Prediction of Incident Heart Failure in Men and Women with a History of Myocardial Infarction
Phoebe Chan, Thomas F Kok, Robert M A van der Boon, Navin Suthahar, Rudolf A de Boer, Eric Boersma, Isabella KardysAbstract
Aims
Individuals with prior myocardial infarction (MI) show increased risk of heart failure (HF), yet current risk prediction models leave room for improvement. We aimed to develop and validate a clinically viable prediction model for incident HF in persons with prevalent MI in a general population, and explore sex- and MI subtype-specific differences.
Methods
We analysed UK Biobank participants with prior MI but absence of HF at baseline. Sociodemographic factors, clinical variables, and 59 blood biomarkers were included. The primary endpoint was the first in-hospital HF diagnosis. Backward selected Cox proportional hazards models were used to identify independent predictors of incident HF. Model performance was assessed via discrimination (C-index), using internal- and hold-out validation, and calibration.
Results
A total of 4,743 participants with prevalent MI; 81.4% male, median (P25-P75) age 62(58–66) years, were included. During a median follow-up of 12.1(11.1–13.0) years, 767(16.2%) developed HF. Sixteen independent predictors were identified and eleven remained significant (P<0.05) after treating all-cause mortality as a competing risk. These included several well-established predictors (age, body mass index, smoking, atrial fibrillation, diabetes) as well as haemoglobin (HR per 1 standard deviation (SD) increase: 0.88(95%CI:0.79-0.96)), mean reticulocyte volume (HR: (1.13(95%CI:1.03-1.23)), and neutrophil percentage (HR:1.12(95%CI:1.03-1.21)), monocyte count (HR:1.11(95%CI,1.06-1.16)). A trend towards an interaction(P<0.1) between sex and MI subtype was present. Model discrimination was modest (C-index=0.67) and calibration was adequate.
Conclusion
Our study identifies several clinically accessible blood biomarkers as important risk factors for HF in post-MI individuals, and suggests interactions between sex and MI subtype. Further model refinement and external validation are needed.