Author + information
We congratulate Shavelle et al. (1)on their study in which they present data comparing the use of electron beam computed tomography (EBCT), treadmill-electrocardiogram (ECG) and technetium-stress testing to identify patients with coronary artery disease (CAD). The investigators conclude that EBCT has a higher diagnostic ability than either treadmill-ECG or technetium-stress for the detection of CAD. There are several important issues, however, that should be addressed when interpreting the results of their study.
First, it is important to note that the authors chose to study a high-risk population in which 69% of patients had CAD. In most populations undergoing screening tests, the disease prevalence is much lower. Some diagnostic test characteristics are affected by disease prevalence, including positive and negative predictive values and accuracy. Consequently, studies of diagnostic tests often present predictive values and accuracy that are calculated for lower-risk populations, as well as likelihood ratios that are independent of disease prevalence. Using data from the report by Shavelle et al., we calculated predictive values and accuracy using a lower disease prevalence than the one reported in their paper (Table 1). Moreover, we calculated likelihood ratios for the three diagnostic tests. Positive predictive values for the three tests were similarly poor, suggesting that each of these tests has a poor diagnostic value for CAD in low-risk populations. At the same time, EBCT had the highest negative predictive value (98%), suggesting that the strength of EBCT lies in its ability to rule out CAD. When we calculated the accuracy of each of the three tests, using a more typical disease prevalence, EBCT was the least accurate. However, EBCT had the best negative likelihood ratio, reaffirming its strength in ruling out CAD. When we calculated positive likelihood ratios, technetium-stress testing was somewhat better at identifying patients with CAD. Thus, EBCT may not be the ideal test to identify patients with CAD in a general population.
A second issue is that the investigators’ study may be limited by the sensitivity and specificity they found for technetium-stress testing. The numbers that they report are lower than those found in previous studies. Because the investigators examined only 97 patients, they may have inadvertently underestimated the sensitivity and specificity of technetium-stress testing. The authors refer to a meta-analysis of studies examining exercise nuclear testing (2). In this pooled analysis, nuclear testing had a sensitivity of 87% and a specificity of 64%. Using this sensitivity and specificity, and assuming a disease prevalence of 20%, we calculated a positive predictive value of 38%, a negative predictive value of 95% and an accuracy of 69% for technetium-stress testing. The positive likelihood ratio was 2.4, and the negative likelihood ratio was 0.2. Thus, when the diagnostic test characteristics are recalculated using the higher sensitivity and specificity from the literature, it appears that technetium-stress testing may be a better screening test than EBCT to identify patients with CAD.
A third issue is that the meta-analysis (2)referred to by Shavelle et al. (1)also published test characteristics for exercise echocardiography, reporting a pooled sensitivity and specificity of 85% and 77%, respectively. Using these data, we calculated the diagnostic test characteristics for exercise echocardiography in a population with a disease prevalence of 20% (Table 1). The positive and negative predictive values were 48% and 95%, respectively; accuracy was 79%; and the positive and negative likelihood ratios were 3.7 and 0.2, respectively. Thus, using the pooled sensitivity and specificity from the literature, exercise echocardiography may well be the best of the four diagnostic tests.
Finally, because of the relatively small number of patients studied by Shavelle et al., it would have been helpful if the authors had included 95% confidence intervals (CIs) for the diagnostic test characteristics they cited. Although the point estimates they reported suggest that EBCT has reasonable test characteristics, CIs for these numbers may be wide, and there may be no statistically significant difference among the various diagnostic tests.
In conclusion, EBCT has several important limitations when used for the diagnosis of CAD. It is a costly procedure, and it provides no information regarding the functional significance of any lesions that may be present. Before EBCT is widely employed as a screening test for CAD, the evidence in favor of this test must be carefully scrutinized. Investigators should report on diagnostic test characteristics in populations where the disease prevalence is reflective of the populations in which the tests will eventually be used. In addition, similar, but larger studies, need to be performed in lower-risk populations to provide precise point estimates of diagnostic test characteristics as well as narrow CIs for these numbers.
- American College of Cardiology