borderline significantp = 0.279
In general linear models fit better than the log-linear model, as shown by the AIC statistics (Table 2 , Additional file 1 : Table S2). Table 3 demonstrates that there is at best borderline significant variation of excess absolute risk by time since exposure ( p = 0.279) or with age at exposure ( p = 0.085); this contrasts with more pronounced and significant variation of relative risk both with time since exposure ( p = 0.034), and with age at exposure ( p = 0.012) with relative risk particularly large for the period 10–14 years after UVR exposure and for exposures under the age of 25 (Table 4 ).