High hemoglobin A1c (HbA1c) levels are strongly associated with an increased risk of cardiovascular disease (CVD) in people with and without diabetes. However, information regarding the relationship between low HbA1c levels and the risk of CVD among people without known diabetes is limited. The aim of this large-scale, prospective, population-based cohort study was to clarify the association between HbA1c levels and CVD risk among people without known diabetes.We followed-up 10,980 men and 18,079 women (46-80 years old and free of CVD and cancer at baseline) in the Japan Public Health Center-based Prospective Study. Using Cox models, we estimated the hazard ratios for CVD risk with adjustments for age, sex, geographic areas, body mass index, smoking status, sports and physical exercise, alcohol intake, systolic blood pressure, non-high-density lipoprotein cholesterol, and high-density lipoprotein cholesterol.During the median follow-up of 9.4 years, 935 CVD events (770 strokes and 165 coronary heart diseases) occurred. We observed a nonlinear association between HbA1c levels and CVD risk in participants without known diabetes. Compared with HbA1c levels of 5.0 to 5.4% (31-36 mmol/mol), the hazard ratios for CVD in participants without known diabetes were 1.50 (95% confidence interval: 1.15-1.95), 1.01 (0.85-1.20), 1.04 (0.82-1.32), and 1.77 (1.32-2.38) for HbA1c levels of <5.0% (<31 mmol/mol), 5.5 to 5.9% (37-41 mmol/mol), 6.0 to 6.4% (42-47 mmol/mol), and ≥6.5% (≥48 mmol/mol), respectively (P value for nonlinear trend: <0.001). In addition, the hazard ratio for CVD was 1.81 (1.43-2.29) in patients with known diabetes compared with participants with HbA1c levels of 5.0 to 5.4% and without known diabetes. This nonlinear relation persisted after excluding people with kidney dysfunction, liver dysfunction, anemia, body mass index <18.5 kg/m, or early events within 3 years of follow-up (P value for nonlinear trend: <0.01 for all tests).In conclusion, both low and high levels of HbA1c were associated with a higher risk of CVD in a Japanese general population without known diabetes.