Class I recommendations for invasive coronary angiography also apply to some patients with nonspecific chest pain or unstable angina (1). In the patient with nonspecific chest pain (in whom the pretest probability of coronary stenoses would usually be low), a high-risk abnormality on noninvasive testing is a class I indication for invasive coronary angiography. Unstable angina is a class I recommendation because of the high pretest probability of a stenotic lesion, which led to revascularization in published studies.

To put the class I indications in an overall perspective, expert panels recommend invasive coronary angiography when there is a high probability of finding stenoses that require revascularization. Guidelines recommend noninvasive tests rather than invasive coronary angiography for myocardial ischemia when pretest probabilities are lower because the post-test probability after these tests may help to decide whether to use invasive coronary angiography. Viewed in this light, an accurate estimate of the post-test probability after a noninvasive test is critical to making a good decision about further testing or treatment. It is important, then, to have accurate estimates of pretest probability, based on careful clinical assessment, and accurate estimates of noninvasive test performance.

In the past several years, 2 new forms of coronary angiography that do not require cardiac catheterization have emerged: multislice computed tomography (CT) and magnetic resonance imaging (MRI). In a meta-analysis (2) of the comparative diagnostic performance of MRI and multislice CT for detection of significant (50% diameter) stenosis, the authors reviewed 28 studies in 903 patients who had MRI and 24 studies in 1300 patients who had multislice CT (invasive coronary angiography was the gold standard test). The weighted average sensitivity and specificity of MRI for detection of high-grade stenoses were 72% and 87%, respectively, with narrow confidence intervals. Weighted sensitivity and specificity of multislice CT were 85% and 95%, respectively, also with narrow confidence intervals. The data suggest that multislice CT is more accurate, but few if any studies directly compared multislice CT and MRI results in patients who had both tests (head-to-head comparison).

The paper by Dewey and colleagues (3) in this issue presents important new information about the accuracy of multislice CT and MRI for detecting coronary stenoses. In this carefully done study, the authors obtained both multislice CT and MRI in patients presenting to a single tertiary care center in Germany for suspected CAD. Both noninvasive tests were performed within a median of 1 day of the invasive coronary angiography. Direct comparisons in the same patient were possible in 108 of the 129 patients. In per-patient analyses, CT outperformed MRI with higher sensitivity (92% vs. 74%) while specificity was comparable for the 2 tests (79% vs. 75%; P = 0.643). In analyses of specific coronary arteries, both tests performed less well: The sensitivities were 82% for multislice CT and 54% for MRI, and the specificities were similar (P = 0.726).

The patient sample in this study is important in understanding the results. Approximately half of the patients had typical angina, which is generally associated with a 90% pretest likelihood of significant CAD in men. The rest had atypical angina (25%), nonspecific chest pain (10%), and no chest pain (18%). In their Figure 3, the authors provide a plot of the post-test probability of CAD for positive and negative multislice CT findings and MRI studies for all pretest probabilities from 0 to 1.0. Since the sensitivity and specificity in this single-center study are very similar to those in the meta-analysis cited earlier (2), the authors' analysis of the effect of these 2 tests on post-test probability is probably reliable. In addition, because their study included different kinds of patients (not only high-risk patients) and because the prevalence of no detectable CAD was fairly high (48%), it provides a useful estimate of the performance of these 2 tests in patients that are typical of those seen in clinical practice.

How can we conclude which test to use? According to the accumulated data (2, 3), multislice CT is more sensitive than MRI and equally specific. Thus, if accuracy were the sole criterion, multislice CT would be the preferred test. However, multislice CT requires a contrast dye load and radiation exposure that are approximately equal to invasive coronary angiography. Accordingly, if many patients need invasive coronary angiography after positive results on multislice CT, the additional contrast and radiation exposure are strong reasons against using the latter test.

The best use of multislice CT (which most would prefer over MRI because of its higher sensitivity, despite its higher risks) would be in a patient whose post-test probability after a negative study would be so low that invasive coronary angiography would be unnecessary. According to Dewey and colleagues' Figure 3, the post-test probability after negative results on multislice CT would be about 40% in a man with typical angina, which is too high to confidently exclude CAD. A more likely candidate is the man with atypical angina or atypical chest pain, in whom the pretest probability is relatively low. Other authors have identified this type of patient as a likely candidate (2, 3). However, this scenario would be relatively uncommon in patients with a class I (and even class IIa) indication for invasive coronary angiography, since most such patients have a high pretest probability of CAD.

Who, then, would qualify as a potential candidate for invasive coronary angiography despite having a relatively low pretest probability of CAD? Certainly not patients with unstable angina, who are at high risk for a subsequent CAD event and are therefore highly likely to undergo invasive coronary angiography and concurrent or subsequent revascularization, or the man with stable angina, whose pretest probability of CAD is high. Even the patient with nonspecific chest pain and high-risk findings on functional noninvasive testing would probably not be a candidate for multislice CT because the high-risk functional test results would most likely increase the probability of CAD to at least 40% to 50% or greater.

Although multislice CT seems to have little value in most candidates for invasive coronary angiography, I believe that it can be useful in an important, albeit small, subset of patients with chest pain. As shown in Dewey and colleagues' Figure 3, negative results on multislice CT are associated with post-test probability of CAD below 10% only in patients whose pretest probability of CAD was less than 50%. Achieving a post-test probability below 5% requires a pretest probability below 30%, as would occur in a man or a woman with nonspecific chest pain and equivocal results on functional noninvasive testing. A CAD diagnosis might be important to establish anatomically, especially if the patient's symptoms require multiple physician or emergency department visits or hospitalizations. According to Dewey and colleagues, such patients commonly undergo invasive coronary angiography in Germany. In these patients with low pretest probability, in whom the response to treatment and other noninvasive tests is inconclusive, multislice CT results will usually be negative and the post-test probability of CAD will be low enough to forgo invasive coronary angiography and yet reassure the patient. Appropriate use of multislice CT in such patients requires a careful clinical assessment of probabilities (4) as well as a properly conducted work-up with functional tests, as outlined in current American College of Cardiology/American Heart Association guidelines (1, 5).

What is the bottom line for clinicians? Clinicians should not use multislice CT routinely in the evaluation of patients with suspected myocardial ischemia. Multislice CT is not useful in patients with a high pretest probability of CAD and may be harmful. It is useful when the diagnosis is uncertain after a careful chest pain history and equivocal results on functional testing for CAD.