Low dose radiation and cancer in A-bomb survivors: latency and non-linear dose-response in the 1950–90 mortality cohort

Table 2 Transient model^a

^a0 – 20 mSv subcohort, Excess Relative Risk ERR = σD_φ $e^{(- τ D_{φ})}$ with τ ≥ 0 and σ,τ constrained by 1 + ERR ≥ 0 for all data cells; for D_φ see Table 1. The linear model is nested by τ = 0; the control model is nested by σ = 0.
^bchosen in the range 5 – 44 years. to minimise transient model Deviance; where italicised φ_m minimises Dev subject to ERR ≥ 0, equivalent to σ ≥ 0.
^cAt φ_m, σ and τ are the fitted Maximum Likelihood Estimate parameter values.
^dAt φ_m, the likelihood ratio test LRT_trans-lin compares the transient with the linear model and LRT_trans-con compares the transient model with the control model. Each LRT is χ² distributed on 1 degree of freedom.
^eAt φ_m, ERR_D,φ is the Excess Relative Risk at D_φ = D, with 95% Profile Likelihood Confidence Interval.
^fAt φ_m, Dmax = 1/τ is the value of D which maximises ERR_D,φ
^gn/a indicates no lower limit could be calculated with 1 + ERR ≥ 0
Other definitions and total cases as Table 1.