This perception is wrong, but the exercise electrocardiogram is in need of rehabilitation. It is time to rethink its overall role for the routine evaluation of patients and for specific diagnostic purposes in a range of highly specific populations. As the exercise test is reexamined, there is little reason why the test's performance should be constrained by empirically derived criteria that have been based on visual estimation of ST-segment depression since mid-century. Progress in exercise testing requires development of new criteria with improved sensitivity and specificity, and it is reasonable to expect the exercise electrocardiogram to play a useful and fundamental role in the diagnostic evaluation of chest pain, the assessment of disease severity, and the prediction of risk for coronary events.

The performance of exercise test criteria can vary strikingly with the diagnostic purpose of the test and with the nature of the study population [1-6]. For example, the high prevalence of "false-positive" results in previous studies of catheterized patients may not be representative of the broad range of patients with noncoronary chest pain symptoms who are seen in practice. Even so, the reduced specificity found in meta-analyses of these studies contributes to the Bayesian calculation of poor predictive value that has withered confidence in the general usefulness of the exercise test. Therefore, clarification of exercise test performance in populations free of workup bias is an important step in the reassessment of test value and applicability.

In this issue, Froelicher and colleagues [7] report major progress in defining and acquiring a group of patients with reduced workup bias. In their carefully planned multicenter study, patients who were referred for evaluation of chest pain were included only if they agreed to have both exercise testing and coronary angiography within 30 days. As proposed, this inclusion requirement should effectively eliminate the outcome of the exercise test itself as a confounding factor in the decision to proceed with diagnostic catheterization for the documentation of coronary artery disease.

More than 800 men with chest pain syndromes in 12 Veterans Affairs medical centers consented to this diagnostic approach, and half were found to have diameter narrowing of more than 50% in one or more coronary vessels. When these men were compared with nearly 700 men in a pilot study in whom the decision to perform angiography was made after exercise testing, the specificity of standard visual criteria was shown to be increased from 69% to 85%. At the same time, the sensitivity of standard criteria decreased from 65% to 45%. An interesting new variable based on recovery-phase ST-segment depression was found to perform almost as well as other standard criteria. Adding such clinical findings as age and type of chest pain into multivariate models improved the diagnostic performance of the test. However, none of several newer, computer-based diagnostic algorithms were found to exceed the performance of simple visual estimation of the ST-segment alone.

The improved specificity of exercise electrocardiography in this study [7] confirms an important reduction in workup bias. Some bias may have remained in the sample because as many as 25% of the patients were actually recruited after exercise tests had been done. Comparison of test performance in this subgroup with that in the other recruited patients and in the men in the pilot study might provide additional insight into the nature of the referral process. Specificity in this study still remains lower than the 96% estimated by Morise and Diamond [8] for men totally free of workup bias. Perhaps novel biases entered the study as a result of the many overt and subtle reasons why patients might consent to have angiography at the outset of evaluation for chest pain. Other bias might enter as a result of selection of some patients but not others for original referral to this protocol. Despite these issues, this population is an important resource for future study.

The poor sensitivity of standard exercise electrocardiographic criteria in this population [7] is another consequence of reduced workup bias. Positive standard test results may occur in only 40% of patients with single-vessel disease but in 80% or more of those with three-vessel or left main coronary disease [1, 2, 9]. The reduced effect of test outcome on the decision to refer patients to this study therefore favors the inclusion of patients whose obstruction is relatively milder than that present in most catheterized populations. Low test sensitivity also highlights the inadequacy of ST-segment criteria alone for the detection of ischemia. "False-negative" results on standard tests are common in patients with limited effort tolerance, those receiving ß-blockers, and those whose test results are defined as negative when the ST-segment depression slopes upward [1, 2]. The large subgroup of this study with false-negative results might also include some patients, free of workup bias, whose chest pain syndromes are causally unrelated to the coronary obstruction detected at angiography, particularly when this obstruction is defined by only 50% diameter stenosis. Exploration of the performance of the various criteria in relation to the underlying severity of coronary disease in this group is therefore of interest. The improved performance seen in these patients with the addition of clinical variables in multivariate models emphasizes the importance of judgment in clinical prediction [2, 4], but this process does not directly contribute to the improvement of electrocardiographic diagnosis itself.

Beyond the important issue of workup bias, this study addresses a current controversy about the value and limitations of newer approaches to analysis of the exercise electrocardiogram [9-12]. These approaches include quantification of non-ST-segment exercise findings, such as QRS duration, QRS scores based on measurement of amplitudes, axis change, QT dispersion, and T-wave alternans, as well as algorithms based on workload adjustment of ST-segment depression. Simple heart rate adjustment of ST-segment depression has been shown to markedly improve the sensitivity of exercise electrocardiography in uncatheterized patients with stable angina as well as in patients with angiographically demonstrated coronary disease, and it is particularly useful for correctly classifying many "false-negative" results due to upsloping ST segments [9, 12]. This method has also been shown to predict future coronary events in asymptomatic, low-risk men and women in the Framingham Offspring Study [13] and in higher-risk men in the Multiple Risk Factor Intervention Trial [14].

The similar performance of the ST/HR [heart rate] index and standard visual criteria in this study is therefore surprising. It might be explained by differences in methods or in study populations. However, a recent analysis by Morise [15] showed that the ST/HR index method actually performs best in patients without workup bias, in whom it may double the poor test sensitivity of standard test criteria at matched high specificity for the identification of coronary disease. Indeed, Froelicher and colleagues' Figure 2 suggests that sensitivity is higher for the ST/HR index than for visual criteria at almost all clinically relevant test specificities above 85%; thus, incremental value for the ST/HR index may emerge in this range as the investigators further examine subgroups of this population.

Whatever the resolution of this disagreement, the recognition of test limitations should not arrest enthusiasm for continued improvement of exercise electrocardiography. The findings of Froelicher and colleagues [7] confirm that specificity of this test is better than we generally think. The poor sensitivity of the test remains a fundamental challenge for the development of new criteria that can improve the test's usefulness in clinically focused populations. Computer-assisted methods should facilitate this development by improving precision of measurement and simplifying calculation of new algorithms, and the database organized by the Quantitative Exercise Testing and Angiography [QUEXTA] investigators can play a major role in this effort. Reports of the death of the exercise electrocardiogram are exaggerated, but rehabilitation of the method now will be far more productive than attempts at resuscitation later.