(C-156) Nuances of Risk Window Selection for Evaluation of Acute Adverse Events

Background: Real-world studies are used to evaluate medication safety. These studies must account for the time after medication discontinuation during which a drug may influence the risk of adverse events (AEs), termed as the risk continuation window (RCW). In practice, choosing a RCW has implications beyond the additional time at risk that are not always obvious.

Objectives: To examine the implications of using multiple RCWs. Serious infection (SI) outcomes between IL-17A and non-IL-17A biologic drug classes, consisting of medications with varying half-lives are considered as an example.

Methods: A cohort study using CorEvitas Psoriasis Registry data (APR 2015 - SEPT 2024) was performed. Subjects aged 18+ with at least one qualifying medication start and follow-up or reported targeted AE were included. Three RCWs (30-, 60-, and 90-day) were evaluated. For each exposure episode, the index date was set to the drug start date. The episode continued until the earliest of discontinuation plus RCW, new drug start, Registry exit, or death. Once exposure windows were defined, the first SI event during the window was identified. Differences in populations and event sets under each RCW were illustrated using graphical case examples and Venn diagrams. Descriptives at index date, crude incidence rates, and 95% Poisson confidence intervals were calculated. Propensity score matching (PSM) was used prior to estimating hazard ratios (HRs) using multivariable Cox proportional hazards models.

Results: In the pre-matched and PSM populations, more exposure episodes were observed with shorter RCWs. Conversely, SI rates were highest with the 90-day RCW. Case examples showed how the RCW influenced exposure eligibility and ultimately the set of SI events included in each cohort, as the RCWs created different exposure episodes. Consecutive exposures to the same drug were considered one continuous episode or separate episodes depending on the RCW. Furthermore, separate exposure episodes resulted in separate index dates with different baselines, affecting PSM. RCW choice had little impact on baseline characteristics as imbalances were observed regardless of the RCW. Despite differences in exposures and events, the HR for SI was similar across RCW approaches.

Conclusions: The choice of RCW impacts eligible and matched exposure episodes and ultimately the events credited to each cohort. Evaluating multiple RCWs in a single study may obfuscate findings and may be unadvisable. Selecting a single RCW reflecting a drug’s half-life is preferred, although challenges may arise when considering multiple drugs with different pharmacokinetic properties.