A machine-learning based bio-psycho-social model for the prediction of non-obstructive and obstructive coronary artery disease

Abstract

Background: Mechanisms of myocardial ischemia in obstructive and non-obstructive coronary artery disease (CAD), and the interplay between clinical, functional, biological and psycho-social features, are still far to be fully elucidated.

Objectives: To develop a machine-learning (ML) model for the supervised prediction of obstructive versus non-obstructive CAD.

Methods: From the EVA study, we analysed adults hospitalized for IHD undergoing conventional coronary angiography (CCA). Non-obstructive CAD was defined by a stenosis < 50% in one or more vessels. Baseline clinical and psycho-socio-cultural characteristics were used for computing a Rockwood and Mitnitski frailty index, and a gender score according to GENESIS-PRAXY methodology. Serum concentration of inflammatory cytokines was measured with a multiplex flow cytometry assay. Through an XGBoost classifier combined with an explainable artificial intelligence tool (SHAP), we identified the most influential features in discriminating obstructive versus non-obstructive CAD.

Results: Among the overall EVA cohort (n = 509), 311 individuals (mean age 67 ± 11 years, 38% females; 67% obstructive CAD) with complete data were analysed. The ML-based model (83% accuracy and 87% precision) showed that while obstructive CAD was associated with higher frailty index, older age and a cytokine signature characterized by IL-1β, IL-12p70 and IL-33, non-obstructive CAD was associated with a higher gender score (i.e., social characteristics traditionally ascribed to women) and with a cytokine signature characterized by IL-18, IL-8, IL-23.

Conclusions: Integrating clinical, biological, and psycho-social features, we have optimized a sex- and gender-unbiased model that discriminates obstructive and non-obstructive CAD. Further mechanistic studies will shed light on the biological plausibility of these associations.

Clinical trial registration: NCT02737982.

Keywords: Cytokines; Frailty; Gender; Inflammation; Ischemic heart disease; Machine learning; Non-obstructive coronary artery disease.

Fig. 1

Definition of the machine-learning classification model. Schematic representation of the working process to define and optimize the ML-based model. After data pre-processing, 75% of the data is used to train an XGBoost model including the tuning of the hyperparameters by means of a fivefold cross-validation procedure. The hyperparameters that give the best average validation values of the key performance indicators are chosen and the model is retrained on the full training set with the optimal setting. Eventually, the model is deployed and used to predict the type of CAD of the patients in the test set (25% of data) and the final performance (accuracy and precision) of the model is determined

Fig. 2

Cytokine profile of obstructive and non-obstructive CAD patients. Violin plots display in logarithmic scale the concentration (pg/ml) of 13 cytokines measured by multiplex bead-based flow cytometric assay in coronary serum samples of non-obstructive (NO; < 50% diameter stenosis) or obstructive (O; ≥ 50% diameter stenosis) CAD patients. A straight line indicates the median and the dotted lines show the interquartile range (IQR) (25th, 75th percentile). The significance of differences between median values was tested by Mann–Whitney tests for independent samples (*p ≤ 0.05, **p ≤ 0.01)

Fig. 3

Cytokine profile of female and male individuals with CAD. Violin plots display in logarithmic scale the concentration (pg/ml) of 13 cytokines measured by multiplex bead-based flow cytometric assay in coronary serum samples of female (F) and male (M) CAD patients. A straight line indicates the median and the dotted lines show the interquartile range (IQR) (25th, 75th percentile). The significance of differences between median values was tested by Mann–Whitney tests for independent samples (*p ≤ 0.05, **p ≤ 0.01)

Fig. 4

Cytokine profile of patients with acute and stable CAD. Violin plots display in logarithmic scale the concentration (pg/ml) of 13 cytokines measured by multiplex bead-based flow cytometric assay in coronary serum samples of CAD patients with acute (A) or stable (S) presentation. A straight line indicates the median and the dotted lines show the interquartile range (IQR) (25th, 75th percentile). The significance of differences between median values was tested by Mann–Whitney tests for independent samples (*p ≤ 0.05)

Fig. 5

Machine-learning model that integrates biological, clinical, functional and psycho-social features to predict obstructive and non-obstructive CAD. SHAP Plot of the best performing ML-based model. Each feature is shown on the vertical axes of a graph where the higher is the position (ranking), the more influential is the characteristic. In the graph for each feature (namely on the horizontal lines) the dots represent patients, and the colour indicates whether the value of the characteristic considered is high or low in relation to the range of values (red refers to high values and blue to low values). The graph has a median line and the farther the point is from the median line, the stronger is the influence on the output, with the points on the right correlating positively with obstructive CAD and the points on the left negatively, thus predicting the opposite outcome (non-obstructive CAD)

Fig. 6

Conceptual framework of the study. Our ML-based model supports the idea that different inflammatory mechanisms underlie different type of CAD. We found that obstructive CAD was associated with increased frailty index, older age and a cytokine signature characterized by IL-1β, IL-12p70 and IL-33, which is pro-atherogenic and promotes plaque instability. Non-obstructive CAD was associated with a higher gender score (i.e., social characteristics traditionally ascribed to women) and with a cytokine signature characterized by IL-18, IL-8 and IL-23, which supports plaque stability and neutrophil recruitment