is calculated, they estimate that for radium-224 this local dose is 9 times the average skeletal dose, on the assumption that half the skeletal radium-224 decays on bone surfaces. In contrast, for ra- dium-226 the average soft tissue surface dose is less than the average skeletal dose by a factor of 0.63. On the basis of the surface dose, there- fore, the lowest dose at which cancer has been observed in man is approximately the same for the two radioisotopes, 810 rads for radium-224 and 570 rads for radium-226, suggesting the possibility that, based on follow-up periods of at least 20 to 30 years, a threshold could exist for carcinogenesis by radium isotopes. As yet, however, no conclusion on this point can be drawn because of the small population at risk in the lower exposure range. Further study of the radium-224 patients should help resolve this problem. An interesting footnote to the work with 224Ra is that in those patients who received the radium injections for shorter peri- ods of time (i.e., about 3 months), the incidence per rad was one-half as great as in those who received more protracted exposure (11,12). It is noteworthy that the new dosimetric models, on which the risk ultimately must be based, are both biological and physical in their approach. It has long been recognized that the radiation dose from "hot spots" in bone, that is, from local relatively high concentrations of radioactivity, appears to be less well correlat- ed with biological effects than is the dose from the more diffusely deposited radionuclides. It has been concluded from this fact that the "hot spot" results in "overkilling"; i.e., it generally causes local cell death and thus irradiates acel- lular portions of the bone, a conclusion which is consistent with the more recent models. In cal- culating bone dosimetry from internal bone- seeking radionuclides, it seems likely that in the future the relevant dose will be the inte- grated dose to the cells at bone surfaces, and considerable efforts are now being made to as- semble relevant metabolic data as a basis for calculating the integrated dose for all of the radionuclides of interest. For example, Mar- shall has summarized the information on sur- face retention for radium-226, using available kinetic data (15), and he has shown that the surface retention for radium-226 would be rela- tively short from a single exposure. On this basis, therefore, the average skeletal dose from radium-226 would be expected to include a sub- stantial fraction of wasted radiation, in that the radiation would not be delivered to the cel- lular elements at the bone surface. The same argument applies also for calcium-45 and americium-241. Plutonium and thorium have been recognized as remaining with the bone surface until resorbed or buried under new bone, and hence they will give higher surface doses per average rad to bone, a conclusion supported by animal studies. Mays (14) esti- mates that "monomeric" plutonium-239 is about 9 times more effective on the basis of average skeletal dose than radium-226, with polymeric plutonium-239 somewhat less effec- tive than the "monomeric" form. This distinc- tion between the forms in which plutonium may reach the bone illustrates the importance of physico-chemical factors in the microdosimetry from these radionuclides. Host Factors in the Relation to Bone Cancer One obvious factor which contributes to the probability of bone cancer development in man is age at the time of exposure. For young indi- viduals, and possibly also in those exposed in utero, the rapid deposition of bone-seeking ra- dioelements during active bone growth might confer a higher risk of cancer than in adults. It should be noted, however, that long-term expo- sures to low levels of long-lived radionuclides may not necessarily lead to a higher risk when the exposure begins prior to birth than when exposure begins at a later age, if the dose is accumulated very slowly. In patients exposed to radium-224, there was no significant difference in the incidence of bone cancer by sex. Since this group of patients 127

also had pre-existing bone disease, Spiess and Mays (11) attempted to determine whether their cancers were more likely to appear in the areas of bone affected by the disease, and they concluded that there was no predilection for cancer to develop in regions with active tuber- culosis or spondylitis. Experimental Studies The body of information which has accumu- lated from experimental studies in a number of species is greater for the bone-seeking ele- ments than for any other group of internal emitting radionuclides. Many of the experi- ments evaluate the effects of low doses in long- lived species. Experiments have been summa- rized recently in a symposium held at Sun Val- ley, Idaho, in 1967 (15), and also by Mays and Lloyd (I6), and by Mays, (unpublished) report for the NCRP committee. The particular ra- dionuclides whose long-term cancerogenic ef- fects have been investigated are plutonium- 239, thorium-228, americium-241, strontium-90, radium-226, radium-228, andcalcium-45. One of the principal purposes of these studies has been to compare the relative radiotoxicity of these different radioelements. The dose-response evidence obtained to date for strontium-90 indicates that as in the case of radium-226 in man there appears to be a lower limit of dose at which no significant cancer effects have yet been observed, and Mole (17) has concluded that a relationship I = KD2 is applicable for 9°Sr. For plutonium-239 and thorium-228 the evidence indicates a signifi- cant probability of cancer induction even at relatively low average skeletal doses (15). These experimental studies reinforce the view that alpha-emitting radionuclides are more effective than are beta-emitters, such as stron- tium-90, and that those radionuclides which tend to be translocated to the interior of bone will show a lower cancer probability for the same total dose to bone than those which re- main on the bone surface. Summary of Human Data and Estimate of Radia- tion Risk for Bone Cancer Table C-1 summarizes the data for the three populations in which risk estimates can be cal- culated. Dose-response curves for the two ra- dium-injected groups are shown in Figs, c-1 and c-2. The data for the 224Ra-injected patients are consistent with the linear nonthreshold dose- response curve within the limits of the dose range available and when the dose is expressed as mean dose to bone. The data for 226Ra are more consistent with a curvilinear relationship between cancer rate and mean bone dose al- though a straight line fit to the data cannot be excluded within the statistical confidence lim- its. This subject has been highly controversial, and it is apparent from the figures that as of now a final determination of the dose-response relationship for 226Ra in bone cannot be made. Until the difference between the two radium groups has been resolved, perhaps by use of a bone cell dose model such as has been developed by Marshall, the risk will be calculated as the average for each entire group. In the following summary table the correc- tion from rad to rem is made by using an RBE (or QF) of 10 for alpha irradiation. SUMMARY OF RISK ESTIMATES FOR BONE CANCER Irradiated as Adults (20 years) Absolute Risk Cases/108/yr per remi mean bone dose Pts. Pts. Spondylitics 0.11 0.55 0.10 Relative Risk % Increase in Rate/year per rem* mean bone dose 0.71 5.5 1.4 Irradiated as Children (1-20 years) Absolute Risk Cases/106/yr per rem* mean bone dose Pts. 0.96 Relative Risk % Increase in Rate year per remi mean bone dose 9.6 iCorrection to rem on assumption that RBE = 10 for alpha particles 128

B T ^ go S £t O """ C n a 4) X CI O iJ r £ H3U001KOD JaqUO JO S330U3OOJ c •*•i i * a 'd i -u 4J 4J a) a * g . T) Li >4 O 41 01 4J V-. o c a. « o • 11 O 0 0 B W sO iM sl 4J y -o o v ) 4) •-i W lix VI for definitions of headings. L U) ' - L ( « O C cn o CN M CO CN CN ioaJ/JaaX/ nT •H o ' '9°T S """ O i-l 0 M /t9BV3 -to sipBdp ^- !? c „ i 00 i-l 0 O o O S 3 sO in m 2 » 0 ""»T*««MtmTe»u.o»a 30 sO in 2 o CO •I in o 0 0 O - - I I cn c g. (a/o) I«TJ aAT3BTS ^ ~ O i-l O 01 0i Xoj3uoo § 1 1 X |: § United States S S *-4 • s u Kl 4 0) xaS 3 •«* o X *S *• X Jl u. X -« ». ac ^ b. [-1 (SJBaA) U*aW S 0S 0 c in iH § ^ ti CN i O i O l iH r-4 r-- — ! CN ^ 0 CN r^ m in 01 0) o, r;;vw4 I i i~i in § 2 CN CN tf) 4) O O ss S3 8§ o 3 m "§ r* g£ o gin* V o ~ vS V V CM -Tzn O in o cn 5 CN E o^ cn S 00 ao" CN sO* CN CN cn CO m r*.. § S § 3° i-*. CN m U 00 o CN r- 0 4-1 O) (ut*X) 1 paBBq aaB c 1 i V -C -C 4J 4J .C X CN II 5S 4 00 -pi-JJT J011B P-lTJL'J i-i in 4) S a 1 <„.,*) S S CN 00 0. 0 « m in M CN CM CN c i i 1 sO m in m m a^nsodxd 30 uo^3BJna cn 1-4 c i •4 » o « iD 0 9 a 0 « 41 X 8" S. l^l g~- v X t) X Hi uopBtpoj, jo adXx e» • e « M 0 m*m 00 r- o o »• p C O -"JET k 1 R 1 T3 -' 3 , ,,3-a • ^4 S -T >, C lft •0 « 01 l Isss-j, B # a * E 4i C in 3 4J V i 1 Ha 3 •« U « i 3 -j- 4-1 Oi i £ S £3 -a iN 41 u 5 oe HP* i-i 2 iS22 •SH-SS: n 22 R O C e- S 8 "3 * c u (3 O O) r- >, O- pi O r- *u a) a u > a- CD ra ia «a) 33 -H £ OO4^ •o *i o Li o 00 C 4 O X g4J J3 X T5 T3 4) U 4) •O 3 V jj — at 4-1 c ) -^ — jr o ed ffl •*-) O O lu -• O "3 >u " "3 § 5 LJ U' Si! •J -J CO > u in i^ ••* c ~* 3 > -H U b iw -* 3 4-i O (X C 4 4) 4J C Li X 4 a) n § ° ° g. £ S5 u friu •^ 03 O •o e •1 S £ •O O U C Cu O > i^ Li C7v i 4) U 5 3 4B9-T97 O - 72 - l0 129

12,000 10,000 RISK OF BONE CANCER RADIUM 226 EXPOSED GROUP "DIAL PAINTERS" FOLLOWUP THROUGH 1971 DC Z EC LLJ 0. 8,000 6,000 0 4,000 2,000 20,000 40,000 60,000 80,000 MEAN BONE DOSE (REM) 100,000 120,000 Figure c-1. Dose-response data for bone cancer in Argonne National Laboratory series of subjects exposed to radium-226 in period from 1915 to 1935 (H, 9). This group includes dial painters and some patients given radium therapeutically. Ordinate: Excess bone cancer cases per million person years Abscissa: Mean bone dose in rem (RBE= 10). The dashed line is drawn by eye through the points; the solid line is the weighted mean slope taken from Table c-1. Error bars based on Poisson statistics and include 90% range (see Appendix IV). 130

RISK OF BONE CANCER T RADIUM 224 EXPOSED 1 GERMAN PATIENTS FOLLOWUP THROUGH 1969 I/ /I 30,000 • / \ / of / \ 1 / \ / r / 4 \ | \ # 1 _i 2 20,000 I 'fcV i UJ i /^ 1 i \ /</ 1 UJ T * JL ; / 10,000 T T / I ' / IRRADIATED AS: 7 T A I -T ADULTS (>20yr.) =jt= I ' / I I I / OLX- CHILDREN «20 yr.) i \j 1 $/\* SLOPE 0.56/1 06/YR/R EM l^Qi IJL i -ft_r . i i 10,000 20,000 MEAN BONE DOSE (REM) 30,000 Figure c-2. Dose-reaponse data for bone cancer in German patients given radium-224 therapeutical!y (11). Ordinate: Excess bone cancer cases per million person years. Abscissa: Mean bone dose in rem (RBE= lO). Open circles and dashed error bars: Patients given doses as children (less than 20 years of age); closed circles and solid er- ror bars: Patients given doses as adults (greater than 20 years of age). Error bars calculated as in previous graph (see Appendix IV). 131

There is not close agreement among the three studies, particularly the two groups exposed to the radium isotopes. If, however, a quality fac- tor of 7 is applied for radium-226 and a quality factor of 50 is applied for radium-224 (to take into account differences in surface dose as well as in LETT), then the relative risks for adults are 1.0,1.1, and 1.4 percent for the three groups of adults, and 1.9 for the children. The absolute risks would be 0.16, 0.11, and 0.10 for the adults. Such an analysis indicates that the sur- face alpha irradiation from 224Ra is about 7 times as effective as the alpha radiation from 226Ra> in reasonable agreement with animal experiments and with the analysis of Spiess and Mays cited above. The risk estimates from the 224Ra-injected patients are probably low because this group is still under study and substantially more cases are likely to appear. On the other hand, the risk estimates in the Argonne series are likely to be too high for low cumulative doses, principally because of the evident non-linearity of the dose-response curve. For the group of x-irra- diated spondylitis patients studied by Court Brown and Doll, more cases will probably occur with longer follow-up, but in this instance there is the problem of comparing x-ray data at high dose rates to data from internal alpha irradiation at low dose rates. It does not appear possible to define the risk of bone cancer with greater precision at this time, but it is worthwhile to emphasize again that each bone-seeking radionuclide will re- quire further evidence on which to base a quali- ty factor in determining the relevant rem dose to the sensitive cells. It is particularly impor- tant to obtain such information for plutonium, which is comparable to 224Ra in the distribu- tion of dose to the surface cells. For this pur- pose, animal experiments may be the only practical way to estimate risks, and we are for- tunate that a growing body of relevant experi- mental data already exists. REFERENCES (1) Bloch, C.: Postradiation Osteogenic Sarcoma. Report of a case and review of literature. Am. J. Roentgenol. 87:1157-1162,1962. (2) Pifer, J. W., Toyooka, E. T., Murray, R. W. Ames, W. R. and Hempelmann, L. H.: Neoplasms in Children Treated with X-rays for Thymic Enlargement, 1. Neoplasms and mortality.}. Nat. Cancer Inst. 31:1333-1356,1963. (3) International Commission on Radiological Protection. Publication 14. Radiosensitivity and Spatial Distribu- tion of Dose, Pergamon Press, Oxford, 1969. (4) Court Brown, W. M., and R. Doll, Mortality from Can- cer and other Causes After Radiotherapy for Ankylosing Sondylitis. Brit. Med. J. 2:1327-1332,1965. (5) Doll, R., Personal communication bringing up to date bone cancer data; December 1971. (6) Evans, R. D., A. T. Keane, R. J. Kolenkow, W. R. Neal, and M. M. Shanahan, Radiogenic Tumors in the Radium and Mesotherium Cases at M. I. T. In: Delayed Effects of Bone-Seeking Radionuclides. University of Utah Press, Salt Lake City, 1969, pp 157-191. (7) Finkel, A. J., C. E. Miller, and R. J. Hasterlik. Radium- Induced Tumors in Man, In: Delayed Effects of Bone- Seeking Radionuclides. University of Utah Press, Salt Lake City, 1969, pp 195-224. (X) Rowland, R. E., P. M. Failla, A. T. Keane, and A. F. Stehney, Sonic Dose-Response Relationships for Tumor Incidence in Radium Patients, ANL-7760, 1970, Part II, pp. 1-17. (9) Rowland, R. E., Personal communication bringing up to date bone cancer data, December 1971. (10) Muller, J., and J. Thomas, Strontium Retention in Man, In Delayed Effects of Bone-seeking Radionuclides, University of Utah Press, Salt Lake City, 1969, pp. 51-59. (11) Speiss, H., and C. W. Mays, Bone Cancers Induced by 224 Ra (THX) in Children and Adults, Health Physics 19:713-729, 1970. (12) Speiss, H., and C. W. Mays, Protraction effect on bone sarcoma induction of 224 Ra in children and adults. Ra- dionuclide Carcinogenesis Symposium, Richland, Wash- ington, May 11-13,1972. (13) Marshall, J. H., A Comprehensive Model of Alkaline Earth Metabolism: Preliminary Report, ANL-7760 1970. Part II, pp. 95-109. (14) Mays, C. W., Draft Report for Committee #31 Nation- al Council for Radiation Protection and Measurements, June, 1971. We are indebted to Dr. Mays for making this draft available. (15) Mays, C. W., T. F. Dougherty, G. N. Taylor, R. D. Lloyd, B. J. Stover, W. S. S. Jee, W. R. Christensen, J. H. Dougherty, and D. R. Atherton, Radiation Induced Bone Cancer in Beagles. In: Delayed Effects of Bone-Seeking Radionuclides, University of Utah Press, Salt Lake City, 1969. pp. 387-405. (16) Mays, C. W., and R. D. Lloyd, Bone Sarcoms Risk from 90Sr. Conference on Biomedical Implications of Radios- trontium Exposure. University of California at Davis, California, February 1971. (17) Mole, R. H., Endosteal Sensitivity to Tumor Induction by Radiation in Different Species: A Partial Answer to an Unsolved Question? In: Delayed Effects of Bone- Seeking Radionuclides, University of Utah Press, Salt Lake City, 1969, pp. 249-258. d. Skin The incidence of cutaneous cancer is in- creased following intensive irradiation of the skin, especially in the presence of chronic ra- diodermatitis. The types of neoplasms most commonly reported vary in frequency, depend- ing on site, dose, dose rate, and type of radia- tion (1-3). Both squamous-cell carcinomas and 132

basal-cell carcinomas have been noted, the lat- ter more commonly on the head and neck. Sar- comas of subcutaneous tissues, which are in- frequent, have been found most often in asso- ciation with long-standing and severe radioder- matitis. Dose Response Traenkle (2) has suggested that total doses greater than 1000 roentgens (R) are required to produce skin cancer. Sulzberger et al. (4), in the only prospective study of the incidence of ma- lignancy in patients receiving superficial ra- diation therapy for both benign and malignant conditions, found epitheliomata in 6 of 1000 U.S. patients irradiated previously and the same lesion in 9 of 1000 patients who had not been irradiated. They reported no sequelae be- low 1000 R and only mild chronic changes be- tween 1000 R and 2630 R. In contrast to the findings of Sulzberger et al. (4), Takahashi (5) has reported data sug- gesting that the relative risk of skin cancer in Japanese may be increased by 500-2000 R (Ta- ble d-1). In a retrospective statistical study on human cancer induced by radiation they ob- served 8923 patients with cancer, 207 of whom had skin cancer, as compared with 289 who had malignant lymphomas (skin cancer is relative- ly rare in Japan). For this entire group of pa- tients, the history of previous radiation was not different from that found in a control group of 11,556 persons. Subsequently, the authors selected 308 cases of skin cancer entering var- ious hospitals. Of these, 14 (4.55%) had re- ceived radiotherapy of the primary site (Table 1), whereas only 6 out of 762 (0.79%) in the con- trol group were so exposed. However, Taka- hashi,s finding of a relatively high risk of ra- diation-induced skin cancer among the Ja- panese (5) stands in contrast to data on the natural incidence of skin cancer, which indicate higher rates in white races than in nonwhite races (21). At the same time, a study of A-bomb survivors at Hiroshima reported in 1961 showed no evidence of radiation-induced accel- eration of age dependent changes in skin as measured by the appearance, elasticity, and looseness of the skin or by graying of the hair (14); moreover, no increase in skin cancer has been reported in atomic bomb survivors (15). Martin et al. (6) reported a relative risk of 3.74 in 649 irradiated U. S. patients but were not able to provide dose estimates. The doses were several thousand R or more in the few cases shown. In 2043 children treated by x-ray epilation for tinea capitis (ringworm of the scalp) Albert et al. (7) found 2 cases of basal cell skin can- cers, both in white males. Dose estimates ranged between 450 and 850 rad (8). No skin cancers were found in 1413 patients with tinea capitis who were not irradiated. A re-evaluation of this population by R. E. Albert in 1972 (9) now reveals 6 basal cell carci- nomas in the irradiated group of which 4 are in or on the edge of irradiated areas correspond- ing to doses of about 450 rad or more. Three of the six irradiated patients have other basal cell tumors not in the irradiated sites and thereare 2 basal cell carcinomas in the 1413 controls. The occurrence of 6/2043 cancers is not statis- tically different from 2/1413 cancers Ridley (10) reports retrospectively 6 cases of basal cell cancer of the scalp in white children aged 5-9. These occurred from 7 to 53 years af- ter treatment at doses of about 475 R. In the British study of long-term effects of irradiation in patients with ankylosing spon- dylitis, no deaths were found from skin cancer, even though the skin was included in heavily irradiated sites (11). Recently, five British pa- tients who had received 1000-8875 R to the spine and other joints for rheumatoid arthritis were found to have developed multiple basal-cell cancers, and in two other cases there were fi- broephitheliomata of Pinkus (12). In a brief note, Meara (13) noted six similar cases with multiple basal-cell epitheliomata, three of whom also had premalignant fibroepithe- liomata. At present, there is no way to deter- mine whether any of these patients were includ- ed in the original studies by Court Brown and Doll (11). In rats, the incidence of skin tumors induced by a single exposure to electrons, ranging from 230 to 10,000 rads, has been observed to rise abruptly between 1000 and 2000 rads, reach a peak of about 3000 rads, and fall rapidly with further increase in the dose (16). In mice, tumor induction following superficial beta irradiation has been reported to be proportional to the square of the dose (17). 133

Table d-1-Reiative Risks for Skin Cancer at Various Exposure Levels After Therapeutic Radiation (External Sources) (16) (Computed from data of Takahashi et al (5)) Estimated exposures (Roentgen) Proportion of cancer cases Proportion of controls Relative risk 95% limits of brackets 0, 500-2,000.. 2,000-4,000.. 4,000-6,000.. 6,000-8,000.. 8,000-10,000. 10,000. 95.45 (294) 0.97 (3) 0.97 (3) 0.65 (2) 0.65 (2) 0.97 (3) 0.32 CD 99.43 (4,044) 0.25 (10) 0.25 (10) 0.05 (2) 0.02 (1) 4.1 (1.2-9.6) 4.1 (1.2-9.6) 13.7 (1.8-100.0) 27.4 (2.5-300.0) 134

Host Factors The frequency of skin cancers appears to be related to the severity of pre-existing radioder- matitis, in that it is far more common (of the order of 10-28%) in severe cases and relatively uncommon (about 1%) in association with mild changes (2). Occasional cases of skin cancer have been reported in irradiated sites in the absence of clinical evidence of radiodermatitis. Whether these represent coincidental occur- rence or an effect of radiation cannot be deter- mined. The influence of pigmentation, which influ- ences susceptibility to ultraviolet irradiation and hence to naturally occurring skin cancer, is not known. Mechanisms The pathogenesis of cutaneous cancer is not fully understood, but clinical and experimental observations imply that gross injury of the skin greatly enhances the process (20). Cancer may thus be viewed as the end result of a series of changes, only some of which are detectable soon after irradiation. These changes, in order of increasing severity, are (1) threshold erythe- ma—a distinct reddening produced by vasodila- tation—(2) dry desquamation—loss of superfi- cial layers of epidermis—(3) moist desquama- tion—exudative reaction with loss of the basal layer of epidermis—and (4) necrosis, from der- mal destruction (18). As long as chronic ulceration is avoided, the skin usually returns to a nearly normal ap- pearance; however, clinically evident and per- manent changes occur after doses which pro- duce only dry desquamation. With severe inju- ry to the dermis, the changes also eventually include dermal fibrosis and endarteritis. Since rate and degree of change are affected by many factors (e.g., dose, dose rate, spatial distribu- tion of dose, region of body exposed, total area involved, blood supply, presence of irradiation, and the influence of drugs or other factors), these variables must be taken into account in considering the probability of injury attributa- ble to a given dose (18, 19). At non-necrotizing dose levels, radiation has been postulated to act as an "initiator" of the cancer process in mice, in a manner analogous to that in which certain carcinogenic chemicals have been observed experimentally to induce cutaneous tumors (22). According to this hy- pothesis, radiation is conceived to cause per- manent changes in cutaneous cells whose sub- sequent expression is enhanced by promoting factors which in themselves may not be carcin- ogenic. In the rat, irreversible radiation injury of hair follicles may be envisioned to act as such a promoting factor, in that the probability of radiation-induced skin tumors has been ob- served to depend heavily upon it (12). Whether an analogous model is applicable to carcino- genesis in human skin is speculative; however, the association between neoplasia and radio- dermatitis tends to argue for the possibility that gross injury of the skin contributes in some way to the evolution of the cancer proc- ess. Risk Estimate Although evidence suggests that the proba- bility of radiation-induced skin cancer is great- ly,increased in the presence of radiodermatitis, the data are insufficient to document the induc- tion of skin cancer at doses below the level re- quired to cause radiodermatitis, suggesting that the susceptibility of the skin to radiation carcinogenesis may be lower than that of cer- tain other tissues, such as the thyroid and the bone marrow. The possibility remains, howev- er, that the absence of recorded cases may be attributable to unusually long latency or to under-reporting of skin neoplasms. In the ab- sence of further data, numberical estimates of risks at low dose levels would not seem to be warranted. REFERENCES (1) Furth, J., and Lorenz, E: Carcinog-enesis by Ionizing Radiations In: Radiation Biology, A. Hollaender, Ed. New York, New York; Toronto, Canada; and London, England: McGraw-Hill Book Company, Inc., Vol. I, pp. 1145-1201, 1954. (2) Traenkle, H.L.: X-ray Induced Skin Cancer in Man. U.S. National Cancer Institute Monograph: Biolo- gy of Cutaneous Cancer 10:423-32,1963. (3) Cipollaro, A.C., and Crossland, P.M.: X-rays and Ra- dium in the Treatments of Diseases of the Skin, 5th Ed. Lea & Febiger, Philadelphia, pp. 391-398,1967. 135

(4) Sulzberger, M. B., Baer, R. L., and Barota, A.: Do Roentgen Ray Treatments as Given by Skin Specialists Produce Cancers or Other Sequelae? Arch. Dermat. & Syph. 65:639-655,1952. (5) Takahashi, S: A Statistical Study on Human Cancer Induced by Medical Irradiation. Acta Radiologica 23:1510-1530,1964 (Nippon Acta Radiologica) (6) Martin, H., Strong, E.; and Spiro, R. H.: Radiation Induced Skin Cancer of the Head and Neck. Cancer 25:61 - 71,1970. (7) Albert, R. E., and Omran, A. R. Follow-up Study of Pa- tients Treated by X-ray Epilation for Tinea capitis. Arch. Environ. Health 17:899-918, 1968. (8) Schulz, R. J., and Albert R. E. Ill-Dose to organs of the head from the X-ray treatment of tinea capitis. Arch. Environ. Health 17:935-950,1968. (9) Albert, R. E. Personal Communication (10) Ridley, C. M. Basal cell carcinoma following X-ray epilation of the scalp. Brit. J. Derm. 74:222-223, 1962. (11) Court Brown, W. M., and Doll, R.: Mortality from Cancer and Other Causes after Radiotherapy for Anky- losing Spondylitis. Brit. Med. J. 2:1327-1333,1965. (also pending UNSCEAR report). (12) Sarkany, I., Fountain, R. B. Evans, C. D., Morrison, R., and Star, L.: Multiple Basal-cell Epitheliomata Fol- lowing Radiotherapy of the Spine. Brit. J. Dermatol. 80:90-96,1968. (13) Meara, R. H., Epitheliomata after Radiotherapy of the Spine. Brit. J. Dermatol. 80:620,1968 (14) Hollingsworth, J. W., Ishii, G., and Conard, R. A.: Skin Aging and Hair Graying in Hiroshima. Geriatrics 16:27-36, 1961. (15) Johnson, M.L.T., Land, C.E., Gregory, P.B., Taura, Tadashi, Milton, R. C. Effects of Ionizing Radiation on the Skin, Hiroshima and Nagasaki. Atomic Bomb Cas- ualty Commission Technical Report 20-69,1969. (16) Burns, F. J., Albert R. E., and Heinbach, R. D.: The RBE for Skin Tumors and Hair Follicle Damage in the Rat Following Irradiation with Alpha Particles and Electrons. Radiation Res. 36:225-241,1968. (17) Hulse, E. V., Mole, R. H., Papworth, D. G.: Radiosensi- tivities of Cells from which Radiation Induced Skin Tumors are Derived. Int. J. Radiat. Biol. 14(5)437-444. 1968. (18) von Essen, C. E.: Radiation Tolerance of the Skin. Acta. Radiologica: Therapy, Physics, Biology, 8:311-330, 1969 (19) Rubin, P., and Casarett, G. W.: Clinical Radiation Pathology. W. B. Saunders, Philadelphia, Vol. I, pp. 62- 119, 1968. (20) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. Supplement 14 (A/5814), p. 105, 1964. (21) Doll, R., Muir, C.S., and Waterhouse, J.A.H.: Cancer Incidence in Five Continents. Vol. II, International Un- ion Against Cancer; Springer-Verlag, Berlin, 1970. (22) Lisco, H. Ducoff, H.S., and Baserga, R.: The Influence of Total Body X-irradiation on the Response of Mice to Methylcholanthrene Bulletin of the Johns Hopkins Hos- pital, 103:101-116. e. Breast Three different populations of women ex- posed to ionizing radiation have revealed an incidence of breast cancer in excess of that found in comparable nonirradiated popula- tions. The first of these populations, a series of tuberculous female patients in a Nova Scotia sanatorium, was first reported by the late Ian MacKenzie (1). At the time of his report, in 1965, his follow-up included 13 cases of breast cancer in 271 patients subjected to repeated chest fluoroscopy for artificial pneumothorax, as compared with only 1 case which developed in 570 patients who were not fluoroscoped. The study of these patients was later extended by Myrden and Hiltz (16), who reported 22 cases of breast cancer in 300 tuberculous women sub- jected to repeated fluoroscopies, as compared with 4 cases in 483 women not fluoroscoped. A significant increase in the incidence of breast cancer in female A-bomb survivors was first reported by Wanebo et al. (2), from a study of 12,003 women in the Adult Health Study sample. Examinations of these women from 1958 to 1966 revealed 5 definite cases of breast cancer in 5,540 women exposed to less than 9 rads or not in the city at the time of the bomb, as compared with 15 definite cases in 3,762 women exposed to doses larger than 10 rads. A chi-square test on the heavily irradiated women (>90 rads) versus the lightly irradiated women (<90 rads) revealed a significant in- crease at the 1% level. This elevated risk of developing breast cancer in the A-bomb survi- vors has now been confirmed in the latest re- port of mortality in the JNIH-ABCC Life Span Study sample (13, 18). However, a significant excess of deaths from breast cancer did not appear until the 1965-70 time period, when 19 deaths occurred in those exposed to doses of 10 rads or more, as compared with an expectation of 4.9 from the rate in the 0-9 rad control group. This 15-20 year minimum latent period is perhaps not surprising in view of the often long history of breast cancer from its detection to its fatal outcome. Finally, a significantly increased rate of breast cancer has been reported in women given localized x-ray treatments for acute post par- turn mastitis (3). In this series of 606 women, 13 cases were reported, as compared with 5.9 expected from New York State incidence fig- ures. Although the number of cases of cancer attributable to radiation in each of these popu- lations is not large, it is likely that the radia- tion was the causative agent, and from each study risk estimates can be derived as dis- cussed below. 136

Corroborative evidence that the increase in the number of breast cancers seen in human populations was induced by radiation comes from the demonstration of a carcinogenic effect of radiation on breast tissue in laboratory animals (4). Dose Response Data on the incidence of breast tumors in irradiated women are too meager to allow a precise evaluation of the dose response. The series with the largest number of cancer cases (and hence the most likely to provide informa- tion on the dose-reponse relationship) is the Nova Scotia study, the data from which are shown in Figure e-1. Although these data are consistent with linearity, they cannot be used as evidence for a linear dose-response curve owing to the extremely fractionated nature of the irradiation in this study. A steep dose-incidence curve has been ob- served for mammary adenocarcinomas and mammary adenofibromas (either alone or com- bined) in rats exposed to x rays or 60Co gamma rays (5,6), and for the overall incidence of all types of mammary neoplasms in rats exposed to fission neutrons (7), as judged one year after exposure. The dose-response curve for the com- bined incidence of all such tumors (malignant and benign) appears linear down to doses as low as 15 R of gamma rays. In all studies, the response tends to plateau in the high dose range. Although it is clear from these studies that radiation hastens the onset of mammary neoplasms, it is not certain whether there is a corresponding dose-dependent increase in the total number of tumors in rats observed throughout their entire life span, since the natural incidence rises sharply in aged con- trols. Mice exposed to whole-body radiation from a nuclear detonation showed an increase in total incidence of mammary carcinomas and sarco- mas at intermediate dose levels; however, at higher dose levels, the incidence plateaued and then decreased (8). Breast tumor development in irradiated mice is complicated in some, but not all, cases by the presence of radiation-in- duced ovarian granulosa cell tumors which may stimulate the growth of mammary tumors through the secretion of estrogen (9). Dose Rate Although there are no good quantitative data concerning the influence of dose rate on induction of breast cancer in women, a compar- ison of the risk estimates in Table e-1 indicates that extreme fractionation of the total dose makes little or no difference in the absolute risk per rad of developing cancer. For example, the risk estimates from the postpartum mastitis and the fluoroscopy series are indistinguisha- ble, despite the fact that the total dose divided by the total time in the latter series was at least an order of magnitude less than in the former. Perhaps an even more appropriate comparison, that of the A-bomb survivors (to whom the dose was delivered within seconds) with either of the two Western series, indicates that fractionation of the dose does not signifi- cantly reduce the absolute risk per rad of de- veloping breast cancer. Corroboration of this tentative conclusion comes from animal data: in rats, for instance, lowering of the dose rate of x or gamma irra- diation causes only minimal reduction of the oncogenicity; a dose rate sparing effect (from l0R/min to 0.03R/min) has been found only for the induction of mammary adenocarcinomas at a total dose of 265 R no dose-rate sparing effect having been found for the mammary fibroadeno- ma or the total mammary neoplastic response to any dose studied (10). Likewise, fractiona- tion of a dose of x rays into successive expo- sures delivered at a high dose rate has been observed to reduce its tumor-producing effec- tiveness only slightly (11). Host Factors Data from the JNIH-ABCC Life Span Study sample for the period 1965-70 reveal a marked decrease in the relative sensitivity of the breast to cancer induction with advancing age at the time of irradiation (18). Figure e-2 shows the ratio of breast cancer mortality in survi- vors exposed to 50+ rads, as compared with that in the 0-9 rad group, in terms of age at the time of the bomb (ATB) and age at death. If this same dependence of relative risk on age at the time of exposure were to hold for Western pop- ulations, the age-specific variation would dis- appear when judged in terms of the absolute 137

Figure e-1: Incidence of breast cancer per 103PYR (1966 data) against the number of fluoroscopies received. The error bars represent 90% confidence intervals, and the line is the best fitting weighted least squares regression line. 138

I I CO t- e 00 c u 5 4) rH 4) 4) s}u,.MiJU1o;> joquo j iL 4J m l 5 O CO 4J i-H 4) C -O iH c - sr c n £ "- C Cl "iJ • S2 a) « a. •O m r-i iu « in in H •H 4_ 3 . C B •- » 4) |s CO 0] OS cg g » o -H to a) ^ i 0 CN .3" O in i O /3-JS3UJ Jo sip gap 5 I n i-J Q O •-i CJv •O CO PH* ^ i in CN ,saa S ix O CN r-^ "d- 00 O TOJ Jad >(STJ -A] 'v[.-i u} .n.-c,.J,-.iiv ni.j.u.1,1 CO CN CO m co 00 O 00 r* O CO CN o o a an 5 r-i in i-* m - * ii O Ov CN sJT 1C ii i in O CN CO CN CN 6 a- 1OJ3uo3 JO m •a m TJ j| in I i 4) CM 4) -H 5. 4J 4J O nj O U s u OX o o M a i z w XaS 2 s Ph h k. (6JBdX) UBSW rn s c^ a CN g -f o 0 + CN CN Os O CO O in rx i-i o iJUS ti I 4 O3 i UV9U CN O •-i CN o 00 CO CN ' TBUJ95X3 § § M o i a8ui^i O vo Of i O 1 1 o I O o o o in r^ o m SJBdjC-UOSJdd io CO p- § vO O JO Jdquinjj s CN CO 00 3 m 01 m CO S3oa["qns 00 00 CO CN w io 00 01 O in in O CO sa^eaiT3sa ^F?I:X iptijrt uo "pe-i.ii -in i ;r po^Jaj r- CN CN O o (««X) U"aH S in in o a dn-noll°J -j 10 uoyiBJna CN CN dSue^ O in o* 5 m CN CN i CN m m m uo?3B1PBJ S b V b V Sou c o (3 10 uoyiLunn a) a) S >i •H 4J 4) S 3 uoT»TP« jo ad/! S t- + c t- + G X X ra335J5S CN CO 00 2 n m n uoT3oindod § + s A Z O «9 Jt U 4) 1 0 X-^ 0 3 Q. U J- •-1 O 4) Ov •r4 4J 4J C ** OQ Oi § + 3 jj -*4 s s o co "-H m 4J Z Oi g. 5- 1 8. 139

30 25 15 20 Q: i 15 10 Ageat Birth 0-9 10-19 Age at Death(Years} 30 -44 20-34 40-54 35-49 55-74 Figure e-2: Mortality ratios and 80% confidence intervals for deaths from breast cancer during 1965-70 in A-bomb survivors exposed to 50+ rads (from Ref. 13). 140

risk (since the spontaneous cancer incidence rises by a factor of approximately 4 in the United States from 30-44 to 55-74). No data are available on the role of hor- mones in the pathogenesis of breast cancer in humans, aside from the marked sex differences, but it is probable that they are important in this regard, in view of their known role in the treatment of the disease and also since it has been shown that estrogen and mammotropic hormone are involved in the pathogenesis of radiation-induced breast cancer in irradiated rodents (12). Mechanisms Although studies on human populations have been too limited thus far to contribute much to our understanding of the mechanism of radia- tion-induced mammary carcinogenesis, two tentative conclusions might be made. The first is that breast neoplasms, whether spontaneous or radiation-induced, appear to have a hormon- al requirement. Thus, in the ABCC Life Span Study sample, no cases of breast cancer devel- oped in the period 1965-70 in those aged 0-9 years ATB and exposed to 10+ rads, although 7 would have been expected had the absolute sen- sitivity been the same as in those 10-19 years ATB. Second, the limited human data imply that the pathogenesis of radiation-induced breast cancer in women may resemble that in animals, in which the findings support the multistage theory of carcinogenesis (15). It has been shown in rats, for example, that the mammary tissue itself must be irradiated for the primary, or initiating, step to occur (14). The secondary step is promoted by proliferative stimulation of the damaged cells by one or more of the mammotropic hormones of the ovarian-anter- ior pituitary axis. The interaction between mammotropic hormone stimulation and x-irra- diation has been shown to be synergistic in the induction of mammary neoplasms in the rat (12). Experimental studies with rats have shown an RBE of approximately 2 for fast neu- trons, for the induction of mammary gland tumors following exposure at relatively high doses. The RBE value for exposure at lower doses is higher, approximately 10 to 20 (7). Risk Estimates The data in the risk estimate table (Table e-1) are drawn from the following sources: A) Breast Cancer in A-bomb Survivors: Lines 1-2 of the table summarize the data ob- tained from death certificate analysis for the period 1960-70 (13, 18). Since there was no ex- cess of breast cancer deaths during 1960-64 in the irradiated (10+ rads) group, the data for 1965-70 have been analyzed separately (line 2). The best relative risk estimate from these data is a 3.5% increase in the cancer rate per rad and an absolute risk estimate of 2.9 deaths from breast cancer/106 women/year/rad if an RBE of 1 for neutrons is assumed. With an RBE for neutrons of 5, these estimates become 2.3% and 1.8/106/year/rad, respectively. B) Breast Cancer Following Multiple Flu- oroscopies: This was first reported by Mac- Kenzie (1) and the study was later extended by Myrden and Hiltz (16). It was found in both studies that women who were subjected to mul- tiple fluoroscopies during artificial pneumotho- rax for pulmonary tuberculosis later developed breast cancer at a much higher rate than did similar women not subjected to the fluorosco- pies. The Myrden and Hiltz study (I6) has a total of 783 tuberculous women in their 15-25 year follow-up, of whom 22 out of 300 given pneumothorax treatment developed breast cancer compared with only 4 cases out of 483 with no pneumothorax treatment. More recent data (Myrden, personal communication) show the necessity of revising these figures to allow for many patients who died within 10 years of treatment and for extra cases of breast cancer which have developed subsequently. Table e-2 shows these data broken down by number of fluoroscopies received and the follow-up period. These data are also shown graphically in Fig- ure e-1. The number of cancers in Table e-2 is correct up to September 1971, but the person years at risk are known at this point only to the time of the original study (16). If it is as- sumed that all patients alive in 1965-66 were still alive in 1971, the number of persons years (PYR) in the non-fluoroscoped group increased from 3,250 to 4,665 and in the fluoroscoped group from 2607.5 to 3707.5. Undoubtedly, the 1971 figures are better for calculating the absolute 141

C 3 e £ s 9 o, •§ & o w 01 I a 41 1 e u 01 e o «e o e 09 • s ft $ 3 >> iS C-J V rH 1 — N <D / N / — v / — v i Jtf) / — v • r? io 0 vO VO io Q> s-i >* ^ • in * io o io en oo in o io r- vo (j Qi pi f 'jj rj * io * rH • rH • iH • CM • | *H C* ""^ -rH i— i i io i OO l vO " io i io CTi u 3 Io s «»> vO OO io rH in io • C U oo • • • £• • in lH •— ' rH Oi o O CM CM rH v — / 15 is en rl in CM O H O iH £ o in in rH LO £ CM CM CM in O r- iH 00 oo CM 0 H io vO in *- CM •5 rt io CM io io r- in w -:* E m 1 in io CM o o O 00 CM io rH i o CM Q CM •g iH CM o io CM rH •H 0 Q HH c* v rt CM 0 rH rH rH CM w • rH ^J (X H PL, iH s io rH in io vO io l l % 00 rH rH rH iH O rH "S CM O CM io O iH nJ •iH i-J Q O ^ Pt) ±3 ctf O 0 O 0 CM O • H U £ 3= *t s iH 1 0 O O O O 0 1 O iH -2 0 CM CM rH in rH •H rH fcmg JO.H Q O O in in p. in 00 vO O rH 4_> io gmS nj 00 in in O • o 3 ^- o o oo io S ^ i-H io o in io in UN rH rH CM V) in in O in CM r-* . 0 (D HH ^H 0 p* rH rH io ^ O O g 3 rH 1 l vO 3 ,— i vO vO 0 CM o rH rH io U) 4-1 U ? 142

risk estimate and so have been used, even though they would tend to produce a slight: underestimate of this risk. The relative risk of developing breast cancer in the fluoroscoped group compared with the controls is 6.73 or 6.79 depending on whether the 1966 or 1971 PYR are used. Since the majority of the pa- tients had unilateral pneumothorax treatment, it follows that the breast on the treated side was exposed to more dose from the fluorosco- pies than was the other. This means that the relative risk of developing breast cancer has to be adjusted to compensate for this inequality of dose. From the data published by Myrden and Hiltz (16), it can be calculated that the re- lative risk should be increased by a factor of 1.551 bringing it to 10.5. The main problem in developing a risk esti- mate from these (and MacKenzie,s) data is the lack of a reliable radiation dose estimate. MacKenzie (1) tried to estimate the typical dose that might have been given and found that at the average setting of the x-ray equipment a dose rate of 22 to 55 R/min was delivered (de- pending on the presence or absence of a filter). Physicians were strongly advised never to ex- ceed an exposure of 10 seconds, but longer ex- posures were apparently not uncommon. It is likely that the majority of patients were sub- jected to radiation during a fluoroscopy of from 10 to 30 seconds (Myrden, personal communica- tion; Skavlem, personal communication). This range together with the range of 22 to 55 R/min leads to a mean dose per fluoroscopy of 10R with standard error limits of 5 to 20R. The mean number of fluoroscopies received by the women treated with pneumothorax and who survived for 10 or more years is 162, which gives an average dose estimate of 1,610 R. Since a few patients received more than 500 fluoro- scopies (=5,000 R or 6,000 rads to the skin using a backscatter factor of 1.2), and three patients developed radiation dermatitis, this dose esti- mate seems reasonable. Owing to the soft na- ture of the x rays, it is probable that a further correction to the dose should be made to correct for the attenuation of the x rays through the tissues overlying the breast tissue. Assuming the breast tissue lies at a minimum of 1 cm be- low the skin surface, the maximum dose to breast tissue would be approximately 75% of the air dose. Hence, the average dose2 estimate becomes 1,215 rads. These calculations lead to a relative risk of a 0.78% increase in the spontaneous rate per rad and an absolute risk estimate of 8.4 cases/106/ year/rad. i The correction factor has to take into account two phe- nomena: (i) The unilaterally treated patients received some radiation to the breast on the non-treated side, (ii) Some patients received bilateral treatment and, hence, had equal exposure to both breasts. The factors for each of these phenomena can be calculated as follows: (i) In the Myrden and Hiltz (16) study, 14 of the 17 patients who developed breast cancer following unilateral exposure, developed the tumor on the treated side. Assuming a linear dose effect curve and equal probability of developing cancer in ei- ther breast, it follows that this reflects the differ- ent doses to the two breasts. Hence, dose to "irra- diated" breast/dose to "unirradiated" breast is 14/3 or 1/0.214. Since we need to calculate the probability of a woman developing cancer, the relative risk for the unilaterally treated patients must be increased by a factor of 2/1.214 or 1.65. (ii) Of the 22 patients who developed cancer following pneumothorax treatment, 5 were bilaterally and 17 were unilaterally treated. Since the unilateral- ly treated patients would be expected to be under- represented in the group of cancer patients since these women received less radiation from the fluo- roscopy exposure, the true bilateral: unilateral treatment ratio should be 5:17 x 1.65 or 5:28. Assuming that the bilaterally treated patients have the correct relative risk, the final correction factor is: 5x1 + 28x1.65 33 2 It should be noted that although fluoroscopic proce- dures varied from place to place in Canada and the United States with respect to the patient either facing the x-ray tube or facing the physicians, in the Nova Scotia series the former position, i.e., the patient facing the x-ray tube, was adopted (1). This might well partially account for the fact that other centers have failed to report high incidences of mammary carcinomas subsequent to artificial pneumothor- ax treatment. 143

C) Breast Cancer Following Treatment for Postpartum Mastitis: Mettler et al. (3) de- scribed the 10 - 25 year follow-up of 606 women treated with x ray for acute postpartum masti- tis. They found 13 cases of breast cancer against 5.86 expected. The majority of women (513/606) were aged 20 - 34 at the time of treat- ment and received an average air dose of 211 R to both breasts. Converting this dose to rads yields an average dose to both breasts of 200 rads, which gives rise to a relative risk esti- mate of a 0.61% increase per rad and an abso- lute risk of 3.1 cases/106/year/rad. However, these calculations ignore the fact that approxi- mately half of the PYR occurred in the first 10 years of follow-up during which time only 2 cases of breast cancer developed (1.6 expected). Since a minimum latent period of 10 years is consistent not only with these data but also with the above studies, it is essential in calcu- lating the risk estimates to derive new data ex- cluding the first 10 years of follow-up both for the PYR and for the expected number of cases. This can be done from the age distribution of patients at first treatment, provided it is as- sumed that each age group has the same mean follow-up. Such a calculation gives a value of 5,606 for the PYR and an expected number of cases of breast cancer of 4.23 (from the age- specific cancer rates in upstate New York for 1958-60). These data lead to a relative risk esti- mate of a 0.80% increase in the cancer inci- dence per rad and an absolute risk estimate of 6.0 cases/106/year/rad. A legitimate objection to the calculation of risk estimates from this study is the uncer- tainty as to whether the general population constitutes an adequate control; or in other words, does acute postpartum mastitis predis- pose to breast cancer? Although such acute infectious processes are not usually believed to be associated with subsequent development of cancer, women with so-called chronic cystic mastitis are more prone than the general popu- lation to have breast carcinomas. In this study, approximately half of the 38 women subjected to breast surgery for neoplasms were reported to have chronic cystic mastitis. What role this plays in the findings is not understood, but for lack of good evidence to the contrary, it has been supposed for purposes of risk estimation that the excess cases were radiation-induced and that the risk estimates are valid. Summary Despite the number of assumptions involved in calculating risk estimates and the paucity of cases in the above studies, the following tenta- tive conclusions can be drawn: (a) If an RBE of 1 for the neutron compo- nent at Hiroshima is assumed, the absolute risk estimates from the 4 studies are remarkably close. For example, if it is assumed that a fac- tor of 2 can be applied to correct deaths from, to incidence of, breast cancer in Japanese wom- en, then the estimated values of the absolute risk, in cases/106women/year/rad, are 6.0 for the Japanese study, and 8.4 for the two studies of Western populations. None of these is signif- icantly different from any other. On the other hand, the relative risk estimates are signifi- cantly higher for the Japanese women, reflect- ing their much lower natural incidence of breast cancer, as compared with Western wom- en. (b) If an RBE of 5 for the neutron compo- nent in Hiroshima is assumed, then neither the absolute nor the relative risk estimates for the Japanese would appear to agree with those of the two Western studies. (c) Since the two Western studies give close agreement, both in absolute and relative risk estimates, and since the major interest of this analysis is the development of risk estimates for the U.S., it seems appropriate to focus on these two series to obtain an overall best esti- mate of the risk. The high degree of uncertain- ty in the dose estimate for the fluoroscopy se- ries makes this estimate less reliable than that from the post-partum mastitis patients, and so the value of 6 cases of breast cancer/106 wom- en/rad (or rem) has been chosen as the best esti- mate of the absolute risk. Since the age-adjust- ed annual incidence of breast cancer in U.S. women is 72/105 (17), the above absolute risk corresponds to a doubling dose of 120 rads, or an 0.83% increase per rad in the spontaneous incidence. Reasonable high and low estimates might lie within a factor of 2 on either side of these values. 144

(17) Preliminary Report of the Third National Cancer Survey, Biometry Branch National Cancer Institute, 1971. (18) Jablon, S. and H. Kato. Mortality among A-bomb Survivors, 1950-1970. JNIH-ABCC Life Span Study, Report 6. Technical Report 10-71,1971. f. Lung Introduction There has been a worldwide increase in the incidence of bronchial cancer within the last few decades, pointing to the sensitivity of the renewal cells of the respiratory epithelium to carinogenic influences in the environment (1). The increase in lung cancer is not uniform throughout the world, nor can it, in all in- stances, be directly correlated with cigarette smoking. Other factors such as air pollution evidently also play a role. These considerations are important because of the evidence that bronchial cancers associated with occupation- al irradiation may vary in frequency, depend- ing on whether other environmental factors are also present. Radiation protection stand- ards for the general public must allow for the possibility that a significant fraction of the human population will be exposed to cigarette smoke by direct inhalation, as well as to the other less well-defined environmental carcino- genic factors. The principal series of radiation-induced lung cancers has been observed in underground miners exposed to radon decay products in the mine atmosphere. From the multiplicity of oc- cupational exposure conditions that have been associated with an increased incidence of bronchial cancer, however, it is evident that many other types of carcinogens, besides ra- diation, can also induce bronchial cancer. Chemical agents included in this category are asbestos, chromium salts, mustard gas, hema- tite, nickel and arsenic compounds, and asphalt derivatives (2-8). Radon daughters and asbes- tos appear to be most strongly carcinogenic in association with cigarette smoking. REFERENCES (1) MacKenzie, L, Breast Cancer Following Multiple Fluo- roscopies. Brit. J. Can. 19:1-8,1965. (2) Wanebo, C. K., K. G. Johnson, K. Sato, and T. W. Thorslund: Breast Cancer after Exposure to the Atomic Bombings of Hiroshima and Nagasaki. New Eng. J. Med. 279:667-671,1968. (3) Mettler, F. A.. L. H. Hempelmann. A. M. Dutton, J. W. Pifer, E. T. Toyooka, and W. R. Ames: Breast Neoplasms in Women Treated with X-rays for Acute Post-partum Mastitis. A Pilot Study. J. Nat. Can. Inst., 43: 803-811, 1969. (4) Casarett, G. W.: Experimental Radiation Carcinogene- sis. Prog. Exp. Tumor Res. 7:49-82,1965. (5) Bond, V. P., E. P. Cronkite, S. W. Lippincott, and C. J. Shellabarger: Studies of Radiation-induced Mammary Gland Neoplasia in the Rat. III. Relation of the Neoplas- tic Response to Dose of Total-body Radiation. Rad. Res. 12:276-285,1960. (6) Shellabarger, C. J., V. P. Bond, E. P. Cronkite, and G. E. Aponte: Relationship of Dose of Total-body 60Co Ra- diation to Incidence of Mammary Neoplasia in Female Rats. In Radiation-Induced Cancer, Intern. Atomic Ener- gy Agency, pp. 167-172, 1969. (7) Vogel, H. H.: Mammary Gland Neoplasms after Fis- sion Neutron Irradiation. Nature 222:1279-1281,1969. (8) Upton, A. C., A. W. Kimball, J. Furth, K. W. Christen- berry, and W. H. Benedict: Some Delayed Effects of Atom-bomb Radiations in Mice. Cancer. Res. 20 (8, pt.2), 1-60,1960. (9) Lorenz, E.: Some Biologic Effects of Long-continued Irradiation. Am. J. Roentg. 63:176-185,1950. (10) Shellabarger, C.J., and R. D. Brown: Rat Mammary Neoplasia Following 60Co Irradiation at 0.03 R or 10R per Minute. Abstract Ed-3, Radiation Research Society Meeting, Portland, Oregon, May 14-18, 1972. (11) Shellabarger, C.J., V. P. Bond, G. E. Aponte, and E. P. Cronkite: Results of Fractionation and Protraction of Total-body Radiation on Rat Mammary Neoplasia Can. Res. 26:509-513,1966. (12) Yakoro, K., J. Furth, and N. Haran-Ghera: Induction of Mammatropic Pituitary Tumors by X-rays in Rats and Mice: The Role of Mammotropes in the Development of Mammary Tumors. Can Res. 21:178-186, 1961. (13) Jablon, S., and H. Kato: Studies of the Mortality of A- bomb Survivors, No. 5. Radiation Dose and Mortality, 1950-1970. Rad. Res. 50:649-698, 1972. (14) Furth, J.: Radiation Neoplasia and Endocrine Sys- tems. In Radiation Biology and Cancer. Univ. of Texas Press, Austin, Texas, pp. 7-25,1959. (15) Bond, V. P., C. J. Shellabarger, E. P. Cronkite, and T. M. Fleidner: Studies on Radiation-induced Mammary Gland Neoplasia in the Rat. V. Induction by Localized- Irradiation. Rad. Res. 13:318-328,1960. (16) Myrden, J. A., and J. E. Hiltz: Breast Cancer Follow- ing Multiple Fluoroscopies during Artificial Pneumo- thorax Treatment of Pulmonary Tuberculosis. Canad. Med. Assn. J. 100:1032-1034,1969. 419-797 O - 72 - ll 145

There has been considerable discussion of the comparability of the different types of tumors associated with environmental agents. The epidemiology and histologic types of these tumors have been reviewed by Berg (9) and Kreyberg (10). The bronchial and parenchymal respiratory cancers in man are generally divid- ed into two major classifications. The first group comprises adenocarcinomas of the bron- choalveolar type, as well as special types of tumors such as carcinoids. The second category includes epidermoid carcinomas and small- and large-cell anaplastic epithelial tumors of the proximal portion of the bronchial tree. The first group of tumors is the most common in nonsmokers, while the latter group of tumors are those particularly associated with ciga- rette smoking. The type associated with expo- sure to radiation, arsenic, nickel, chromium, hematite, mustard gas, and asbestos is similar to that associated with cigarette smoking (4, 11), which is not astonishing inasmuch as ciga- rette smoking generally plays an important contributory role in their development. It should be emphasized, however, that there is considerable overlap in the distribution of the different types of lung cancers, regardless of the presence or absence of environmental fac- tors. A recent analysis by Saccomanno and col- leagues (12) of 150 cases of lung cancer among uranium miners has shown that the predomi- nant cancer types among individuals with the highest radon-daughter exposures are the small-cell and undifferentiated types, consti- tuting about 75% of all lung cancers in the higher dose categories. The possibility exists that the cells of origin of the epidermoid can- cers are different from those of the small-cell cancers (13), but the existence of two such pop- ulations of origin in normal tissue remains to be established. Because of the presence of a number of po- tential occupational carcinogens in the dust of underground mines, there has been some ques- tion as to whether radon and radon daughters constitute the principal cause of increased risk among these miners. Pertinent to this issue is the fact that underground mining per se does not necessarily lead to an increase of lung can- cer risk, a fact that has been well documented for underground coal miners in the United Kingdom (14). A recent study has investigated 5,500 potash miners in New Mexico, working in mines not associated with elevated concentra- tions of radon-daughter products in the air, and has shown no increased risk in such below- ground miners as compared with above-ground workers (15); (in both groups excess cigarette smoking could account for the increased lung cancer compared to the general population). It is pertinent to point out that in those mining operations where a significant increase in re- spiratory cancer has been associated with in- halation of radon and its daugher products, the mineral constituents being mined were widely variable. Besides the uranium miners in Europe and the U.S.A. (15, 16), excess respiratory can- cer risk has been found among underground metal miners (IT), fluorspar miners (18, 36), and hematite miners (5). In each of these popu- lations, there was occupational exposure to increased concentrations of radon which was also present in the mines. Thus, whether or not other agents such as arsenic, uranium, or fluoride may have been present in the air, the one constant relationship in these groups has been radon-daughter exposure and the inci- dence of lung cancer (15). In the early studies of the Bohemian pitchblende industry (19), some of the employees in milling operations devel- oped lung cancer, as did miners, but their expo- sures to radon and radon daughters, while probably significant, are not known with accu- racy. In the U.S., uranium mill workers have not experienced an increased risk of lung can- cer (15), presumably because good ventilation minimizes their exposure to radon daughters. Relationship of Cancer Rate to Radiation Expo- sure In view of the importance of control of lung cancer among underground miners, especially in the uranium industry, vigorous efforts at establishing a dose-response relationship have been undertaken. In the U.S. a study of approx- imately 4,000 uranium miners has been carried out by the U.S. Public Health Service, particu- larly dated from 1957. A current report of this continuing study was submitted to an ad hoc 146

subcommittee of the Advisory Committee to the Federal Radiation Council by Lundin, Wagoner and Archer (20). In addition, a report to the Interagency Uranium Mining Review Group has recently been prepared (15), with a summa- ry of cases through September 1969. Finally, Dr. Archer has made available data on all cases of cancer in the uranium mining study group identified through March 1971. Although most of the evidence relating radia- tion exposure to lung cancer in man pertains to internally deposited alpha-emitting radionu- clides, such as radon daughters and thoron and its short-lived daughters, as summarized by Lundin and coworkers (15), there is some evi- dence of an excess lung cancer rate in individu- als exposed to gamma and x radiation. Among the survivors of the atomic bombing in Hiroshi- ma and Nagasaki, data are now available for the period up to 1970 (21), which show the rela- tive risk of cancer of the tracheobronchial tree for the period of 1955 to 1970 to be 1.4 times higher for doses of 10 rads or more than for lower doses. Difficulties exist in interpreting these data, however, one of which is the fact that in the control group (i.e., those farthest from ground zero) the observed cancer rate was about 50% higher than that expected for the Japanese at large. In addition, there is the question of neutron irradiation in the exposed individuals, which may have contributed signif- icantly to the observed effects in view of the possibly high RBE of this component of the total dose. An approximately two-fold increase in the relative risk of lung cancer was observed in the study by Court Brown and Doll (22) of patients with ankylosing spondylitis treated with x-ray therapy. In these cases large doses of x rays were delivered to the spine, and doses to the bronchial epithelium were estimated to aver- age about 400 rads (23). In a study of patients with tuberculosis, whether active or inactive, an increase in lung cancer of from 5- to 10-fold was found in com- parison to the incidence in the general popula- tion (24). The possibility has been raised that the patients may have been exposed to fluoros- copy during treatment of the disease, and that this may account for their increased risk (24). In the absence of specific exposure information, however, and in view of the fact that there could also be a relationship between tubercu- losis itself and the likelihood of developing lung cancer (25), little emphasis can be given to this study at present. The incidence of lung cancer in x-ray techni- cians has been compared with that in pharma- cy and medical technicians in the U. S. military service during World War II (26). Out of ap- proximately 13,000 individuals who were pre- sent in both groups, 17 deaths from respiratory cancer were observed among the x-ray techni- cians as compared with four among the other groups. This difference is highly significant, but when the groups were compared with ap- propriate U. S. mortality statistics, a total of 12.4 cancers was expected for the x-ray techni- cians, which was not significantly lower than the 17 cases observed. Thus the difference be- tween groups may be due primarily to a de- creased lung cancer incidence among the phar- macy and medical technicians, which is para- doxical and which complicates interpretation of the data. The experience through 1967 for all of the various underground mining groups in which an increased risk of cancer has been found, has been summarized by Archer and Lundin (27), and Archer has updated this summary to Sep- tember 1969 in the report for the Interagency Uranium Mining Review Group (15). Central to an interpretation of data from underground miners are a number of fundamental issues, which include the following: (a) What expo- sures to radon daughters have actually oc- curred? (b) What is the rad dose to the critical cells from radon daughters in the air? (c) Is an increased risk observed at a dose rate below that equivalent to continuous occupational exposure to one working level of radon and radon daughters? (d) Is the dose-response curve at low doses linear, is it concave downward (i.e., giving a higher risk per rad at lower cumula- tive doses than at higher cumulative doses) or does a true threshold for cancer production from a cumulative dose exist? Considerable effort has been made to eval- uate the radiation exposures of the various groups of miners in the Colorado Plateau area, with particular emphasis on previous under- ground mining experience not included in the category of uranium mining (a substantial number of the miners had such experience). 147

Absent or infrequent sampling of air of some of the mines, especially in the early exposure prior to 1950, makes estimates of cumulative dose only approximate at best, but it is unlikely that these estimates can greatly be improved at this time, and it is probable that in the ag- gregate the estimates of exposure are adequate to determine trends in the data. It should be emphasized that among these miners the dose rate was quite high in comparison to that in some of the other mining groups (about 10 Working Levels on the average, see below). With regard to the relationship between the WLM and the rad dose to the basal cell layer of respiratory epithelium from inhalation of ra- don and radon daughters, the literature has been recently reviewed by Walsh (28) and by the Interagency Uranium Mining Review Group (15), with essential agreement between both reviews. One "Working Level" (WL) in air is defined as any combination of short-lived radon daughters (through polonium-214, RaC,) leading to total emission of 1.3 x 105 Mev of alpha energy per liter, and the cumulative measurement of Working Level Month (WLM) is denned as exposure at the rate of 1 WL for 170 hours. There' has been criticism of the WL as an exposure index, because the state of equi- librium of the various nuclides in the chain is critical, especially with regard to the fraction present as free ions. This latter criticism re- mains valid, but it is fair to say that samples of mine air usually show relatively little contribu- tion of unbound radon daughters. Estimates of the rad dose/WLM for basal cell layers of different segments of the bronchial epithelium have varied widely, from less than 0.1 rad/WLM to as much as 20 rad/WLM (37). A critical factor in these estimates is the thick- ness of the epithelial and mucous layers, an uncertain quantity in smokers with some de- gree of chronic bronchitis. The unpublished studies of Gastineau (20) indicate that the normal epithelium of segmental and more prox- imal bronchi, where most radiogenic cancers have arisen, is thicker than had previously been assumed. On the basis of the present evidence, 1 rad/ WLM is probably close to the upper limit for a reasonably uniform dose to the basal cell layer of the epithelium of the larger bronchi on a probabilistic basis. In the presence of existing chronic bronchitis, the dose factor may well be substantially lower, owing to increased thick- ness of the mucous layer as well as of the epi- thelium, and thus a figure of 0.5 rad/WLM has been adopted for this report. It should be em- phasized that uncertainties in this value are probably greater than for the working level measurements themselves in determining risks per rad for the mining populations. So far as a limiting dose rate is concerned, the question is whether continuous exposure to less than 1 WL has been found in miners to be associated with increase in lung cancer risk. The problem is related to the possible influence of dose rate on latent period, and if latent peri- ods of 20 to 30 years are found at the lowest exposures, no mining group has been under observation with known exposures at these levels for a long enough time to provide a defi- nitive answer. The metal miners studied by Wagoner et al. (17) showed a cancer rate about three times that expected, with exposures at the time of the study well below a concentra- tion of 1 WL, but these authors indicate that earlier exposures before the mines were venti- lated may well have been higher. The hematite miners studied by Boyd et ai. (5), who have shown a risk of about 1.7 compared with con- trols, worked in mines where the radon concen- trations are equivalent to WL concentrations of 1 WL or less, but until measurements of ac- tual radon daughter exposures and the influ- ence of the hematite itself are determined, no final conclusion is possible. For the Colorado Plateau uranium miners in the lowest cumula- tive WLM exposure category whose dosage was usually received from several short periods of high working level exposures, no significant excess of cancer has appeared as yet (see be- low). The question of the linearity of the dose-re- sponse relationship and whether a true thresh- old is present has been discussed thoroughly by the ad hoc Committee report (20) and the Inter- agency Uranium Mining Review Group report (15). At present, the fact that the lowest expo- sure group shows only a slight increase in can- cer rate above that expected makes the Colora- do Plateau group inadequate to resolve this issue. Inspection of the composition of the study population indicates that the population at risk in this dose range (120 WLM) is now so small as to make it unlikely that even future follow-up will settle the matter. 148

There has been observed in the U.S. Colorado Plateau workers an inverse relationship be- tween cumulative radiation dose and the latent period for cancer after initial exposure in the mines, but this effect is not very striking at the present time. The relationship of cigarette smoking to the latency period for lung cancer among uranium miners is not known. Experimental Bronchial Cancer in Animals A large body of experimental work has now been assembled relating the occurrence of lung cancer to ionizing radiation in animals, and has been summarized by Sanders, et al. (29) and by the Interagency Uranium Mining Review Group (15). Although lung tumors are readily induced in animals by radiation exposure, not all of these may be relevant to the human dis- ease, since peripheral adenocarcinomas are much more likely to occur in animals from whatever inciting stimulus is applied than are tumors comparable to squamous cell tumors in man. For alpha-emitters the lowest cumulative dose at which a rise in lung cancer has been observed experimentally was in rats given po- lonium-210 with a sodium chloride aerosol by inhalation (30). In this experiment one squa- mous cell cancer occurred after 70 rads cumula- tive mean lung dose, and the dose-incidence re- lation was approximately linear at higher dos- es. For beta-emitters the lowest dose associat- ed with cancer induction was approximately 600 rads, in rats given cesium-144 salts by in- tratracheal instillation (31). In these experi- ments the dose-response curve appeared to be curvilinear (concave upward). An inherent diffi- culty in animal experiments, of course, is the short life span of the small rodents usually used and thus the fact that only the cancers with short latent periods may be detected by this approach, a limitation which might be ex- pected to produce a curvilinear dose-response curve of the kind observed. An important issue is whether local, or "hot spot", doses are more effective in producing cancer in the respiratory tract than is uniform radiation exposure to the entire epithelium. Experiments of Grossman and Little (32), in which polonium-210 chloride was given intra- tracheally, with and without hematite parti- cles, are pertinent to this issue. Previous expe- riments have shown that when polonium-210 was given simultaneously with hematite, the incidence of tumors was related to the polonium concentration, with a latent period as short as 15 weeks, depending on the total cumulative alpha radiation dose (33). In the more recent study, polonium and hematite were given either on alternate days or no hematite was given at all. Since polonium solution alone was as effec- tive as polonium given with hematite, it may be inferred that a higher localized dose from al- pha particles was not more cancerogenic than the same mean tissue dose delivered more uni- formly to critical cells. Host Factors and Mechanism of Action of Radiogenic Lung Cancer It has been pointed out above that a number of environmental factors may influence the development of bronchial cancer in individuals exposed to radiation. The lower incidence of lung cancer in females than in males may pre- sumably be due in part to differences in expo- sure to these factors, the most obvious of which is cigarette smoking. In addition, however, the contribution of other environmental factors, such as carcinogens in air pollution, occupa- tional inhalation of asbestos fibers, or systemic carcinogenic factors such as nitrosamines must be considered (34). Other host factors, such as may influence susceptibility to chronic lung disease, for example, o^-antitrypsin defi- ciency, may be mentioned but are not yet ade- quately evaluated in relation to cancer. It has been postulated that cancer produc- tion in the bronchial epithelium involves meta- plastic changes in the tissue induced by non- specific irritants, such as phenols or sulfur dioxide exposure. That is, a chronic inflamma- tion is established in which the carcinogenic potential of inhaled carcinogens is subsequent- ly brought out. In this case the transformation by initiators, such as ionizing radiation, be- comes manifest in an overt cancer, possibly arising at several points in the lungs simulta- neously. An important unresolved issue is the question of whether the radiation exposure to local areas is the critical datum or whether an effect extending over the entire respiratory epithelium is more likely to lead to cancer. This is important because lung cancers usually ar- ise at bifurcations of the bronchial tree. Most 149

analyses have concluded that the issue is basi- cally a probabilistic one, in that a more widespread exposure is likely to subject more cells to the carcinomatous transformation. Also at issue is the critical number of cells which must be affected within a single region, and various theoretical models have been applied to this, such as that of Bevan and Haque (35), whose speculative analysis concludes that somewhere between 15 and 20 cells in a particu- lar region must be traversed by alpha radia- tion in order to produce the cancer transforma- tion. Summary of Human Data and Estimates of Risks of Bronchial Cancer From Radiation Tables f-1 and f-2 summarize data obtained in six human populations in which it is possible to estimate the risk of lung cancer from radia- tion exposure. The data for the U.S. uranium miners, Newfoundland fluorspar miners, and the Hiroshima and Nagasaki survivors have been analyzed in terms of dose-response rela- tionships, since it has been possible to subdi- vide these groups by dose categories. The dose response data are given in Figs, f-1, f-2, and f-3. In calculating the slope of the curve on each figure, the data points are weighted for the number of person-years at each point. For the first two mining groups, a straight line through the origin provides the best fit of the data, as might be expected for alpha-radiation expo- sure. In the case of the Japanese survivors, the four dose levels give somewhat erratic results, but the lowest dose range (10-49 rads) gives a higher risk than would be predicted for a linear fit to all the points, and thus there is no evi- dence of a "threshold" for this group. The underground metal miners and the thoro- trast patients are not considered to be as relia- ble for risk estimates as the other groups, be- cause the dose estimates to the bronchial epi- thelium are even more uncertain than in the other study groups. For this reason they have been excluded from the following summary (al- though their inclusion would not greatly alter the results). SUMMARY OF RISK ESTIMATES FOR BRONCHIAL CANCER Adults only, and with cigarette smoking assumed to be characteristic of these populations. Absolute Risk Cases/106/years per remi Mean Bronchial Dose Relative Risk % Increase in Rate/Yr. per remi Mean Bronchial Dose Uranium Miners (white only) Fluorspar Miners Spondylitis Patients Hiroshima & Nagasaki Survivors 0.63 1.61 1.2 0.60 0.18 0.61 0.19 0.19 Average 1.0 0.29 *Conversion to rem based on an RBE of 10 alpha particles (miners) and 5 for neutrons (Hiroshima and Nagasaki survi- vors), with the fraction of the rad dose assigned to neutrons taken from the T65 calculations. 150

j £ on s See footnote (b) S See footnote (b) eiuaimioD 131)30 10 sa3oujooj 9 o. g a i 5, i i I «/«-V9OT . gg I -. p* 2 R •a O CO /»•» Jo sqnop ^ j|H 'Wa Mn10.qV ^ 3° ltH o sO n ii o cn 5 m 3 sag § m cn r^ sO rx 4-1 m e so * c 0aJ Jdd HSIJ tiAI4l-IOJ OJ aSBflJJUT JU.1ZUOJ OO H 00 0 iH O 0 •o «-l m 3 o 0 0 O 0 1-1 a I Olo; ^ • ^ * •Q (3/0) 1«TJ aAT3e-[»H o m so < :• 1 o S sO *"i 1 N CU Arizona. Colorado. N. Mexiyo, Utah jli i U •i II Eleytion. S JO n • *a VH 2i C • K bo • S < o z => B-S*.S « r-l -H < 0 X O w > m "S 2 X X X x X • r-i 0S «5 u 3 •n u ** oS C CM 8$ Rl R| i 1 (spea) asoa •mrrn 5 S r». 3 CN i-1 n 0 '••MI gl t| 2 1 «) c V* m * si m *3 "o jo JaqnmN s 1. N M \ I o n 1 R O •Q ^i C jo S J % •i 1 •n u •«" •1) + (sJBdX) pJSBq M* i p. i il § p* X X m II £3 CO <1e9H s - H H N o cn 1 ' -.— *, g S CO I m 1 IX. r- i sO 3Inso3xa 2 s B 1 i JO UOT3BJIXJ 1 X 4) uoi;ieTpBa JO oHAj c 8 8 a B 0 Q "U"'J'H M m 2 2 s H J= i0 5 2 S 1 5s? IS?? L. s B 25 Xpiug £ 0 C O c a -H o a B S C C C 0 1 -rl O iN Hi B" 2 ! • *2 i 9 &Sc- * 0 « -H - •O -^ b 0 T3 4 0 ss f r 0 ii C £ •a o o c m Li O 188.2 «- = « £ o a £ 00 0 QD ^t i so S- §2 . d) -H • M m ^IS g. M ° • eS 8 S « - f- *i 0 *s| *^ -H 5 CN W n ia i^ C - 01 4J -H r- 13 •« 09 £0 0 Li • £ a i-H SH - J 4J •-• • b C io 0 0 4 •a *H c 0 u C c > •»< • Ll « -^ o a c O r4 C *O 0 0 •3 £ a '^ J 0 0 •O -O iH - 8C 3 44 -^ O 4J Ll 3 — CO 4J 4J 0 0 U U -ct ' U & 60 C i-• T3 a c *H a) iij no — 5 4t Jt §0 « a i o B T3 4 4-1 C - C 03 U 9* c « o m •-• -H QJ -^ C • O 4J 0 O 4-1 O u ao n — sir a. ". q} > ( i a) c i J5 o -mm ^'S it ai . " ~ * S "5 i t S S - § S« s Mll'i I <u u -•5»8S|. .M5&S"! U 0 O 2* c£ J /^ 3 ^ - m io -H M i-» qj 4) 3 kl r-i * 0 3- S g T-S •° -2 o -" 3 § c C » V • •••%£ c -4 « x a 01 O • lO 4J O. iw Jl -i J g. 5 151

M O u s * "S 1 > CO 5 *e a n M | C -i BJUOiUUOO J3I{3O JO S8]OU300^ 41 U i IM JJ V 1 i-l o * 41 V CO M I L 9 J5 * * ,*. ^ *. 00 CN /Q-JtE3 JO aqlC.lp _6 | iJ 5 C CO 8 g S r-l 0 3tda Os VI M ;. s rM Cf OSB.U3UJ 3l1a3Ja,1 00 m o 0 CN i-i m 2. 2 ~ - « • 1 CN u~i -» N. (3/0) T-"'M -iAHL-I;'>1 I r-4 • r^ -cj Portugal -0 « m JO c u •-' « -< C3 8 xaS 3 x 4 fc I J) s x •* *• (•aeaX) UBa» B u in ^ S 3B »*» 98UBH CN 3 J.A * V •-i ift 03 i ULV-H S US g CO o c- oS si in r^ 1 CN * m JO Jdqiunf^ CT. a f« 00 S •-* s3oafqno jo laqomfj 0 r-i CN n CN paseq aJ8 01 i Os SJ3BU1J3SS 1(87J qjIqn r-. 3 «fl ss uo 'pojiJ| J33JB po?Jaj o io •-O i*.- UBaH S o 3 •n CN CN jo uoi3BJna a^uF^ iTi •-i r*. « CN m aJnaodxa 5 e 5,0 S « <j « s JO UOT3HJnQ >. * 5L V uoT^oTpDj jo adXi c a X ^ + o aauaaajaH 3 io in - ThoroCraat systemic Injeytion 1935-. 53- Xpn3S X io O ^ |- 3 •^ oS5 .A & S S < as S a? • w iJ 4J c n a) a V *-! O.iu 00 n a o c lp.H O U 4) » •^1 bo B *i-i 4J 3 b .^ n ~* o 4J O 4 •^4 iM 4J i• • O -HO J iw O o 4J 3 U O 9 ^ a 4) o a x 4J u a) ^ *O v 4 ) -i-i C U £ U • s.H« 1. .1 S.2l« iN O •O b iM « CO C0 .3 ^ g g- g^'g 4 •-* 4) ^ ;: u f6 ,j O -u o Stf S *M Ii §4) **i U b iJ 4) •O « 4 ^ U 3 4J 4) 0) C 4i Pi « 4) £ 152

300 O CO cc LU D. 200 LU Q 100 7 RISK OF LUNG CANCER HIROSHIMA AND NAGASAKI SURVIVORS 1955-1971 O O O CONTROL: EXPOSED GROUP 0-9 RAD 100 200 300 DOSE (RAD) Fig- f-': Dose-response data for lung cancer in Hiroshima-Nagasaki survivors (21). Ordinate: Excess deaths per million person years. Abscissa: T65 mean skin dose in rad. Correction of the dose for attenuation and for a neutron RBE of 5 gives rem values close to the doses shown. These data differ from those given in Table f-1 because they cover a different time period, and are given only as an approximate indication of the dose-response experience up to the present time. 153

RISK OF LUNG CANCER U.S. URANIUM MINERS 1951-1971 20,000 - -3,000 -2,000 HI 15,000- s § X UJ 2,000 4,000 DOSE REM 10,000 J 5,000 - WHITE MINERS 5,000 10,000 15,000 20,000 25,000 CUMULATIVE BRONCHIAL DOSE (REM) 30,000 Fig. f-2: Dose-response data for lung cancer in U. S. uranium miners (Ref. 15, with added cases from Dr. Victor Archer). Ordinate: Excess cases per million person years. Abscissa: Rem dose to bronchial epithelium, calculated on basis that 1 \\ I.M =.'. rent. Insert: Lowest dose range for white miners. Error bars for white miners include 90% range for Poisson statistics (Appendix IV). 154

15,000 10,000 5.000 RISK OF LUNG CANCER NEWFOUNDLAND FLUORSPAR MINERS 1952-1968 2,000 4,000 MEAN BRONCHIAL DOSE (REM) 6,000 Fig. f-3: Dose-response data for lung cancer in Newfoundland fluorspar miners (36, Fig. 2). Ordinate: Excess cases per million person years. Abscissa: Mean bronchial dose, calculated on basis that miners were exposed at 5 \VI., and 1 WLM = 5 rem. Error bars include 90% range based on Poisson statistics (Appendix IV). 155

There is fairly good agreement among these four studies in the absolute risk calculated, when the RBE corrections are applied. On the other hand, the relative risk estimates for the fluorspar miners are somewhat divergent. The fluorspar miners give higher risk values than the U.S. uranium miners but a number of fac- tors may account for this difference. These in- clude: a) the fluorspar miners have been fol- lowed for a longer period, b) they are probably heavier smokers than the U.S. miners, and c) the U.S. data includes the period 5-10 years after beginning uranium mining, at a time when the risk is lower, while the risk estimates for the fluorspar miners are obtained over a peri- od beginning, on the average, ten years after beginning underground mining. The possibility also exists that fluorspar acts as a cocarcino- gen to increase the apparent risk in the fluor- spar miners. All four of these groups are still under inves- tigation, and it is probable that because of the relatively long latent period for lung cancer, the rates calculated will rise as further cases develop. This is particularly true for the spondy- litis patients. It is possible, therefore, that in the final analysis the absolute risk in these groups will approach 2/106/year/rem and the relative risk reach 0.5% or higher. For the three groups (miners and Japanese survivors) in which up-to-date information is available, it is significant that many new cases have been added during the past few years. REFERENCES (1) Smoking and Health: Report of the Advisory Commit- tee to the Surgeon General of the Public Health Service. Publication No. 1103, U.S. Public Health Service, Wash- ington, D.C., 1964. (2) Selikoff, I.J., E. C. Hammond, and J. Churg. Asbestos exposure, smoking and neoplasia. J. A. M. A. 204: 106, 1968. (3) Baetjer, A. M. Pulmonary carcinoma in chromate workers. I. A review of the literature and report of cases. II. Incidence on the basis of hospital records. Arch. Ind. Hyg. and Occ. Med. 2:487,1950. (4) Wada, S., M. Miyanishi, K. Nichimoto, S. Kambe, and R. W. Miller, Mustard gas as a cause of respiratory neo- plasia in man. Lancet 1:1161,1968. (5) Boyd, J. T., R. Doll, J. S. Faulds, and J. Leiper. Cancer of the lung in iron ore (haematite) miners. Brit. J. Indust. Med. 27:97,1970. (6) Doll, R. Cancer of the lung and nose in nickel workers. Brit. J. Indust. Med. 15:217,1958. (7) Buchanan, W. D. Toxicity of Arsenic Compounds. EI- sevier Publishing Co., Amsterdam, 1962. (8) Doll, R., R. E. W. Fisher, E. J. Gammon, W. Gunn, G. O. Hughes, F. H. Tyrer and W. Wilson. Mortality of gas workers with special reference to cancers of the lung and bladder, chronic bronchitis and pneumocom-osis. Brit. J. Indust. Med. 22:1,1965. (0) Berg, J. W. Epidemiology of the different histologic types of lung cancer. In: Morphology of Experimental Respiratory Carcinogenesis. U. S. Atomic Energy Com- mission Symposium Series No. 21, 1970, pp. 93-101. (10) Kreyberg, L. Histological Lung Cancer Types. A Morphological and Biological Correlation. Norwegian Universities Press, Oslo, 1962. (11) Churg, J. and M. Kannerstein. Occupational exposure and its relation to type of lung cancer. In: Morphology of Experimental Respiratory Carcinogenesis. U.S. Atomic Energy Commission Symposium Series No. 21, 1970, pp. 105-113. (12) Saccomanno, G., V. E. Archer, O. Auerbach, M. Kus- chner, R. P. Saunders, and M. G. Klein. Histologic types of lung cancer among uranium miners. Cancer 27:515, 1971. (13) Tamplin, A. and J. Gofman. The Colorado Plateau: Joachimsthal Revisited? An analysis of the lung cancer problem in uranium miners. Hearings of the Joint Com- mittee on Atomic Energy on Environmental Effects of Producing Electric Power. Part 2, Vol. 11, 1970, pp. 1994- 2033. (14) Goldman, K. P. Mortality of coal-miners from carci- noma of the lung. Brit. J. Indust. Med. 22:72,1965. (15) Lundin, F. E., Jr., J. K. Wagoner, and V. E. Archer. Radon Daughter Exposure and Respiratory Cancer: Quantitative and Temporal Aspects. NIOSH-NIEHS Joint Monograph No. 1, U.S. Public Health Service, 1971. (1ff) Pirchan, A. and H. Sikl. Cancer of the lung in miners of Jachymov. Am. J. Cancer 16:681-722,1932. (17) Wagoner, J. K., R. W. Miller, F. E. Lundin, Jr., J. F. Fraumeni, Jr. and M. E. Haij. Unusual cancer mortality among a group of underground metal miners. New Eng. J. Med. 269:284,1963. (18) deVilliers, A. J. and J. P. Windish. Lung Cancer in a fluorspar mining community. Brit. J. Indust. Med. 21:94, 1964. (19) Peller, S. Lung cancer among mine works in Joachim- sthal. Human Biol. 11:130,1939. (20) Lundin, F. E., Jr., J. K. Wagoner, and V. E. Archer. Draft report of the Interagency Uranium Mining Review Group: Summary prepared for Advisory Committee to Federal Radiation Council, March 1971. (21) Jablon, S. and H. Kato. Radiation Dose and Mortali- ty of A-bomb Survivors, 1950-1970. ABCC TR 10-71. JINH-ABCC Life Span Study. Report No. 6, 1972. (22) Court Brown, W. M. and R. Doll. Mortality from Can- cer and other causes after radiotherapy for ankylosing spondylitis. Brit. Med. J 2:1327,1965. i,23) See Appendix V. (24) Steinitz, R. Pulmonary tuberculosis and carcinoma of the lung. A survey from two population-based disease registers. Am. Rev. Resp. Dis. 92:758, 1965. (25) Campbell, R. E. and F. A. Hughes, Jr. Development of bronchogenic carcinoma in patients with pulmonary tuberculosis. J. Thor. Cardiovasc. Surg. 40:98,1960. 156

(26) Miller, R. W. and S. Jablon. A search for late radia- tion effects among men who served as x-ray technolo- gists in the U.S. Army during World War II. Radiology 96:269,1970. (27) Archer, V. E. and F. E. Lundin, Jr., Radiogenic lung cancer in man: exposure-effect relationship. Env. Res. 1:370,1967. (28) Walsh, P. J. Radiation dose to the respiratory tract of uranium miners- a review of the literature. Env. Res. 3:14,1970. (29) Sanders, C. L., Jr., R. C. Thompson and W. J. Bair. Lung cancer: dose response studies with radionuclides. In: Inhalation Carcinogenesis. U.S. Atomic Energy Commission Symposium Series No. 18. Oak Ridge, 1970, pp. 285-302. (30) Yuile, C. L., H. L. Berke and T. Hull. Lung cancer fol- lowing polonium-210 inhalation in rats. Radiation Res. 31:760,1967. (31) Cember, H. Empirical establishment of cancer-asso- ciated dose to the lung from i44Ce. Health Physics 10:1177,1964. (32) Grossman, B. N., J. B. Little and W. F. OToole. Role of carrier particles in the induction of bronchial cancers in hamsters by polonium-210 alpha particles. Radiation Res. 47:253,1971. (33) Little, J. B., B. N. Grossman, and W. F. OToole. Respiratory carcinogenesis in hamsters induced by po- lonium-210 alpha radiation and benzo(a)pyrene. In: Morphology of Experimental Respiratory Carcinogene- sis. U. S. Atomic Energy Commission Symposium Series No. 21, Oak Ridge, 1970, p. 383. (34) Mohr, U. Effects of diethylnitrosamine in the respira- tory system of Syrian golden hamsters. In: Morphology of Experimental Respiratory Carcinogenesis. U.S. Atom- ic Energy Symposium Series No. 21, Oak Ridge, 1970, pp. 255-264. (35) Bevan, J.S. and A.K. M. M. Haque. Some speculations on the carcinogenic effect of inhaled alpha-active materi- al. Phys. Med. Biol. 13:105,1968. (36) deVilliers, A. J., J. P. Windish, F. de N. Brent, B. Hol- lywood, C. Walsh, J. W. Fisher, and W. D. Parsons. Mor- tality Experience of the Community and of the Fluorspar Mining Employees at St. Lawrence, Newfoundland. Occ. Health Rev., Ottowa, 1969, pp. 1-15. (37) Haque, A.K.M.M. and A.J.L. Collinson. Radiation dose to the respiratory system due to radon and its daughters products. Health Physics 13:431-443,1967. Data on stomach cancer (Table g-1) may be drawn from atomic bomb survivors and pa- tients treated with x rays for ankylosing spon- dylitis. Analysis of the A-bomb data shows that there was no evidence of any radiation induced cases of stomach cancer in those survivors exposed to 10 or more rads in the period from the 16th to the 25th year after irradiation. Restriction of the analysis to the latter half of this period (i.e., 20-25 years after the bomb) still fails to indicate any excess cases (the rela- tive risk for the 10+ rad group being 0.98). Anal- ysis of the data from patients treated with x rays for ankylosing spondylitis shows a signif- icant excess of stomach cancers occurring 6-27 years after irradiation. The best estimate of the absolute risk from these data is 0.32 to 0.64 deaths/106/year/rem depending on whether a value of 500 rads or 250 rads is used for the mean dose to the stomach. However, the possi- bility remains that the excess number of cases was not due to radiation but arose from selec- tive factors associated with the disease process or its treatment. An analysis of all G.I. cancers excluding those of the stomach is shown in Table g-2. The data again are taken from the atomic bomb survivors and the patients treated for ankylos- ing spondylitis. The mean dose to the relevant organs in the spondylitics patients is assumed to be the same as that for the stomach, i.e., lying between 250 and 500 rads. This dose range gives rise to a best estimate of the abso- lute risk varying from 0.22 to 0.44 deaths/106/ year/rem. Again, the same limitation as dis- cussed above applies to the data from the spon- dylitics patients. g. Other Neoplasms of Specific Types A variety of neoplasms other than those mentioned above have been reported to occur in excess following irradiation. The neoplasms include lymphomas, carcinomas of the phar- ynx, carcinomas of the stomach, carcinomas of the pancreas, carcinomas of the paranasal and mastoid sinuses, cholangiomas and heman- gioendotheliomas of the liver, tumors of sali- vary glands, and miscellaneous neoplasms of other types and sites (1,2). REFERENCES (T) Upton A.C. Effects of Radiation on Man. Ann. Rev. Nucl. Sci. 1S: 495-528,1968. (2) International Commission of Radiological Protection Radiosensitivity and Spatial Distribution of Dose. ICRP Publication 14, Oxford, Pergamon Press, 1969. (3) Jablon, S. and H. Kato. JNIH-ABCC Life Span Study Report 6, Radiation Dose and Mortality of A-bomb Sur- vivors - 1950-1971, Atomic Bomb Casualty Commission Technical Report #10-71. (4) Court-Brown, W.M. and R. Doll. Mortality from Can- cer and Other Causes After Radiotherapy for Ankylos- ing Spondylitis. Brit. Med. J. 2:1327-1332,1965. 157

s es O x 09 I s JC e go *4 9 n BiujumoD jaqjo 10 •a^otUOOj 4) n o M 03 • a) c UJ ^ rt Sid 4V > O M C S2 00 C/l •— - ttf 10 4 maJ/JBaX/ QI « S ° 9 5 ™ "" "u- oo ^H l/l j /aasBD JO sqiBdp H s II •-i o O i/l ° o o o o tO ^H i 8 S Sr-i n -c d o o o • V "V.wL.K.T »„.«.* o o 0 O o o e m rH in n X n c 0 (3/0) W SAT4B73H r "- r - 00»0-^ | 1 i K) CN eleytion. . •[CU3U03 V) CTi O §m •O 4> »i « S * x»S Z UO U. X «r u. o (8JBaA) ""H a S £ '• »*V aSuEH o 00 in in S o (N o i o in i o .2 05 -UB3W 00 rN i m § -S m O l"- 0 m CN 4 "T^ O 00 9 9 •^3afqns !* 1 | JO 3 O. g paseq am S9}BmT]ta W ipPl" CN Ct 1 UO *pBJJT J93JW P°TJad 3 (.«,*) U"H a Ll 30 uo^3oaixi j3noy N LL i aanscxlxa in 4 •o V) V) Ll o X o rt •o x V . 2 uoi3*7pB:i jo atl/i >- + c X • »ou«a^a A n - H m 4 •^sr •i x -H 1.3 IS O 4J 4J i*) O.-H CC CT* "-')•' I^H -- 5-S t. E U -< U iJS •O 41 > M C O CC «• f -w 9 L. 4) O -O o! 3 Ml o « » g. H --i m t- "7 -Q ;"; -r 158

N 41 I b U 3 * •9 •p« * S CO 1C t3oaoBK» Jaipo Jo taiou3ooi 14) Le . l« r-i Is r , 1C 4 10 ^ lu - i m •-i 0*0 K O C B iu cd S SS m 00 CO *-* 6C Ig « CO "" 00 i CN •3 9 5 p' m d o 5 iM /83BB3 JO SVpeap .B Jj ^ O 1 O • 1 • 3 O O 0 1 0 1 0 CD • c o « 1 00 I s i 3«ag i-. i 3 o* o O 1 O •(4 raJ Jad i(sT-> aATJBjaJ ui asBaJ3u? qua3Jaj CN 1 00 0 0 0 O 01 r-1 1 O • 1 • o O 1 O Id l — M X o 1i O M (3/0) 1»fJ OAJ3BIaH ii • in ^ i— i CN & o o CN i-H • •o o 1 B O JO co m i-H Q) 4-1 MBMg UJ Tl - - C C OS U u « s , CJ xag *-* S^D. b (BJBdA) "an (N •0 t >• -» s 4.1 O 00 n m 01 3I1USI3 O 1 O m i o 00 C--1 1 u~i ' " "" "TS o o m O a i m) O P^ o in - •4 -~":™ m CO ° B to 3 CO •ri tJ s3oafqns JO it in in C R o o (sJBaX) pasaq aa. m 0, CM 8 1 m u ig 4J no -pBJJT J»3JB poTJaj kfi I u«an (•MMQ S 3 | JO UOT3BJT1G B ajnsodxa m I b io o a) •a 4J >i •O jo uoi3tanf! a UOT3B1PBJ JO 3tUl * + * X «ua«jaa & * * ** -«r Hi i *j .3- :-. cm •o « o i c -^ ••-* m a.-*-) « ov •n - ft. — SiB ' f •-• > „ «, I A) O a) O ju in B iO q c -r: Cd « fi !3 l^ •8 H S 159

h. All Cancers Other Than Leukemia Tables h-1 and h-2 show the data for all cancer excluding leukemia, the sources again being the A-bomb survivors and the patients treated with x rays for ankylosing spondyli- tis. The data from the A-bomb survivors are derived from deaths occurring from 1960 through 1970 and are grouped by age at the time of the bomb (ATB). These data are also shown in graphical form in Figures III-1 and III-2, and demonstrate the apparent higher relative sensitivity to cancer induction of persons irradiated when very young (0-9 year old). The risk estimates obtained from the patients treated for ankylosing spondylitis can be seen to be compatible with the pooled data from all survivors who were aged 10 or more ATB. These observations indicate that the eventual total number of deaths from sol- id tumors induced by a given dose may well exceed those from leukemia by a factor of 5 or more. 2. Cancer Following Irradiation Before Conception or During Intrauterine Life First Reports In 1956 Stewart and her associates (1) pub- lished a preliminary report describing a two- fold excess of leukemia and other cancer among children whose mothers received diag- nostic x radiation during the relevant preg- nancy. Two years later a definitive report was published (2). It showed that the risk of cancer among the irradiated group was about doubled for six of eight categories of child- hood neoplasia, the exceptions being myelob- lastic leukemia and lymphoma. The history of irradiation was obtained by interview from mothers of a) 619 children who died of leukemia, b) 680 who died of other cancer, and c) an equal number of controls (children without cancer) matched by age, sex, and locality. All deaths in the case-group occurred before 10 years of age, 1953-1955. Confirmation In 1962 MacMahon (3) reported similar results from a study with objective evidence of maternal radiation exposure rather than reliance on perhaps unavoidably biased inter- views. Through the use of obstetric records in 37 large maternity hospitals in New England, for the period 1947-54, he determined the fre- quency of diagnostic x-ray study of the moth- er,s abdomen during pregnancy for 569 child- ren who subsequently died of cancer, as com- pared with a 1% systematic sample of all other births in the same hospitals. (Table 2- 1). The results indicated that following such exposure the relative risk of childhood leuke- mia was increased by 40%, cancer of the cen- tral nervous system by 60% and all other cancer by 50%. No relation was found be- tween neoplasia and recorded complications of pregnancy. The oncogenic effect of x ray seemed to be exhausted by eight years of age in contrast to its persistence through the entire span (up to 10 years of age) in the study of Stewart et al. MacMahon and Hutchison (4) later noted, in comparison these results with all others available to that date, that those showing no relation to x-ray exposure, including a prospective study by Court-Brown, Doll, and Hill (5), lacked power to reveal an increase in rela- tive risk of only 40-60% because of their re- latively small sample sizes. Subsequently, a prospective pilot study by Diamond and Lil- ienfeld (6), involving follow-up of about 20,- 000 children exposed to diagnostic radiation in utero and 40,000 controls, revealed among whites "a nearly two-fold increased risk of dying (from all causes) during their first 10 years of life." No excess occurred among the black children, who comprised about half the sample studied. Leukemia caused the deaths of six white children as compared with two expected; no such deaths occurred among in utero exposed black children. Neither ethnic group experienced an excess of other can- cers. Both Stewart and MacMahon showed that the relative risk of developing cancer follow- ing fetal diagnostic irradiation is elevated even in the first two years of life. The relative risk in both studies rises after this time, to reach a maximum for children dying at ages 6-7. In the MacMahon study (3), the oncogenic effect of x rays seemed to be exhausted by 8 years of age (analysis of the data indicates that one can state with 95% confidence that the relative risk for those dying at 8 years old or later does not exceed 1.05), in contrast to its persistence through the entire span (up 160

« -i X U J e ST. BilUJUlllOD JdqjO JO saiou^oojj c M b rt - in C M M m iO 6 m B in g 8^ f* 0l M 0) • M 4) i/i UJ J4 Cd CO CO OS i-• >- O M UJ .u cd V m (U jtf i ill £2 ^ 03 « • §g CN CN : 2 •o r- 1 maJ/JBaX/ 9 CN i-H CN -H oo in CN ~4 9D £ ~* o o d O O vo o CN ^t CN c-4 0 O CN ^ c 'i\ail i>] m os 3 l§ d d d d CM ^H c o 00 ^ 10 io r- o i-H d -H O •* io oo in ao in W io c iO iO 00 - »0 CN 00 CM ao m o o maa aad H^l- l SA;4E?-[dJ CN pH o o d d o d o o m o UJ OKE.UDl d d « a an - - i-• in - m ^ in -H in rf « (a/o) I«TJ a,-.i am da Ol cn io CM CN in 1 CN O vO r-4 00 CM — . "**. in in r-i io m ^4 O s vO CN (N ii 00 •-• CN CN ^ CO iO H iO CA {ono°g 1 M 1 cd 1H •3 eleytion. C7. Ol Ol H Ol 0 0 1 cn i o XaS Z M» tt. I »s U. «.«, • «« Z «ff U. Z •0U. fe -Sir UU.JH in m 00 CN CM s M t; — Oi »o CM in . 1 1 o 0 o o CN o qj CO m • 05 -UEdW CM 00 00 9 (speJ) asoa O 00 § nH IF"a|u^ o o i o o o 3 . iO 1 i «B 0 0 o o o % sJBaA-uosJad jo aaqmnii o in 00 CO o Ok 01 *o in CN CN m 8 1 10 0O CO 00 m o" m n r*. •a c V t3oafqns m " This table does not provide a yomplete com[ J" Jaqmnji (sacdX) pasBq 9JB in m m m in CN 1 CN CN CN CN i i uo "pBJJT ao: JB poTJaa m in in LO m U8dH m in in <HBaX) CM CN CN CN CN in a3llBH in m m in CN CN CN CN CN CN m CN aJnsodxa O b O o b b JO HOT ii-MTHI V V V V V T IP aouajiajau >- + C >- + C >- + C >• + S >- + c >- + C CM CN CN CN CN CN dO j-g f* jO E Z uoT38indod Apn^s Jz m o m o in ^3 i -• A + •+ JS + ^ 1 O) < x 2 < X -H < X - 9-797 O - 72 - 12 161

b o a t. e es M "s (B 1 gquaraooa Jsqjo V 4* O JO ••3OU300J M c r- O s & p- C b A O O 4 iH 5 SS Is ao m w « -O B •H «0 4) 0 -O •aJ/J«a*/ oi ^ IN 3 "o "S •i /ta»3 JO sqtlBJp 3 lg o Oi 4 £ ib > b .^•04 U b "-Ha-ii aaniosqy J 9> in N ,.»« I tv flO 4-1 3 4) C >i UBJ Jad us JJ a A[3^TOJ CO 5 a - P. U O -H A iX UI SSBJiDUJ juaojaj d c) _ u-j a) O 4-1 1 iO 4J am 3 - I iN 00 « CN 4) CB 4 6 5. 3 — (a/o) w* aAT3»ian io o -t •? -H *S "O JOJ3UO3 CO 00 *O CN 4) 4. 0. U) 3° -^ ^ £,,3 C 4J b O O4) H-i -O iu pH £ Si 4J 4) O. U *i • O 4) U b 4) i-" 4) 4) O. x,S r4 Z d b b O. C O (BJeaA) u.aH n X - Ci-i 4) V O u X -- O * -" c P x 3 iu — d 4J b • if • «i in in ai CX b b 0} X Q) O » 4) X X rH in o a ' usax 5 o i o m i O •- 0) *U O CN i in a) 4-1 m i0 H-i g BUJa3xg ,8 3 0 0) *0 4J X b n m 4 a) o O r» • -H O 4J 4 sJBaA— no9 Jad i O O CO U •o 2 ^ § M •2 5*5 m ._ ¥ so- C .0 r-i O ! a 4 P-i x 3° JaqmnN 4O 0 X 4 U U 4) C 4) X 4) O b 4J 4J *r4 4) 4J 4) O O. — u) g 3 f 2 paSeq aJ» Is rs m U i-H 4 4 n -o » 4 U H-l m 4) C 2 4) M 4J 3 0 •H 4 -4 ^ u.aH S a > -a M o O - C CM (M.aX) Li cd 4 4) O. 4J X n 4J ^ an *i * b t« « ,^ CN c t) 4 c ^ un 1 41 4) -H in CO 4J X 4J C ^n 8b 4 O TqB^pBJ § >. 4 -O O, CN O 4J 4) 0. 4 JO V OTlEJna 4 O 4) T3 4J >, 4) Li - 4 « ^ •-• b 3 •O 41 4 4J U uoT«TP« jo adXj. 3 X 4J 4-i b Q) 4 Q) U B) g i-i >, a3uaaajaH S CO X b -H jg H b '^ 4J uoJ3Eindod z Spondy- litia Patients 1935-54 4 £* « "* ~ QO iu a J 4) nj bo T3 a c I £ it s - J o y ^ " 1 I Ib 1 "^ i 'i. ,- b u -5 £ O S*1^ b **2i v o X "£ -H rf O •-C js o o H « H 9* m v 3 0.0 S • U . g t B 4 b u 0 • 0 « § M m cu 4 B 4 CN CL. 4) B) >s-^ *' ^ ~ = Ml!- £ b b b> O O U) 4J i; iw C ca 3 •-• C 1 *' — " *-. * -g -o ^ "o w 91 * « 4 ^ -9 CN U b " gS| ii « c 1 o a c b 01 • ** • 00 n 8.H m a ) a) a) a) A 0 CO O 4J iJ o w • •c X b 4 ii §1-0 o c ,21 *i ^ "^ o" « n 4 n ^ sK — 4 O £ b 4) CO O *> v So 1 X X 0 •r* g • t ^ SliO •r- 10 Ov X 'IsJ OJCT. X " a) -H * X « u in f) 1: • *i *j 3 -*4 0 M CO •j • m C i^ -^ •f X io C 4) a) -i 4 O. n ^ 4) B) ^ sis J4 QO a o 4 b < -a ij n o. O 4) n O CN •-i 3 X 1 " b 1 4) T3 -O a) « *j a) +i w • q X 3 -H 4 S -^ — O ^_| 05 .3 r-. . ojm O *J 4 4 « 4) ^ L. b 4) X a ) b H > CN J8 U X d 4 4 CN 4) -404) -H IX OX 4J w * V " > C C 3 m * ^ •S "a U U i usd to de le hypotheti m « « B» X 4) CN *J 4 -O CO 8|CN CO 5 B "S 2 2 H olo H • O 41 T3 CO O 813 °J° 4J 3 4) "i O ij •iJ m 4) 4) 4 • sis • A *H 4) *D f*|0* •8 » 4 ^ 4 X • C ^ U * 0 4 ) 00 4 •*i t3 in >. c SO a) CN c b *J X iJ - a) iu w w 0 3 58-S5 ° 11 2 •" 4J a) b 4J X 4J •s-g^ig 4J 0-1 4 B 01 -^ 4 0 4) « 0 x x t 4) •a 4J O 4J ^H « •O b 4 -H 4) •o •^ .2 S 4) C > •3 X H-C 4 a) > 3 m u X O (U f' c.s a) £ a) -o A 5 3 5 -a 4 5 * O f- J5 H « u c 4J O O • 4 O 4J Sii *o 4) « 0 si 8 3 162

e w u. b i—, s r I e I S S (a JO ••30113004 N se footnote (b) ralue of E alyulated 'rom 1950-52 rital atatia- iys for Japan. 3 Oi •aJ/J«aX/ OT „ S 2 9 « o" IN f) m /tas»3 JO oi() i, )[- ,< |g ' nTOBQV * CN 04 o 3 sag s ps. CN s o mai Jad >[(TJ aAT3V[aJ B S • o m M - - - 5 * • o (3/0) W aAT3«ian § in ci s •§ CN jo if (- li! in C 1 id o o. m *»S 5 s*,^ .*., ==*„i UB3H s O o O 3* a»v aSuEy c- O o O SflBBTT 03 • UBBH ; 00 O o (gp>J) asog _„-.—- — o o CN CN o c 1 o n i m o UI o 8 o JO Jaq1mijj o i O ln s3DOfqns JO 0- • P^ P^ in (BJeaX) paBBq 3J* p o o o o o UO -pOJJ-i J04JB poTJ.IJ o uwajj S , CO o ,0 U0,3.lna ^ j— 1 O o m aJnBodxa r-- >• c c b V 3 •rl B B uoijBtpeJ jo adXi S X X ,.. . a3uaaaja* p- 0 "« n n uoH^indcHl Apn^S a) m » 1 « -n a S z ^ •- I aS 4J •-•'?• 3 1+3 £ -iaS 4) * i7* Ui UJ " •" Cu 3 •-i 163

CN See footnote (b) Value of E estimated from 1950-52 vital statistiys for Japan maa/JBaX/ QT S S C a a o 9 3 ^ "s o /sasCo JO 8l43BaP E C _ J 00 ^3 CT* CN 00 o 3sag 2 O s ^ maJ Jad isii aAT3>-[aJ 00 CD in in 1i1 aSBaJl3UI iluaaJaj am 5 H - - (3/0) *,SP :'Aiqu[.iM § CN i£3 in n 00 CN O -[CU3UO3 JO CO *O CN iti 4) in ui CO >i C 1 1 3 4 4 O 4J 4J li p. in •atom 4) 0 Cb C X 4 a> xas 3 Z *» fc as ia EH »«» U8aW 2 O o o •M O o o 03 i ueaw 00 o o ^H o in o 0 4d3uFH e CN CN 1 8 m d in o in i sjeaA-uosJad O O o in jo aaqmnu S i o r- o o o o - m 3° S i CS- paseq ,uw o o O sa^BttiT3sa wiz ipj.il" r-- 3 o S <saBaX) USaH S • CO o TO 1i07 icaim SSJUB-H g i o o lO aJnsodxa JO uoyji-Miifi 5 c c •fi r o ••H B 8 V uo'puTpiu jo ad/x S X X ^+c a DU ai a j ay S «s o co CN uo^3e^ndod m *O in n •* CO m CQ i 1 .S s "is » 1 = s 3 *H cO 31+3 « C°^ B, 3 rt 4) 1 OV 9 i 1 — Q) X •2 S m 41 > -x ij • c CO c c I 1r O — 5 . 4-1 01 i " s £jg a -2 U 4-1 CO CJ >, ro C V C 4) 4) iu 44 O B •& ° ca 3 -H O" « O 4) C CM & U U i— CO £ O a) n s •o h O 4J c •^ U O "o 4J 4 B 4-1 —- R3 Cu C 3 a) ai a) JO «J ^ ^ J if, a •O O <J 4J 23 lu o ca CO C T O 0 iu 44) — . CU O U CO O > ^ W) s« SJ .a *- s si II •O CO 1 ll -e 5 J « Sca m u S "S c 4 4-1 iw w o *al u-i •^ >. r- t4 tf 0 "8 £ « 4) O) B < j=: u - M U C) 4J C B y iu C eo u 4 W to >, CO 4*4 JJJ •r4 ^ a 0 a) c — O §c Go <a U M 4-1 0) *4 4 i H 4J CO g-S m "•§ S " § ^H m 4 4) 5 ° C LJ E -u C iu 0l O 4) ij c <= 1 CD AJ CQ 40 n 4J 4 ^i 4J •" * •-" 4J a) m ^ « JH 8 •O U •O 9 4) cn 3 Ji Li *•> iO O* f 4J CO 4-1 r - "* I i i d" •* C* --i 4J 4-) •-• MO) « C * 3 0 m CN ^ 3 - ~ - - 3 -^ CN vO C7N BO 8 - 4 Ob 4J O iu 3 CO -4 C JO 7 DO iu m *-* 4J CU i-) O 01 O 4 a) 4 O §3 X 4) t3 4) •* 4 a a. m -^ » CN — £ 4) CO O O £ 1 •0 0) o c •O 4J —i 4) 4J "O C -44 a -H 4) 4J 4 « « « •= = ss D. •O 4) £ 4-1 B C iu m 4) m m O 4 -H Cw co m cn fc "•* y o -* i- d- o •-• B m ^ 0 B " S 1-3 4) C C - 4J &0 00 U ,-T 1-1 "^ 4) -H 4 •-i t^1 S c CO CJ U S T3 O ^ C 4J y cm £J T3 ^ ^* CO > •r4 -4 * C iu 3 « >S 4) Jrf ** C m m 0 y -o •a •* £ ca H-i a) y «.S •rl 4-13 y erf « > y n 0 4) U CL b ^ 44) CO m 4 4) 4) ,C -^ ^ U > w H jf! & m — < p 4) §o o -H y U H 4J C Ji 4J 4 -4 • fa 4J HA 3 •-< « O 4J 4-i 4) O 4) 4) 4J JO Of t 4 4) ^% - m C 4 B 4J in m 0 4 O 4J 4 Ov io •O 3 - •H iO m a) >* ^ jj 3 ^ fl~ *-! -ao A 3 t4 ft w CN i-i O ••-' 4) r-H *7 4J CO C O s.1- a) 4J in o n my ^ 3 !fi J3 •O •^ ON •• 4-i r-l O H H C 4 "o CO CN 4 A 01 0 . CO Z) 4-) 4J !33 164

to 10 years of age) in the study of Stewart et al. A possible reason for at least part of the discrepancy between the two studies might be in the methodology used: Stewart et al. (2) relied in the main on mothers, memories whereas MacMahon (3) used hospital records for the source of the data. It is conceivable that if the mothers whose children died of cancer were more efficient in remembering whether they had diagnostic irradiation than the control mothers, this discrepancy would increase, the longer the interval between the pregnancy and the death of the child. In other words, the relative risk estimate would in- crease due to this systematic error, the older was the child at death. Until evidence is con- clusive, it is prudent to assume that the la- tent period is effectively zero years and that the period of risk is at least 10 years in dura- tion. Equal Induction of Each Form of Childhood Cancer and Interpretation of the Data In 1968 Stewart and Kneale (7) published a further report on their data, which showed that each of six categories of childhood neo- plasia was equally induced by maternal ab- dominal exposure to x ray during pregnancy (Table II of their paper). The classes of neo- plasia involved (leukemia, lymphoma, neu- roblastoma, cerebral tumors-, Wilms, tumor and all other cancer) differ markedly from one another with respect to epidemiologic characteristics (8); hence the plausibility that low-dose intrauterine irradiation would increase the frequency of each by about 50% was questioned by Miller (9). Establishment of a causal relationship would have been aid- ed by showing a specificity of effect (for one rather than all forms of childhood cancer), and a consistency with data from animal or other laboratory experimentation. Interpre- tation of the results seemed further compli- cated by a previous report by a group of U.S. epidemiologists (10.) who found in a collabora- tive (Tri-state) study conducted in Baltimore, Minneapolis, and New York State excluding New York City, that a similar (60%) excess of leukemia occurred in children whose mothers reported diagnostic x-ray exposures up to 10 or more years before conception of the child- ren. Half as great anincrease (30%), of bor- derline significance, was also observed with respect to paternal diagnostic irradiation before conception of the child. One must con- clude that such an effect, if real, was herita- ble; however, it seems unlikely to be due to a genetic influence since a comprehensive study involving six indicators of genetic damage in the F! generation in Hiroshima and Nagasaki showed ,no detectable influence of radiation exposure (11, 12). Also, no excess of leukemia was observed in the F\ generation (13) al- though the radiation dose-range was much greater than in the diagnostic x-ray studies. In addition, cytogenetic studies of atomic- bomb survivors in Japan showed complex chromosomal abnormalities if exposure oc- curred in utero or in later life, but not in the F! generation (conceived after the explo- sions) (9, 14-17). Thus, the claim that a pre- conception radiation is leukemogenic cannot now be linked to a genetic mechanism or chro- mosomal abnormality such as that which characterizes persons known to be at high risk of leukemia (IS). The authors of the Tri-state study, upon further examination of their data on precon- ceptional exposures, concluded that other factors interacted with radiation to increase the risk of leukemia appearing at 1-4 years of age (19). Thus, for example, the relative risk rose about four-fold when such irradiation occurred in conjunction with a maternal his- tory of miscarriages or stillbirths when, and in addition, the child had at least one viral infection more than 12 months before the onset of leukemia. The numbers of cases in- volved in this particular estimate, however, were very small. Statistical significance at the 5% level was attained when two radiolog- ic factors (preconception and intrauterine exposures) were related to one or two path- ologic factors (childhood virus infection or maternal miscarriages and stillbirths), or when one radiologic factor was related to both pathologic factors. Surely, as the inves- tigators themselves said, these observations require confirmation (which will not come easily because of the massive effort required to collect such data). A further report from the Tri-state study by Bross and Natarajan 165

(in press) finds that the relative risk of leuke- mia induction from previous diagnostic x rays to the mother during pregnancy is high- er among children prone to allergy and some other diseases than among children whose immune and repair mechanisms, apparently, are in better order (22). As stated earlier, the finding that precon- ceptional irradiation is associated with an excess of childhood leukemia complicates in- terpretation of cancer occurrence following small intrauterine doses of x rays. It is diffi- cult to rationalize why such similar results should occur whether exposure was prior to conception, as found in one well-executed study, or intrauterine—and if the latter, whether the outcome observed was leukemia or any other childhood cancer. Linear Dose-Response Relationship Stewart and Kneale (20) subsequently de- scribed a linear relationship between radia- tiondose(0.5+ abdominal films) and the excess in cancer risk under 10 years of age. The au- thors estimated that "among one million children exposed shortly before birth to one rad of ionizing radiations there would be an extra 300-800 deaths before ten years of age due to radiation-induced cancer (mean 572 deaths, standard error 133)." This estimated number of extra cancers per million rad of intrauterine exposure could be tested in an- other situation: the survivors of Hiroshima and Nagasaki. A study, recently reported by Jablon and Kato (21), concerned 1250 children exposed in utero to less than 500 rad. The accumulated dose was 34,933 person-rad. Under the conservative assumption that half of the dose was attenuated by the mothers, bodies, 18.4 extra cancer deaths under 10 years of age would have been expected ac- cording to the Stewart and Kneale estimate (lower limit= 5.2), whereas essentially no ex- tra cancer deaths were observed among the children exposed in utero to the atomic bombs. To explain the lack of agreement be- tween the two studies, Jablon and Kato sug- gested that Stewart and Kneale may have overestimated the cancer induction rate; that the dose-response curve may be linear at low doses and concave downward at higher ones, as might occur if abortions were induced by radiation, i.e., a competing risk; that high energy atomic radiation may be a less effec- tive carcinogen than is low energy x ray as used for diagnosis; or that factors other than x ray distinguish the irradiated from the non- irradiated fetus; i.e., low-dose x-ray exposure is not the cause of the childhood cancer, and the diagnostic procedure merely indicates that the pregnancy differed from normal. Another possibility is that the Japanese have a lower sensitivity to fetal irradiation than do Caucasians. Conclusion The studies reported to date indicate that diagnostic exposures during fetal life are associated with an increase in cancer deaths under 10 years of age. Whether or not radia- tion is causally related to the increase in can- cer is open to question, since neither labora- tory research nor clinical observations as yet support the concept that very low doses of irradiation might increase the relative fre- quencies of all categories of childhood cancer by about 50%. In any event, it is difficult to extrapolate from childhood cancers to adult cancers because of differences in type and epidemiology—and hence possibly in etiolo- gy- The risk of childhood leukemia has been reported to be similar whether diagnostic x irradiation occurred during pregnancy or, in the only such study reported to date, as long as 10 years before conception. Study of the Fj generation of Japanese survivors, howev- er, has failed thus far to show an excess of leukemia following preconceptional irradia- tion. Also, comprehensive studies of the Fj generation, using six indicators of genetic damage, have failed to reveal an effect. Hence, the interpretation of those studies which have reported an association between leukemia and preconceptional irradiation re- mains uncertain. Despite uncertainty about the oncogenic effects of intrauterine exposures, we presume for purposes of conservative overall risk evaluation that such exposures do increase the risk of cancer in the child until 10 years of age but not thereafter. 166

3. Total Cancer Risk In view of variations among the different types of cancer in their relative rate of induc- tion by irradiation (see Table 2., Appendix III), which are apparently unrelated to variations in the respective natural incidence levels, there is no basis for assuming that the incidence of alj types of cancer will be increased by the same magnitude, in either absolute or relative terms, in response to a given dose. Hence esti- mates of overall cancer risk must be based ei- ther on direct observation of overall radiation- induced cancer excess, which as yet are incom- plete due to the limited duration of follow-up of exposed populations, or on the total of the ex- cess rates of different types of cancer, data for which are also incomplete at present. In the atomic bomb survivors, the cumulative excess of all forms of cancer, including leuke- mia, corresponds to 50 to 781 deaths per 106 exposed persons per rem during the 20 year period from 1950-1970; i.e., from 5 to 25 years after exposure (1). Stewart (2) has raised a question concerning the general applicability of the ABCC results to radiation effects in man, arguing that ob- servations of mortality were not made during the first five years after the bombings (prior to October 1950) and cites Bennet,s report on the Bristol floods of 1968 (3) as evidence that a dis- aster may increase the number of cancer deaths in the year immediately following, pre- sumably with a consequent lowering of cancer mortality in subsequent years. Jablon (4) has argued in rebuttal that even if such a disaster effect be conceded, the number of deaths in- volved would be too small to be apparent in a follow-up extending over a twenty year period from the fifth to twenty-five years after the bombings. This Subcommittee concludes that although some effect of the kind suggested by Stewart may be present in the ABCC data it would, at worst, be quantitatively very small and would have no practical effect on the risk estimates derived from the ABCC data. REFERENCES (1) Stewart, A., J. Webb, D. Giles, and D. Hewitt: Ma/te- nant Disease in Childhood and Diagnostic Irradiation In Utero. Lancet 2:447,1956. (2) Stewart, A., J. Webb, and D. Hewitt: A Survey of Child- hood Malignancies. Brit. Med. J.I: 1495,1958. (3) Mac-Million, B.: Prenatal X-ray Exposure and Child- hood Cancer. J. Nat. Cancer Inst.28:1173,1962. (4) MacMahon, B., and G. B. Hutchison: Prenatal X-ray and Childhood Cancer: A .Review. Acta Unio Contra Can- crum 20:1172-74,1964. (5) Court Brown, W.M., R. Doll, and A. B. Hill: The Inci- dence of Leukaemia Following Exposure to Diagnostic Radiation In Utero. Brit. Med. J. 2:1539,1960. (6) Diamond, E. L., and A. M. Lilienfeld: Pilot Study Con- cerning the Relationship of X-ray Pelvimetry to the Subsequent Development of Leukemia in Children. (To be published) (7) Stewart, A., and G. W. Kneale: Changes in the Cancer Risk Associated With Obstetric Radiography. Lancet 1:104-107, 1968. (8) Miller, R. W.: Relation Between Cancer and Congenital Defects: An Epidemiologic Evaluation. 3. Nat. Cancer Inst. 40:1079, 1968. (9) Miller, R. W.: Delayed Radiation Effects in Atomic- bomb Survivors. Science 166:569,1969. (10) Graham, S., M. L. Levin, A. M. Lilienfeld, L. M. Schu- man, R. Gibson, J. E. Dowd, and L. Hempelmann: Precon- ception, Intrauterine, and Postnatal Irradiation as Re- lated to Leukemia. Nat. Cancer Inst. Mongr. 19:347, 1966. (11) Neel, J. V., and W. J. Schull: The Effect of Exposure to the Atomic Bombs on Pregnancy Termination in Hiroshi- ma and Nagasaki. Nat. Acad. Sci.-Nat. Res. Counc. Publ. No. 461,1956. (12) Kato, H., W. J. Schull, and J. V. Neel: A Cohort-type Study of Survival in Children of Parents Exposed to Atomic Bombings. Amer. J. Hum. Genet. 18:339,1966. (If) Hoshino, T., H. Kato, S. C. Finch, and Z. Hrubec. Leu- kemia in Offspring of Atomic Bomb Survivors. Blood 30:719,1967. (14) Bloom, A. D., S. Neriishi, N. Kamada, T. Iseki, and R. J. Keehn. Cytogenetic Investigation of Survivors of the Atomic Bombings of Hiroshima and Nagasaki. Lancet 2:672,1966. (15) Bloom, A.D., S. Neriishi, A. A. Awa, T. Honda, and P. G. Archer: Chromosome Aberrations in Leucocytes of Older Survivors of the Atomic Bombings of Hiroshima and Nagasaki. Lancet 2:802,1967. (16) Bloom, A. D., S. Neriishi, and P. G. Archer: Cyto#e- netics of the In Utero Exposed of Hiroshima and Naga- saki. Lancet 2:10,1968. (17) Awa, A. A., A. D. Bloom, M. D. Yoshida, S. Neriishi, and P. G. Archer: Cytogenetic Study of the Offspring of Atom-Bomb Survivors. Nature 218:367, 1968. (18) Miller, R. W.: Persons With Exceptionally High Risk of Leukemia.Cancer Res. 27:2420,1967. (19) Gibson, R. W., I. D. J. Bross, S. Graham, A. M. Lilien- feld, L. M. Schuman, M. L. Levin, and J. E. Dowd: Leuke- mia in Children Exposed to Multiple Risk Factors. New Eng. J. Med. 279:906,1968. (20) Stewart, A., and G. W. Kneale: Radiation Dose Effects in Relation to Obstetric X-rays and Childhood Cancers. Lancet 1:1185-1188,1970. (21) Jablon, S., and H. Kato: Childhood Cancer in Relation to Prenatal Exposure to Atomic-Bomb Radiation. Lancet 2:1000,1970. (22) Bross, Irwin D. J., and N. Natarajan: Leukemia from Low-level Radiation: Identification of Susceptibles. Mms. 1972. iThis range is the result of assuming an RBE for the neu- tron component at Hiroshima of 1 to 5. 167

In patients treated with fractionated x-ray exposures for ankylosing spondylitis (5), the excess mortality corresponds to a cumulative total of 92 to 1652 deaths from cancer per 106 persons per rem during the 27 years immediate- ly following irradiation. If rates of radiation-induced cancer mortali- ty similar to those above are assumed to apply generally and at the low dose levels approach- ing natural background radiation, then contin- ual exposure of the U.S. population to a dose of 0.1 rem per year (approximately equivalent to natural background radiation) would be ex- pected ultimately to cause approximately 1,350 - 3,300 cancer deaths per year, provided that the risk attributable to radiation did not per- sist for more than 27 years after any given in- crement of exposure. However, since the radia- tion-induced excess for many types of cancer is likely to persist longer than this period, as well as the fact that this simple calculation ignores other cogent variables, a somewhat more de- tailed approach has also been used to estimate the exccess cancer deaths in the U. S. popula- tion from continuous exposure to 0.1 rem/year. This approach, together with the various as- sumptions used, are outlined in full on the fol- lowing pages. Such calculations must remain highly tentative in the absence of more com- plete data, but represent the best estimates that can be made at present. The numbers yielded by these calculations overlap those presented above, i.e., 2,000 to 9,000 annual can- cer deaths in the U.S. population from 0.1 rem per year (Table 3-1). The wide spread in these estimates arises from our lack of knowledge of the long term consequences of irradiation of young children. According to the A-bomb data, these individuals appear to have a very high relative sensitivity (but a low absolute sensi- tivity) to cancer induction. Thus, if relative rather than absolute sensitivity is the appro- priate way of determining risk and the risk is maintained throughout life, the upper figure of 2This range is the result of the lack of certainty in the dose to the heavily irradiated sites of these patients. The dose estimate used in this report is 250 and 500 rads (see Appendix V) and the dose to the spinal marrow has been taken at 880 rads. Part of the difference between this range of values and that given above for the ABCC experience might be due to the fact that a skin dose was used in the ABCC analysis and would be subject to appreciable atten- uation, as mentioned earlier. 9,000 annual extra cancer deaths in the U.S. population exposed to 0.1 rem/year is predicted from the relative risk model. With this limitation in mind, the Committee considers the most like- ly value to be approximately 3,000-4,000 cancer deaths (or a 1% increase in the spontaneous rate). These figures must not be taken to represent more than crude estimates of risk, based on the incomplete nature of the data at present avail- able. Several factors, not taken into account in the calculation of these estimates, exist which compound the uncertainty of these numbers. First, no allowance has been made for the likli- hood that the carcinogenic effectiveness of low- LET radiation is reduced at low dose rates through the action of biological repair process- es. Second, the individual cancer risks used in the derivation of the numbers may rise or fall as the follow-up of the study groups is extended to longer periods. Third, the risks have been derived for the most part at high total doses, which may have been sufficient to kill a large proportion of the normal or susceptible cells from which a cancer might result. Finally, the risk estimates themselves are crude and often have wide statistical confidence limits which are made even wider than is indicated in the tables by uncertainty about the dose-effect re- lation and the RBE values that must be used for neutron and alpha radiation for some of the exposed groups. One further consideration is that these num- bers reflect mortality data and do not, there- fore, represent the number of individuals af- fected. If expressed in terms of incidence, in- cluding nonfatal cancers, estimates of risk could be higher by a factor of roughly 2. In ad- dition to this, it can be calculated, using the relative risk model, that roughly 2,000-4,000 cases per year of thyroid cancer will be pro- duced in the U. S. population from continuous exposure to 0.1 rem/year, using a 5-year latent period and either a 30-year or a lifetime pla- teau region. Detailed examination of the table summariz- ing the calculations of the excess deaths in the U.S. population (Table 3-1) reveals the follow- ing: 1. When the absolute risk model is used, there is a small difference between the two as- sumptions of either a 30 year plateau (a) or 168

Table 3-1 Estimated numbers of deaths per year in the U. S. population attributable to continual exposure at a rate of 0.1 rem per year, based on mortality from leukemia and from all other malignancies combined. Irradiation During Period ABSOLUTE RISK MODEL3 RELATIVE RISK MODEL3 Excess Deaths Due to: Excess Deaths Due to: Leukemia All other Cancer Leukemia All other Cancer In Utero 75 75 56 56 0-9 years 164 (a) 73 (b) 122 93 (a) 715 (b) 5,869 / 10 + years 277 (a)l,062 (b) 1,288 589 (a) 1,665 (b) 2,415 Subtotal 516 (a) 1,210 (b)l,485 738 (a) 2,436 (b) 8,340 TOTAL (a) 1,726 = 0.6% incr. (b) 2,001 = 0.6% incr. (a) 3,174 - 1.0% incr. (b) 9,078 - 2.9% incr. a The figures shown are based on the following assumptions: (1) 1967 U.S. vital statistics can be used for age specific death rates from leukemia and all other cancer, and for total U.S. population (2) Values for the duration (a or b) of the latent period (the length of time after irradiation befor any excess of cancer deaths occur), duration of risk ("plateau region"), and magnitude of average increase in annual mortality for each group are as shown in Table 3-2. 169

a lifetime plateau (b) when applied to all cancers other than leukemia. Each assump- tion yields a calculated excess number of deaths of approximately 2,000, with leuke- mia deaths constituting about one fourth of the total. 2. Contrary to the absolute risk model, the relative risk model generates quite differ- ent numbers (by a factor of about 3) for the two assumptions (a) and (b) of the length of the plateau region for all cancers excluding leukemia. Examination of Table 3-1 reveals that almost the entire difference is due to the deaths generated from those irradiated when 0-9 years of age. It is caused by the assumed high relative risk (a 2% increase per rem) of cancer induction in this young age group being projected onto the over 50 year age group when the spontaneous can- cer death rate is very high (see Table 3-3). No data are available as yet to test wheth- er this assumption is true or false. 3. Agreement between the absolute and rela- tive risk models is reasonably close except for the calculated excess of all other can- cers arising from the age group 0-9 years at the time of irradiation. The reason for the high numbers generated by the relative risk model for this case is discussed in the preceeding paragraph. Other differences are due either to the fact that no attempt was made to produce absolute internal con- sistency between the relative and absolute risk estimates (e.g., the 2% increase per rad for leukemia induction in the 10 + yr. olds represents an absolute risk estimate of 1.6/ 106/year/rem in the U.S. population, not 1.0), or to the fact that the relative risk model tends to give somewhat higher num- bers as the plateau region is projected into older age groups. 4. If the projections of these models are to be used for individuals occupationally ex- posed to radiation it is important to note that only the risk estimates for the "10 + year olds" will be relevant. The table below summarizes the projections of the two mod- els assuming exposure beginning at 20 years of age and ending at 65, first in terms of excess deaths in the U.S. population to 0.1 rem/year, and second in terms of the excess deaths from cancer per million peo- ple assuming exposure to 5 rem/year (the current standard for occupational expo- sure). This latter expression of the risk incorporates the assumption that the mil- lion people have an age and sex distribu- tion identical to that of individuals 20 years and older in the U.S. population (1967 statistics). Thus, the risk is obtained from the U.S. population figure by simple divi- sion by the number of people over 20 years of age (in millions) and multiplication by 50 (to convert from 0.1 rem/year to 5 rems/ year). The figures do not represent an indi- vidual,s chance of eventually dying from a radiation-induced cancer. Calculation of the excess annual number of cancer deaths for individuals exposed from 20 to 65 years of age. Exposure Conditions ABSOLUTE RISK MODEL RELATIVE RISK MODEL Excess Deaths Due to: Excess Deaths Due to: Leukemia All other Cancer Leukemia All other Cancer U.S. Pop,n 0.1 rem/yr 195 (a) 721 (b) 808 436 (a) 1,444 (b) 1,793 10 people: 5 rem/yr. 81 (a) 300 (b) 336 181 (a) 601 (b) 746 170

Table 3-2 Assumed values used in calculating estimates of risk shown in Table 3-1. Risk Estimate Duration Duration Absolute of Latent of Plateau Riskb Age at Ir- Type of Period Region (deaths/106/ radiation Cancer (years) (years)3 yr/rem) Relative Risk (% incr. in deaths/rem) In Leukemia 0 10 25 50 Utero All other cancer 0 10 25 50 0-9 Leukemia 2 25 2.0 5.0 Years All other cancer 15 (a)30 (b)Life 1.0 2.0 10 + Leukemia 2 25 1.0 2.0 Years All other cancer 15 (a)30 (b)Life 5.0 0.2 a Plateau region = interval following latent period during which risk remains elevated. b The absolute risk for those aged 10 or more at the time of irradiation for all cancer excluding leukemia can be broken down into the respective sites as follows: Type of Cancer Breast Lung GI incl. Stomach Bone All other cancer Total Deaths/106/year/rem 1.5* 1.3 1.0 0.2 1.0 5.0 * This is derived from the value of 6.0 quoted in Appendix II, Section Ale corrected for a 50% cure rate and the inclusion of males as well as females in the population. 171

m O £ C1iN i-iCNd- •—OOOd-l/l---s-JjCOWrMCN o 41 ss s^ i-iCNr^ioaDOrNC4i-^r*.Cio« ^ c o &S 00 m (M • £ 09 O OJ a a P1CN ^-CNJjOi-*OjJOvnin«CNlnrNCN 1— iWnCNi— -CNCNO4CN*— i C-l o • K O d- w z 0 • a vrv \fv n in m n m vi in n M in in m 5 in 5 m » aj r*.i*.. minOi-iCNrOd"in^f-aoosOr--rNnd'in 5 CO s*1 •&- mm o«CNCNCNCN4CNCNCNCNCNCNiommconin w H H o z • 01 in in ininmminnoooooooooo n) < z o S" L, ifl r)€0 0•-iCNrMCNCN•-.r-OooooOOOOO o -^ H C M M s mmminminmininininminininin O---CNcod'in*or*.aoiy*Or— cNcod-m id 2 <e O CN X Ll s u X ininminininOOOOooOooO m w •o + ca o ^ oooooooooooooooo « CO t »- o 0*8 u ai H •* X mnOooooOooOooooooO VO ij s- o O•-iCNCNCNcNCNrNrNCNCNrMCNCNCNrNCNCN m ~a 4) & CO « 0 CO O^rNCNCNrN^-O p.. u o ^ c o u in in M S* C 4) M 4J r*. !*• • 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 A p ro en m m CM n o x: S - to cNin cod-*aoaOoc o io u O 0 Z Q eas < 3 "1 CNiTiO--*i-i0Od"»Ooooooooooooo a 4J ^iomior*.voin^u-iinmu-iininininininmmin o M O H H £ M 1il T3 U >. c^^r--i^cnooooooooooooo ai u 0 -H H ^ 0) 4) P- O 4- O Oi—CNn^d-minmininmminminininm m 3 H _3 ^ ^ 4) -^ U >. in ln m r*. m Li d- r- ao o o i*. M n a 3 O CN ^t in 5 n •-i gfi O c - o u iD in r* . r*. i i i i i i i i i 1 i i 1 i 1 i i i i rt io Hi *O 4 m "S C 4) ln CO M 4J • io V) o • 3*1 mOiiOONo* — vOCNCNinaOCN^r*.O^WOOO^W iu — o*o*cocNinCNCNnocNaocNaoaocomvoCN"--aoao J- X ia a) ^H^CN^CTimcntNaoMCOr^inCNio •C » 0 . . LJ O J,s rHCNCOCOJ'd"nCN--- CN 1 35 iM ^5 jj^aor-id-CNrncNCticDy3inn*oCOP^rMOr^^i/i aoOi*.•-iioCNm^COaocNcorNn00o* OP*.CN a n O * * ^, tki 0 a) 4 > z _,£ 1 j'CTid-os-drONd"i»id"ON^iONd'ON^ONd"» + d-o*^-ir-ir*jcNnO^^tininoiori*ii*.aoao»» 1 1 1 1 1 1 1 1 i 1 1 1 1 1 1 1 1 1 1 1 O CU ou"ioui mOinomoinOinOinOinOino O 172

ce s. en O e 1 CJ I o ~ CU 4) -M co .5 X^ "S "=* II h 01 e ! o c cs .S •M — CO CO r-H GO ^ CO QJ 4J ^— , vO ON -^ i-4 r^ CN CO lA lA io O\ CT> in 4J CJ CO J3 rn c*-i i i— i c'j1 l0 O O r j r*^ --C CO co 0 X CU *.-- cvj m cn cs •-i ,^ H W a r-i CO r-l CO CO CO CU 4J ^ vQ oN *.^ f^ I*** CN en rn m 10 u O CO CO o 0 X CU *w rncni i— < C.i L", Q cr». ^c i*.~ ov 00 00 H W Q CN CM CN •-i 00 *H CM r-l CO u 60 I-l 00 M co O M o r-i CN CO (N r-l CN 2 a < § CO 55 1-l E l 1 l l I *— < <T r~* . 00 vO r* . 00 i— < in i— i cy» CM co in ON oo CM o M CO vO a r^ CN CM r-4 O M ,^ rH •o i— 1 I-l CO IJ 1-1 CO S o ^S. 1ll i-l r-H r- 1 CM CM i-i •-* CN CN 4J OV — ' r-l l CU o 3 w a CO 3J CO E 1ll r-lr-Hr-4CNi-«llIl *.3- ,-( CN i— 4 <t •— * l^ H 4J CO CO 0 CU o Q CO CO o •J a) c rf u cu vO ON in X 4-1 CO CO l l l l l t l l l l l a r-i CO CO co jr ON 04Oio>d-fnini~i|vovooNm co CO CU 4J co vo "^ ij" *^ ^f co \o in -^ CN f— ' vo 4-1 U CO r-i O X CU in H W Q CO b c5 l lCNOU-lCI>CNOM3vOCT.U-im rt i — O "'' CN 4J T3 o ^ HI -H CO 3 M iH Q CU PL, ^ CO CO OO U~l CJ> *^ i— 1 i— 1 CN cn m CN CN r-i i i i i i *.-4. vO y CO O) 1-4 .C C l U co 0 w a -c o •a J-l CO lJ 4J \o &* in u cu i-i 3 COCOlllllllll1l 1-1 0 M X fi CO alScnoAfiinSrocol^lfSl?!? -c § m r- oo r-ic^oovoincTioooominvocisr-i vO •o . Q. -H 00 O i^ o*oo>r**~d-r-iorocNt^r-iin^i r-*" r-i CU i— l •*H ^ r— i .0,52^^^553^5 + j (U S H i i i i i i i i i i i i omoinoinoininmininin -"•' "-""•*- °° H 173

REFERENCES (1) Jablon, S., and H. Kato: Radiation Dose and Mortality of A-bomb Survivors 1950-1970. JINH-ABCC Life Span Study. Report 6 (TR 10-71). 1971. (2) Stewart, A.: Radiation Cancers and A-bomb Survi- vors. Lancet 2:1203.1971. (3) Bennet, G.: Bristol floods of 1968. Controllert survey ol effects on health of local community disaster. Brit. Med. J. 3:454-458,1970. (4) Jablon, S.: Radiation Cancers and A-Bomb Survivors. Lancet 1:375,1972. (5) Court-Brown, W. M., and R. Doll: Mortality from Can- cer and Other Causes After Radiotherapy for Ankylos- ing Spondylitis. Brit. Med. J. 2:1327-1332,1965. B. Mortality From Causes of Death Other Than Cancer 1. Adult Experimental work has shown that single, sublethal doses of whole body radiation, and various regimens of divided doses, shorten the life expectation of exposed rats, mice, dogs, and other animals, and efforts to determine cause of death have suggested that excess mortality is not confined to cancer but extends rather generally over the spectrum of disease normally observed in these animals (1). The increase usually seems to be proportional to dose (1-3). On the basis of such data, estimates have been made of a life-shortening effect not only at the comparatively high doses and dose- rates used in experimental Work, but also at relatively low dose levels and for man; e.g., that life-expectation may be reduced by one to five days per roentgen (4,5). Information on man is comparatively sparse, and observations on dose and dose rate less adequate. Human data pertain to occupational exposure, diagnostic or therapeutic exposure, and the Japanese A-bomb survivors, and relate to a time during which mortality rates have been falling rapidly. There is now little doubt that human life can be, and has been, shortened by exposure to ionizing radiation, in that the evidence is clear that such exposure is leukemogenic and, more generally, carcinogenic. The issue here concerns the identification of other specific diseases caused by ionizing radiation, and the likelihood that there has been, or may be, a nonspecific life- shortening, i.e., a reduction in life expectation from diseases generally, not merely from malignant neoplasms. The great difficulty with the human data, of course, is that they derive not from experiments, but from diagnostic or therapeutic situations, or from occupational choices, the effects of which cannot surely be separated from the effects of ionizing radiation. The data on A-bomb survivors, furthermore, may not be entirely free of such confounding, since socio-economic characteris- tics were not distributed uniformly over each city and hence also over each dose level. Also whether the heavy acute mortality, and the widespread deprivation and disease associated with the disorganization of each city following the bombings, in any way modified the late ef- fects of radiation, remains unknown. By 1950, papers by March (6), Henshaw and Hawkins (7), Ulrich (8), and Dublin and Spiegelman (9), had clearly established that U. S. radiologists were at higher risk of death from leukemia than other physicians and than the general population. Dublin and Spiegelman, reporting on the mortality of U. S. medical specialists during 1938-1942, noted that specialists generally had lower age-standardized mortality ratios (SMR) than non-specialists (0.78 vs. 1.10), with specialists in roentgenology and radiology occupying a relatively high position among specialists (SMR = 0.90, third highest in their list of 12 specialists). With only 95 deaths observed among radiologists, they found only leukemia to have a remarkably high incidence in these physicians, in comparison with other specialists. In 1955 Warren (10) observed that the average age at death was about five years lower for U. S. radiologists generally than for other physicians. Since this was true not only for all causes of death but also for each of the many other causes of death he examined, Warren argued for a non-specific life- shortening effect of radiation in man. Although Warren gave some consideration to the comparability of radiologists and other physicians as to age, his analysis included no adjustment for differences in age structure, and the finding was challenged (29) on the basis that the five-year differential might reflect no more than differences in age composition. Stimulated by Warren,s report, Court Brown and Doll (11) examined the mortality of 1,377 174

British radiologists initially resident in Britain or Eire or belonging to the Colonial or Armed Services, over the period 1897-1956. With 463 deaths observed and expectation variously estimated on the basis of the general population, men in Social Class I, and physicans generally, they were unable to find evidence of non-specific life-shortening, but did observe a significant excess of mortality from cancer among men entering the practice of radiology before 1921, when serious attention seems first to have been paid to protective measures. Deaths from other causes were not elevated in any of the comparisons made. Selter and Sartwell reported a similar but preliminary study in 1959 (12), comparing 869 U. S. radiologists with 1,170 pathologists and bacteriologists as to mortality at ages 35 to 79 over the interval 1905-1956. With 235 deaths observed among the radiologists, and 244 among the pathologists and bacteriologists, and with incomplete reporting of cause of death, they found leukemia deaths to be definitely increased among radiologists, and deaths from other forms of cancer suggestively increased for radiologists elected to membership in their professional society in the period 1905-1914. In 1965 these same authors (13) reported more fully on the influence of occupational exposure to radiation, comparing the mortality of members of the Radiological Society of North America with that of members of the American College of Physicians and the American Academy of Ophthalmology and Otolaryngology, 16,339 men in all, with 3,421 deaths observed in the interval 1935-1958. In an age-controlled analysis they found the mortality of radiologists higher than that of the other groups throughout the period of study, but especially in 1935-1944. Most significantly, their analysis by cause of death showed that radiologists, especially at ages 65- 79, had higher death rates not only from cancer but also from cardiovascular-renal diseases and from all other causes combined. Relative to the mortality risk of members of the American Academy of Ophthalmology and Otolaryngology, mortality ratios for radiologists in 1935-1958 were 1.4 times expectation for all causes, 2.5 for leukemia, 1.6 for other forms of cancer, 1.2 for cardiovascular-renal diseases and 1.6 for all other causes. At ages 65-79 these ratios were 1.5 for all causes, 1.9 for leukemia, 1.5 for other forms of cancer, 1.4 for cardiovascular-renal diseases, and 2.0 for all other causes. In matched-pair analyses they found a relative mortality risk of 1.3 for radiologists entering the specialty in 1921-1939 and no excess among men entering thereafter. A decreased rate of mortality of radiologists in recent years, vis a vis the general population, has also been reported by Warren (14). In 1965 Court-Brown and Doll (15) reported their 5-to-25 year mortality follow-up on 14,554 patients with ankylosing spondylitis treated by x rays during the period 1935-1954. They classified the 1,582 deaths observed into (a) those directly attributable to arthritis and other forms of rheumatism, excluding rheumatic fever, (b) those attributed to conditions known to be associated with ankylosing spondylitis, e.g., ulcerative colitis, (c) leukemia, aplastic anemia, and cancer of heavily irradiated sites, and (d) others considered to be unrelated to the underlying disease and cancer of lightly irradiated sites. Expected deaths were calculated from the corresponding national mortality rates for England and Wales. Deaths observed in Group D numbered 812 vs. 608 expected, a mortality ratio of 1.3. Court-Brown and Doll found it difficult to interpret this excess mortality from unrelated causes, but were reluctant to consider it evidence of nonspecific aging in view of the other possible explanations. Miller and Jablon (16) recently reported an 18-year follow-up of 6,560 men who served as x- ray technologists in the U. S. Army during World War II, compared with 1,522 pharmacy and 5,304 medical technologists. With 289 deaths observed among the x-ray technicians, and 256 among the other two groups of technicians, they found only a questionable increase in bronchogenic carcinoma among the x-ray technicians, and no evidence of any general mortality increase. In his recent report on 3,239 Massachusetts dentists (17), most of whom received some exposure to radiation in their work, Warren found no evidence of excess mortality in comparison with U. S. white males of comparable age. Duncan and Howell, studying the experience of employees of the United Kingdom Atomic Energy Authority during the period 1962-1968 (18), when exposure averaged 0.3 man-rad/year, found no 175

association between the working environment and morbidity, and fewer deaths than expected from the mortality rates of the general population of the same age and sex (18). Although the mortality comparison is based on 200,000 man-years of exposure, use of death rates of the general population seems inappropriate for an employed population. The morbidity data pertain to 69,000 man-years of observation during 1964-1968 and are internally controlled. Tachikawa and Kato (19) have recently reported the mortality experience of Hiroshima A-bomb survivors ascertained by means of a Hiroshima City survey in August 1946. Mortality from 1 October 1946 to 1 October 1950 was highest for those closest to the hypocenter at the time of the bomb (ATB). However, only for leukemia and deaths of unknown cause was there any relationship with distance ATB. The cohort of 82,271 A- bomb victims under continuing mortality surveillance at the Atomic Bomb Casualty Commission (20) experienced 13,093 deaths from 1 October 1950 to the end of September 1966. These deaths have recently been analyzed by Beebe et al. (21) in relation to the revised (T- 65) dose as well as to distance ATB, and with regard to about 50 cause-of-death groups. The hypothesis of accelerated aging was one of many considered in this analysis. Apart from the excess mortality observed for leukemia and for diseases of the blood and blood-forming organs throughout the 16-year period, and for forms of cancer other than leukemia, especially in the 1962-1966 interval, systematic mortality differentials associated with distance or dose were not seen. Jablon and Kato have re- examined this material (28) and extended the period of observation through 1970. In the six- year period, 1965-1970, they find no more than suggestive evidence (P = 0.06) of an increase in mortality from all the diseases except neoplasms among A-bomb survivors exposed to more than 100 rads. The suggestion rests on an estimated excess of 24 deaths above the expected 218. Their analysis by age and by disease-groups and systems throws no further light on the source of the possible discrepancey. Thus far the experience of the A-bomb survivors does not confirm the hypothesis of accelerated aging, but it remains possible that the youngest victims of the bombs will eventually show a disturbance of mortality patterns consistent with the hypothesis of accelerated aging. By 1970 those under age 10 ATB were under 36 years of age. The mortality differences seen thus far seem better explained in terms of more specific relationship between ionizing radiation and individual diseases or groups of diseases, especially the leukemias, other malignant neoplasms, and diseases of blood and blood-forming organs. The argument for a nonspecific aging effect of radiation in man rests on an extrapolation from animal data, and on inferences from comparisons of occupational groups and patient-groups which are open to the possible influence of other factors, notably those associated with occupational choice and with the diseases being treated by radiation. In contrast, the exposure of the Japanese in Hiroshima and Nagasaki is relatively free from such influences, but only relatively because socio-economic characteristics are not uniformly distributed over each city and hence also over all dose levels. It differs also in having been a single, whole-body dose ranging from the neighborhood of 0 at several km from hypocenter to supralethal amounts in the vicinity of the hypocenter. Immediate mortality from the bombs exceeded 50 percent at about 1.25 km in Hiroshima and 1.35 km in Nagasaki (22); 20 percent mortality, in turn, corresponds to about 1.75 km in Hiroshima and 1.80 km in Nagasaki. Whether the initial mortality was selective in the sense that survivors would be less vulnerable to late chronic effects remains unknown. Nor is the evidence drawn from occupational and patient- group comparisons uniformly suggestive of the existence of a nonspecific aging effect. The hypothesis remains unproved but the evidence in its favor is strong enough to require further investigation. The age-adjusted data on U. S. radiologists (13, 29) have not been reported in great detail by cause, but a relative risk estimate of 1.2 is given for cardiovascular-renal diseases, and mortality from these causes is also elevated among the patients with ankylosing spondylitis compared with national death rates. These diseases have been intensively studied at ABCC on the basis of both clinical and autopsy observations (23-26), with no 176

suggestion of a radiation effect ever having been seen. It should be n< ted, however, that some of the specific diseases included in the cardiovascular-renal complex have very different mortality rates in Japan in comparison with the U. S. (27). Apart from cancer and the cardiovascular-renal diseases, it is only for diseases of blood and blood- forming organs that impressive differentials in mortality have been associated with radiation (11, 21) but it seems likely that this is no more than a small part of the leukemogenic effect, misclassified. REFERENCES (1) Lindop, P. J., and Sacher, G. A. (eds.): Radiation and Ageing, Proceedings of a Colloquium Held in Semmering, Austria, June 23-24, 1966, London, Taylor and Francis Ltd., 1966. (2) Mole, R. H.: Shortening- of Life by Chronic Irradiation: The Experimental Facts, Nature 180:456-460 (Sept.) 1957 (3) United Nations: Report of the Scientific Committee on the Effects of Atomic Radiation, New York, United Na- tions, 1962. (4) Jones, H. B.: The Nature of Radioactive Fallout and Its Effects, Washington, D. C., U. S. Government Print- ing Office, 1957. (5) Failla, G., and McClement, P.: The Shortening of Life by Chronic Whole-Body Irradiation, Amer. J. Roentgen- ol. Rad. Therap. Nucl. Med. 78:946-954,1957. (6) March, H. C.: Leukemia in Radiologists, Radiology 43:275-278 (Sept), 1944. (7) Henshaw, P. S., and Hawkins, J. W.: Incidence of Leu- kemia in Physicians, J. Nat. Cancer Inst. 4:339-346 (Feb.) 1944. (8) Ulrich, H.: Incidence of Leukemia in Radiologists, New England J. Med. 234:45-46 (Jan.) 1946. (9) Dublin, L. I., and Spiegleman, M.: Mortality of Medical Specialists 1938-1942, JAMA 137:1519-1524 (Aug.) 1948. (10) Warren, S.: Longevity and Causes of Death from Ir- radiation in Physicians JAMA 162:464-468 (Sept.) 1956. (11) Court Brown, W. M., and Doll, R.: Expectation of Life and Mortality from Cancer Among British Radiologists, Brit. Med. J. 2:181-187 (July) 1958. (12) Seltser, R., and Sartwell, P. E.: The Application of Cohort Analysis to the Study of Ionizing Radiation and Longevity in Physicians, Amer. J. Public Health 49:1610- 1620 (Dec.) 1959. (13) Seltser, R. and Sartwell, P. E.: The Influence of Occu- pational Exposure to Radiation on the Mortality of American Radiologists and Other Medical Specialists, Amer. J. Epid. 81:2-22 (Dec.) 1965. (14) Warren, S.: The Basis for the Limit on Whole-Body Exposure-Experience of Radiologists, Health Physics 12:737-741,1966. (15) Court Brown, W. M., and Doll, R.: Mortality from Cancer and Other Causes After Radiotherapy for Anky- losing Spondylitis, Brit. Med. J. 2:1327-1332 (Dec.) 1965 (16) Miller, R.W., and Jablon, S.: A Search for Late Radia- tion Effects Among Men Who Served as X-ray Technolo- gists in the U.S. Army During World War II, Radiology 96:269-274 (Aug.) 1970. (17) Warren S., and Lombard, O.M. Mortality and Radia- tion Exposures of Massachusetts Dentists, J. Amer. Dent. Assc. 80:329-330 (Feb.) 1970. (18) Duncan, K.P., and Howell, R.W.: Health of Workers in the United Kingdom Atomic Energy Authority, Health Physics 19:285-291, (Aug.) 1970. (19) Tachikawa, K., and Kato, H.: Mortality Among A- bomb Survivors, October 1945-September 1964, based on 1946 Casualty Survey, Hiroshima, ABCC Technical Re- port 6-69, Hiroshima, Japan. (20) Jablon, S., Ishida, M., and Yamasaki, M.: Studies of the Mortality of A-bomb Survivors. 3. Description of the Sample and Mortality, 1950-1960, Radiat. Res. 25:25-52 (May) 1965. (21) Beebe, G. W., Kato, H., and Land, C.E.: JNIH-ABCC Life Span Study, Hiroshima and Nagasaki. 5. Mortality and Radiation Dose, Oct. 1950-Sept. 1966, ABCC Techni- cal Report #11-70, Hiroshima, Japan. (22) Oughterson, A.W., and Warren, S. (eds.): Medical Effects of the Atomic Bomb in Japan, New York, Mc- Graw-Hill Book Co., Inc., 1956. (23) Yano, K., and Ueda, S.: Coronary Heart Disease in Hiroshima, Japan: Analysis of the Data at the Initial Examination 1958-1960, Yale J. Biol. Med. 35:504-522 (June) 1963. (24) Johnson, K.G. et al: Coronary Heart Disease in Hiro- shima. Japan: A Report of a Six-Year Period of Surveil- lance, 1958-1964, Amer. J. Public Health 38:1355-1367, (Aug.) 1968. (25) Johnson, K.G., et al: Cerebral Vascular Disease in Hiroshima, J. Chron. Dis. 20:545-559 (July 1967) (26) Beebe, G.W. et al.: ABCC-JNIH Pathology Studies, Hiroshima-Nagasaki Report 2, Oct. 1950 - Dec. 1965, ABCC TR-8-67, and Hiroshima Igaku 21:729-735,1968. (27) Gordon, T.: Mortality Experience Among the Ja- panese in the U.S., Hawaii and Japan, Public Health Rep. 72:543-553 (June) 1957. (28) Jablon S., and Kato, H.: JNIH-ABCC Life Span Study Report No. 6. Radiation Dose and Mortality of A-bomb Survivors, 1950-1970, ABCC Technical Rpt. #10-71, Hi- roshima, Japan. (29) Seltzer R. and P.E. Sartwell: Ionizing- Radiation and Longevity of Physicians. JAMA 166:585-587,1958. 2. Infant Mortality and Ionizing Radiation In 1954, Yamazaki et al. (1) reported in- creased rates of fetal death and of infant mor- tality among children who were in utero at the time of the atomic bombing of Nagasaki. They divided the population into three subgroups: (a) 30 fetuses whose mothers were exposed within 2000 meters and had signs of major ra- diation injury, (b) 68 whose mothers were ex- posed within 2000 meters without evidence of 489-797 0-72-l3 177

radiation injury, and (c) 133 "controls" whose mothers were between 4000 and 5000 meters distant from the hypocenter. In these three groups there were, respectively, 7 (23%), 3 (4%) and 3 (3%) fetal deaths. Among livebirths in the three groups, there were, respectively, 6 (26%), 3 (5%), and 4 (4%) neonatal and infant deaths. The numbers are, of course, very small. In addition, as the authors pointed out, it is difficult to interpret the high fetal and infant loss rates in the heavily exposed group since factors other than radiation, such as trauma, burns, infections, etc., may have played a role. A substantially larger series, encompassing the data both from Hiroshima and Nagaski, has been reported recently by Kato (2). The dis- tribution of exposed fetuses by estimated ma- ternal dose received, and the observed and ex- pected infant deaths (the expected being based on the death rates in the group as a whole, standardized for sex and trimester of exposure) are as follows: Estimated maternal dose (rads) 0-9 10-39 40-79 80+ Unknown TOTAL Infants exposed in utero Infant deaths: Obs. year)Exp. Observed/Expected 795 60 62.9 0.95 223 11 18.0 0.61 180 19 15.1 1.26 68 13 6.1 2.14 26 1 1.9 1292 104 104.0 1.00 There is little doubt that the more heavily exposed infants experienced higher infant mor- tality rates than the lightly exposed. There are, however, some features of the data that sug- gest that factors other than radiation may have been responsible. Specifically, the in- crease in mortality was noted only for infants exposed in the third trimester of pregnancy, and there was a striking increase in the propor- tion of low birth weight infants in the highest exposure group (35% under 2500 meters, com- pared to 9% in the 0-9 rad group). Factors oth- er than radiation operating in the third trimes- ter could have affected birth weight possibly by induction of premature labor. Hence some of the concomitants of heavy exposure other than the radiation per se require further investiga- tion before a judgment can be made as to the role of irradiation in the observed excess of mortality (3). A genetic mechanism for association between environmental radiation levels and infant mor- tality has been proposed recently by Stern- glass (4). The evidence presented to support this hypothesis consists of temporal and geograph- ic correlations of infant mortality rates with levels of radioactive fallout from atomic weap- ons testing or from nuclear power installa- tions. In addition, experimental data on mice are cited as evidence that genetic effects of strontium-90 have been observed. That the mechanism of increased infant mortality asso- ciated with increase in environmental radia- tion is genetic, rather than somatic, is deduced from a lag of five years between the fallout from the first weapons test in New Mexico in 1945 and appearance of the increased mortali- ty. The evidence assembled by Sternglass has been critically reviewed by Lindop and Rotblat (5) and by Tompkins and Brown (6). It is clear that the correlations presented in support of the hypothesis depend on arbitrary selection of data supporting the hypothesis and the ignor- ing of those that do not. In several regards, the data used by Sternglass appear to be in error. One of the most vital assumptions in the model —that without the atomic tests the infant mor- tality rate would have continued to fall in a geometrically linear fashion—is without basis either in theory or in observation of trends in other countries and other times. The doses of strontium-90 used in the experiments referred to by Sternglass were of the order of 100,000 times greater than those received by humans from all the atomic tests and were associated with extremely small differences in infant mor- tality (8.7% in the irradiated vs. 7.5% in the control mice) (5). 178

In short, there is at the present time no con- vincing evidence that the low levels of radia- tion in question are associated with increased risk of mortality in infancy. Hence, for the purposes of this report, no estimate of risks are considered to be applicable. REFERENCES (1) Yamazaki, J. N., Wright, S. W., and Wright, P. M. Outcome of pregnancy in women exposed to the atomic bomb in Nagasaki. (2) Kato, H. Mortality in children exposed to the A-bombs while in utero 1945-1969. ABCC Tech. Rep. 23-70,1970. (3) MacMahon, B. Radiation exposure in utero and mortality. Amer. J. Epidemiol. 95:3,1972. (4) Sternglass, E. J. Evidence for low-level radiation ef- fects on the human embryo and fetus. In: Radiation Biol- ogy of the Fetal and Juvenile Mammal (edited by M.R. Sikov and S. D. Mahlum, U.S.A.E.C., Div. of Technical Infor., Oak Ridge, 1969, pp. 693-719 (See also, General Bibliography) (5) Lindop, P. J., and Rotblat, J., Strontium-90 and infant mortality. Nature 224:1257-1260,1969. (6) Tompkins, E., and Brown, M. L. Evaluation of a Possi- ble Causal Relationship Between Fallout Deposition of Strontium-90 and Infant and Fetal Mortality Trends. DBE 69-2, Clearinghouse for Federal and Scientific Tech- nical Information, Springfield, Virginia 22151. October 1969. C. Morbidity from Causes Other Than Cancer 1. Cataracts Although it is generally accepted that the human lens is sensitive to opacification by fast neutrons, the dose-response relation for induc- tion of cataracts severe enough to impair vi- sion seems clearly to be sigmoid, at least for low-LET radiation (T). In the case of x rays and gamma rays, the threshold varies from 200 to 500 rads, when delivered in a single brief expo- sure, to 1000 rads or more when delivered over a period of months (1, 2). For high-LET radia- tions, the data are fragmentary; however, ob- servations suggest that there is a threshold for fission neutrons in the vicinity of 75-100 rads, which depends less on intensity than does the threshold for low-LET radiations (1). Despite evidence for the existence of a threshold for opacities severe enough to impair vision, minute amounts of radiation (less than 5 rads) may suffice to cause microscopically detectable lens changes in radiosensitive spec- ies such as the mouse (3, 4). Whether corre- sponding changes might ultimately be elicited by comparable doses in the human lens after a suitably long latent period remains to be deter- mined. REFERENCES (1) International Commission on Radiological Protection. Publication 14. Radiosensitivity and Spatial Distribu- tion of Dose. Pergamon Press, Oxford, 1969. (2) Langham, W. H., editor, Radiobiological Factors in Manned Space Flight, National Academy of Sciences - National Research Council, Publication 1487, National Academy of Sciences, Washington, 1967. (3) Upton, A. L., Christenberry, K. W., Melville, G. S., Furth, J., and Hurst, G. S.: The Relative Biological Effectiveness of Neutrons, X-Rays, and Gamma Rays for the Production of Lens Opacities: Observations on Mice, Rats, Guinea-Pigs, and Rabbits, Radiology, 67:686-696, 1956. (4) Bateman, J. L., and Bond, V. P.: Lens Opacification in Mice Exposed to Fast Neutrons. Radiation Res. Supl. 7:239-249,1967. 2. Central Nervous System Effects The central nervous system is relatively re- sistant to induction of structural changes by irradiation at dose levels below several hundred rems except early in its growth and development (1); however, it has been reported to show transitory functional changes in re- sponse to acute doses as low as 1 rem (2). It has not been demonstrated that these changes pro- duce any injurious effects. Effects on the developing nervous system are surveyed elsewhere in this report (see Chapter VI). REFERENCES (1) Rubin, P., and Casarett, G. W.: Clinical Radiation Pathology W. B. Saunders, Philadelphia 1968. (2) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. General Assembly. Official Records: 24th Session, Supplement No. 13 (A/7613), New York, 1969. 3. Impairment of Fertility Testis Among the most radiosensitive cells of the body are primitive sperm precursors, the sper- matogonia; by contrast, mature spermatozoa 179

are relatively resistant to killing. Spermato- gonia are drastically depleted by small amounts of radiation; i.e., a dose of 50 rem de- livered in a single brief exposure may result in cessation of sperm formation. Fertility is not impaired, however, until the pre-existing sperm cells and those formed from the maturation of surviving spermatocytes and spermatids are eliminated from the genital tract, which takes several weeks (1). The sterility ensuing after such a dose may be expected to be only tempo- rary, since enough spermatogonia survive to restore spermatogenesis through eventual re- generation of the seminiferous epithelium (2). In men who have received testicular irradia- tion in criticality accidents or radiotherapy, the time required for the sperm count to return to normal has varied from about one year after a dose 100 rem to more than three years after near-lethal exposure (2). Because acute whole- body irradiation has not been observed to cause permanent sterility, the sterilizing dose for man, as for other male mammals, is thought to exceed the lethal dose if applied to the whole body in a single, brief exposure (3). Protracted or fractionated whole-body irra- diation, on the other hand, has caused perma- nent sterility in male laboratory animals, de- pending on the dose rate and quality of radia- tion (4). In dogs exposed daily to x radiation for the duration of life, no change in sperm count was detectable at or below a dose rate of 0.6 rem per week, whereas sterility occurred within months at a dose rate of 3.0 rem per week (5). It is conceivable, therefore, that permanent steri- lization of the human testis might also result from protracted exposure, although such an effect has yet to be reported. In any event, it seems unlikely that impairment of fertility would occur at dose levels compatible with ex- isting radiation protection standards (6). Ovary Unlike the testis, the ovary possesses its en- tire supply of germ cells, or oocytes, early in life and lacks the ability to replace them as they are lost subsequently. Hence, since ooc- ytes are relatively radiosensitive, irradiation causes a lasting reduction in the reproductive potential of the affected ovary, varying in se- verity with species, age, and other factors (7). Data on the response of the human ovary come chiefly from observations of the effects of therapeutic irradiation, from studies of Ja- panese atomic bomb survivors, and from inves- tigation of Marshallese women exposed to fall- out (7). The data suggest that a minimum of 300-400 rems must usually be given in a single exposure to insure permanent sterility and that an even larger dose (i.e., 1000-2000 rems) is required for sterilization if administered to young women in fractionated exposures over a period of 10-14 days (6-7). That ovarian sensi- tivity may vary, however, is implied by the suggestion that a single dose of 170 rems is "not without risk of permanent sterilization" in some young women, although the same total dose given in two or three weekly fractions may apparently increase fertility in others who have a history of infertility (6). Follow-up stud- ies of Japanese atomic-bomb survivors and Marshallese women exposed to nuclear fallout have disclosed no lasting impairment of fecund- ity (6). REFERENCES (1) Langham, W. H., editor: Radiobiological Factors in Manned Space Flight National Academy of Sciences- National Research Council Publication 1487. National Academy of Sciences, Washington, 1967. (2) Upton, A. C.: Effects of Radiation on Man. Ann. Rev. Nucl. Sci., 18:495-528,1968. (3) United Nations Scientific Committee on the Effects of Atomic Radiation. Report. General Assembly. Official Records: 17th Session, Supplement No. 16 (A5216), Unit- ed Nations, New York, 1962. (4) Rubin, P., and Casarett, G. W.: Clinical Radiation Pathology, W. B. Saunders, Philadelphia, 1968. (5) Casarett, G. W., and Eddy, H. A.: Fractionation of Dose in Radiation - Induced Male Sterility. In: Dose Rate in Mammalian Radiation Biology, edited by D. G. Brown. R. G. Cragle, and R. R. Noonan, CONF-680410. U.S.A.E.C. Div. of Tech. Information, Oak Ridge, Tennes- see, 1968, pp. 14.1-14.6. (6) International Commission on Radiological Protection. Publication 14. Radiosensitivity and Spatial Distribu- tion of Dose, Pergamon Press, Oxford, 1969. (7) Baker, T. G.: Radiosensitivity of Mammalian oocytes with particular reference to the human female. Amer. J. Obstet. and Gyn. 110:746-761, 1971. 180

Appendix III. Analysis of Viewpoints on Record A. United Nations Reports The United Nations Scientific Committee on the Effects of Atomic Radiation in its 1964 re- port (1) estimated the increased risk of radia- tion-induced leukemia to be one to two cases per million people exposed per rad per year. For thyroid cancer, the estimate was one case per rad per year per million. On the basis of the data then available, it was estimated that ap- proximatley one other cancer (excluding thy- roid) would occur for every leukemia induced by irradiation. The Committee cautioned that these estimates were reliable only for the dos- age range where the data were reasonably good (above about 100 rads) but that extrapolation to low doses could be considered to give an up- per limit to the risk. It was pointed out that the duration of the period of increased risk was not known and that since cancers other than leuke- mia might well have longer latent periods for their induction, the ratio of other cancers to leukemia might increase as the period of fol- low-up was extended. In both the 1962 (2) and 1964 (1) reports, the United Nations Committee emphasized that radiation delivered at low dose rates was probably much less effective than that given at high rates but that the data were inadequate to make a numerical estimate of the factor of change in effectiveness. B. Reports of the International Commission on Radiological Protection In 1965, a Task Group of the International Commission on Radiological Protection (ICRP) submitted a report to ICRP,s Committee I on "The Evaluation of Risks from Radiation" (3). In essence, the estimates of risk of radiation- induced neoplasms were identical to those of the U. N. Committee. Using the linear, nonthreshold hypothesis, the Task Group esti- mated that if one million persons were exposed to one rad the total excess number of cancers over the lifetime of this population would be: 20 leukemias, 20 other cancers (except thyroid) and 10 to 30 thyroid cancers. They assumed the period of increased risk would extend from 10 to 20 years which, of course, makes their esti- mate correspond to 1 to 2 leukemias per million persons exposed per rad per year, etc. They in- dicated that there might subsequently be found an increased risk of other cancers (because of the longer latent period) but they considered it unlikely that the increase would reach a factor of ten. The Task Group considered these as upper limits of risk and pointed out that at very low doses and dose rates some of these effects might not occur at all. The shortcomings and uncertainties in the risk estimates from the UN and ICRP commit- tees were well recognized and discussed in de- tail in their reports. Both reports depended heavily on data from the Atomic Bomb Casual- ty Commission (ABCC) and from patients treated with radiation for ankylosing spondyli- tis. One of the uncertainties has been and will continue to be reduced with the passage of time, namely, the question of whether the ratio of other neoplasms to leukemia will increase as the follow-up progresses. The uncertainties concerning the effect of dose rate and the shape of the dose-response curve will require data for their resolution and the necessary data do not appear in the offing. More recently, two Task Groups of Commit- tee I of the ICRP have reported on "Radiosensi- tivity and Spatial Distribution of Dose" (4). The Group reporting on radiosensitivity at- tempted to estimate the order of sensitivity to cancer induction of the various organs and parts of the body. They were able to do this in a tentative way but pointed out that the data were adequate for risk estimates (slope of the assumed linear dose-response relationship) only for two tissues, namely, bone marrow and thyroid. They considered the lymph nodes and reticular tissue to have a high sensitivity but properly based risk estimates were not availa- ble. Other tissues that might have a high sensi- tivity included the pharynx, bronchus, pancre- as, stomach, and large intestine, but the evid- ence was considered to be "incomplete or capa- ble of other interpretations." A few organs (ovary, breast, prostate, uterus, bladder) were not classified, and the remaining tissues were believed to have a low sensitivity. The question of the ratio of other cancers to leukemias was again considered and the Task Group on Spatial Distribution of Dose estimat- ed that by 27 years after exposure to radiation this ratio might be as high as 5 or 6 to 1. This 181

conclusion was based on data from Court Brown and Doll (5) on the spondylitics. The pa- tients with the longest period of follow-up showed no evidence of a decrease in the excess incidence of "other cancers" whereas the peri- od of high risk from leukemia appeared to end at about 14 years. Perhaps the principal criticism of this conclusion arises from the fact that the majority of the excess cancers originated in the bronchus or stomach. The Task Group on Radi- osensitivity commented on the excess of these two types of cancer in the spondylitics as fol- lows: ". . .the large quantities of drugs, espe- cially analgesics, which spondylitics inevitably consume, may conceivably have a direct carcin- ogenic effect on the stomach or an indirect and synergestic action together with irradiation. Similarly, the change in respiratory dynamics associated with increased rigidity of the chest may increase the carcinogenicity of cigarette smoking and so make invalid the calculation from national statistics of the expected number of cases of bronchial carcinoma in the absence of irradiation. Similar reasoning might be used to explain away the excess of cancer of the pharynx. There are no data on unirradiated spondylitics which would allow these sugges- tions to be checked." It is pertinent to note that the combination of aminopyrine, an analgesic drug formerly in wide use, with sodium nitrite, a food additive used extensively, yields nitrosa- mine. The reaction takes place at pH values found in the stomach (6). Nitrosarriines have been shown to be potent carcinogens in ani- mals, yielding malignancies of the stomach and lung. The incidence of these cancers in the ex- posed patients was high. It would be of consid- erable importance to determine how many of the patients had, in fact, used aminopyrine. C. Report by Dolphin and Marley Dolphin and Marley (7), drawing on essential- ly the same data as the Task Groups of ICRP, provided risk estimates that were very similar to those of the Task Groups. They estimated the total risk of leukemia as 20 cases per 106 man-rads, and that the risk of thyroid cancer is 30 and 10 per 106 man-rads for children and adults, respectively. They noted that thyroid cancer is not usually fatal and suggested that for risk calculations only 10% should be consi- dered as fatal. Dolphin and Marley also specifi- cally recommended that these risk coefficients not be applied to dose levels below 10 rads. D. Report by Mole Mole (8) has pointed out that since the ratio of other cancers to leukemia will probably reach 4 or 5 to 1 in the spondylitics who re- ceived partial-body radiation the ratio might reach 10 tol in persons receiving total body radiation. The view extends the conclusion of the ICRP Task Group Report (ICRP 14), which attempted to evaluate from independent sources the radiosensitivity of organs not in- cluded in the irradiation field in the case of spondylitis. The Task Group was unable to es- tablish a high sensitivity except in the case of the thyroid gland. E. Report by Baum Despite the seemingly rather general ac- ceptance of the linear hypothesis for protec- tion purposes, there have been recent reports suggesting certain alternatives. For example, Baum (9) has elected to fit a power function to certain selected data, and from the slope con- stant he has sought to deduce whether the curve (on a linear plot) is convex, straight, or concave. The data for neoplasms in human populations that he used to illustrate the pow- er function were all from studies of the ABCC. These included: all malignancies (Hiroshima), lung cancer (both cities), stomach cancer (both cities), breast cancer (both cities), acute leuke- mia (Nagasaki), and all leukemias (Hiroshima) (10-16). In every case, he found the exponent of the power function to be less than 1.0, implying a convex curve on a linear plot (greater effec- tiveness per rad at lower doses). Data for lung, stomach, and breast cancer are too fragmen- tary and the numbers of cases too small to jus- tify the fitting of any regression equation (Miller (10), ICRP 14 (4)) to say nothing of choosing among alternatives. 182

F. Reports by Gofman and Tamplin Estimates have been made that if the current FRC guidelines for the maximum "allowable" dose from all nonmedical sources of 0.17 rem per year to the general population is reached, some 32,000 extra deaths per year from all ma- lignancies in the U. S. population will result (17-19). More recently, the same authors have refined their estimates, concluding that the additional cancer death rate may be as high as approximately 104,000 per year or an increase of 34.3% over the current rate (20). The basis for the estimates rest on three generalizations: GENERALIZATION I All forms of cancer, in all probability, can be increased by ionizing radiation, and the correct way to describe the phenomenon is either in terms of the dose required to double the spontaneous mortality rate for each can- cer or, alternatively, of the increase in mor- tality rate of such cancers per rad of expo- sure. GENERALIZATION II All forms of cancer show closely similar doubling doses and closely similar percen- tage increases in cancer mortality rate per rad. GENERALIZATION III Youthful subjects require less radiation to increase the mortality rate by a specified fraction than do adults. In addition to these three generalizations, two further assumptions are made, viz: (i) The radiation dose-response curve is line- ar at all dose levels. (ii) There is no dose-rate effect for any type of radiation-induced malignancy. Analysis of certain data on radiation-induced neoplasia in the light of these assumptions enables the authors to construct the following table of the age-specific radiosensitivity of cancer induction (20). Table III-1: Variation in Cancer Induction per rad (Table 4 of ref. 20) Age at Irradiation Increase in Cancer Mortality Rate per rad (in plateau region) (percent) In utero 0-5 years 6-10 11-15 16-20 21-30 31-40 41-50 51-60 61 + 50 10 8 6 4 2 1 O.o 0.25 Assumed negligible These relative risk estimates are then ap- plied to the hypothetical case of exposure of the U. S. population to 0.17 rem/year using three models differing in the lengths of the la- tency period and plateau region. The most "pes- simistic" case, which assumes latency periods of 5 and 15 years for in utero irradiation and all subsequent irradiation respectively, and a plateau region which never returns to the spon- taneous rate, predicts an annual radiation- induced cancer mortality rate of 104,259 cases, or a 34.3% increase in the present rate. The most "optimistic" case, which assumes 5 and 10 year latency periods for in utero and all other irradiation, together with a plateau region of 20 years, predicts an increased cancer mortali- ty rate of 9,428 cases or a 3.1% increase. Careful analysis both of the data used by the authors in their generalizations and risk esti- mates and of other relevant studies has led to the following conclusions: (a) The epidemiological studies of Stewart and Kneale (21) and of MacMahon (22) on the incidence of malignancy in offspring of mothers given diagnostic irradiation during pregnancy support generalizations I and II and also the risk estimate of a 50% increase in cancer mor- tality per rad for in utero irradiation. The fact that it cannot be proved conclusively that the diagnostic irradiation was the causative agent in producing the increased cancer mortality cannot be used to reject the data. (b) The data of Court Brown and Doll (23) on the follow-up of patients treated in early adult- hood for ankylosing spondylitis constitute a 183

O Age specific rate Mean for all ages Age at Birth O-9 Age at Death 2O-34 1O-19 30-44 20-34 40-59 35-49 55-74 Figure Ill-l. The relative risk of induction of all malignancies (except leukemia) in the A-bomb survivors versus age at the time of the bomb (ATB). These data are for both cities and sexes combined for deaths between 1960 and 1970. No neutron RBE has been applied. 184

critical part of the evidence used in support of generalizations I and II. Basically, these two rules would predict that in a population receiv- ing whole-body irradiation the ratio of the ex- cess number of cancers to excess number of leukemias should be the same as the ratio of spontaneous rates experienced by a non-irra- diated population. Since the spondylitics did not receive whole-body irradiation, a correction factor has to be applied to the so-called "heavi- ly irradiated" sites to equalize their dose to that of the bone marrow (presumably the site of origin of the radiation-induced leukemias). This correction factor was obtained from esti- mates made by Dolphin and Eve (24) of the mean dose received by the stomach in the spon- dylitics. These authors stated that the mean stomach dose might well be only 0.07 of the mean dose to the spinal marrow. Applying this factor brings the stomach dose to 61.6 R com- pared with the estimated dose to the entire marrow of 352 R. This reasoning means that the ratio of excess stomach cancer deaths to excess leukemic deaths should be multiplied by 352 61.6 (=5.71) to correct for the estimated differ- ence in radiation dose. It is clear from the above that the estimation of the mean stomach dose from the mean dose to the spinal marrow is crucial. The committee has reconsidered this question and has concluded that the previous estimate made by Dolphin and Eve is seriously in error. A full discussion of this evaluation is given in Appendix V. It is sufficient to state here the conclusion, namely, that the minimum value of the mean dose to the stomach is 250 rads, and a value of double this is not unlikely. If it is assumed that this dose estimate applies to all "heavily irradiated" sites (as Gofman et al. assumed) then the percent increase in can- cer mortality per rad becomes 0.2 to 0.4, or a factor of 5 to 10 less than that predicted (for persons aged 21-30). (c) The most unequivocal data on the induc- tion of neoplasms in humans comes from the ABCC studies. Analysis of the death certificate data up to 1970 reveals that the doubling dose for the induction of all malignancies (excluding leukemia) in those exposed when aged ten or more is at least 500 rads (Table h-2), represent- ing a 0.2 percent increase per rad. Table III-l shows that Gofman et al. (20) assumed a per- cent increase per rad varying from 6 for 11-15 year olds to 0.25 for 51-60 year olds. In calcu- lating lifetime risk estimates, this sliding scale of relative sensitivity is equivalent to a con- stant value of at least 10 times higher than that derived from the most recent ABCC stud- ies. Figure III-1 shows the ABCC data up to 1970 in terms of percent increase in the death rate from all malignancies except leukemia against age ATB. (d) Analysis of the ABCC data shows that those survivors who were exposed when less than ten years old suffered a higher relative risk of developing a malignancy than did those who were older than ten at the time of the bomb. That is, in terms of relative risk, gener- alization III is correct; youthful subjects do require less radiation to increase the mortality rate from cancer by a specified fraction than do adults. However, the best estimate of the per- cent increase in cancer mortality per rad that can be made for all cancer except leukemia (Table h-1) shows that the values calculated by Gofman et al. (20) (shown in Table III-1) are overestimates of the risk by a factor of approx- imately 4. (e) Doubling doses for induction of various different neoplasms in the Japanese survivors of the atomic bombs are not equal to each other as can be seen from Table III-2 below: Table II1-2: A comparison of relative and absolute risk estimates of the major types of malignancies induced among the A-bomb survivors (all ages). Type of Neoplasm Period of Observations Spontaneous Incidencei Absolute Risk Est.2 Doubling Dose2 Lung Cancer Breast Cancer 1955-70 1965-70 17.8 6.7 0.85 2.43 215 28 185

Table 111-2: A comparison of relative and absolute risk estimates of the major types of malignancies induced among the A-bomb survivors (all ages)—continued Type of Neoplasm Period of Observations Spontaneous Incidence1 Absolute Risk Est.2 Doubling Dose2 All GI less Stomach Cancer Leukemia 1960-70 1950-70 61.8 4.5 0.76 1.71 813 26 Calculated from the rate in the 0-9 rads group and expressed in deaths/105/year. 2 Estimates based on a comparison of the 0-9 rad with the 10 + rad group. The absolute risk estimate is in deaths/106/year/rad. No RBE correction for dose attenuation or for the RBE of neutrons has been applied. 3 This is the period over which the rates of the respective cancers were found to be increased. It is apparent from this comparison that the thesis that doubling doses for different types of cancer are roughly equal does not apply to the A-bomb survivors, at least as far as the data are at present available. If any general rule can be drawn from this small sample, it would be that the absolute risk estimates are roughly equal despite the very different spontaneous incidences of the different types of cancer. This conclusion, together with the lack of any suggestion of an excess number of cancers of the stomach, cervix and uterus, throws grave doubt on the validity of generalizations I and II when they are analyzed in terms of the data collected up to 1970 on the A-bomb survivors. (f) Data on death of U. S. radiologists (25) from leukemia and all other cancer do not sup- port the thesis that the relative risk of leuke- mia induction is the same as that for all other cancer. For example, the relative risk of leuke- mia death in the 50-79 age group of RSNA members compared with AAOO members is 2.9 (17 observed compared with 5.8 expected at AAOO rates), whereas for all other cancer deaths in the same age group the relative risk is 1.6 (126 observed compared with 79.6 expect- ed). Unfortunately, these data can be analyzed only on a semi-quantitative basis due to a lack of detailed information on dosimetry and on the kinetics of development of the excess deaths in each case and also due to the possible unsuita- bility of the controls, but on this basis alone the data imply that the relative risk of leuke- mia induction is roughly 3 times greater than that for all cancer induction. (g) Despite the above criticism, it is felt that it may be better in some cases to express risk in relative rather than absolute terms. This is due to the fact that a constant doubling dose for a particular neoplasm may be the best way to deal with changes in the spontaneous rate of that malignancy during adult life. This has been shown to be true for leukemia induction in spondylitics (26) and it also appears to be true for lung cancer in uranium miners in whom the effect of mining (radiation) is seen as a multi- plication of the spontaneous incidence by a factor which is the same whether the miners were smokers or not (27). Whether or not this holds for the induction of solid tumors in the A- bomb survivors is a moot point as can be seen from Figure III-2. A constant absolute risk would yield a horizontal line, whereas a con- stant relative risk would yield a line parallel to the age specific cancer death rate curve. As can be seen from this figure neither prediction is strictly followed. (h) More critical than the actual values of the risk estimates are the assumptions that have to be made on the lengths of the latency and plateau regions before any overall evaluation of risk can be made. Unfortunately, there are very few data on which to base realistic esti- mates of the lengths of these two periods. Anal- ysis of the evidence that does exist suggests that the most likely estimates are as follows: (i) Irradiation in utero: a latency of zero years and a plateau region of 10 years for all malignancies including leukemia. 186

O Age specific absolute risk A Age specific cancer death rate Age at birth O-9 1O-19 Age at Death 2O-34 3O-44 2O-34 4O-59 35-49 55-74 Figure 111-2. The absolute risk of induction of all malignancies (except leukemia) in the A-bomb survivors versus age ATB. The data are for both cities and sexes combined, for deaths from 1960 to 1970 and no neutron RBE correction has been ap- plied. The age specific death rate has been taken from the death rate in the 0-9 rad control group. 187

(ii) Irradiation at all subsequent times: a latency period of 2-3 years followed by a plateau length of 20-30 years for leuke- mia development, and for all other malig- nancies a latency period of 10 to 15 years (depending on the type of cancer) fol- lowed by a plateau region of between 20 years and the remaining lifetime of the individual. The crucial difference in terms of predicting the annual excess cancer deaths between these estimates and those used by Gofman et al. (20) is the length of the plateau region following exposure in utero. For example, extension of the plateau region after irradiation in utero from 10 years to lifetime adds an additional 11,500 cancer deaths to the relative risk model of evaluating the effect of a continual exposure of 0.1 rem/year to the U. S. population. The reasons for the choice of a 10-year plateau re- gion following in utero exposure have been dis- cussed previously (IIA-2). The conclusion, therefore, is that the figures generated by Gofman et al. (20) are overesti- mates: The reasons for their overestimates are: (i) An overestimation of the relative risk of solid tumor induction following irradia- tion of 0-9 year olds by a factor of 4-5, and by a factor of 10 for all other ages, (ii) The unreasonable assumption of a life- long plateau region following in utero irradiation. No conclusion can be made at this time on the absolute versus relative risk -dilemma. Should the absolute risk model be closer to reality, no numbers greater than approximately 3,000 extra cancer deaths from exposure of the U. S. population to 0.1 rem/year could be obtained given the present risk estimates, regardless of the lengths of any of the plateau regions for cancers other than leukemia. REFERENCES FOR SECTION III (1) United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) Report. Supplement #14 (A/5814), New York, United Nations, 1964. (2) United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) Report. Supplement #16 (A/5216), New York, United Nations, 1962. (3) International Commission on Radiological Protection. The Evaluation of Risks from Radiation. Health Physics 12:239-302,1966. (4) International Commission on Radiological Protection. Radiosensitivity and Spatial Distribution of Dose. ICRP Publication 14. Oxford, Pergamon Press, 1969. (5) Court Brown, W. M. and Doll, R. Mortality from cancer and other causes after radiotherapy for ankylosing spondylitis. Brit. Med. J. 2:1327-1332,1965. (6) Lijinski, W. and Greenblatt, M. Carcinogen dimethyl- nitrosamine produced in vivo from nitrite and aminopyr- ine. Nature, April 12,1972, 236:177-178. (7) Dolphin, G. W. and W. G. Marley: Risk evaluation in relation to the protection of the public in the event of accidents at nuclear installations. Environmental Con- tamination by Radioactive Materials, International Atomic Energy Agency. Vienna, 1969, pp. 241-253. (8) Mole, R. H. Radiation Effects in Man: Current views and prospects. Health Physics 20: 485-490,1971. (9) Baum, J. W. Compound exponential theory of radiation induced carcinogenesis. Presented before the 16th An- nual Meeting of the Health Physics Society, New York, July 12-15,1971. (1O) Miller, R. W. Delayed radiation effects in atomic- bomb survivors. Science 166:569-574,1969. (11) Maki, H., Ishimaru, T., Kato, H., and Wakabayashi, T. Carcinogenesis in atomic bomb survivors. Atomic Bomb Casualty Commission Report TR24-68,1968. (12) Wanebo, C. K., Johnson, K. G., Sato, K., and Thor- slund, T. W. Breast cancer after exposure to the atomic bombings of Hiroshima and Nagasaki. New Eng. J. Med. 279:667-671,1968. (13) Harada, T., and Ishida, M. Neoplasms among A-bomb survivors in Hiroshima: First report of the Research Committee on Tumor Statistics, Hiroshima City Medical Association. Hiroshima, Japan. J. Natl. Cancer Inst. 25:1253-1264,1960. (14) Auxier, J. A., Cheka, J. S., Haywood, F. F., Jones, T. I'., and Thorngate, J. H. Free-field radiation-dose distri- butions from the Hiroshima and Nagasaki bombings. Health Physics 12:425-429.1966. (15) Tomonaga, M. Statistical investigation of leukemia in Japan, Atomic Bomb Casualty Commission Report TR28-66,1966. (16) Ishimaru, T., Hoshino, T., Ichimaru, M., Okada, H., Tomiyasu, T., Tsuchimoto, T., and Yamamoto, T. Leuke- mia in atomic bomb survivors, Hiroshima and Nagasaki, 1 October 1950 - 30 September 1966. Radiation Res. 45: 216-233,1971. (17) Gofman, J. W., and Tamplin, A. R.: Low Dose Radia- tion and Cancer Institute for Electrical and Electronic Engineers, Transactions on Nuclear Science, Part I, NS 17:1-9, 1970. (18) Gofman, J. W., and Tamplin, A. R.: Federal Radiation Council Guidelines for Radiation Exposure to the Popu- lation-at-large-Protection or Disaster? Underground Uses of Nuclear Energy, Part I (Hearings before the Subcommittee of Air and Water Pollution, U. S. Senate, 91st Congress, Nov. 18,1969) (19) Tamplin, A. R., and Gofman, J. W.: Population Con- trol Through Nuclear Pollution. Chicago, Illinois, Nel- son-Hall Company, 1970. (20) Gofman, J. W., Gofman, J. D., Tamplin, A. R., and Kovich, E.: Radiation as an environmental Hazard. Pap- er presented at the 1971 Symposium on Fundamental Cancer Research, M. D. Anderson Hospital and Tumor Institute, Houston, Texas, 1971. (21) Stewart, A., and Kneale, G. W.: Changes in the Can- cer Risk Associated with Obstetric Radiography. Lan- cet, January 20,1968,104-107. (22) MacMahon, B.: Prenatal X-ray Exposure and Child- hood Cancer. Journal of the National Cancer Institute 28:1173-1191,1962. 188

(23) Court Brown, W. M-, and Doll, R.: Mortality fron. Cancer and Other Causes after Radiotherapy for Anky- losing Spondylitis. Brit. Med. J., Dec. 4, 1965, pp. 1327- 1332. (24) Dolphin, G.W., and Eve, I. S.: Some Aspects of the Radiological Protection and Dosimetry of the Gastroin- testinal Tract. In: Gastrointestinal Radiation Injury, edited by M. F. Sullivan, Excerpt Medica Foundation, Amsterdam, 1968, pp. 465-474. (25) Seltser, R., and Sartwell, P. E.: The influence of Occu- pational Exposure to Radiation on the Mortality of American Radiologists and other Medical Specialists. Am. J. Epidemiology 81:2-22,1965. (26) Doll, R.: Interpretations of Epidemiologic Data. Can- cer Research 23:10,1613-1623,1963. (27) Lundin, F. E., Jr., Lloyd, J. W., Smith, E. M., Archer, V. E., and Holaday, D. A.: Mortality of Uranium Miners in Relation to Radiation Exposure, Hard-rock Mining and Cigarette Smoking—1950 through September 1967. Health Physics 16:571-578. Appendix IV. Calculation of Confidence Limits for Risk Estimates There are two classes of risk estimates: 1) Those based on comparison of the experi- ence of a radiated group with expectation derived from national vital statistics or from other sources which are, in contrast with the data for the irradiated group, essentially constants, free from sampling variability. 2) Those based on the comparison of the ex- perience of a radiated group with that of an unirradiated group. In this case the data for the comparison group must also be considered as subject to sampling fluc- tuation. Class 1 For the irradiated group, let x = number of deaths P = number of person-years D = average radiation dose and let R be a population death rate appropri- ate for comparison, that is, standardized to the irradiated group for age, sex, and epoch. Then: Relative risk = (if >0) Doubling dose = D / Absolute risk = (x- P.R)/(D«P)xl06 (if >0) In each of the above expressions, x is the only stochastic variable of interest. If upper and lower confidence limits for the true value of x can be found and substituted into these expres- sions, then the results will be confidence limits on the expressions. Confidence limits for x are derived by consi- dering x to be a realization of a Poisson varia- ble, for which confidence limits on the mean can easily be obtained. Class 2 In the unirradiated comparison group let: y = number of deaths Q = number of person-years while notation for the irradiated group re- mains as before. Then: RR = Relative Risk = — / -%- P I Q Q DD = Doubling Dose = AR = Absolute Risk= D * P x 106 (if > 0) In the expressions for relative risk and for doubling dose, the stochastic elements enter only as the ratio x/y; while in the expression for V V absolute risk the stochastic term is p 7?, which can be seen writing the absolute risk expression as: We suppose x and y are independently dis- tributed as Ppisson_variables with unknown mean values X and Y. Then the joint distribu- tion of x and y is: exp(-X - Y) Xx • Yy x! * y! 189

We impose the condition that x + y be constant and, summing the corresponding probabilities, obtain: Appendix V. Radiation Dosimetry of Heavily- Irradiated Sites in Patients Treated for Anky- losing Spondylitis x + y exp(-X - Y) (X + Y) (x + y)! as the total probability of x + y. Whence, is the conditional probability of x and y, on the condition that x + y is constant. This is a bi- nomial distribution with parameters (x + y) andX/(X + Y). Let: p = X X / + YJ, so that / (1 - P)- We obtain binomial confidence limits on p and this leads immediately to corresponding limits on the ratio X A, which in turn leads to limits on the relative risk and doubling dose. As to the absolute risk, the true value is: X 106 and we can rewrite this as: AR = / RR - i \ y RR . P + Q/ (x + Y) D 106 Confidence limits for the left-hand factor can be obtained easily from the previously ob- tained limits on RR and D is, of course, a known constant. The factor X -I- Y, the expected num- ber of cases in the combined control and test group is, however, unknown. We obtain approx- imate limits by substituting the observed total number of cases, x + y, for the unknown expect- ed total. A. Introduction The problem of radiation dose to the stomach and to the bronchus in patients treated with radiation for ankylosing spondylitis (1) centers on three main considerations: The position of the critical organ or tissue during treatment, the relationship of these structures to the irra- diated fields and to depth dose data, and the frequency of radiation treatments received by the thoracic and lumbar spine. These values are important in the risk estimate for radia- tion-induced carcinoma of the stomach and carcinoma of the bronchus, and subsequently in the estimate of risk for cancers in heavily irradiated sites in these patients treated for ankylosing spondylitis. B. Stomach The stomach is obliquely placed in the left upper abdominal cavity; when the patient is prone, the fundus is posterior to the liver, and often adjacent laterally and sometimes poster- iorly to the lower thoracic and upper lumbar vertebral bodies. The stomach may be of an asthenic, hyposthenic, sthenic or hypersthenic type, depending on the person,s habitus (Figure V-1). The body of the stomach overlies the ver- tebral bodies to an appreciable degree only in the hypersthenic type. In the sthenic type, the pylorus overlies the vertebral bodies of the lumbar spine, and not the lower vertebral bod- ies of the thoracic spine. In hyposthenic and asthenic males, the pylorus may be deep in the pelvis; here, the stomach will frequently cross the midline and, thus, the vertebral bodies of the lumbar-5-sacral-1 level. Here, the lesser curvature of the body and fundus of the stom- ach runs in the paravertebral gutter, laterally adjacent to or posterior to the lumbar verte- bral bodies (Figure V-1). When thin patients lie in the prone position, the stomach and the duodenum press against the vertebral bodies; the gastrointestinal structures spread transversely, and surround the vertebral bodies laterally and anteriorly. 190

D OJ t=- .2 ODD 1 7 •8 | I 191

On radiographic and fluoroscopic examination in which a barium meal is used, a pressure fill- ing defect is observed. The variation in the anteroposterior thickness of the abdomen be- tween the erect and prone positions may be great; the change in diameter is usually 3 to 5 cm in patients with average build, less in thin patients, and considerably more in obese pa- tients. The range for the anteroposterior diameter for most patients lying in the prone position is between 17 and 26 cm. In cadavers, fixed in the supine position, transverse sections at the level of the thoracic-12 vertebral body demon- strate the stomach to be lateral to and slightly anterior to the vertebral body (Figure V-2). If the mean thickness is approximately 21 cm, the midpoint of the relevant vertebral body from the surface of the back is 7.5 cm. The stomach will lie between 4.5 and 12 cm from the surface of the back, i.e., part of the f undus of the stom- ach will lie more posterior than the vertebral body. Most frequently, the stomach lies below the level of the thoracic-11 vertebral body; it is therefore adjacent to the lumbar spine, not the thoracic (dorsal) spine. The lumbar spine is convex towards the anterior aspect of the trunk. In most patients, particularly in thin individuals in the prone position, the distance between the vertebral bodies and the anterior abdominal wall is less than, or at most equal to, the distance between the vertebral bodies and the posterior skin surface of the trunk. In such cases, the midpoint of the vertebral bodies may be 10 cm from the dorsal skin surface, and the stomach may lie from 6 cm to 16 cm from the dorsal surface of the trunk. The thoracic spine on the other hand is mildly convex towards the posterior aspect of the trunk. The rib cage pre- vents compression of the upper abdomen when the patient is prone, but the stomach may move upwards 4 to 8 cm, according to the patient,s habitus. Here, the stomach will spread trans- versely with a comparable decrease in its length, and more of the gastric structure will be closer to the vertebral bodies of the lower thoracic spine. The important conclusions are as follows: (a) The stomach is much more closely related to the lumbar spine than to the thoracic spine, (b) In thin patients in the prone position, the stomach is lateral and slightly anterior to the lumbar spine. Frequently, the fundus may be posterior to the midpoint of the lumbar vertebral bodies, (c) The radiation dose received by some por- tions of the stomach would be very little less than the closest vertebral body, (d) Depth- dose data derived from water and tissue-equiv- alent phantoms would not necessarily apply directly to each patient, and appropriate modi- fications would be necessary for the extent of anatomic variation in each individual patient. In view of these anatomic observations, the calculations of Dolphin and Eve (2) must be revised; they underestimate the doses to heavi- ly irradiated tissues. The stomach mucosa re- ceived a much higher dose relative to the adja- cent vertebral bodies than estimated by Dol- phin and Eve (2). Not all patients treated were included in their estimate, many more received exposure to the lumbar spine. Their estimate of mean dose to the vertebral bodies was based on the thoracic spine irradiation field, but should include the lumbar spine fields. Only one ra- diotherapeutic technique was used in their de- terminations, whereas a variety of techniques, filtration and field sizes was used in clinical practice. From data of Court Brown and Doll (1), the mean radiation dose to the thoracic spine of treated patients determined as mean exposure to the bone marrow throughout the entire skel- eton and as the maximum exposure at a point in the spinal marrow, was approximately 880 R. Based on depth-dose data for various orthovol- tage and certain higher energy machines (3) the range of fall-off of dose with increasing depth was about 50% at 10 cm to 15 cm from the pos- terior skin surface of the trunk. When the low- er thoracic spine was treated with field sizes greater than 8 to 10 cm in diameter the portion of the fundus of the stomach exposed would receive approximately the same dose as the lat- eral part of the adjacent vertebral body. In the orthovoltage range, this dose would fall from about 44% at 10 cm to 26% at 15 cm; in the su- pravoltage range, the decrease would be from 55% to 45%, respectively. Thus, the dose re- ceived by the mucosa of the exposed stomach when the patient was prone was approximately 40% to 50% of that of the vertebral body. More appropriate calculations, based on the radioth- erapy dosimetry, would indicate a dose of greater than 250 R, and possibly 500 R. In this 192

Liver Stomach Transverse Section of Cadaver at Thoracic -12 Figure V-2. Transverse section of cadaver at thoracic-12 vertebra «»-79T O - 72 - l4 193

7 1 •? 194

report, the mean dose to the stomach in this series has been taken as 250-500 rads. c. Lung The bifurcation of the trachea at the carina in the adult is at the level of the thoracic-6 ver- tebral body, the main stem bronchi at approxi- mately the thoracic-7 vertebral body, and the secondary bronchi at approximately the thora- cic-7-thoracic-8 vertebral bodies (Figure V-3) (4). The mean width for the right and left hili is approximately 5.5 cm (over-all 11.0 cm) and over 85% of average persons demonstrate little difference, less than 1 cm, between the width of the two hili. The carina and hili lie in the mid- mediastinum. In general, the posteroanterior thickness for most patients in the prone posi- tion is between 25 and 35 cm. If the mean is 30 cm, the midpoint of the thoracic-7 vertebral body from the skin surface of the back is ap- proximately 9 cm, and the carina and mainstem bronchi are approximately 15 cm. The thoracic spine is mildly convex towards the posterior aspect of the trunk. The rib cage prevents compression of the chest when the patient is in the prone position during radiotherapy. The important conclusions are as follows: (a) In general, radiotherapy fields to the thoracic spine of patients with ankylosing spondylitis were approximately 10 cm. wide and 10 to 30 cm long (1). Divergence of the orthovoltage beam would readily include the carina, the mainstem bronchi, and the secondary bronchi in all pa- tients, (b) In most patients, these structures are approximately 6 cm anterior to the verte- bral bodies of the thoracic (dorsal) spine. Therefore, the bronchial mucosa received a relatively high dose in relation to the vertebral bodies, (c) Probably all patients with ankylos- ing spondylit i who received thoracic spine ir- radiation during treatment also received bronchial irradiation. on dupth-dose for various orthovol- awl some higher energy machines (3) the range of fall-off of dose with increasing depth ix approximately 50% at 9 cm to 15 cm from the skin surface of the trunk. In the or- range, this decrease is approxi- mately 49% at 10 cm to 26% at 15 cm; in the »U(>ri»V'/lttt||fe range, It is from 58% to 35%. , U»* /Jos* rw-ei ved by the bronchial mucos- al epithelium, when the patient was prone dur- ing treatment, would be at least 40% to 50% of the dose to the closest thoracic vertebral body. If it is assumed that the mean dose to the thoracic vertebral bodies was 880 R, that all patients received full radiation exposure to the critical bronchial structures, and that the dose to the bronchus ranged from 40% to 50% of the dose to the thoracic spine vertebral bodies, then the dose to the bronchus in these patients would be approximately 400 R. Thus, the calcu- lations of Dolphin and Marley (5) are underesti- mates of the dose to the lung, possibly by a fac- tor of 5, and should be revised to take into ac- count the proper therapeutic radiation dosime- try. The figure of 400 rads has been used in this report to calculate the risk estimate for lung cancer from this series. REFERENCES (1) Court Brown, W. M., and R. Doll: Leukemia and aplas- tic anemia in patients irradiated for ankylosing spondy- litis. Medical Research Council, Special Report Series No. 295, H. M. S. O., London, 1957. (2) Dolphin, G. W., and I. S. Eve: Some aspects of the ra- diobiological protection and dosimetry of the gastroin- testinal tract. (In) Gastrointestinal Radiation Injury. Sullivan, M. P., et al., pp. 465-474, Excerpta Medical Foundation, Amsterdam, 1971. i;.-) Depth Dose Tables for Use in Radiotherapy. British Journal of Radiology, Supplement 10, London, 1961. (4) Lusted, L. B., and T. E. Keats: Atlas of Roentgenogra- phic Measurement. Year Book, Chicago, 1959. (5) Dolphin, G. W., and Marley: AHSB (RP) R95, Harwell, England, April 1969. Appendix VI: Definitions and Notes to Accom- pany Reference Tables Summarizing Quantita- tive Data on Carcinogenic Effects of Ionizing Radiation The following notes and definitions are keyed to the numbered columns of the standard table format: 1. Study population - Each experience used here to estimate the risk of cancer attributable to exposure to ionizing radiation is briefly characterized as to reason for radiation or status at irra- diation, geographic source if impor- tant, and calendar period of irradiation. 195

Hiroshima (H) and Nagasaki (N) A-- bomb survivors are sometimes shown separately. 2. Reference - The bibliographic refer- ences from which the information on risk is drawn, keyed to the bibliography of the specific section. When additional information, or an assumption, is used in making the estimates, the additional source or the rationale is given in a footnote to the table. 3. Type of Radiation - Type of radiation is coded as follows: a s y n x Alpha Beta Gamma Neutron X ray 4. Duration of radiation exposure - The total length of time over which radia- tion was received, not necessarily con- tinuously. This will range from less than 10 seconds (<10") for A-bomb sur- vivors to years for some forms of inter- mittent radiotherapy. Duration may be shown merely as minutes (min), or as minutes to day, etc., or with the abbre- viations wk (week), mo (month), and yr (year) being used. 5.-6. Reported duration of follow-up (years) - This is the length of time patients were followed up, with the range being the interval from the shortest to the long- est period of follow-up when subjects are arrayed as to their individual years of follow-up. Time is usually counted from radiation exposure. In the 1965 report on ankylosing spondylitis pa- tients it is indicated that the follow-up period ranges, essentially, from 5 to 27 years after radiation. The arithmetic mean of the frequency distribution of follow-up intervals is shown as the mean, and is the same as the range for A-bomb survivors. 7. Period after irradiation on which risk estimates are based (years) - Here time (t) is always measured from initial (?) radiation exposure as t0, and for A- bomb survivors whose follow-up began 1 October 1950 and ended 31 December 1970 the interval is given as 6th-25th year and is about 20 years in length. The interval for risk calculation need not be the reported interval of observa- tion, however, as when a period of la- tency is excluded. For example, in Hem- pelmann,s follow-up on thymus irradia- tion (high-risk), duration of follow-up ranges from 17 to 32 years, but the risk of thyroid cancer is calculated for the interval 6-32 because the minimum la- tent period was estimated at five years in that study. Similarly, in tables on the risk of all cancer other than leukemia among A-bomb survivors, a latent peri- od of 15 years is used, and data taken for 1960-1970 only. 8. Number of subjects - This is the num- ber on which risk estimates are actual- ly based. If only subjects aged 0-9 at exposure are used in the risk estimate, for example, only the number aged 0-9 at exposure is shown. In prospective, or cohort, studies the irradiated subjects (N) are followed to define the subset of subsequent deaths from cancer, and it is N that is tabled here. In retrospective studies, where cancer cases and controls are studied to define the subsets with, and without, prior radiation, numbers of cases are not shown in the tables, but equivalent details are given in a special footnote to the table. 9. Number of person-years - This is the number of person-years at risk. This number is the denominator for absolute risk estimates. That is, the attributable incidence, I, may be found as I = _ Cases attributed to radiation Person-years at risk which in these tables is expressed as cases/106/year/rem. For retrospective studies the concept of person-years at risk does not apply, and no entry is made in this column. 10.-11. Dose (rads) - In most publications dose is given in rads. However, radiations of different quality or linear energy transfer (LET) may be of different rela- tive biological effectiveness (RBE), and in this report risk estimates are given in rems insofar as possible, with the 196

dose in rads multiplied by the RBE fac- tor. The latter is obtained for a speci- fied form of radiation as the ratio of the dose (in rads) of high energy x or gam- ma radiation required to produce a giv- en level of biological effect, to the dose (in rads) of the specified form of radia- tion required to produce the same ef- fect. Thus the RBE of x or gamma ra- diation is taken as 1. The rem dose is not shown explicitly in the tables, but is either equivalent to the dose in rads or may be found as the dose in rads x an RBE (column 17) if the dose in rads re- fers to a single type of radiation. Such is not the case for Hiroshima and Naga- saki where the total dose consisted of gamma and neutron radiation, the components of which are in these tables weighted 1:1 for RBE (neutrons) of 1, and 1:5 for RBE (neutrons) of 5. An approximate range (column 10) and the arithmetic mean (column 11) are given separately, in recognition of the frequently important variation in individual dose. The range is the exter- nal dose. The mean dose shown in column 11 is usually the relevant tissue dose but for the A-bomb survivors it is the whole- body, free field, or "air," dose. Attenua- tion factors for atomic bomb survivors have not been published, and hence no effort is made here to provide tissue doses, which might be 60 to 75 percent of the dose given in rads, depending on the tissue and the proportions of neu- tron and gamma radiation. For the ankylosing spondylitis patients the mean dose to the marrow of the spine, given as 880 rads, is here converted to a mean (whole body) marrow dose of 372 rads on the assumption that 42.3 per- cent of the entire marrow was irradiat- ed. In the table on lung cancer, the re- estimated average tissue dose to the lung of patients treated for ankylosing spondylitis is 400 rads, an"d for the stomach it is 250 to 500 rads (see App. 2.5). 12 -13. Age at irradiation (years) - Both the range and the arithmetic mean are shown. A major distinction is made in these tables between A-bomb survivors under 10 at the time of the bomb and those aged 10 or more because of the greater relative sensitivity of the younger victims. Fetal ages at expo- sure are shown as 0, with no attempt to distinguish among the trimesters of gestation. In the pelvimetry series 87- 89 percent were irradiated in the third trimester, in the A-bomb survivors, 27 percent. 14. Sex - Sex is coded as M (male) or F (fe- male). 15. Nature of control - Since the risk that is sought in these tables is relative to a norm, or is the excess above that norm attributable to radiation, a normal basis of expectation is required. This is an intrinsic control (low-dose, 0-9 rads) group among the A-bomb survivors, or a calculation based on the vital statis- tics of the country or other jurisdiction where the experience took place. Most of the A-bomb survivors in the 0-9 rads range had 0 rads; their mean is estimat- ed at 1.4 rads. When national mortality or incidence statistics are used as the basis of expectation, the name of the country is shown, in abbreviated form. Other geographic abbreviations are N. Y. (New York State), Ariz. (Arizona), Colo. (Colorado), N. Mex. (New Mexico). Other intrinsic controls are "sibs," sib- lings of treated cases, and "untreated" patients with the same disease but not treated by radiation. In retrospective studies the controls are defined differently from the above, which applies to prospective studies, and the information is shown, not in column 15, but in a special footnote to the table. 16. Relative risk (O/E) - For the period de- fined in column 7, the relative risk (RR) is expressed as observed (0) cases/ex- pected (E) cases, where the expected cases represent the normal expectation estimated from the group defined in column 15. Thus, if U. K. mortality sta- tistics were used to calculate expected deaths in the irradiated group this was often done by multiplying the person- years at risk for each sex and age 197

group, and calendar period by the cor- responding U. K. mortality rates for the cancer of interest and summing the products. In this calculation the RR is equivalent to the standardized mortali- ty ratio (SMR). In the case of the A- bomb survivors a calculation of this sort was done for each of several dose- groups, including the 0-9 rads group, and the corresponding SMR,s were manipulated to produce the tabled val- ue of E. Thus, for example, the SMR,s for leukemia in subjects 0-9 ATB are as follows: 0-9 Rads 10+Rads Observed deaths Japanese expectation SMR 7 4.572 1.531 19 1.909 9.953 Then E for 10+ rads = 1.909 x 1.531 = 2.93, and O/Efor 10+ rads = 19/2.93 = 6.50 In this case the ratio 0/E is also the ratio of the two SMR,s, i.e., 9.953/1.531 = 6.50. The foregoing methods apply to prospective, or cohort, studies, but not to retrospective case/control studies. There is one such in the table on leukemia following exposure in utero or before age 10. In this study the data may be arranged as follows: In utero No in utero radiation radiation Total Cases (leukemias) Controls Total a c a+c b d b + d a + b c+d a+b+c+d Then the relative risk is, approximately, ad/ bc.i 17. RBE - Relative biological effectiveness (RBE) is defined above under columns 10- 11, dose. In the sections on bone cancer and on lung cancer the RBE for alpha radiation is taken as 10 in calculating increase in risk per rem. In the tables on A-bomb survivors two calculations are routinely made, the RBE for neutrons being taken first as 1 and again as 5. 18. Percent Increase in Relative Risk per Rem - The percentage increase in rela- tive risk (RR) is, of course 100 x (RR-1), or 100 x (Column 16-1). Then the percentage increase per rem is found as 100 x (RR-1) Mean dose expressed in rem For x- and y -radiation, and for neutron radiation with RBE = 1, this will be 100 x (Column 16-1). Column 11 For a -radiation this will be 100 x (Column 16-1). 10 x Column 11 For the A-bomb experience the entry in Column 11 is a mixture of a and neutron radiation, and the details of each would have to be tabled separately in order to illustrate the calculation at RBE = 5. If a mean dose (column 11) for the two cit- ies were, for example, 69 rads made up of 10 rads of neutron radiation and 59 of gamma, then the calculation in column 18 would be 100 (Column 16-1). 59 + 5 x 10 The doubling dose is not used here (cf. pp. 99) but may be found from column 18 as iCornfield, J. and Haenszel, W.: Some aspects of re- trospective studies, J. Chron. Dis. 11:523-534 (May) 1960. doubling dose = 100 Column 18 198

19-20. Absolute Risk, Deaths or Cases/I0<V Year/Kern - This is the absolute risk attributable to radiation, i.e., the excess above normal expectation, and repre- sents the slope of a linear dose-response curve giving the control incidence (or mortality rate) at zero dose. Three val- ues are given: the best (or mean) esti- mate calculated from the data; a lower 90 percent limit on this estimate; and an upper 90 percent limit on it. These two limits define an 80 percent confidence interval. For prospective (cohort) stud- ies the best estimate is found, for x-, a-, and n- radiation of RBE = 1, as O-E x 106 Column 9 x Column 11 In retrospective studies, where person- years are not available, the best esti- mate is found by multiplying the pro- portionate increase in risk per rem by the appropriate estimate of natural in- cidence or mortality, i.e., xnaturalrate- 21. In Table a-7 (page 117), for example, the best estimate of 27 for the U. K. in utero study of Stewart et al. is found as 0.79 x the rate of 35 leukemia deaths per mil- lion/yr at ages 0-9 in U.K. The calculation of confidence limits is described in Appendix IV. For the re- trospective study (Table a-7, page 117) the method developed by Woolf1 and modified by Haldane2 was used. Footnotes or other comment - Here will be found references to footnotes a, b, etc. to the particular table and other comments relating to the source-materi- al or the calculation, e.g., adequacy of the dosimetry. Most series provide data on mortality; the exceptions are coded here as "morbidity." i Woolf, B.: On estimating the relation between blood group and disease. Ann. Hum. Genet. 19:251-253,1954. 2Haldane, J. B.: The estimation and significance of the logarithm of a ratio of frequencies. Ann. Hum. Genet. 20:309-311,1955. 199