Table 3. The accuracy by which molecularly selected patient trials can be distinguished from all-comer trials using two alternative endpoints.
Endpoint with the most accurate cutoff level | Molecularly selected patient trials (n = 8) | All-comer trials (n = 10) | Sensitivity | Specificity | AUC | |
---|---|---|---|---|---|---|
PFS-HR | 100% | 100% | 1.00 | |||
< 0.60 | 8 | 0 | ||||
≥ 0.60 | 0 | 10 | ||||
Odds ratio for the overall response | 88% | 90% | 0.95 | |||
≥ 3.40 | 7 | 1 | ||||
< 3.40 | 1 | 9 | ||||
Both the PFS-HR and odds ratio for the overall response | ||||||
< 0.60 and ≥ 3.40 | 7 | 0 | ||||
Other | 1 | 10 | ||||
Either the PFS-HR or the odds ratio for the overall response | ||||||
< 0.60 or ≥ 3.40 | 8 | 1 | ||||
Other | 0 | 9 |
AUC, area under the receiver operating characteristic curve.