Objective According to most prospective studies, being underweight (BMI<18·5 kg/m2) is associated with significantly higher mortality than being of normal weight, especially among smokers. We aimed to explore in a generally lean population whether being underweight is significantly associated with increased all-cause mortality. Design Prospective cohort study. Setting Korea Medical Insurance Corporation study with 14 years of follow-up. Subjects After excluding deaths within the first 5 years of follow-up (1993-1997) to minimize reverse causation and excluding participants without information about smoking and health status, 94 133 men and 48 496 women aged 35-59 years in 1990 were included. Results We documented 5411 (5·7 %) deaths in men and 762 (1·6 %) in women. Among never smokers, hazard ratios (HR) for underweight individuals were not significantly higher than those for normal-weight individuals (BMI=18·5-22·9 kg/m2): HR=0·87 (95 % CI 0·41, 1·84, P=0·72) for underweight men and HR=1·12 (95 % CI 0·76, 1·65, P=0·58) for underweight women. Among ex-smokers, HR=0·86 (95 % CI 0·38, 1·93, P=0·72) for underweight men and HR=3·77 (95 % CI 0·42, 32·29, P=0·24) for underweight women. Among current smokers, HR=1·60 (95 % CI 1·28, 2·01, P<0·001) for underweight men and HR=2·07 (95 % CI 0·43, 9·94, P=0·36) for underweight women. Conclusions The present study does not support that being underweight per se is associated with increased all-cause mortality in Korean men and women.