Joint longitudinal trajectories of the triglyceride-glucose index combined with BMI and waist-to-height ratio and incident cardiovascular disease: a prospective cohort study from the English longitudinal study of ageing

The triglyceride-glucose (TyG) index and its composite obesity indices have been linked to cardiovascular disease (CVD) risk. However, most prior studies relied on single baseline measurements, and few have employed group-based multi-trajectory modeling to capture concurrent longitudinal changes in metabolic and anthropometric indicators. This study aimed to identify joint longitudinal trajectory groups of TyG combined with body mass index (BMI) and waist-to-height ratio (WHtR) using parallel approaches and evaluate their associations with incident CVD in middle-aged and older adults. This prospective cohort study included 1808 CVD-free participants aged ≥ 50 years from the English Longitudinal Study of Ageing. Group-based multi-trajectory modeling was applied to jointly identify latent trajectory classes using repeated measurements of TyG with BMI and TyG with WHtR across three waves over approximately 8 years. Cox proportional hazards models with four sequential adjustment models estimated hazard ratios (HRs) and 95% confidence intervals (CIs) for incident composite CVD, heart disease, and stroke. Subgroup analyses were stratified by age, gender, smoking, diabetes, and hypertension status. Nine sensitivity analyses were conducted to assess robustness. During follow-up, 263 participants (14.5%) developed incident CVD. Four trajectory groups were identified for each approach. In fully adjusted BMI + TyG models, compared with the normal weight–low TyG reference group, composite CVD risk increased progressively across the overweight–moderate TyG (HR 2.31, 95% CI 1.40–3.83), obese–high TyG (HR 2.72, 95% CI 1.63–4.52), and severely obese–high TyG groups (HR 5.06, 95% CI 3.01–8.51). The WHtR + TyG approach demonstrated a consistent dose–response pattern, with HRs of 2.22, 2.75, and 5.12 for ascending risk groups. For stroke, only the highest-risk groups reached statistical significance (HR 6.74 and 5.00, respectively). Formal discriminative comparison showed no significant difference between the two approaches (C-statistic difference 0.003, P = 0.723). All nine sensitivity analyses consistently corroborated the primary findings. Both approaches yield robust and comparable dose–response gradients, supporting further validation of serial TyG-related composite index monitoring for cardiovascular risk stratification in aging populations.