Abstract
BACKGROUND: It is unclear whether the time-varying mediation effects of glucagon-like peptide-1 receptor agonists (GLP-1RAs) on the three-point major adverse cardiovascular event (3P-MACE) outcomes, including nonfatal stroke, nonfatal myocardial infarction, and cardiovascular death, via HbA(1c), urine albumin-to-creatinine ratio (UACR), and systolic blood pressure (SBP) vary by types of GLP-1RAs, and patient subgroups. METHODS: A causal mediation analysis was applied to assess the mediation effects of HbA(1c), UACR, and SBP on the 3P-MACE with liraglutide and semaglutide over time, using the LEADER and SUSTAIN-6 trials. We further explored the heterogeneity in mediation effects in subsets of vulnerable patients, including those with an estimated glomerular filtration rate (eGFR) level of < 60 ml/min/1.73 m(2) or established cardiovascular diseases (CVDs). RESULTS: Individual-level data consisting of 9340 (liraglutide: 4668; placebo: 4672) and 3297 (semaglutide: 1648; placebo: 1649) subjects from the LEADER and SUSTAIN-6 trials, respectively, was utilized for this study. The study population aged around 64.2-64.4/64.4-64.8 years old, with 64.0%-64.5%/58.5%-63.0% of males, and having a diabetes duration of 12.8-12.9/13.2-14.3 years in the LEADER/SUSTAIN-6 trials, respectively. Among study populations, those established CVDs (e.g., heart failure, stroke, myocardial infarction) accounted for 14.0%-35.4% in the LEADER trial, whereas 13.7%-61.9% in the SUSTAIN-6 trial. At the end of each trial, HbA(1c) contributed the most to liraglutide's/semaglutide's effect on 3P-MACE (38.2%, 95% confidence interval: 13.8%-194.5%/51.8%, 29.6%-86.7%), followed by UACR (17.9%, 8.5%-61.9%/12.4%, 6.1%-32.4%) and SBP (6.8%, 2.2%-31.0%/6.7%, 2.9%-11.7%), respectively. Over the trial period, the effect of liraglutide mediated by HbA(1c) increased from 31.3% to 38.2%, whereas the effects of semaglutide mediated by HbA(1c) remained stable from 52.5% to 51.8%. For patients with eGFR < 60 ml/min/1.73 m(2), HbA(1c) was the leading mediator (35.2%) of semaglutide-associated 3P-MACE, while the mediation effects via HbA(1c) (7.2%), UACR (11.8%), and SBP (5.1%) on liraglutide-associated 3P-MACE seemed similar and smaller. For patients with established CVDs, HbA(1c) was the leading mediator of liraglutide- (25.3%) and semaglutide-associated 3P-MACE (51.2%). CONCLUSIONS: The HbA(1c)-mediated effects of liraglutide and semaglutide varied over time and across patient subgroups. HbA(1c) consistently explained a larger proportion of mediation compared to UACR and SBP, with patterns suggesting that mediation through HbA(1c) may differ by underlying risk factors.