Dr. Palen wonders about "the manner in which information from the traditional clinical evaluation was incorporated into the clinical model" of cardiac risk. For each cardiac event, we used stepwise multivariable regression to identify the clinical factors that were the best predictors of the outcome of interest and then added the echocardiographic data to that "best" clinical model to see whether the diagnostic test added any incremental information. For example, our clinical model of all cardiac outcomes included vascular surgery, history of dysrhythmia, definite coronary artery disease, and use of digoxin (see Tables 4 and 7 in our report). In our cohort, the clinical risk models we presented were better predictors of adverse events than were the Goldman cardiac risk index [1], Detsky risk index [2], and American Society of Anesthesiologists classification (as we previously reported [3]). Even so, analyses that used the aforementioned cardiac risk indices produced similar findings. The echocardiogram provided no significant prognostic information in addition to that predicted by the cardiac risk indices.

Dr. Klein is intrigued that we assessed left ventricular ejection fraction "visually." Echocardiographic determination of ejection fraction is most commonly estimated visually, as was done in our study. More sophisticated quantitative and computerized techniques for assessing ejection fraction are time consuming, require ideal endocardial resolution (which is difficult to achieve in many instances), and are of limited use in clinical practice. In addition, studies suggest that quantitative and "visually estimated" echocardiographic ejection fraction have comparable correlation, limits of agreement, and reproducibility [4]. Furthermore, the correlation and reproducibility between visually estimated echocardiographic ejection fraction and the gold standard, quantitative radionuclide angiography, are also quite high [5]. As we indicated in our methods, the interobserver, intraobserver, and interexamination agreement rates for echocardiographic interpretations in our laboratory all exceed 90%. Finally, had our measurements been prone to error or whimsy (as Dr. Klein implies), we could not have reported that ejection fraction was a highly significant univariate predictor of total postoperative cardiac events, congestive heart failure, and ventricular tachycardia.

We stand behind our conclusion that echocardiography had limited prognostic value and suboptimal operating characteristics in patients having noncardiac surgery. Nor did the echocardiogram add any incremental predictive information to that readily available from known clinical risk factors. The clinical history and physical examination still remain the cornerstones of cardiac risk assessment.