Exercise electrocardiography testing has been a valuable noninvasive tool for many decades in the evaluation of patients with suspected or known coronary artery disease (CAD) and remains widely utilized. The primary goals of noninvasive stress testing are to aid in the diagnosis of obstructive CAD and to provide risk stratification in an attempt to estimate the probability of myocardial infarction (MI) or death. Although cardiac stress imaging can improve the diagnostic yield and enhance risk stratification beyond exercise electrocardiogram (ECG) variables alone, the standard exercise ECG stress test should play a key role in the evaluation of patients with suspected CAD. The exercise ECG continues to be of great value, especially as the CAD paradigm has shifted from an emphasis on diagnosis to the importance of risk stratification.

It is beyond the scope of this chapter to review the instrumentation, patient preparation, technical considerations, detailed instructions, and interpretation of exercise ECG testing; much of this material is discussed in Chapter 8. Additional information is available for review in prior publications from the American Heart Association.^1–4 As a general rule, exercise testing is most commonly performed on a treadmill using a standardized protocol that incrementally increases both the speed and grade of the treadmill in an attempt to achieve maximal cardiac stress. A variety of exercise protocols have been developed, with the Bruce protocol being the most commonly used. Despite the strong clinical validation of the Bruce protocol, the use of a singular protocol for all patients is not appropriate.

Exercise testing is generally a safe and well-tolerated procedure, yet clinical judgment and appropriate supervision should be employed. Death or MI may occur in up to 1 per 2500 tests.^2,5 Absolute and relative contraindications to exercise testing and indications for the termination of exercise testing are summarized in Chapter 8.²

The standard criterion of an abnormal exercise ECG response is ST-segment depression (horizontal or down-sloping) ≥1 mm 80 ms after the J-junction. Using the standard 12-lead ECG, ST-segment depression in lead V₅ has the greatest diagnostic value for CAD, whereas ST-segment depression confined only to the inferior leads is of little value. ST-segment depression does not localize areas of ischemia. ST-segment elevation in leads without Q waves occurs infrequently. This finding represents transmural ischemia and, as a result, localizes ischemia and the culprit vessel. Exercise-induced ST-segment elevation in lead aVR has been utilized to improve detection of left main or ostial left anterior descending artery disease.⁶

It is well recognized that certain drugs (beta-blockers and nitrates) reduce test sensitivity, and resting baseline ECG abnormalities (left ventricular hypertrophy with strain, digoxin effect, and ST-segment depression) reduce test specificity and should be taken into consideration for test interpretation. Bayes’ theorem states that the greatest yield of testing for diagnostic purposes occurs in patients who have intermediate (10–90%) pretest probability of CAD. Individuals at the extremes of pretest probability, either very low (younger women with atypical chest pain) or very high (older men with typical angina), derive little benefit from testing. The ACC/AHA Exercise Test guidelines² recommend against performing an exercise test for diagnostic purposes in these patient subsets. The ACC/AHA guidelines as well as other organizations generally discourage exercise testing with our without imaging in asymptomatic patients.^7,8

There are multiple meta-analyses reviewing the diagnostic accuracy of exercise treadmill testing.^2,9–11 Sensitivity is generally in the range of 65% and specificity 80%. It is important to appreciate that these apparent values for sensitivity and specificity have not been corrected for referral bias (also called verification bias). Referral bias is a concept that addresses the preferential referral of patients with positive test results to coronary angiography, the subset of patients in whom sensitivity and specificity are determined. The net impact of referral bias is to artificially inflate test sensitivity and to artificially decrease test specificity. Approaches to overcome referral bias include catheterizing the entire population of patients who undergo evaluation for CAD¹² or adjusting the sensitivity and specificity values by applying mathematical corrections.¹³ Posttest referral bias alters not only standard exercise treadmill test (ETT) results, but also stress imaging results.^14,15

A multitude of valuable prognostic information can be obtained during an exercise ECG test. Variables obtained through the ETT may reflect overall cardiovascular and physical fitness as well as function of the autonomic nervous system. Exercise variables that have been shown to predict outcome include: exercise duration, chronotropic incompetence, heart rate recovery, exercise hypotension, exercise hypertension, and ventricular ectopy.^16,17 Exercise ST-segment depression contains prognostic information but generally is a weaker parameter than these other variables. Exercise duration has been demonstrated in many studies to be the strongest prognostic variable (Fig. 26-1).^2,18–23 Several studies have demonstrated the relationship of exercise duration to myocardial ischemia and subsequent events. Patients achieving ≥10 metabolic equivalents (METS) have demonstrated a lower prevalence of myocardial ischemia by stress SPECT compared to patients achieving <7 METS and also low rates of nonfatal myocardial infarct (0.7%/year) or cardiac death (0.1%/year).^24,25 Available data on whether exercise hypertension (commonly defined as a systolic blood pressure during exercise >190–220 mm Hg) corresponds to an increased risk of cardiac events are unclear,^26,27 whereas exercise hypotension has been associated with an increased (threefold higher) risk of future cardiovascular events over a 2-year period.²⁸ Chronotropic incompetence is the failure of the heart rate to increase appropriately with exercise (defined as <80% of the predicted value) and the proportion of heart rate reserve used during exercise can be calculated by the formula: (heart rate^peak − heart rate^rest)/(220 − age − heart rate^rest).¹⁷ Chronotropic incompetence has been shown to predict all-cause mortality and cardiac death.^29,30 Impairment of heart rate recovery and heart rate variability have also been shown to predict all-cause mortality and cardiovascular events.^31,32 While sustained episodes of ventricular arrhythmias are uncommon and can be mediated by ischemia, shorter periods of ventricular ectopy (isolated ventricular premature complexes, couplets, or nonsustained ventricular tachycardia) during exercise or in the recovery period occur more often. The prognostic importance of these “short” episodes of ventricular ectopy is uncertain.³³ The relationship between exercise-induced ventricular ectopy, myocardial ischemia, and left ventricular systolic function remains ill defined.

Many prognostic scores have been developed and subsequently validated using combinations of the variables described above. The most commonly used score is the Duke treadmill score that is calculated using three exercise parameters (exercise duration, ST-segment depression, and angina). The Duke treadmill score = exercise time (minutes based on the Bruce protocol) − (5 × maximum ST-segment deviation [in millimeters]) − (4 × exercise angina [0 = none, 1 = non-limiting, and 2 = exercise limiting]). The Duke treadmill score stratifies patients into low-risk (score ≥+5, annual cardiovascular mortality 0.25%), intermediate-risk (score +4 to −10, annual cardiovascular mortality 1.25%), or high-risk (score <−10, annual cardiovascular mortality 5%) categories (Fig. 26-2).^34,35 This score was initially developed using the Duke University treadmill database and has subsequently been validated by several other investigators.^36,37 Applying all available prognostic information and not just the three variables that comprise the Duke score enhances the accuracy of risk stratification.³⁸

Comparison of ECG–ETT with Stress Imaging

The standard ETT and stress imaging study each has advantages compared to the other (Table 22-1). The major advantage of ETT is lower cost. The charge for an exercise stress SPECT study is approximately five to seven times higher than a standard ETT. Stress imaging does have several advantages over the standard ETT. In patients with significant abnormalities on the resting ECG (left bundle branch block, paced ventricular rhythm, ventricular pre-excitation, and ST-segment depression ≥1 mm), the stress ECG will be either uninterpretable or inaccurate. Patients with percutaneous coronary intervention (PCI) and/or coronary artery bypass graft surgery (CABG) are more likely to have significant resting ECG abnormalities. In these patients another important clinical issue is localization of ischemia if further revascularization is being considered. Finally, the sensitivity of the stress ECG in patients who undergo pharmacologic stress, especially vasodilator stress, is very low. In these patients pharmacologic stress needs to be combined with imaging. The ACC/AHA guidelines^2,39 acknowledge these issues and recommend that stress imaging preferentially be performed instead of standard ETT as the initial stress modality in the following patients: (1) high pretest probability of CAD; (2) inability to exercise; or (3) significant resting ECG abnormalities.

Many studies have demonstrated that stress imaging also has higher sensitivity⁴⁰ and provides incremental prognostic accuracy^36,41–45 compared to the standard ETT. However, these studies did not indicate which subsets of patients benefit or what percentage of the population can be more accurately risk stratified. An important question is whether the standard ETT can identify a subset of patients whose event rate is so low that imaging fails to add additional prognostic information or is not cost effective.

To address this issue, several studies have compared the ability of an approach using just clinical and exercise ECG variables versus an approach using clinical, exercise ECG, and nuclear imaging variables to identify patients with severe (left main and/or three-vessel) CAD at angiography and to predict clinical outcome.^46–49 The study populations were restricted to patients with a normal resting ECG. The rationale for studying only patients with a normal resting ECG was twofold: (1) the majority of patients with a normal resting ECG have normal left ventricular ejection fraction (LVEF) and (2) the exercise ECG is more accurate if the resting ECG is normal. Approximately 95% of patients with a normal resting ECG undergoing evaluation for CAD have a normal LVEF when directly measured by a variety of imaging techniques.^50–53 In addition, in patients with a normal resting ECG, the specificity of the exercise ECG is much higher compared to those with resting ST-T abnormalities.²