Beyond Adherence: Unraveling Methodological Complexity in Comparative Weight Change Among Antidiabetic Agents

We read with great interest the recent article by Sneed et al. titled “Comparative Weight Change With Initiation and Adherence to Common Medications for Type 2 Diabetes 1,” published in Obesity. The authors employed an impressive target trial emulation framework using electronic health records (EHR) from over 22,000 patients, providing valuable real-world evidence regarding weight trajectories associated with second-line glucose-lowering agents in adults with type 2 diabetes treated with metformin. Their findings, particularly the sustained weight reduction observed with glucagon-like peptide-1 receptor agonists (GLP-1RAs) and sodium-glucose cotransporter 2 inhibitors (SGLT-2is), have significant implications for clinical decision-making. However, from a clinical standpoint, several methodological considerations may warrant further discussion, as they could substantially influence the interpretation of the reported effects. First, while the authors' use of inverse probability weighting and marginal structural models represents a sophisticated approach to emulate randomization, the treatment exposure definition—specifically “initiation and adherence”—may introduce a composite exposure misclassification that undermines causal inference. In clinical practice, adherence to GLP-1RAs and SGLT-2is is often constrained by side effects (e.g., nausea, genitourinary infections) and cost, which selectively affects patient continuation. Patients who remain adherent for 24 months likely differ systematically from those who discontinue early—not only in socioeconomic status but also in metabolic phenotype (e.g., baseline BMI, insulin resistance, motivation for weight control). By modeling “adherence” as a time-varying covariate rather than a stratified behavioral phenomenon, the study may have inadvertently captured the effect of a “high-adherence phenotype” rather than the pharmacologic weight effect itself. This may inflate the apparent weight loss associated with GLP-1RAs and SGLT-2is, particularly relative to medications that are less self-limited by cost or tolerance. Second, the analytical framework presumes that the recorded baseline and time-varying covariates sufficiently account for confounding by indication; yet several clinically salient confounders—particularly the temporal evolution of glycemic control—were likely incompletely captured. In routine diabetes care, escalation from metformin to agents such as basal insulin or sulfonylureas is typically triggered by poor glycemic control, which itself correlates with catabolic weight loss prior to treatment intensification 2. Without incorporating longitudinal hemoglobin A1c trajectories as a dynamic confounder, inverse probability weighting may overestimate post-initiation weight gain among insulin or sulfonylurea users, falsely magnifying the relative weight advantage of GLP-1RAs and SGLT-2is. This bias is subtle yet critical, as EHR data often lag behind true metabolic shifts leading to therapy changes, challenging the assumption of exchangeability at treatment initiation. Third, the decision to pool individual medications within drug classes (except sulfonylureas) may obscure clinically relevant heterogeneity. The GLP-1RA group likely included both short- and long-acting formulations with distinct pharmacodynamics and dosing burdens, influencing both efficacy and adherence. For instance, semaglutide produces substantially greater weight reduction than exenatide or albiglutide 3. Aggregating these drugs may lead to “dilution” of effect estimates and limit interpretability for clinicians choosing among specific agents. Furthermore, given that semaglutide and tirzepatide have now reshaped treatment paradigms, their underrepresentation in the dataset may result in conservative estimates that underestimate the real-world effect of modern incretin therapy. In summary, Sneed et al. make a valuable contribution by applying target trial emulation to compare population-level weight outcomes of common antidiabetic medications. Nevertheless, we believe that the conflation of adherence and pharmacologic exposure, incomplete adjustment for glycemic trajectory–related confounding, and intraclass heterogeneity collectively warrant cautious interpretation of the magnitude and generalizability of the reported effects. As clinicians increasingly individualize therapy to optimize both glycemic and weight outcomes, clarifying these nuances will be essential for translating real-world data into precision treatment strategies for type 2 diabetes. The authors have nothing to report. The authors declare no conflicts of interest. Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.