Author + information
- Received May 13, 1998
- Revision received July 7, 1998
- Accepted July 24, 1998
- Published online November 15, 1998.
- Karen P Alexander, MD∗,* (, )
- Leslee J Shaw, PhDa,
- Elizabeth R DeLong, PhD†,
- Daniel B Mark, MD, MPH, FACC∗ and
- Eric D Peterson, MD, MPH∗
- ↵*Address for correspondence: Dr. Karen P. Alexander, Box 3411, Duke University Medical Center, Durham, North Carolina 27710
Presented at the 69th Annual Scientific Session of the American Heart Association, New Orleans, Louisiana, November 9–13, 1996.
Objectives. We sought to determine the ability of a treadmill score to provide accurate diagnostic and prognostic risk estimates in women.
Background. Treadmill testing has been reported to have a lower accuracy for diagnosis of chest pain in women. The diagnostic and prognostic value of the Duke Treadmill Score (DTS) in women is unknown.
Methods. We determined the diagnostic and prognostic value of the DTS in 976 women and 2,249 men who underwent both treadmill testing and cardiac catheterization in a single institution from 1984 to 1994.
Results. Women and men differed significantly in DTS (1.6 vs. −0.3, p < 0.0001), disease prevalence (32% vs. 72% significant coronary artery disease [CAD], p < 0.001), and 2-year mortality (1.9% vs. 4.9%, p < 0.0001). The DTS provided information beyond clinical predictors of both coronary disease and survival in women and men. Although overall women had better survival, the DTS performed equally well in stratifying both genders into prognostic categories. The DTS actually performed better in women than in men for excluding disease, with fewer low risk women having any significant coronary disease (≥1 vessel with ≥75% stenosis) (20% vs. 47%, p < 0.001), or severe disease (3-vessel disease or ≥75% left main stenosis) (3.5% vs. 11.4%, p < 0.001).
Conclusions. By combining several aspects of treadmill testing, the DTS effectively stratifies women into diagnostic and prognostic risk categories.
Women presenting with chest pain differ from men in aspects of their clinical presentation, performance on diagnostic tests and in the prevalence of coronary artery disease. These differences alter pretest likelihood, referral patterns and the diagnostic ability of tests. The fact that most of the available data on the noninvasive diagnosis of chest pain is based on studies in men provides as additional challenge in the evaluation of women (1,2). The standard exercise electrocardiogram (ECG) is the most commonly used and least costly of noninvasive tests for the assessment of ischemic heart disease. This test, however, appears to be less accurate in women for the diagnosis of coronary disease due in part to an increase rate of false positives (3–9). For example, recent meta-analyses have found a statistically significant lower specificity of ST segment depression on treadmill tests in women compared with men (10,11). As a consequence, many clinicians believe that a stress-imaging study should be the test of first choice for the diagnostic evaluation of chest pain in women.
Although the interpretation of the exercise test as “positive” or “negative” has traditionally been based on the presence of ST segment depression, the exercise test provides a variety of other diagnostic and prognostic indicators, including exercise capacity and symptoms, which are useful in test interpretation (12). By interpreting all the information as a composite score, the diagnostic ability of treadmill testing in women may increase. The Duke Treadmill Score (DTS), a weighted index combining ST segment deviation, treadmill time and exercise-induced angina, was developed and validated as a risk-prediction instrument in a predominately male population (13,14). Recently, it has also been shown in a predominately male population to stratify risk of significant and severe coronary disease (15). Because of its demonstrated prognostic ability, the DTS has been included in the recommended screening algorithms of the Agency for Health Care Policy and Research (AHCPR) Unstable Angina Guidelines and the newly revised ACC/AHA (American College of Cardiology/American Heart Association) Guidelines for Exercise Testing (9,16). To date, however, the diagnostic and prognostic accuracy of the DTS in women has not been fully evaluated. The purpose of the present study was to assess the diagnostic and prognostic performance of the DTS in almost 1,000 women undergoing treadmill testing and cardiac catheterization.
We identified 3,225 patients referred for evaluation of chest pain who underwent exercise treadmill testing and subsequent diagnostic cardiac catheterization at Duke University Medical Center from 1984 to 1994. Of these, 30% were women (n = 976). Inclusion criteria were cardiac catheterization performed within 90 days of exercise treadmill testing. Exclusion criteria were prior cardiac catheterization or revascularization procedure, significant valvular or congenital heart disease, acute MI (myocardial infarction) on presentation, or resting ST-T wave changes from conditions such as bundle branch block in the baseline 12-lead ECG that would interfere with the interpretation of the standard exercise stress test.
Clinical, catheterization and follow-up data
Baseline clinical, exercise stress test and catheterization results were collected prospectively and entered into a computerized data base. Follow-up data were obtained by mailed questionnaire or telephone interview at 6 months and 1 year postcatheterization, and yearly thereafter as previously described (13,14). All cardiac deaths were confirmed with clinical data or death certificate by reviewers unaware of the clinical, angiographic or exercise stress data. Two-year follow-up for survival was 97% complete.
Exercise treadmill testing
All patients underwent symptom-limited treadmill testing using the standard Bruce protocol (13,14). A 12-lead ECG was recorded before exercise, at the end of each exercise stage, at peak exercise and at 2-min intervals during recovery. Three standard ECG leads were continuously monitored during exercise. The test was discontinued for limiting symptoms (angina, dyspnea, fatigue), abnormalities of rhythm or blood pressure, or marked and progressive ST segment deviation (>0.2 mV in the presence of typical angina or in the first stage of exercise). The ECG criterion for a positive test was 1 mm or more of exercise-induced ST segment deviation at 0.06 after the J point, relative to the PR segment.
Duke Treadmill Score
The Duke Treadmill Score (DTS) was calculated by inserting a patient’s test results into the following formula: Exercise time is measured in minutes, ST deviation is the largest net deviation, either depression or elevation in any lead except (aVR), and treadmill angina is graded on the following scale: 0 = no angina during exercise, 1 = nonlimiting angina during exercise, and 2 = exercise-limiting angina. The distinction between exercise-induced angina and nonanginal chest pain is based on the supervising clinician’s judgment with particular emphasis on reproduction of the patients presenting symptoms and on the classic features of typical angina. The typically observed range for the DTS is −25 (highest risk) to +15 (lowest risk). In this study, we used a previously proposed subgrouping system for the DTS: low risk (DTS score ≥5), moderate risk (DTS score between 5 and −11) and high risk (DTS score ≤−11) (13).
For descriptive purposes, continuous variables were presented as medians (25th and 75th quartiles) and discrete variables as percentages. Statistical comparisons were made between men and women. The Wilcoxon rank-sum test was used to assess differences in continuous variables, and chi-square tests were used for discrete variables. Outcomes evaluated were significant coronary artery disease, severe coronary artery disease and survival. Significant coronary artery disease was defined as ≥75% diameter stenosis in one or more vessels, and severe disease, a subset of significant disease, was defined as ≥75% stenosis in three vessels or ≥75% left main stenosis.
The prognostic value of the DTS was examined in three ways. First, to examine the empirical risk stratification provided by the DTS, Kaplan-Meier survival curves were generated separately for women and men stratified by the three DTS risk categories (17). In this analysis, patients who subsequently underwent coronary revascularization were censored at the time of their procedure. Second, to test whether the DTS performed differently in women and men, we introduced an interaction term (gender by DTS) into the Cox model with DTS and gender and tested for its significance (18). Finally, we added independent clinical history and physical examination prognostic factors to the Cox model, and we tested for the predictive contribution of the DTS by comparing the model containing only clinical data to the model containing both clinical data and the DTS (Appendix) (19). The difference in the overall likelihood ratio chi-squares between the models was used to quantitate the incremental contribution of the DTS to the overall model of prognosis in women and men.
The diagnostic value of the DTS was examined in three ways. First, the probability of significant and severe coronary disease in women and men was plotted from logistic regression models including DTS and gender. The extent of coronary disease found at catheterization was correlated with DTS risk categories. Second, to compare diagnostic accuracy by gender, we used these logistic regression models, and we tested for the significance of an interaction term (gender by DTS) that would indicate differential effects of DTS on likelihood of significant or severe coronary disease due to gender. Finally, we added independent clinical history and physical exam diagnostic factors to test whether the DTS added diagnostic information to the baseline clinical data in men and women (Appendix) (20,21). We compared the models containing only clinical data to those containing both clinical data and the DTS. The difference in the overall likelihood ratio chi-squares between the models was then used to quantitate the incremental contribution of the DTS to the overall model for significant and severe disease in women and men.
Baseline characteristics and test results
The baseline characteristics of the women and men in the study cohort are shown in Table 1. Women were slightly older (51 vs. 50 years, p = 0.02), more likely to have hypertension (39% vs. 34%, p = 0.004) and were more likely to report atypical chest pain (53% vs 34% p < 0.001). Men were more likely to have had a prior MI (34% vs. 14%, p < 0.001) and to report typical angina (55% vs. 28%, p < 0.001).
The baseline exercise stress test variables and angiographic data are shown in Table 2. Women had shorter exercise duration (5.5 min vs 7.0 min, p < 0.001), a lower prevalence of ST segment depression (19% vs. 37%, p < 0.001) and less exercise-induced angina than men (42% vs. 54%, p < 0.001). The majority of women and men had moderate risk treadmill scores (63% and 54%), with most of the remaining women and men in the low risk group (33% and 34%), and only a few of either gender in the high risk group 4% and 12%). Overall, the distribution of scores for women was shifted up (toward lower risk) relative to that for men (median treadmill score 1.6 vs −0.3, p < 0.0001). At cardiac catheterization, women were much less likely to have significant coronary disease when compared with men (32% vs. 72%, p < 0.001).
Prognostic ability for gender
The 2-year mortality was 1.9% for the study women compared with 4.9% for the men. Mortality increased for higher-risk DTS groups in both genders (Fig. 1). Two-year mortality for women was 1.0%, 2.2% and 3.6%, respectively for low, moderate, and high risk groups. Two-year mortality for men was 1.7%, 5.8% and 16.6%, respectively for low, moderate, and high risk groups. Both gender and the DTS were independent predictors of survival, but the interaction term between gender and the DTS in this model was not significant (p = 0.093), suggesting a similar relative relationship between DTS and prognosis for both genders (Table 3).
When the DTS was added to a Cox model containing the independent clinical predictors of survival, it contained 7.5% of the total prognostic information available in women and 20.8% of that available in men (Table 4).
Diagnostic ability by gender
At every DTS risk level, women had less coronary artery disease than did their male counterparts. Over 80% of the women with low risk scores had no significant coronary artery disease at catheterization compared with 52.6% of the men with low risk scores (Table 5). Similarly, only 3.5% of low risk women had severe coronary disease, compared with 11.4% of low risk men. The majority of women with moderate risk scores frequently had no significant coronary disease (65.1%), but over one-third also had significant disease (34.9%), or severe coronary disease (12.4%). By contrast, only a few men with moderate risk scores had no coronary disease (17.8%), whereas the majority had significant disease (82.2%) or severe disease (38.7%). Patients of either gender with high risk scores frequently had severe coronary disease (46% of women and 71.5% of men). Graphs of the probability of significant and severe coronary artery disease plotted for DTS by gender reveal similar shapes to the curves, although women had less disease at every DTS value compared with men (Figs. 2 and 3). ⇓⇓Both DTS and gender contributed significantly independent information for the prediction of significant and severe coronary artery disease (Table 3).
An interaction term between the DTS and gender of borderline significance was found for the prediction of significant disease (p = 0.046), but not for severe disease (p = 0.211). When the components of the DTS were tested individually to discover the source of this gender interaction, an interaction term also of borderline significance was found between gender and the angina component of the treadmill score (p = 0.048). This suggests a differential relationship exists between angina and the prediction of significant coronary disease in women compared with men. Women have a similar frequency of angina on the treadmill as do men, but exercise angina in women was less often correlated with the presence of coronary artery disease. When the angina component of the DTS was removed, the score’s ability to diagnose significant disease in women was essentially unchanged (area under ROC [receiver operating characteristic] 0.694 vs. 0.697). In men, however, deletion of the angina component diminished the predictive value of the score for significant coronary disease (area under ROC 0.806 vs. 0.792).
When the DTS was added to the model with clinical predictors of coronary disease, it provided significant incremental information in both men and women. For significant disease, DTS provided 11.3% of the total information in women, and 13.9% of the total information in men. For severe disease, DTS added 18.5% of the total predictive information in women, and 26.9% in men (Table 4).
The evaluation of chest pain in women can be simple and cost-efficient. In the largest study of treadmill testing in women to date, we demonstrated equivalent diagnostic and prognostic ability to the Duke Treadmill Score (DTS) in women and men. As has been shown in previous populations, women in our study presenting for chest pain evaluation had less coronary artery disease and performed differently on the treadmill from men (5,7). Despite this, we found that the DTS, by combining several standard measures available from exercise treadmill testing, is able to accurately risk-stratify both genders for the presence of disease and survival and adds information beyond clinical factors in both women and men.
Comparison to prior studies
Because standard interpretation of exercise treadmill testing has lower specificity and positive predictive value in women, many investigators have attempted to improve the diagnostic accuracy of treadmill testing in women by creating new variables such as ST/HR (heart rate) slope, computer-generated algorithms and gender-specific guidelines for interpretation (4,22–24). Because these methods have not found widespread clinical acceptance, others have suggested that initial testing strategies in women exclude standard treadmill testing in favor of exercise or stress-imaging studies (25–28). In the current study, we found that the combination of several variables from treadmill testing into a single composite risk score provided equivalent risk stratification in men and women. There was no gender interaction between the DTS and the diagnosis of severe coronary disease and the prediction of survival. The borderline gender interaction between the DTS and the diagnosis of significant disease was largely explained by the weaker relationship between angina and the presence of coronary artery disease in women. Removing the angina component from the DTS did not significantly change its diagnostic ability in women because treadmill time and ST deviation substitute more of the diagnostic information in women. This emphasizes the advantage of interpreting results of treadmill testing as a composite score.
Clinical significance of treadmill scores
Our analysis showed that predictions for women should be interpreted within the context of lower pretest risk for both diagnostic and prognostic risk stratification. Because of differences in disease prevalence, women had better survival at all values of the DTS. Risk categories were essentially shifted down one level of severity in women. The three treadmill risk categories of low, moderate, and high risk in men corresponded to very low, low, and moderate risk in women. In essence, owing the lower prevalence of disease in women, a low risk DTS was actually better at excluding coronary artery disease in women than in men. Although many low risk men can be managed without additional invasive testing, this is true for both low and moderate risk women. Therefore, renewed confidence in the initial use and interpretation of treadmill testing in women should be encouraged, especially for the purpose of excluding coronary artery disease.
The population in our study consisted of inpatients who underwent cardiac catheterization as part of their initial cardiac evaluation. Although this is necessary to determine diagnostic accuracy of treadmill testing, it creates a potential bias toward patients with a higher likelihood of disease. The prognostic accuracy of the Duke Treadmill Score has been previously tested and validated in an outpatient population, but this latter population did not have enough women to examine this subgroup separately. Furthermore, though our population of 976 women was smaller than comparable studies in men, it still represents one of the largest assessments of treadmill accuracy in women. In addition, our findings only apply to patients who are candidates for exercise treadmill testing, who are physically able to walk on a treadmill and have no resting abnormalities on their ECG.
In the largest study to date of women undergoing treadmill testing, we demonstrated that the DTS can accurately stratify diagnostic and prognostic risk in women. Our results support the routine initial use of the exercise treadmill test in suitable candidates of both genders presenting with suspected coronary artery disease, as recommended by the new ACC/AHA Exercise Testing Guidelines.
The authors would like to acknowledge the excellent editorial and technical support of Ms. Tracey Simons in the preparation of the manuscript.
Clinical index models for the prediction of disease and survival
|1. Predictive Characteristics and Coefficients for Cox Proportional Hazards Model Predicting Survival (19),legend|
|Prognostic pain index (episodes of daily angina ∗ (6 ∗ unstable angina + 2 ∗ progressive angina + nocturnal angina + 3 ∗ presence of ST-T wave changes∗ [1–0.5 ∗ (ECG Q waves)]))||0.0364|
|Myocardial index (History of CHF + 2 ∗ class IV CHF + cardiomegaly + ECG PVCs + ventricular gallop + 2 ∗ [history of previous MI or Q waves])||0.4506|
|Vascular disease index (history of peripheral vascular disease + history of cerebrovascular disease + presence of carotid bruits)||0.5333|
|Conduction index (4 ∗ ECG LBBB + 1 ∗ ECG RBBB + 2 ∗ ECG LAD + 4 ∗ ECG IVCD)||0.8975|
|Sex (0 = male, 1 = female)||−0.6732|
|2. Predictive Characteristics and Coefficients in Logistic Multiple Regression Model for the Prediction of Significant Coronary Disease (20)|
|Sex (0 = male, 1 = female)||−0.328|
|Age ∗ Sex (interaction)||−0.0301|
|Typical angina (1 if present)||2.581|
|Atypical angina (1 if present)||0.976|
|History of MI (1 if present)||1.093|
|ECG Q waves (1 if present)||1.213|
|History of MI ∗ Q waves (interaction)||0.741|
|Smoking (1 if present)||2.596|
|Hyperlipidemia (1 if present)||1.845|
|Diabetes (1 if present)||0.694|
|ECG ST-T wave changes (1 if present)||0.637|
|Age ∗ Smoking (interaction)||−0.0404|
|Age ∗ Hyperlipidemia (interaction)||−0.0251|
|Sex ∗ Smoking (interaction)||0.550|
|3. Predictive Characteristics and Coefficients in Logistic Multiple Regression Model for the Prediction of Severe Coronary Disease (21)|
|Log10of duration of CAD + 1||0.3424|
|Type of pain (0 = nonanginal, 1 = atypical, 2 = typical)||0.3014|
|Log10of duration + 1) ∗ type of pain (interaction)||0.1559|
|ECG Q waves (1 if present)||0.3513|
|Pain index (typical angina ∗ episodes weekly angina [maximum, 35]) ∗ 1 ∗ progressive pain + 4 ST-T waves but no Q waves + 2 ∗ presence of nocturnal angina)||0.0054|
|Sex (0 = male, 1 = female)||−0.3823|
|Risk factor index (hyperlipidemia + diabetes + hypertension)||0.1734|
|Vascular disease index (history of peripheral vascular disease + history of cerebrovascular disease + presence of carotid bruits)||0.2402|
↵legend Appendix abbreviations: CHF = congestive heart failure; IVCD = intraventricular conduction defect; LAD = left axis deviation; LBBB = left bundle-branch block; MI = myocardial infarction; PVCs = premature ventricular contractions; RBBB = right bundle-branch block.
- American College of Cardiology
- American Heart Association
- Agency for Health Care Policy and Research
- Duke Treadmill Score
- receiver operating characteristic
- Received May 13, 1998.
- Revision received July 7, 1998.
- Accepted July 24, 1998.
- American College of Cardiology
- Okin P.M.,
- Kligfield P.
- Barolsky S.M.,
- Gilbert C.A.,
- Faruqui A.,
- Nutter D.O.,
- Schlant R.C.
- Detry J.R.,
- Kapita M.B.,
- Cosyns J.,
- Sottiaux B.,
- Brasseur L.A.,
- Rousseau M.F.
- Gibbons R.J.,
- Balady G.J.,
- Beasley J.W.,
- et al.
- Kwok Y.S.,
- Kim C.,
- Grady D.,
- Redberg R.F.
- Zhao D.,
- Freeman D.H.,
- deFilippi C.R.
- Fletcher G.F.,
- Froelicher V.F.,
- Hartley L.H.,
- Haskell W.L.,
- Pollock M.L.
- ↵Shaw LJ, Kesler K, Peterson ED, et al. The Duke Treadmill Score is an accurate predictor of the extent of coronary artery disease and 5-year survival. Circulation 1998; in press.
- Braunwald E, Mark DB, Jones RH, et al. Unstable angina: diagnosis and management: Clinical Practice Guideline Number 10. Rockville, MD: Agency for Health Care Policy and Research: 1994. Publication 94-0602.
- Cox D.R.
- Robert A.R.,
- Melin J.A.,
- Detry J.R.
- Hachamovitch R.,
- Berman D.S.,
- Kiat H.,
- et al.