Novel predictive model of obstructive coronary artery disease combining clinical and biohumoral data

Background. The probability of coronary artery disease (CAD) is currently estimated by the analysis of symptoms characteristics, age, sex, cardiovascular risk factors and stress tests. We sought to investigate the incremental diagnostic value of circulating biomarkers in reclassifying the probability of CAD. Methods. A total of 449 patients (60±9 years, 61% males) with angina-like chest pain or equivalent symptoms, enrolled in the EVINCI study, underwent clinical examination, exercise electrocardiogram, biohumoral characterization (36 biomarkers associated with atherosclerosis) and invasive coronary angiography. Primary end point was obstructive CAD, defined as >70% coronary stenosis in at least one major epicardial vessel, or 30-70% stenosis with fractional flow reserve <=0.8. Diamond & Forrester (D&F) predictive model, including age, sex, type of symptoms and exercise ECG, was used as reference. A new model was built-up adding biohumoral variables selected by multivariate logistic regression to D&F variables. Diagnostic performance was assessed by the area under the Receiver Operating Characteristics curve (AUC) and the clinical utility by estimating net reclassification improvement (NRI), using 50% probability of CAD as cut-off. Results. The probability of obstructive CAD was 65% [34-78%] by D&F model while the actual prevalence of the disease was 30%. High-density lipoprotein cholesterol, aspartate transaminase, homeostasis model assessment index, interleukin-6 and osteopontin were independent predictors of obstructive CAD at multivariate logistic regression. The addition of these biomarkers to D&F variables determined a significant increase of diagnostic accuracy [Area Under the Curve from 0.72 (SE 0.03) to 0.82 (SE 0.03)] (p <0.001). The integrated model allowed a correct reclassification of 18% of patients [NRI= 18% (SE 0.05), p=0.001]. Conclusions. The D&F model overestimates the probability of obstructive CAD. Updating the model to a contemporary European population and adding biohumoral markers improves the predictive accuracy and correctly reclassifies a significant number of patients.