NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Institute of Medicine (US) Committee on Thyroid Screening Related to I-131 Exposure; National Research Council (US) Committee on Exposure of the American People to I-131 from the Nevada Atomic Bomb Tests. Exposure of the American People to Iodine-131 from Nevada Nuclear-Bomb Tests: Review of the National Cancer Institute Report and Public Health Implications. Washington (DC): National Academies Press (US); 1999.

Cover of Exposure of the American People to Iodine-131 from Nevada Nuclear-Bomb Tests

Exposure of the American People to Iodine-131 from Nevada Nuclear-Bomb Tests: Review of the National Cancer Institute Report and Public Health Implications.

Show details

2Review of the NCI Radiation Dose Reconstruction

The National Cancer Institute's report consists of two volumes. One sets forth the history of the study, its methods, and findings, and the second consists of extensive tabulations of the data on which the results and findings are based. The second volume does not include the gummed-film data that are said to be archived at the Health and Safety Laboratory (U.S. Department of Energy Environmental Measurements Laboratory) in New York City.

The purposes of this chapter are to determine

  • Whether what NCI was asked to do had a reasonable chance of success, given the sparseness of the data and the difficulties involved in reconstructing events that occurred 40-50 years prior.
  • Whether there are identifiable weaknesses in the techniques or assumptions used to perform the dose reconstruction.
  • Whether the amount of detail presented in the report about thyroid dose estimates of representative individuals by county, by age group, and by milk-drinking habits is meaningful and useful to the public and the public-health establishment.
  • Whether the dose reconstruction produced gives useful information about an individual's thyroid dose and, by implication, his or her personal risk of thyroid cancer.
  • In addition, we present a summary of estimated doses from the NCI report to illustrate variation in exposure by geographic location and by year of birth.

In this chapter, the committee first lists the contents of the NCI report and gives general impressions of the thoroughness of the study and of the soundness of the study's results. Then there is a general description of the pathways of iodine-131 (I-131 or 131I) exposure in humans. Most of the chapter is devoted to a review of the NCI report, and specific comments are made about NCI's methods, their appropriateness, and their shortcomings.

Box 2.1 lists chapter contents from Volume 1 of the NCI report and the appendices of Volume 2. Separate from the published report, NCI staff prepared estimates of the number of people who might be expected to develop fallout-related thyroid cancer (see Appendix B of this report). Additional information is also available on the NCI Web site (

Box Icon

BOX 2.1

Summary Contents of the 1997 National Cancer Institute Report Estimated Exposures and Thyroid Doses Received by the American People from Iodine-131 in Fallout Following Nevada Atmospheric Nuclear Bomb Tests.


Given the challenges posed by the paucity of the data relevant to dose reconstruction, the disparity between the original purpose of the data collection and the requirements for dose reconstruction, the need to derive amounts of I-131 ingested on the basis of measured deposition of gross beta activity, and the problems associated with drawing conclusions about counties and states based on incomplete information, this committee was generally impressed by the clarity and thoroughness of the NCI report and the obvious attention to detail. This committee does note, however, that some of the procedures used in dose reconstruction were not sufficiently detailed (for example, the discussion of kriging) for the committee to judge fully the efficacy of NCI's work and the accuracy of the results. The task confronting the NCI investigators was obviously formidable, requiring the reconstruction of events—some of which occurred almost a half-century ago—for which direct measurements of dose or even environmental contamination were often limited or nonexistent, particularly in areas of the continental United States not immediately adjacent to the Nevada Test Site.

There is an important way in which the presentation of NCI's findings is inadequate. The report gives very little statistical tabulation of dose estimates that is useful from a scientific or epidemiologic standpoint. The important numbers, averaged dose by birth cohort, used in the Land presentation, for example, cannot be easily verified as being consistent with the report. The report, the annexes, and Internet Web pages do not appear to allow for the ready computation of average doses for broad categories of people or for the uncertainties in those average doses. This makes it difficult to apply the findings of the report to compute, for example, doses for birth cohorts of people living in the regions covered by tumor registries in two different states or cities, as would be needed in the design stages of an epidemiologic analysis. The numbers provided in the map captions could be used to tabulate average dose by birth cohort, for example, but it would be laborious and possibly inaccurate to use the maps to do this by region. The report and Web pages are geared toward calculations at the county or individual level, and can be used for that purpose. However, the ability to make confident distinctions in the differences of estimates for individual persons or counties is in doubt, given the very sizable between-county uncertainties. This problem could be fixed in future work on the Web pages by allowing for the computation of regional doses, as well.

For most locations the NCI dose analysis begins with the gummed-film measurement data. This is appropriate despite the uncertainties in the film's retention of deposited radionuclides, possible losses due to washoff, and the like. That said, the estimates are most useful for average doses for representative groups of persons; the wide range of individual conditions causes estimates for specific individuals to be highly uncertain. The state with the greatest reported range in per capita exposure is Idaho, with an estimated average thyroid dose of 0.157 gray (Gy, 15.7 rad) in Custer County and 0.017 Gy (1.7 rad) in Bingham County. In addition to the dose estimate, there is an estimate given of the variability of the dose estimates, as the geometric standard deviation (GSD) for each county average dose estimate (3.9 and 1.4, respectively, for the two counties).

As described in the NCI report, GSD can be used to provide approximate 95 percent confidence intervals for county dose estimates by multiplying (to give the high end) and dividing (to give the low end) the dose estimates by the GSD raised to the power 1.96. Thus, in Custer County the range of the likely average dose is (0.011-2.26 Gy), the range in Bingham County is 0.0088-0.0329 Gy, and there is an overlap in the likely range of dose of 0.011-0.033 Gy between these two extreme counties.

One important question left unanswered by the NCI report is, When ranges overlap like this, is there really any statistical difference between the county estimates? The answer depends on the proportion of the GSD due to uncertainties shared among counties (a systematic component) and the proportion of the GSD that varies from one county to another (between-county variation). For example, suppose all of the uncertainty (the entire GSD) in the dose estimates in these two counties is caused by uncertainties in the total amount of I-131 coming from the cumulative tests. In this case, the uncertainties in the county dose estimates reflect systematic variation only, and even though the 95 percent confidence intervals for the county estimates overlap, it is likely that if the dose estimate in Custer County is too high, the dose estimate in Bingham County is too high as well. In this case, there is no chance that the dose in Bingham County is actually higher than it is in Custer County. If, on the other hand, all of the uncertainties in the county dose estimates are due to between-county variation, there is a small chance (about 6.5 percent, based on standard statistical calculations) that the doses in Bingham County are actually higher than are those in Custer County. The NCI report is silent regarding the differentiation between systematic and within-county variations in dose. This distinction is, nevertheless, important for the epidemiologic and public-health implications of the report and for whether the amount of detail offered by the report is meaningful. In its appraisal of the NCI report, this committee has attempted to determine what effect the techniques used may have had on the estimates of uncertainty provided in the report, and on whether between-county or systematic components of uncertainty would predominate.


Although the NCI report focuses on exposure to I-131, as its mandate required, it warrants noting that about 200 radionuclides are produced and released from atomic weapons tests, and exposure to these radionuclides could pose health risks to organs other than the thyroid. For the very large tests that were conducted by Russia and the United States, but not at the NTS, there was sufficient energy released to carry most of the radioactive debris into the stratosphere where it resided with a half-time of a few years. This was the mechanism that gave rise to “global fallout,” and because the debris came down to earth months to years after its creation, nearly all of the short-lived radionuclides had already decayed. Some of the more prominent surviving radionuclides in global fallout are carbon-14, strontium-90, and cesium-137 that are incorporated into environmental media and foodchains and persist for decades. Global fallout is dispersed broadly over the hemisphere in which it was injected. Effective dose commitments from global fallout have been evaluated by UNSCEAR (1993). The total effective dose commitment to individuals in the north temperate zone is 4.4 mSv (0.44 rem) per person, of which 2.6 mSv (0.26 rem) is from carbon-14 and only 0.08 mSv (0.008 rem) from iodine-131. Of more interest within the present context is the fallout from the relatively low-yield shots at the NTS. Much of this debris remained in the troposphere, and according to Beck and colleagues (1990) about 25 percent of this material was deposited within the United States within a few days. In this case there was not time for the many additional short-lived radionuclides to decay before being deposited. The primary manifestation of the deposition of the additional radionuclides was a contribution to external gamma dose during the first month after each event. No organ other than the thyroid would have received a larger dose from internal radiation than it received from external radiation (Ng and others 1990). Organs receiving larger internal doses—but less than that from external exposure—were the lower large intestinal wall, bone surface, upper large intestinal wall, red marrow, and stomach wall. Thus, given the public concern about doses from iodine-131, further understanding of the risks to the U.S. population could be gained by a detailed evaluation of the dose and risks from the other radionuclides released by the nuclear tests at NTS. Given that the work has already been done for iodine-131, it would be relatively easy to adapt the method for the additional radionuclides of interest though this exercise is not viewed as an urgent public health necessity.


Each nuclear weapons test produced a large cloud of radioactive fission and activation products and residual nuclear materials, such as plutonium-239. The cloud also included entrained debris from the weapons' casing and varying amounts of other debris, principally soil and the shot tower material if the test was detonated on a steel tower or in such a manner that the fireball touched the ground. Tests conducted aloft, either tethered to a balloon or dropped from the air, contained substantially less debris, and the nearby fallout pattern for balloon or air shots was much less than it was for tower shots. Although the fallout debris was distributed through many layers of the atmosphere, the major part for tower shots would exist in two layers, one at lower elevations that distributed into the lower level wind flow at the time of the test and the other at much higher elevations. Many of those lower level flows moved east into Utah and northward into Idaho. The large particles entrained in the lower stratum fell out over several hours. The upper level debris contained a large fraction of small particles that settled out over large areas as the material moved eastward. Although the patterns varied somewhat, they followed the path of the major air mass present, forming fairly uniform trajectories that compared well with the high-level weather data. Exposure in some counties in the west was attributable to only a few weapons tests, but many counties received deposition from numerous events.

The NCI report (Section 3.6) estimates that only 25 percent of the activity produced was deposited in the United States. Of the remaining 75 percent, much remained in the troposphere and stratosphere and was carried around the globe. However, that portion in the stratosphere decayed before returning to Earth because of the long residence time relative to the half-life of I-131; that in the stratosphere remained entrained for several weeks and a sizable fraction decayed while still airborne; the remainder fell out nonuniformly over the Northern Hemisphere. Even though rainout produced nonuniform patterns and substantial depositions in some locations, the major proportion of the I-131 was deposited in the oceans.

Most of the I-131 deposited in the United States fell on soil or vegetation other than pasture grasses; much that fell on grass was not consumed by cows, and some decayed during the processing and distribution of the milk.

Iodine-131 attaches readily onto dust and other particles and onto vegetation. It is also readily entrained in precipitation and falls to Earth with rain. Once it reaches ground level, a fraction is deposited on grass, dust particles, or by direct absorption into leafy structures. The time-integrated concentration of I-131 on vegetation (per kilogram dry weight) will be several hundred to many thousands of times the time-integrated concentration in ground level air. The large grazing range of cows further integrates or amplifies their accumulated activity into a substantial intake, and the I-131 is taken up in the cow's thyroid and secreted directly into the milk.

A person who consumes cows' milk will concentrate about 25 percent of the intake into the thyroid, which has a mass of about 2 g in a 1 year-old child and about 20 g in an adult. Any I-131 that reaches the thyroid is likely to decay in the gland because the biologic clearance time (the biologic half-life), which is 80-120 days for adults and likely much shorter for young children and infants, is substantially longer than the radioactive half-life (8 days).

Consequently, I-131 deposited over a large area can be concentrated by the pasture-cow-man food pathway into the very small human thyroid to produce relatively high doses of radiation. Any delay or shortened time in any step of this pathway has an equally significant effect on the eventual dose: Time to deposition on the pasture, storing of hay or delaying of pasturing, and time between milking and consumption all affect the radiation dose actually delivered. These are highly variable factors that depend greatly on individual circumstances and the ruminant from which the milk is derived. For example, goats concentrate more I-131 in milk than cows do, and if goats' milk or cows' milk is consumed from private stock, doses will be higher because the milk will be consumed quickly and will not be diluted by other less contaminated sources during pasteurization and distribution. Therefore, the most important determinant of dose for each person was the amount of milk consumed within a period of a few days after the fallout from each test.


Determination of the doses to members of the public is subject to the highly variable behavior of individuals and the uncertainties in the calculated pathway levels. For these reasons, the NCI report presents doses as averages for representative groups. This is convenient and somewhat useful, but it begs the issue of the variability of dose for individuals as opposed to county or group averages. Only limited information is given in the report on the variability of individual dose estimates around the group or county averages. It is clear, however, that individual dose estimates must be more uncertain than are county averages. As seen in Chapter 6 of the NCI report, age at exposure is important in determining thyroid dose estimates, as is individual milk consumption. Thus, individual doses will vary considerably about the county averages, especially in children. There are two independent components to this variability. First, there will be variation in average doses within counties because of different exposure conditions at different locations and because of differences in the size of the counties and the size and intensity of a storm causing the deposition. Second, there will be variation between individuals with the same exposure conditions because of differences in consumption.

The first source of variation was shown in Scotland to be at least 1 order of magnitude (a factor of 10 times) for the distribution of the isotopes of cesium after the Chernobyl accident (Sanderson and others 1994). Local differences in precipitation rate and runoff on slopes led to variation in cesium contamination of soils and pastures. The variation in I-131 contamination of grasses could be less than that for cesium because soil-to-leaf transfer, and hence the effect of runoff of rain on slopes, is much less important for I-131 than is direct deposition on leaves. Nevertheless, the Scottish experience indicates that marked differences in contamination on pasture can be expected within an area the size of a county.

Variation in absorbed dose from a given exposure situation attributable to differences in the behavior of individuals can be estimated from measurements made after the Chernobyl accident (Likhtarev and others 1995). Measurements of activity in the thyroid glands of exposed persons indicate that, for children of the same age and living in the same settlement, doses can range over at least 2 orders of magnitude (a factor of 100 times) (Goulko and others 1998). Some of this variation will be a function of measurement error (about 25 percent, according to Likhtarev and others 1995), but mostly it will be ascribable to differences in behavior, mainly in the amount of milk consumed. The combined effect of the two sources of variation will set large uncertainties on individual doses even where reliable estimates of contamination at the county level are available.

The NCI report indicates in two places (pp. 9.3 and 9.6) that individual thyroid doses have uncertainties within a factor of about 5 in either direction. If this assertion can be interpreted as a 95 percent confidence interval for true dose, then it corresponds to an individual dose GSD of 2.27, which seems inconsistent with the range of uncertainties given for the county estimates; the average GSD over all counties is 2.24. The near equivalence of these county and individual numbers might indicate that the county GSD actually incorporates uncertainties in individual doses. If this is true, it is not clearly stated in the report or generally reflected in the way the uncertainty analysis appears to have been performed.

The remainder of this committee's commentary parallels the structure of the NCI report. The estimation of thyroid dose due to the cows' milk pathway was broken down into the following basic steps:

  • Estimation of the total amount of I-131 released (Chapter 2).
  • Estimation of I-131 deposition (Chapter 3).
  • Estimation of concentration of I-131 in fresh cows' milk (Chapter 4).
  • Estimation of human intake of I-131 in cows' milk (Chapters 5 and 6).
  • Estimation of thyroid doses to people from the ingestion of I-131 in cows' milk (Chapter 6).


I-131 Released by Tests

Estimation of the errors inherent in the data on the deposition of I-131 on the ground necessarily begins with the reliability of the estimation of the source term, the I-131 actually released in the course of the various atomic weapons tests. One would then proceed to estimate the deposition of radioactive nuclides from the cloud stemming from each detonation and use the deposition estimates to infer individual doses. However, in the NCI report, formal estimation of the source term is limited to those tests for which no network of measurement stations was in operation. For most of the tests, the point of departure in estimating doses to individuals rests on the measured deposition of radionuclides on gummed-film at a network of stations across the United States.

The reliability of I-131 dose reconstruction is obviously greatly influenced by the accuracy of the estimate of the amount released to environmental pathways. The NCI report has prepared an excellent and thorough compilation of the events at the Nevada Test Site, including relevant configurations (air bursts, tower shots, underground tests) and explosive yields in kilotons (kt). Because the explosive power of a device is determined by the number of fissions, the number of atoms of I-131 formed and hence the activity are easily calculated. The NCI report uses a plutonium fission chain yield for mass 131 of 3.72 percent, from which is calculated an activity of 150,000 curies (Ci) of I-131 per kiloton of yield. This is a bit higher than the value of 140,000 Ci/kt calculated by this committee, which used more recent fission yield data (4.3539 percent) provided by the National Nuclear Data Center (England and Rider 1994). It is, however, an adequate and conservative value for the dose reconstruction source term for atmospheric tests. The estimation of the source terms for underground tests is much more complex because varying degrees of venting might have occurred. In this instance, the amount of I-131 released is almost solely that formed by the direct independent fission yield of I-131 rather than the cumulative mass chain yield; the nonvolatile precursors were most likely contained in the ground. Estimated releases for underground tests are, therefore, much more dependent on the air-monitoring data obtained by sampling aircraft and ground stations immediately after the test. These are believed to be good measurements, and the uncertainty in the reported data has been clearly reported.

The NCI report presents an appropriate discussion of the factors that influenced the amount of I-131 present in fallout clouds that left the Nevada Test Site, in particular the influence of the height of the burst on the amount of entrained debris that caused close-in deposition where few, if any, people were present; increased fallout in the near field offsite; and the amounts of activity entrained into the stratosphere as small particles. These variables tend to reduce the amount of radioactivity reaching the more populated downwind areas. Precipitation, or washout, dramatically influenced the deposition patterns as the fallout clouds passed over the United States, and these considerations were also addressed appropriately.


The source term for the environmental assessment is known to within approximately 10 percent because it is based on weapon yields, which were carefully measured because the yields were fundamental to each test.

Environmental levels of direct relevance to human exposures are less clear because they were modeled, and even though various components of the models are validated with environmental measurements, it is important to note that the data for each pathway to humans were generally not based on comprehensive environmental measurements made for the period of interest.

Measurement of Deposition and Estimation of I-131

Estimates of the deposition of fallout radioactivity are primarily based on data from the gummed-film monitoring network in existence from 1951 until the end of that decade (Beck and others 1990) and, for locations close to the test site, for hand-held-monitoring data. It is estimated that 85 percent of the total release of I-131 occurred while the gummed-film network was in operation. There were 40-95 monitoring stations in operation during most of the period of aboveground testing. Also, during the testing period, ground level measurements of exposure rate were made by the Public Health Service along roads and towns in southern Nevada and Utah. In the mid-1980s, these data were used to form the Town Data Base by the Offsite Radiation Exposure Review Project (ORERP), a study that reconstructed radiation doses from all radionuclides in the area within several hundred kilometers of the Nevada Test Site (Church and others 1990). For a minority of shots, representing a relatively small portion of the I-131 releases, no or few gummed-film measurements were available. For these tests, deposition outside the ORERP area was primarily estimated by meteorologic monitoring and modeling of the passage of clouds, combined with precipitation records from the days of and immediately following the shots. Available data come largely from the activities of the Health and Safety Laboratory (HASL) of the Atomic Energy Commission's New York Operations Office. The hand-held measurements were made by program-associated technicians, scientists, and engineers and should be trustworthy within the limits of the techniques used. Many of the gummed-film analyses were conducted at HASL, which was probably the best such laboratory in the world at the time.

The gummed-film measurements, however, are not really direct measurements of I-131, but rather of gross (total) beta activity. Moreover, the efficiency of the collectors (fraction of fallout measured relative to total deposited on the earth) depended on several factors. Based on an analysis described in Beck and others (1990), the deposition efficiency was assumed in the NCI report to range from 30 percent to 70 percent, depending on daily rainfall. The collection efficiency of the gummed film probably depended not only on daily rainfall but also on the size and solubility of the particles to which the I-131 was attached. The distance from ground level of the nuclear explosion also would influence particle size distribution. Tower shots, for example, would produce larger and less soluble particles than would balloon tests. For most of the soluble substances, a large part of the activity is found in standing water; thus, the extent to which standing water was or was not lost at the time of collection could have influenced the amount of activity recorded (Hoffman and others 1992).

The estimation of the total release of I-131 as well as the conversion of gross beta activity to I-131 was made possible by tables constructed and published by Hicks of Lawrence Livermore National Laboratory (Hicks 1982; 1981). These data were essential to the task of the NCI and have been used for similar purposes by other dose-reconstruction studies (Stevens and others 1992; Ng and others 1990).

Imputation of Deposition When No Direct Measurements Exist

Only a relatively limited number of gummed-film measurements were available during the period of testing, so interpolation methods were devised to give county level estimates of deposition on a test-by-test basis. The simplest of these, the area-of-influence precipitation-corrected (AIPC) method, was used for estimation when too few positive gummed-film measurements were available for the use of the more elaborate kriging algorithm described below. The AIPC method is very simple: the gummed-film measurements from the nearest stations were used in all other counties after applying a correction based on the difference in precipitation rate in the county compared with that at the location of the gummed-film measurement.

The majority of deposition estimates were interpolated from the gummed-film measurements using a mathematical technique called kriging. The real distinction between interpolating using kriging versus using the AIPC method is that kriging involves fitting parameters in a statistical model for the gummed-film measurements prior to performing the interpolation. Because kriging is the most important method used to obtain deposition estimates, much of this section is devoted to reviewing the NCI report's application of this method.

Overview of Kriging

Various substances are spatially distributed, that is they have unique values at each point in space or at each geographic location. Examples are the amount of minerals in the ground or the amount of fallout contamination deposited on the soil. Various methods can be used to estimate the amount of minerals, fallout, etc. at locations where no measurement data exist. The simplest such technique is the average of the measurement data nearby to the site of interest. More sophisticated estimation techniques are usually termed as interpolative methods. One such technique is kriging, named for D.G. Krige, a South African mining engineer and a pioneer in the application of spatial interpolation. Kriging, developed by Prof. Georges Matheron and his associates at the Institute for Mathematical Morphology in Fontainebleau, France is the process and method of estimating the local value of a spatially distributed quantity while considering the interdependence of the measurements within the region. The mathematics of this technique are designed to minimize the error in estimation by using information about the similarity (“covariance” in mathematical terms) of adjacent data points. Interested readers can consult various introductory texts on this subject (see, for example, Isaaks and Srivastava 1989).

Kriging is considered an appropriate technique for the interpolation of data measured at a small number of locations and where some degree of spatial correlation is exhibited; that is, when values at locations in close proximity are assumed to be more similar than are values at locations far apart. As discussed by Beck and others (1990), the HASL gummed-film stations provide a set of measurements that have both characteristics, and thus, kriging is a reasonable technique for estimating deposition of I-131 from the Nevada Test Site.

Kriging, in any specific setting, involves considerable modeling before interpolations are computed. The answers supplied by kriging will depend in important ways on choices made by the analyst, generally with incomplete knowledge about which choice is correct for a given situation. The modeling issues to be dealt with include:

  • The possible selection of a transformation of the measured data, typically used to improve the symmetry or normality of the overall distribution of measurements.
  • The use of additional covariates available at the data locations and at the points of interpolation to supplement or improve the interpolations.
  • The form of the function or the variogram used to describe the pattern of spatial correlation between measurements.

In practice, the variogram model is a positive linear combination of a small number of standard models that still allows for variation in the parameters of the standard models. The range of dependence is one parameter to be estimated. Having made these modeling choices, parameters in the regression function and variogram were estimated for each day. The resulting kriging estimator or predictor at a specific point is a linear function of the data involving the spatial correlation. In this sense, kriging is similar to multivariate regression analysis, but with the extra complexity involved in modeling spatial correlation.

In most uses of kriging, the kriging estimator is exact; if the data value is known at a location then the kriging estimate at that location will coincide with the data value. This property was used by the authors of the NCI report to validate the variogram model. To estimate the magnitude of the error in the interpolations, the authors sequentially deleted each location from the data set and used only the remaining data locations to estimate the value at the deleted one; the estimated values were then compared with observed values.

Kriging and the NCI Study

The kriging methods used in the NCI report used the following choices: a logarithmic transformation of all the nonzero measurements was made to improve the normality of the distribution of the gummed-film measurements. Adjustments were then made to incorporate information pertaining to precipitation in each county for the day in question. If more than one measurement of rainfall existed for a particular county, the arithmetic average was used, and if no measurements were available, assignment of a precipitation value was based on the closer measurements in the adjacent counties. It appears that residuals were computed and then kriged and added to the precipitation component, but details are not given in the report. A data-driven approach was used in modeling the spatial correlations; variograms were chosen from among several candidates to give the best fit to a single day's gummed-film measurements using an average error criterion (C. Gogolak, private communication).

The value of kriging for estimating I-131 deposition depends, first, on the quality of the measured data—here the gummed-film estimates of deposition at HASL stations—and, second, on the number and location of the sampling stations. Accordingly, the evaluation of the adequacy of the gummed-film data and the number of monitoring stations is essential to an overall appraisal of whether the task set for NCI was likely to succeed.

The principal discussion of the quality of the gummed-film measurements appears in a publication by Beck and others (1990). Table 7 in that report gives data that compare estimates of cumulative cesium-137 derived from gummed-film beta-activity data with direct measurements of cesium from analyses of soil samples. The correlation between the two estimates is extremely high (∼0.95). Data are given only for 11 stations, not all of which have complete data, but the comparison certainly adds credence to the deposition estimates based on the gummed-film data. Beck and others (1990) estimate that the errors in deposition estimates for a given site and shot are probably no worse than a factor of 2 and that uncertainties in estimates of cumulative deposition are much lower. Nevertheless, that discussion—along with the data in Table 7 of that report—indicate that, even for total deposition, the fraction of variability attributable to measurement error could still be a significant portion of total deposition variability, perhaps 50 percent, depending on how the rather vague comments about uncertainty are interpreted. Further, Beck and co-workers claim that errors in the gummed-film estimates are essentially random. Kriging can be modified to incorporate random error in the measurements themselves. This was not done by NCI collaborators, although some ad hoc adjustment to the kriging estimates of variability (see Table 3.7, NCI report) were made to account for errors in both the gummed-film measurements and the precipitation index used as a covariate in the analysis. A more systematic incorporation of measurement error into the kriging would have altered the interpolation estimates as well as the variability of the estimates. In this case, the kriging estimator would not be exact at data locations, but theoretically would improve the predictions relative to methods that ignore the measurement error. Incorporation of measurement error into the kriging generally requires outside information concerning the magnitude or variance of measurement error at each data location and can complicate assessments of prediction error.

Turning to the number and location of the HASL stations, the adequacy of the monitoring system for the production, at the county level, of reasonable estimates either of deposition from a specific shot or of total deposition, depends greatly on the basic structure of the spatial correlation among measurements. If spatial correlations are high, then few measurements will be needed to characterize deposition, even in counties far from the locations for which measurements are available. If they are low, then trying to form county-specific estimates is an exercise in futility, even though estimates over much wider areas—states, regions, or the country as a whole—can be reasonably accurate. The report gives no specifics concerning the spatial correlation pattern of the gummed-film data. Beck and others (1990) give gummed-film data for several locations for several important tests (Table 5, Beck and others 1990), so some spatial correlation is observed in the data, although no specific analysis is provided.

Magnitude of Error

Pages 3.29 and 3.30 in the NCI report present some general remarks about the magnitude of error in the estimates of deposition obtained by kriging. The report finds that the predicted values, as estimated by a variety of means, are generally accurate within about a factor of 3 and do not appear to contain any significant bias in either direction. This factor-of-3 estimate is by itself difficult to interpret. Apparently, it refers to the magnitude of error in the estimates of a single day's deposition in a specific county. It is not clear, however, whether the factor of 3 is to be regarded as a 95 percent confidence interval or as a single standard deviation. The committee assumes that it is a 95 percent confidence interval, specifically, that the GSD for a single day's county estimate is equal to 1.75, as this usage is consistent with the rest of the NCI report.

The NCI estimate of the magnitude of error is apparently a single summary statistic based on the magnitude of the prediction error from the cross-validation of the kriging estimates. The approach by which this estimate of uncertainty was obtained seems reasonable. However, throughout the NCI report there is no attempt to apportion the uncertainty estimate (GSD = 1.75) into systematic versus between-county variability, although it is likely that each type of variability exists to some degree. In general, one could expect that adjacent counties will have considerable sharing of uncertainties, so they are not independent. For counties separated by larger distances, errors in the county estimates should be independent. Without further information about the type of variogram models used—the rate of decay of the spatial correlations—it is difficult to determine from the report whether counties within the same state are correlated enough that systematic or between-county estimates, as described above, dominate.

Findings and Recommendations about Kriging

This committee finds that kriging is appropriate for the task of interpolating the gummed-film measurements. There appears to be little question about the general validity of wide-area deposition estimates, such as regional or nationwide total deposition estimates. Estimates made for local areas for which good coverage by the HASL gummed-film stations exists, such as the cities reported in Table 7 of Beck and others (1990), also are regarded as reliable. Nonetheless, it is possible that different analysts using the same general class of techniques would derive different answers by choosing different models. Although it seems unlikely that important differences, such as those in wide-area average deposition estimates, would result, estimates at the small scale could be affected considerably. For example, the explicit incorporation of random error in the gummed-film estimates at the HASL locations might lead to much less variability in the interpolation estimates at the county level.

There is insufficient information given in the report or appendices to determine whether county level deposition estimates are supportable and, in particular, whether variation at the county level in the deposition estimates represents anything but statistical noise.

Recommendation: It would be desirable for NCI to ensure that the data are preserved in electronic form and made available for possible re-analysis.

Recommendation: To the extent possible, more details should be published on the variogram modeling as well as on the actual application of kriging, and the computer codes used in the NCI report should be made available.


Figure 3.34 in the NCI report is a map of I-131 deposition estimates for all tests for all counties in the contiguous United States. Figure 4.25 of the report gives the estimates of total time-integrated concentrations in fresh cows' milk, IMC. The primary path by which cows' milk is contaminated with I-131 is ingestion of I-131 from contaminated pasture. Unlike the situation with deposition, few measurements of I-131 (see Knapp 1963; Campbell and others 1959) or total beta decay in cows' milk or in pasture were made during the testing. To offset this paucity of measurements, the approach taken to estimating the concentration of I-131 in cows' milk was to break down the transfer process into discrete steps for which experimental or observed data either exist or could be inferred based on expert knowledge.

For a given county, i, moving from deposition to milk concentrations through the pasture pathway involves the following primary estimation steps, done for each day, j, after deposition from any particular test for 60 days thereafter.

  • Estimate the initial concentration (day 0) of I-131 in pasture.
  • Estimate the remaining concentration of I-131 in pasturage on day j + t (t = 0, … , 60).
  • Estimate the amount of pasturage eaten per cow on day j + t.
  • Estimate the fraction of I-131 ingested with pasturage intake that is secreted into milk.

Initial Concentration

The parameter used to define the initial concentration in pasture is the mass interception factor F*(i,j), the ratio of the fraction of deposition retained in vegetation to the standing plant biomass (kg/m2). Multiplying F*(i,j) by deposition, DG(i,j), on day 0, gives the initial concentration, Cp(i,j,0), of I-131 retained in pasture (nCi/kg dry mass). This factor was treated separately for wet and dry deposition. Estimates of F*(i,j) for periods of dry deposition were based on a review of experimental work by Chamberlain (1970). The value of F*(i,j) for dry conditions, F*dry, used in the report depends on the value of standing crop biomass, Y, and on the distance from the Nevada Test Site. The reason F*dry is assumed to depend on distance is that distance is treated as a surrogate variable for particle size, which has been found to affect mass interception. The relationship between distance from the test site and particle size is based on a limited amount of data, which are summarized by Simon (1990).

The mass interception fraction for wet deposition was found to depend on the amount of rainfall or standing crop biomass. Empirical values were in an experimental program (Hoffman and others 1992; 1989) set up by the NCI though the NCI dose estimates in the 1997 report used an earlier formula derived by Horton (1919) for that parameter. The distribution of total standing crop biomass needed for calculating F*dry was assumed to be lognormal with a median of 0.3 kg/m2, and GSD of 1.8 (see p. 4.4, NCI report), based on data from Baes and Orton (1979).

Concentrations in Pasture on Day j + t

Radioactive decay and environmental removal processes are assumed to reduce the initial amount of radioactivity deposited on pasture grass. The NCI report takes as the combined effect of decay and environmental removal an environmental half-life, λe, of 4.5 days. Experiments by Hoffman and others (1989) found no important systematic differences in λe by physico-chemical form; that is, particle or soluble I-131, plant growth rate, or the fall of subsequent uncontaminated rain.

Amount Eaten by Dairy Cows

The NCI report notes that pasture practices of the 1950s differed from those of today, with more reliance then on outdoor grazing. Reconstruction of pasturing practice of the 1950s relied primarily on Dairy Herd Improvement Association data for several variables, including average weight of cows, milk yield, and dry-matter intake. Expert opinion was obtained to fix the beginning and end of the pasture season and the fraction of dry-matter intake attributable to pasture. Those data were developed for each of 67 pasture regions (NCI report, Figure 4.10) in the contiguous United States. The database was used to estimate the amount of pasture eaten by cows in each county on each day of interest.

Those data were modified for backyard cows by extending the pasture seasons for each pasture region and assuming that backyard cows were fed entirely on pasture during that extended season. However, this committee finds the definition of a backyard cow confusing and possibly misleading. Careful reading of the NCI report is required to understand that Category 1 milk (see below) includes milk from small farms that is consumed on the farm; this is not, in general, the same as milk from a backyard cow. In the 1950s there were just as many, or possibly many more, people living on small dairy farms as there were people who owned one or two cows. It is important to distinguish more clearly the consumption of fresh milk from cows on small dairy farms from milk that came from true backyard cows. On the dairy farms, families drank fresh milk within hours of collection, but the cows would have had standard pasturage and higher milk yield. In the 1950s in high-milk-producing states and areas (for example, in upstate New York) the relatively small dairy farm was common. As the report stands now, it would be easy for someone growing up on a small farm to estimate his or her dose erroneously because of confusion about what constitutes a backyard cow.

Intake from pasturage is converted into a daily pasture intake equivalent, PI*(i,j), which is the ratio of the total activity of I-131 in pasturage actually eaten to the time-integrated concentration in pasture at the time of deposition.

Percentage I-131 Ingested with Pasturage Intake Secreted into Milk

The cumulative fraction of I-131 after a single episode of intake that is secreted into cows' milk (on a per-liter basis) is estimated to be equal to 5 percent, with a range of 1-20 percent, and it varies not only between animals but also in the same animal at different times. The NCI report describes the transfer of I-131 from pasture to milk primarily in terms of an intake-to-milk transfer coefficient, fm, defined as the time-integrated concentration of I-131 activity in milk (nCi d per L) per unit of I-131 activity consumed (nCi). Much of the published data (summarized in Table 4.6, NCI report) give fm estimates based on dozens of feeding experiments, with each experiment reporting results from a relatively small number of animals. The NCI report (p. 4.21) assumes that fm for I-131 in Nevada Test Site fallout is distributed as lognormal random variables with a geometric mean of 4 × 10-3 d per L for any county in the contiguous United States for any time of year.

The time-integrated concentration of I-131 in milk resulting from a deposition on day j in county i is computed as

IMCp(i,j) = DG(i,j) x F*(i,j) x te x PI*(i,j) x fm

Similar calculations are provided for other exposure routes.


The basic NCI approach to estimating, for each county, the median time-integrated concentration of I-131 in milk after a given deposition is sound. Although few or no relevant measurements of I-131 contamination of milk from the Nevada tests exist, the NCI review of experimental work on fm and F*(i,j) is extensive, and much work has been done on the major components of the pasture-cow-milk pathway.

The assumption of a uniform weathering half-life of 10 days in all situations is undoubtedly a crude approximation; some evidence that λe is less than this (perhaps 3-4 days) in the spring months is provided by the International Atomic Energy Agency (IAEA 1996). The committee would expect fm to depend on particle size, with higher milk transfer rates for smaller particles, so that eastern areas distant from the Nevada Test Site might have had relatively more of the I-131 in pasturage transferred into cows' milk. This is supported by the relatively high fm (0.0061-0.0127 d per L) reported by Weiss and others (1975) and Voillequé and others (1981), working with fallout from far distant atomic weapons tests. The reconstruction of pasture practices of the 1950s is inherently less certain than are the rest of the calculations, the likelihood that they are grossly in error seems low, given the extensive effort that has gone into the process. The assumption that hay, rather than fresh pasturage, is uncontaminated, is crucial to the calculations of cows' intake of I-131. This appears to be partly confirmed by the Chernobyl experience; BIOMOVS (1991) found a factor-of-10 difference between I-131 concentration in milk from cows on and off pasture.

There is a point of confusion regarding the uncertainty analysis described at the end of Chapter 4 of the NCI report. The uncertainty analysis is designed to estimate the variability of the county average concentrations of I-131 in fresh cows' milk, not to give more-individualized estimates of variability. Because it is the variability of the county level estimates that is at issue, the method used seems to exaggerate the importance of uncertainties in the milk transfer coefficients and probably in the mass interception factors as well. This error results from the failure to differentiate between the uncertainties of averages relative to the uncertainties of individual items in the average. The report assumes that milk transfer coefficients are distributed as lognormal variables with a geometric mean, 4 × 10-3 d per L, and GSD, 2.1. This appears to describe reasonably well the data in Table 4.6 of the NCI report, but each experiment reported in this table is based on a small number of animals. However, a given county could have contained hundreds or thousands of dairy cows. The variability of average fm over these larger herds is sure to be smaller than would be the variability in fm in a single cow.

Nevertheless, there is uncertainty as well as variability between cows in the average fm that should be assigned to the animals making up the dairy herds of the 1950s. It is likely that values based on relatively recent data are not appropriate for the 1950s. It is also probable that true backyard cows would have had lower milk production and higher fm than would commercial cows of that time. A recent survey (Brown and others 1997) of 10 experts' interpretations of the likely average value of fm for future reactor accidents yielded estimates that range from 0.004 d per L to 0.011 d per L, with a wide range of uncertainty attributed to the estimates. This subjective probability assessment, although intended for reactor accidents rather than for weapons fallout, indicates that the value of fm used in the NCI report (0.004 d per L) is uncertain to a possibly important degree and that not all of this uncertainty is caused by random variability among averages of small groups of animals.

Similar remarks can be made about the mass interception factors. A single county can contain a variety of pastures, and the average F*(i,j) over these types could be much less variable than would be the mass interception factors obtained in single experiments. Although the NCI report does not explicitly incorporate uncertainty in the fraction of intake from pasture, errors in the fraction of intake from pasture could produce both systematic errors affecting large numbers of counties and between-county errors. In general, however, systematic uncertainties, after proper accounting for averaging in this part of the calculations, are likely to be small based on the extensive experimental data available.


Figure 6.2 in the NCI report is a map of estimates of integrated concentrations of I-131 in volume-weighted milk for all tests. Figure TS.3 gives the estimated time-integrated concentrations of I-131 in volume-weighted mixed milk. Going from Figure 4.25, which shows integrated concentrations of I-131 in fresh cows' milk, to Figure 6.2, involves multiplying the estimates of each county's integrated concentrations in fresh cows' milk by the estimates described in Chapter 5 of the NCI report for the fluid-use milk production of each county. The estimates in Figure TS.3 for integrated concentrations in volume-weighted mixed milk are obtained from volume-weighted fresh milk by application of milk transfer functions, VOL(i,j), which describe the transfer of milk between counties.

Milk Production

Chapter 5 of the NCI report describes the modeling of the milk production and distribution system for the 1950s, which was used to develop the VOL(i,j) values. Milk production for fluid use and size of milk surplus for each county, assumed constant over the years of interest, were estimated using U.S. Department of Commerce data, U.S. Department of Agriculture (USDA) data, and census data as follows:

  • The number of cows in each county in 1954.
  • State statistics on the amount of milk used for non human consumption on farms.
  • State statistics on the amount of milk used in manufacture of food products.
  • Per capita milk consumption in each state.
  • Population in each county.

The county level data were imputed by apportionment of state data.

The distribution of fluid milk from counties with surpluses to counties with deficits was modeled. This was done by classifying fluid milk production in each county into four categories corresponding to four population groups:

  • Category 1, those living on farms in the county where the milk is produced.
  • Category 2, those living in the county where the milk is produced but not on farms.
  • Category 3, those living in a group of neighboring counties within a designated milk region.
  • Category 4, those living at greater distances in other milk regions.

The model for the distribution system first moves milk from counties with surpluses to neighboring counties with deficits (Category 2 milk), then to other counties in the milk region with deficits (Category 3 milk), and finally from milk regions with surpluses to neighboring milk regions with deficits (Category 4 milk). Time-integrated concentrations of I-131 in milk coming out of the distribution system for human consumption are calculated based on an assignment of delay time between production and consumption of 1 day for Category 1 milk, 2 days for Category 2, 3 days for Category 3, and 4 days for Category 4. The information on which the estimates of volumes and the direction of milk flow within and between milk regions and the data on delay times between production and consumption are based is described as “qualitative.”


The largest source of uncertainties in the county milk production estimates is most likely in the apportionment of state data on the amount of milk used in manufacture to form county estimates. Errors in such apportionment add to between-county uncertainties in the size of county surpluses or to deficits in fluid milk production but would not add to systematic uncertainties because the state statistics were known.

The errors in I-131 concentration coming from incomplete information in the modeling of milk distribution would add to systematic error only if the delay times between production and consumption assumed for each category of milk are incorrect. Even the addition of 2 extra days between production and consumption in all of the categories would reduce average I-131 concentration levels only by about 15 percent.

The uncertainty analysis provided on pages 6.4-6.6 of the NCI report starts with the uncertainties in the county estimates of time-integrated concentrations of milk. These initial uncertainties, as discussed above, might be too large because they appear to neglect the effect of pooling in averaging out variations in such factors as milk transfer coefficients, fm, except for Category 1 milk, where it is more reasonable that intake depends on the production of only a few cows.

The uncertainty analysis does not explicitly deal with the effect of errors in the VOL(i,j) terms that define the milk distribution system. These would appear to add significantly to between-county uncertainty but not to overall systematic error. Given that kriging has produced estimates of deposition that are generally accurate only to within a factor of 3, much of the variability of which must be within-county rather than purely systematic, the level of detail used in the reconstruction of the milk distribution system seems unwarranted. Nevertheless it reasonably represents many broad features of milk distribution in the country.

Individual Consumption of Milk

Milk consumption is assumed in the NCI report to vary by age, race, and sex, but not by time, over the period 1950-1965. Detailed descriptions of infant, child, and adult milk consumption are based on several reports (Yang and Nelson 1986; Rupp 1980; Durbin and others 1970; Thompson 1966; PHS 1963b; 1963a). Data on milk consumption patterns at the county level are not available. Differences among regions of the country and in the degree of urbanization (city versus country) in milk consumption are used in the milk consumption model based on statistical data for 1954 (USDA 1955). Considerable data on the distribution of milk consumption (based on PHS 1963b), in addition to average milk consumption, are presented in the report and were used in the calculations.


Uncertainties in the milk consumption data exist principally because they are based on self-reports of individual or family members' consumption, for limited periods (3-7 days), and for finite (although relatively large) samples of subjects. The committee noted some inconsistencies in the self-reports. For example, USDA (1955) reported per capita milk consumption rates by region, using information collected over 1 week in 1955, that were somewhat higher than other USDA estimates based on total amount of milk for fluid use for that year. The reliance on regional statistics for use at the individual county level adds some error in individual county estimates. The likelihood of systematic error affecting regional estimates of consumption appears small, however, reported milk consumption rates were adjusted to correspond to production. The data used in estimating infants' and children's consumption of milk are necessarily somewhat less certain than are data for overall per capita consumption, but again the likelihood of important systematic errors in these data seems small.


National Cancer Institute county estimates of thyroid dose resulting from consumption of cows' milk are developed according to age and sex, which influence both the amount of milk consume which a given time-integrated concentration of I-131 in milk is converted to thyroid dose.2 Determination of a dose conversion factor involves analysis of the fraction of ingested I-131 that is taken up by the thyroid, the size and geometry of the thyroid, the average energy of beta rays from the physical decay of I-131, and the biologic half-life of I-131 in the thyroid.

Appendix 6 of the NCI report gives a summary of the available data regarding weight of the thyroid (in children, adults, and fetuses), fractional uptake of I-131 to the thyroid, the biologic half-life of I-131, and the homogeneity of the distribution of dose within the thyroid. Information about population averages of the basic determinants of dose varies from good (for thyroid weight and uptake in adults) to poor (for uptake and biologic half-life in fetuses, for which models, but few data, are available for dosimetry analysis).

The NCI reports dose conversion factors as ranging from 1.3 mrad/nCi (3.5 × 10-7 Gy/Bq) for adults to 12-15 mrad/nCi (3.2 × 10-6 to 4.1 × 10-6 Gy/Bq) for infants. The variability of such fundamental factors as the size of the thyroid (perhaps a 20-fold variation around the median values in adults and children at any given age, Figure A6.1, NCI report) is large at the individual level. Greater thyroid mass translates to lower dose (dose = energy absorbed per gram of tissue), so individual doses in response to the same intake of I-131 are potentially highly variable. (Differences in thyroid size between individuals could be offset by compensating factors in fractional uptake.) This likely interindividual variability is a source of additional skepticism concerning the factor-of-5 estimate of the variability of individual thyroid dose estimates cited in Chapter 8 of the NCI report.

The population of the United States between 1950 and 1960 was about 160 million, and the derived average thyroid dose from the I-131 produced by the collective tests (150 MCi or 5.6 × 1018 Bq) was estimated to be about 0.02 Gy (2 rad). It should be noted, however, that doses to specific persons can vary substantially about this mean, depending on age at exposure, diet, and, to a lesser extent, location at the time of the tests. To give perspective to the estimated average thyroid dose, this committee observes that standards for radiation exposure of the public evolved during the 1950s perhaps because of the presence of radioactive fallout from weapons testing in the United States and elsewhere. As developed in Appendix E of this report, the standards available during the Nevada Test Site operations were for protection of radiation workers, with limits for thyroid exposure that ranged from 15 to 30 rem per year. It became general practice in the 1950s to limit public exposure to one-tenth that for occupational workers, and on this basis exposures of 1.5 to 3 rem per year to members of the public would have been considered safe with no requirement for public intervention. In particular, children and other persons regularly consuming milk from backyard animals could have received doses that exceeded these limits. Athough the doses to most people over the age of 15 years were below the limits, it should be recognized that the limits applied to all individuals, not population averages.


It is especially important, from the point of view of assessing the overall validity of the NCI report, to confirm the collective and average dose to the American people and the extent of the uncertainty in these estimates. This committee has, therefore, developed a “top down” approach to validate NCI's “bottom up” approach. It should be stressed that this committee's approach should not be seen as a substitute for the rigorous analysis applied by NCI, but it does provide the opportunity to confirm the general correctness of NCI's primary estimate.

Only iodine that undergoes radioactive decay in the thyroid glands of U.S. citizens contributes to the collective dose. If NCI's estimate of the amount of I-131 released in the total testing period (150 MCi) and the collective dose of 4 × 108 person-rad are correct, only 1 in 1 million I-131 decays contributes to the collective dose; the remaining 999,999 decay harmlessly on the ground, in the atmosphere, in cattle, and so forth. The committee therefore assumed that the pasture-cow-milk-man pathway was the dominant route of exposure and considered 6 “filtration” steps along the route through which those I-131 decays contributing to collective dose would have to pass. The steps are as follows:

  • Step 1, the fraction of I-131 deposited on the mainland of the United States.
  • Step 2, the fraction retained on pasturage.
  • Step 3, the fraction consumed by cattle before it decayed.
  • Step 4, the fraction of total pasturage consumed by cattle.
  • Step 5, the fraction the cattle consumed and that entered the milk.
  • Step 6, the fraction in total produced milk that was consumed.

By then multiplying the amount of residual activity that remains after Step 6 by the total released activity, and converting that to dose, the collective dose of U.S. citizens is estimated. For example, of the amount released, 25 percent was estimated to be deposited in the 48 contiguous states and 75 percent either decayed in the atmosphere, was carried out of the contiguous states, was deposited on nonpasture land, or was otherwise filtered out of the chain. At Step 2, of the 25 percent deposited, the fraction intercepted by the pasture is estimated. This process continues through the remaining steps. Full details of the calculation are given in Appendix C.

At the completion of this “chain of filters” the estimated collective dose is 8 × 108 person-rad, twice the value determined by NCI. Given the rough nature of this committee's calculations, this is considered by the committee to be good agreement and should provide confidence that the NCI estimate is not grossly under or over the actual value.

This committee is, however, concerned that NCI's estimate of the uncertainty in the collective dose is too low. The only discussion to date of this figure has been in the material presented to the committee by Dr. Charles Land of NCI. The material states that the GSD for the collective dose estimate is 1.4, so the per person thyroid dose could vary by a factor of 2 in each direction (that is, the average individual doses lie between 0.01 and 0.04 Gy). Based on estimates of the uncertainty in each step in the filtration chain above, the 5 percent and 95 percent confidence limits for the collective dose are 5 × 107 and 3 × 109. This would give the range for average individual doses as 0.0025-0.15 Gy (0.25-15 rad). In fact, to arrive at an uncertainty factor of 2 in either direction, using the simplified model above, would require an uncertainty of 30 percent at each step in the filtration chain. Given the overall effect of this filtration process, the committee is concerned that the NCI estimate of uncertainty is be underestimated.

The range of likely collective dose estimates implies that there is a probability (see Chapter 3 of this report) that substantial elevations in thyroid cancer incidence were produced by the Nevada tests for large groups of Americans. At the same time, there is a corresponding probability that the NCI report's calculation is too large and that the increase in thyroid cancer incidence is quite small. Because of the sensitivity of the thyroid to radiation, epidemiologic methods (the study of observed thyroid cancer rates) could in fact provide an important indication of the size of the collective dose, and such analyses can and should be used to reduce the uncertainties in the collective dose estimate.

This committee, by independently validating the “order of magnitude” of the NCI's estimate of collective dose, is confident that there has been no gross over- or underestimation, but it is less confident that the uncertainty in the estimate has been realistically determined by NCI. It should, however, be noted that within these collective and average doses are concealed large differences in individual doses that depend on factors such as lifestyle and age at the time of testing. In this aspect lie important public-health implications of the Nevada atomic weapons tests.

Variation of Estimated Doses by Geographic Location and by Year of Birth

Considerable variation of possible doses is obvious from inspection of the data from the NCI report at a selection of locations across the country. Table 2.1 provides a summary of thyroid doses for 20 cities across the United States for four different consumption scenarios. In addition to the wide variation, it is obvious that some northeast locations have predicted doses similar in magnitude to those in the mountain states.

TABLE 2.1. Thyroid Dose (cGy or Rad) for an Individual Born on January 1, 1952.


Thyroid Dose (cGy or Rad) for an Individual Born on January 1, 1952.

Considerable variation in dose is also predicted as a consequence of date of birth for a single location (Denver, CO is used as an example). Table 2.2 gives the estimated dose for individuals born in 5 year increments from 1937 to 1962. The date of birth resulting in greatest exposure is about 1 January 1952.

TABLE 2.2. Variations in Dose as a Consequence of Date of Birth.


Variations in Dose as a Consequence of Date of Birth.

This table shows the relative difference in dose due to the age at time of testing. For example, an individual born in 1932 would have received 9 percent of the dose of someone born in 1952; someone born in 1947 would have received about one half that dose.


The NCI report presents a comprehensive rationale for assuming that significant thyroid doses were experienced by many in the U.S. population, particularly for the youngest birth cohorts, as a result of the fallout of I-131 from the nuclear weapons testing program. The overall report is thorough, but complicated, and complexity is the price paid for the amount of detail in the results presented, for example, in the estimates for representative age groups and for milk intake patterns in each county within the continental United States. Between-county and systematic uncertainties are important in determining the public-health significance of the estimates. Although it is recognized—both in the report and in the material made available on the NCI Web site—that uncertainties in county dose estimates are important, no clear statements in the report distinguish between systematic and between-county uncertainties. It is likely that within-county uncertainties produce a significant portion of overall uncertainty in thyroid dose, which means that the level of detail presented (county estimates, rather than estimates over much broader regions) is probably inappropriate.



To find information on the 1997 NCI report, it is necessary to follow on-screen links at the NCI Web site to “What Was New in 1997?”


The units of mrad/nCi as used in the NCI report are conventional units; the same quantity couldalso be expressed in international units of gray per bequerel (Gy/Bq) though the numerical valuewould be different (1 mrad/nCi = 2.7 × 10−7 Gy/Bq).

Copyright © 1999, National Academy of Sciences.
Bookshelf ID: NBK100834


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (6.6M)
  • Disable Glossary Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...