To estimate a cutoff level for cumulative number of cigarettes smoked critical for diabetes prevention, we examined health-check data for 121 male residents in Shimane Prefecture, Japan, from 1998 to 2005. We used the Brinkman Index (BI) calculated as the number of cigarettes smoked per day multiplied by the number of years of smoking for the index of cumulative number of cigarettes smoked. Multivariate logistic-model analysis was conducted to determine the relationship between BI and the diabetes risk. We documented 26 new cases during the observation period. Multivariate-adjusted odds ratio (OR) for diabetes risk compared with BI of 0 was 3.52 (95% CI: 0.84–14.71) for BIs of 1–600 and 10.19 (95% CI: 2.38–43.64) for BI ≥ 601 (p-trend = 0.002). We found that the BI is an effective marker for predicting future risk for diabetes, and indicated a BI of 600 as a useful cutoff value.