Measurement differences explain much of the variation in adherence and persistence after ischemic stroke, and standardized estimates are lower: a systematic review with meta-regression

Why clinicians should know about it

• Picked for Epidemiology (paper of the day, 25 October 2026): Methods-focused review quantifies measurement-related heterogeneity and bias

Abstract

OBJECTIVE: Continued use of secondary prevention medications after ischemic stroke or transient ischemic attack is an important component of long-term management. Reported estimates of adherence and persistence vary considerably across studies, limiting comparison and interpretability. Although some variation may reflect differences in patient behavior, inconsistencies in operational definitions and measurement approaches are also likely to contribute. We sought to examine sources of heterogeneity and to derive standardized estimates using a common reference measurement framework. STUDY DESIGN AND SETTING: We searched bibliographic databases, dissertations, and trial registries from inception to September 10, 2025, for routine-care studies reporting adherence or persistence to secondary prevention medications after ischemic stroke or transient ischemic attack. Cutoff-based proportions were extracted. We used 3-level hierarchical Bayesian models to characterize heterogeneity and multivariable meta-regression to evaluate associations with measurement specifications, follow-up window, medication class, region, and study-level characteristics. From the fitted model, we derived estimates standardized to a prespecified reference specification: pharmacy or administrative data, moderate definition stringency, and a 9- to 15-month follow-up window. RESULTS: The review included 141 studies; 101 (432 estimates; 976,020 participants) contributed to the quantitative synthesis. Unstandardized pooled estimates suggested that 74.1% of patients were adherent and 76.8% were persistent. Heterogeneity was substantial, with explained variation concentrated more in the measurement-related Assessment and Administration domains than in reported Patient characteristics. Under the common reference specification, the model-based estimates were lower, at 53.7% for adherence and 48.2% for persistence. Standardized persistence declined from 63.3% at 1 to 4 months to 28.1% beyond 30 months. Adjusted estimates were generally lower in Asia; persistence was numerically lowest for anticoagulants, although credible intervals were wide and overlapping. CONCLUSION: A large proportion of the heterogeneity reported in medication adherence and persistence after ischemic stroke or transient ischemic attack appears to be attributable to methodological non-equivalence in how these outcomes are defined and measured. When evaluated within a standardized framework, suboptimal medication continuation appears more common than conventional pooled estimates indicate. Meaningful surveillance and benchmarking require longitudinal, standardized, and transparent measures of medication continuation rather than reliance on discharge prescribing alone.

Abstract as published, via PubMed.