Summary: In this post hoc analysis of the EXSCEL trial (adults with type 2 diabetes), a validated diabetes outcomes model (UKPDS-OM2) explained only modest proportions of the observed relative risk reductions seen with once-weekly exenatide: MACE 29%, all-cause mortality 15%, CV death 18%, and stroke 29%, with larger proportions for heart-failure hospitalisation (67%) and myocardial infarction (200%, against a very small observed effect). Mediation analysis found that changes in HbA1c, blood pressure, heart rate, LDL cholesterol, triglycerides, and weight did not explain the mortality reduction.
PICO Summary
| Element | Detail |
|---|---|
| Population | Participant-level data from EXSCEL, an international placebo-controlled cardiovascular outcomes RCT in adults with type 2 diabetes (around 14,750 participants, mixed primary and secondary prevention). This report is a post hoc simulation and mediation analysis. |
| Intervention | Once-weekly subcutaneous exenatide 2 mg added to usual care, as randomised in the parent trial. |
| Comparison | Placebo plus usual care, as randomised in the parent trial. The present analysis compares model-simulated event rates against the trial-observed rates rather than testing a new intervention. |
| Outcome | Proportion of the observed relative risk reduction explained by simulated conventional risk-factor changes: MACE 29%, all-cause mortality 15%, CV death 18%, stroke 29%, heart-failure hospitalisation 67%, and myocardial infarction 200% (the model predicted a roughly 2% reduction against an observed ~1%, on a near-null base). Mediation analysis (Cox models) found that baseline-to-6 or 12-month changes in HbA1c, blood pressure, heart rate, LDL cholesterol, triglycerides, and weight did not mediate the effect on all-cause mortality. No new treatment-effect estimates, confidence intervals, or p-values for exenatide versus placebo are generated by this analysis. |
Expert Commentary
This is a mechanistic post hoc analysis, not a fresh efficacy trial, and it should be read as hypothesis-generating rather than practice-changing. The authors fed participant-level risk-factor trajectories into a validated type 2 diabetes outcomes model and asked a focused question: how much of the cardiovascular signal seen in EXSCEL can be accounted for by conventional risk-factor changes? The honest answer is that most of it cannot. Simulated changes in glycaemia, blood pressure, lipids, heart rate, and weight explained only a minority of the observed reductions in MACE, mortality, CV death, and stroke, and a formal mediation analysis found none of these factors mediated the mortality benefit at 6 or 12 months. The apparently large figures for heart-failure hospitalisation and myocardial infarction warrant caution: the myocardial infarction value of 200% reflects a model estimate set against a near-null observed effect, so it is statistically fragile rather than impressive. A key limitation is that the conclusions inherit every assumption of the simulation model and of the parent trial, and modelled risk factors are an incomplete proxy for biology. Can I use this with my patients? Not directly; it does not change who should receive a GLP-1 receptor agonist, but it does reinforce that the cardiovascular benefits of these agents are not fully explained by the metabolic numbers we routinely track. Clinicians should keep prescribing based on the trial outcomes themselves, while researchers pursue the residual mechanisms this analysis leaves unexplained.
References
Coleman RL, Adler AI, Mentz RJ, Fudim M, Sattar N, Holman RR. Impact of changes in conventional risk factors induced by once-weekly GLP-1 receptor agonist exenatide on cardiovascular outcomes: an EXSCEL post hoc analysis. Cardiovasc Diabetol. 2025;24(1):347. doi:10.1186/s12933-025-02866-7