Abstract Background The trajectories of anthropometric and body composition measures (important predictors of diabetes) are rarely explored before diabetes diagnosis. Our study aimed to compare trajectories of fat mass (FM), fat-free mass (FFM), body mass index (BMI), and waist circumference (WC) preceding type 2 diabetes mellitus (T2DM) to aging trajectories of individuals without diabetes during follow-up. Methods We used data from the Whitehall II study, a prospective cohort of British civil servants. 5-yearly BMI and WC were available for up to 20 years, while 5-yearly FM and FFM measures were available for up to 10 years. Linear mixed models with a backward timescale (from diabetes diagnosis or end of follow-up) were performed stratified by sex and adjusted for age, occupational grade, ethnicity, and lifestyle factors. Results A total of 1674/990 (anthropometric/body composition analysis) women (233/81 incident diabetes) and 3917/2710 men (479/217 incident diabetes) 49.81 [0.08]/60.91 [0.9] (mean [SE]) years of age at baseline were included. All outcomes were higher in cases compared to controls. Women’s FM, BMI, and WC followed a quadratic increase in both groups with a faster increase among incident diabetes cases (dBMI 0.04 [0.01] kg/m2/year, dFM 0.19 [0.09] kg/year, dWC 0.2 cm/year [0.05]). FFM decreased linearly with similar slopes in cases and controls. Men’s FM, BMI, and WC also showed a quadratic increase with faster increase in incident cases compared to controls (dBMI 0.03 [0.01] kg/m2/year, dFM 0.23 [0.04] kg/year, dWC 0.08 [0.02] cm/year). FFM followed a quadratic decrease in both groups with a slower rate (0.06 [0.03] kg/year) in incident cases. Conclusions Incident diabetes cases have higher anthropometric and body composition measures 10-20 years before diabetes diagnosis compared to controls. Furthermore, incident cases showed faster increases in these measures except for FFM that decreased during follow-up with similar or lower speeds in cases than controls. Key messages • Traditional anthropometric measures do not capture the underlying changes in body composition. • The inclusion of body composition in risk calculators may result in more precise diabetes prediction.