Question 15 of 20 Intermediate Interpretation
How do intercurrent events change the treatment effect a trial estimates?
Short answer
An intercurrent event, such as stopping treatment, taking rescue medication or dying, happens after randomization and affects how the outcome is measured or interpreted. How the estimand handles it decides which question the trial answers. The five ICH E9(R1) strategies give five different treatment effects from the same data.
Why
ICH E9(R1) lists strategies, and each answers a different clinical question. Treatment policy uses the outcome regardless of the event, so it estimates the effect of being assigned the drug, including switching and rescue. A hypothetical strategy asks what would have happened if the event had not occurred, for example if rescue medication had not been available.
A composite strategy folds the event into the outcome, so stopping for toxicity counts as failure. A while on treatment strategy uses the outcome only up to the event. A principal stratum strategy looks at the effect in a subset defined by whether the event would occur, such as people who could tolerate the drug.
Take a diabetes trial where 20 percent of the placebo arm takes rescue insulin. A treatment policy estimate of HbA1c at week 26 will be smaller, because rescue lowers placebo values. A hypothetical estimate of the effect without rescue will be larger.
Neither is wrong. They answer different questions, which is why the strategy is fixed before analysis. Treatment policy cannot be used for death, because no later value exists.
A strong answer covers
- Definition of an intercurrent event
- Five strategies and the question each answers
- Same data, different effects depending on strategy
- Treatment policy unavailable for terminal events
- Strategy fixed before analysis
Common mistakes
- Treating intercurrent events as missing data to impute.
- Assuming one strategy must be used for every intercurrent event.
- Thinking the choice of strategy is a purely statistical decision with no clinical input.
The interviewer might ask next
Which strategy would you choose for treatment discontinuation due to adverse events, and why?