Abstract
<jats:p><p dir="ltr">This study aimed to identify integrated body-composition patterns and evaluate their sex-specific associations with incident type 2 diabetes (T2D). Among 258,500 UK Biobank participants free of diabetes at baseline, principal component analysis (PCA) was applied to 28 anthropometric, fat, lean, bone, and muscle-strength traits. Cox models assessed associations with incident T2D, and LASSO-penalized Cox and weighted quantile sum (WQS) regression identified key components. During a mean follow-up of 11 years, 33,636 incident T2D cases were identified. PCA yielded seven interpretable patterns. LASSO consistently identified generalized adiposity, leg-dominant lean distribution, and central leanness as robust components across sexes. Compared with the low-score reference group, generalized adiposity was associated with higher T2D risk in men and women, respectively (HR 1.73 [95% CI 1.33–2.24] and 2.21 [1.55–3.16]), whereas leg-dominant lean distribution was associated with lower risk (HR 0.57 [0.48–0.68] and 0.50 [0.40–0.63]). Central leanness was associated with lower risk in women (HR 0.41 [0.34–0.51]) and showed a non-monotonic association with risk in men. In WQS analyses, central leanness contributed more strongly to the inverse mixture association in men, whereas leg-dominant lean distribution contributed more strongly in women. These findings highlight sex-specific fat–lean configurations in diabetes risk.</p></jats:p>