Dr. Hurwitz would like the scope of our study to be broader. Our study was designed to address a specific issue—the identification of three-vessel or left-main disease in patients with a normal resting electrocardiogram who had exercise testing. This particular scenario in exercise electrocardiography can be expected to be at its best (yielding interpretable ST segments in the setting of severe disease), whereas thallium-201 scintigraphy is not (a relative perfusion technique where abnormalities in all three territories occur in less than one third of patients with three-vessel disease) [1]. We agree that there are many other useful applications of perfusion imaging in which Bayesian principles are used. However, we feel that it is important to determine in which cases such technology is effective and in which cases similar performance can be provided at less expense.

Although Drs. Halkar, Tarcan, and Bodenheimer all believe that post-test referral bias was a major factor in the results of our study, they do not describe the magnitude or direction of the bias. We agree that referral bias (patients with positive test results are preferably referred to the gold standard test) is important, as we stated [2]. Referral bias not only decreases specificity but increases the sensitivity of the test [3]. If the results of thallium scintigraphy were the only factor determining referral to angiography (70% of patients had the thallium study before angiography), its sensitivity should be better than using clinical and exercise variables and its specificity should be worse. The receiver-operator characteristic curves shown in Figure 2 of our article do not suggest such an effect. Clinical and exercise variables probably also influenced the decision to proceed to angiography. As a result, the major effect of referral bias is probably to overestimate sensitivity and underestimate specificity for both exercise testing and thallium imaging.

Dr. Bodenheimer is concerned about how these results relate to revascularization. This was not the focus of our study. Although 59% of patients classified as having a high risk by clinical exercise and thallium variables were referred for revascularization, 63% of patients classified as having a high risk by clinical and exercise variables alone were also referred. This suggests that the thallium scan results were only one factor in this complex clinical decision, along with clinical variables and exercise test results.

We disagree with Dr. Bodenheimer's claim that our study was not an incremental analysis. In a perfect study, clinicians and institutional review boards would agree to have all patients receive coronary angiography regardless of the noninvasive test results, and external sources would agree to fund the coronary angiography. Given our current inability to do such a study, we believe that our analysis is the best approach available and that it does carefully consider how much information thallium imaging results add to clinical and exercise results, in this specific situation. As indicated above, we believe that there are many other useful applications of perfusion imaging. We are not trying to throw the baby out with the bath water, but rather are trying to avoid drowning the health care system in unnecessary expense in this specific situation.