Derivation of low-dose extrapolation factors from analysis of curvature in the cancer incidence dose response in Japanese atomic bomb survivors.

PURPOSES To assess the degree of overestimation in low-dose cancer risk when models linear in dose are fitted to cancer data. MATERIALS AND METHODS Examination of the quadratic and the linear coefficients (in dose) in the Japanese atomic bomb survivor cancer incidence data, based on follow-up to 1987 and taking separate account of random errors in DS86 neutron and DS86 gamma dose estimates and systematic errors in Hiroshima DS86 neutron dose estimates. RESULTS When the 0-4Gy dose range is used, the low-dose extrapolation factor for all solid tumours, assessing the degree to which low-dose cancer risks are overestimated by fitting a model linear in dose, is 1.06 (95% CI 0.78, 1.62) and so is not significantly different from 1; the associated ratio of the quadratic to the linear coefficients is 0.06 Sv(-1) (95% CI -0.22, 0.67). The best estimate of the low-dose extrapolation factor for leukaemia is 2.47 (95% CI 1.24, > 1,000). The ratio of the quadratic to the linear coefficients is 1.81 Sv(-1) (95% CI 0.21, > 1,000). When various types and groupings of solid tumours are considered there is not generally any strong evidence for upward curvature; for only two out of the six solid tumour groupings are there indications of appreciable upward curvature in the dose response, and in no instance is this statistically significant. Consideration of a lower dose range (0-2Gy rather than 0-4 Gy) results in the low-dose extrapolation factor for solid tumours increasing to 1.21 (95% CI 0.81, 2.45), with corresponding increases for solid tumour subtypes; the corresponding quantity for leukaemia decreases to 1.73 (95% CI 0.79, 147.67). Three out of the six solid tumour subtypes now show appreciable upward curvature. If there is additional adjustment of the Hiroshima neutron dose estimates over the 0-2 Gy dose range, the low-dose extrapolation factor for all solid tumours increases still further, to 1.43 (95% CI 0.97, 2.72), whereas for leukaemia this quantity is further reduced, to 1.58 (95% CI 0.90, 10.58). CONCLUSIONS There is marked upward curvature in the dose response for leukaemia. For solid tumours there is little indication of upward curvature in the dose response over the 0-4 Gy dose range, although over the 0-2 Gy dose range and after adjustment of Hiroshima DS86 neutron dose estimates the upward curvature is more pronounced. Uncertainties in the likely adjustments to the Hiroshima DS86 neutron dose estimates imply that these findings should be treated with caution.