Development and Validation of a Nomogram to Predict Ventricular Fibrillation During Percutaneous Coronary Intervention in Patients With Acute Myocardial Infarction

Abstract

Background: Ventricular fibrillation (VF) is a life-threatening complication of acute myocardial infarction (AMI), particularly in patients undergoing percutaneous coronary intervention (PCI). Early identification of high-risk patients is crucial for implementing preventive measures and improving outcomes.

Methods: This retrospective study analyzed clinical, laboratory, and angiographic data from 155 AMI patients to identify predictors of VF during PCI. Variable selection was performed using least absolute shrinkage and selection operator (LASSO) regression, elastic net regression, and random forest. Independent predictors were identified through multivariable logistic regression, and a nomogram was developed and validated to predict VF risk. Model performance was assessed using receiver operating characteristic (ROC) and calibration curves.

Results: Independent predictors of VF included diabetes (OR = 3.676 (1.365-10.668); p = 0.012), neutrophil-to-lymphocyte ratio (NLR) (odds ratio (OR) = 1.149 (1.053-1.265); p = 0.002), right coronary artery (RCA) intervention (OR = 3.185 (1.088-9.804); p = 0.037), Gensini score (OR = 1.020 (1.007-1.033); p = 0.003), and absence of beta blockers (OR = 0.168 (0.054-0.472); p = 0.001). The nomogram, incorporating these predictors, demonstrated a strong discriminative ability with an area under the ROC curve (AUC) of 0.882 (0.825-0.939) and good calibration (Hosmer-Lemeshow test, p = 0.769). The calibration curve showed a strong alignment between predicted probabilities and observed outcomes, with a mean absolute error of 0.033.

Conclusions: This study identified diabetes, NLR, RCA intervention, Gensini score, and absence of beta-blocker use as key predictors of VF during PCI in AMI patients. A nomogram incorporating these factors showed strong predictive performance, aiding clinicians in identifying high-risk patients for targeted preventive strategies.

Keywords: acute myocardial infarction (AMI); nomogram; percutaneous coronary intervention (PCI); ventricular fibrillation (VF).

Fig. 1.

Flowchart of this retrospective case-control study. ROS, receiver operating characteristic; LASSO, least absolute shrinkage and selection operator.

Fig. 2.

Variable selection methods—elastic net, random forest mean decrease accuracy, and LASSO. This figure presents the results of three variable selection methods—LASSO regression, elastic net regression, and random forest—used to identify significant predictors of ventricular fibrillation (VF). In the LASSO regression analysis (A,B), the left panel (A) illustrates the coefficient paths for variables as the regularization parameter (log lambda) changes. As lambda increases, more coefficients shrink to zero, leaving only the most important predictors, ensuring model simplicity while retaining predictive accuracy. The right panel (B) shows the cross-validation errors for different lambda values, with the optimal lambda (indicated by the dashed line) minimizing the binomial deviance. Elastic net regression results (C) are displayed as a bar plot of regression coefficients, where positive coefficients indicate variables associated with an increased risk of VF (e.g., hypokalemia and intervention on the RCA), while negative coefficients represent protective factors (e.g., beta-blockers and aspirin). Elastic net combines the strengths of LASSO and ridge regression, allowing for the retention of correlated variables while reducing overfitting. The random forest analysis (D) ranks the top 30 variables based on their importance, measured by the mean decrease in accuracy. Variables such as hypokalemia, aspirin, and ventricular tachycardia are identified as the most important predictors, highlighting their strong influence on VF risk. RCA-related ventricular fibrillation was included, it ranked 42nd in the random forest plot but was ranked higher in two other methods. Random forest is particularly valuable for capturing complex, non-linear relationships and interactions between variables, making it a powerful complement to regression-based methods. Together, these three approaches provide a robust framework for identifying and prioritizing predictors of VF. RCA, right coronary artery; NLR, neutrophil-to-lymphocyte ratio; MAP, mean arterial pressure; LDL-C, low-density lipoprotein cholesterol; PT, prothrombin time; AST, aspartate aminotransferase; CO₂, carbon dioxide; ARB, angiotensin II receptor blockers; ACEI, angiotensin-converting enzyme inhibitors; Beta blocker, beta-adrenergic blocking agents; FDP, fibrin degradation products; LASSO, least absolute shrinkage and selection operator; CCB, calcium channel blocker therapy; AV, atrioventricular; APTT, activated partial thromboplastin time; hs-CRP, high-sensitivity C-reactive protein.

Fig. 3.

A nomogram was constructed to facilitate the prediction of VF during PCI. This nomogram serves as a visual tool designed to predict the probability of a specific outcome, such as the risk of a clinical event. It combines multiple predictors into a single scoring system, enabling individualized risk assessment. Each predictor variable is represented on a separate scale, with its corresponding value mapped to a “Points” scale at the top. The variables included in this nomogram are as follows: Intervention on RCA: Indicates whether an intervention was performed on the RCA. Binary values (0 or 1) contribute different points to the total score. Diabetes: Represents the presence or absence of diabetes, with binary values (0 or 1) contributing to the score. NLR: The ratio of neutrophils to lymphocytes, ranging from 0 to 35. Higher values contribute more points, reflecting increased systemic inflammation. Gensini score: A measure of coronary artery disease severity, ranging from 20 to 240. Higher scores indicate more severe disease and contribute more points. Beta blocker: Indicates whether a beta-blocker is being used, with binary values (0 or 1) contributing to the score. To use the nomogram, the value of each variable is located on its respective scale, and the corresponding points are determined by projecting upward to the “Points” scale. The points for all variables are then summed to calculate the Total Points, which are mapped to the linear predictor and subsequently to the predicted probability at the bottom of the nomogram. RCA, right coronary artery; PCI, percutaneous coronary intervention; NLR, neutrophil-to-lymphocyte ratio; VF, ventricular fibrillation.

Fig. 4.

ROC curve was employed to validate the nomogram. This figure presents the ROC curve for the nomogram model, illustrating its diagnostic performance. The curve demonstrates a strong ability to distinguish between positive and negative outcomes, with an area under the ROC curve (AUC) of 0.882 (95% CI: 0.825–0.939). This high AUC value indicates excellent predictive accuracy, as the model achieves a good balance between sensitivity (true positive rate) and specificity (1-false positive rate). The curve’s proximity to the top-left corner further highlights the model’s robust discriminative power. ROC, receiver operating characteristic.

Fig. 5.

A calibration curve was employed to validate the nomogram. This calibration plot evaluates the agreement between the predicted probabilities and the observed outcomes for the nomogram model. The red dashed line represents the ideal calibration line, where predictions perfectly match the observed probabilities. The black solid line shows the bias-corrected performance of the model after 40 bootstrap repetitions, while the dotted line represents the apparent performance without correction. The close alignment of the bias-corrected line with the ideal line indicates good calibration, suggesting that the model’s predictions are reliable. The mean absolute error of 0.033 further supports the model’s accuracy in predicting outcomes.