Table 6.

Simulated power (%)^a results for the less severe placebo group on day 7 of follow-up ($p_{07}$^b  = 0.82).^c

Model	T1: Constant benefitd	T2: Benefit only for first three days	T3: Linear decrease in benefit	T4: Larger benefit for hospitalized patients	T6: Benefit only for last three days	T7: Linear increase in benefit
No adjustment for baseline status
Simple Logistic	54.07	63.20	63.42	62.44	36.23	45.35
Sliding Dichotomy	47.75	58.77	56.92	55.50	32.77	39.06
Win Ratio	75.87	76.15	76.16	76.08	76.09	75.59
Proportional Odds	75.90	76.20	76.16	76.13	76.11	75.63
Longitudinal Ordinal Outcome	64.44	44.81	52.52	62.78	65.88	70.26
Cox Proportional Hazards	70.74	99.04	95.81	80.04	15.94	31.88
Accelerated Failure Time (Exponential)	55.18	97.11	90.71	66.84	7.51	17.25
Accelerated Failure Time (Weibull)	75.21	99.11	96.52	83.48	18.67	37.27
Adjustment for baseline status
Simple Logistic	59.39	68.44	69.06	68.29	41.08	50.68
Sliding Dichotomy	49.97	60.64	58.93	57.89	34.40	41.08
Win Ratioe	73.82	72.61	72.31	73.28	77.76	75.46
Proportional Odds	80.52	81.15	80.80	80.68	80.12	79.94
Longitudinal Ordinal Outcome	45.15	6.13	14.09	40.69	71.35	70.70
Cox Proportional Hazards	79.02	99.79	98.50	87.80	17.60	37.29
Accelerated Failure Time (Exponential)	58.68	98.82	94.48	70.81	5.83	16.23
Accelerated Failure Time (Weibull)	82.50	99.79	98.80	90.09	20.03	41.97

Due to the skewness of the less severe placebo group distribution, a treatment effect corresponding to T5 that approximated an odds ratio of 1.77 on day 7 could not be found.

^aPower (%): percentage of the 10,000 simulated datasets in which the Wald test statistic for the estimated treatment effect of the fitted model was significant at the two-sided 0.05 level. Power for the top three models for each treatment effect with or without adjustment for baseline health status is printed in bold.

^b$p_{07}$ denotes the probability of discharge (i.e., categories 5 and 6 of the ordinal endpoint combined) by day 7 of follow-up for the placebo group.

^cSupplementary Table 3 displays the category percentages for the less severe placebo group on day 7 and each of the six IVIG groups.

^dBenefit refers to the differences in transition probabilities between randomized groups across treatment effects.

^eUnlike all other models which include a covariate to adjust for baseline status, the win ratio stratifies by only comparing patients who started at the same baseline category between randomized groups.