I was honored to be asked to contribute a chapter to the collected works within Tortured Science: Health Studies, Ethics and Nuclear Weapons in the United States, compiled and edited by Dianne Quigley, Amy Lowman, and Steve Wing, published 2012 by Baywood Publishing Company, Inc, Amityville, New York. I am very grateful that Baywood has allowed me to reprint the chapter in its entirety. The chapter is reprinted as a series of excerpts.
Excerpt #10
Strength of the HTDS Findings Overstated
The subcommittee is concerned that the results of the study were reported— and interpreted—in black and white terms of whether a statistical test was passed or failed. It recommends that confidence limits be provided through- out the report to allow the readers to judge how large a radiation effect might be consistent with the data. It feels that the HTDS investigators probably overstated the strength of their finding that there was no radiation effect [60, p. 81].
The expert technical review of the HTDS commended HTDS authors for responding to the NRC Subcommittee’s recommendation to include confidence intervals [59]. Yet, the final report failed to make full use of the confidence intervals in interpreting study results. Had confidence intervals been used as counterpoint to HTDS authors’ reliance on statistical significance in drawing conclusions, HTDS authors’ results would have demonstrated inconclusiveness [59, p. 18].
Correct interpretation of lack of significance in the HTDS report
. . . is simply that, when examined using the models and methods of the authors, the data do not overwhelmingly favor any alternative over the null. This correct interpretation leaves open the possibility that the evidence favors the alternative (that there is an effect) albeit not very strongly when using the authors’ approach [59, p. 18].
HTDS researchers’ statements such as “the results of the HTDS provide no evidence of a statistically significant association” conflate absence of statistical significance with absence of evidence [59, p. 18; 72]. The two concepts must be separated, at which point, it can be seen that there is a lack of significance, but that some of the evidence provides weak support for the possibility of small effects [59, p. 18].
The problems that have been discussed regarding power and dosimetry uncertainty in HTDS require more thorough analysis, which would have had a “serious impact on the interpretation of the study, leading to even more ambiguous results, which in turn should lead to even more cautious and limited interpretation” [59, p. 19].
HTDS: Consistent With Other I-131 Studies Showing Risk
The HTDS final report stated, “There is little evidence in the literature to suggest that people exposed to I-131 at the levels found in this study over a period of months or years would experience higher rates of thyroid or parathyroid disease as a result of their exposure” [59, p. 19; 71, p. 543].
Technical reviewers challenged this statement and found no studies published in the scientific literature that study protracted exposures at the levels found within the HTDS. The reviewers concluded that “The reason that there is little evidence is due to the absence of evidence, not to the existence of studies that actually show no risks from protracted exposures” [59, p. 20]. Furthermore, the technical reviewers concluded that
The results and conclusions of the Final Report of the Hanford Thyroid Disease Study (HTDS) . . . cannot be used to rule out important risks for thyroid cancer, neoplasms, or hypothyroidism from exposures to iodine-131 (I-131) from the Hanford nuclear facility [73].
Considering the HTDS limitations in measurement and resulting uncertainties, expert review of the study found that even though HTDS findings did not show statistically significant elevations in risk, HTDS results are not inconsistent with other published studies supporting risks for certain thyroid diseases from I-131 exposures, if the upper bounds of the reported confidence intervals are considered [74].
HTDS authors used three different approaches to analyze whether their results were inconclusive because of dose uncertainties in the primary analyses [71, p. 603]. They found similar results in each of their analyses, which the technical review found to be consistent with low statistical power in each of the three approaches, rather than consistent with a hypothesis that there is no relation between exposure and disease [59, p. 20].
The final sentences of the HTDS report state:
These findings do not definitively rule out the possibility that Hanford radiation exposures are associated with an increase in one or more of the outcomes under investigation. However, it does mean that if such associa- tions exist, they were likely too small to detect using the best epidemiologic methods available [72, p. 21].
HTDS findings do not rule out the possibility that Hanford I-131 exposures are associated with an increase in thyroid health outcomes [75]. HTDS findings are compatible with significant increase in health risk from these exposures as well as with no increase. “Even the best epidemiologic methods are not enough to compensate for a study population that is too small and measurements too uncertain to detect even large risks” [59, p. 21].
Notes
59. See Ruttenber, A. J., et al. 30 March 2004. A technical review of the final report of the Hanford Thyroid Disease Study. p. 3, citing Hoffman 1991, Hoffman et al., 1993, Hoffman et al., 1996, Hoffman 1999. Prepared as an expert report for the Hanford litigation, to respond to suggestions and recommendations of the NRC Subcom- mittee’s 2002 Hanford Thyroid Disease Study Draft Final Report. See also Hoffman, F. O., Ruttenber, A. J., Apostoaei, A. L., Carroll, R. J., and Greenland, S. 2007. The Hanford Thyroid Disease Study: An alternative view of the findings. Health Physics Journal 92(2):99-111.
a. HTDS results do not show statistically significant elevations in risk but are not inconsistent with other published studies supporting risks for certain thyroid diseases from I-131 exposures, if the upper bounds of the calculated confidence intervals are considered (as pointed out for the situation of thyroid cancer and I-131 exposure from NTS fallout by the NRC). [emphasis added]
b. Applying a strict interpretation of the lack of statistical significance found in the results of HTDS, we find that the results of HTDS are consistent with the finding of thyroid risk in some but not all studies of I-131 exposures, but not consistent with those studies which show elevated thyroid risk, or;
c. If the results are subjected to strict interpretation based upon statistical signifi- cance, then HTDS is not inconsistent with other studies which show that chronic exposure to I-131 is not associated with thyroid disease.
60. NAS-BRER. 2000. Review of the Hanford Thyroid Disease Study draft final report.
Washington, DC: National Academy Press, p. 7.
66. Greenland, S. 1977. Response and follow-up bias in cohort studies. American Journal of Epidemiology 106(3):184-187.
67. Fred Hutchinson Cancer Research Center/CDC. 1998. Preliminary technical review of the Hanford Thyroid Disease Study draft final report, p. 6.
68. Greene, G. 1999. The woman who knew too much: Alice Stewart and the secrets of radiation. Ann Arbor, MI: University of Michigan Press, p. 216.
69. The incidence of thyroid cancer found within HTDS was very close to that which was predicted by the medical monitoring program plan. See Spengler, R. F. July
1997. Hanford medical monitoring program: Background consideration document and ATSDR decision. US DHHS, ATSDR.
70. HTDS Newsletter, February 1997.
71. Davis, S., Kopecky, K. J, and Hamilton, T. E. 1999. Hanford Thyroid Disease Study final report, p. 326, 370.
72. Davis, S., and Kopecky, K. 1998. HTDS final report, summary and conclusions, p. 5
73. Ruttenber, A. J., et al. 30 March 2004. A technical review of the final report of the Hanford Thyroid Disease Study. This technical report was written in support of Hanford plaintiffs in litigation. The report as such is not independent, but it has no relationship to the authors of HTDS nor CDC.
74. This was the most supportable interpretation, particularly because of the HTDS overestimates of statistical power and the incomplete characterization of the effects of dose uncertainty.
75. Considerably more goiter and other thyroid disease was reported in the R-11
Hanford Survey than in national survey data. See Report of R-11 survey results.
14 November 1995. JSI Center for Environmental Health Studies. See also Grossman, C. M., Morton, W. E., and Nussbaum, R. H. 1997. Malignancies among Hanford downwinders. Presented at the Health of the Hanford Site Conference, 3 December
1997.
Leave a Reply