Restricted mean survival time: an alternative to the hazard ratio for the design and analysis of randomized trials with a time-to-event outcome

Table 2 Operating characteristics of the test of RMST difference

Model	Recruit	Follow-up	$t_{des}^{*}$	Sample	RMST test		Logrank test
	(K ₁, yr)	(K ₂, yr)		size	Power	Sig. level	Power	Sig. level
PH	1	7	8	424	89.6	4.6	90.7	4.6
	3	5	8	432	90.1	4.6	90.3	4.7
	5	3	7.5	463	90.9	5.1	90.7	4.9
	7	1	6.7	532	90.5	5.4	90.3	5.1
Non-PH	1	7	4.4	324	89.4	4.5	81.2	4.8
	3	5	4.4	325	90.0	5.0	83.2	5.2
	5	3	4.3	328	89.5	4.3	84.4	4.5
	7	1	3.8	351	91.6	4.8	83.5	5.5

Significance level and power are presented as percentages. Sample size calculations are for hypothetical trials with proportional hazards (PH) or non-proportional hazards (non-PH) of the treatment effect. Results are given for the RMST test and for the logrank test using the RMST designed sample sizes. Details are as given in Table 1.

ISSN: 1471-2288