Aims/hypothesis High fasting blood glucose is one of the well-known risk factors for CHD. However, in certain settings, patients cannot always be expected to fast. For example, community screenings for cardiovascular disease (CVD) risk factors in Japan are performed under non-fasting conditions to achieve high participation rates. Thus, we examined a representative cohort of the Japanese population ( n  = 9,444, follow-up period 17.3 years) to clarify whether high casual blood glucose (CBG) can predict CVD mortality. Methods We defined CBG groups as follows: high CBG ≥ 11.1 mmol/l or participants with a history of diabetes mellitus; borderline high, 7.77 ≤ CBG < 11.1 mmol/l; higher normal, 5.22 ≤ CBG < 7.77 mmol/l); and lower normal, CBG < 5.22 mmol/l. The multivariate-adjusted hazard ratios (HRs) for CHD, CVD and all-cause mortality were calculated. Results The crude CHD mortality rate was 0.84 per 1,000 person-years. Age- and sex-adjusted HRs for CHD mortality were high among participants with CBG levels  ≥ 7.77 mmol/l, regardless of time since last meal. Multivariate-adjusted HRs (95% CI) of CHD mortality in high and borderline high CBG groups were 2.62 (1.46–4.67) and 2.43 (1.29–4.58), respectively. Similar results were observed for both CVD and all-cause mortality. Even within the normal blood glucose range, each 1 mmol/l increase in CBG was associated with a statistically significant increase in the HR for CVD mortality (1.12, 95% CI 1.02–1.22). Population-attributable fractions of the combined groups of high and borderline high CBG for CHD, CVD and all-cause mortality were 12.0, 4.9 and 3.5%, respectively. Conclusions/interpretation Increases in CBG, even within the normal range, predict CVD mortality.