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.
7 Comparison with Other Studies CONCORDANCE WITH THE UTAH STUDY OF THYROID DISEASE FROM NEVADA TEST SITE FALLOUT Section TX of the Draft Final Report describes in some detail whether the HTDS study confirms the results of the Utah study of exposure to i3~l from the NTS (Kerber and others, 1993~. With a generally negative dose-response relationship, especially for thyroid carcinoma, the HTDS cannot be regarded as confirming the Utah findings of increased risk of thyroid neoplasia. However, another aspect of the comparison of the two studies is not fully treated: the degree to which the results of the HTDS directly contradict those of the Utah study. One approach to answering that question is to assess the degree to which confidence intervals of risk estimates from the two studies overlap. Even if one study's results are positive and another is negative, it does not mean that they necessarily are irreconcilable. If the positive study is barely positive (p ~ 0.05) and the negative study has wide confidence intervals, there might be no fundamental disagreement. For the HTDS analysis of thyroid carcinoma, the estimate and the confidence interval for the linear slope term for thyroid carcinoma are not reported, because the maximum- likelihood estimates failed to converge. We can, however, work backward from other information in the report to estimate the standard error of the slope term. The attained power to detect a
I12 Review of the HTDS Draft Final Report slope term of 2.5% per Gy is states} to be 0.96 (section VITT), so we will have (at least approximately) 0.025/(standard error of B) = (A 0.05 - ZI-0.96) = 3 396 Therefore, the standard error of slope term B must have been 0.007 per Gy. To approximate the value of the estimate of B. we note that a logistic mode! did converge and gave an estimated slope that was about ~ standard error below zero. We can assume roughly that the linear slope estimate would also have been about ~ standard error less than zero, or about -0.007 per Gy. If (as hypothesized) males had a background risk of 0.004, the upper confidence limit for the risk at ~ Gy is -0.007 + 2~0.007) = 0.007, so the upper limit of the excess relative risk (ERR) at ~ Gy is 1.75 for males. For females, assuming a background of 0.007, the ERR at ~ Gy is I, so the average of the two is ERR = ~ .375 per Gy. For the Utah study, the estimate was 7.7 with a lower 95% confidence limit of 0.74 per Gy. It seems, then, that the confidence intervals for the risk of thyroid cancer overlap to some degree. Moreover, on the basis of the considerations above, it is evident that the confidence intervals for the HTDS in fact depend on dosimetry- error assumptions; if the pure Berkson mode} of errors in the dosimetry does not hold, the confidence intervals for the HTDS could be considerably wider. Thus, there does not appear to be a fundamental incompatibility between the two studies. CONCORDANCE WITH STUDIES OF EXTERNAL RAI)IATION TO THE THYROID It may be an oversimplification to say that the HTDS, because it found no significant dose-response relationship for any disease end point, is in direct contradiction with the cohort studies of external radiation exposure and risk of thyroid cancer. Of the five cohort studies of external radiation and thyroid cancer that were analyzed by Ron and others (1995), one yielded estimated dose-response relationships considerably stronger than the others.
Comparison with Other Studies ~3 To combine the results of the five cohort studies, Ron and others used a random-effects model. The dose-response relationship was allowed to vary from study to study, and the average dose-response relationship for a hypothetical population of studies was estimated. The average estimate was equivalent to an ERR of 7.7 times the age-specific baseline per Gy with a confidence interval of 2. I-28.7 times the baseline. As described above, the HTDS is probably consistent with an upper ERR of about T.4 per Gy, which is not statistically compatible with the estimate for external radiation. It is not known whether the two estimates could be statistically compatible if uncertainties in dosimetry were factored into the confidence interval for the HTDS. COMPARISON WITH CHERNOBYL STUDIES The effectiveness of 13~} in causing thyroid cancer has been shown by the Chernobyl experience. The first increases, reported in ~ 992, of thyroid cancer attributed to the accident were challenged as possibly the result of intensive screening. More recently (Astakhova and others, 1998), however, a case-control study in Belarus has found highly significant differences between cases and controls in estimated ]3~} dose to the thyroid, even when controls with similar presenting complaints or screening circumstances were selected. But the durations of exposures were shorter for Chernobyl than for the Hanford downwinders, and the doses were higher, so the dose-rate issue still is unresolved with respect to the epidemiologic data. Furthermore, the dose reconstruction for the study in Belarus was based on actual measurements of ground deposition of ]3~{ and cesium-137, a data bank of 1986 thyroid-radiation measurements, and interviews and questionnaires. Therefore, doses were probably better estimated for individuals in the Belarus study than in the HTDS.