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 XIV RECAPITULATION THE present chapter will be devoted to a brief recapitulation of the findings of this study, and a comparison of these results with those of other studies utilizing mammalian material. 14.1 The findings. â Information has been collected concerning the relationship between irradiation and the outcome of 76,626 preg- nancies terminating in Hiroshima and Nagasaki between 1948 and 1953. The following aspects of pregnancy outcome were considered: sex of infant, occurrence of major malformation, vi- ability at birth, birthweight, and occurrence of death during the first six postpartum days. In addition, on a random sample of 19,818 preg- nancy terminations followed up at age nine months, data were obtained on infant survival up to nine months postpartum, malformations which had become apparent since the previous examination, and growth of the child as re- flected in certain anthropometric measurements. The variation in these potential indicators of radiation-induced genetic change has been ana- lyzed in relation to the experience of the parents at the time of the atomic bombings. For analytical purposes, five categories of radiation, defined in Table 4.7, have been recognized. Each infant studied may thus fall into any one of 25 categories when classified according to the radiation history of both father and mother. In evaluating the significance of the findings, it is important to bear in mind that there are, in effect, two sets of control parents, namely, those parents (group 1) who were not in either city at the time of the bombings, and those parents (group 2) who, although in one or the other of the two cities, were at distances of 3,000 or more meters from the hypocenter or, if nearer, were significantly shielded, and presumably re- ceived negligible amounts of irradiation. In Chapters III and V an effort has been made to explore dissimilarities between Hiro- shima and Nagasaki parents which might be responsible for differences in indicator findings in the two cities. It is noteworthy that despite the several demonstrated or possible dissimilari- ties, the two cities did not differ significantly with respect to any of the indicators save birth- weight and the anthropometric measurements at age nine months. This may be taken as some measure of "relative stability," within this range of environmental variation, on the part of the indicators of potential genetic damage utilized in this study. This has some bearing on one's attitude towards the appropriateness of radia- tion category 1 parents as the source of control material. As mentioned in Section 5.10, there is a relatively high proportion of repatriates or persons of rural background among category 1 parents. It seems unlikely, however, that cate- gory 1 parents, as a group, differ biologically from the "exposed" group any more than the inhabitants of Hiroshima differ from those of Nagasaki. If this is correct, then theoretically the differences, if any, between the offspring of category 1 parents and of categories 2, 3, 4, and 5 parents may be used as a measure of com- bined radiation and "disaster" effect, with the radiation effect then best measured by progres- sive differences between the offspring of parents falling into categories 2, 3, 4, and 5. There are significant sex differences as regards the indicators birthweight, neonatal death rate, and anthropometric measurements at age nine months, but not with respect to major mal- formations or stillbirths. Such sex differences are well recognized in the literature. It is per- haps worth pointing out that the level of the significance with which these previously recog- nized differences emerge in this analysis in- creases one's confidence that valid differences of appreciable magnitude between the various analytical subclasses would be detected. The reader's attention is directed to Figures 14.1 and 14.2 which present graphically the 192
Recapitulation 193 50 ~-^ . f H PIRCCNT MALFORMCO !WANTS MOTMCn'j EXPOSURE 2 I MOTHER'S EXPOSURE PERCENT STILLBmTMi a â¢ MOTMER'l CXPOSUKC PCMCCNT NCONATAL DEATHS I 1 MOTHCK'S cxro$uÂ« ADJUSTED MCAN OF BIKTM. WCICHTS (OECAGBAMS) t i WOTMCH'S cxroumc RCSIOUAL BIBTMWttCMT MCAN SOUAUCS 2 1 MOTHER'S CXPOSUHC FIGURE 14.1 â A graphical representation of the effect of parental exposure on: (1) the sex ratio; (2) the frequencies of malformed infants, stillborn infants, and infants dying in the neonatal period; (3) birthweight means; and (4) birthweight variances.
194 Chapter XIV Genetic Effects of Atomic Bombs distributions of the indicators by parental ex- posure, and to Table 14.1 for a summarization of the tests of significance with reference to exposure. With regard to the effect of parental exposure on the indicators, we find the fol- lowing: (1) Sex ratio. âThe one indicator wherein differences of opinion with regard to the effect of parental exposure may arise is the sex ratio. PERCENT MALFORMATION AT ACE NINE MONTHS terminations where the father was in class 1, provided the mother was not also in this class, there emerges a significant effect, on a one- tailed test, of mother's exposure, but not of father's exposure. Two questions immediately arise, namely, (a) is this a legitimate compari- son, and (b) if so, does there exist ancillary evidence which supports the reality of this difference? PERCENT DCATHS BY AGE MULTIVARIATE MEAN LOG Of GENERALIZED VARIANCE f â¢ 2 MOTHER'S 3 EXPOSURE 4-5 FIGURE 14.2 â A graphical representation of the effect of parental exposure on: (1) the frequency of malformed infants alive at age nine months; (2) the frequency of death in the first nine months of life; and (3) the anthropometric measurements of weight, height, head circumference and chest circumference. This stems from the fact that significance or non-significance of a maternal exposure effect is, in part, a function of that portion of the data which one elects to analyze. If all exposure classes are used, there is no demonstrable effect of mother's exposure or father's exposure on a two-tailed, or for that matter, a one-tailed test of significance. Similarly, if all terminations occurring to parents one or both of whom are in exposure class 1 are rejected, there is no effect of mother's exposure or father's exposure. If, however, one rejects those terminations where the mother was in class 1 but retains the It is difficult to appraise the legitimacy of this comparison; however, two observations seem pertinent to any appraisal. Firstly, there is an element of arbitrariness in a procedure which rejects as unsuitable mothers in one exposure class but not fathers on such vague grounds as dissimilarity in place of origin, in the absence of demonstrable differences with respect to concomitant variables which are known or can be shown to appreciably affect the indicator. Secondly, rejection of mothers in class 1 leads to the use of a "control" approxi- mately 25 per cent of which is contributed by
Recapitulation 195 a single exposure cell (mothers 2, fathers 1, Hiroshima) exhibiting a sex ratio which is higher than any observed in these cities (or all of Japan) for any one of the fourteen years in the interval 1935-1952 for which data exist. Figure 14.1 fails to reveal consistent changes in the "exposure surface." Data continue to be collected which bear on the relationship of maternal exposure to the sex ratio.1 TABLE 14.1 A SUMMARIZATION OF THE COMPARISONS OF THE VARIOUS INDICATORS WITH PARENTAL EXPOSURE WHEN (a) ALL EXPOSURE CELLS ARE CONSIDERED (THE 4X4 CASE), AND (b) ONLY THOSE CELLS WHERE BOTH PARENTS WERE EXPOSED ARE CONSIDERED (THE 3X3 CASE) (The general direction of change is indicated by an arrow pointing upwards if the observed frequency of departure from "normality" increases with increasing parental exposure, and by an arrow pointing downwards if the frequency of the event decreases with increasing exposure. The tabular entries are probabilities.) Parental exposure Indicator Sex ratio 30-.50 f Malformation At birth 70-.80 f At 9 months Stillbirth 20-.30 f "Neonatal" death Death in 9 months Birthweight means Males-Hiroshima 10-.25 J, Females-Hiroshima > .25 i Males-Nagasaki > .25 f Females-Nagasaki > .25 f Anthropometries generalized means.. < .001 â¢No general test (see Chapter XI). Fathers Mothers 4X4 3X3 4X4 3X3 case case case case .30-.50T .90-.95 t .10-.20i .95-.9S I .70-.80 t .80-.90 1 .30-.50J .50-.70 t .80-.90 f .02-.05 i .20-.30 t .80-.90 t .001-.01 t .20-.30 J, a .20-.30 | a .02-.05 t .95-.9S f .50-.70 f .25-.50 >.25i .05-.10 j, .10-.25 I .10-.25 f .02-.05 Combined parental exposure 4X4 case Indicator Birthweight variances Males-Hiroshima 10-.25 Females-Hiroshima < .001 Males-Nagasaki 10-.25 Females-Nagasaki 10-. 2 5 Anthropometrics generalized variances Males-Hiroshima 10-.25 Females-Hiroshima 10-.25 Males-Nagasaki 10-.25 Females-Nagasaki 02-.05 3X3 case >.25 >.25 >.25 .05-.10 .05-.10 With regard to whether there exists ancillary evidence suggesting an effect of mother's ex- posure, it may be stated that neither the limited observations available on early pregnancy terminations, nor the stillbirth data, nor the neonatal death data, indicate an interaction of sex with mother's exposure such as one might expect if the death of male infants is a function of mother's exposure. Moreover, inspection of (2) Malformation. â There is no significant effect of mother's or father's exposure on the frequency of congenitally malformed infants irrespective of whether exposure class 1 parents are or are not included in the analysis. Further classification and analysis of these data to take 1 The reader may wish at this point to refer to the discussion of the findings in this supplementary material, at the end of Chap. VII.
196 Chapter XIV Genetic Effects of Atomic Bombs into account known differences between ex- posure classes in the age of the mother at the birth of the infant fails to reveal consistent, significant differences in the frequency of con- genital malformation attributable to parental exposure. There is evidence, however, of an effect of mother's exposure on the frequency of malformation among the children of very young mothers (<21 years of age). This effect, how- ever, seems largely one of an inordinately low frequency of malformation among young, unex- posed mothers rather than an elevation of the frequency among exposed mothers since there is no demonstrable difference among mothers in exposure classes 2, 3, and 4â5. Finally, analy- sis of additional data on the frequency of congenital malformation obtained at nine months of age discloses no significant effect of parental exposure. (3) Stillbirths. â Analysis of the data un- adjusted for differences in parity and maternal age between exposure cells fails to reveal an effect of paternal exposure, regardless of whether category 1 fathers are or are not in- cluded in the analysis. However, there does exist an effect of maternal exposure (significant at the 5 per cent level) when category 1 mothers are included in the analysis. This effect does not persist when mothers in exposure category 1 are excluded. Further classification of the data by parity and subsequent analysis reveals no effect of mother's exposure when the category 1 mothers are included. It would seem that the apparent mother's exposure effect is explicable in terms of differences between exposure classes in birth ranks. The latter analysis suggests an effect of paternal exposure limited to first-born infants. An explanation of this finding is not immediately apparent. (4) Neonatal death. â Both with and with- out allowance for differences in parity among the mothers falling into the various radiation categories, there is no significant effect of mother's or father's exposure on the frequency of neonatal death. Moreover, the analysis of deaths during the first nine months of life fails to reveal a significant effect of parental exposure. (5) Birthweight. â When allowance is made for differences between exposure cells in ma- ternal age and parity, there exists no significant difference between classes of father's exposure or classes of mother's exposure in mean birth- weight. Furthermore, it was not possible to demonstrate a consistent effect of parental ex- posure on (a) the relationship between birth- weight and concomitant variables, notably ma- ternal age and parity, or (b) the residual birthweight variances, that is, on the hirthweight variances following removal of maternal age and parity effects. Figure 14.1 illustrates the lack of a consistent exposure effect on the birthweight means or birthweight variances. (6) Anthropometrics. â Neither the gen- eralized means nor the generalized variances can be shown to differ among classes of paternal or classes of maternal exposure. This remains true following removal of differences between ex- posure cells in the age of the infant at the time the measurements were obtained. 14.2 The question of evaluating the over-ail direction of the indicators. â As noted in Sec- tion 6.2, the plan of analysis adopted ensured non-overlapping indicators of genetic effects. By suitable transformations it is readily possible to combine the results of such independent tests, given a theory which permits specifying the direction of the differences between control and irradiated material. In this particular study, however, the findings are sufficiently small and inconsistent, that we have not felt an attempt at a combined treatment to be justifiable. There is in the analysis of these data another problem, closely related to the foregoing. There are segments of the analysis which suggest the possibility of a radiation effect (e.g., certain selected comparisons within the sex ratio data; frequency of malformations in relation to ma- ternal radiation history when only very young mothers are considered; frequency of stillbirths in relation to paternal radiation exposure his- tory, when only the first-born children of these fathers are considered). Such findings tend to stand out in one's mind as he thinks back over the analysis as a whole. It is important in this connection not to lose sight of the fact that even in the absence of any effects whatsoever, by definition 1 in 20 independent tests or subtests may be expected to exceed the 5 per cent level of significance. Thus, while these "hints" can- not be entirely disregarded, they must be viewed against the background provided by the totality of the significance tests to which the data were subjected. 14.3 The confidence limits determined by these observations.âThus far in our analysis we have been concerned with attempts to demonstrate a positive effect of exposure to the
Recapitulation 197 atomic bombs on the indicators selected for study. There is, however, another aspect of these data. They permit us to place upper limits on the effects which may have been induced but not demonstrated by these studies. In other words, we can place confidence limits on our observations. One may proceed to establish these limits by several different approaches. To avoid misunderstanding, it should be clearly pointed out at this juncture that we are not primarily concerned with placing limits on a series of estimates, since estimates of abnormalities in pregnancy terminations derived from these data would probably have little meaning to other populations, in view of the racial differences and the variety of "experimental" conditions which could arise in man. Our primary concern here is to determine the adequacy of our test procedures. The ideal solution, perhaps, would be the computation of the power curves asso- ciated with each of the individual tests of sig- nificance. In multi-way analyses, however, the computation of the power function is, to say the least, tedious, and for a number of the tests performed here the power function is not known. An approximate solution to the problem of the adequacy of our data, sufficient for our purposes, can, however, be obtained. We shall begin by restricting our attention to the ade- quacy of the data for detecting differences of a specified size between a control population and a single, moderate-to-heavily irradiated popula- tion. Thus we shall seek to determine the prob- ability of rejecting the null hypothesis (no difference between groups), given a fixed sam- ple size, when the true situation is one wherein the exposed group differs from the control group by some amount, say d, for varying values of d. The groups with which we shall be con- cerned will be those terminations occurring to parents (a) neither of whom had exposures in excess of class 2 â the control group, and (b) one or both of whom had exposures equal to or greater than class 3 â the moderate-to-heavily irradiated group. We shall make the following simplifying assumptions: 1. The proportion of "successes" in the control population is known without error. 2. The normal approximation to the binomial distribution has the requisite accuracy over the critical range, which, for this study, includes those values for the exposed group which de- viate from the control by a factor less than two. The latter assumption requires no special comment, and seems warranted in this situation. The former assumption is obviously not strictly satisfied by these data; however, the sample number available on the control group is suffi- ciently large that the error associated with the estimate of the true proportion of successes for this group is small under all circumstances, and can for our purposes be disregarded when con- trasted with the error associated with an estimate based on the sample sizes available for the exposed groups. Now to determine the probability of rejecting the null hypothesis under a variety of alternative situations we proceed as follows: We shall assume that the level of confidence with regard to the null hypothesis is 0.05. There exist, then, two limiting values corresponding to this level of confidence such that the number of successes, say x, among N trials will lie within these limits 95 times in 100 if the null hypothesis is, in fact, true, that is, if there is no difference be- tween the proportion of successes in the control, say pc, and the proportion of successes in the heavily exposed group, say pe. These limiting values, Lt and L2, are chosen so that If now N and pc (or pe) are sufficiently large such that N/>c> 5, this probability can be reason- ably approximated by the normal distribution. Under the latter circumstance, we obtain LÂ± and L2 from the relations L1= -1.96VN/>C(1-/.C) + Npe L2= 1.96V Npc(l-p If pf^p,., but the number of successes, x, among N trials from a population characterized by pe should lie within the limits Lt and L2, we would erroneously accept the null hypothesis as true (the type II error) . The probability of the latter can be computed for any value of pe, and is merely '
198 Chapter XIV Genetic Effects of Atomic Bombs and, again, this probability can be reasonably approximated by the normal distribution. If this procedure were repeated for every possible value of pe, we would generate a continuous curve resembling either the normal cumulative prob- ability curve (if we were interested solely in one-tailed tests of significance) or the normal frequency curve (if we were interested in two- tailed tests of significance). These curves are termed the power function of the test or the tests of significance. For the remaining operat- ing characteristic curves the sample size was the number of parents both of whom experienced exposures equalling or exceeding class 3. Before we turn to a consideration of these five curves, it is worth noting that the amount of informa- tion actually available with regard to these four indicators is in excess of that used to compute the OC curves. Accordingly, we may view these curves as underestimating the adequacy of the OC FOR MALE BIRTHS - MOTHERS EXPOSED, FATHERS NOT OC FOR MALE BIRTHS - FATHERS EXPOSED, MOTHERS NOT .4996 .5078 .5158 5258 .5338 .5418 .5498 .5038 .5118 .5198 .5298 .5378 .5458 .5538 FREQUENCY FIGURE 14.3 â The adequacy of the data with regard to sex ratio as indicated by the operating char- acteristic (OC) curves for analyses based on sample sizes of 5,629 mothers and 2,453 fathers, where the true proportion of successes is assumed to be 0.5198, the value observed in the control population. operating characteristic of the test depending upon whether we are interested in the prob- ability of rejecting the null hypothesis when it is, in fact, false, or the probability of errone- ously accepting the null hypothesis when some alternative hypothesis is true. Figures 14.3 and 14.4 present five operating characteristic curves for sex ratio, malformation, stillbirth, and neonatal death. The two curves for sex ratio are based on sample sizes corre- sponding to the number of mothers in classes 3, 4, and 5 whose spouses were in classes 1 and 2, and the number of fathers in classes 3, 4, and 5 whose spouses were in classes 1 and 2. In both instances, we are concerned with one-tailed data to detect departures of specified size from the null hypothesis. Before turning to Figures 14.3 and 14.4, we must define what is meant by an adequate test. We shall say that a test is adequate with regard to the alternatives, peâpc^|s|, if the prior probability of detecting departures from the null hypothesis of size s, or greater, equals, or exceeds, some specified probability. Otherwise stated, a test is adequate with reference to all those alternatives for which the power (the probability of correctly rejecting the null hy- pothesis when it is, in fact, false) is not less than some specified amount. The reader may, of course, impose any power restriction he
Recapitulation 199 wishes, but we shall define adequate power, here, as being 0.90. That is to say, if the prior probability of detecting a difference of size j- is at least nine chances in ten we shall say that the test is adequate with regard to departures of size s or greater. This value may well be too high, since, as we have indicated, our procedure takes into account appreciably less than the total number of observations which are germaine. Within this frame of reference, we can say that our data are adequate to give reasonable as- n o heavily exposed, or 4 per cent if the father only was exposed; 2. Changes of less than 100 per cent (absolute) in the frequency of malformed infants; 3. Changes of less than 80 per cent (absolute) in the frequency of stillborn infants or infants dying during the neonatal period. 14.4 A resume of work on mammalian ma- terial pertinent to the interpretation of these findings. â Before proceeding to a final inter- . 1 00 â¢ OC FOR STILLBIRTHS AND NEONATAL DEATHS -- OC FOR MALFORMATION 2.5P 2P 0.67P FIGURE 14.4 â The adequacy of the data with regard to malformations, stillbirths, and neonatal deaths as indicated by the operating characteristic (OC) curves for analyses based on samples of size 1,097 parents, where the true proportion of successes are assumed to be 0.0090 for malformations and 0.0142 for stillbirths and neonatal deaths, the control values. surance that we would be able to detect the following: 1. A decrease in the sex ratio, following ma- ternal exposure, in excess of an absolute change of 1.6 per cent; 2. An increase in the sex ratio, following pa- ternal exposure, in excess of an absolute change of 4 per cent; 3. An alteration of the malformation rate in excess of two times the control value; 4. An alteration of the stillbirth rate (and neo- natal death rate) in excess of approximately 1.8 times the control value. Alternatively stated, we should be unable to detect: 1. Changes of less than 1.6 per cent (absolute) in the sex ratio when the mother only was pretation of these findings, we would do well to consider at some length the results of more or less comparable studies on other animal spe- cies. Although the many important studies which have been carried out on the genetic effect of irradiation on Drosophila clearly indi- cate the types of results to be expected in mam- mals, the differences between mammalian and insect physiology are such that any semiquanti- tative extrapolation is at best hazardous. Ac- cordingly, we will for the purposes of this discussion confine our attention to studies on mammalian material. The current status of our knowledge of mammalian radiation genetics has recently been comprehensively summarized by Russell (1954). There is actually a rather strik- ing dearth of information available for com-
200 Chapter XIV Genetic Ejects of Atomic Bombs parison with the findings of the present study. Such information as there is will be presented in the following six subsections. In most of the work to be cited, the house mouse has been the experimental animal. Some two to four weeks after male mice are irradiated with a dosage in excess of 400r, they become sterile, the time of onset of sterility depending on dosage. In those animals which survive, fer- tility usually returns in four to twelve weeks, again depending on dosage. There are impor- tant quantitative and qualitative differences be- tween the observable genetic effects associated with sperm released in the pre-sterile period â corresponding to the result of irradiation of spermatocytes, spermatids, and spermatozoa â and the effects associated with sperm released during the post-sterile period â corresponding to the irradiation of spermatogonia. We will consider primarily data collected during the post-sterile period, since this corresponds to the period during which observations have been made in Japan. 14.4.1 Effects on sex ratio. â It will be recalled that genetic theory suggests that the irradiation of males should result in an increase in the proportion of males in the next genera- tion. Parkes (1925), studying the offspring of male mice exposed to doses of X-ray below those which produce temporary sterility, re- ported (a) a moderate but non-significant in- crease in the proportion of males born to mice mated 0-4 days after irradiation (59.4 per cent of 133 offspring of irradiated males vs. 51.6 per cent of 735 controls), (b) a significant de- crease in the proportion of males born from matings made 5-18 days after irradiation, and (c) no change in the sex ratio among the off- spring of matings made 19-57 days post irradia- tion, this group apparently reflecting the results of the irradiation of spermatogonia. The ob- servations of Hertwig (1938) on the results of irradiation of spermatogonial stages in the house mouse are reproduced in Table 14.2. Although the difference is in the postulated direction, the results of a comparison of "all irradiated" vs. "controls" are not at the level of significance nor, for that matter, were the results obtained during the pre-sterile period significant. How- ever, when only the two groups receiving the most irradiation are considered, there is a sig- nificant difference. Russell (1954) reports that in his experiments with the exposure of male mice to 600r, there were 50.35 per cent males in 72,472 post-sterile period offspring, as against 51.00 per cent males in 55,828 controls. Finally, Kalmus, Metrakos, and Silverberg (1952) re- ported a significant decrease in the frequency of females among the offspring of male mice mated immediately following treatment with 150 or 300r, but were unable to confirm this observation in a later study (Trasler and Metrakos, 1953). Even if clear-cut effects had been demon- strated for the house mouse, the advisability of extrapolating to man would be rendered ques- tionable by the much higher apparent proportion of inherited defects which are sex-linked in TABLE 142 THE EFFECT OF IRRADIATION OF MALE MICE ON THE SEX RATIO OF OFFSPRING CONCEIVED DURING THE POST-STERILE PERIOD (After Hertwig, 1938) Dose in Number of Number of r 400 offspring 415 males 197 males 47.47 500 167 86 51.50 600 167 86 51.50 800 683 M7 50.80 1,000 . . . . 574 291 50 69 1 200-1 400 171 102 59.65 1 500-1 600 63 37 58.72 All doses 2 240 1,146 51.16 Controls 2.595 1.290 49.71 X* (all doses vs. controls)=: 1.011 DF=1 0.50 >P> 0.30 man than in the mouse, suggesting the possi- bility of a genetically more active differential segment of the X-chromosome in man. The corollary of this is, of course, a probable higher expectation of induced sex-linked lethals per unit irradiation in man. Recently Macht and Lawrence (1955) have published a detailed comparison of the sex ratio, the frequency of twinning, the frequency of fetal death, and the occurrence of congenital defect among the children of radiologists as contrasted with the children of physicians in other specialties (pathologists, psychiatrists, anesthesiologists, plastic surgeons, and ophthal- mologists). The information was obtained by questionnaires mailed to the subjects. Answers were obtained from 74.1 per cent of the radiolo- gists queried, with information on 5,461 chil- dren, but from only 53.8 per cent of controls, with data on 4,484 children. Because of the
Recapitulation 201 publicity given the matter of the genetic risks of radiation, one immediately wonders if radi- ologists as a group were especially motivated to reply, perhaps particularly if the reproductive history was in some way unusual. The authors are aware of this source of bias, and quite prop- erly point out that inasmuch as the distribution of single persons, married couples without chil- dren, and married couples with children is similar in the two groups, this suggests that radiologists with (abnormal) children were not especially motivated to reply. On the other hand, they present data suggesting that recall for abnormal terminations is better for later than earlier pregnancies. Despite this, they combine the (early) reproductive history of radiologists results obtained on the two appeals. But that this may be an important source of bias is indi- cated by the fact that whereas after correction for parity differences, there were 3.10 per cent more "abnormal offspring (fetal death and con- genital defects)" in the children of radiologists than non-radiologists on the basis of first ques- tionnaires, this difference dropped to 0.76 per cent on second questionnaires. This is a striking and disturbing discrepancy. Inasmuch as their study will be widely quoted, a detailed comparison between their findings and our own seems indicated. With regard to the indicator under consideration, sex ratio, the findings are as shown in Table 14.3. With respect to livebirths, the proportion of TABLE 143 THE SEX RATIO AMONG LIVEBIRTHS AND FETAL DEATHS IN THE STUDY OF MACHT AND LAWRENCE (1955) __ Livebirths Fetal deaths Offspring of , â¢Â« persons Males per Males Females Unknown 100 females Exposed throughout marriage 1,104 1,115 293 99.01 Exposed part of mar- riage 986 922 299 106.94 Unexposed part of mar- riage 526 461 152 114.10 Unexposed throughout 106.62 102.60 Males per Males Females Unknown 100 females marriage 1,240 All exposed 2,090 All unexposed 1,766 1,163 2,037 1,624 404 592 556 108.74 Total 3,856 3,661 1,148 105.33 53 55 22 42 108 64 172 23 19 14 23 42 37 316 300 102 345 616 447 79 1,063 230.43 289.47 157.14 182.61 257.14 172.97 217.72 prior to entering this field with the control data. A study of their Table 13 suggests that this may be an important source of bias. A second source of bias in this study stems from the fact that "persons who did not respond (to the questionnaire) within approximately two months were sent a second questionnaire together with a further appeal that it be re- turned." Of the total completed questionnaires available for analysis, 78.5 per cent of those completed by radiologists were obtained on first writing, as contrasted to 63.8 per cent of those from non-radiologists. The possible significance of this to the study lies in the fact that there was, in general, a higher percentage of abnormal outcomes reported on "second-appeal" question- naires than on "first-appeal" questionnaires. Specification of the direction of the bias intro- duced is difficult, since it depends on one's interpretation of the reasons for the different males is decreased, although not significantly so, a finding opposite in direction to expectation when fathers are irradiated unless â as they suggest â the male is more sensitive than the female to the effects of induced mutations in the autosomes. Among fetal deaths (their term; includes miscarriages and stillbirths), the data are in the direction of expectation but so very limited in number that even the large difference observed is not significant; it is worth noting that sex was not recorded for 81 per cent of this material. 14.4.2 Effects on malformation frequency. â Dominantly inherited mutant forms, some of which in our terminology would be classified as congenital malformations, have been detected in small numbers by Charles (1950) among the offspring of male mice exposed to 60r (7 in 3,072 offspring; 0 in 2,755 controls), and by Russell (1954) among the post-sterile period
202 Chapter XIV Genetic Ejects of Atomic Bombs offspring of male mice exposed to 600r (5 in approximately 30,000 offspring; number of controls not stated). The examination given the offspring by Russell was much less searching than that by Charles since Russell's work was primarily oriented in other directions. From the quantitative standpoint, these findings have debatable carry-over value for man if only be- cause of the tendency observed in most labora- tories for female mice to devour dead and defective embryos immediately following par- turition. The older medical literature contains a num- ber of studies, based on the questionnaire ap- proach, of the relationship between therapeutic irradiation and the "health" of children con- ceived subsequently (review in Murphy, 1928). For a variety of reasons, and particularly with reference to the problems of diagnostic stand- ards and control observations, these data do not lend themselves to critical inferences. More re- cently, however, Macht and Lawrence (1955), in the study described earlier, after correcting for parity differences between the two sets of data, record 5.99 per cent congenital defect among 5,461 children of radiologists and 4.84 per cent among 4,484 control children. The figures cannot be compared with those of the present study because (a) minor defects are included, (b) autopsy findings are included with no analysis of comparability of autopsy frequency in the two groups, (c) congenital conditions not usually regarded as malforma- tions, such as erythroblastosis fetalis and ate- lectasis of the lung, are included in the list, (d) multiple defects in one individual are scored separately, and (e) disease states are listed which very probably developed and were detected sometime postpartum, such as muscular dystrophy, Oppenheim's disease, spondylolisthe- sis, celiac syndrome, melanoma, lymphosarcoma, retinoblastoma, Tay-Sach's disease, missing teeth, and defective night vision. There are differences between the children comprising the two groups which are difficult to understand: There are 14 cases of "erythroblastosis" and one "died 4 days Rh positive problem" among the children of radiologists but only three cases of "erythroblastosis" among the controls. The in- terpretation of this discrepancy is complicated by the fact that the children born to radiologists before they entered this specialty are included among the controls, and, as is well-known, erythroblastotic infants are found among later pregnancies. This fact would tend to bias the results in the observed direction, although whether to the extent observed is debatable. There are nine cases of atelectasis of the lungs among the exposed group but only three among the controls. These two differences alone ac- count for approximately one-third of the ob- served 1.15 per cent increase in malformation among the children of radiologists. Although Macht and Lawrence believe the difference be- tween the two groups is significant, we do not feel they have established this fact. 14.4.3 Effects on stillbirth -frequency.â Mutations with a dominant lethal effect can, TABLE 14.4 THE EFFECT OF IRRADIATION OF MALE MICE ON THE FREQUENCY OF STILLBIRTHS AMONG OFFSPRING CONCEIVED DURING THE POST-STERILE PERIOD (After Hertwig, 1938) Dose in r 400 .. 500 .. 600 .. 800 . Number of still- offspring Stillborn born ..423 21 4.97 . . 170 11 6.47 . . 175 12 6.85 .. 693 12 1.73 1,000 590 1,200-1,400 171 1,500-1,600 65 All doses 2,287 Controls 2,616 36 0 11 103 90 6.10 0.0 16.92 4.50 3.45 X* (all doses vs. controls) = 3.649 DF=1 0.10 >P> 0.05 by definition, produce death any time from shortly following fertilization to just prior to reproduction (cf. Hadorn, 1955). If the domi- nant lethal exerts an effect during the last tri- mester of pregnancy, the result would be a stillbirth. Evidence concerning an increase in stillbirths among the pre- and post-sterile period progeny of irradiated male mice is unsatisfactory because of the tendency, referred to in the preceding section, for females following parturi- tion to devour stillborn young. Hertwig (1938) has published data, reproduced in Table 14.4, which show a small and insignificant increase in stillbirths in the offspring of post-sterile period irradiated male mice, and Strandskov (1932), in a pioneer experiment in which the dose varied from 173 to 2,592r, has recorded similar data for the guinea pig, although the latter's results must be interpreted with caution
Recapitulation 203 because the effect was present in post-sterile period but not pre-sterile period offspring. However, the number of the latter observed was quite small (39). Macht and Lawrence (1955) report 13.77 per cent "fetal death" in the children of the exposed, as contrasted to 12.50 per cent in the controls. The term fetal death "includes all re- ported cases where the product of conception was not born alive. It therefore includes mis- carriages and stillbirths irrespective of the length of uterogestation and regardless of whether or not the embryo or fetus was other- wise normal at this stage" (p. 447). From the frequency of the event in the controls, as well as the wording of the questionnaire, this term must include abortions as well. The difference between the two series was not statistically significant. Crow (1955), in a similar ques- tionnaire-type study, also failed to detect any statistically significant differences between fre- quencies of stillbirths and miscarriages among the children of radiologists and in the children of a control group of pathologists. 14.4.4 Effects on birthweight. â Strand- skov (1932) observed that when correction was made for the smaller mean litter size of females mated to irradiated male guinea pigs, mean birthweight was decreased in the post-sterile period offspring of his experiment from 90.93 gms. in the controls to 86.35 gms. in the post- sterile period offspring. The number of animals involved is not stated but can be estimated from the data in the paper to be approximately 260 for the controls and 75 for the irradiated. The data are not presented in such a way that the significance of the findings can be analyzed. The interpretation is further complicated by the fact that only the data for animals surviving at least 30 days are presented. Variances are not given for the birthweights of guinea pigs con- ceived during the post-sterile period, a point of especial interest in view of the findings of our study. There do not seem to be comparable data for the house mouse. 14.4.5 Effects on neonatal death rates. â Hertwig (1938) observed a slightly but not significantly increased death rate during the first 75 days following birth in the offspring of male mice receiving 800-1,600r, the survival figures being 79.14 Â±1.80 per cent in 508 progeny of irradiated mice as contrasted to 82.43 Â±1.38 per cent in 757 control offspring. Strandskov's data, on the other hand, involving approxi- mately the numbers referred to earlier, reveal no difference in post-natal death rate in the off- spring of irradiated guinea pigs up to 30 days postpartum. Crow (1955), in the questionnaire study referred to earlier, found that "the infant mortality rates also were not significantly differ- ent in the children of the two groups, but the numbers were very small." 14.4.6 Effects on growth and development. â The only pertinent data appear to be those of Strandskov (1932). At 30 days postpartum the mean weight of the post-sterile period off- spring of irradiated male guinea pigs, corrected for litter size, was 265.73 gms., against a con- trol value of 281.66 gms. The number of ani- mals involved is presumably the same as for the birthweight figures given above. 14.5 Interpretation of the findings. â The interpretation of the findings of the present study can now be very simply stated. The fore- going section has brought out the fact that when one considers data of a type which it is feasible to collect on human populations, the available information on other animals is scattered, frag- mentary, and often contradictory. However, this information, taken in conjunction with the much more extensive data on Drosophila, cer- tainly does not suggest a high likelihood of demonstrating clear-cut effects in Hiroshima and Nagasaki. The divergence of opinion with regard to what one might expect in human populations is sufficiently large as to make it unprofitable to explore all presently held opin- ions with regard to all of the indicators. How- ever, it does seem worth-while to explore a single class of mutations, say the sex-linked lethals. Evidence from Drosophila suggests that the yield in sex-linked lethals is approximately 3 per cent per 1,000 roentgens (Timofeeff- Ressovsky, 1937; Spencer and Stern, 1948; Muller, 1954). The yield in the mouse, on the basis of the data on induced autosomal muta- tions, might, if the mouse X-chromosome had the same number of loci capable of giving rise to lethal mutations as the Drosophila X, be greater by a factor of 15-20. Now, if the dosage estimates advanced for the various exposure classes are reasonably accurate, then the average dose in rep's received by the 5,629 mothers with exposure 3 or greater (whose spouses were in
204 Chapter XIV Genetic Effects of Atomic Bombs classes 1 or 2) is approximately 100 rep's. If we impose the restrictions indicated in the previ- ous paragraphs with regard to an adequate test, then differences in the sex ratio between the control (average exposure 0 rep's) and the ex- posed group (100 rep's) as large as 1.6 per cent (absolute change) would quite probably not be detected. Accordingly, we may estimate that the yield, in man, might be as high as roughly 2 per cent per 100 rep's and we would not detect it. This is a value six times that in Drosophila, but approximately one-half to one- third the value to be expected if human genes were as sensitive to irradiation as the small series of tested mouse genes (extrapolating from autosomal visibles to sex-linked lethals) and the X-chromosome of man had the same genetic length as that of Drosophila. This is, of course, the upper limit; the yield could be, and quite probably is, much lower. Similar con- jectures could be made for a number of the other indicators. Accordingly, we can say of the present study that under circumstances where, on the basis of what is known concerning the radiation genetics of mammals, it appeared unlikely that conspicu- ous genetic effects of the atomic bombs could be demonstrated, such effects have in fact not been demonstrated. The present study can in no way be interpreted to mean that there were no mutations induced in the survivors of the atomic blasts. Neither, on the other hand, is the reverse interpretation â that of mutation production â permissible from this series of observations, al- though, on the basis of all that is known of radiation genetics, there is no real reason to doubt that mutations were produced in Hiro- shima and Nagasaki. We are left with incon- clusive findings, albeit findings which permit us to set confidence limits.