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 6 HR, RMST and derived statistics on survival for four randomized controlled trials in various cancer sites conducted by the Medical Research Council

Statistic	ASTEC	BA06	ICON4	OE02
t _max (yr)	6.8	6.7	6.4	10.8
$t_{final}^{*}$ (yr)	6.1	6.7	6.3	10.8
S (t ^∗)	0.72	0.39	0.082	0.11
N (events ^a)	1394 (188)	962 (483)	749 (417)	802 (655)
Design HR	0.75	0.75	0.75	0.75
Achieved HR	1.16	0.85	0.82	0.85
(Cox model)
P-value for HR	0.3	0.07	0.04	0.03
(logrank test)
P-value for non-PH	0.06	0.7	0.6	0.1
(G-T test ^b)
RMST in control	5.39	3.69	2.46	2.68
arm ( ${\hat{μ}}_{0}$ )
RMST in research	5.28	4.02	2.79	3.13
arm ( ${\hat{μ}}_{1}$ )
Diff. $\hat{Δ} =$	-0.11 (0.10)	0.33 (0.18)	0.33 (0.17)	0.46 (0.26)
${\hat{μ}}_{1} - {\hat{μ}}_{0}$ (SE)
P-value for	0.3	0.07	0.05	0.08
$\hat{Δ}$ (RMST test)
P-value for	0.07	0.4	0.7	0.04
non-PH ^c

See text for details.
^aEvents occurring in the interval (0,t ^*).
^bGrambsch-Therneau test
^cFrom a flexible parametric hazards model with a time-dependent treatment effect.

ISSN: 1471-2288