AbstractObjective: To investigate whether the association between BMI and all‐cause mortality could be disentangled into opposite effects of body fat and fat‐free mass (FFM).Research Methods and Procedures: All‐cause mortality was studied in the Danish follow‐up study “Diet, Cancer and Health” with 27, 178 men and 29, 875 women 50 to 64 years old recruited from 1993 to 1997. By the end of year 2001, the median follow‐up was 5.8 years, and 1851 had died. Body composition was assessed by bioelectrical impedance. Cox regression models were used to estimate the relationships among body fat mass index (body fat mass divided by height squared), FFM index (FFM divided by height squared), and mortality. All analyses were adjusted for smoking habits.Results: Men and women showed similar associations. J‐shaped associations were found between body fat mass index and mortality adjusted for FFM and smoking. The mortality rate ratios in the upper part of body fat mass were 1.12 per kg/m2 (95% confidence interval: 1.07, 1.18) in men and 1.06 per kg/m2 (95% confidence interval: 1.02, 1.10) in women. Reversed J‐shaped associations were found between FFM index and mortality with a tendency to level off for high values of FFM.Discussion: Our findings suggest that BMI represents joint but opposite associations of body fat and FFM with mortality. Both high body fat and low FFM are independent predictors of all‐cause mortality.