Heterogeneity of variation of relative risk by age at exposure in the Japanese atomic bomb survivors.

General reductions in cancer relative risk with increasing age at exposure are observed in the Japanese atomic bomb survivors and in other groups. However, there has been little evidence of heterogeneity in such trends by cancer type within the Japanese cohort, nor for cancer-type variations in other factors (sex, attained age) that modify relative risk. A recent report on the Japanese atomic bomb survivors published by Preston et al. in 2007 suggests that solid cancer relative risk exhibits a U-shaped relationship with age at exposure, and is initially decreasing and then increasing at older exposure ages. In this report, we reanalyse the latest Japanese atomic bomb survivor solid cancer mortality and incidence data analysed by Preston and co-workers, stratifying by cancer subtype where possible, the stratification being both in relation to the baseline and the radiation-associated excess. We find highly statistically significant (P < 0.001) variations of relative risk by cancer type, and statistically significant variations by cancer type in the adjustments for sex (P = 0.010) and age at exposure (P = 0.013) to the relative risk. There is no statistically significant (P > 0.2) variation by cancer type in the adjustment of relative risk for attained age. Although, for all incident solid cancers, there is marginally statistically significant (P = 0.033) variation of relative risk with a quadratic log-linear function of age at exposure, there is much weaker variation in the relative risk of solid cancer mortality (P > 0.1). However, the manner in which relative risk varies with age at exposure is qualitatively similar for incidence and mortality, so one should not make too much of these differences between the two datasets. Stratification by solid cancer type slightly weakens the evidence for quadratic variation in relative risk by age at exposure (P = 0.060).