Table 3.
Illustrative example to show differences between three statistical approaches comparing the effect of treatments on functional outcomes “truncated due to death.” We consider the population of patients eligible for the ABC trial and for each patient, generate the potential outcome under control by mimicking characteristics of the control group from the ABC trial such that 62% of the patients receiving the control would experience death by 12 months, the mean cognition among survivors is 42, and based on the composite outcome, as described in the manuscript, 50% of the patients will survive past 72 days (25% of patients will survive to 12 months with cognition score >39). Assuming the monotonicity assumption holds, potential outcomes under intervention were derived based on the effect of the intervention on mortality and the functional outcome among always survivors (for a total of six scenarios). For each scenario, the true population value of the treatment effect defined by the survivors analysis, the SACE and composite endpoint approach are displayed
| Intervention has no effect on mortality | Intervention promotes survival in patients with robust or high cognition | Intervention sustains patients who are less healthy or have poor cognition | ||||||
|---|---|---|---|---|---|---|---|---|
| Treatment improves cognition among always survivors | No | Yes | No | Yes | No | Yes | ||
| Scenario | 1 | 2 | 3 | 4 | 5 | 6 | ||
| Always survivors | 38% | 38% | 38% | 38% | 38% | 38% | ||
| Mortality benefiters | 0% | 0% | 21% | 21% | 21% | 21% | ||
| Always diers | 62% | 62% | 41% | 41% | 41% | 41% | ||
| Treatment effects: | ||||||||
| Survivors analysis | 0 | 15 | 0 | 15 | -7 | 2 | ||
| Survivor average causal effect | 0 | 15 | 0 | 15 | 0 | 15 | ||
| Composite endpoint* | ||||||||
| Rank statistic | 0 | 0.12 | 0.36 | 0.55 | 0.29 | 0.41 | ||
| Median under intervention | Survive >72 days | Survive >72 days | Survive 12 months with cognition >34 | Survive 12 months with cognition >49 | Survive 12 months with cognition >20 | Survive 12 months with cognition >20 | ||
| 75th centile under intervention | Survive 12 months with cognition >39 | Survive 12 months with cognition >54 | Survive 12 months with cognition >41 | Survive 12 months with cognition >56 | Survive 12 months with cognition >39 | Survive 12 months with cognition >54 | ||
For composite endpoint analysis, the median under control is survive >72 days, the 75th centile under control is survive to 12 months with cognition score >39.