The Predictive Value of the Combined FVC/DLCO Ratio and 1-Minute Heart Rate Recovery After Exercise in Stable Chronic Obstructive Pulmonary Disease with Pulmonary Hypertension.

Objective: To evaluate the predictive value of combining the pulmonary function index forced vital capacity to diffusing capacity of the lung for carbon monoxide ratio (FVC/DLCO) with 1‑minute heart rate recovery (HRR1) in stable chronic obstructive pulmonary disease (COPD) complicated by pulmonary hypertension (PH), and to develop a machine learning‑based prediction model. Methods: A total of 159 stable COPD patients (COPD group) and 52 stable COPD patients with PH (COPD+PH group) were enrolled. Baseline data were collected. Predictive performance of FVC/DLCO and HRR1 alone and in combination (logistic regression) was assessed. Four machine learning models (Random Forest, XGBoost, LightGBM, SVM) were trained using all features. Results: The COPD+PH group showed significantly higher FVC/DLCO and lower HRR1. The combined FVC/DLCO+HRR1 model achieved an AUC of 0.781 (95% CI: 0.709-0.853), outperforming either marker alone. The Random Forest model performed best (AUC = 0.923, 95% CI: 0.881-0.961; accuracy 0.892; sensitivity 0.871; specificity 0.912), with good calibration. FVC/DLCO and HRR1 were the top two predictors. Conclusion: Combining FVC/DLCO with HRR1 enhances predictive accuracy for PH in stable COPD. The Random Forest model shows excellent performance, but external validation is needed before clinical implementation.