Prognostic Value of a Treadmill Exercise Score in Symptomatic Patients With Nonspecific ST-T Abnormalities on Resting ECG FREE

Repolarization abnormalities—ST-segment depression and T-wave inversion or flattening—are the most common abnormalities on the resting electrocardiogram (ECG) in the general population and in patients who present with chest pain.^1- 4 Patients with ST-T abnormalities have a higher prevalence of coronary artery disease (CAD), severe CAD, left ventricular dysfunction, and higher cardiac morbidity and mortality than those with normal results on resting ECG.^1- 10 Although previous studies have shown that treadmill exercise testing is useful for the diagnosis of CAD in patients with ST-T abnormalities, the exercise ECG is considerably less specific then in patients with normal findings on resting ECGs.^11- 16 Thus, exercise testing for prognostic purposes in patients with resting ST-T abnormalities is problematic, since these patients are at higher risk yet have a less specific exercise ECG response for the diagnosis of CAD.

Previous studies have identified the most important prognostic variables from exercise testing and have derived scores for risk stratification.^17- 26 The most thoroughly validated of these is the Duke treadmill exercise score, which summarizes prognostic information in a simple quantitative equation.^25- 26 National guidelines currently recommend the use of standard exercise treadmill testing with application of the Duke treadmill score for prognostication in patients with normal results on resting ECG and those with mild (<1-mm ST-segment depression) ST-T abnormalities.^27- 29 Stress imaging is recommended for patients with more severe (≥1-mm ST-segment depression) ST-T abnormalities. However, the prognostic capability of the Duke treadmill score in patients with normal findings on resting ECG compared with those with ST-T abnormalities has not been evaluated. Patients with baseline ST-T abnormalities were included in the study populations from which the Duke treadmill score was derived and validated, but these patients were not analyzed separately to confirm the prognostic value of the Duke treadmill score in this specific subset. The objective of this study was to test the prognostic value of the Duke treadmill exercise score in symptomatic patients with nonspecific ST-T abnormalities on a resting ECG compared with patients with normal ECG results.

All patients with symptoms of chest pain or dyspnea who had undergone treadmill thallium testing between January 1, 1989, and December 31, 1991, at the Mayo Clinic, Rochester, Minn, and who had ST-T abnormalities on a resting ECG were retrospectively identified using the nuclear cardiology laboratory database. Patients with valvular, congenital, or cardiomyopathic disease, previous cardiac surgery, or previous coronary angioplasty were excluded. The resting ECG was interpreted by the physician or nurse supervising the exercise test. Resting ST-T abnormalities were defined as the presence of ST-segment depression of any magnitude in at least 1 lead or T-wave abnormalities, or both. The number of patients who met these initial criteria was 1122. Patients were excluded if they had ST-T changes secondary to left-ventricular hypertrophy by vectorcardiographic criteria³⁰ (16 patients), digoxin therapy (87 patients), complete left- or right-bundle–branch block (18 patients), preexcitation syndrome (10 patients), permanent pacemaker (6 patients), or technically unsatisfactory results or missing exercise data (45 patients). This study was approved by the Mayo Clinic Institutional Review Board. Minnesota state law requires that patients must authorize use of their medical records for research studies. One patient who refused research authorization was excluded. As a result, the study population consisted of 939 symptomatic patients with nonspecific ST-T abnormalities on a resting ECG. A total of 1466 symptomatic patients with normal resting ECG results who had undergone exercise thallium testing during the same period and who met the same criteria were the control population. Clinical information on each patient was collected before exercise testing and entered into a computer database. Each patient's chest pain was coded as typical angina, atypical angina, or noncardiac pain according to the criteria set out by Diamond.³¹

Resting heart rate, blood pressure, and findings of 12-lead ECG were recorded before exercise. Treadmill exercise testing was performed in all subjects according to the Bruce or Naughton protocols to end points of severe fatigue, moderate angina, or 2-mm or greater ST-segment depression. During exercise 3 ECG leads were continuously monitored. A 12-lead ECG was recorded every minute, at peak exercise, and at 3 and 6 minutes of recovery. Blood pressure was recorded by cuff sphygmomanometry at the end of each exercise stage and at peak exercise. Magnitude and slope of maximum ST-segment deviation (depression or elevation), treadmill angina index,²⁵ and exercise duration were recorded. ST-segment deviation was interpreted visually by the physician or nurse supervising the exercise test. ST-segment depression was considered present if the ST-segment demonstrated horizontal or downsloping depression 0.08 seconds after the J point in leads with or without baseline ST depression. Maximum ST-segment depression was calculated by subtracting the ST-segment depression during or after exercise from the resting ST-segment level in the lead with the greatest change. Maximum ST-segment depression was coded in the database into several groups: less than 1.0 mm, 1.0 to 1.4 mm, 1.5 to 1.9 mm, 2.0 to 2.4 mm, and greater than 2.4 mm. ST-segment elevation of 1 mm or greater (in ECG leads without pathological Q waves and in the absence of ST depression of 1 mm or more in any other lead) was regarded as 1-mm ST-segment deviation. Exercise time (in minutes) was defined as the duration of exercise for patients following the Bruce protocol; for those following the Naughton protocol, a conversion factor was applied to equate exercise duration with the Bruce protocol.³² The Duke treadmill score^25- 26 was calculated for each patient as follows: exercise time − (5 × maximum ST deviation) − (4 × angina index).

Mailed questionnaires, telephone interviews, and reviews of patients' medical records or physician contacts were used to acquire information for subsequent events. Events were defined as death, nonfatal myocardial infarction, and late revascularization by percutaneous transluminal coronary angioplasty (PTCA) or coronary artery bypass grafting (CABG). Death was confirmed by reviewing the hospital chart, a death certificate, or a clinician's report. The cause of death of each patient was coded as cardiac or noncardiac by a reviewer blinded to the baseline information and the exercise results. Myocardial infarction was defined on the basis of chest pain, ECG changes, and elevated serum creatine kinase isoenzyme levels. Revascularization procedures were defined as early (<3 months after thallium imaging) or late (≥3 months after thallium imaging). Early revascularization procedures generally are performed based on the results of the exercise thallium test; conversely, decisions to perform late revascularization procedures usually are not significantly influenced by the results of the test but instead by worsening clinical status, such that late revascularizations represent a surrogate for disease progression. Patient follow-up was 94% complete at a median duration of 7.0 years.

Clinical and exercise variables of the study and control populations were compared using the χ² test for independence for categorical variables and the Wilcoxon rank sum test for the continuous variables.

Patients were stratified into prospectively defined risk groups by the treadmill score according to the Duke classification.²⁶ Patients with a treadmill score of 5 or higher were classified as low risk; those with a score lower than 5 but −10 or higher were designated as intermediate risk; and those with a treadmill score lower than −10 were classified as high risk. The Kaplan-Meier method was used to generate survival curves for 4 end points: overall mortality (no patient censored from analysis); cardiac death (censoring of patients with noncardiac death or PTCA/CABG at any time after the exercise test); combined cardiac death and nonfatal myocardial infarction (censoring of patients with noncardiac death or PTCA/CABG at any time after the exercise test); and combined cardiac death, nonfatal myocardial infarction, and late revascularization (censoring of patients with noncardiac death or early PTCA/CABG). Group comparisons of survival and the relationships of the Duke treadmill score as a continuous variable to the 4 end points were evaluated separately in study and control populations using Cox proportional hazards analysis. Differences among risk groups and between the study and control populations for the 4 end points were investigated using the log-rank test. Statistical significance was defined as P<.05.

The clinical characteristics and exercise results of the study and control populations are compared in Table 1. Patients with ST-T abnormalities were older, had a higher prevalence of hypertension and previous myocardial infarction, and more commonly had a history of typical angina and use of antianginal medications.

There were significant differences in exercise testing results between patients with ST-T abnormalities and those without. Patients in the study population achieved a lower exercise workload, lower exercise heart rate, and lower Duke treadmill score. In the study population, 426 patients (45%) were in the low-risk group; 464 (49%) in the intermediate-risk group; and 49 (5%) in the high-risk group. In the control population, 811 patients (55%) were classified as low risk, 621 (42%) as intermediate risk, and 34 (2%) as high risk (Table 2, P<.001 for control population vs study population). The clinical and exercise characteristics of the study population demonstrated that patients with ST-T abnormalities on a resting ECG were a "sicker" cohort.

During follow-up there were 150 deaths (58 cardiac deaths), 49 nonfatal myocardial infarctions, and 83 early and 144 late revascularization procedures in the study population. For the control population, there were 95 deaths (29 cardiac deaths), 64 nonfatal myocardial infarctions, and 111 early and 197 late revascularization procedures. Seven-year outcome was significantly worse (P<.001) in the study population than in the control population for all 4 outcome end points (Table 2).

In both the study and the control populations, there were significant associations between the Duke treadmill score (as a continuous variable) and outcome (in the study population, P<.001 for all 4 end points; in the control population, P≤.002 for all 4 end points). When patients with ST-T abnormalities were classified into risk groups according to the Duke treadmill score, there were significant overall differences among the risk groups for all outcome end points. Seven-year cardiac survival for the low-risk group was 97% (annual cardiac mortality rate, 0.4% per year); for the intermediate-risk group, 92% (annual cardiac mortality rate, 1.1%); and for the high-risk group, 76% (annual cardiac mortality rate, 3.4%) (P<.001). For the end point cardiac death or nonfatal myocardial infarction, 7-year event-free survival rate was 94%, 88%, and 69% for the low-, intermediate-, and high-risk groups, respectively (P<.001). Risk stratification for the other end points (combined cardiac death, nonfatal myocardial infarction, and late revascularization, and for overall mortality) was also well demonstrated (Figure 1).

The study and the control populations had different risk-group distributions as well as different event rates in the same risk group (Table 2). Compared with patients with normal results on resting ECG, fewer patients with ST-T abnormalities were low risk (45% of the study population vs 55% of the controls; P<.001). For the end point cardiac death, 7-year survival was lower (97% vs 99%; P=.008) in the low-risk group of the study population. The intermediate-risk group was a larger proportion (49%) of the study population than of the control population (42%; P=.007), and the 7-year survival was significantly lower than that of the control population (92% vs 97%; P<.001). More patients with ST-T abnormalities were classified as high risk (5% vs 2%; P=.001) and their 7-year survival was lower than that of the control population high-risk patients (76% vs 93%). This difference was not statistically significant (P=.36) due to small patient numbers. Similar findings were observed for the other 3 end points.

A post hoc analysis was performed to determine if the performance of the Duke treadmill score was dependent on the type of ST-T abnormality on the resting ECG. We could retrieve results of resting 12-lead ECGs from electronic storage for most (907/939 [97%]) of the study population. We categorized these as follows: T-wave abnormality without ST-segment depression was present in 599 patients (26 of whom had ≥1-mm ST-segment elevation); ST-segment depression less than 1 mm in 289 patients (12 of whom had ≥1-mm ST-segment elevation); and ST-segment depression of 1 mm or more in 19 patients (1 of whom had ≥1-mm ST-segment elevation). When we eliminated the 19 patients with ST-segment depression of 1 mm or greater and repeated the statistical analysis, the Duke treadmill score remained significantly associated (P<.001 as a continuous variable) with all 4 outcome end points. For the end point cardiac death, 7-year survival rates were 87%, 92%, and 97% of the high-, intermediate-, and low-risk groups, respectively (P=.005). The population was further analyzed for cardiac mortality on the basis of the resting ECG. For patients with T-wave abnormality alone (n=599), 7-year survival was 92% for high-risk, 91% for intermediate-risk, and 97% for low-risk groups (P=.048). For patients with ST-segment depression of less than 1 mm (n=289), there was a nonsignificant trend: 7-year survival was 80%, 95%, and 99% for the high-, intermediate-, and low-risk groups, respectively (P=.10).

The results of this study support national guidelines^27- 29 that recommend standard exercise treadmill testing with use of the Duke treadmill score for risk stratification in patients with normal findings on resting ECG or mild (<1-mm ST-segment depression) ST-T abnormalities. Of the 906 patients in the study population who had a resting ECG available for interpretation, 98% had mild ST-T abnormalities. The Duke treadmill score can be used to effectively risk-stratify symptomatic patients with nonspecific ST-T abnormalities on a resting ECG. There were significant differences in event-free survival rates for all end points when the treadmill score was treated as a continuous variable and the study population was divided into low-, intermediate-, and high-risk groups using previously published²⁶ cutoff values for the treadmill score. The low-risk group had an annual cardiac mortality rate of 0.4%, comparable to that of the low-risk cohort reported by Mark et al (0.25% per year).²⁶ The high-risk group had an annual cardiac mortality rate of 3.4%, clearly identifying them as high risk, although their cardiac mortality rate was slightly lower than that of the high-risk cohort reported by Mark et al (5.3% per year).²⁶ Our data demonstrate that the Duke treadmill score is capable of predicting not only cardiac death but also total cardiac events.

The Duke treadmill score was first derived from an inpatient cohort of angiography patients and then validated in an outpatient cohort of nonangiographic patients. Since the angiographic cohort was a sicker population, 9% of the inpatients were classified as high risk, compared with 4% of the outpatients.^25- 26 In the present study, 5% of patients with resting ST-T abnormalities were classified as high risk, compared with only 2% of the control population. As expected from earlier studies,^1- 10 event rates in patients with ST-T abnormalities were higher than in those with normal results on resting ECG. The 7-year cardiac survival rate was 94% in the study population compared with 98% in the control population. For the end point cardiac death and nonfatal myocardial infarction, 7-year survival was 90% in the study population vs 94% in the control population. These results agree with those of the Framingham Heart Study⁵ in which age-adjusted cardiac mortality and morbidity of patients with ST-T abnormalities were approximately twice that of patients with normal results on resting ECG. Other epidemiologic studies have shown similar findings.^1- 3,7- 8

In the original inpatient study by Mark et al,²⁵ the prognostic implications of the treadmill score were greatest in the subset of patients with 3-vessel disease. That study reported differences in survival rates in patients with the same treadmill score but different extent of CAD and suggested that exercise testing results should be interpreted in the context of other available information. In our study, patients with resting ST-T abnormalities had consistently higher event rates than patients with normal findings on resting ECG. Our results are consistent with previous studies showing that 95% of patients with normal resting ECG findings have normal left-ventricular function^33- 34 and an excellent prognosis.^35- 36 The Duke treadmill score allows risk stratification of both the study and control populations, despite their different overall event rates. The results of the present study extend the findings of previous studies^25- 26 and emphasize the importance of considering the resting ECG when using exercise testing for risk stratification.

Previous studies used different diagnostic criteria to define an ischemic exercise ECG response in patients with resting ST-T abnormalities based on additional ST depression ranging from 0.5 to 2 mm.^11- 16 Because each of these studies consisted of a heterogenous group of patients (some included patients with conduction abnormalities or left-ventricular hypertrophy) with considerable referral bias (all patients had undergone coronary angiography), there were considerable discrepancies in the study results. Patients with secondary ST-T abnormalities are different from those with nonspecific ST-T abnormalities, and Meyers et al³⁷ demonstrated that ST-segment analysis with exercise testing is not reliable in such patients. We therefore included only patients with nonspecific ST-T abnormalities in the present study.

This study has several limitations. First, the patients being studied were referred for exercise thallium imaging instead of treadmill exercise testing, and so the results of this study need to be confirmed in patient populations referred for standard exercise testing before they can be generalized to the entire population of patients with ST-T abnormalities. We cannot determine the number of patients who met the same inclusion and exclusion criteria who underwent standard exercise treadmill testing at our institution during this time frame. It is important to emphasize that all patients in this study were symptomatic and not undergoing evaluation solely for abnormal results on resting ECG. Baseline ST-T abnormalities probably influence a clinician's decision to refer a patient for radionuclide imaging instead of standard exercise testing, and this selection process could have been 1 of the factors responsible for the different outcomes between the study and the control populations. Second, 2-mm ST depression is generally considered an end point to terminate exercise in our laboratory. Since some patients may have developed greater ST depression if allowed to exercise longer (the longer duration would partially offset the greater ST depression when calculating the Duke treadmill score), this practice may have reduced the number of patients in the high-risk groups. Finally, this study was not a true "natural history" study. Patients were treated with medications and underwent revascularization procedures during the follow-up period, the net impact of which probably lowered cardiac event rates. Despite relatively low cardiac event rates in both the control and study populations, patients with low- or intermediate-risk Duke treadmill scores had significantly worse outcomes in the study population compared with the control population.

Nonspecific resting ST-T abnormalities are common in patients undergoing evaluation of suspected CAD. However, there has been concern that ST-segment analysis with exercise testing may not be reliable in patients with resting ST-T abnormalities. Additionally, there is no consensus on the ECG diagnostic criteria for ischemia in these patients. This study clearly demonstrates the value of an exercise ECG for risk-stratifying these patients. Almost one half of the study population could be classified as low-risk with an annual cardiac mortality rate of only 0.4%. Since these patients would be highly unlikely to benefit from revascularization with respect to survival, cardiac catheterization could be safely deferred until it was justified on the basis of symptoms. In contrast, the small group of high-risk patients (annual cardiac mortality, 3.4%) merit early cardiac catheterization to define their potential for revascularization. The value of a test to effectively risk stratify a population depends on the percentage of the population that can be categorized as either low risk or high risk, since management of patients classified as intermediate risk remains problematic. Although the percentage of patients classified as intermediate risk was significantly greater (P<.001) in the study vs the control population, the difference between the study population and the control population was modest (49% vs 42%). The thallium imaging data that were collected on this study population will be analyzed in a future study to determine if we can further refine risk stratification of these patients. The purpose of this study was specifically to evaluate the value of the Duke treadmill score in patients with abnormal results of resting ECG. National guidelines recommend standard treadmill testing as the initial noninvasive method of evaluating patients with either normal results of resting ECG or mild (<1-mm ST-segment depression) ST-T abnormalities.^27- 29 The results of this study document that the standard exercise treadmill test with use of the Duke treadmill score is an effective initial method for prognostication in these patients.