Establishment and internal-external validation of a clinical-adipometric prognostic model for heart failure with preserved ejection fraction

Computed tomography (CT) is the reference standard for quantitative body composition assessment. Heart failure with preserved ejection fraction (HFpEF) carries a poor prognosis, but practical risk prediction tools remain limited. This study aimed to develop and validate a prognostic nomogram integrating CT‑derived body composition metrics with clinical characteristics for predicting the composite endpoint in HFpEF patients. This multicenter retrospective cohort included 459 HFpEF patients from two tertiary hospitals (January 2021-March 2023). From one center, 328 patients were randomly split 7:3 into training ( n = 232) and internal validation ( n = 96) cohorts; the other 131 patients from the second center formed the external validation cohort. The composite endpoint was defined as all-cause mortality or heart-failure-related rehospitalization. Body composition parameters were extracted from T12 vertebral level CT images. Independent prognostic factors were identified via LASSO and multivariable Cox regression, then incorporated into a nomogram. Model performance was assessed by time‑dependent ROC curves, calibration plots, restricted cubic spline (RCS), decision curve analysis (DCA), and Kaplan-Meier analysis with log‑rank tests. Among 459 patients (50.11% male; mean age 61.30 ± 10.93 years), the composite endpoint occurred in 252 (54.90%) over a median follow‑up of 50.26 months. Five independent predictors were selected: fasting plasma glucose, intermuscular adipose tissue index (IMATI), logBNP, total cholesterol, and skeletal muscle index (SMI). The nomogram achieved AUCs for 2‑ and 3‑year predictions of 0.728 and 0.772 in the training cohort; 0.714 and 0.762 in internal validation; and 0.724 and 0.769 in external validation. Calibration and DCA demonstrated good fit and net benefit. RCS showed a positive association between IMATI and the composite endpoint risk and a negative association for SMI. Kaplan-Meier curves revealed significant prognostic stratification (high IMATI or low SMI associated with worse survival; low nomogram scores with better outcomes). SMI and IMATI were independent predictors. The nomogram showed favorable discriminative ability compared with the MAGGIC risk score. In conclusion, the proposed nomogram incorporating CT‑based body composition metrics and clinical factors provides robust the composite endpoint prediction in HFpEF. If prospectively validated, this tool may facilitate risk stratification and aid in identifying high‑risk patients for more intensive monitoring in future clinical practice.