Comprehensive Summary
This study investigates whether the cardiovascular benefits of GLP‑1 receptor agonists (GLP-1RAs) seen in randomized cardiovascular outcome trials (CVOTs) for people with type 2 diabetes (T2D) are transferable to real-world populations, and whether specific patient subgroups derive greater benefit. The authors combined individual-level trial data from the LEADER trial and SUSTAIN‑6 trial with real-world electronic health-record data from the DARWIN‑T2D study, using machine learning (the PRISM framework) to identify subgroups with differential response, and then externally validated findings in a large claims database. They found that the hazard‐ratio reduction for 3-point major adverse cardiovascular events (3P-MACE) observed in trials could be closely transposed to the real-world cohort. Importantly, the algorithm identified a subgroup of patients aged > 71 years without prior myocardial infarction or stroke (about 41% of the real-world cohort) who experienced the greatest absolute risk reduction (ARR ~4.5%) and HR around 0.46 (95% CI 0.24–0.89) in the test set. In external validation new-users of GLP-1RA versus comparators had HR 0.67 and ARR ~3.8% in this subgroup. In the discussion the authors emphasise that (1) trial benefits are transferable to routine care, (2) no subgroup was found that derived no benefit, but some subgroups derived greater absolute benefit, and (3) older patients without established cardiovascular disease (CVD) may need greater recognition as candidates for GLP-1RA therapy.
Outcomes and Implications
This work is important because it supports broader implementation of GLP-1RAs for cardiovascular prevention in people with T2D, bridging the gap between trial populations and “real‐world” patients. Clinically, the finding that elderly patients (age > 71) without prior CVD derive cardiovascular benefit akin to those with known CVD suggests that guidelines might expand priorities to include such “primary prevention” older adults, thereby improving outcomes and optimizing therapy targeting. Since the study used existing datasets, the timeline for clinical translation is relatively near-term: physicians and policymakers could incorporate these stratification insights in decision-making now, though further prospective studies across more diverse populations and settings would strengthen implementation.