A comparison of treadmill scores to diagnose coronary artery disease

Background: Recently, several treadmill scores have been proposed as means for improving the diagnostic accuracy of the exercise treadmill test (ETT). Questions remain regarding the diagnostic accuracy of treadmill scores when applied to a different patient population than that from which they were derived; furthermore, many treadmill scores have not been compared with one another in the same population. Hypothesis: The diagnostic accuracy of treadmill scores may not be the same. Methods: A retrospective analysis of data collected prospectively was performed on consecutive patients referred for evaluation of chest pain. All patients underwent a standard ETT followed by coronary angiography. Using angiographic evidence of coronary artery disease (CAD) as a reference, the area under the curve (AUC) of receiver operator characteristic (ROC) plots of the ST response alone, the Duke Treadmill Score (DTS), the Morise score, the Detrano score, the VA score, and a Consensus score consisting of the Morise, Detrano, and VA scores together were calculated and compared. The predictive accuracies of the DTS and the Consensus score to stratify patients for the likelihood of CAD were calculated and compared. Results: In all, 1,282 patients without a prior myocardial infarction had an ETT and coronary angiography. The AUC (± standard error) was 0.67± 0.01 for the ST response, 0.73±0.01 for DTS, 0.76±0.01 for Detrano score, 0.77±0.01 for Morise score, 0.78± 0.01 for VA score, and 0.78±0.01 for Consensus score. The AUC for each treadmill score was significantly higher (z-score > 1.96) than for the ST response alone. The AUC of DTS was significantly lower than all other treadmill scores (z-score > 1.96). The predictive accuracy (± 95% confidence interval) of the DTS to risk stratify patients into high and low likelihood for CAD was 71 (65-77)%, versus 80 (74-86)% for the Consensus score (p < 0.0001). Conclusion: In this population, the DTS remains useful for diagnosing CAD and stratifying for the likelihood of CAD, although it is less accurate than other treadmill scores.