E model discrimination:

We assess model discrimination by estimating t-year (cumulative/dynamic) ROC curves, which display true positive and false positive rates across a range of threshold values used to classify an observation as ‘high-risk’. Area under the ROC curves (AUC) are also estimated.

We show ROC curves for 2-year, 5-year and 10-year prediction times. First we estimate ROC/AUC for the entire cohort, and then we stratify on categorized age groups.

Estimation methods employ a nearest neighbor approach that accounts for competing risks. More details can be found in the manuscript referenced below.

Table of AUC estimates:

Sex	2 year	5 year	10 year
Female	0.619	0.623	0.618
Male	0.67	0.652	0.647

Stratified Performance

By Arm

Table of AUC estimates:

Arm	2 year	5 year	10 year
Control	0.643	0.637	0.635
Intervention	0.599	0.612	0.607

By Screening

Table of AUC estimates:

Screening	2 year	5 year	10 year
0	0.614	0.638	0.631
1	0.638	0.625	0.631

Age (Males)

Table of AUC estimates:

Age	2 year	5 year	10 year
<= 60	0.641	0.639	0.65
60-65	0.657	0.646	0.621
65-70	0.67	0.65	0.622
70+	0.65	0.592	0.585

Age (Females)

Table of AUC estimates:

Age	2 year	5 year	10 year
<= 60	0.578	0.651	0.621
60-65	0.608	0.64	0.62
65-70	0.622	0.555	0.558
70+	0.582	0.594	0.582

References

Saha, P. and Heagerty, P. J. (2010). Time-Dependent Predictive Accuracy in the Presence of Competing Risks. Biometrics 66, 999-1011.