OBJECTIVES:  To illustrate selection bias when choosing inappropriate anti-diabetes medications (ADM) as comparison to incretin enhancers in cardiovascular outcome studies using real-world data. METHODS:  Adults with type 2 diabetes (T2D) who initiated ADM between 2006 and 2012 and enrolled in Humedica HER database were included in this retrospective cohort study. Initiators of incretin enhancers were compared to initiators of insulin, metformin, sulfonylureas, and thiazolidinediones (TZD). Incretin enhancers included glucagon-like peptide-1-receptor agonists (GLP-1-RA) and dipeptidyl peptidase-4-inhibitors (DPP-4-I). In order to describe the effect of T2D severity progression on future treatment choices, study follow-up was divided into two 12-month periods prior to ADM initiation date (index date), outcome period (12 months closest to index date), and baseline period (12 months farthest from index date). T2D severity and complications were measured during the baseline period, and incident hospital-related myocardial infarction (MI) events were measured during the outcome period. Stepwise proportional hazards regression was used to estimate the pre-exposure risk of MI between ADM groups, with 95%CI upper limit ≥1.0 as selection bias indicator. RESULTS:  61,262 GLP-1-RA/DPP-4-I enhancer initiators were compared to 80,465 insulin; 184,655 metformin; 104,905 sulfonylureas; and 47,812 TZD initiators. On average, initiators of GLP-1-RA/DPP-4-I had T2D severity and complication history comparable to insulin initiators, but significantly different from other ADM initiators (poorly controlled T2D—HbA1c>7.5% and microvascular complications, respectively: GLP-1-RA/DPP-4-I, 11.2% and 6.2%; insulin, 10.8% and 7.9%; metformin, 2.6% and 1.8%; sulfonylureas, 5.3% and 3.4%; and TZD, 5.9% and 3.0%). Unlike other ADM, compared to insulin initiators, GLP-1-RA/DPP-4-I initiators didn’t have pre-exposure risk of MI (95%CI upper limit <1.0): insulin HR=0.5, 95%CI=0.42-0.6; metformin HR=0.97, 95%CI=0.81-1.16; sulfonylureas HR=1.24, 95%CI=1.03-1.5; and TZD HR=0.76, 95%CI=0.56-1.04. CONCLUSIONS:  To minimize selection bias in real-world cardiovascular outcome studies, GLP-1-RA and DPP-4-I should be compared to insulin rather than to other ADM.