Bias and precision of methods for estimating the difference in restricted mean survival time from an individual patient data meta-analysis

Table 1 Simulation results for comparisons of methods in estimating the difference in restricted mean survival time. Scenario with 5 trials and 200 patients per trial and with proportional hazards

Heterogeneity scenario	Methods	β = 0						β = −0.7
		t ^* = 5 True rmstD = 0			t ^* = 10 True rmstD = 0			t ^* = 5 True rmstD = 0.8			t ^* = 10 True rmstD = 2.0
		Bias	ESE	ASE	Bias	ESE	ASE	Bias	ESE	ASE	Bias	ESE	ASE
(σ²,τ²) = (0.01;0.01)	Naïve Kaplan-Meier	0.00	0.12	0.11	0.00	0.28	0.24	0.01	0.12	0.11	0.01	0.28	0.23
	Pooled Kaplan-Meier	0.00	0.12	0.13	0.00	0.29	0.30	0.01	0.12	0.13	0.02	0.30	0.29
	Pooled Exponential	0.00	0.10	0.11	0.00	0.27	0.28	0.01	0.11	0.11	0.02	0.27	0.27
	Peto-quintile	0.00	0.11	0.10	0.00	0.24	0.21	−0.04	0.12	0.10	−0.21	0.26	0.23
(σ²,τ²) = (0.01;0.10)	Naïve Kaplan-Meier	0.00	0.19	0.11	0.01	0.49	0.24	0.00	0.19	0.11	0.00	0.48	0.23
	Pooled Kaplan-Meier	0.00	0.20	0.18	0.01	0.50	0.47	0.00	0.20	0.18	0.01	0.48	0.44
	Pooled Exponential	0.00	0.19	0.17	0.01	0.48	0.45	0.00	0.19	0.17	0.01	0.47	0.44
	Peto-quintile	0.00	0.18	0.10	0.00	0.43	0.22	−0.04	0.19	0.10	−0.20	0.46	0.23
(σ²,τ²) = (0.10;0.01)	Naïve Kaplan-Meier	−0.01	0.12	0.11	−0.01	0.28	0.24	0.00	0.14	0.11	0.01	0.28	0.24
	Pooled Kaplan-Meier	−0.01	0.12	0.13	0.00	0.28	0.29	0.00	0.14	0.13	0.01	0.28	0.29
	Pooled Exponential	0.00	0.10	0.11	−0.01	0.26	0.28	0.00	0.12	0.12	0.02	0.27	0.27
	Peto-quintile	−0.01	0.11	0.10	−0.01	0.23	0.22	−0.03	0.13	0.11	−0.15	0.26	0.23
(σ²,τ²) = (0.10;0.10)	Naïve Kaplan-Meier	0.00	0.18	0.11	0.00	0.45	0.24	0.01	0.19	0.11	0.01	0.44	0.24
	Pooled Kaplan-Meier	0.00	0.18	0.18	0.00	0.45	0.45	0.01	0.19	0.18	0.02	0.44	0.43
	Pooled Exponential	0.00	0.17	0.17	0.00	0.44	0.44	0.01	0.19	0.18	0.02	0.44	0.42
	Peto-quintile	0.00	0.17	0.10	0.00	0.40	0.22	−0.02	0.19	0.11	−0.13	0.43	0.24

A DerSimonian-Laird random effects meta-analysis model was used for Pooled Kaplan-Meier, Pooled Exponential and Peto-quintile
β Size of treatment effect (=log(HR)), σ ² baseline hazard heterogeneity, τ ² treatment effect heterogeneity, ASE average standard error, CI Confidence interval, ESE empirical standard error, rmstD difference in restricted mean survival time, t ^* time horizon

ISSN: 1471-2288