Table 3A. Variable Impact on the Apparent C-Statistic for the Colorectal Cancer Model in Men

Variable	C-Statistic	Change in C-Statistic	Variables (n)
No Model	0.5	—	0
Age	0.663280	0.163280	1
Race/ethnicity	0.672989	0.009710	2
Pack-years of smoking	0.678214	0.005224	3
Alcoholic drinks per day	0.681442	0.003229	4
Body mass index	0.684263	0.002821	5
Years of education	0.686596	0.002333	6
Regular use of aspirin	0.688405	0.001809	7
Family history of colon cancer	0.689931	0.001526	8
Regular use of multivitamins	0.691143	0.001212	9
Red meat intake (oz) per day	0.691765	0.000622	10
History of diabetes	0.692365	0.000600	11
Moderate physical activity per day (hours)	0.692879	0.000513	12
History of cancer^*	0.693176	0.000297	13
Regular use of NSAIDs^*	0.693397	0.000221	14
Preference for well-done meat^*	0.693538	0.000141	15 (Full)

The outcome was colorectal cancer.
↵* The variable was excluded from the final model because of the relatively small amount of additional prediction accuracy that it could have contributed to the model. The final model, which included 12 variables, had an apparent c-statistic of 0.6929, which was within 0.001 of the accuracy associated with the c-statistic of the full model, which was 0.6935.
NSAIDS, nonsteroidal anti-inflammatory drugs.