Question 4 of 9 Intermediate Scenario
What is an estimand and why must it be defined before analysis?
Short answer
An estimand describes exactly what treatment effect a trial wants to estimate. It specifies the population, treatments, outcome, handling of events after randomization and how treatment effects are summarized. Defining it early aligns design, data collection and analysis with the clinical question.
Why
An estimand is the target quantity, not a particular statistical estimator. Under ICH E9(R1), its attributes specify the treatment conditions, target population, outcome variable, population-level summary and strategy for intercurrent events. Intercurrent events are things occurring after treatment initiation that affect interpretation or existence of measurements, such as treatment discontinuation, rescue medication or switching therapies.
A treatment-policy strategy asks about outcomes regardless of a specified intercurrent event. A hypothetical strategy might ask what the effect would be if rescue medication were not used. These can answer different clinical questions and often require different data collection or assumptions.
For example, if many participants switch treatment after toxicity, simply censoring them changes what question can be answered. Specify the estimand in the protocol before outcomes are known, then choose an appropriate estimator and sensitivity analyses to assess robustness of assumptions. Missing data and intercurrent events are related challenges but not interchangeable terms.
Common mistakes
- Calling a regression model the estimand, or automatically treating treatment discontinuation as ordinary missing data
The interviewer might ask next
How does a treatment-policy estimand differ from a hypothetical estimand after rescue medication?