The inclusion of cancer in heart failure risk assessment tool for patients with type-2 diabetes mellitus: a precision medicine approach

Abstract Introduction Cancer survivors are at increased risks of developing heart failure (HF) due to shared risk factors and cancer treatment-induced cardiac dysfunction. However, current HF risk assessment tools often overlook cancer history, potentially missing early intervention opportunities. Including cancer risk into HF risk models provides a more defined approach, improving personalised care. Aim To integrate cancer history into HF risk assessment tool for improved early detection in patients with T2DM. Methods We obtained electronic medical records (EMRs) from a diverse type-2 diabetes mellitus (T2DM) Chinese cohort, including demographics, laboratory measurements, medications and prior comorbidities. Prevalent cancer and incident HF diagnoses were determined based on the International Classification of Diseases (ICD) Ninth version codes. Cox regression and time-dependent machine-learning models were implemented for patients with and without prevalent cancer. State-of-the-art artificial intelligence interpretability with clinical expertise were used to provide a support tool able to predict the risk of HF in a population with an established cancer. Results Out of 262,687 patients with T2DM aged 50 years, 8,515 (3%) had incident HF between 2009-2019. 11,433 (4%) had a prior history of cancer. The prevalent cancer inclusion model c-statistic score was 0.91 and the brier score was 0.02 for predicting incident HF. The key factors in order are shown in Table 1: use of insulin, estimated glomerular function ratio (eGFR), use of loop diuretics, history of cancer and coronary artery disease (CAD) (myocardial infarction and angina). Lower levels of haemoglobin, serum albumin, lymphocytes, higher levels of alkaline phosphatase (ALP) and potassium were the main factors associated with incident HF. The model that excluded prior cancer history showed decreased in performance (c-statistic: 0.88 and brier: 0.03). Individuals with a history of cancer had a higher risk of incident HF compared to those who did not (HR: adjusted for age and sex 1.27 95% CI 1.15 – 1.40 (0.005)). Conclusion Including cancer history into HF risk assessment enhances prediction accuracy, highlighting its importance in personalised care for T2DM patients.Table 1:Cancer History in HF Risk