Method for evaluating prediction models that apply the results of randomized trials to individual patients

Table 5 Net benefit for the Moertel trial.

T	Treat all patients	Treat by prediction	Advantage of prediction
0.5%	0.133	0.139	0.007
1.0%	0.128	0.132	0.004
2.0%	0.118	0.116	-0.001
3.0%	0.108	0.108	0.000
4.0%	0.098	0.102	0.005
5.0%	0.088	0.091	0.003
7.5%	0.063	0.064	0.001
10.0%	0.038	0.065	0.028
12.5%	0.013	0.035	0.022
15.0%	-0.013	0.029	0.029
20.0%	-0.063	0.021	0.021
25.0%	-0.113	0.011	0.011
30.0%	-0.163	-0.002	-0.002
35.0%	-0.213	0.000	0.000

T is the treatment threshold corresponding to the minimum reduction in absolute risk required to consider opting for treatment. The units are the equivalent of the number of deaths per patient. The "advantage of prediction" column shows the increment in net benefit of prediction compared to the better of "treat all" and "treat none".

ISSN: 1745-6215