A predictive model for distinguishing ischemic from non-ischemic cardiomyopathy

Abstract

Objectives: To develop a predictive model to distinguish ischemic from non-ischemic cardiomyopathy Material and Method: The authors randomly assigned 137 patients with LV systolic dysfunction into two subsets - one to derive a predictive model and the other to validate it. Clinical, electrocardiographic and echocardiographic data were interpreted by blinded investigators to the subsequent coronary angiogram results. Ischemic cardiomyopathy was diagnosed by the presence of significant coronary artery disease from the coronary angiogram. The final model had been derived from the clinical data and was validated using the validating set. The receiver-operating characteristics (ROC) curves and the diagnostic performances of the model were estimated. Results: The authors developed the following model: Predictive score = (3 x presence of diabetes mellitus) + number of ECG leads with abnormal Q waves - (5 x presence of echocardiographic characteristic of non-ischemic cardiomyopathy). The model was well discriminated (area under ROC curve = 0.94). Performance in the validating sample was equally good (area under ROC curve = 0.89). When a cut-off point ≥ 0 was used to predict the presence of significant coronary artery disease, the model had a sensitivity, specificity and positive and negative predictive values of 100%, 57%, 74% and 100%, respectively. Conclusion: With the high negative value of this model, it would be useful for use as a screening tool to exclude non-ischemic cardiomyopathy in heart failure patients and may avoid unnecessary coronary angiograms.