(165) A hypothetical case study describing different intercurrent events in an observational prospective study: Differential interpretation of results

Background: Estimands are a relatively new concept in observational studies. This abstract explores how different strategies for handling intercurrent events (ICEs) can lead to multiple interpretations of a research question. ICEs are complex aspects of estimands that need careful consideration.

Objectives: To evaluate how different descriptions of ICEs impact the interpretation of results in a hypothetical prospective study.

Methods: We applied the ICH-E9 R1 addendum on estimands to one example of an observational prospective study. We considered different scenarios for the description of ICEs.

Results: For the research question “What is the improvement (i.e., reduction) in exacerbations with a prescribed treatment (PT) after 12 months of follow-up compared to the 12 months pre-treatment period in adults with severe eosinophilic asthma (SEA)?”, we considered two possible estimands that differed in their approach to handling ICEs. We assumed a population of adults with SEA, the endpoint to be number of exacerbations after 12 months, and the population-level summary to be change in exacerbation rate between 12 months post treatment compared to previous 12 months. Discontinuation of PT, switch to a rescue medication (e.g., inhaled corticosteroids), and death (a “while on treatment” strategy for death was assumed for all scenarios) were considered ICEs.

In scenario 1, we apply the “treatment policy” for the ICEs (i.e., ICEs are ignored, and all observations collected after the ICE are used). This scenario does not take into account the use of rescue medication or patient persistence on treatment. These results would be applicable to patients with varying levels of rescue medication use and persistence to PT, which may be reflective of the real-world.

In scenario 2, we apply the “composite strategy” for the use of rescue medication (i.e., the ICE becomes part of the endpoint definition itself). With this strategy, we consider the use of rescue medication to be a treatment failure. These results would be applicable to patients who do not use rescue medications, which might not reflect the real-world.

Conclusions: The strategies for dealing with ICEs should consider different perspectives from different functional groups or stakeholders and take into account how the study results will be used and for what purpose. Our hypothetical example shows that even if a research question is clearly defined, the findings can be prone to misinterpretation if the strategies for the ICEs are not clearly chosen and described as part of the estimand.