next up previous
Next: Results of the Up: Dose-Response Analyses. Previous: Dose-Response Analyses.

Selected Results for Main Effects Model Using ERR.

The marginal distribution (dose group by facility) of person years, observed and expected cancer death, and average dose is given in Appendix Table AIV. The parameter estimates for the ERR main effects model (see Eq. 4) with external rates and using unadjusted doses with a ten year lag are given in Tables VI and VII. A detailed description of the data analysis procedure and example of how to interpret the parameter estimates is given in the Appendix. The referent group is nonmonthly workers employed at least one year at the X-10 facility who were eligible but not monitored for internal radiation exposure and had zero external occupational dose. The parameter estimates for the factor B (rows 2-6 of Tables VI and VII) are estimates of the log SMR for the referent group for each birth cohort for each cause of death. The large negative values for all causes of death (column 2 of Table VI) indicate a strong ''healthy worker" selection effect for the X-10 facility, and that the effect is strongest among workers born after 1930 (SMR = exp [-0.63] = 0.53). Review of the birth cohort effect estimates shows a similar pattern for diseases of the circulatory system and external causes of death. For all cancer causes lower rates were observed primarily for those born after 1920 (see column 4 of Table VI). By contrast, the birth cohort effect estimates for selected cancer causes in Table VII show a much different pattern. The birth cohort effect estimates for those born before 1930 are large and positive for cancer of the prostate and leukemia, indicating that for the internal referent group these death rates are higher that those for U. S. white males.

The SES effect estimates (row 8 of Table VI) are large and negative indicating that death rates are substantially lower for monthly workers than for nonmonthly workers. For example, for all cancer the relative risk for monthly versus nonmonthly workers is exp(-.419) = 0.66 with (95% CI: 0.56, 0.78). The SES effect is most pronounced for lung cancer where the relative risk for monthly workers is 0.43 with (95% CI: 0.31, 0.59).

The log relative risk for all cause mortality for short term workers is 9.9L%, mostly due to higher rates for external causes. Differences in mortality rates among the three levels of the internal radiation risk group (IG) were unstable and in varying directions (see lines 11-13 of Tables VI and VII). LRT statistics (not shown in the table) for the null hypothesis of no difference among the levels of this factor were all below the 95 percentile of the chi-square distribution with 2 df.

There are differences in the death rates in the facility groups in the X-10/Y-12 subcohort. All cause death rates for those employed only at Y-12 are 12.4L% higher than X-10 only workers, and multiple facility workers are higher by 8.4L% (see lines 14-16 of Tables VI and VII). These differences appear to be primarily due to higher lung cancer rates for the Y-12 workers (log relative risk = 46L%) and multiple facility workers (log relative risk = 33L%). A notable exception occurs for leukemia where the rates are substantially lower for the Y-12 (-121.4L%) and multiple facility (-64.0L%) workers than for X-10 workers.

The last lines in Tables VI and VII provide an estimate of and 95% CI for the ERR per Sv. The estimated dose-response relations are represented graphically in Figs. 2 and 3 ( see the Next Section for additional discussion of Figs 2A-D and Figs 3A-D ). The average lifetime dose for X-10/Y-12 subcohort is about 10 mSv ( see Fig.4 ). This coefficient can be used to compute the relative risk of the ''average" worker in the subcohort of Oak Ridge workers with potential for exposure to external penetrating radiation. This estimate can be compared with the other estimates of the log relative risk parameters for each cause of death. Using all cancer as an example( see Fig. 2A ), the log relative risk for the average worker is about 100*log(1+ 0.0145)= 1.44L%. By contrast, the effect of being a nonmonthly worker is 41.4L %, suggesting that SES factors are far more influential than external radiation exposure in determining cancer risk in this group of workers.



next up previous
Next: Results of the Up: Dose-Response Analyses. Previous: Dose-Response Analyses.