Development and Validation of a Prediction Model for Cardiovascular Events in Exercise Assessment of Coronary Heart Disease Patients After Percutaneous Coronary Intervention

Abstract

Objective: This study aimed to develop a model for predicting cardiovascular events in the exercise assessment of patients with coronary heart disease after percutaneous coronary intervention (PCI) based on multidimensional clinical information.

Methods: A total of 2,455 post-PCI patients who underwent cardiopulmonary exercise testing (CPET) at the Peking University Third Hospital from January 2016 to September 2019 were retrospectively included in this study; 1,449 post-PCI patients from January 2018 to September 2019 were assigned as the development cohort; and 1,006 post-PCI patients from January 2016 to December 2017 were assigned as the validation cohort. Clinical data of patients before testing and various indicators in the exercise assessment were collected. CPET-related cardiovascular events were also collected, including new-onset angina pectoris, frequent premature ventricular contractions, ventricular tachycardia, atrial tachycardia, and bundle branch block during the examination. A nomogram model for predicting CPET-related cardiovascular events was further developed and validated.

Results: In the development cohort, the mean age of 1,449 post-PCI patients was 60.7 ± 10.1 years. CPET-related cardiovascular events occurred in 43 cases (2.9%) without fatal events. CPET-related cardiovascular events were independently associated with age, glycosylated hemoglobin, systolic velocity of mitral annulus, ΔVO₂/ΔWR slope inflection, and VE/VCO₂ slope > 30. The C-index of the nomogram model for predicting CPET-related cardiovascular events was 0.830, and the area under the ROC curve was 0.830 (95% CI: 0.764-0.896). For the validation cohort of 1,006 patients, the area under the ROC curve was 0.807 (95% CI: 0.737-0.877).

Conclusion: Post-PCI patients with older age, unsatisfactory blood glucose control, impaired left ventricular systolic function, oxygen uptake parameter trajectory inflection, and poor ventilation efficiency have a higher risk of cardiovascular events in exercise assessment. The nomogram prediction model performs well in predicting cardiovascular events in the exercise assessment of post-PCI patients and can provide an individualized plan for exercise risk prediction.

Keywords: cardiac rehabilitation; coronary heart disease; exercise risk; exercise safety; prediction model.

FIGURE 1

Coefficients of baseline variables in the Lasso regression. Each curve in the figure represented the trajectory of each independent variable coefficient. The ordinate was the value of the coefficient, and the lower abscissa was log(λ). The upper abscissa was the number of coefficients in the model. According to different λ values, individual coefficients without a coefficient value of 0 were variables included in the model.

FIGURE 2

Tuning parameter (λ) selection in the Lasso regression model. Models were fitted using cross-validation, and then, the optimal λ value was selected to have a more accurate estimate of the performance of the model. For each λ value, around the mean of the target parameter shown in the red dot, we could obtain a confidence interval of the target parameter. The two dashed lines indicated two particular λ values, namely, the one with the smallest model error, and the one with the best model performance but the smallest number of independent variables.

FIGURE 3

Nomogram model for predicting CPET-related cardiovascular events in patients with CHD after PCI. The first line was the points of risk factors, and the second to sixth lines were risk factors. By summarizing each risk factor point, total points could be obtained. Then, the event probability (line 8) was calculated according to the total points of line 7. HbA1c, glycosylated hemoglobin A1c; Sm, systolic velocity of mitral annulus; VE/VCO₂ slope > 30, minute ventilation/carbon dioxide output slope > 30; ΔVO₂/ΔWR slope inflection, inflection of oxygen uptake relative to work rate; Events, CPET-related cardiovascular events.

FIGURE 4

Calibration curve of nomogram model in the development cohort. The standard curve was a straight line passing through the origin of the coordinate axis. If the predicted calibration curve is closer to the standard curve, the better the prediction ability of the nomogram. “Bias-corrected” was the correction curve, while “Ideal” was the standard curve. The graphical predicted and measured values are basically consistent, indicating that the nomogram model of this study has a good calibration.

FIGURE 5

Calibration curve of nomogram model in the validation cohort. As demonstrated in the calibration curve of the nomogram model in the validation cohort, the predicted value was basically consistent with the measured value. “Basically consistent” means that the actual solid line and the ideal dashed line were basically parallel, indicating that the nomogram model of this study has a good calibration.