Table 1.
Mean OS estimated from each survival model
| Model | Section | AIC | BIC | AUC (K-M 2.25; years) |
AUC (K-M new 3.5; years) | AUC K-M + (years) |
AUC (lifetime; years) |
|---|---|---|---|---|---|---|---|
| K-M, 2.25 years [95% CI] | 1.05 [0.96–1.14] | ||||||
| K-M new DCO, 3.5 years [95% CI] | 1.08 [1.00–1.16] | 1.33 [1.21–1.45] | |||||
| Minimum across all fitted models | 1.03 | 1.14 | 0.05 | 1.19 | |||
| Maximum across all fitted models | 1.11 | 1.37 | 9.9 | 11.22 | |||
| Weibull | 3.3 | 3793 | 3801 | 1.03 | 1.14 | 0.05 | 1.19 |
| Exponential | 3791 | 3795 | 1.03 | 1.15 | 0.07 | 1.22 | |
| Gompertz | 3792 | 3800 | 1.04 | 1.20 | 0.38 | 1.59 | |
| Generalised gamma | 3779 | 3792 | 1.05 | 1.25 | 0.58 | 1.83 | |
| Log-normal | 3777 | 3786 | 1.05 | 1.26 | 0.60 | 1.86 | |
| Log-logistic | 3783 | 3791 | 1.04 | 1.24 | 0.87 | 2.11 | |
| Spline 3 months and 1 year | 3.4 | 3780 | 3796 | 1.04 | 1.21 | 0.31 | 1.52 |
| Spline 1 year | 3779 | 3792 | 1.05 | 1.25 | 0.52 | 1.77 | |
| Spline 3 months | 3779 | 3792 | 1.05 | 1.26 | 0.59 | 1.85 | |
| Spline 5 knots 3-month intervals | 3783 | 3812 | 1.06 | 1.31 | 2.00 | 3.31 | |
| MCM log-normal UK | 3.5 | 3772 | 3785 | 1.04 | 1.25 | 0.51 | 1.76 |
| MCM log-normal USA | 3771 | 3783 | 1.04 | 1.25 | 0.50 | 1.75 | |
| MCM Weibull UK | 3780 | 3792 | 1.05 | 1.34 | 4.47 | 5.81 | |
| MCM Weibull USA | 3779 | 3791 | 1.05 | 1.33 | 4.28 | 5.61 | |
| nMCM log-normal UK | 3773 | 3786 | 1.04 | 1.26 | 0.84 | 2.10 | |
| nMCM log-normal USA | 3772 | 3784 | 1.04 | 1.25 | 0.84 | 2.09 | |
| nMCM Weibull UK | 3779 | 3791 | 1.05 | 1.31 | 3.71 | 5.02 | |
| nMCM Weibull USA | 3777 | 3790 | 1.05 | 1.31 | 3.60 | 4.91 | |
| PMM Weibulla | 3.6 | 3787 | NA | 1.04 | 1.32 | 9.90 | 11.22 |
| PMM log-normala | 3779 | NA | 1.04 | 1.28 | 2.68 | 3.96 | |
| Landmark Gompertzb | 3.7 | 2910 | 2922 | 1.05 | 1.23 | 0.22 | 1.45 |
| Landmark generalised gammab | 2912 | 2928 | 1.05 | 1.23 | 0.23 | 1.46 | |
| Landmark exponentialb | 2908 | 2916 | 1.05 | 1.23 | 0.25 | 1.48 | |
| Landmark Weibullb | 2910 | 2922 | 1.05 | 1.24 | 0.24 | 1.48 | |
| Landmark log-logisticb | 2920 | 2932 | 1.07 | 1.32 | 1.40 | 2.72 | |
| Landmark log-normalb | 2952 | 2964 | 1.09 | 1.37 | 1.74 | 3.11 |
AIC Akaike information criterion, AUC area under the curve, BIC Bayesian information criterion, CI confidence interval, DCO data cut-off, K–M Kaplan–Meier, K-M +K-M out-of-sample fit, MCM mixture cure model, NA not available, nMCM non-mixture cure model, PMM parametric mixture model, wAIC Watanabe–Akaike information criterion
awAIC rather than AIC; without background mortality
bLandmark AIC/BIC are lower than the other AIC/BIC, because they only assess goodness of fit from a landmark time point onwards