Predictors of incident heart failure in people with and without type 2 diabetes free from baseline cardiovascular disease: a UK Biobank study

Background Heart failure (HF) remains a major cause of morbidity and mortality, especially in individuals with type 2 diabetes (T2D). Our ability to identify at-risk individuals, especially those without known cardiovascular disease (CVD), is still limited. Purpose The purpose of this study was to identify early indicators of new-onset HF hospitalisation (HFH) in adults free of CVD and to determine whether these indicators differed in individuals with T2D. Methods Adults from the UK Biobank without CVD at baseline were included. Thirty-six clinically relevant predictors (demographic, medical, biochemical, phenotypic, and lifestyle) were selected a priori. Microalbuminuria was log10-scaled, and diabetes status was stratified by glucose-lowering therapy. Ethnicity, chronotype, alcohol frequency, and walking pace were categorised using standard UK Biobank fields. The apolipoprotein A:B ratio was derived to capture atherogenic balance. Flexible parametric survival models were used to estimate associations with first HFH, with sequential inclusion of sociodemographic, biochemical, phenotypic, and lifestyle domains. Model fit was assessed using AIC and BIC. Analyses were repeated in the sub-group of T2D-only diabetes with duration and diabetes medication included as additional predictors. Results Over a median follow-up of 11 years, in 443244 adults without CVD (58 ± 7years, 58% male, 96% white-European, 4%T2D) HFH incidence was over 2.5-fold higher in individuals with T2D (0.8% vs. 2.4%) (Table 1). In the full cohort, 20 predictors were independently associated with HFH, including traditional risk factors, liver enzymes, lifestyle behaviours, and microalbuminuria. Microalbuminuria was the strongest predictor (HR 2.83 [95%CI 2.59,3.09], p<0.001), conferring a nearly three-fold increase in risk per 10-unit increment. In the T2D subgroup, five independent predictors identified the full cohort (e.g., use of beta-blockers, SES, hsCRP, haemoglobin and handgrip) were not found to be important for predicting risk (Figure). Instead, the apolipoprotein A:B ratio emerged as a key independent predictor, with higher ApoA:B ratios associated with lower HFH risk (HR 0.56 [0.33,0.95], p=0.001). In the final model, for both whole cohort and T2D-only, alcohol frequency and walking pace were independent predictors of risk with protective properties. Conclusion Early HF risk in CVD-free adults is driven by a combination of traditional, metabolic, and lifestyle factors, while in T2D, lipoprotein balance emerged as a significant clinical predictor with lifestyle factors remaining important. These findings highlight opportunities for earlier, targeted HF prevention and requirement for T2D-specific risk stratification.