The U-shaped relationship between body mass index (BMI) and all-cause mortality has generated uncertainty about optimal BMI. For clarification, we have related BMI to both mortality and medical expenditure. The MJ Health examination cohort of 111,949 examinees established during 1994-1996 was followed with endpoint information derived from death certificates and National Health Insurance records from 1996 to 2007. Age- and gender-specific relative risks between BMI groups were estimated by Cox and logistic regressions. The BMI and all-cause mortality relationship is U-shaped with the concave regions sitting in the region of BMI 22-26, butshifted rightward for the elderly. After excluding smokers and cancer patients at baseline, the low mortality region moved leftward to BMI 20-22. Cause-specific mortalities from respiratory disease, injury, and senility increased in the underweight group (BMI <18.5). Above 18.5, BMI was negatively associated with mortality from respiratory diseases and senility, but not with others. In contrast, irrespective of age and gender, the overall median and mean medical expenditures progressively increased with BMI, particularly beyond 22. Expenditures for injury, respiratory, circulatory diseases and senility all increased with BMI. The U-shaped BMI-mortality relation was a result of elevated death rate at both ends of the BMI scale. Increased mortality at the low end did not contribute to higher medical expenditure, maybe because the lean and frail deceased tend to die abruptly before large amount of medical expenditure was consumed. Our findings suggest that current recommendations to maintain BMI at the lower end of the desirable range remain tenable for the apparently healthy general public.