-
1
Standardized pneumonia-response criteria do not, by themselves, define the treatment effect estimated in a randomized trial.
-
2
Among 1,603 pneumonia episodes, 302 had indeterminate outcomes, including 240 deaths; classifying death as failure increased the failure proportion from 22.2% to 34.3%.
-
3
Because death can be affected by treatment and precludes later response assessment, comparing only patients alive and classifiable on day 10 may not preserve the randomized causal contrast.
-
4
The response criteria could be incorporated into a global outcome or DOOR that includes death, complications, and adverse events; a survivor average causal effect requires additional assumptions.
-
5
Interventional protocols should prespecify the target population, treatments, response variable, handling of death and other assessment-precluding events, and treatment-effect summary measure.
