Exploratory clinical predictors and internally validated models of in-hospital mortality after PCI in obese patients with acute myocardial infarction and documented cardiogenic shock: a single-center retrospective study

Obese patients with acute myocardial infarction complicated by cardiogenic shock (AMI-CS) remain at high risk of early mortality despite percutaneous coronary intervention (PCI). Obesity and related cardiometabolic disorders may aggravate vascular and microcirculatory dysfunction, potentially involving impaired endothelial cell function. However, prognostic evidence specifically focusing on obese patients with AMI-CS undergoing PCI remains limited. This exploratory, single-center study investigated clinical factors associated with in-hospital mortality and compared the predictive performance of logistic regression and eXtreme Gradient Boosting (XGBoost) using rigorous internal validation. The archived analytic dataset contained 268 patients. After eligibility re-audit, three patients with body mass index (BMI) < 28 kg/m² were excluded, leaving 265 patients with documented AMI-CS treated with PCI. Seven routinely available candidate variables were evaluated in multivariable association and prediction analyses. Logistic regression and regularized XGBoost were compared using 20 repetitions of stratified five-fold nested cross-validation, with all model tuning confined to the inner training folds. Supplemental data on shock severity, serial lactate levels, vasoactive therapy, mechanical circulatory support, ventilation, renal replacement therapy, and hemodynamic variables were additionally evaluated descriptively and in a post hoc sensitivity analysis. Among 265 patients, 202 survived and 63 died during hospitalization. Higher heart rate, hypertension, diabetes mellitus, malignant arrhythmia, and multivessel disease were associated with in-hospital mortality. The seven-variable logistic regression model achieved an area under the receiver operating characteristic curve (AUC) of 0.767 (95% CI, 0.696–0.831) and a Brier score of 0.152, compared with an AUC of 0.712 (95% CI, 0.641–0.779) and a Brier score of 0.169 for regularized XGBoost. Serial lactate concentrations at admission, peak, 6 h, 12 h, and 24 h were consistently higher in non-survivors, and cardiac power output was modestly lower. A post hoc severity-augmented logistic model achieved an AUC of 0.922 (95% CI, 0.888–0.951), but its calibration slope was 2.077, indicating that the model was not sufficiently calibrated for clinical deployment. In obese patients with AMI-CS undergoing PCI, heart rate, hypertension, diabetes mellitus, malignant arrhythmia, and multivessel disease were associated with in-hospital mortality. Logistic regression demonstrated more favorable internally validated predictive performance than regularized XGBoost using the same routinely available predictors. The associations of diabetes and hypertension, together with abnormalities in systemic perfusion, are biologically compatible with a potential contribution of endothelial cell dysfunction and impaired microvascular regulation; however, endothelial function was not directly measured in this study. External multicenter validation and studies incorporating endothelial biomarkers are warranted to further clarify the role of endothelial dysfunction in risk stratification for this high-risk population. These findings are exploratory and internally validated only; the modest sample size, absence of external validation, and originally outcome-informed nomination of the seven candidate predictors limit generalizability and preclude claims of clinical readiness.