Improving Medication Classification in GLP-1RA Research: Feasibility of a Photo-Review Protocol

Glucagon-like peptide-1 receptor agonists (GLP-1RAs) are dominating the type 2 diabetes and obesity intervention landscape, yet their real-world use is increasingly difficult to characterize amid widespread availability of non-FDA-approved, compounded products obtained outside traditional healthcare systems. These users may not be captured by the clinical and claims databases commonly used in real-world GLP-1RA research. Online surveys offer a scalable alternative for reaching this population; however, existing survey studies depend on self-reported medication use, the reliability of which has not been well characterized.

An online study was conducted among U.S. adults recruited via Prolific who reported ≥ 3 months use of once-weekly injectable semaglutide or tirzepatide. Participants completed a screening survey requiring a photo of the prescribed GLP-1RA, followed by a main survey for eligible participants assessing self-reported GLP-1RA type and sociodemographic and health characteristics. Two blinded, independent coders classified photos as one of four FDA-approved branded formulations or a compounded product; discrepancies were resolved by the second author. Agreement was assessed using Cohen's kappa.

Of the 419 screening survey participants, 322 participants comprised the analytic sample; 22 of the total exclusions were based on medication photo review. Self-reported and photo-coded medication types showed substantial agreement (82.61%; Cohen's κ = 0.78, 95% CI [0.73, 0.83]). Photo review classified 83 participants (25.78%) as using compounded GLP-1RAs; among these, 54 (65.06%) self-reported using one of the four branded GLP-1RAs and 29 (34.94%) selected a "none of the above" response option, operationalized as compounded use.

This study demonstrated that photo review may be a feasible method for improving medication classification in research and offered recommendations for future studies. Additional work is needed to evaluate this protocol in larger and more diverse samples and to refine formulation-specific self-report items.