Table 1.
RPSFTM, IPCW and two‐stage estimates of overall survival treatment effect
| Description | HR | CF HR comparisona | Median (chemo group, days) | ||
|---|---|---|---|---|---|
| Point estimate | Lower 95% CI | Upper 95% CI | |||
| METRIC primary efficacy population | |||||
| ITT analysis | 0.72 | 0.52 | 0.98 | – | 338.0 |
| RPSFTM | 0.38 | 0.15 | 0.95 | 1.00 | 220.0 |
| RPSFTM without recensoring | 0.49 | 0.25 | 0.96 | 1.00 | 220.0 |
| IPCWb | 0.48 | 0.25 | 0.91 | – | – |
| Two‐stage method | 0.43 | 0.20 | 0.96 | – | 244.2 |
| Two‐stage method without recensoring | 0.53 | 0.29 | 0.97 | – | 244.2 |
| METRIC first‐line metastatic primary efficacy population | |||||
| ITT analysis | 0.67 | 0.45 | 1.00 | – | 338.0 |
| RPSFTM | 0.33 | 0.11 | 1.00 | 1.00 | 196.0 |
| RPSFTM without recensoring | 0.44 | 0.20 | 1.00 | 1.00 | 207.1 |
| IPCWb | 0.33 | 0.16 | 0.68 | – | – |
| Two‐stage methodb | 0.51 | 0.26 | 1.00 | – | 256.5 |
| Two‐stage method without recensoringb | 0.55 | 0.30 | 1.00 | – | 268.5 |
CI, confidence interval; HR, hazard ratio; IPCW, inverse probability of censoring weights; ITT, intent‐to‐treat population; LDH, lactate dehydrogenase; OS, overall survival; RPSFTM, rank‐preserving structural failure time models.
The ITT HRs presented in the table are estimated using Cox PH models, stratified for LDH. GSK use a Pike estimator to calculate ITT HRs. This resulted in a HR of 0.72 (95% CI, 0.52–1.01) for the Primary Efficacy Population and a HR of 0.74 (95% CI, 0.49–1.12) for the first‐line metastatic subpopulation. Results are given to 2 decimal places.
CF HR comparison represents the comparison of counterfactual survival times in each randomized group if no patients in either group received any treatment, given the treatment effect estimated by the method. Successful estimation would result in a CF HR of 1.00.
Analyses did not fully converge when all covariates were included. Results presented represent those from most complete application that converged (see Data S1. for full details).