Medication adherence is dynamic and varies over time, yet many studies use the proportion of days covered as a static measure summarized across the entire observation period, which can obscure meaningful changes in adherence behavior over time. Group-based trajectory modeling (GBTM) has become an increasingly popular method for examining longitudinal adherence patterns and for identifying distinct subgroups of individuals with similar medication-taking behavior. Although this approach provides important advantages, it also introduces additional methodological complexity that has implications for reproducibility, interpretation, and bias. This Viewpoint article provides practical guidance for researchers, clinicians, editors, and peer reviewers who evaluate or apply GBTM to study medication adherence. We describe common modeling decisions, including selecting the time scale, choosing between binary or continuous adherence measures, determining the number and shape of trajectories, and handling enrollment requirements, while highlighting the benefits and limitations associated with each approach. We also identify major threats to study validity, including selection bias from restrictive continuous enrollment criteria, immortal time bias, and reverse causality. To support transparency and consistency in reporting, we summarize key components that should be documented in manuscripts using GBTM. In sum, we aim to strengthen the quality of evidence informed by medication adherence studies.
Rizk et al. (2026) studied this question.