Exercise Electrocardiogram Testing

Abstract

HomeCirculationVol. 114, No. 19Exercise Electrocardiogram Testing Free AccessReview ArticlePDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissions ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toFree AccessReview ArticlePDF/EPUBExercise Electrocardiogram TestingBeyond the ST Segment Paul Kligfield, MD and Michael S. Lauer, MD Paul KligfieldPaul Kligfield From the Division of Cardiology, Weill Medical College of Cornell University and the Cornell Center of the New York–Presbyterian Hospital, New York, NY (P.K.); and the Departments of Cardiovascular Medicine, Quantitative Health Sciences, and Epidemiology and Biostatistics, Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, Ohio (M.S.L.). Search for more papers by this author and Michael S. LauerMichael S. Lauer From the Division of Cardiology, Weill Medical College of Cornell University and the Cornell Center of the New York–Presbyterian Hospital, New York, NY (P.K.); and the Departments of Cardiovascular Medicine, Quantitative Health Sciences, and Epidemiology and Biostatistics, Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, Ohio (M.S.L.). Search for more papers by this author Originally published7 Nov 2006https://doi.org/10.1161/CIRCULATIONAHA.105.561944Circulation. 2006;114:2070–2082Exercise testing remains the most widely accessible and relatively inexpensive method for initial evaluation of suspected coronary disease and for evaluation of its severity.1–3 Clinical usefulness has been limited, however, by poor sensitivity of standard ST-segment depression criteria for assessment of anatomic and functional coronary disease severity and for prediction of risk.1,2,4–6 Recent data make it clear that symptomatic obstructive plaques that typically result in exercise-mediated ischemia may be less relevant to infarction and sudden death than less obstructive unstable plaques.7 These limitations mandate a rethinking of the exercise ECG along 2 distinct lines: First, is it possible to improve the diagnostic value of the exercise ECG? Second, separate from its ability to diagnose obstructive coronary artery lesions, can the exercise test be used as a prognostic tool that can encourage effective prevention of premature deaths or coronary events? Both goals take us beyond the ST segment.Beyond the ST Segment: What Are We Looking for?Reversible ST-segment depression is the characteristic finding associated with exercise-induced, demand-driven ischemia in patients with significant coronary obstruction but no flow limitation at rest. This process differs from the flow-limited acute coronary syndromes because exercise-related ischemia is generally limited to the subendocardium and is proportional to increases in myocardial oxygen demand. Ventricular waveforms of the ECG can be related to the net uncanceled transmural gradients between endocardial and epicardial myocardium, as extrapolated from the work of Holland and Brooks, among others.8–10 Accordingly, isoelectric TQ and ST segments in normal and in nonischemic patients can be related to comparable resting membrane and action potential plateau voltages in endocardial and epicardial action potentials. During exercise, progressive ischemia results in changing endocardial action potentials during both diastole and systole. Less negative endocardial cell resting membrane potential leads to current flow across the ischemic boundary during diastole, leading to elevation of the TQ segment on the ECG. Lower endocardial plateau voltage leads to current flow during systole, leading to ST-segment depression.11These combined diastolic and systolic effects of subendocardial ischemia produce ST-segment depression. The magnitude of ST depression during ischemia is related to spatial and nonspatial factors.12 The spatial factor is roughly the area of ischemic myocardium; the larger the involved area, the greater is the ST depression. The first nonspatial factor is the physiological severity of ischemia within the affected myocardium,12 which becomes progressively greater during exercise and is largely responsible for the changing amount of ST depression during the test. An additional nonspatial factor that may affect ST-segment depression during exercise-induced ischemia is changing intraventricular conductance.9,12It is immediately apparent from these considerations with regard to mechanisms that many conditions can confound the sensitivity and specificity of the ECG for the detection of coronary disease and for the anatomic, functional, and prognostic predictive value of the ST-segment response. These confounders include the presence of nonobstructive but vulnerable lesions in the coronary arteries, subthreshold nonspatial severity of ischemia with inadequate effort tolerance or blunted heart rate response to exercise, any cause of high endocardial wall stress that can limit subendocardial flow reserve, and electrolyte and drug effects on the action potentials, as well as ST-segment cancellation from opposingly oriented discontinuous areas of ischemia, among others.6,11,13 These considerations and the past 50 years of experience have taught us that we need to go beyond the ST segment.Before alternative exercise test methods are considered, it is important to acknowledge inherent limitations in diagnostic accuracy of any noninvasive test. Alternative methods of imaging during ischemia, such as echocardiography and perfusion scanning, reach beyond the ST segment and have inherent limitations; a detailed discussion of these tests extends beyond the scope of this report.Tests should be evaluated in the population to which they will be applied. However, most of the diagnostic literature is based on patient samples in which all subjects had both the diagnostic test and a coronary angiogram.6 Because patients who did not undergo angiography were not considered, the reported test accuracy is inaccurate. This phenomenon of “referral bias,” also known as “verification bias” or “workup bias,” occurs because the decision to perform the gold standard test (eg, angiography) is heavily influenced by the results of the diagnostic test and physicians’ faith in them. Verification bias inflates estimated sensitivity and deflates estimated specificity.14–17Test sensitivity is also dependent on the study population and generally increases with the severity of disease. The standard exercise test tends to correctly identify patients with high-grade proximal multivessel or left main coronary disease but to incorrectly miss patients with less severe obstruction.18 Test performance is dependent on the admixture of disease in the population, making comparison of test performance difficult across broadly defined groups. Too much effort has been expended in comparing one test with another.19 We are not necessarily looking for one ECG feature that will improve identification of disease but perhaps a logical combination of findings.Another important interpretive limitation of the exercise test has been its dependence on coronary angiography as the diagnostic gold standard. Prognostically important coronary disease may be present in the absence of hemodynamically obstructive lesions.7,20 Under conditions of stress, inadequate coronary vasodilatation or paradoxical constriction may generate ischemia without any requirement for fixed resting stenoses.21,22 Thus, the prognostic value of a test for a clinical outcome may differ strikingly from its sensitivity and specificity characteristics for the presence of underlying obstructive disease.Beyond the ST Segment: Diagnostic Value of the ECG for Detection of Ischemia or Obstructive Coronary Artery DiseaseDependence on a discrete ST-segment threshold for the definition of exercise-induced myocardial ischemia reduces the sensitivity of the standard exercise test.1,4,23 As a single continuous variable, the magnitude of measured ST depression at peak exercise limits test performance to that defined by the intrinsic reciprocity of sensitivity and specificity. Improvement in test performance of the exercise ECG cannot occur without the incorporation of additional information beyond the ST segment. As a corollary, new criteria and combinations of findings should be capable of altering the performance of any diagnostic test.3,19To go beyond limitations of the ST segment, it is necessary to consider the effects of exercise-related, demand-induced ischemia on other features that can be extracted from the ECG. Heart rate normally increases through exercise in proportion to myocardial oxygen demand24 and is therefore related to the onset and severity of ischemia. Even if not contained within a single cardiac cycle, heart rate is an intrinsic part of the ECG and can be measured easily. Other ischemic ECG findings include changes in QRS duration, QRS amplitudes, QT intervals, observed QT dispersion, and subintervals of repolarization, such as the duration from the peak to the end of the T wave.Heart Rate Adjustment of ST DepressionST depression during exercise-induced ischemia is dependent not only on the presence of coronary obstruction but also on the increase in excess myocardial oxygen demand as workload increases.25–28 This suggests a physiologically sensible principle: Because ST segment depression changes throughout the course of exercise, it must reflect more than just coronary obstruction.23 Because changes in heart rate are related to changes in myocardial oxygen demand,26–29 in the presence of limited coronary blood flow there should be a progressive relationship between the degree of ST depression and increasing heart rate.23,30,31 Adjustment of ST depression for changes in myocardial oxygen demand related to heart rate is therefore physiologically logical.18,23Two methods of heart rate adjustment of ST-segment depression during exercise have evolved (Figure 1). Methodologies of the linear regression–based ST segment/heart rate (ST/HR) slope and the simpler ST/HR index have been detailed elsewhere.18,23,32–35 Much of the improved sensitivity results from correct classification of threshold levels of upsloping ST-segment depression that is classified as “equivocal” because upsloping depression is common in normal subjects.18,36 Heart rate adjustment of subthreshold ST depression <0.1 mV also results in correct classification of “false-negative” tests, but this will only occur when ST-segment measurement is precise. With high sensitivity but lower specificity, these methods serve as a reasonable screen for identification of 3-vessel or left main coronary artery disease37 and are more accurate than standard criteria for identifying extensive obstruction as defined by high Duke jeopardy or Gensini scores38 or by larger reductions in exercise ejection fraction during exercise radionuclide cineangiography.39Download figureDownload PowerPointFigure 1. Schematic definition of the ST/HR slope and the ST/HR index. Because the rate of change of ST depression in relation to change in heart rate during ischemia is greatest at the end of exercise, the ST/HR index systematically underestimates the ST/HR slope. (ST depression is plotted as positive on the y axis.)Watanabe et al40 found useful predictive value of the ST/HR index during supine bicycle testing for identification of 3-vessel disease and significant correlation with the Gensini score. More recently, favorable reports from Lee et al41 and Hsu et al42 have supported the ST/HR index for improved detection of disease. The ST/HR index alone and a summation of index values have also been demonstrated to have enhanced value for prediction of restenosis after percutaneous transluminal coronary angioplasty.43,44 Hamasaki et al45 recently developed a criterion of subtracting the ST/HR index from the ST/HR slope, which has allowed for detection of coronary artery disease in patients on digoxin. However, improved performance with the use of heart rate–adjusted measures of ST depression has not been found in all studies, perhaps in part because of methodological and population differences.23,46–49QRS DurationSympathetic stimulation increases conduction velocity, whereas ischemia tends to decrease conduction velocity by slowing the rapid upstroke (phase 0) of the ventricular action potential. It has been postulated that differences in QRS duration from rest to exercise might serve as a marker of ischemia. A subtle prolongation of QRS duration during exercise was demonstrated by Ahnve et al50 in 1986. Modest exercise QRS shortening in normal subjects was found by Michaelides et al.51 The magnitude of change in these studies was small, in the range of 3 ms of shortening in normal subjects and 6 to 8 ms of lengthening in coronary disease patients. Berntsen et al52 were able to associate more marked exercise-induced QRS prolongation, in the range of 15 ms, with increased risk for subsequent ischemia-related ventricular tachycardia. Computer-based optical scanning for more precise measurement of QRS duration during exercise testing was introduced by Cantor et al53 and was found to outperform standard ST-segment criteria for identification of disease in women54,55 and for the detection of post–percutaneous transluminal coronary angioplasty restenosis.56 These methods are amenable to computer-based implementation in digital ECGs.QRS AmplitudesQRS amplitudes have been examined in several ways as markers for ischemia. According to the “Brody hypothesis,” other things being equal, the R-wave amplitude recorded by the surface ECG should be proportional to chamber size and the ischemic dilatation of the left ventricle may thereby be recognized. Bonoris et al57 demonstrated in 1978 that failure of R-wave amplitude to decrease during exercise was associated with modestly useful sensitivity and specificity. Adjustment for R-wave amplitude improved the performance of the Hollenberg treadmill score.58,59 Other studies have been less positive, including the finding of opposite directional changes.60 Detrano et al32 found that R-wave behavior with exercise alone had limited performance but that there was some diagnostic advantage in normalizing ST depression for R-wave amplitude. Although Ellestad et al61 found some value in R-wave normalization, this was limited to patients with extreme amplitudes.Another algorithm, the Athens QRS score, has had increasing support in recent reports.62,63 The score examines net amplitudes in leads aVF and V5 by subtracting Q and S amplitudes from the R wave in each lead at rest and during exercise and then subtracting the exercise result from the rest result in each lead. Values for these rest-exercise differences in each lead are then added to produce the resulting score. Initial findings suggested that the score was inversely related to the anatomic extent of disease, always associated with disease when negative, and independent of the presence or absence of ST depression.62 It was also found to be inversely related to the extent of ischemic wall motion abnormalities63 and to the magnitude of exercise-induced, handgrip-induced, and dipyridamole-induced perfusion abnormalities.64,65 Subsequent studies indicated value of the Athens score for the detection of restenosis and for ischemia after bypass surgery.66 Koide et al67 demonstrated that the Athens QRS score added complementary diagnostic information to ST-segment depression.High-frequency components of the QRS complex that are not usually represented in the routine ECG tracing, between 150 and 250 Hz, have been shown to decrease in the presence of acute ischemia.68 An initial report found that reduction of high-frequency forces during exercise may have useful sensitivity and specificity for the detection of coronary disease.69QT Interval and T-Wave SubintervalsEvaluation of the QT interval during exercise is subject to a number of methodological problems. Noise and baseline drift increase the difficulty of determination of the end of the low-frequency T wave. Fusion of the end of the T wave with the subsequent P wave as heart rates rise further obscures the end of repolarization. QT-interval restitution is not instantaneous with change in cycle length but depends on the rate and direction of changing cycle length, which varies throughout exercise. Moreover, the rate of QT-change with exercise is different in men and in women.70,71 Limitations notwithstanding, a number of studies have suggested that lengthening of the rate-corrected QT interval with exercise identifies myocardial ischemia,72–74 particularly in patients with exercise-related ventricular arrhythmias.75An immediately confounding difficulty with QTc, however, is the generally different peak exercise heart rates that are found in patients with and without disease and the general inapplicability of traditional rate correction algorithms within the exercise environment. Lax et al76 demonstrated that Bazett-corrected QTc interval prolongation (the measured QT divided by the square root of the RR interval) was only minimally greater in patients with coronary disease than in normal subjects when measured at corresponding subthreshold heart rates. The small differences present in QTc at peak exercise were also present at rest, arguing against important diagnostic value for the QTc during exercise testing.On the other hand, differences between normal subjects and patients with coronary disease in a marker of precordial QT dispersion in the study by Lax et al76 diverged throughout exercise at all heart rates and provided information that was complementary to standard measures of ST-segment depression. QT dispersion can be defined simply as the difference between the longest and shortest measured QT in any lead or, alternatively, as a standard deviation of all measured QT intervals in the ECG. A number of studies have indicated useful predictive value of QT dispersion for the identification of disease and ischemic contractile dysfunction during exercise67,77–81 and after adenosine infusion.82 Although considerable controversy exists about the relationship of QT dispersion to heterogeneity of repolarization, it is clearly dependent on underlying T-wave morphology, which might be expected to be sensitive to exercise-induced ischemia. Further examination of T-wave subintervals during exercise is warranted, such as the T peak to T end duration.83,84 Other T-wave measures that have been found to have prognostic value in the resting ECG, such as principal component analysis,85,86 require clarification in the exercise test environment.Recovery Phase MethodsBecause postexercise ST-depression resolution is asymmetrical with respect to heart rate in patients with myocardial ischemia, the pattern of ST-segment change in relation to heart rate during the recovery phase of exercise also has diagnostic value.23,87–90 Data can be examined qualitatively as the simple rate-recovery loop,91 with findings from the first minute of recovery, when patients with ischemia generally have greater ST-segment depression than was present at the corresponding heart rate during exercise before peak effort (Figure 2). In contrast to standard ST-depression criteria and heart rate–adjusted criteria derived purely from exercise phase data, the sensitivity of the rate-recovery loop appears to be relatively independent of the extent of disease. The rate-recovery loop has been found to be more useful than ST-depression criteria for the identification of restenosis after angioplasty.92Download figureDownload PowerPointFigure 2. Normal and abnormal rate-recovery loops. In patients with ischemia, ST depression relative to heart rate is greater during recovery than during exercise. (ST depression is plotted as positive on the y axis.) Reproduced from Okin and Kligfield23 with permission from the American College of Cardiology Foundation. Copyright 1995.Recovery phase and exercise phase ST behavior were incorporated by Hollenberg et al58,59 into a treadmill exercise score that is adjusted for heart rate. A number of more recent studies have confirmed the diagnostic value of combining exercise and recovery phase ST-segment data in the heart rate domain. A relatively simple quantification of the rate-recovery loop involves calculation of the ST-segment “deficit” between recovery phase ST depression at 3.5 minutes and the ST depression at the corresponding heart rate during exercise.89,93 ST/HR hysteresis, as developed by Lehtinen et al,88,94,95 integrates the area of ST-segment depression with respect to heart rate that is included in the exercise and recovery loop over the heart rate range included in the first 3 minutes of recovery. This integral is then divided by the heart rate difference (ie, the maximum heart rate during exercise minus the minimum heart rate during recovery) of the integration interval to normalize the result with respect to the postexercise heart rate decline.In one large study, comparison of receiver operating characteristic curves demonstrated that performance of ST/HR hysteresis exceeded that of the ST/HR index and standard test criteria for the detection of coronary artery disease.88 A recent evaluation by Svensburgh et al90 has addressed the effect on test performance of the exact range of heart rates included in the rate-recovery area calculation, noting that a large percentage of the loop needs to be considered to optimize diagnosis, especially in women. Bigi et al96 examined the entire rate-recovery loop by defining a stress-recovery index as the difference in areas under the full exercise and recovery phase ST/HR plots. The stress-recovery index has been found to be more accurate than other standard ST segment– and heart rate–adjusted test methods for the identification of anatomically extensive disease after myocardial infarction,96 for the prediction of mortality after myocardial infarction,97 and for the prediction of all-cause mortality in hypertensive patients with chest pain.98Thus, combination of exercise and recovery phase ST-segment data as an area of the ST/HR loop appears to be the most accurate and predictive of the current heart rate–adjusted methods in routine exercise testing. Recovery phase behavior of other potential ECG markers of ischemia, such as QRS duration, Athens QRS score, and QT-interval derivatives, also deserve careful attention.90,96–98 As shown by Koide et al67 for complementary performance of ST depression and Athens QRS score, it is reasonable to predict that combinations of these methods should be capable of improving the performance of the exercise ECG for the detection of those patients who have demand-induced ischemia. Accordingly, we believe that continued development and multicenter evaluation of the applied value and limitations of these newer methods are warranted.Application in Current Clinical PracticeWe recommend the incorporation of the simple ST/HR index into routine exercise test evaluation, and we believe that other measurements and combinations of measurements require further evaluation and/or technical implementation before application becomes practical.19 Reasons for use of the ST/HR index include simplicity of calculation, improvement in test sensitivity by resolution of otherwise “equivocal” test responses, and demonstrated prognostic value in Framingham men and women and in higher-risk Multiple Risk Factor Intervention Trial (MRFIT) participants.18,99–101 The ST/HR index is derived by dividing the maximal additional change in ST-segment depression at end exercise, measured in microvolts (where 1.0 mm at standard gain=100 μV) at a constant 60 ms after the J point,46 by the corresponding change in heart rate from upright control. Additional ST depression compared with upright control is determined only by deviation below the isoelectric line, with excursion from any ST elevation to baseline not included.23 This requires no automated computer algorithm for calculation as needed for routine application of the more complex ST/HR slope and ST/HR hysteresis determinations, but precision of measurement contributes to test accuracy, particularly when subthreshold ST depression is present.49 In practice, averaging of 3 successive complexes is effective. Alternatively, computer-averaged ST-depression values can be accurate to 0.1 mm (10 μV), but these always require visual verification. Values of the ST/HR index >1.6 μV/bpm are abnormal.Although the linear regression–based ST/HR slope and ST/HR hysteresis have been implemented in some computer-based algorithms,35 these are not yet ready for widespread implementation. This is due to technical calculation issues and other methodological factors.23,47 Clinicians who find value in the simple ST/HR index may wish to explore the ST/HR slope and hysteresis in greater detail. Implementation of QRS duration, QRS score, and high-frequency QRS findings awaits confirmatory studies and practical application of automated signal-averaged QRS measurements in routine exercise ECG recording equipment. QRS amplitude change, as well as QTc, has been of disappointing value when used alone; whether these will find a role in expanded algorithms remains to be seen. Newer algorithms examining QT hysteresis101a and T-wave shape require further evaluation and are not yet ready for practice.Beyond the ST Segment: Prognostic Value of the Exercise ECGPrediction of future coronary events and mortality can be separated from the identification of obstructive disease as an important use of the exercise ECG. We will focus on methods that are intrinsic to the information contained in the exercise test, such as effort tolerance, and within the ECG signal itself, including heart rate changes and ventricular arrhythmias. Methods with demonstrated prognostic value include simple heart rate adjustment of ST-segment depression, measurements of functional capacity, chronotropic competence and incompetence, heart rate recovery, and frequent ventricular ectopy during recovery (Table 1). Exercise test scores, which incorporate clinical and demographic risk factors not based on the exercise test, will be examined briefly. TABLE 1. Summary of Major Diagnostic and Prognostic Exercise Test MeasuresMeasureDescriptionPopulations StudiedCommentsCAD indicates coronary artery disease; HR, heart rate.ST/HR indexMaximum change in ST depression/change in HR: (Peak Exercise−Upright Control ST)/(Peak Exercise−Upright Control HR)Catheterized patients, clinical CAD without catheterization, asymptomatic high-risk men and low-risk men and womenIncreases sensitivity for the detection of CAD and predicts mortality and cardiovascular eventsConsider abnormal if >1.6 μV/bpmST/HR slopeGreatest statistically significant slope by linear regression relating ST depression to HR during exerciseCatheterized patients, clinical CAD without catheterization, clinically normal men and womenIncreases sensitivity for the detection of CAD and for identification of anatomically and functionally severe CAD when markedly abnormalConsider abnormal if >2.4 μV/bpmConsider markedly abnormal if >6.0 μV/bpmEstimated functional capacity in METsBased on protocol and exercise timeSymptomatic and asymptomatic men and womenStrongly predictive of mortality and cardiovascular events (although prognostic value of <85% of predicted has only been validated in women)Predicted value in men: 14.7−0.11×AgePredicted value in women: 14.7 −0.13×AgeConsider abnormal if <85% of predictedChronotropic responseProportion of HR reserve use calculated as (Peak HR−Resting HR)/(220 −Age−Resting HR)Symptomatic and asymptomatic men and womenPredictive of mortality and cardiovascular events; limited evidence regarding usefulness with β-blockersConsider abnormal if ≤80% (≤62% for patients on β-blockers)HR recoveryDifference between HR at peak exercise and HR 1 or 2 min laterSymptomatic and asymptomatic men and womenPredictive of mortality, cardiovascular events, and sudden cardiac deathWith upright cool-down period, abnormal if ≤12 bpm 1 min into recoveryWith immediate supine position, abnormal if ≤18 bpm 1 min into recoveryWith sitting, recovery abnormal if ≤22 bpm 2 min into recoveryVentricular ectopy during recoveryFrequent ventricular ectopics (>7 bpm), couplets, bigeminy, trigeminy, ventricular tachycardia, or fibrillationSymptomatic and symptomatic men and womenUncommon but predictive of all-cause mortalityDuke treadmill scoreMinutes (Bruce Protocol)−5×ST-Segment Deviation−4×Angina ScoreSymptomatic and asymptomatic men and womenPredictive of cardiovascular mortality and all-cause mortalityIf protocol other than Bruce used, convert to estimated Bruce minutes based on METsST segment must be ≥1 mm horizontal or sloping away from the isoelectric line to be countedAngina score=1 if not test-limiting, 2 if test-limitingValue of ≥5 low-risk, between −10 and 5 intermediate risk, and <−10 high riskFunctional CapacityIt may be argued that the most important prognostic marker obtained by the exercise test is functional capacity or the amount of work completed before exhaustion. Extensive literature in asymptomatic cohorts has demonstrated that functional capacity is a powerful independent predictor of all-cause and cardiovascular mortality.102–106 More recently, functional capacity has been studied within a clinical context. For example, a study of >3000 patients undergoing stress testing with single-photon emission computed tomography nuclear myocardial perfusion imaging demonstrated that functional capacity was at least as strong a predictor as perfusion defects for prediction of all-cause death.107 Similarly, a study of an angiographic cohort showed that functional capacity was a stronger predictor of death than angiographic severity of coronary disease and ST depression.108 When functional capacity and the presence or absence of angiographic coronary di