Author + information
- Received March 4, 2003
- Revision received July 23, 2003
- Accepted September 9, 2003
- Published online January 21, 2004.
- Indu G. Poornima, MD*,
- Todd D. Miller, MD, FACC*,
- Timothy F. Christian, MD, FACC*,
- David O. Hodge, MS†,
- Kent R. Bailey, PhD† and
- Raymond J. Gibbons, MD, FACC*,* ()
- ↵*Reprint requests and correspondence:
Dr. Raymond J. Gibbons, Nuclear Cardiology, Charlton 2, Mayo Clinic, 200 First Street SW, Rochester, Minnesota 55905, USA.
Objectives The purpose of this study was to determine whether a previously validated clinical score (CS) could identify patients with a low-risk Duke treadmill score who had a higher risk of adverse events and, therefore, in whom myocardial perfusion imaging would be valuable for risk stratification.
Background Current American College of Cardiology/American Heart Association guidelines recommend using a standard exercise test without imaging as the initial test in patients who have an interpretable electrocardiogram and are able to exercise.
Methods We studied 1,461 symptomatic patients with low-risk Duke treadmill scores (≥5) who underwent myocardial perfusion imaging. The CS was derived by assigning one point to each of the following variables: typical angina, history of myocardial infarction, diabetes, insulin use, male gender, and each decade of age over 40 years. A CS cutoff ≥5 or <5 was used to categorize patients as high risk (n = 303 [21%]) or low risk (n = 1,158 [79%]). Perfusion scans were categorized as low, intermediate, or high risk on the basis of the global stress score (GSS).
Results High-risk scans were more common in patients with a high-risk CS (26.4% vs. 9.5%, p < 0.0001). The CS and GSS were significant independent predictors of cardiac death. However, in patients with a low CS, seven-year cardiac survival was excellent, regardless of the GSS (99% for normal scans, 99% for mildly abnormal scans, and 99% for severely abnormal scans). In contrast, patients with a high CS had a lower seven-year survival rate (92%), which varied with GSS (94% for normal scans, 94% for mildly abnormal scans, and 84% for severely abnormal scans; p < 0.001).
Conclusions In symptomatic patients with low-risk Duke treadmill scores and low clinical risk, myocardial perfusion imaging is of limited prognostic value. In patients with low-risk Duke treadmill scores and high clinical risk, annual cardiac mortality (>1%) is not low, and myocardial perfusion imaging has independent prognostic value.
Current American College of Cardiology/American Heart Association (ACC/AHA) guidelines recommend using the standard exercise test without imaging as the initial test in the noninvasive evaluation of patients with suspected coronary artery disease (CAD) who have an interpretable electrocardiogram (ECG) and are able to exercise (1). In patients who subsequently have a low-risk Duke treadmill score, the guidelines recommend only medical management. However, even though the event rate in patients with low-risk treadmill scores is very low, myocardial perfusion imaging can identify a subset of such patients who are at higher risk (2). However, this use of perfusion imaging is controversial. Two-thirds of patients undergoing an exercise treadmill test have low-risk Duke treadmill scores (3,4), and it may not be cost-effective to perform perfusion imaging on all of these patients to identify a small number of high-risk individuals.
The purpose of this study was to investigate whether a simple clinical score (CS), derived from commonly used clinical variables, could be applied to symptomatic patients with low-risk treadmill scores to identify the small cohort of high-risk patients likely to yield prognostic information from the nuclear scan. We hypothesized that perfusion imaging would have incremental value only in the group whose clinical characteristics put them at higher risk.
The study population was identified from the database of the Mayo Nuclear Cardiology Laboratory. A total of 3,251 patients referred for evaluation of chest pain or dyspnea underwent exercise thallium-201 imaging between January 1, 1989, and December 31, 1991. Of these, 1,461 patients (mean age 58.6 ± 11.1 years) were found to have low-risk Duke treadmill scores. Exclusion criteria included known cardiomyopathy, valvular heart disease, previous percutaneous transluminal coronary angioplasty or coronary artery bypass graft surgery, recent (within three months) myocardial infarction (MI), or ECG findings that precluded calculation of the Duke treadmill score (e.g., left bundle branch block). Chest pain was defined as typical angina, atypical pain, or noncardiac pain, using the criteria of Diamond et al. (5), on the basis of chart review and supplemental patient interview, as required.
All patients underwent standard symptom-limited treadmill testing using the Bruce, modified Bruce, or Naughton protocol. Near peak exercise, 3 to 4 mCi of thallium-201 was injected, and the patient exercised for an additional minute, after which the single-photon emission computed tomography (SPECT) was initiated. Redistribution images were obtained 4 h later. For patients who underwent the test after January 1, 1990, an additional 1.0 to 1.5 mCi was injected 30 min before the redistribution images. The magnitude of horizontal or downsloping ST-segment depression, treadmill angina index (no angina = 0; angina during test = 1; test stopped due to angina = 2), and exercise duration was recorded. For those who exercised using the Naughton or Bruce protocol, a conversion factor was applied to the exercise duration to equate the exercise duration with the Bruce protocol (6). The Duke treadmill score was calculated for each patient as follows: exercise time (Bruce protocol) − (5 × maximum ST-segment deviation) − (4 × angina index). Patients were classified as low risk, based on the Duke classification, with a score of ≥5 (3).
Stress perfusion images were obtained 10 min after completion of exercise, and redistribution images were obtained 4 h later. Anterior planar images were first obtained (30% window centered at 68 to 80 keV and 20% window centered at 167 keV), followed by the SPECT acquisition using a 64 × 64 matrix of 30 views at 40 s/view. Imaging with SPECT was performed with a large field-of-view camera (Elscint 409, Haifa, Israel) equipped with an all-purpose collimator. After filtered back-projection, the images were reconstructed to obtain the short-axis, vertical, and horizontal long-axis images.
Interpretation of SPECT images
The images were scored using a five-point scoring system by the consensus of two experienced observers. The short-axis slices were divided into 14 segments corresponding to the anterior, lateral, septal, and inferior walls at the apex, mid, and basal levels of the left ventricle. At the mid and basal levels of the left ventricle, the septum was further divided into anterior and inferior septal walls (7). The five-point scoring used to grade each segment was as follows: 0 = absent perfusion; 1 = severe hypoperfusion; 2 = moderate hypoperfusion; 3 = mild hypoperfusion; and 4 = normal perfusion. A fixed defect was considered present if the degree of hypoperfusion was at least moderate and equal on both rest and stress images. A score of 3 on both rest and stress images was considered to represent an attenuation artifact. An increase of >1 grade on the delayed images was considered a reversible defect. Defects in the short-axis slices were confirmed by the other two planes.
Nuclear variables were defined using the five-point scoring system. A global stress score (GSS) was obtained by adding the scores on all the stress short-axis images. A global rest score (GRS) was obtained by adding the scores of all the redistribution short-axis images. A global difference score (GDS) was obtained by subtracting GSS from GRS. The maximum that could be obtained for a completely normal set of images was 56 (14 × 4). A score of 53 to 56 was defined as low risk (normal or mildly abnormal), a score of 48 to 52 as intermediate risk (moderately abnormal), and a score of <48 as high risk (severely abnormal). This classification of scores was defined prospectively, based on previous work from Cedars Sinai (2,8).
Hubbard et al. (9)developed the CS retrospectively in a population undergoing evaluation for CAD as a predictor of three-vessel or left main disease. A simple five-point scoring system was developed after consideration of 16 clinical and ECG variables. The variables included in the five-point scoring were male gender, history of MI (clinical event and Q waves on ECG), diabetes, insulin use, and typical angina. This score was an accurate predictor of severe CAD in different age groups. Additional points can be assigned to age, with 1 point for age 40 to 49 years, 2 points for age 50 to 59 years, 3 points for age 60 to 69 years, 4 points for age 70 to 79 years, and 5 points for age >80 years. A male >80 years old with typical angina, a history of MI, and diabetes requiring insulin would receive the maximum of 10 points.
In a prospective study of 2,255 patients, a score ≥5 identified patients with a higher risk of adverse events (10). Therefore, this cutoff value was used in this study.
Personnel blinded to the results of the nuclear study conducted the follow-up by telephone interviews and chart review. Hard events were defined as cardiac death (confirmed by death certificate or medical records) or nonfatal MI (documented by history, ECG changes, and cardiac enzymes). Late revascularization was defined as percutaneous intervention or coronary artery bypass grafting more than three months after the stress test. Follow-up was 99% complete at a mean of 7 ± 1 year.
The end points analyzed by the Kaplan-Meier method were: 1) cardiac survival (patients with noncardiac death or revascularization were censored); 2) cardiac survival free of nonfatal MI (patients with noncardiac death or revascularization were censored); and 3) cardiac survival free of nonfatal MI and late revascularization (patients with noncardiac death or revascularization within three months of the thallium study were censored). The association between the CS and GSS and each of these end points, during the entire follow-up period, was evaluated by Cox proportional hazards analysis on both a univariate and bivariate (each variable adjusted for the other) basis. The GSS was selected prospectively as the single nuclear variable to be included in the analysis, based on previous studies (9,11). We tested the proportional hazards assumption by including the product of GSS with time and CS with time in the analysis. The proportional hazards assumption was confirmed for CS. There was a modest time dependence of the hazard ratio for GSS, which achieved significance only for the end point of cardiac survival free of nonfatal MI, and was likely of limited clinical significance. Annual event rates were calculated by dividing the event rate over the follow-up period of seven years by seven.
Each of the three GSS groups described earlier was further subdivided by a CS of <5 or ≥5, thus creating a total of six risk groups. Group comparisons for the three end points were made by using the log-rank test. Statistical significance was defined as p < 0.05.
The baseline characteristics of the study group are presented in Table 1. The majority of patients were male. The prevalence of diabetes was low (8.2%), but hypertension and hypercholesterolemia were present in about 45% of patients. The majority had atypical angina.
Risk stratification by CS and nuclear variables
The number of patients in each of the six risk groups is shown in Table 2. In this low-risk population, 72% of the scans were low risk, 15% were intermediate risk, and 13% were high risk. The CS identified 1,158 patients as being low risk. Of these, the majority, 885 (76%) also had low-risk GSS, whereas 163 (14%) had intermediate-risk, and 110 (9.5%) had high-risk GSS. The distribution of GSS in the 303 patients with a high-risk CS was significantly different (p < 0.0001): 167 (55%) had low-risk, 57 (19%) had intermediate-risk, and 79 (26%) had high-risk GSS. Patients with a low-risk CS were far less likely to have a severely abnormal GSS than those with a high-risk CS (9.5% vs. 26%, p < 0.001).
The total number of events was 211: 30 deaths, 55 nonfatal MIs, and 124 late revascularizations. Overall, seven-year cardiac mortality in this cohort was low at 2%. On univariate analysis, both the CS and GSS were predictive of each of the end points (p < 0.0001) (Table 3). The GRS was predictive of each of the end points (p < 0.001). The GDS was predictive of cardiac death and cardiac death/nonfatal MI/late revascularization (both p < 0.001) but less predictive of cardiac death/nonfatal MI (p = 0.08). On bivariate (two-variable) analysis, the independent predictive power of CS appeared to be greater than that of GSS (Table 3). However, the GSS was independently significant for the end points of cardiac death and cardiac death/nonfatal MI/late revascularization. The overall seven-year cardiac survival in the two CS groups was 99% for the low-risk group and 92% for the high-risk group, regardless of the GSS (p < 0.0001). Also noted were significant differences in seven-year cardiac survival free of nonfatal MI (96% vs. 86%, p < 0.0001) and nonfatal MI and late revascularization (88% vs. 72%, p < 0.0001) (Table 4).
Interaction between perfusion score and CS
The impact of GSS on outcome appeared to be related to the CS, although the formal test of interaction only achieved significance (p = 0.009) for cardiac survival free of nonfatal MI and late revascularization. Patients with a low CS had an excellent survival, regardless of GSS, with an annual cardiac mortality rate of <0.5% (Fig. 1). In contrast, survival in the group with a high CS worsened with increasing abnormality on the perfusion scan (p < 0.001) (Fig. 2). The results were similar for the end point of cardiac death or nonfatal MI (Table 4). Annual event rates for cardiac death/nonfatal MI, stratified on the basis of CS and GSS, are shown in Figure 3. Patients with a low CS had an annual rate of cardiac death/MI of <1%, regardless of GSS. In contrast, the GSS predicted event rates in the high CS group.
The main goal of this study was to assess the value of nuclear perfusion imaging in an overall low-risk population defined by the presence of low-risk Duke treadmill scores. The results suggest that a simple CS has significant prognostic value in this population and can categorize low- and high-risk subgroups that have considerably different outcomes. Nuclear imaging does not provide any additional prognostic information for predicting hard events in the group with a low CS, as there is no difference in event rates among those with a low-, intermediate-, or high-risk GSS. In contrast, in the group with a high CS, whose overall survival is worse, the prognosis worsens as a function of the abnormality on the perfusion scan.
The ACC/AHA guidelines recommend the treadmill test as the first step in patients with interpretable baseline ECGs who are able to exercise, with the exception of those who have undergone revascularization (1). Our study group with low-risk treadmill scores had a very low seven-year cardiac mortality rate of 2% (annual mortality 0.28%). The performance of nuclear imaging in all of these patients after treadmill testing, to identify a few higher risk individuals, does not appear to be efficient. However, those patients with a high-risk CS and low-risk treadmill score are candidates for perfusion imaging, as their seven-year mortality rate is 8% (annual mortality 1.1%), and they can be further risk-stratified by perfusion imaging.
Our results further suggest the possibility that clinical risk (as reflected by the CS) should have a role in the selection of the initialdiagnostic test. A CS <5 mandates only a regular exercise treadmill test in patients who can exercise and have an interpretable baseline ECG. If such patients have a low-risk treadmill score, they have an annual cardiac mortality rate <0.25%, and perfusion imaging does not provide enough additional information to justify its added cost. However, in patients with a CS ≥5, a low-risk treadmill score would still imply an annual cardiac mortality rate of 1.1%. Unless the physician and patient are willing to accept this level of risk, exercise perfusion imaging may be a more appropriate initialtest in such patients, as it can better risk-stratify the patient and avoid the inefficiency of two tests. However, our study was not designed to address this issue, which would require consideration of all patients with a high CS, not just those with a low-risk treadmill score.
Role of CS
The CS initially developed by Hubbard et al. (9)paralleled efforts by Pryor et al. (12), the Coronary Artery Surgery Study (CASS) (13)investigators, and others (14–17)to develop a clinical model to predict disease severity and prognosis in patients with CAD. Hubbard et al. (9)demonstrated that the probability of severe CAD increased as a function of the CS. This study utilizes the CS as a measure of pretest probability of severe CAD (and of adverse outcomes) in an overall low-risk population. The score clearly identifies a higher risk subgroup (even in this low-risk population) that merits a more aggressive diagnostic approach with perfusion imaging.
Comparison with previous studies
This is the first study to focus specifically on perfusion imaging in patients with low-risk treadmill scores. Using the Duke treadmill score, Mark et al. (3)classified 62% of 613 outpatients as low risk, with an annual mortality rate of 0.25%. However, this group is heterogeneous. In our study, 21% of this population had a high-risk CS and an annual cardiac mortality rate of 1.1%.
Our results are consistent with previous studies (11,18,19)that have examined the incremental value of exercise perfusion imaging in the risk stratification of patients with a normal ECG (and therefore intrinsically at low risk). In these studies, clinical variables were prognostically important, independent of the exercise ECG.
Our results are also consistent with Hachamovitch et al. (2), who reported the incremental value of perfusion imaging in a population with low-risk Duke treadmill scores. In their study, the hard event rate (including cardiac death and MI) increased as a function of scan results (p < 0.05 across scan results). Their subgroup with high-risk scans comprised 51 patients of the total of 926 (5.5%); among these patients, there were four hard events occurring over a period of 1.5 years. In comparison, our study group had a higher prevalence of high-risk scans (12.9%), which may be attributable in part to the inclusion of patients with previous MIs. The total number of hard events among the patients with high-risk scans, 19, in our study was higher, reflecting the larger subgroup and longer follow-up, but the event rates were not significantly different.
A later study by Hachamovitch et al. (20)was restricted to patients with normal resting ECGs. The incremental value of perfusion imaging was greatest (and most cost-effective) in patients with an intermediate to high pretest likelihood of CAD and an intermediate to high post-test likelihood of CAD. The prevalence of high-risk scan (16%) and hard events, 10, in patients with low-risk treadmill scores was closer to that in the current study.
These results are from a symptomatic patient population seen in a tertiary care setting. The applicability of these results to asymptomatic patients or other settings needs to be confirmed. The results of this study should not be applied to the patients in whom the Duke treadmill score cannot be calculated or in whom it has not been validated. The prevalence of diabetes in our population was modest (8%) and lower than the prevalence we reported in other studies, suggesting that diabetics are less likely to have low-risk Duke treadmill scores. Further studies are needed in diabetics.
The cutoff value for the CS was identified by post hoc analysis from a previous study and needs to be validated in other groups. Because we had only modest numbers of patients with a CS near the cutoff value (i.e., 4 or 5), we could not accurately estimate the annual mortality in such patients.
Information on other potentially useful parameters, such as peripheral arterial disease, carotid disease, and heart failure, was not available. Because these patients were studied before the introduction of gated SPECT, we could not assess the impact of ejection fraction measurements.
Another limitation is that we did not consider other exercise test findings to identify the low-risk group, including chronotropic incompetence (21), abnormal heart rate recovery (22), or ventricular ectopy (23). We used only the Duke treadmill score, which is based on exercise capacity, ST-segment depression, and exercise-induced angina (3). Recent work from our institution has suggested that exercise capacity is the only prognostically important component of the Duke treadmill score for prediction of mortality and cardiovascular events in the elderly (24).
Patients with a low CS may still have CAD (e.g., a 55-year-old male with typical angina but no diabetes or previous MI). Our results suggest that such a patient has a low risk of subsequent events if he or she has a low-risk treadmill score.
Despite these limitations, the results show that in symptomatic patients with low-risk Duke treadmill scores and high clinical risk, annual cardiac mortality (>1%) is not low, and myocardial perfusion imaging has independent prognostic value. Myocardial perfusion imaging may be a more appropriate initial stress test in patients with high clinical risk, but further studies are needed to assess this possibility.
- American College of Cardiology/American Heart Association
- coronary artery disease
- clinical score
- global difference score
- global rest score
- global stress score
- myocardial infarction
- single-photon emission computed tomography
- Received March 4, 2003.
- Revision received July 23, 2003.
- Accepted September 9, 2003.
- American College of Cardiology Foundation
- Gibbons R.J.,
- Chatterjee K.,
- Daley J.,
- et al.
- Hachamovitch R.,
- Berman D.S.,
- Kiat H.,
- et al.
- Diamond G.A.,
- Staniloff H.M.,
- Forrester J.S.,
- et al.
- Pollock M.L.,
- Wilmore J.H.
- Hachamovitch R.,
- Berman D.S.,
- Shaw L.J.,
- et al.
- Fisher L.D.,
- Kennedy J.W.,
- Chaitman B.R.,
- et al.
- Diamond G.A.,
- Staniloff H.M.,
- Forrester J.S.,
- Pollock B.H.,
- Swan H.J.
- Hachamovitch R.,
- Berman D.S.,
- Kiat H.,
- Cohen I.,
- Friedman J.D.,
- Shaw L.J.
- Shetler K.,
- Marcus R.,
- Froelicher V.F.,
- et al.