Multivariable Modeling of Postoperative Risk in Infant Cardiac Surgery: Integrating Clinical Variables and 20 Inflammatory Biomarkers
Abstract
Introduction
Cardiac surgery in infants often triggers a severe inflammatory response. The role of biomarkers in predicting clinical outcomes in this group of patients has been debated in the literature. This study aimed to investigate the predictive value of 20 inflammatory biomarkers, in combination with clinical data, for acute kidney injury, ventilator support duration, and inotropic score following infant cardiac surgery by developing and comparing three models: Clinical-Data-Only, Biomarker-Only, and Combined.
Methods
This secondary analysis of the MiLe-1 study included infants undergoing surgery with cardiopulmonary bypass. Biomarkers were measured before and after CPB. Using BIC-guided logistic regression, we developed and compared three multivariable models—Clinical-Data-Only, Biomarker-Only, and Combined—for each outcome. Model performance was assessed using c-statistics and p-contrast tests.
Results
Regarding AKI risk prediction, the c-statistics for Biomarker-Only, Clinical-Data-Only, and Combined Model were 0.79, 0.60, and 0.78 respectively. The difference in performance between the Combined and Clinical-Data-Only Models was statistically significant (p < 0.001). Concerning ventilator support time prediction, the c-statistics were 0.80, 0.72, and 0.77 for the models respectively (p-contrast = 0.10). As for inotropic score prediction, the c-statistics were 0.83, 0.77, and 0.85 for the models (p-contrast = 0.007).
Conclusion
Inflammatory biomarkers may enhance risk stratification for postoperative outcomes in infant cardiac surgery. However, given the exploratory nature of this study, further validation in larger and more diverse cohorts is needed.