Obesity and Mortality in a Pooled Analysis of Three Prospective Cohorts of Korean Adults: Is Obesity Paradox the BMI Paradox?

OBJECTIVES: It is unclear whether the obesity paradox (observation of lower mortality in overweight vs. normal-weight adults) consistently exists in Asian population after accounting for body composition (fat vs. fat-free mass). We examined the associations of body mass index (BMI) with mortality in Korean adults while carefully controlling for various sources of biases, including those from body composition. METHODS: We conducted a pooled analysis of three prospective cohort studies from the Korean Genome and Epidemiology Study Project (n = 153,248; mean age = 53.8 years; mean follow-up = 9.6 years; number of total death = 6,061). Weight and height were directly measured at baseline. Fat-free mass index (fat-free mass/height squared) was derived using a validated prediction model. Using Cox proportional hazards models, we estimated multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for the associations of BMI with all-cause, cardiovascular, and cancer mortality. The nonlinear shape of the association was assessed using restricted cubic spline models. We compared the models after excluding older adults (age >70 years), ever smokers, and deaths occurred during the first 5 years of follow-up. We also compared the models after further adjustment for fat-free mass index. RESULTS: Among participants without a history of cancer or cardiovascular disease, the relation between BMI and all-cause mortality was J-shaped. Compared with normal-weight adults (BMI 18.5–22.9), the lowest risk of all-cause and cardiovascular mortality was observed among overweight adults (BMI 23.0–24.9: HR [95% CI] = 0.89 [0.83–0.96]; 0.94 [0.79–1.13]; respectively), while for cancer mortality among grade I obesity (BMI 25.0–29.9: 0.91 [0.83–1.01]). After excluding older adults, ever smokers, and deaths occurred during the first 5 years of follow-up, the overall patterns of the associations did not change. After adjustment for fat-free mass index, the J-shaped relation changed into a monotonically positive association, showing the lowest mortality among underweight and normal-weight adults. CONCLUSIONS: Our findings suggest that the obesity paradox in Asian population may be at least partially explained by confounding due to fat-free mass. FUNDING SOURCES: This work was supported by National Research Foundation of Korea (2019R1G1A1004227, 2019S1A3A2099973).