ETD

• case-control study
• coronary disease
• risk prediction

Current clinical practice requires coronary angiography (CAG) for all patients with suspected left main coronary artery disease (LMCAD). Given that LMCAD is diagnosed in a low percentage of CAGs, a less invasive diagnostic method would be highly desirable.
We conducted a multicenter retrospective case-control study and developed a prediction model based on clinical and exercise stress test (EST) variables to diagnose LMCAD and severe LMCAD. We performed internal validation of the model by bootstrap resampling and external validation by splitting the dataset by center and by time.
For the LMCAD model, the Area Under the Curve (AUC) of the internally validated prediction model was 0.72. Assuming a LMCAD prevalence of 5% among those referred to CAG and a 1:100 misclassification cost between false negatives and false positives, the model had a sensitivity of 85.3% and a negative predictive value of 98.2%. The application of this model in the clinical practice would safely spare CAGs to 58 true negatives per missed LMCAD diagnosis. The external validation provided similar results.
These findings suggest that a risk prediction model based on clinical and EST parameters could allow a less expensive and non-invasive management for patients with chronic coronary syndrome.