Although there are many reasons for performing the diagnostic techniques mentioned above, most fall into the following categories: 1) to determine the presence or absence of significant coronary artery obstructions in a patient with signs or symptoms consistent with this disease; 2) to determine the severity and extent of coronary stenosis; 3) to assess prognosis in patients with known or suspected coronary disease; 4) to assess the physiologic significance of coronary stenoses visualized by angiography; 5) to assess the "viability" of a region of the myocardium before revascularization; 6) to gather information that can guide treatment options; and 7) to assess response to a treatment strategy. Physicians must understand the specific clinical questions to be addressed before choosing specific diagnostic techniques because those that adequately address one question (such as the presence of significant coronary artery stenosis) may be less well suited to address another (such as the prognosis in a patient with previously shown coronary disease) [1].

One example of this phenomenon is the use of nuclear imaging studies to diagnose significant coronary stenosis. Although cardiac catheterization with coronary angiography is the "gold standard" for detecting coronary stenosis, many less invasive and less costly methods for identifying patients with either a high or a low likelihood of significant stenosis are available. One of the most common noninvasive methods is planar thallium perfusion scintigraphy. The sensitivity and specificity of planar thallium imaging for diagnosing coronary artery stenosis are usually reported to be about 85% [2]. The use of single-photon emission computed tomography technology, which enhances the spatial resolution of thallium scintigraphy, has recently been reported to increase sensitivity to more than 95% [3]. In comparison, exercise electrocardiography has good sensitivity for detecting severe coronary artery disease (that is, severe stenosis of the left main coronary artery or all three main epicardial vessels), but it has significantly lower sensitivity for detecting limited coronary artery disease (for example, disease confined to a single, non-left main vessel) [4]. With its increased sensitivity, stress thallium imaging may offer an advantage in detecting limited coronary artery disease. If, however, the clinical question is not "Is there a high likelihood of any significant coronary stenosis?" but rather "Is there a high likelihood of severe coronary artery stenosis?" the incremental value of thallium imaging in addition to exercise electrocardiography is significantly decreased because of the higher sensitivity of clinical variables and stress electrocardiography alone in detecting severe disease.

In choosing the most appropriate diagnostic technique, it is also important to consider the characteristics of the individual patient. For example, stress electrocardiographic testing is good at predicting future adverse cardiac events in patients with normal resting electrocardiograms but is much less successful in patients with previous Q-wave infarctions, left ventricular hypertrophy, conduction abnormalities, or other factors resulting in an abnormal resting electrocardiogram [5]. Likewise, the stress electrocardiogram is less valuable in women, patients receiving digoxin, and persons with renal insufficiency [1, 6]. Patients with one or more of these characteristics are therefore more likely to obtain incremental benefit from thallium scintigraphy in terms of prognostic or diagnostic accuracy than are patients ideally suited to exercise electrocardiography.

In a well-done study published in this issue, Christian and colleagues [7] compared the incremental value of thallium single-photon emission computed tomography scintigraphy with that of clinical variables and stress electrocardiography in predicting three-vessel or left main coronary artery disease in selected patients with normal resting electrocardiograms [7]. They found that variables obtained from the stress electrocardiogram (magnitude of exercise-induced ST depression and peak heart rate-blood pressure product during exercise) increased the ability of clinical variables (age, sex, presence of diabetes, and history of typical angina) to categorize patients as having low, intermediate, or high risk for having three-vessel or left main disease on angiography. Thallium scintigraphy added little to the combination of clinical and stress electrocardiographic variables and did so at a high cost. Using a cross-validation model, which more accurately reflects the performance of their model when applied to future patients in clinical practice, the authors found that only 2 of 411 patients were correctly reclassified when thallium-derived variables were added to clinical and exercise electrocardiogram variables. The cost of thallium imaging per additional correct classification was estimated to be more than $143 000.

Several aspects of the study by Christian and colleagues show the appropriate use of important methodologic approaches that are frequently neglected in evaluating diagnostic tests. First, the authors account for patients not included in the study, which supplies information on the generalizability of their conclusions. Second, the regression modeling strategy builds on clinical information that was available before the test and thus provides a critical link between statistical evaluation of the technique and actual clinical practice. In evaluating a patient, the clinician has access to the history, physical examination, and resting electrocardiogram before imaging studies are done. By integrating this "free" information into an initial estimate of diagnosis and prognosis, the clinician can judge the value of the test according to whether it incrementally improves the accuracy of the original estimate compared with other technologies. Third, the use of model validation techniques allows an accurate estimate of the value of diagnostic or prognostic testing. Although validation in an independent sample is the best test of a statistical model, internal validation methods, such as the jack-knife technique used by Christian and colleagues, provide a critical method of countering the excessive optimism that results from overfitting a statistical model to a particular data set. Finally, their efforts to simultaneously evaluate both the information content and the cost of thallium exercise testing are laudable.

The data provided by Christian and colleagues, when combined with that reported in previous studies [5, 8], shows that tomographic thallium imaging adds little information to that obtained from a history and exercise electrocardiogram in selected patients for a specific question of interest. It is important to realize that the authors' patients were highly selected: Of the 2638 patients evaluated at the Mayo Clinic who had both an exercise thallium study and coronary angiography within 6 months, only 411 (16%) were included in the current analysis. The rest were excluded for various reasons, including a history of myocardial infarction, current digoxin use, or an abnormal resting electrocardiogram (anomalous ventricular conduction, atypical QRS axis or voltage, pathologic Q waves, nonspecific ST-segment or T-wave aberrancy, or a rhythm other than normal sinus or sinus bradycardia). Patients with valvular heart disease or a history of coronary revascularization (either bypass surgery or angioplasty) were also excluded. In addition, more than 80% of patients studied were male. The importance of this observation is shown in a previous study by Morise and colleagues [1] that examined the value of thallium scintigraphy in addition to clinical and exercise electrocardiographic data for detecting coronary artery disease. This report suggests that thallium scintigraphy adds significant diagnostic information in women but is less useful in men.

It is also important to consider that the clinical question in Christian and colleagues' study—"What is the likelihood of this patient having either three-vessel or left main coronary disease?"—is but one of many possible questions about patients who have noninvasive cardiac evaluations. The value of thallium imaging may differ if the physician is interested primarily in the presence or absence of any significant coronary disease (as could be the case when trying to determine the likelihood of coronary insufficiency as an explanation for a patient's symptoms) or the physiologic significance of a documented coronary stenosis while considering a revascularization procedure.

Although we must be careful not to overgeneralize the conclusions, Christian and colleagues' study substantially adds to our understanding of who benefits (and who does not) from the addition of thallium scintigraphy to the simpler, safer, and less costly stress electrocardiography in evaluating the risk for severe coronary stenosis. Frequently, the value of a test as reported by investigators seems to be directly proportional to the expertise of the investigators in doing that test. Christian and colleagues therefore are to be especially commended for their objective evaluation of a technology that they helped develop. Their study shows that in men without a history of cardiac disease and with a normal resting electrocardiogram for whom an estimation of the likelihood of left main or three-vessel disease is needed, thallium scintigraphy provides little additional information to an exercise electrocardiogram alone. Previous articles published by this group have shown the limited value of exercise radionuclide angiography (RNA or multiple gated image acquisition analysis) in a population similar to the one described in this issue [6] and have documented the usefulness of thallium imaging in other populations [9]. As physicians are increasingly forced to confront the incremental cost compared with the incremental value of expensive technology, similar analyses should be done in other populations using other end points and comparing other diagnostic techniques. Only through this type of rigorous inquiry can we ultimately answer the question of which test is best for a particular patient.