Model fit characteristics by outcome, exposure (preferred lag or window), and model type
| Outcome Adjust for | Non-CLL (n=264) race, sex and hire date | AML (n=150) birth date | CML (n=52) race, sex and hire date | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Model | Lag (2 years) | Windows (6–14 years) | Lag (1.1 years) | Windows (6–14 years) | Lag (6.3 years) | Windows (6.3–14 years) | ||||||
| Fit statistic* | AIC | LRT p value | AIC | LRT p value | AIC | LRT p value | AIC | LRT p value | AIC | LRT p value | AIC | LRT p value |
| Linear | 850.9 | NA | 847.6 | NA | 480.9 | NA | 473.7 | NA | 168.0 | NA | 166.2 | NA |
| Linear-quadratic | 852.1 | 0.382 | 846.9 | 0.104 | 481.7 | 0.281 | NC | NC | 170.0 | 0.962 | NC | NC |
| RCS | 844.7 | 0.006 | 845.6 | 0.046 | 479.4 | 0.065 | 473.5 | 0.136 | 170.6 | 0.488 | NC | NC |
| PLS† | 846.2 | 0.013 | 847.6 | 0.135 | NM | NM | NM | NM | ||||
| Categorical | 852.2 | NA | 849.8 | NA | 486.1 | NA | 477.7 | NA | 170.9 | NA | 169.6 | NA |
*LRT degrees of freedom (df): Linear-quadratic all=1; RCS lag=2, window=1; and PLS lag=4, window=2.
†Best fitting PLS models were obtained using two knots (8.3 and 46.9 mGy) in the 2-year-lagged model and one knot (20.1 mGy) in the windows-based model.
AIC, Akaike Information Criterion; AML, acute myeloid leukaemia; CML, chronic myeloid leukaemia; CLL, chronic lymphocytic leukaemia; LRT, likelihood ratio test; NA, not applicable; NC, not calculable; NM, not modelled; PLS, piecewise linear spline; RCS, restricted cubic spline.