would be needed. Such an effort would easily be on the scale or greater of that which NCI accomplished in support of its 1997 report. This ought to only be considered as a research recommendation if the government or public believes attaining a fuller understanding of nuclear testing to be an issue that supercedes other more basic health issues.
The report of the National Cancer Institute (NCI 1997a), together with the presentation material provided by Charles Land (Appendix B, this report), gives a theoretical basis for positing that nuclear-weapons testing at the Nevada Test Site led to exposures to the thyroid that resulted in significant excess risk of thyroid disease for some members of the population of the United States, depending on age, geography, and individual consumption and source of milk. As indicated in Land's analyses, and as further discussed in Chapters 2 and 3 of this report, there are still important uncertainties in the assessment of the size of the exposures and their influence on thyroid cancer risk. No epidemiologic evidence was presented either in the NCI (1997a) report or in concomitant releases of information that thyroid cancer risk differs significantly by birth cohort, geography, or behavior as a consequence of the iodine-131 fallout from weapons testing. It would be premature in this committee's view to initiate changes in medical practice based solely on the theoretical calculations unless there is also empirical evidence of an increased risk. Accordingly, a comprehensive study of past incidence of thyroid disease—using the full range of resources from existing tumor registries, including but not restricted to those associated with the Surveillance, Epidemiology, and End Results (SEER) program—is clearly important and essential to more realistic risk estimates. Such studies have the statistical power to detect increases in risk of the size predicted by Land, by birth cohort, and possibly by geography.
Studies of current incidence (case-control studies) are feasible and would be useful in evaluating whether behavioral factors such as milk consumption in the 1950s are indeed related to individual risk. For example, a study of 500-1,000 cases matched to an equal number of controls would have good statistical power to detect increases due to behavioral factors (milk consumption), of 50 percent or more. Such a difference in risk would correspond to a difference of 0.08 Gy in dose using the Ron and others (1995) risk estimates for those exposed as very young children. The most heavily exposed birth cohort (persons exposed between the ages of 0 and 4 in 1952) consists of approximately 18 million people. At an average current rate of thyroid cancer of 11 per 100,000 person-years (assuming a 40 percent increased risk due to exposure), this entire cohort would produce about 2,000 cases of thyroid cancer in a year. Case-control studies, however, are likely to be subject to recall biases—especially given the degree of publicity the release of the NCI report received. Nevertheless, case-control studies would appear to be the only reasonable way of empirically relating behavior to risk.