To further evaluate the effect of the dose adjustment procedure at each facility on the ERR dose-response coefficients, a more detailed analysis was done using two year, ten year, and twenty year lag intervals for all cancer causes. Facility differences in the impact of the dose adjustment procedures are of interest because external radiation monitoring differed in coverage and other respects at X-10 and Y-12 before 1961. Most dose adjustment at Y-12 resulted from the replacement of missing values for unmonitored workers with one of three assumed values, whereas at X-10 most workers were monitored for external radiation after 1947 and individual yearly doses for each worker were adjusted using the preliminary ad hoc procedure [2,3]. In the analysis with follow-up through 1984, the cumulative adjusted dose assignments for the twenty year lag are based almost entirely on annual doses obtained during the period when the adjustments were made.
The first row of panel A in Table X shows the estimated ERR per Sv for each lag using the unadjusted doses. The second line gives the value of the LRT statistic comparing the estimated ERR to the assumption of zero slope. Lines three, four and five show the estimated ERRs for each facility separately, and the final line is the LRT for facility differences in slope compared to the assumption of a common slope. Panel B gives the same results for adjusted doses. The estimated ERRs for all X-10 and Y-12 workers combined based on unadjusted doses are approximately 50% larger than those based on adjusted doses, and the largest ERR and LRT values are observed for the ten year lag.
The lower lines of each panel show that the change in ERR with increasing lag differs between X-10 workers and the other groups. Estimated ERRs at X-10 increase with increasing lag for both unadjusted and adjusted doses, while estimates for Y-12 and multiple facility workers are positive for two and ten year lags but negative for the twenty year lag, the period when the majority of Y-12 workers were not monitored for external radiation and adjusted annual doses for Y-12 workers took on one of three assigned values. The LRT for facility differences in ERR increases with increasing lag. The difference between ERRs based on adjusted and unadjusted doses is proportionately larger for Y-12 than for X-10 workers. Unadjusted values for multiple facility workers are closest to zero and change the least with adjustment assumptions.