Comparative prognostic value of simple, formula-derived, and biomarker-based adiposity indices for cardiovascular disease and mortality: a UK Biobank cohort study

Abstract Background Whether differences in adiposity–outcome associations translate into better prediction remains uncertain. We primarily compared the strength and patterns of associations of 12 adiposity-related indices with cardiovascular and mortality outcomes, and secondarily evaluated their discriminatory and incremental predictive performance. Methods We analysed a complete-case sample of 278,695 UK Biobank participants without baseline coronary heart disease (CHD) or stroke. Multivariable Cox models, restricted cubic splines, and time-varying analyses associations. The primary outcome was incident cardiovascular disease (CVD; CHD or stroke); component outcomes and all-cause and broadly defined CVD mortality were also examined. Discrimination was compared using five-fold cross-validated integrated time-dependent areas under the curve (iAUCs) over 1–17 years, with alternative indices evaluated in place of body mass index (BMI) within a common covariate framework. Incremental prediction was assessed using 10-year truncated Harrell’s C-index, integrated discrimination improvement (IDI), and continuous net reclassification improvement (NRI) after adding non-BMI indices to a Framingham risk-factor benchmark containing BMI and conventional cardiovascular risk factors. Results During a median 15.6 years, 27,947 incident CVD events and 24,464 deaths occurred. Waist-to-hip ratio (WHR) was positively associated with CVD, CHD, and stroke, with adjusted hazard ratios per standard deviation of 1.11 (95% CI 1.09–1.13), 1.12 (1.10–1.14), and 1.07 (1.03–1.10), respectively. Association patterns varied by outcome, and several CVD and CHD associations attenuated over follow-up. The highest observed iAUC gains over BMI ranged from 0.0001 to 0.0025 across outcomes. Weight-adjusted waist index yielded the highest observed total NRI for CVD and CHD, WHR for broadly defined CVD mortality, and A Body Shape Index for all-cause mortality. For CVD, adding weight-adjusted waist index yielded an NRI of 0.077 (95% CI 0.063–0.093), primarily attributable to the non-event component, with small changes in C-index (0.00076) and IDI (0.000455). Conclusions Association differences did not consistently translate into substantial discrimination gains, and no index was uniformly superior across outcomes. Selected non-BMI indices provided modest, outcome-specific incremental information, based on apparent performance within this cohort. External validation and assessment of calibration and clinical utility are needed before these findings can support clinical recommendations.