Two-stage estimation to adjust for treatment switching in randomised trials: a simulation study investigating the use of inverse probability weighting instead of re-censoring

Table 2 Scenarios 25 and 26 – performance measures for estimation of control arm RMST

Scenario details	Method	Percent bias	Empirical SE of % bias	RMSE of % bias	Coverage (%)	Convergence (%)
Scenario number: 25 True RMST: Control: 228 Experimental: 269 Mean switch: 25% True ave. HR: 0.78 True ave. AF: 1.30 Mean censored: 18%	No switching	0.1	5.8	5.8	94.9	100
	ITT	2.7	5.8	6.4	92.9	100
	TSE	0.3	6.4	6.4	95.6	100
	TSEnr	1.0	5.9	6.0	94.8	100
	TSEipcw	1.0	6.6	6.7	95.3	100
	min/max MC error	0.2/0.2	0.1/0.1	0.1/0.2	0.7/0.8	–
Scenario number: 26 True RMST: Control: 228 Experimental: 269 Mean switch: 57% True ave. HR: 0.78 True ave. AF: 1.30 Mean censored: 18%	No switching	0.0	5.7	5.7	95.6	100
	ITT	6.2	5.6	8.4	85.4	100
	TSE	0.3	7.1	7.1	95.2	100
	TSEnr	1.5	6.4	6.6	95.3	100
	TSEipcw	1.8	8.3	8.5	95.2	100
	min/max MC error	0.2/0.3	0.1/0.2	0.1/0.3	0.6/1.2	–

Note: RMST restricted mean survival time, HR hazard ratio, AF acceleration factor, SE standard error, RMSE root mean squared error, MC Monte-Carlo; ITT intention to treat, TSE two-stage estimation, TSEnr two-stage estimation without re-censoring, TSEipcw two-stage estimation with inverse probability of censoring weights

ISSN: 1471-2288