For the ``external/internal'' analysis we have
where is the SMR for the stratum and A = (age - 52.5)/100.
The main differences between the internal and external analysis are in birth cohort and age terms. The estimates for referent factor B (birth cohort) in column 4 lines 2 through 6 of Table AV represent the SMR for each birth cohort in L% units, i.e. exp(-7.7/100) = 0.919 is the estimated SMR for the 1915 L% birth cohort at the reference level of each of the other factors. It is easy to interpret the levels of the referent factor. These estimates (-9.2,-4.3,-10.1,-23.0,-66.2) show, for example, that the all cancer mortality rate for X-10 only nonmonthly long term workers that were eligible but not monitored for internal radiation exposure are less than the U. S. white male rates, and that the deficit is larger for younger workers, i.e. the more recent birth cohorts. The parameter estimates for the factors S, L, IG, and F (in column 4) have the same interpretation as in the internal analysis, and are almost identical in numerical value to the corresponding estimates from the internal analysis in column 2. The estimate for the age term in column 4 describes (and adjusts for) any systematic age-related difference in the external rates and the study cohort in percent per year units.
The relative risk estimates for the exposure variables IG, F, and D are adjusted for effects of the confounding variables age, B, S, and L. Their values are readily combined to obtain an estimate of the SMR for any combination of factor levels and dose. Consider four individuals in the 1915 birth cohort, who were long-term Y-12 only workers and were monitored for internal radiation exposure. Further, suppose that the first two were nonmonthly, and that the second two were monthly, and that two of them received cumulative external radiation doses of 0.01 Sv (1 rem). Using the parameter estimates from column 4 of Table AV we obtain:
This illustrates how to read the summary tables given in the Results Section of this report. Note in particular that the coefficients for the birth cohort factor represent a comparison (an SMR in L% units) of the internal control group with the external referent group. The coefficients for the potential confounding variables (S, L) and the exposure variables of (IG and F)are relative risk estimates (in L% units). The coefficient for D is the estimated ERR per Sv which is equivalent to percent per 10 mSv (rem).