Table 3.

e-Score model comparisons and LRT.

Model	χ 2 (df) P	AIC	AUC	LRTa
Primary analytic modelsb				GECCO e-Score: 8.608 > 9.49; P = 0.07 Recalibrated e-Score: 23.969 > 9.49; P < 0.001
Base model	56.610 (4) <0.0001	7339.076	0.579	7329.076
GECCO e-Score expanded model	65.218 (8) <0.0001	7338.468	0.589	7320.468
Recalibrated e-Score expanded model	80.579 (8) <0.0001	7323.107	0.600	7305.107

Abbreviation: AIC, Akaike information criterion.

^aLRT: If difference between the −2LogL for the base model and expanded model is greater than the χ² statistic for difference in degrees of freedom based on an α = 0.05, then the expanded model is significantly better.

^bPrimary analytic models use the male validation sample in the following form for the base model: colorectal cancer ∼ age + race + family history; with the addition of the GECCO e-Score and recalibrated e-Score quintiles in separate expanded models.