Prescribing Trajectories in Type 2 Diabetes in the United States, 2019–2024

IMPORTANCE: Clinical guidelines for type 2 diabetes (T2D) provide population-level recommendations, but real-world treatment patterns evolve dynamically and vary across patients. Understanding longitudinal prescribing trajectories may reveal heterogeneity in care not captured by cross-sectional analyses. OBJECTIVE: To identify and characterise real-world T2D prescribing trajectories using nationwide electronic health record (EHR) data. DESIGN, SETTING AND PARTICIPANTS: This retrospective cohort study used de-identified EHR data from the TriNetX Research Network. Adults newly diagnosed with T2D who initiated their first glucose-lowering medication in the first half of 2019 were followed from 2019 through 2024. EXPOSURES: Longitudinal glucose-lowering prescribing trajectories derived from prescription orders across nine drug classes (metformin, sulfonylureas, thiazolidinediones, dipeptidyl peptidase-4 inhibitors, sodium-glucose cotransporter 2 inhibitors, glucagon-like peptide-1 RAs (GLP-1 RAs), dual glucose-dependent insulinotropic polypeptide/GLP-1 RAs (GIP/GLP-1 RAs), insulin and other antidiabetic agents), encoded in 12 semiannual intervals and identified using hierarchical agglomerative clustering. MAIN OUTCOMES AND MEASURES: Prescribing trajectory cluster membership; adjusted longitudinal changes in body mass index (BMI) and haemoglobin A1c (HbA1c); and associations between cluster membership and patient characteristics assessed using multivariable logistic regression. RESULTS: Among 9327 patients, 30 distinct prescribing trajectory clusters were identified and grouped into monotherapy, dual therapy, complex therapy, variant therapy and GLP-1 RA-based trajectories. Treatment intensification patterns, BMI and HbA1c trajectories and demographic composition varied substantially across clusters. Within GLP-1 RA-based trajectories (11 clusters), substantial within-group reductions in BMI and HbA1c were observed; these clusters generally comprised patients with higher baseline BMI and most commonly involved transitions following metformin. GLP-1 RA prescribing was more frequent among younger patients and those with higher baseline BMI. In early GLP-1 RA transition trajectories, Asian and Hispanic patients had lower odds of cluster membership compared with non-Hispanic White patients (Asian: odds ratio [OR] 0.40; 95% CI 0.23-0.63; Hispanic: OR 0.57; 95% CI 0.40-0.79). CONCLUSIONS AND RELEVANCE: Data-driven clustering of longitudinal EHR medication data identifies substantial heterogeneity in real-world T2D treatment trajectories. Differences in the timing and adoption of newer therapies were observed across demographic groups, underscoring the value of longitudinal approaches for evaluating real-world diabetes care patterns.