Trial design considerations in the face of overdispersion and time-varying event rates: a meta-regression approach with an application in asthma

Abstract

**Purpose:** Relative changes in event rates are a common primary endpoint in confirmatory asthma trials. Reliable assumptions about baseline event rates and extra-Poisson variability are essential for sample size planning. We synthesize placebo-group exacerbation data, evaluate calendar-time trends, and assess implications for trial design. **Methods:** We systematically reviewed randomized controlled trials reporting recurrent asthma exacerbations and extracted control-group outcomes classified by placebo type or asthma severity. Hierarchical negative binomial models accommodated heterogeneous reporting formats and jointly modeled event rates and overdispersion. We compared separate subgroup models with covariate-based models and used meta-regression to adjust for calendar-time trends. Trial design implications were illustrated using the NAVIGATOR trial. **Results:** Sixty-two trials met the inclusion criteria. The estimated relative change in annualized exacerbation rates was -1.8% per year (95% CrI: -7.5%, 4.3%), corresponding to a 16.5% decrease per decade (95% CrI: -54.2%, 52.2%). Substantial between-study heterogeneity in event rates and overdispersion was observed. Within-trial overdispersion was estimated as 1.53 (95% CrI: 0.60, 2.83). In an illustrative phase III design scenario, sample size requirements were sensitive to the target subgroup and uncertainty in the predicted event rate and overdispersion. **Conclusion:** Event rates in clinical trials exhibit considerable heterogeneity and potential temporal variation, and overdispersion is a critical design parameter. Comprehensive modelling of recurrent-event endpoints allows for evidence-based and robust sample size calculations and trial design.