OBJECTIVE: This study aimed to examine the impact of obesity, as defined by body mass index (BMI), and a metabolically unhealthy phenotype on the development of coronary artery disease (CAD) according to glucose tolerance status. METHODS: This population-based retrospective cohort study included 123,746 Japanese men aged 18-72years (normal glucose tolerance: 72,047; prediabetes: 39,633; diabetes: 12,066). Obesity was defined as a BMI≥25kg/m2. Metabolically unhealthy individuals were defined as those with one or more of the following conditions: hypertension, hypertriglyceridaemia and/or low HDL cholesterol. A Cox proportional hazards regression model identified variables related to CAD incidence. RESULTS: The prevalences of obese subjects with normal glucose tolerance, prediabetes and diabetes were 21%, 34% and 53%, whereas those for metabolically unhealthy people were 43%, 60% and 79%, respectively. Multivariate analysis showed that a metabolically unhealthy phenotype increases hazard ratios (HRs) for CAD compared with a metabolically healthy phenotype, regardless of glucose tolerance status (normal glucose tolerance: 1.98, 95% CI: 1.32-2.95; prediabetes: 2.91, 95% CI: 1.85-4.55; diabetes: 1.90, 95% CI: 1.18-3.06). HRs for CAD among metabolically unhealthy non-obese diabetes patients and obese diabetes patients with a metabolically unhealthy status were 6.14 (95% CI: 3.94-9.56) and 7.86 (95% CI: 5.21-11.9), respectively, compared with non-obese subjects with normal glucose tolerance and without a metabolically unhealthy status. CONCLUSION: A metabolically unhealthy state can associate with CAD independently of obesity across all glucose tolerance stages. Clinicians may need to consider those with at least one or more conditions indicating a metabolically unhealthy state as being at high risk for CAD regardless of glucose tolerance status.