A Firth Logistic Regression Analysis of Cardiac Abnormalities and Inflammatory Biomarkers in Peadiatric HIV Patients: A Cross-Sectional Study

Abstract

ABSTRACT Cardiac abnormalities are increasingly recognized among paediatric HIV patients and are thought to be driven partly by chronic inflammation. Standard logistic regression may yield biased estimates in rare-event outcomes. This study applied Firth’s penalized logistic regression to identify predictors of cardiac abnormalities and improve inferential robustness. A cross-sectional study of 126 paediatric HIV patients was conducted. Cardiac abnormalities were defined using a composite of ECG and echocardiography findings. Agreement between ECG and ECHO was assessed using Cohen’s kappa. Predictors including C-reactive protein (CRP), troponin, BMI, social class, age, sex, and ART duration were analyzed using Firth logistic regression. Model performance was evaluated using AIC and BIC. The prevalence of cardiac abnormalities was 29.4% (95% CI: 21.9%–38.0%). ECG and ECHO showed moderate agreement (κ = 0.43, p < 0.001). The Firth model demonstrated superior fit (AIC = 95.56; BIC = 115.42) compared with standard logistic regression. Elevated CRP (OR = 5.42, p = 0.039), elevated troponin (OR = 11.50, p = 0.002), and overweight BMI (OR = 27.60, p < 0.001) were strong predictors of cardiac abnormalities, while higher social class was protective. Inflammatory biomarkers and obesity are major predictors of cardiac abnormalities in paediatric HIV patients. Firth logistic regression provides a robust framework for modelling rare cardiac outcomes.