Background Body mass index (BMI) is the most frequently used adiposity measure, yet it is unable to differentiate fat mass from lean mass. Relative fat mass (RFM) has been proposed as alternative. The aim of this paper is to study RFM and BMI association with mortality in a general Italian population and potential mediators. Methods 20,587 individuals from the Moli-sani cohort were analysed (mean age = 54 ± 11, 52% women, median follow up = 11.2years, interquartile range 1.96 years). Cox

more »

... ions were used to assess BMI, RFM and their interactive association with mortality. Dose-response relationships were assessed with spline regression, mediation analysis was performed. All analyses were separated for men and women. Results Men and women with BMI > 35 kg/m2 and men in the 4th quartile of RFM showed an independent association with mortality (HR = 1.71, 95%CI = 1.30–2.26 BMI in men, HR = 1.37, 95%CI = 1.01–1.85 BMI in women, HR = 1.37 CI95%=1.11–1.68 RFM in men), that was lost once adjusted for potential mediators. Cubic splines showed a U-shaped association for BMI in men and women, and for RFM in men. Mediation analysis showed that 46.5% of the association of BMI with mortality in men was mediated by glucose, C reactive protein, forced expiratory volume in 1 second (FEV1) and cystatin C; 82.9% of the association of BMI in women was mediated by HOMA index, cystatine C and FEV1; lastly 55% of RFM association with mortality was mediated by glucose, FEV1 and cystatin C. Regression models including BMI and RFM showed that RFM drives most of the risk in men, but is not predictive in women. Conclusions The association between anthropometric measures and mortality is U shaped and it is largely dependent on sex. Associations were mediated by glucose metabolism, renal and lung function. Public health should mainly focus on severely obese people, or people with impaired metabolic, renal, or respiratory function.