Exploratory Modeling of Postoperative Atrial Fibrillation After Cardiac Surgery with Cardiopulmonary Bypass Using Inflammatory Biomarkers and Clinical-Surgical Factors

Figures

• 

  Figure 1. Methodology flowchart. The upper diagram outlines the patient enrollment process, types of cardiac surgery, and selection criteria for assembling the patient cohort. The bottom diagram details the sampling timeline throughout the surgical process and the inflammatory biomarker analysis. CABG: Coronary Artery Bypass Grafting; Valve: Valve replacement surgery; Combined: CABG-Valve surgery; AF: Atrial Fibrillation; T1: Patient plasma sampling at preoperatory; T2: Patient plasma sampling at 24 h postoperative; T3: Patient plasma sampling at 48 h postoperative; IL-6: Interleukin-6; IL-8: Interleukin-8; IL-10: Interleukin-10; CRP: C-Reactive Protein.

• 

  Figure 2. Perioperative distributions of inflammatory biomarkers by postoperative atrial fibrillation status. Box plots show log2-transformed concentrations of (A) IL-8, (B) IL-10, (C) IL-6, and (D) CRP measured preoperatively (T1) and at 24 h (T2) and 48 h (T3) after surgery. Boxes represent the interquartile range, horizontal lines indicate the median, whiskers extend to 1.5 times the interquartile range, and points represent individual observations. Postoperative plots include only patients who remained free of POAF before blood collection at each landmark time. CRP = C-reactive protein; IL = interleukin; POAF = postoperative atrial fibrillation.

• 

  Figure 3. Receiver operating characteristic curve for the final postoperative atrial fibrillation model. The model included log2-transformed IL-10 and magnesium concentrations at 24 h and was fitted using Firth’s bias-reduced logistic regression. The solid line represents the receiver operating characteristic curve, and the shaded area represents the 95% confidence band. The apparent area under the curve was 0.776 (95% CI: 0.604-0.948). AUC = area under the curve; CI = confidence interval; IL-10, interleukin-10.

• 

  Figure 4. Calibration plot for the final postoperative atrial fibrillation model. Predicted probabilities are plotted against observed POAF frequencies. The diagonal line indicates perfect calibration; the solid line is the fitted logistic calibration curve, and the dotted line is the nonparametric calibration estimate. Rug marks show the distribution of predicted probabilities. Calibration metrics indicate good agreement between predicted and observed probabilities. POAF = postoperative atrial fibrillation.

• 

  Figure 5. Decision curve analysis of the final postoperative atrial fibrillation model. Net benefit is shown across a range of predicted-probability thresholds. The solid line represents the final model, the dashed line the treat-all strategy, and the horizontal dotted line the treat-none strategy. The model showed greater net benefit than the reference strategies at thresholds of approximately 0.10 to 0.40.

• 

  Figure 6. Receiver operating characteristic curves for the preoperative and postoperative 24-h models. The T1 model included baseline hemoglobin and IL-10 and had an apparent AUC of 0.711 and a bootstrap optimism-corrected AUC of 0.695. The T2 model included IL-10 and magnesium measured at 24 h and had an apparent AUC of 0.776 and a bootstrap optimism-corrected AUC of 0.762. The apparent AUCs did not differ significantly by DeLong’s test (p = 0.526). AUC, area under the curve; IL-10, interleukin-10.

• 

  Figure 7. Receiver operating characteristic curves for predictor-domain models. Predictor selection was repeated within each domain using 1000 bootstrap replicates. The bootstrap optimism-corrected AUC was 0.739 for the inflammatory biomarker domain (IL-6, IL-8, and IL-10 at T2), 0.682 for the clinical-surgical domain (hypothyroidism), 0.652 for the complete blood count domain (hematocrit and platelet count at T2), and 0.639 for the blood chemistry domain (chloride at T2). AUC = area under the curve; IL = interleukin.