Purpose Within OECD countries, Japan has the lowest percentage (3.4%) of the people with obesity (30≦ body mass index, BMI). This study aims to reveal the relationship between BMI and mortality among older Japanese.

Method We started a cohort study (Aichi Gerontological Evaluation Study, AGES) on non- institutionalised elderly aged 65 and over in 2003. We excluded the subjects with missing data and who died within 1 year in 4 year follow-up period considering reverse causation. Consequently, 17 017 subjects among 21 047 respondents for baseline survey were analysed. Cox's proportional hazard model was applied to calculate HR and 95% CI of BMI for all cause and cause specific mortality. Age, disease history, weight loss, smoking, alcohol drinking, and exercise were used as covariates.

Result Respondents with lowest category of BMI (under 18.5) had the highest all cause mortality among both sexes. Compared to the respondents with BMI 23.0–24.9, men respondents with under 18.5 BMI and BMI 18.5–22.9 had significantly higher BMI (2.16 (1.58–2.95, p<0.001) and 1.30 (1.04–1.63, p=0.015) respectively). For cause specific mortality, men with BMI under 18.5 had a significantly higher HR of malignant neoplasm mortality (HR=1.68 (1.05–2.67, p=0.030)) and respiratory disease mortality (HR=5.63 (2.57–12.34, p<0.001)).

Conclusion Older Japanese with lower BMI had a significantly higher risk of all cause mortality in both sexes. Not only obesity, but also underweight prevention is needed for older Japanese.