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.
4 Evaluation of Dosimetric Methods and Results This section presents a brief critique of HTDS dosimetric methods and evaluates the quality of the dosimetric component. Our purpose is to assess whether there are any flaws in the estimation of the individual thyroid doses calculated by the HTDS, whether uncertainties are properly accounted for, and how much the NTS and global fallout might have affected the results of the HTDS. BACKGROUND . One of the disputed aspects of the estimates of exposures of children downwind from the Hanford site pertains to the amount of AT released from the site, which in this case originated in the T and B reactors on the Columbia River; these reactors became operational in late 1944 or 1945. The fuel rods were "cooled" for various periods to let the short-lived fission products decay and then carried to separation plants, where the plutonium was ultimately reclaimed and the remaining ill{ effluent was a byproduct of the chemical separation procedures. The quantities of Lit released to the air from the separation processes used to obtain plutonium from the uranium fuel rods of the reactors at the Hanford Atomic Product Operations were estimated, and meteorologic information and environmental-transfer models were used to assess the environmental concentrations of ]3~{ in air, milk, and other foodstuffs. Personal information on the whereabouts and 61
62 Review of the HTDS Draft Final Report dietary habits of the study subjects led to the estimation of the thyroid doses from islet. When the operation being described is 50 years in the past and the release and deposition of a radioisotope with only an 8-day half-life are being estimated, reconstructing the dose to the thyroids of a somewhat mobile population surrounding the effluent site is extremely difficult, even if it is based on the best records to be found and if due care is used in the calculations and interpretation of the scant data that are available. in the HTDS Draft Final Report, the thyroid doses were calculated on the basis of information provided by the HEDR project. It is the subcommittee's understanding that the HEDR project estimated the daily environmental concentrations of ]3~} (in indoor and outdoor ground-level air, in milk from a backyard cow and from commercial sources, in leafy vegetables, and so on) for the period from December 26, 1944, to December 3l, 1957, and for geographic locations in the study domain where the subjects could have been exposed from the Canadian border to central Oregon and from the eastern border of the Idaho panhandle to the center of the Cascade mountains. To do that, the HEDR project performed 100 realizations of its computer codes. In each realization, the input parameter values were randomly selected from subjective probability-distribution functions; 100 results (~3~T concentrations in air, milk, leafy vegetables, and so on) were provided for each day for each location of interest. Uncertainties were assigned according to histograms of the results. Using the calculated HI environmental concentrations and the answers provided by the subjects as to their residence and dietary histories, the HTDS estimated their i3~T intakes via inhalation and via ingestion as a function of time. The products of the AT intakes and the AT thyroid dose coefficients for inhalation and for ingestion yielded the thyroid dose estimates. It is unclear to the subcommittee how the uncertainties in 13~] intakes were estimated by the HTDS. One of the difficulties encountered in reviewing the work done by the HTDS is that the method used to calculate doses
Evaluation of Dosimetric Methods and Results 63 is not clearly described in the documents that were provided to the subcommittee. This infonnation is presumably scattered in the large number of documents that were prepared by the HEDR project and the HTDS in the course of their work. It is recommended that the HEDR project develop a single document in which the methods, assumptions, coefficients, and shape and magnitude of the coefficient uncertainties are summarized, the nature of the results provided to the HTDS is clearly descnbed (with their strengths and wealmesses), and the ways in which the HTDS made use of the HEDR results to estimate individual thyroid doses are described in detail. During the 1950s and early 1960s, two other environmental sources of ]3~{ contributed to the thyroid doses received by the populations around Hanford: the nuclear-weapons tests canted out at the NTS, mainly in 1952, 1953, 1955, and 1957; and the nuclear-weapons tests conducted outside the United States, which gave rise to "global" fallout, mainly in 1954, 1956, 1957, 1958, 1961, and 1962. Only the first of those two sources was taken into consideration by the HTDS. A crude assessment of the thyroid doses resulting from "global" fallout near Hanford has been prepared for the purposes of this review (appendix C). HEDR CALCULATIONS OF ENVIRONMENTAL CONCENTRATIONS OF 3~} FROM HANFORD AND RESULTING DOSES: ASSESSMENT OF KEY PARAMETER VALUES Because of the paucity of ]3~] environmental data, HEDR used a suite of environmental-transfer models to simulate the transfer of ]3~{ Dom the point of release to the dietary products consumed by the subjects and to ground-level air. The HEDR models have been subjected to numerous reviews, for example, the Technical Steering Pane} (TSP) and CDC review of the RATCHET model, the TSP external review of the environmental-accumulation dosimetry code development, two workshops held by CDC to address issues related to the dose and risk assessments, and a recent review prepared by this
64 Review of the HTDS Draft Final Report subcommittee. In addition, the various computer codes have been tested by staff independent of the developers to ensure correct implementation of the models. The dose assessment has been found, on the whole, to be structurally sound for the estimation of thyroid doses to representative, hypothetical persons, but minor errors have been found and doubts have been raised about the validity of the results for particular environmental conditions and for the estimation of thyroid doses to specific people. In its letter report dated January 25, 1999, this subcommittee recommended that . The HEDR investigators supplement their description of the mode} with an account of the origin of the errors made with regard to the estimation of the i3~} concentrations in pasture grass on the basis of measurements, the impact on the predicted values when the errors are corrected, and a preliminary assessment of the effect of reparameterization on estimates of absorbed dose to the thyroid. Until such a reassessment has been made it is difficult to know whether the current estimates of dose to representative individuals need recalculation. As of September 1999, the subcommittee had not received any response to its recommendation. Hourly releases of AT based on recorded data on the average reactor power levels and on the burnup of the fuel at discharge are provided in PNWD-2033 HEDR (volumes ~ and 2) (Heeb, 1993) for 1944-1947 and in PNWD-2222 HEDR (Heeb, 1994) for 1944-1949. For 1950-1972, only monthly release estimates are reported in PNWD-2222 HEDR (Heeb, 1994~. The computer mode! for the source term (STRM) generated ~ 00 realizations of the emissions for each hour for 1944-1949, which were aggregated into daily releases. The emissions were aggregated into monthly releases for 1950-1957.
Evaluation of Dosimetric Methods arid Results 65 The RATCHET model for environmental transport has a significant source of uncertainty built in that is not accounted for: the interpolation of the wind fields. This uncertainty is not spatially uniform. There should at least be a study of the sensitivity of the deposition estimates with respect to local errors in the wind field. According to Ramsdell and others (1994) (table D, inverse-square- distance weighting was used for the interpolation. That method does not produce any information about the uncertainties associated with the interpolated values. The paper does not justify the choice of the method; in particular, it is known to be very sensitive to the location pattern of the data points. Inverse-square- distance weighting is known to be "isotropic"; that is, it incorporates the distance from the data points to the interpolation point but not the directions. In the case of wind fields, one would expect the direction from an interpolation point to a data point to be important. The algorithm is also insensitive to the directions between the pairs of data points; it incorporates only their relative distances from the interpolation point. The environmental-accumulation mode} developed by the HEDR project (DESCARTES) provides the concentrations of Al in different plant and animal products for numerous space-time combinations, whereas the individual-dose mode} (CIDER) is used to estimate the Fit body burdens and doses in humans of various ages and both sexes on the basis of ingestion and inhalation pathways. The equations used in the DESCARTES and CIDER models and the key input-parameter values and their distributions can be found in PNWD-2023 HEAR (Snyder and others, 19941. The DESCARTES code was operated with a daily time step for 1944-1949 and with a monthly time step for 1950-1957 (Farns and others, 1994; Price, 19941. The thyroid dose via ingestion is due primarily to the consumption of milk produced by cows that fed on pasture grass.
66 Review of the HTDS Draft Final: Report For the purposes of this report, the thyroid dose from this pathway can be expressed as follows for an acute release of 13lT: Dm= (ST)(VEG/ST)(Tveg)(Pl)(Fm)(MDF)(MCR)(DCF)' where Dm = thyroid dose, Gy, ST = source term, Bq, VEG = concentration of i31} in pasture grass at time of fallout, Bq kg-~, TVeg = mean time of residence of 13lT on pasture grass, d, PI = pasture intake by cows, kg d-~, Em = transfer coefficient of AT to cow's milk, d L-~, MDF = milk-distribution factor (which tales into account the fact that milk consumed might not be of local · ~ ongm, MCR = milk-consumption rate by subject, L d-l, and DCF = thyroid dose coefficient, Gy Bq~l. Those key parameters will be considered in turn. Most of the parameter values are presented in Snyder and others (19941. Source term: The source term for 1944-1947, before the installation of removal devices for i31l, appears to have been well estimated on the basis of the available historical data. It is during these early years that the largest nit releases occurred, but there is concern
Evaluation of Dos~metric Methods arid Results 67 that the 13lI releases for later dates have been underestimated by the HEDR project. The following are Hof~nan's analysis of the HEDR estimates and not necessarily those of the committee. Several reasons for such an underestimation are listed in a draft report prepared by Hoffman and others (1999~: · The HEDR project estimates of the amounts of 13lT processed and released in 1959 and 1960 are substantially less than the amounts reported by Warren (1961), which are the source of HEDR release estimates. · The HEDR project documents use Warren (1961) as a source but do not evaluate the credibility of his values; for example, he projected an unrealistically high scrubber efficiency. · The HEDR project misapplied measured release-factor data from 1959-1960 to the period 1951-1957, when less emission-control equipment was in place. The HEDR project incorrectly accounted for operation of the silver reactors in the B and T plants by inexperienced personnel during the first IS months after installation in 1951. · The HEDR project substantially underestimated the source-term uncertainties for the B. T. and RED OX plants. The HEDR project inadvertently used the medians instead of the arithmetic means of the monthly source terms for the air-concentration and ground-deposition calculations. · The HEDR project did not propagate the source-term uncertainties to air concentrations, ground deposition, and doses.
68 Review of the HTDS Draft Final Report The subcommittee did not investigate whether the reasons for a possible underestimation of the source term, as proposed in Hoffman and others (1999), are valid. Hermann and Hermann (1996) evaluated the I'll releases in 1948-1960; they based their estimates of emission- contro! efficiency on investigations conducted at fuel-reprocessing plant WAK in Karisruhe, Germany, and concluded that the HEDR project had overestimated the efficiency of emission-control equipment for Gil. They also concluded that the cooling time (time elapsed between fuel discharge from the reactor and fuel dissolution in the chemical plant) might not have been properly taken into account. Hermann and Hermann (1996) estimated an ~3~} release of about 250,000 Ci in 1948-1960, whereas the HEDR estimate for the same period is about 54,000 Ci. In comparison, the AT release in 1944-1947, which seems to have been well estimated, was about 685,000 Ci. The estimates of the total releases of AT in 1944-1960 are therefore 740,000 Ci according to Heeb (1994) and 940,000 Ci according to Hermann and Hermann (1996~. Concentration of HI in pasture grass at time of fallout: The concentration of i3~T in pasture grass at the time of fallout is characterized by the value of VEG/ST. It is derived from the daily ground-deposition densities of AT that are provided by the RATCHET code (Ramsdell and others, 1994~. The deposition pattern is based on mathematical modeling, with little validation of the calculated deposition pattern of AT. As shown by Napier and others (1994), there are discrepancies between the measured and calculated values of vegetation samples that have been collected since 1945. It might have been helpful to measure the }29{ pattern of deposition in the geographic domain considered by the HTDS. Because ]29], a radioactive isotope of iodine with a very long half- life, behaves in the same way as }3~{ and because Hanford was a dominant source of AT as long as the chemical plants were in operation, measurements of AT concentrations in soil would have
Evaluation' of Dosimetric Methods and Results 69 produced valuable information regarding the overall geographic deposition pattern of its, even though the day-to-day temporal pattern of releases of ]29] iS not highly correlated with those of 13~. The derivation of the 13~{ concentration in pasture grass Tom the ground-deposition densities makes use of a relationship that is valid only for dry deposition processes. It is not clear whether that relationship was used when precipitation occurred. Hoffman and others (1992) showed that the parameter values are substantially different for dry and wet processes. Because precipitation amounts in counties remote from Hanford are about 2-3 times higher than for counties near the site (see appendix C), this could have resulted in differences in a degree of overestimation or underestimation of the thyroid doses. Mean time of residence of AT on pasture grass: The mean time of residence of ]3~] on pasture grass seems to have been well estimated. Because of the short radioactive half-life of Ill, the thyroid dose is relatively insensitive to the choice of value of this parameter. Pasture intake by cows: The values chosen for pasture intake by cows are not presented in Synder and others (1994), because they are extracted Tom a database described in Beck and others (1992) but later revised in an unpublished report submitted to the chair of the HEDR Technical Steering Pane} (oh. Till) in 1994. It would have been helpfi~} to the subcommittee to have a report available on the topic. Transfer coefficient of 13~} from cow's intake to milk: The central estimate of the transfer coefficient of AT from cow's intake to milk is taken to be I.2 x lo-2 ~ LO for a dairy herd and 9.2 x 10~3 ~ is-] for a single cow (Snyder and others, 1994~. Those values seem to be too high by a factor of about 2,
70 Review of the HODS Draft Final Report compared with the central estimates used for NTS fallout 4 x ~ 0-3 L-r in NCT (1997) and 4.2 x 10~3 ~ it-} in Simon and others (1990) and with values obtained after the Chernobyl accident, which ranged from 2 x 10~3 to 6 x 10~3 ~ ~-l, according to Hoffman and others (1988~. However, this is an open issue, in that the measured values of this transfer coefficient vary over a large range for reasons that remain largely unexplained. Milk-distribution factor: Data on the distribution of milk within the project domain are not presented in Synder and others (1994), because they are extracted from a database described in Deonigi and others (1994~. It would be helpful to know how the uncertainties and the correlations in the distribution of milk were taken into account by the HTDS. Milk-consumption rate: The HEDR project selected milk-consumption rates for representative people in 12 age and sex categories (Anderson and others, 1993; Fares and others, 1994) on the basis of the results of the 1977-1978 Nationwide Food Consumption Survey and extrapolation back to 1945-1957. The HTDS used these rates when reliable information could not be determined from the personal interviews. However, how they were applied is unclear in the Draft Final Report. Because a number of assumptions are made (without quantifying the uncertainty) in Anderson and others (1993) and extrapolations are made from current food-consumption habits back to the relevant period, largely on the basis of broad-scale measures rather than measures particular to the people in the study, there will be at least an uncertainty associated with the extension of population or group practices to specific persons. it would be useful to know whether the milk-consumption rate distributions given in Anderson and others (1993) were used by the HTDS and whether milk-consumption rates were assumed to be constant in a
Evaluation of Dosimetric Methods and Results 71 given age category or were taken to vary with age within an age category. Thyroid dose coefficient: The value of the thyroid dose coefficient depends on three parameters: the uptake of IT from blood by the thyroid, the time of retention of 13~{ in the thyroid, and the mass of the thyroid gland. The values of those three parameters depend to some extent on the amount of stable iodine in the diet. The thyroid dose coefficient varies strongly as a function of age, the average value for infants being about ~ O times greater than that for adults. That is caused by the variation of the mass of the thyroid gland as a function of age; the uptake of ill} from blood by the thyroid and the time of retention of }3~} in the thyroid do not vary substantially with age. The central estimates of the thyroid dose coefficients that are used by HEDR and the HTDS are taken from the international Commission on Radiological Protection BURP. ~ , 1990) and assume a sufficient amount of stable iodine in the diet; the uncertainties are taken from Dunning and Schwarz (1981~. Those sources of information are generally accepted as the best available. However, it would be useful to know whether the thyroid dose coefficient was assumed by the HTDS to be constant within a given age category. HEDR DOSIMETRY-MODEL VALIDATION For an ideal validation of the dosimetry model, one would like to have measured doses for at least a subset of the people in the study; then one would compare the doses predicted for those people with the measured doses. But those data do not exist. instead, the validation study used other people for whom some measurements were available. - -- 1- - -1-- The validity of the comparisons hinges on the cross-applicability of various assumptions and parameter values between the test population and the study population. Shipler and others (1996), of the HEDR
72 Review of the HTDS Draft Final Report project, acknowledge the lack of fully appropriate data for the validation exercise. However, a substantial effort was made to validate the HEDR models with the data available. The model-validation exercise consisted of comparing computational-mode} estimates with limited historical field measurements and experimental measurements that were independent of those used to develop the HEDR models. The results are provided in PNWD-2221 HEDR (Napier and others, 1994~. The following review focuses on the results of the model-validation exercise and on the selection of the values used for the key parameters. The thyroid doses received via ingestion were essentially due to the consumption of cows' milk, resulting from the cows' intake of fresh pasture grass. The historical measurements that are the most interesting for testing the validity of the calculated ingestion doses are the thyroid burdens, the milk concentrations, and the pasture-grass concentrations of ]3~{, in that order. Similarly, the historical measurements that are the most interesting for testing the validity of the calculated inhalation doses are the thyroid burdens and the ground-level air concentrations of silt. Thyroid burdens Nearly 7,900 thyroid burdens of Hanford workers are available for the period June 1945-August 1946 (Napier and others, ~ 9941. The report did not indicate how many workers were included in the 7,900 measurements nor how the sample of workers was drawn. it was assumed that all workers had environmental exposures and that occupational exposures did not affect the mean thyroid burdens. The mean monthly thyroid burdens were Tower by a factor of 2-4 than the values calculated by the HEDR mode} for 1945 but were in agreement with the calculated values for 1946. The calculated values were obtained by using the assumptions that the workers lived in Richiand and that the milk that they consumed originated in the area around
Evaluation of Dosimetric Methods and Results 73 Sunnyside. Under those conditions, the HEDR investigators believed that most of the thyroid burden was due to inhalation. in view of the importance of the thyroid-burden data set to validate the HEDR models, it seems that a larger effort could have been devoted to the comparison of the observed and calculated values. For example, whether all workers actually lived in RichIand and drank commercial milk originating in Sunnyside could have been checked. If it was true, the variability of the observed values, which is not provided in PNWD-2221 HEDR (Napier and others, 1994), could have been used to estimate the random uncertainties in individual doses, and research could have been earned out to investigate the bias observed for 1945. If not, the data set could have been stratified according to place of residence or origin of milk, and calculations done for each subgroup. For people with many thyroid-burden measurements, it also would have been worth while to analyze the temporal variation of the results. Two more thyroid burdens are available for 1963. An acute, inadvertent release at the PUREX plant in September 1963 was extensively studied. In particular, pasture grass and milk at two farms~alled Farm A and Farm B were measured regularly, and thyroid counts were taken on two children who were consuming milk from a family cow at Farm B. The thyroid doses that were calculated for the two children with the HEDR models are in very good agreement with those derived from the thyroid burdens: there is a difference of only about 20% between the "measured" doses and the medians of the calculated doses. However, because the thyroid counts were taken ills months after the PUREX acute release, it is likely that a fraction of the thyroid burdens arose from releases that occurred between the PUREX acute release and the time when the thyroid counts were made (Goble, 1968~.
74 Milk concentrations Review of the HTDS Draft Final Report The only milk concentrations reported in PNWD-2221 HEAR (Napier and others, 1994) are those measured at Farm A and Farm B in September ~ 963. The measured milk concentrations at the two farms differed by a factor of about 3. The calculated milk concentrations were the same at the two farms, and they were included in the same calculation cell, although Farm A was on the western edge of the cell and Farm B was on the eastern edge. The measured and calculated milk concentrations were in very good agreement for Farm A, whereas the measured milk concentrations in Farm B were about 3 times higher than the calculated milk concentrations. As indicated in appendix C, concentrations of 13~} in milk were measured around Hanford as early as 1958. It would have been valuable to investigate how the calculated concentrations agree with the measured values. Concentrations in vegetation Historical measurements of 13~{ in vegetation at the Hanford site were available to the HEDR project for the period beginning in the middle of ~ 945 and continuing to the present. For example, over 3,500 samples were reported for 1946. The samples in which 13~{ was measured were most often labeled "vegetation", but it was noted that the samples were usually sagebrush. Sagebrush is very different from pasture grass, so these measurements might not be very useful for deriving the concentrations in pasture grass. And errors were made because wet weights were sometimes used instead of dry weights; these errors are being corrected by Pacific Northwest National Laboratory, which will issue a revised analysis. The subcommittee will review the revised analysis when it is available. Concentrations in air Relatively recent Hanford data on atmospheric dispersion include a dataset of coupled monthly source terms and
Evaluation of Dosimetric Methods and Results 75 environmental measurements of atmospheric krypton-85 (8sKr). The number of locations sampled per year increased from four in 1984 to I] in 1987; up to 12 samples per year were taken from each location. Detailed results were presented for three of the sampling stations that were in the same grid cell (North of Hanford Site 300 Area). Air concentrations were calculated from two sets of meteorologic data: those available in the 1940s and those, more complete, available in the 1980s. The calculations were in good agreement with the measurements when the 1980 meteorologic data were used and were about 3 times higher than the measurements when only the meteorologic data available in the 1940s were used. Much less detailed information is presented for the other sampling stations where 85Kr was measured. It would have been valuable to present the results in the same fashion as in the "North of Hanford Site 300 Area" grid cell to identify possible location-related biases in the calculated values. Conclusions On the whole, the results presented are good, inasmuch as there is agreement between measured and median calculated values to within a factor of 3. However, the thyroid burdens and the ~sKr air concentrations could have been analyzed more thoroughly. In addition, errors in the assessment of the concentrations in vegetation should be corrected, and many isolated data were not used in validating the models. Possible Underestimation or Overestimation of Doses Hof~nan and others (1999) and He~mann and Hermann (1996) have reviewed the HEDR dosimetry and claimed that thyroid doses were substantially underestimated by the HEDR project. However, the divergences that they identified occurred during years when 13~t releases were rather small, and their findings and HEDR results differed by only about a factor of I.3 in total releases.
76 Review of the HODS Draft Final Report If doses were underestimated, it is important to consider the implications. To the degree that doses are underestimated, the imputed risk estimates will be too large, but systematic, across-the- board dose underestimation does not alter the statistical significance of dose-response trends. Hence, in the HTDS-in which, as it turned out, the primary issue became whether there is any association between ]3~} exposure and thyroid diseases-the effect of possible dose underestimation might be minimal. If there was systematic underestimation of doses, the true statistical power of the study would be greater than one would estimate it to be given the dose distribution used by the HTDS, in which case negative results would be more persuasive than they are. However, if the doses were underestimated more for some study subjects than for others or if the effect of dose underestimation was greater for some subjects than for others (for example, because some are younger than others), variation in the underestimation or in its effect would act as another source of measurement error and tend to cancel out the gain in statistical power achieved by having generally higher doses. Therefore, a simple generalization about the effects of dose underestimation cannot be given. In contrast, the HEDR project might have overestimated the transfer coefficient from cow's intake to milk by a factor of 2, and some of the validation comparisons also suggest an overestimation of doses by the HEDR calculations. The effect of dose overestimation could have implications for the interpretation of HTDS results because overestimation would mean that the statistical power of the study was also overestimated; this would imply that the negative results were not as definitive as the investigators asserted. Doses might have been underestimated in some areas and overestimated in others. Possible reasons are errors in the deposition pattern of ]3~{ due to inaccuracies in wind flows or precipitation amounts and errors in the milk-distribution pattern. it is beyond the mandate of the subcommittee to assess the direction and magnitude of possible dose misestimation -
Evaluation of Dosimetric Methods and Results 77 definitively. As stated in a previous letter report (National Research Council Letter Report to CDC: HEDR-Issues Regarding the Hanford Environmental Dose Reconstruction (HEDR) Atmospheric I-131 Pathway Models dated January 25, 1999), the subcommittee found the dose-reconstruction effort to be, by and large, a credible effort, although there were unresolved questions and considerable uncertainty in doses because of the lack of adequate histoncal measurements and other relevant data. DOSE UNCERTAINTIES Considerable efforts were made by the HEDR project to mode! the uncertainties attached to estimated thyroid doses. Uncertainties for many parameters used in the environmental pathways and radiologic dose modules (DESCARTES, CIDER, and CRD codes) are listed in Snyder and others (19941. The HEDR investigators assigned probability distributions to the various parameters on the basis of their assessment of the uncertainties in the parameter values. They then obtained 100 estimates based on the computer codes. in each estimate, for each parameter the values were randomly selected from the probability-distribution functions, so 100 results (~3~{ concentrations in air, milk, lead vegetables, and so on) were generated for each day for each square in their grid that defined a location in their geographic domain. Uncertainties were assigned according to the frequency distributions ofthe 100 results. A key point here is the possible distinction between variability (added to the data) and uncertainty inherent in the data. If one has an estimate for a parameter value and adds a random component, such as meteorologic data (Ramsdell and others, 1994, p. 571), this does not characterize the uncertainty in the original estimate (although the authors claim that it does). Adding the random component does not result in a random component with the same distribution as the added part, but rather is the convolution of the distribution of the original and the distribution of the random component. Regrettably, the distribution of the
78 Review of the HTDS Draft Final Report original estimate, the distribution of the error of estimation, is urlkno~vn. in other places in the publications mentioned above, the authors indicate that they have sampled from a subjective distribution, as opposed to using the estimate plus a random component. Aside from the question of whether a proper subjective distribution has been chosen, this approach is more appropriate. However, it is difficult to quantify the uncertainty associated with choosing or fitting the subjective probability distribution, and the authors have not incorporated this uncertainty into the calculations. There might be adequate explanations for these discrepancies that do not appear in the publications or in the Draft Final Report. The omission and the lack of clarity of what the HEDR project dice constitute inadequacies in communication and perhaps in scientific methods. Without adequate explanations in the Draft Final Report, the committee cannot confirm that the propagation of error in the dose reconstruction (the source term in particular) was the best choice of methods and is correct. This situation adds to our overall concern about the integrity of the uncertainties of the dose estimates and how they could affect the power calculations. By "integrity of the uncertainties" the subcommittee means the inclusion of all of the uncertainties and then understanding and using them in an appropriate way in the epidemiologic analyses. It is not clear to the committee what was done by the HEDR project about the parameters that are not listed in Snyder and others (1994) or about the mode} uncertainties. For example, . According to Hoffnan and others (1999), the HEDR project did not propagate the source-term uncertainties to air concentrations, ground deposition, and doses for the 1950-1972 period. Evidence of uncertainty propagation could not be found in the HEDR reports made available to the subcommittee.
Evaluation of Dosimetric Methods and Results 79 · No information could be found on the uncertainties related to the cows' feeding practices or to the commercial distribution of milk. . No information could be found on the correlations between parameter values. · No information could be found on the uncertainties related to the use of models to calculate thyroid doses from exposures. The uncertainty estimates attached to the HTDS thyroid doses appear to be too low. The geometric standard deviations (GSDs) of some of those doses are estimated to be about 1.4. The GSD of the thyroid dose-conversion factor alone is taken to be 2.0 (Snyder and others, 1994), so, the GSDs of thyroid dose estimates should be at least 2. The authors of the report should explain why GSD values of less than 2 are found for many HTDS thyroid-dose estimates. It is to be noted that the GSDs of the thyroid doses calculated by the HEDR project are 2 or greater (Harris and others, 1996~. An increase in the GSD from I.4 to 2.0 quadruples the width of the 95°/0 confidence interval. The HTDS did not take into account the uncertainties associated with the recall 5 decades after the period of exposure-of the origin of milk, the milk-consumption rate, or changes in residence. That is a serious flaw of the methodology, which is discussed further in chapter 6. In summary, although the subcommittee believes that the methods used by the HEDR project to estimate thyroid doses and their uncertainties are on the whole structurally sound, there remain quite a few errors or omissions that should be corrected, and supplementary work is desirable. in a National Research Council letter report dated July 27, 199S, the committee stated that the resulting dose estimates are highly uncertain, and the use of self-reported information in the construction of dose estimates
80 Review of the HTDS Draft Final Report presents an opportunity for bias, such as overreporting or underreporting of activities that might substantially affect exposure estimates. The investigators are urged to evaluate further the use of some of the interview data to corroborate the HEDR exposure estimates and the default options. Modeling has inherent limitations that need to be acknowledged candidly. Given the insufficient data available on Hanford, there will always be considerable imprecision in the individual doses, and it is unlikely that further refinements in the mode} will increase the precision of these doses substantially. It warrants noting, too, that although the uncertainties in some of the components of the mode! have been accounted for, this is not true for all of them, and the total uncertainty is likely to be greater than has been publicly stated. This subcommittee stresses again that a review of the dosimetry work conducted by the HEDR project and the HTDS is difficult in the absence of a single dosimetry document in which the HEDR work is summarized, the nature of the results provided to the HTDS is clearly described (with their strengths and weaknesses), and how the HTDS made use of the HEDR results to estimate individual thyroid doses is described in detail. In the absence of such a document, however, it seems that the uncertainties in the thyroid doses were underestimated by the HEDR project and consequently by the HTDS. ~ FROM THE NEVADA TEST SITE AND GLOBAL FALLOUT During the 1950s and early 1960s, two other environmental sources of i3~{ contributed to the thyroid doses received by the populations around Hanford: the nuclear-weapons tests carried out at the NTS, mainly in 1952, 1953, 1955, and 1957; and the nuclear-weapons tests conducted outside the lower 48 states of the United States, which gave rise to so-called global fallout, mainly in ~ 9 5 4, ~ 9 5 6, ~ 9 5 7, ~ 95 8, ~ 96 I, and ~ 962. if global or NTS fallout resulted in significant thyroid doses in the counties used in the HTDS, and if the variation in doses among counties used in the HTDS was large for the sources of additional
Evaluation' of Dos~metric Methods and Results 81 exposure, then it would be important to take the global and NTS fallout into account in the HTDS. Only the first source was taken into consideration by the HTDS. For the purposes of this review, a crude assessment of the thyroid doses resulting from global fallout near Hanford has been prepared in order to obtain preliminary evidence as to whether global fallout is likely to be an important confounding variable in the analyses pertaining to Hanford fallout and thyroid disease. Details of the assessment are provided in appendix C. The following paragraphs present a brief summary of the methods used and the results. The global tests were carried out primanly in the Pacific at Bikini and Eniwetok and in the Soviet Union at Novaya Zemlya. The total fission yield from the tests was over ~ 50 megatons, compared with a total of about ~ megaton for the tests cattier out at the NTS. However, fission yield is only a crude indicator of the fallout deposition in the continental US because about 80°/O of the debris was injected into the stratosphere, where i3~} decayed and therefore did not contribute to human exposure. Of the 20% of the debris injected into the troposphere, a considerable fraction was probably deposited locally or regionally; the remainder was deposited primarily in the latitudinal band of the location of the test site and was mainly associated with rain. It would be extremely difficult with atmospheric-dispersion and- deposition models to predict with reasonable accuracy the fallout deposition at a given site in the lower 48 states on the basis of the yield of a particular test. The dose-assessment methods used were based on the sparse fallout data and on the small number of IT concentrations in milk in the Hanford area for some of the testing period that are available. The basic methodology used by Beck (see appendix C) to estimate the 13~} doses near Hanford was as follows: . For 1961 and 1962, the doses for Franklin, Adams, and Benton counties were estimated directly from the measured ]3~{
82 Review of the HTDS Draft Final Report in milk at farms near Hanford after the milk data were corrected for Hanford plant contributions. All three counties are part of the same milk shed and receive similar amounts of rain, so the milk-concentration estimates were assumed to apply to all three counties, and a single set of dose estimates was made for these three counties. (The doses in Yakima and Kittitas Counties, which are in the milk-producing area, would also be similar to those in Franklin, Adams, and Benton counties because of the similar rainfall pattern.) . The strontium-89 (89Sr) depositions in 1957, 195S, 1961, and 1962 were then estimated from the 89Sr depositions at sites in the western United States and the measured monthly rainfall in counties in the Hanford area. The ratios of the deposition in WalIa WalIa and Stevens counties to deposition in Franklin, Adams, and Benton counties were used to estimate the milk concentrations in those counties in 1961 and 1962 from those measured in milk near Hanford. The calculated deposition in 1958 relative to 1961 and 1962 was then used to estimate the relative concentrations in milk in 1957 and 1958 in three counties. . The 89Sr deposition in 1956 and 1954 relative to 1958 was estimated from the gummed-fiIm data, and the concentrations in milk were assumed to vary with the estimated deposition. · Finally, the thyroid doses in each county were estimated by using the conversion factors for milk concentration to dose that were used for that county in the National Cancer Institute study (NCT, 1997) for the estimation of )3~{ thyroid doses resulting from tests earned out at the NTS.
Evaluation of Dosimetric Methods and Results 83 The thyroid doses from global fallout calculated as described above for the Hanford area are lower but of the same order of magnitude as those from NTS fallout, as shown in table 4.~. Considering the uncertainties in both the NTS and global fallout dose estimates, the differences between the two sets of dose estimates are probably not statistically significant. Thyroid doses for infants, children, and teenagers are also provided, by year of fallout and by county, in appendix C. Table 4.~ Comparison of average estimated 13~] doses (mGy) from Hanford, the Nevada Test Site (NTS), and global sources for study counties County HTDS Global fallouta NTS fadeouts Benton 172 3.8 12 Franklin 248 3.8 ~ ~ Adams 169 3.8 12 Walla WalIa 86 8.7 30 Fee Stevens 39 ~ I.8 14-19 Okanogan ~ ~not available a Estimated doses for teenagers (see appendix C). b Estimated doses for children born January I, ~ 945. in theory, global fallout is a potential confounding variable in the analyses of thyroid-disease risk posed by Hanford fallout. But in reality, the degree to which global fallout could confound the analytic results appears to be limited. The range of variation in global fallout between the Hanford high-dose and low-dose counties is small compared with the variation in Hanford dose (see table 4.~. in addition, the global-fallout exposures occurred during the teenage years and early 20s for the study population. Data from the best studies of external radiation exposure to the thyroid strongly indicate that the thyroid is much more sensitive to radiation-induced cancer in childhood, especially before the age of 5 years, than in adolescence or adulthood (Ron and others, 1995~.
84 Review of the HTDS Draft Final Report Hence, the effect of the Hanford exposures on thyroid-disease risk would be expected to be much greater per unit dose than exposures from the NTS or global fallout. Nevertheless, since no actual analysis of the impact of global fallout on the results has been done, the committee recommends that the HTDS investigators do so. OTHER SOURCES OF RADIATION EXPOSURE Ideally, estimates of the diagnostic and therapeutic medical radiation exposures received by each person in the HTDS should be factored into a dose-response assessment. The investigators did include questions about history of substantial medical radiation exposures in the questionnaire administered to subjects. in a supplementary set of analyses, they evaluated whether those exposures were confounders or effect modifiers of the dose-response results for id versus thyroid diseases. However, they did not state how the radiation exposures were coded for analysis or how reliable the responses were, and this could be problematic: a crude, dichotomous coding or unreliable reports might adjust only partially for potential confounding. Nevertheless, it is not very likely that medical radiation exposure would be a confounding variable, in that the frequency and intensity of such exposures would have to be correlated with the magnitude of the Hanford IT doses for confounding to occur. The thyroid doses from the medical radiation exposures asked about would probably vary appreciably among medical procedures and radiological practices. For example, in the 1940s - a Or r ~-a ___-~--~- r~ and 1950s, the radiation exposure from taking a full-mouth diagnostic x-ray picture was often several hundred milligrays to the cheek. The machines often had poor collimation (Budowsky and others, 1956; Nolan, 1953) (that is, there was appreciable scatter of the radiation), and the adequacy of neck shielding was probably variable. Hence, some persons could have received thyroid doses from diagnostic dental procedures that exceeded those from Hanforc! fallout.
Evaluation' of Dosimetric Methods arid Results 85 Questionnaire information on historical exposures is usually insufficient to impute thyroid doses to individuals, so one would prefer to have documented medical radiation doses to use in the analyses. However, it would be impossible to retrieve records to document many radiation exposures that occurred 2-5 decades ago, so it would not be feasible to estimate lifetime medical radiation exposures. In lieu of that, the investigators did list (in table VIl]:-46 of the Draft Final Report) the reported medical diagnostic and therapeutic radiation exposures for each thyroid- cancer case, and this is informative.
