Table 1.

Sensitivity Analysis for the EMPULSE trial Win Ratio Approach

A. Different Ways of Handling HF Events
Criterion to Declare a Win	% of Wins using HF Events		Overall Result
	on empa	on placebo	Win Ratio (95% CI)	P	% tied
Number of HF events	10.59%	7.65%	1.36 (1.10, 1.69)	0.0051	6.5%
Time to First HF event	10.79%	8.25%	1.36 (1.09, 1.68)	0.0057	6.4%
Both in Sequence*	10.83%	8.21%	1.36 (1.09, 1.68)	0.0054	6.4%
B. Different Ways of Handling KCCQ-TSS change in 90 days
Criterion to Declare a Win	% of Wins using KCCQ		Overall Result
	on empa	on placebo	Win Ratio (95% CI)	P	% tied
any difference	38.94%	30.35%	1.34 (1.09, 1.64)	0.0051	0.5%
≥2 points	38.16%	29.54%	1.34 (1.09, 1.65)	0.0050	2.1%
≥5 points*	35.91%	27.48%	1.36 (1.09, 1.68)	0.0054	6.4%
≥10 points	32.39%	24.06%	1.39 (1.10, 1.75)	0.0050	13.3%
≥15 points	28.83%	20.70%	1.42 (1.11, 1.81)	0.0046	20.3%
KCCQ-TSS not used	0%	0%	1.50 (0.99, 2.26)	0.055	69.5%

^*the pre-specified primary analysis

all analyses are stratified, with multiple imputation for missing KCCQ at 90 days

Table 1A and 1B shows how the win ratio results are affected by a variety of sensitivity analyses. Table 1A shows that the win ratio is virtually unchanged if we use either (i) number of HF events, or (ii) the time to first HF event, or (iii) both in sequence (the pre-specified primary analysis in EMPULSE). Table 1B shows how the win ratio is affected by varying the winning margin required for the KCCQ-TSS or by omitting this level completely. The win ratio increases with larger margins though the p-value remains fairly unchanged.