Hajin Jang,¹ Rockli Kim,^1,2 Jong-Tae Lee,^1,2 Dong Hoon Lee,³ Edward L. Giovannucci^3,4 and Hannah Oh^1,2* 

¹Interdisciplinary Program in Precision Public Health, Department of Public Health Sciences, Graduate School of Korea University, Seoul, Republic of Korea
²Department of Health Policy and Management, College of Health Sciences, Korea University, Seoul, Republic of Korea
³Department of Nutrition, T.H. Chan Harvard School of Public Health, Boston, MA, USA 
⁴Department of Epidemiology, T.H. Chan Harvard School of Public Health, Boston, MA, USA 

^*Corresponding author: Hannah Oh

Background: Studies found a J-shaped association between body mass index (BMI) and mortality. However, it is unclear whether the association is driven by biases, particularly confounding by fat-free mass.

Methods: We conducted an individual-level pooled analysis of three cohorts of Korean adults (aged ≥ 40 years; n = 153 248). Mortality was followed up through December 2019. Anthropometric data were directly measured at baseline. Fat and fat-free mass were predicted using validated prediction models. Using Cox proportional hazards models, we estimated the associations of BMI and waist circumference (WC) with all-cause and cause-specific mortality. To account for biases, we excluded participants aged ≥ 70 years, deaths that occurred within 5 years of follow-up and ever smokers, and adjusted for fat-free mass index (FFMI).

Results: During the follow-up of up to 18 years, 6061 deaths were identified. We observed J-shaped association of BMI (nadir at 22-26) and monotonically positive association of WC with all-cause, cardiovascular, and cancer mortality among Korean adults without a history of cancer or cardiovascular disease. In the BMI analysis, excluding ever smokers and adjusting for FFMI attenuated the excess mortality in underweight participants and transformed the J-shaped association into a monotonically positive shape, suggesting an increased mortality at BMI > 22.0. Excluding participants aged ≥ 70 years and deaths that occurred within 5 years of follow-up did not change the results. In the WC analysis, the monotonic positive associations did not change after the control. Similar results were observed among participants with a history of cancer or cardiovascular disease.

Conclusions: Our data suggest that both overall and abdominal body fat are associated with increased mortality in Korean adults