Table 4 Reasons why attributes of estimands were inferable or non-inferable
From: Estimands in published protocols of randomised trials: urgent improvement needed
Question | Trials—n/N (%) |
|---|---|
Population | |
 Inferable (N = 32) | |
  Inferred as all eligible participants based on ITT description | 32/32 (100%) |
 Not inferable (N = 18) | |
  Analysis population not clearly described | 11/18 (61%) |
  Participants with treatment deviations excluded from analysis, but unclear whether target population was all patients (under hypothetical compliance) or subset of compliers | 7/18 (39%) |
Treatment condition(s) | |
 Inferable (N = 40) | |
  Inferred treatment policy based on ITT description | 33/40 (83%) |
  Inferred intended treatment based on the exclusion of certain deviations | 7/40 (18%) |
 Not inferable (N = 10) | |
  Unclear how treatment deviations will be handled in analysis | 9/10 (90%) |
  Unclear which treatment strategy planned analysis corresponds toa | 1/10 (10%) |
Population-level summary measure | |
 Inferable (N = 33) | |
  Inferred from the type of regression model | 17 (52%) |
  Stated type of summary measure they would estimate | 16 (48%) |
 Not inferable (N = 17) | |
  Analysis strategy not clearly described | 8 (47%) |
  Statistical test only | 9 (53%) |