A practical risk model based on Lasso-Cox regression for angina with nonobstructive coronary arteries (ANOCA) patients

Abstract Background Coronary flow reserve (CFR) reduction is currently one of the main criteria for risk-stratifying patients with angina and nonobstructive coronary artery disease (ANOCA). However, CFR depends on resting and vasodilated coronary hemodynamics, and a theoretical concern remains that physiological changes may inadvertently affect the result. Purpose We aim to fit different risk models for patients with ANOCA, which further provides applicable guidance to clinical doctors for their patients. Methods Between May 2018 and October 2023, 251 consecutive patients who were diagnosed as ANOCA were enrolled from Qilu Hospital of Shandong University, and the median follow-up was 800 (493, 1221) days. Demographic characteristics, cardiovascular risk factors, cardiac structure and function index, echocardiography-derived coronary flow velocity, coronary flow velocity-time integral, and CFR were collected. Least absolute shrinkage and selection operator (LASSO) and multivariable Cox regression analyses were used to screen for significant parameters that were included to develop a novel predictive nomogram model. Receiver operating characteristic curve and calibration curve were used to evaluate the performance and clinical utility of the nomogram. Results In LASSO regression selection, five variables, diabetes, global work efficiency (GWE), global wasted work (GWW), decreased hyperemic coronary flow velocity (<0.63m/s), and peak flow velocity-derived CFR<2.5 (Figure 1) showed their correlation to the occurrence of major adverse cardiovascular events (MACE). Univariate Cox regression analysis shows that diabetes (hazard ratio [HR] = 4.108; 95% confidence interval [CI], 2.296–7.350; p = 0.001), GWE (HR = 0.890; 95% CI, 0.856–0.926; p = 0.001), GWW (HR = 1.005; 95% CI, 1.004–1.007; p = 0.012), decreased hyperemic coronary flow velocity (HR =3.405; 95% CI, 1.730–6.702; p = 0.001) and peak flow velocity-derived CFR<2.5 (HR =3.456; 95% CI, 1.898–6.292; p = 0.001) were predictors of recurrent chest pain. Multivariate analysis showed that the association of diabetes (HR = 2.735; 95% CI, 1.452–5.152; p = 0.002), GWW (HR = 1.004; 95% CI, 1.001–1.007; p = 0.010), and peak flow velocity-derived CFR<2.5 (HR = 2.650; 95% CI, 1.306–5.380; p = 0.007) with MACE remained statistically significant after adjusting for other variables. A nomogram incorporated the predictors mentioned above; it is presented in Fig.2. Internal validation showed that the model has good predictive performance. The areas under the curve (AUCs) for prediction of the risk of MACE was 0.849 (Fig.3). The calibration curve for predicting chest pain recurrence shows a good agreement between prediction and observation (Fig.4). Conclusions Diabetes, GWE, GWW, decreased hyperemic coronary flow velocity (<0.63m/s), and peak flow velocity-derived CFR<2.5 are significant in identifying the occurrence risk of MACE for ANOCA patients.