. Author manuscript; available in PMC: 2017 Jul 17.

Published in final edited form as: Stat Med. 2016 Nov 2;36(4):592–605. doi: 10.1002/sim.7157

Table I.

The impact of mis-specifying t^* on the power calculation using APPLE.

λ₂	APPLE¹	True t^*2 = 3 months		True t^* = 9 months

	Mis-specified t^m⁵ = 6	PW-Logrank³ t^m = 6 (%)	S-Logrank⁴ (%)	PW-Logrank t^m = 6 (%)	S-Logrank (%)
0.6	260	79	76	56	40
0.64	335	79	76	56	41
0.68	441	78	76	57	42
0.72	599	78	76	57	42
0.78	1025	77	76	58	43

The sample size calculated using APPLE is based on the target power at 80%, the proportion of subjects who can survive beyond the delayed phase at p = 70%, total accrual duration A = 1 year, total study duration τ = 3 years, and Type I error rate α = 0.05.

t^*: True hazard ratio changing time.

PW-Logrank: Piecewise weighted log-rank test.

⁴

S-Logrank: Regular log-rank test.

⁵

t^m: mis-specified hazard ratio changing time.