Electrocardiographic Exercise Stress Testing for Cardiac Risk Assessment in Patients Undergoing Noncardiac Surgery

Over a 2-yr period, 204 consecutive patients with or at high risk for coronary artery disease who underwent elective noncardiac surgery requiring general anesthesia were enrolled prospectively in the study. Written informed consent was obtained, and the study protocol was approved by the Ethics Committee of the University of Ulm. Detailed criteria for entry into the study included at least one of the following: definite coronary artery disease as indicated by previous myocardial infarction, typical angina pectoris 12or angiographically assessed significant stenosis of coronary arteries, or high risk for coronary artery disease suggested by previous or current vascular surgery, or the presence of at least three of the following cardiac risk factors: hypertension, smoking, dyslipidemia (serum cholesterol > 6.7 mm/l), or diabetes mellitus.

We defined a previous myocardial infarction as either a positive history, electrocardiographic evidence of a previous myocardial infarction, or arteriographic evidence of coronary occlusion or narrowing and corresponding ventricular asynergy or wall motion abnormalities not caused by ischemia alone. 12Patients with left bundle-branch block or cardiac pacemakers and patients who could not perform bicycle ergometry were not eligible. A total of 19 patients were excluded from the study after performance of the exercise electrocardiography: eight because of subsequent withdrawal of consent before surgery, three because of planned surgery was canceled, and eight because of limited resources of ambulatory electrocardiographic recordings. Six of these 19 excluded patients had an abnormal electrocardiography exercise test result. The subsequent outcome of the excluded patients could not be evaluated because further investigation and data collection regarding cardiac events were not performed. The final evaluation was thus based on 185 patients.

Exercise stress testing was performed a median of 1 day before surgery without discontinuing cardioactive medication. A 12-lead resting electrocardiogram was recorded before exercise. The resting electrocardiogram was classified as abnormal if abnormal Q waves were present (≥ 0.04-s duration, ≥ 25% R wave) or if ST-segment depressions of 0.1 mV or more (60 ms after the J point) were present at rest. Exercise stress testing was performed on a supine bicycle ergometer (EL 900 B, Madaus Schwarzer, Munich, Germany).

The examination was begun at 25 W and increased in 2-min steps of 25 W. A 12-lead electrocardiogram was recorded once a minute at 50 mm/s during exercise and during the recovery period until the electrocardiogram reverted to its pre-exercise morphology, but at least 5 min after exertion. Blood pressure measurements were performed during the pretest period before exercise, at the end of each workload, and every 2 min thereafter during the recovery period. Exercise stress testing was symptom-limited. 12The exercise electrocardiography was classified as abnormal if new horizontal or down-sloping ST-segment depressions of 0.1 mV or more (60 ms after the J point, unless that point fell within the T wave, in which case it was measured at a minimum of J + 40 ms) were seen in three consecutive beats without baseline variation. 12,13If ST-segment depression was present at rest, an increase of the depression by 0.1 mV or more was classified as pathologic. 13ST-segment elevations of 0.2 mV or more were regarded as pathologic except if they occurred in leads with signs of previous infarctions. 13Each rest and exercise electrocardiogram was evaluated by two independent investigators who were blinded to patient identity and clinical course. If the interpretations of the two investigators did not agree, a third investigator performed an independent evaluation, with a majority of two concordant interpretations prevailing.

Patients were monitored with a two-channel Holter electrocardiography recorder (LRE, Technology Partner GmbH, Munich, Germany), from the evening before the operation until the morning of the second postoperative day (total recording time, 55.5 ± 10 h). Two bipolar leads (modified V⁴and V⁵) were used. The data were evaluated by computer with an evaluation software (CAPS-L, LRE Technology Partner GmbH). The baseline ST-segment level was defined at the start of the monitoring period as the average ST-segment during a stable period (usually 10 min). Prolonged episodes of ischemia were defined as reversible ST-segment changes lasting more than 30 min 14and involving a shift from baseline of either 0.1 mV or more of horizontal or downsloping ST-segment depression (60 ms after the J point, unless that point fell within the T wave, in which case it was measured at a minimum of J + 40 ms), or 0.2 mV or more elevation (at the J point), separated from a preceding episode by at least 1 min. Every computer diagnosis of an ST-segment depression was visually confirmed to exclude false-positive ST depressions associated with artifacts or abnormal QRS complexes such as ventricular ectopic beats or conduction abnormalities. Furthermore, T-wave changes that were incorrectly detected as ST changes by the computer software were identified. Holter tapes were analyzed by two independent investigators who were blinded to patient identity and clinical and any other electrocardiographic data. The interpretation of the results was performed analogously to that of resting and exercise electrocardiograms.

For each patient, the American Society of Anesthesiologists physical status classification, 15the original cardiac risk index, 16and the modified multifactorial index 17was determined.

Two-dimensional surface echocardiography with color Doppler spectral analysis was performed by an experienced cardiologist before surgery. Echocardiographic assessment included visual determination of global left ventricular performance (graded as normal or mildly, moderately, or severely reduced) and observation of valvular and wall motion abnormalities. Regional left ventricular function was assessed from parasternal short-axis and long-axis views and apical two- and four-chamber views. Motion was graded as normal, hypokinetic, akinetic, or dyskinetic. Preoperative use of any medication and serum levels of creatinine and potassium were also analyzed (table 1).

Surgery was classified as major vascular (aortic repair, aortoiliac or iliac endarterectomy), minor vascular (carotid endarterectomy, peripheral vascular), major intraabdominal (gastric, liver, pancreatic, small bowel, colonic, or rectal) or minor abdominal (cholecystectomy, laparotomy, herniotomy).

Balanced anesthesia was used with induction of anesthesia by 0.3 mg/kg etomidate combined with 0.003–0.01 μg/kg fentanyl. Intubation was performed after application of succinylcholine (1.5 mg/kg) or vecuronium bromide (0.1 mg/kg). Bolus injections of fentanyl (0.1–0.2 mg), nitrous oxide, and volatile anesthetics (enflurane or isoflurane) were used for maintenance of narcosis. Postoperative pain control was managed with nonopioid and opioid analgesics, without the use of regional anesthesia.

A history and a physical examination were performed by one of the authors every day from the first to the fifth postoperative day and on the day of discharge. Approximately 30 days after the operation, a telephone interview was performed to obtain the cardiac history. A 12-lead electrocardiogram was obtained on the day before surgery, just before induction of anesthesia, after surgery, daily for the first 5 days after surgery, and on the day of discharge.

Twelve-lead electrocardiograms were analyzed by two independent investigators who were blinded to patient identity, clinical data, Holter electrocardiography, as well as creatine kinase and troponin-T data. The interpretation of results was performed analogously to the evaluation of the resting and exercise electrocardiograms.

Serum creatine kinase, creatine kinase isoenzyme MB, cardiac troponin-T, and creatinine levels were measured before surgery and daily for the first 5 days after surgery. For the troponin-T measurement, serum was frozen at −18°C and analyzed some weeks later. Troponin-T (reference range, 0–0.1 ng/ml) was measured by an enzyme-linked immunosorbent assay (ELISA Troponin-T, Boehringer Mannheim, Germany). Because the cause of elevated troponin-T in patients with chronic renal failure is not clarified 18,19and the ability of troponin-T to predict cardiac risk is reduced in the presence of renal insufficiency, 20troponin-T concentration more than 0.1 ng/ml was only classified as positive if renal insufficiency was excluded.

The physicians caring for the patients had access to the preoperative resting and exercise electrocardiogram and were aware of the clinical findings and any preoperative cardiac evaluation. They had no access to the final evaluation of the preoperative resting and exercise electrocardiogram, the results of Holter electrocardiography, and troponin-T measurements.

The major analyzed outcome variables were cardiac mortality and cardiac morbidity during the first 30 postoperative days. Adverse events noted by the study physicians during research visits and the abnormal findings of the 12-lead electrocardiograms, Holter electrocardiograms, and the creatine kinase and troponin-T analysis were validated separately by two investigators. These investigators were blinded to the patients’ preoperative clinical and electrocardiography data; disagreements were resolved by a third investigator.

Cardiac death was diagnosed if the patient died from a myocardial infarction, dysrhythmia, or congestive heart failure caused primarily by a cardiac condition. Perioperative cardiac morbidity was defined as the appearance of perioperative myocardial infarction or myocardial injury, unstable angina pectoris, 12congestive heart failure, or ventricular tachyarrhythmia.

The diagnosis of perioperative myocardial infarction required one of the following criteria:

• new large Q waves (≥ 0.04-s duration, ≥ 25% R wave), even in the absence of symptoms or abnormalities in serum enzymes or troponin-T levels (Q-wave infarction) 12 

• compatible history and characteristic enzyme or troponin-T changes without electrocardiography abnormalities, provided that new wall motion abnormalities are identified on echocardiography (non–Q-wave infarction) 12 

• S-T–segment and/or T-wave changes in two consecutive 12-lead electrocardiograms or an episode of new 0.1-mV or more horizontal or downsloping ST segment depression in the Holter electrocardiography more than 30 min in duration without abnormal Q waves, each of them coupled with specific enzyme or troponin-T alterations (creatine kinase MB > 20 U/l and > 6% of the creatine kinase or troponin-T > 0.1 ng/ml on at least 2 successive days). All T-wave changes in the 12-lead electrocardiogram had to be localized and not widespread (non–Q-wave infarction).

Minor perioperative myocardial cell injury was defined as troponin-T elevation more than 0.1 ng/ml on at least 2 successive days in the absence of symptoms or electrocardiography abnormalities.

Congestive heart failure was defined as the need for sympathomimetic support associated with hemodynamic and pulmonary findings consistent with the diagnosis or at least the need for diuretics because of signs or symptoms of pulmonary congestion. Ventricular tachyarrhythmia was defined as a documented ventricular tachycardia or fibrillation with hemodynamic compromise.

Sample size calculation was based on the estimated event rate of meaningful cardiac events of approximately 10% for the selected study population. 21,22The sample size was chosen in a way that the limits of the 95% confidence interval for this rate were in a range of ± 5% around the 10% mark. For a sample size of 200 subjects, the 95% confidence interval will be 6.22–15. Another reason for choosing the sample size of approximately 200 subjects was the fact that in the planned study duration of 2 yr this size would be manageable because of our clinical experience.

Potential risk factors for an adverse event were identified by univariate and multivariate logistic regression analysis. To avoid overadjustment by using too many variables in the multivariate model, all variables were subjected to univariate analysis in a first step. Variables with a P  value less than 0.1 in the likelihood ratio test were considered as potential risk factors in the multivariate model. To avoid multicollinearity, only one variable in a set of variables with a correlation coefficient greater than 0.7 was used for to the multivariate analysis. Multivariate analysis was performed by multiple logistic regression, including backward selection. Stopping criterion was a P  value of 10%. Odds ratios and their 95% confidence intervals were calculated from the model. Sensitivity, specificity, and predictive values were calculated by contingency table analysis. All analysis was performed using SAS software 6.12 (SAS Institute, Cary, NC).

Demographic and clinical data for the 185 patients and the numbers with cardiac event and variable are shown in table 1. Forty-nine percent of the patients presented with definite coronary artery disease. Vascular surgery was performed in more than 80% of the patients (23% aortic or iliac operations, 45% carotid endarterectomy). Eighty-three percent displayed at least one of Eagle’s clinical markers. 6All 185 patients had at least an intermediate risk according to the American College of Cardiology–American Heart Association guidelines for preoperative clinical risk assessment. 5With respect to the original cardiac risk index, 16approximately 70% of our patients were classified as class I. The preoperative resting electrocardiography was abnormal in 22% of patients, and the exercise stress electrocardiography was abnormal in 30%. Twenty-four percent of patients achieved 85% of their predicted maximal heart rate (table 2).

Sixteen patients (9%) had perioperative adverse cardiac events (table 3). Six patients (3%) developed a perioperative myocardial infarction; in two of these patients, the myocardial infarction was combined with acute congestive heart failure. One patient presented with acute heart failure and pulmonary edema combined with troponin-T elevation. Another nine patients had troponin-T elevations more than 0.1 ng/ml on 2 consecutive days without symptoms or electrocardiographic abnormalities. Three patients showed prolonged ischemia more than 30 min in the Holter electrocardiography. All of these three patients had troponin-T elevations more than 0.1 ng/ml on 2 consecutive days, and one of them developed a Q-wave infarction. There were two deaths within the first 30 postoperative days, but in each case, there was a noncardiac cause.

The preoperative variables significantly associated with an adverse cardiac event in the univariate analysis at P  less than 0.1 are presented in table 4. Other exercise test parameters such as abnormal blood pressure responses or shortness of breath proved to be nonpredictive in the univariate analysis. The multivariate correlates of adverse cardiac events were definite coronary artery disease, major vascular or visceral surgery, reduced left ventricular performance, and ST-segment depression of 0.1 mV or more in the exercise electrocardiography (table 5). ST-segment depression of 0.1 mV or more during exercise and ST-segment depression of 0.1 mV or more before achieving 85% of the predicted maximal heart rate, as well as ST-segment depression of 0.1 mV or more before achieving a workload of 100 W were highly correlated. Therefore, only the variable with the lowest P  value (ST-segment depression ≥ 0.1 mV during exercise) was included in the main multivariate model. Given the fact that previous studies 11,23had found that the combined variable (ST-segment depression ≥ 0.1 mV before achieving a certain percentage of the predicted maximal heart rate) was predictive of cardiac complications, we also conducted a secondary multivariate analysis including the combined variable instead of the single variable (ST-segment depression ≥ 0.1 mV during exercise). However, the results were comparable with the main model: independent predictors of outcomes were coronary artery disease, major surgery, reduced left ventricular performance, and the combined exercise stress test variable.

Risk stratification was possible with use of the two clinical independent risk factors of coronary artery disease and major surgery. Sixty-three patients without any risk factor had a risk of 0% for cardiac complications. Eighty-five patients with one of the two risk factors had a risk of 7%. The remaining 37 patients with both risk factors had an aggregated risk of 27% for cardiac complications. These latter groups were further differentiated by the findings of exercise stress electrocardiography (fig. 1). In particular, risk stratification of the group presenting with both clinical risk factors was improved: 20 of these patients who had normal exercise electrocardiographic findings had a risk of 10%, whereas 17 patients with abnormal exercise electrocardiographic findings had a risk of 47%.

The present study is the first prospective investigation to show that the occurrence of an ST-segment depression of 0.1 mV or more in a preoperative exercise stress electrocardiography is a statistically significant independent predictor of perioperative cardiac events. This finding is significant, because 76% of the patients were not able to reach 85% of their maximal predicted heart rate, and 80% of our patients received vascular surgery. This result should be emphasized because it has been assumed up to now that exercise stress electrocardiography is of little value for the preoperative risk stratification for such patients. 6,7,24 

Review articles with discussions of the role of preoperative exercise stress electrocardiography generally cite many studies 3,5,6; however, the majority of the investigations are retrospective or have an unclear study design. 23,25–31In addition, in some studies only a part of the patients were evaluated perioperatively 26,29,31,32or only a small number of patients were investigated. 25,29,31–34In one of the largest studies with unclear study design, Cutler et al.  23investigated 130 patients and found the highest rate of cardiac complications in patients who had an ischemic response on their exercise electrocardiography at less than 75% of their predicted maximal heart rate.

To date, only three large prospective studies have been published on the role of preoperative exercise stress electrocardiography. 8,10,11Carliner et al.  8investigated 198 patients, von Knorring and Lepäntalo 10investigated 105 patients, and McPhail et al.  11investigated 100 patients. In two of these studies, several different types of exercise stress testing were used, 8,11and in one study, only a single modified V⁴-electrocardiography lead was recorded. 10None of these studies 8,10,11involved the evaluation of the exercise stress electrocardiography by two independent investigators blinded to patient identity and clinical course, 35and perioperative Holter monitoring and newer biochemical markers were not used for the detection of cardiac complications. 8,10,11Two of these prospective studies demonstrated that preoperative stress-induced ischemia is a significant predictor in univariate analysis 8,10; in the study by Knorring and Lepäntalo’s, 10only univariate analysis was performed. McPhail et al.  11found that patients who had a positive stress test (ST depression ≥ 1 mm) and impaired exercise tolerance indicated by their inability to achieve less than 85% of their predicted maximal heart rate had a significantly higher rate of cardiac complications than patients with a positive stress test who achieved more than 85% of their predicted maximal heart rate. Multivariate analysis of the data from the study by Carliner et al.  8showed only preoperative resting electrocardiography as a significant predictor of risk. In contrast, the study by McPhail et al.  11showed that the ability to reach a high maximal heart rate and a high cardiac work load maximal oxygen uptake was an independent predictor of a low rate of cardiac complications. 11Gerson et al.  27,28reported similar results in two retrospective studies. By multivariate analysis, inability to perform 2 min of supine bicycle exercise increasing the heart rate to greater than 99 beats/min was the best predictor of perioperative cardiac complications. 27,28Contrary to these previous studies, 11,27,28our study did not show that inability to reach a certain heart rate is predictive. This could be a result of the large proportion of vascular surgery patients with limited exercise capacity in our study, and the fact that 30% of our patients were receiving treatment with β blockers. However, consistent with the results of Cutler et al.  23and McPhail et al. , 11we also found that ST-segment depression of 0.1 mV or more during exercise before achieving a specified percentage of the predicted maximal heart rate was predictive.

The second important result of our study is the definition of the role of exercise stress electrocardiography: in contrast to the preoperative resting electrocardiography, the exercise electrocardiography was found to be an independent predictor and represented a significant enhancement for risk stratification. The independent predictive value of resting electrocardiography was demonstrated by Carliner et al.  8and confirmed in recent studies. 9,36Our study is the first to demonstrate the superiority of preoperative exercise stress electrocardiography over resting electrocardiography.

Palda and Detsky 7advised against the use of exercise stress electrocardiography for preoperative risk stratification and instead recommended myocardial imaging or stress echocardiography; therefore, a direct comparison of these methods would be interesting. To our knowledge, there were only two smaller studies, in which dipyridamole-thallium scintigraphy was compared with exercise electrocardiography. 34,37In the study by McPhail et al. , 37sensitivity of scintigraphy was significantly higher than that of exercise electrocardiography; whereas specificity was lower. Regarding the predictive value, there was no difference between both tests. 37In the regression analysis of the study by Leppo et al. , 34only the presence of thallium redistribution was significant. However, the predictive value of thallium imaging has been questioned in two newer prospective studies. 38,39The positive predictive value of dipyridamole-thallium scintigraphy ranges from 4% to 20%5,6,40and that of stress echocardiography from 7% to 38%, 5,6,40–42so that the two methods are comparable. 40The predictive value of exercise stress electrocardiography in our study was 20%, and published values range from 5% to 25%5,8,11; however, it should be noted that the predictive value is dependent not only on the incidence of cardiac complications, which is influenced by the special risk of the investigated study population, but also on the definitions of adverse outcome. In contrast to other studies, we have included myocardial cell injury as an outcome parameter in addition to myocardial infarction and death. Thus, comparisons of the predictive values from different studies must be interpreted with care. Myocardial perfusion imaging and stress echocardiography have the disadvantage of limited availability, in addition to being expensive and time-consuming. Stress echocardiography also requires very experienced medical staff. In contrast, exercise stress electrocardiography is simple, widely available, and has become a well-standardized technique. Especially if a clinical parameter (definite coronary artery disease) and the operative risk are combined with the results of the exercise stress electrocardiography, preoperative risk stratification is possible and seems comparable to scintigraphy-based risk stratification 24with the advantage of a more cost-effective and easy accessible approach. 3It is unclear if patients who cannot perform bicycle ergometry or treadmill stress testing can be evaluated with arm ergometry 13,29,33as an alternative to scintigraphy or stress echocardiography. Exercise electrocardiography is not useful for patients with left bundle branch block or paced ventricular rhythm. 43Consistent with previously published data, definite coronary artery disease 39,44–46and major surgery 16,44,46proved to be independent predictors.

Although, because of the clinical markers and the type of operation performed, all 185 patients had at least an intermediate risk, 5approximately 70% of our patients were classified as relatively healthy given their classical risk classifications. 16,17The main reason for this discrepancy is that in contrast to the American College of Cardiology–American Heart Association guidelines, 5the risk indices 16,17do not contain mild angina pectoris, diabetes, compensated congestive heart failure, peripheral vascular surgery, and carotid endarterectomy as risk factors. Nonetheless, our population was not a particularly sick one, and this might be one reason why we found risk indices not to be independent predictors.

First, the major limitation of our study is the relatively small sample size and thus the low number of severe cardiac complications. In anticipation of this, we had defined the occurrence of troponin-T elevation more than 0.1 ng/ml on at least 2 successive days without electrocardiographic abnormalities or symptoms as a cardiac event, even though this is a relatively “soft” outcome parameter compared with myocardial infarction or cardiac death. However, the definition of minor postoperative myocardial injury as a perioperative event is supported by data from the last 10 yr, which show that postoperative myocardial ischemia, 47–49especially if prolonged, 14,21is significantly associated with perioperative complications and can influence late outcome. 49Abnormal postoperative troponin-T levels are also a predictor of 6-month prognosis. 50In addition, reduction of perioperative myocardial ischemia by medical treatment can influence early and late mortality. 49,51Second, in the period between the sixth postoperative day and the telephone interview 30 days after surgery, minor myocardial injury and even myocardial infarction could have been missed. Third, the physicians caring for the patients were not blinded to the preoperative resting and exercise electrocardiography. This could have influenced anesthetic management and postoperative care. However, in patients with an obvious abnormal stress test, this fact should have rather reduced than increased adverse outcome. Fourth, our finding that preoperative reduced left ventricular performance is an independent predictor of perioperative cardiac complications should be interpreted cautiously as echocardiography was performed by one unblinded cardiologist only. In contrast, studies especially evaluating this issue have shown that preoperative assessment of left ventricular performance by echocardiography or radionuclide angiography have only limited prognostic value. 39,52 

In conclusion, this prospective study shows that a ST-segment depression of 0.1 mV or more in the exercise stress electrocardiogram is an independent predictor of perioperative cardiac complications in intermediate-risk patients undergoing noncardiac surgery, whereas an abnormal resting electrocardiogram did not prove to be an independent predictor. A combination of a clinical parameter (definite coronary artery disease) with the type of surgery proposed and the results of the exercise stress electrocardiography allowed sufficient risk stratification.

The authors thank the Department for Biometry and Documentation, University of Ulm (Prof. Dr. Gaus and Dr. R. Muche), for advice and assistance in the statistical analysis of the data.