Author + information
- Received October 17, 2002
- Revision received April 7, 2003
- Accepted April 24, 2003
- Published online September 3, 2003.
- ↵*Reprint requests and correspondence:
Dr. Anthony P. Morise, Section of Cardiology, HSC - South, WVU, Morgantown, West Virginia 26506, USA.
Objectives To determine how well recently developed multivariables scores assess for all-cause mortality in patients with suspected coronary disease presenting for exercise electrocardiography (ExECG).
Background Recently revised American College of Cardiology/American Heart Association guidelines for ExECG have suggested that ExECG scores be used to assist in management decisions in patients with suspected coronary artery disease. Recently developed scores accurately stratify patients according to angiographic disease severity.
Methods To determine how well these scores assess for all-cause mortality, we utilized 4,640 patients without known coronary disease who underwent ExECG to evaluate symptoms of suspected coronary disease between 1995 and 2001. Previously validated pretest and exercise test scores as well as the Duke treadmill score were applied to each patient. All-cause mortality was our end point.
Results Overall mortality was 3.0% with 2.8 ± 1.6 years of follow-up. All three scores stratified patients into low-, intermediate-, and high-risk groups (p < 0.00001). No differences were seen when patients were evaluated as subgroups according to gender, diabetes, beta-blockers, or inpatient status. Low-risk patients defined by the Duke treadmill score had consistently higher mortality and absolute number of deaths compared with low-risk patients using other scores. In addition, the Duke treadmill score had less incremental stratifying value than the new exercise score.
Conclusions Simple pretest and exercise scores risk-stratified patients with suspected coronary disease in accordance with published guidelines and better than the Duke treadmill score. These results extend to diabetics, inpatients, women, and patients on beta-blockers.
The simple exercise electrocardiogram (ExECG) is alive and well, but living and existing in the shadow of exercise imaging. Recent consensus guidelines for chronic stable angina recommend the simple treadmill exercise ECG in a variety of diagnostic and prognostic situations (1). American College of Cardiology/American Heart Association (ACC/AHA) guidelines for exercise testing suggested that pretest and exercise scores be utilized in the interpretation of exercise tests and in clinical decision-making (2). Recently developed and validated pretest and exercise scores have demonstrated accuracy in stratifying men and women according to the likelihood of the presence of any and severe angiographic coronary disease (3–5). Given that these scores were developed in angiographic populations, the purpose of the present study was to validate them in a clinically relevant unselected population presenting with suspected coronary disease using the end point of all-cause mortality rather than angiographic disease. Secondarily, we wished to validate these scores in clinically relevant subpopulations defined by gender and diabetic, inpatient, and beta-blocker status.
Between May 1995 and February 2001, we screened all patients ≥18 years of age referred by primary-care physicians and cardiologists to the stress laboratory for their first exercise test. First exercise tests could include ExECG, nuclear, or echocardiographic studies. We included only symptomatic patients referred with the express purpose of evaluating for the presence of coronary disease. Included in this population were inpatients admitted for chest pain. These inpatients were observed for at least 24 h and had infarction excluded by assessment of serum markers. Those physicians caring for them as well as those in the exercise laboratory felt they were appropriate for exercise testing. We excluded asymptomatic patients, those receiving digitalis preparations, those with a history of coronary artery disease (prior myocardial infarction or coronary angiography), and those with resting electrocardiograms (ECGs) that were considered uninterpretable (left ventricular hypertrophy, left bundle branch block, Wolff-Parkinson-White Syndrome, or other significant downward displacement of the ST-segment) (2).
Baseline clinical information
We collected the following data from patients during a pre-ExECG interview: age, symptoms, medication usage at the time of the ExECG, and other coronary risk factors. Patients had height and weight recorded. We classified chest pain using the three categories of Diamond (6): typical angina, atypical angina, and non-anginal chest pain. Risk factors included the following: current or prior cigarette smoking, history of hypertension (on antihypertensive therapy), history of insulin- or noninsulin-requiring diabetes, history of high cholesterol or on cholesterol-lowering therapy, a family history of premature (<60 years of age) coronary disease (infarction, coronary bypass or angioplasty, sudden death) in first-degree relatives, and obesity defined as a body-mass index (kg/m2) >27. We determined estrogen status using previously published criteria (7,8). Women were estrogen status negative if they were postmenopausal and not receiving estrogen replacement therapy. If they were premenopausal or receiving estrogen replacement therapy, they were considered as estrogen status positive. Women who underwent hysterectomy without oopherectomy were considered estrogen status positive if they were under the age of 50 and without symptoms of estrogen deficiency. Otherwise, they were considered estrogen status negative.
All patients exercised using the Cornell treadmill protocol. We did not use predetermined peak heart rates to determine when to stop exercise. We read all studies in a blinded fashion.
Using the 12 standard leads, we measured peak exercise or 3-min recovery ST-segment changes (60 ms following the J-point compared to the baseline between 2 PR segments). We qualitatively categorized peak exercise ST slope as upsloping, horizontal, or downsloping. Positive ST-segment criteria consisted of ≥1 mm horizontal/downsloping ST depression. For purposes of the Duke treadmill score calculation, we considered any mm (≥1 mm) of exercise-induced ST-segment depression (over baseline measurements) that was associated with horizontal or downsloping ST segments. Exercise-induced ST-segment elevation was not considered in this analysis.
We also recorded resting and peak exercise heart rate and blood pressure, exercise capacity estimated in metabolic equivalents (METs), and exercise-induced angina. Because the Duke treadmill score employs exercise duration using the Bruce protocol, METs were converted to minutes of exercise duration using the Bruce protocol by use of the speed and elevation of the treadmill.
Angina during testing was classified according to the Duke Treadmill Angina Index (2 if angina required stopping the test, 1 if angina occurred during or after treadmill test, and 0 for no angina) (9).
Score and end point determination
Utilizing the scores presented in Figures 1A to 1C,
each patient had the following scores determined with respective assignment to low-, intermediate-, and high-risk subgroups for each score:
Pretest score: low 0 to 8 points, intermediate 9 to 15 points, high >15 points (3).
Exercise test scores for men and women: low 0 to 39 points, intermediate 40 to 60 points, high >60 points (4,5). For the purposes of this study, these two scores will be considered as one score and referred to as the new exercise score.
Duke treadmill score: low ≥+5 points, intermediate −10 to +4 points, high ≤−11 points (9).
The Duke treadmill score was calculated using the following equation: Patients had vital status and date of death determined by a search of the Social Security Death Index. Approval for collection of follow-up data was obtained from our institutional Human Subjects Committee.
The NCSS 2001 software (Number Cruncher Statistical System) was used for all statistical analyses. Comparison of frequencies was accomplished using chi-square testing. Comparison of means was accomplished using nonpaired ttesting. Normality of data distribution was determined using the Wilk-Shapiro test as well as observation of box and normal probability plots. When data were not distributed normally, the Mann-Whitney Utest was utilized. Survival analysis was accomplished using Kaplan-Meier curves and Cox proportional hazards analysis. A p value <0.05 was considered statistically significant.
Incremental stratifying value of the exercise scores over the pretest score was assessed as follows. To demonstrate incremental value, two things should occur: 1) the number of correct classifications (low-probability patients who are alive plus high-probability patients who are dead) should significantly increase, and 2) the number of unclassified patients (those with an intermediate pretest probability) should significantly decrease.
During the period of interest, 4,640 symptomatic patients with suspected coronary disease and interpretable ECGs underwent exercise testing with or without simultaneous imaging (echocardiographic 992, or 21%; and nuclear 2,206, or 48%). See Table 1for summary of clinical and exercise test characteristics of the entire population as well as selected subgroups defined by their clinical relevance, such as inpatient/outpatient status at the time of the exercise test, diabetic status (as defined earlier), and beta-blocker status at the time of the exercise test. Prior reports have presented the characteristics of the men and women (4,5).
Table 2summarizes the pretest, new exercise test, and Duke treadmill score results. The outpatient, diabetic, and beta-blockers subgroups had consistently higher pretest and exercise test scores than their respective alternate subgroups. The distribution of low-, intermediate-, and high-risk patients differed for all subgroup comparisons and scores, except the inpatient/outpatient comparison using the Duke treadmill score.
The mean follow-up for the group was 2.8 ± 1.6 years (alive: 2.8 ± 1.6 years, and dead: 2.5 ± 1.7 years). Table 3summarizes the all-cause mortality data for all subgroups and for risk groups for each of the three scores. Not shown are the results for men and women, which mirrored nearly exactly the results seen by the entire group. The men had slightly but significantly higher overall mortality than the women (3.7% vs. 2.3%; p < 0.01). For the pretest and both exercise scores, there was significant stratification within all subgroups. The mortality of the low-risk Duke treadmill score groups varied from 1.8% to 3.7%, whereas the mortality of the low risk exercise score groups was consistently below 2.0%.
Figure 2displays the Kaplan-Meier curves for each of the three scores comparing the low-risk group to a combined intermediate- and high-risk group. These latter groups were combined because of the relatively small number of high-risk patients. For each score there was significant stratification (p < 0.000001); however, the clearest separation of the low risk group from the other groups occurred with the new exercise score. Figures 3 and 4⇓are similar to Figure 2and display ⇓Kaplan-Meier curves for selected subgroups using only the new exercise score. In each case, a low-risk group is clearly identified.
Cox proportional hazards analysis was performed with scores as the independent variables. Univariate analysis demonstrated that the pretest score (p = 0.00001) and exercise score (p = 0.01) were significant predictors of the time to death. However, the Duke score was not (p = 0.10). When both the pretest and exercise scores were included in the model, both scores remained significant (pretest score p = 0.0003, exercise score p = 0.043). When METs was included in the model, these results did not change (pretest score p = 0.0003, exercise score p = 0.02, METs p = 0.18).
The ACC/AHA guidelines suggest that the value of exercise testing differs according to pretest probability. They also assigned a Class I indication to the intermediate pretest probability group and only a Class IIb indication to the low and high pretest probability groups. When an analysis like that performed in Table 3was limited to the intermediate pretest probability subgroup, the patterns of stratification were similar to what was seen on Table 3.
Table 4presents mortality as a function of both the pretest and the new and Duke exercise scores. Whereas there was significant stratification using this approach especially for the intermediate pretest probability group, this display does not completely reflect the incremental value of the respective exercise scores over the pretest score.
Table 5presents the results of our incremental value analysis. The new exercise score demonstrated significant incremental stratifying value over the pretest score given the substantial increase in correctly classified patients and decrease in unclassified patients, as well as a decrease in incorrectly classified patients. The Duke treadmill score demonstrated a significantly smaller percent increase in correctly classified patients and smaller percent decrease in unclassified or intermediate-risk patients. The remaining columns all assessed the incremental value of the new exercise score over the pretest score. Except for two, all subgroups demonstrated significant and substantial incremental value, that is, an increase in correctly classified patients and a decrease in unclassified patients. Compared to the results in men, the use of the new exercise score in women was associated with more modest increase in incremental stratifying. In addition, the use of the new exercise score in diabetics was associated with less incremental value than in nondiabetics. Diabetics had a substantial increase in correctly classified patients, but no change in the unclassified patients.
It is our observation that, in many institutions, ExECG has been replaced by noninvasive imaging because of the perception that it is of less value in clinical decision-making. This is despite the presence of consensus guidelines that clearly advocate its use as the initial evaluation in many clinical situations, including suspected coronary disease. Its jaded reputation in women, patients on beta-blockers, and those with either poor exercise capacity or chronotropic incompetence are familiar to all who utilize exercise testing in any form. The guidelines for ExECG place no such restrictions on its use in these populations. In fact, there are recent studies suggesting that it is valuable in those populations (5,10–13).
The present study suggests that the utility of ExECG can be enhanced by the use of multivariable scores. In addition, its use and value in risk assessment can be extended to diabetics, women, inpatients, and those on beta-blockers and the designation of low risk extends out to at least three years across all subgroups. The ACC/AHA guidelines for exercise testing (2)assigned a Class I indication to patients with an intermediate pretest probability and a Class IIb indication to those with a low or high pretest probability concerning the diagnosis of coronary disease. From a prognostic perspective, our results as reflected in Table 4would suggest that these assignments are appropriate.
For low pretest probability patients, exercise testing will most often confirm what is already appreciated from the clinical assessment and will sometimes raise a question for further testing (imaging). In other words, if a stress test is considered for low probability patients, ExECG is the appropriate first choice given its strong negative predictive value (11).
For high pretest probability patients, only a minority will be reclassified as low risk, so in the majority, the exercise test will confirm what is already appreciated from the clinical evaluation. A more appropriate initial strategy might be to have these patients undergo either angiography or imaging (14,15).
For intermediate pretest probability patients, our results suggest that a low-risk ExECG result is associated with a low risk of death. Even though these patients are at low risk of death, they are not necessarily at low risk of having significant coronary disease as a cause of their symptoms (11). Therefore, in selected intermediate pretest probability patients, an initial imaging strategy or follow-up imaging will be needed to clarify what might be causing their symptoms. Clearly more study is needed to determine the incremental value of such strategies and how they might be best applied.
The Duke treadmill score was derived using cardiovascular death as its end point. On the other hand, both the pretest score and the new exercise scores were derived using the presence or absence of angiographic coronary disease as their end point. The present study utilized all-cause mortality rather than cardiovascular death or angiography. This would seem to put all of these scores at some degree of disadvantage. This may explain some of the differences seen between the two exercise scores. This study also did not consider other relevant end points such as cardiovascular death, nonfatal myocardial infarction, stroke, and coronary revascularization. We chose all-cause mortality because of its ease of determination and lack of bias (16). We did not consider the results of noninvasive imaging that were available in many of our patients. Our mortality data were not censored for revascularization because these data were not collected in this study.
Exercise capacity as expressed by METs was not included in the new exercise score because, at its derivation, it was not selected as a predictor of angiographic coronary disease. Despite this, the new exercise score stratified our population well and was at least as good and perhaps slightly better than the Duke treadmill score, which does include exercise capacity. Given that the new exercise score does not include exercise capacity, how did it perform as well as a score that did include METs? Two explanations are possible. First, it is possible that other pretest clinical data not included in the Duke treadmill score made up the difference. However, a more likely possibility is that exercise heart rate, a known predictor of death, has sufficient predictive power to negate the absence of exercise capacity (17).
We divided our population according to peak METs achieved: <10 and ≥10 METs. As expected, the group with lower peak METs was larger (3,018 vs. 1,622) and had a higher mortality (3.8% vs. 1.5%; p < 0.01). However, the pretest and new exercise scores stratified both of these groups well and in a manner mimicking what was seen in Table 3. In addition, the incremental value of the new exercise test was substantial in each (% increase in corrects: low 57 vs. high 80, and % decrease in unclassifieds: low 25 vs. high 71), with the advantage going to the higher exercise capacity group. Cox analysis also indicated that the predictive power of the exercise score was not negated when METs were included in the analysis. Therefore, although the new exercise score does not contain exercise capacity as a predictor variable, it stratifies at least as well as a score that does contain exercise capacity, and performed well in groups with differing exercise capacity.
The new exercise score, although simple by design, does not consider many variables. For that reason, it might be less precise in defining risk than a score that considered many more variables. A preliminary report (18)of a more complicated exercise score suggests that such a score is accurate, but comparison to the present new exercise score has not been undertaken to date.
In conclusion, simple pretest and exercise scores risk stratify symptomatic patients with suspected coronary disease well in accordance with published ACC/AHA guidelines. A new exercise score risk stratifies patients at least as well if not better than the Duke treadmill score. These results extend to diabetics, inpatients, women, and patients on beta-blockers.
- American College of Cardiology/American Heart Association
- exercise electrocardiogram
- metabolic equivalents
- Received October 17, 2002.
- Revision received April 7, 2003.
- Accepted April 24, 2003.
- American College of Cardiology Foundation
- Gibbons R.J,
- Chatterjee K,
- Daley J,
- et al.
- Gibbons R.J,
- Balady G.J,
- Beasley J.W,
- et al.
- Diamond G.A
- Morise A.P,
- Haddad W.J
- Patterson R.E,
- Eisner R.L,
- Horowitz S.F
- Hachamovitch R,
- Berman D.S,
- Kiat H,
- et al.
- Lauer M.S,
- Blackstone E.H,
- Young J.B,
- Topol E.J
- Lauer M.S,
- Okin P.M,
- Larson M.G,
- Evans J.C,
- Levy D
- Morise A.P,
- Blackstone E.H,
- Lauer M.S