Figure 1.

Effects of liraglutide (vs. placebo) on CV outcomes among patients with and without baseline metformin use, adjusted for baseline covariates with inverse probability weighting. Shaded squares = metformin use at baseline; empty diamonds = no metformin use at baseline. HRs derived using a Cox proportional hazards regression model with randomization group, baseline metformin exposure, and the interaction of both as factors, and diabetes duration, eGFR, and age at baseline as additional covariates, adjusted for baseline covariates: age, sex, region, diabetes duration, HbA_1c, antihyperglycemic medication, eGFR, smoking, prior myocardial infarction, heart rhythm disorders, heart failure, left ventricular systolic dysfunction, left ventricular diastolic dysfunction, prior ischemic stroke, prior transient ischemic attack, prior hemorrhagic stroke, prior percutaneous coronary intervention, prior coronary bypass surgery, intracranial artery stenosis, carotid artery stenosis, peripheral arterial disease, and ≥50% stenosis of coronary, carotid, or other arteries. MI, myocardial infarction; n, number of patients with event; PYO, patient-years of observation; %, proportion of patients in subgroup with event.