- Email: [email protected]
A model to predict multivessel coronary artery disease from the exercise thallium-201 stress test
A Model to Predict Multivessei Coronary Artery Disease from the Exercise Thallium-201 Stress Test STEWARTG. POLLOCK,
M.D., ROBERT
D. ABBOTT,
Ph.D., Charlottsville, Virginia,
CHARLES A. BOUCHER, M.D., Boston, Massachusetts, DENNY D. WATSON, Ph.D., SANJIV KAUL, M.D., Charlottsville, Virginia
PURPOSE: The aim of this study was to (1) determine w h e t h e r nonlmnging variables add to the diagnostic information available from exercise thallinm-201 images for the detection of multivessel coro-Ary a r t e r y disease; and (2) to develop a model based on the exercise thallium201 stress test to predict the presence of multivessel disease. P A T I E N T S A N D METHODS: The study populations included 383 patients referred to the University of Virginia and 325 patients referred to the Massachusetts General Hospital for evaluation of chest pain. All patients underwent both cardiac catheterization and exercise thallium-201 stress testing between 1978 and 1981. RESULTS: In the University of Virginia cohort, at each level of thallium-201 abnormality (no defects, one defect, more than one defect), ST depression and patient age added significantly in the detection of multivessel disease. Logistic regression analysis using data from these patients identified three independent predictors of multivessel disease: initial thallium-201 defects, ST depression, and age. A model was developed to predict multivessel disease based on these variables. As might be expected, the risk of multivessel disease predicted by the model was similar to that actually observed in the University of Virginia population. More importantly, however, the model was accurate in predicting the occurrence of multivessel disease in the unrelated population studied at the Massachusetts General Hospital. CONCLUSION: It is, therefore, concluded that (1) nonlmaging variables (age and exercise-induced From the Divisionsof Cardiologyand Biostatistics, Departmentof Medicine, University of Virginia School of Medicine. Charlottesville,Virginia, and the Cardiac Unit. Department of Medicine, MassachusettsGeneral Hospital, Boston, Massachusetts. This work was supportedin part by grants (RO1-HL-26205and ROI-HL262].5) from the National Institutes of Health, Bethesda, Maryland. Dr. Kaul is the recipient of the Clinical Investigator (KO8-HL-01833) and FIRST(R29-HL-38345) Awardsof the NationalInstitutes of Health. Requestsfor reprints shouldbe addressedto SanjivKaul, M.D., Department of Medicine, Box ]58, University of Virginia School of Medicine, Charlottesville, Virginia 22908. Manuscript submitted July 6. 1990, and accepted in revised form November 19. 1990.
ST depression) add independent information to thalllum-201 imaging data in the detection of multivessel disease; and (2) a model ha~ been developed based on the exercise thAnium-201 stress test that can accurately predict the probability of multivessel disease in other populations.
here are two principal aspects to the manage-
T ment of patients with coronary artery disease. The first is treatment of symptoms and the second is prolongation of life. Although patients with single-vessel disease may undergo a revascularization procedure (bypass surgery or balloon angioplasty) for control of intractable symptoms, it is patients with multivessel disease who benefit most from revascularization in terms of prolongation of life [1,2]. It is also such patients who, if treated medically, experience adverse cardiac events [3-5]. Classification of patients with symptomatic coronary artery disease into those with and those without multivessel disease is, therefore, important for optimal patient management. Since an estimated 5 million people in the United States have symptomatic coronary artery disease [6], identification of patients with multivessel disease by cardiac catheterization would be logistically burdensome. A more feasible approach is to first select patients likely to have multivessel disease by noninvasive tests, and then p e r f o r m c o r o n a r y a n g i o g r a p h y to g u i d e revascularization. Exercise thallium-201 imaging is, arguably, one of the more reliable noninvasive methods for the detection of coronary artery disease. Its sensitivity, using quantitative analysis, is close to 90% and its specificity is close to 80% [7], with most of the falsenegative results occurring in patients with less severe disease [8]. Because exercise electrocardiography is less sensitive than thallium-201 imaging [913], its additional value in the diagnosis of coronary artery disease, compared with thallium-201 imaging alone, is questionable. Whereas some studies have demonstrated that the results of exercise electrocardiography are additive to those of thallium201 imaging, the specific manner in which this is 8o has not been defined [12,14]. Furthermore, alMarch 1991
The American Journal of Medicine Volume 90
345
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL Exercise E l e c t r o c a r d i o g r a p h y
TABLE I Characteristics of the Two Patient Cohorts
Variable
UVA (n = 383)
MGH (n = 325)
Age Males No CAD or SVD
58 • 10 326 (85%) 165 (43%)
53 4- 10 272 (84%) 131 (40%)
Multivessel CAD
218 (57%)
194 (60%)
p Value 0.001 0.60 0.49
UVA= Universityof Virginia; MGH = MassachusettsGeneralHospital:CAD= coronar artery disease;SVD= single-vesseldisease.
though clinical variables, such as age, gender, and type of chest pain, have also been shown to add marginally to the imaging data for predicting the presence or absence of coronary artery disease [12,15], their value in predicting multivessel disease is unknown. Since multivessel disease is one end of the spectrum of coronary artery disease rather than a separate disease entity, it stands to reason that the same factors predictive of any coronary artery disease would be predictive of multivessel disease. We hypothesized, therefore, that the noninvasive diagnosis of multivessel disease could be optimized by combining thallium-201 imaging variables with those of exercise electrocardiography and clinical history. We evaluated this hypothesis in a population of 383 patients referred to the University of Virginia Medical Center for evaluation of chest pain. We derived a model to predict the probability of multivessel disease based on this patient cohort. We then tested the validity of this model by applying it to a similar but unrelated population of 325 patients evaluated at the Massachusetts~ General Hospital. PATIENTS
AND METHODS
The patients and methods from both the University of Virginia and Massachusetts General Hospital have been described in detail in previous reports [8,16]. A limited summary of patient characteristics and methods is provided here.
All patients underwent symptom-limited exercise testing terminated for fatigue, claudication, angina, dyspnea, hypotension, or ventricular tachycardia [8,16]. The electrocardiogram, heart rate, and blood pressure were recorded every minute during baseline, exercise, and the 5-minute recovery period. All test results were classified as either positive or negative. When the baseline electrocardiogram was normal, a test was considered positive if it had 1 mm of flat or downsloping ST depression, or 2 mm of upsloping ST depression measured 0.08 msec after the J point [17]. In the presence of baseline ST segment depression, but in the absence of left bundle branch block, left ventricular hypertrophy, or digitalis use, the test result was considered positive if there was an additional 2 mm of ST depression. Any other test result was considered negative. Nondiagnostic test results due to left bundle branch block, left ventricular hypertrophy, or digitalis use were interpreted as negative.
Exercise Thallium-201 Imaging At peak exercise, 1.5 to 2.0 mCi of thallium-201 was injected intravenously and the patient was encouraged to exercise for another 30 to 60 seconds. Images were obtained in the anterior and 45 ° and 70 ° left anterior oblique views. In the University of Virginia cohort, quantitative analysis was performed in seven segments in the anterior and 45 ° left anterior oblique views at the time when these data were collected [11]. Although the quantitative analysis at the Massachusetts General Hospital was performed in 15 segments from all three views [8], for this study only the seven segments corresponding to those quantitated at the University of Virginia were analyzed. The quantitative lung/heart ratio of thallium-201 was not analyzed since it was not routinely performed in the patients studied at the University of Virginia when data for this study were collected. Since the purpose of this study was to develop a model for the detection of multivessel disease; rather than prognosis, only initial defects were analyzed; redistribution was not included in the analysis.
Patient Population Both cohorts were comprised of patients referred for evaluation of chest pain who underwent both exercise thallium-201 imaging and cardiac catheterization within 3 months of each other between 1978 and 1981. Although the patients from the Massachusetts General Hospital were younger, there were no significant differences between the two cohorts in terms of age and the incidence of multivessel disease (Table I).
346
Cardiac Catheterization
The coronary angiographic data were analyzed in a blinded manner. They were analyzed by one experienced observer at the University of Virginia by means of calipers and a ruler. At the Massachusetts General Hospital, the data were subjectively analyzed by three observers and a consensus opinion was obtained. Multivessel disease was defined as stenosis (50% or greater luminal diameter narrow-
March 1991 The American Journal of Medicine Volume 90
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL
ing) of at least two of the three main arteries or their major branches [18]. The major branch of the left anterior descending artery was the diagonal artery. When present, an intermedius artery was included as a branch of the left anterior descending artery. The major branch of the left circumflex artery was the obtuse marginal artery and the major branch of the right coronary artery was the posterior descending artery. In left dominant systems, the posterior descending artery was considered a major branch of the left circumflex artery. Patients with coronary artery disease involving less than two major arteries or their major branches or showing no significant stenosis were coded as not having multivessel disease. Left main stenosis was considered equivalent to multivessel disease. Statistical Analysis
Logistic regression models were used to help evaluate the performance of findings from exercise electrocardiography, exercise thallium-201 imaging, and clinical risk factors in predicting the presence or absence of multivessel disease [19]. Tests of significance were based on likelihood ratio statistics [20]. All tests of significance were two-sided. Estimates of odds ratios were derived from the regression coefficients (also based on maximum likelihood calculations) by comparing meaningful groups (the relative odds of multivessel disease in men versus women for instance). Ninety-five percent confidence intervals for the odds ratios were also included. To evaluate the model's fit, observed rates of multivessel disease were compared with predicted rates. Predicted rates were calculated by using a logistic model to estimate the probability of multivessel disease for each patient. These rates were calculated from this model by ranking the individual probabilities. The ranked probabilities were then grouped into quintiles where patients who fell in the lowest quintile were predicted to have the lowest rate of multivessel disease, whereas patients in the highest quintile were predicted to have the highest rate. Within each quintile, the predicted rate of multivessel disease was determined by averaging the individual probabilities. The predicted values were then compared with the actual rates of multivessel disease using a chi-square goodness-offit test. To further evaluate the model's performance, the logistic model derived from the University of Virginia data was used to predict multivessel disease in the patients from the Massachusetts General Hospital. As with the University of Virginia patients, those from the Massachusetts General Hospital
TABLE II Description of Patients from the University of Virginia Cohort Based upon the Number of Diseased Vessels Variable Males >58 years Typical angina >_2 segments with thallium-201 defects ST depression on EKG Exercise-induced VA Abnormal BP response* <5 METSworkload Exercise-induced angina
NoCAD(%) (n = 93)
SVD(%) (n = 72)
MVD(%) (n = 218)
68 (73) 37 (40) 39 (42) 14 (15)
63 (88) 32 (44) 50 (69) 42 (65)
195 (89) 141 (65) 160 (73) 163 (74)
21 (23) 26 (28) 12 (13) 13 (14) 13 (14)
28 (39) 38 (53) 15 (21) 17 (24) 13 (18)
135 (62) 98 (45) 48 (22) 64 (29) 50 (23)
BP = bloodpressure:CAD= coronaryartery disease:EY,G = electrocardiogram:MVO= multivesseldisease:SVD= single-vesseldisease:VA = ventriculararrhythmias. ° Definedas fall or failureto riseof systolicbloodpressureduringexercise.
were also ranked and grouped into quintiles based on patient-specific risk factors. The predicted rates of multivessel disease in each quintile were then compared with the actual rates using a chi-square goodness-of-fit test. RESULTS Univariate Analysis
The distribution of coronary artery disease in relation to clinical, exercise electrocardiographic, and exercise thallium-201 variables in the University of Virginia population is displayed in Table II. Except for exercise-induced ventricular arrhythmias, there is a progressive rise in the frequency of occurrence of every variable from no coronary artery disease, to single-vessel coronary artery disease, to multivessel disease. Nevertheless, the frequency of exercise-induced ventricular arrhythmias is higher in patients with coronary artery disease than in those without this condition. Patients over 58 years in age constitute fewer than half of those with no coronary artery disease and single-vessel coronary artery disease, but account for the majority of those with multivessel disease. Patients with more than one thallium defect are uncommon in the no coronary artery disease group, b u t make up approximately three fourths of the multivessel disease group. The occurrence of ST depression on the exercise electrocardiogram increases steadily among the groups. Approximately one fifth of the patients without coronary artery disease have ST depression compared with two fifths and three fifths with singlevessel and multivessel disease, respectively. The odds ratios (with their 95% confidence intervals) derived from univariate logistic analysis are shown in T a b l e III. Women have half the odds of multivessel disease relative to men (p <0.01), while
March 1991 The American Journal of Medicine Volume 90
347
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL
TABLE III
Logistic Regression Analysis with Multivessel Disease Predicted by Each Risk Factor
Risk Factor Gender Age (years) Type of chest pain METS achieved Abnormal BP response ST depression on EKG Exercise-induced angina Thallium-201 defects Ventricular arrhythmias
Comparison
Odds Ratio
Female versus male 20-year difference Typical versus atypical 5 MET difference Yes versus no Yes versus no Yes versus no Two-segment difference Yes versus no
0.5 = 2.3 t 2.41 0.St 1.5 3.91 1.6 3.Or 1.3
UnivariateAnalysis 95% Confidence Interval 0.3-0.8 1.5-3.5 1.5-3.6 0.4-0.8 0.9-2.5 2.5-5.9 0.9-2.7 2.2-4.1 0.9-2.0
Odds Ratio
Multivariate Analysis 95% Confidence Interval
0.6 1.8~ 1.5 0.6 0.9 3.5I I. 1 2.51 0.8
0.3-1.1 1.1-2.4 0.9-2.6 0.3-1.1 O.5-1.7 2.1-5.6 0.6-2.1 1.8-3.6 0.5-1.3
ender = l-female. O-male:type of chest pain = l-typical. O-nottypical: abnormalBP (blood pressure)responseto exercise = l-yes, O-no:ST depression on the EKG(electrocardioram) during exercise = 1-yes. O-no:exercise-inducedangina = l-yes; O-no:exercise-inducedventricular arrhythmias = l-yes. O-no. = Women had significantly less multivesseldisease(p
older pptients have double the odds of this condition compared with those 20 years younger (p <0.001). Patients with typical angina pectoris have twice the odds of multivessel disease in contrast to those with atypical angina (p <0.001). A 5 MET difference in work load attained during exercise doubles the odds of multivessel disease for those who performed less work (p <0.05). The occurrence of ST depression on exercise is associated with a nearly fourfold excess of the odds of multivessel disease as compared with its absence (p <0.001). The odds of having multivessel disease triples in the presence of two segmental defects on initial thallium-201 images (p <0.001). The occurrence of angina, ventricular arrhythmias, or an abnormal blood pressure response during exercise fails to raise the odds of multivessel disease in a significant manner. Multivariate Analysis
The results of the multivariate analysis, with the presence or absence of multivessel disease as the dependent variable, are also shown in Table III. After controlling for all variables, the occurrence of initial thallium-201 defects and exercise-induced ST depression on the electrocardiogram remain significantly and strongly associated with multivessel disease (p <0.001). The odds of multivessel disease increases 2.5 times for patients with two thallium201 defects compared with those having one or no defects. In the presence of ST depression, the odds of multivessel disease more than triples. Age is the only other covariate that remains a significant predictor of multivessel disease (p <0.05) on multivariate analysis, although its association with multivessel d i s e a s e is w e a k e r t h a n t h a t of i n i t i a l thallium-201 defects and ST depression. The joint effects of age, ST depression, and the number of initial thallium-201 defects are depicted
348
in F i g u r e 1. The observed rate of multivessel disease rises as the number of initial thallium-201 defects increases, regardless of age and the presence of ST depression. The lowest rate of multivessel disease, however, is seen in patients younger than 58 years who do not exhibit initial thallium-201 defects or ST depression. In this patient subset, the occurrence of multivessel disease is unlikely. The highest rate for multivessel disease is noted in patients older than 58 years who exhibit multiple initial thallium-201 defects and ST depression. In this subset of patients, the occurrence of multivessel disease is highly likely. Model to Predict Multivessel Disease
Based on the results in Table III, age, ST depression, and the number of initial thallium-201 defects were chosen to estimate the probability of multivessel disease in a logistic regression model. The resultant coefficients and intercept yielded the following expression for the estimated probability of multivessel disease:
March 1991 The American Journal of Medicine Volume 90
p(MVD) -- 1/(1 + exp[3.09 - 0.032*age - 1.28"ST - 0.507*defects]) where M~VD -- multivessel disease, age = age in years, ST = one for an abnormal ST-segment response during exercise and zero otherwise, and defects -- number of defects (range: zero to seven) on initial thallium-201 images. To examine the model's goodness of fit, the observed and predicted rates of multivessel disease in the University of Virginia cohort were compared. As illustrated in F i g u r e 2, there were no differences between the observed and predicted rates of multivessel disease in each quintile (p = 0.97). This finding is not unexpected since the model was derived from the identical population used to examine
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL
Figure 1. The additive value of including information regarding age (either less than 58 years or 58 years or older) and presence or absence of exercise-induced ST depression on the electrocardiogram (x-axis) to each level of abnormality on initial thallium-201 imaging (no defect, one defect, more than one defect) (z-axis) on the detection of c o r o n a r y a r t e r y disease (y-axis). See text for details,
/
/
/
~= =
100
L
80 6o
~_ ~ ~
" >1
4o
I
20
)EFECT
0 >
58
yrs
<
~. ST d e p
58
yrs
+ ST dep
Observed
.¢ uJ e~
Figure 2. The rate of multivessel disease in each quintile as predicted by the model derived using logistic regression analysis from the U n i v e r s i t y of Virginia p o p u l a t i o n and the actual rate of multivessel disease as noted on coronary angiography in the same patients. See text for details.
u) u)
DEFECT
DEFECT
> -
58
yrl - S T dep
yrs
<
ST d e p
~
58
Predicted
100 ]
80 60 40
m
I-
=i ~e
20 0 1
2
it. To examine the model's utility, it was applied to the unrelated patient cohort studied at the Massachusetts General Hospital. The observed and predicted rates of multivessel disease in this population are illustrated in F i g u r e 3 by quintile. Similar to the data in Figure 2, there were no differences between the observed and predicted rates of multivessel disease in this patient population in each quintile (p = 0.91). COMMENTS
In the current study, we hypothesized that despite the superior ability of exercise thallium-201 imaging to diagnose coronary artery disease, clinical and exercise electrocardiographic variables would provide additional independent diagnostic information for the detection of multivessel disease. Through logistic regression analysis, we demonstrated this to be true and developed a model to predict multivessel disease using selected thallium201, exercise electrocardiographic, and clinical vari-
3 4 QUINTILE OF RISK
5
ables. The model was found not only to be accurate in the population from which it was derived, but it was also valid when applied to an unrelated, but largely similar, patient population. Value of Nonimaging Variables in Predicting Multivessel Disease
Since exercise electrocardiography is less sensitive than thallium-201 imaging [9-13], its additional value in determining the presence or absence of coronary artery disease compared with thallium201 imaging alone is not clear. For example, if the thallium-201 test is positive and the exercise electrocardiogram is negative, one usually assumes the latter to be a false-negative result. Similarly, if the thallium-201 test is negative and the exercise electrocardiogram is positive, one might consider the latter to represent a false-positive result. What does it mean, however, when both tests provide concordant information? Our results indicate that when both tests are negative in a patient under 58 years of
March 1991 The American Journal of Medicine Volume 90
349
PREDICTION OF MULTIVESSELDISEASEFROMTHE EXERCISETHALLIUM-201STRESSTEST / POLLOCKET AL
Observed
< uJ u)
~
Predicted
lOO 80
,,-I,
60
~ r~
4o
~
0
2o 1
2
3 QUINTILE
4 OF RISK
age, multivessel disease is unlikely. Similarly, when both tests are positive in such an individual, multivessel disease is highly likely. In the latter situation, the probability of multivessel disease is high even if multiple thaUium-201 abnormalities are not seen. This information is of obvious clinical utility. The results from logistic regression analysis demonstrate that age, number of initial thallium-201 defects, and the presence of exercise-induced ST depression are independent predictors of multivessel disease. When these factors are used jointly, it is possible to enhance the prediction of multivessel disease. Since age, ST depression, and number of thallium-201 defects are known to be useful in predicting the presence or absence of coronary artery disease [12,14], their predictive value for the occurrence of multivessel disease is not unexpec~d. The importance of these variables as predictors of multivessel disease can also be regarded from a bayesian point of view where age provides a pretest probability that is modified by the results of exercise electrocardiography and thallium-201 imaging [14,21]. The finding that only age, ST depression, and number of initial thallium-201 defects are independent predictors of multivessel disease should not be construed as meaning that the other variables are not useful. From univariate analysis, both gender and type of chest pain (typical or not) were, as would be expected [21], also significant predictors of multivessel disease. Multivessel disease was also significantly associated, on univariate analysis, with exercise-induced ventricular arrhythmias and poor exercise performance, findings that have been previously shown to be associated with more severe disease [22-25]. In the current study, however, all
350
Figure 3. The predicted rate of multivessel disease in each quintile of the Massachusetts General Hospital population based upon the model derived from the University of Virginia population and the observed rate of multivessel disease in the same population as noted on coronary angiography. See text for details.
these factors were not significant univariate predictors of multivessel disease because they were superseded by the overriding influence of age, ST depression, and number of initial thallium-201 defects. Value of Other Thallium-201 Variables in Predicting Multivessel Disease
In this analysis, we did not include redistribution on delayed thallium-201 images or increased lung uptake of thallium-201 to determine the probability of multivessel disease. Redistribution on the delayed thallium-201 images reflects reversible ischemia and myocardial viability [7]. Although useful in assessing the risk of future ischemic events [16,26-29], it probably does not have a role in assessing the extent of coronary artery disease. Abnormal lung uptake of thallium-201 has been shown to be predictive of both multivessel disease and future events [28-30]. Quantitative analysis of lung thaUium-201 uptake, however, was not routinely performed at the University of Virginia during the period when data for this study were acquired, and is, therefore, not included in the analysis. It is likely that had-this variable been included in the analysis, it w o u l d h a v e p r o v i d e d u s e f u l a d d i t i o n a l information. Comparison with Previous Studies
That multivessel disease is best predicted by including clinical and exercise electrocardiographic variables with thallium-201 imaging data is in general agreement with previous reports. Maddahi et al [31] calculated changes in sensitivity and specificity for the identification of three-vessel and left main disease by combining exercise electrocardiographic and exercise thallium-201 variables. They
March1991 The American Journal of Medicine Volume 90
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL
found that the optimal identification of high-risk patients was obtained when quantitative thallium201 analysis was combined with the blood pressure and ST-segment response during exercise. McCarthy and colleagues [32] performed stepwise discriminant analysis to predict multivessel disease. Their analysis yielded a complex equation comprising nine clinical, electrocardiographic, and thaUiurn-201 variables. They, however, interpreted thallium-201 images using Fourier transformation, a method that is not widely practiced. Hung and coworkers [33] performed stepwise logistic regression analysis to predict multivessel disease. Similar to our results, they found the number of abnormal thallium-201 segments and ST depression to be useful predictors of multivessel disease. Unlike our results, however, they found that the type of chest pain (typical or not), exercise duration, and a history of hypertension were independent predictors of the severity of coronary artery disease. Their population, however, did not include women and was considerably younger, which m~y account for the differences in the results between their study and ours. The unique aspect of our study is that logistic regression derived from one sizeable population of 383 patients was tested in another unrelated population of 325 patients. Correct prediction for the test population is evidence that the prediction of the probability of multivessel disease is robust and may be extended to other similar populations. Limitations of This Study Both the University of Virginia and the Massachusetts General Hospital cohorts include select populations due to the tertiary nature of these hospitals and to the study requirement that the patients undergo cardiac catheterization. This is evident in the high rate of multivessel disease even in the lowest risk group without thallium-201 defects, ST depression, or older age. Although the multivariate analysis did not indicate gender as an important independent variable, the populations studied were dominantly male, and the prediction could be considered somewhat biased toward the male patient. Conclusions We hypothesized that despite the outstanding ability of thallium-201 imaging to diagnose coronary artery disease, clinical and electrocardiographic variables would provide additional independent diagnostic information. Our results substantiate this hypothesis. In addition, we have developed a statistical model to predict multivessel
disease. This model was found to be valid in another unrelated population. Our results imply that this approach may be useful in identifying patients with probable multivessel disease who could then be referred for cardiac catheterization with a view toward possible revascularization.
REFERENCES 1. The Veterans Administration Coronary Artery Bypass Surgery Cooperative Study Group. Eleven-year survival in the Veterans Administration randomized trial of coronary bypass surgery for stable angina. N Engl J Med 1984; 11: 13339. 2. European Coronary Surgery Study Group. Long term results of prospective randomized studyof coronary artery bypass surgery in stable angina pectoris. Lancet 1982: 2: 1173-80. 3. Reeves TJ, Oberman A, Jones WB, Sheffield LT. Natural history of angina pectoris. Am J Cardiol 1974; 33: 423-30. 4. Humphries JO, Kuller L, Ross RS, Freisinger GC, Page EE. Natural history of ischemic heart diseasein relation to angiographic findings. Circulation 1974; 49: 489-97. 5. BrushkeAV, Proudfit WL, SonesF. Progress studies of 490 consecutive nonsurgical cases of coronary disease followed for 5-9 years. II. Ventriculographic and other correlations. Circulation 1973; 47: 1154-63. 6. 1989 Heart facts. Dallas: American Heart Association, 1988. 7. Kaul S. A look at 15 years of planar thallium-201 imaging. Am Heart J 1989; 118: 581-601. 8. Kaul S, Boucher CA, Newell JB, et al. Determination of the quantitative thallium imaging variables that optimize detection of coronary artery disease. J Am Coll Cardiol 1986; 7: 527-37. 9. Kaul S, Keiss MC, Liu P, et al. Comparison of exercise electrocardiography and quantitative thallium imaging for one-vessel coronary artery disease. Am J Cardiol 1985; 56: 257-61. 10. EsquivelL, Pollock SG, Belier GA, Gibson RS, Watson DD, KaulS. Effect of the degree of effort on the sensitivity of the exercise thallium-201 stress test in symptomatic coronary artery disease. Am J Cardiol 1989: 63: 160-5. 11. Berger B, Watson DD, Taylor GJ, et al. Quantitative thallium-201 exercise scintigraphy for detection of coronary artery disease. J Nucl Med 1981; 22: 585-93. 12. Kaul S, Newell JB, Chesler DA, et aL Value of computer analysis of exercise thallium imaging in the noninvasivedetection of coronary artery disease. JAMA 1986; 255: 508-I I. 13.0kada RD, Boucher CA, Strauss HW, Pohost GM. Exercise radionuclide imaging approaches to coronary artery disease. Am J Cardio11980; 46:1188204. 14. Pollock SG, Watson DD, Gibson RS, Belier GA, Watson DD, Kaul S. A simplifled approach to evaluating multiple test outcomes and multiple disease states in relation to the exercise thallium-201 stress test in suspected coronary artery disease. Am J Cardiol 1989; 64: 466-70. 15. Gitler B, Fishbach M, Steingart RM. Use of electrocardiography and thallium exercise testing in clinical practice. J Am Coll Cardiol 1984; 3: 262-71. 16. Kaul S, Lilly DR, GaschoJA, et al. Prognostic utility of the exercise thallium201 test in ambulatory patients with chest pain: comparisonwith cardiac catheterization. Circulation 1988; 77: 745-58. 17. Sheffield LT. Exercise stress testing. In: Braunwald E, ed. Heart disease: a text book of cardiovascular medicine. Philadelphia:WB Saunders, 1988: 22341. 18. AustenWG, EdwardsSE, FryeRL, et aL American HeartAssociation Committee Report. A reporting system on patients evaluated for coronary artery disease. Circulation 1975; 51: 7-40. 19. Walker SH, Duncan DB. Estimation of the probability of an event as a function of several independent variables. Biometrica 1967; 54: 167-79. 20. Mood AM, Graybill FA. Introduction to the theory of statistics. New York: McGraw-Hill, 1963. 21. Diamond GA, Forrester JS. Analysis of probability as an aid in the clinical diagnosis of coronary artery disease. N Engl J Med 1979; 300: 1350-8. 22. Udell JA, Ellested MH. Predictive implications of ventricular premature con-
March 1991
The American Journal of Medicine
Volume 90
351
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL tractions associated with treadmill stress testing. Circulation 1977; 56: 985-9. 23. Ellestad MH. Stress testing: principles and practice, 3rd ed. Philadelphia: FA Davis, 1983: 209. 24. Marieb MA, Belier GA, Gibson RS, Lerman BB, Kaul S. Clinical relevance of exercise-induced ventricular arrhythmias in suspected coronary artery disease. Am J Cardiol 1990; 66: 172-8. 25. Califf RM, McKinnis RA, McNeer JF, eta/. Prognostic value of ventricular arrhythmias associated with treadmill exercise testing in patients studied with cardiac catheterization for suspected ischemic heart disease. J Am Coil Cardiol 1983; 2: 1060-7. 26. Brown KA, Boucher LA, Okada RD, eta/. Prognostic value of exercise thallium-20] imaging in patients presenting for evaluation of chest pain. J Am Coil Cardiol 1984; t: 999-1001. 27. Ladenheim MC, Pollock SH, Rozanski A, eta/. Extent and severity of myocardial hypoperfusion as predictors of prognosis in patients with suspected coronary artery disease. J Am Coil Cardiol 1986; 7: 467-71. 28. Kaul S, Finkelstein DM, Homma S, Leavitt M, Okada RD, Boucher CA. Superiority of quantitative exercise thallium-201 variables in determining long-term
352
March 1991 The American Journal of Medicine Volume 90
prognosis in ambulatory patients with chest pain: a comparison with cardiac catheterization. J Am Coil Cardiol 1988; 12: 25-34. 29. Gill JB, Ruddy TD, Newell JB, Finkelstein DM, Strauss HW, Boucher CA. Prognostic importance of thallium uptake by the lungs during exercise in coronary artery disease. N Engl J Med 1987; 317: 1485-9. 30. Homma S, Kaul S, Boucher CA. Correlates of lung/heart ratio of thallium201 in coronary artery disease. J Nucl Med 1987; 28: 1531-5. 31. Maddahi J, Abdulla A, Garcia EV, Swan HJC, Berman DS. Noninvasive identification of left main and triple vessel coronary artery disease: improved accuracy using quantitative analysis of regional myocardial stress distribution and washout of thallium-201. J Am Coil Cardiol 1986; 7: 53-60. 32. McCarthy DM, Sciacca RR, Blood DK, Canon P. Discriminate function analysis using thallium-20l scintigraphy and exercise stress test variables to predict the presence and extent of coronary artery disease. Am J Cardiol 1982; 49: 1917-26. 33. Hung J, Chaitman BR, Lain J, et al. A logistic regression analysis of multiple noninvasive tests for the prediction of the presence and extent of coronary artery disease in men. Am Heart J 1985; 110: 460-9.
M.D., ROBERT
D. ABBOTT,
Ph.D., Charlottsville, Virginia,
CHARLES A. BOUCHER, M.D., Boston, Massachusetts, DENNY D. WATSON, Ph.D., SANJIV KAUL, M.D., Charlottsville, Virginia
PURPOSE: The aim of this study was to (1) determine w h e t h e r nonlmnging variables add to the diagnostic information available from exercise thallinm-201 images for the detection of multivessel coro-Ary a r t e r y disease; and (2) to develop a model based on the exercise thallium201 stress test to predict the presence of multivessel disease. P A T I E N T S A N D METHODS: The study populations included 383 patients referred to the University of Virginia and 325 patients referred to the Massachusetts General Hospital for evaluation of chest pain. All patients underwent both cardiac catheterization and exercise thallium-201 stress testing between 1978 and 1981. RESULTS: In the University of Virginia cohort, at each level of thallium-201 abnormality (no defects, one defect, more than one defect), ST depression and patient age added significantly in the detection of multivessel disease. Logistic regression analysis using data from these patients identified three independent predictors of multivessel disease: initial thallium-201 defects, ST depression, and age. A model was developed to predict multivessel disease based on these variables. As might be expected, the risk of multivessel disease predicted by the model was similar to that actually observed in the University of Virginia population. More importantly, however, the model was accurate in predicting the occurrence of multivessel disease in the unrelated population studied at the Massachusetts General Hospital. CONCLUSION: It is, therefore, concluded that (1) nonlmaging variables (age and exercise-induced From the Divisionsof Cardiologyand Biostatistics, Departmentof Medicine, University of Virginia School of Medicine. Charlottesville,Virginia, and the Cardiac Unit. Department of Medicine, MassachusettsGeneral Hospital, Boston, Massachusetts. This work was supportedin part by grants (RO1-HL-26205and ROI-HL262].5) from the National Institutes of Health, Bethesda, Maryland. Dr. Kaul is the recipient of the Clinical Investigator (KO8-HL-01833) and FIRST(R29-HL-38345) Awardsof the NationalInstitutes of Health. Requestsfor reprints shouldbe addressedto SanjivKaul, M.D., Department of Medicine, Box ]58, University of Virginia School of Medicine, Charlottesville, Virginia 22908. Manuscript submitted July 6. 1990, and accepted in revised form November 19. 1990.
ST depression) add independent information to thalllum-201 imaging data in the detection of multivessel disease; and (2) a model ha~ been developed based on the exercise thAnium-201 stress test that can accurately predict the probability of multivessel disease in other populations.
here are two principal aspects to the manage-
T ment of patients with coronary artery disease. The first is treatment of symptoms and the second is prolongation of life. Although patients with single-vessel disease may undergo a revascularization procedure (bypass surgery or balloon angioplasty) for control of intractable symptoms, it is patients with multivessel disease who benefit most from revascularization in terms of prolongation of life [1,2]. It is also such patients who, if treated medically, experience adverse cardiac events [3-5]. Classification of patients with symptomatic coronary artery disease into those with and those without multivessel disease is, therefore, important for optimal patient management. Since an estimated 5 million people in the United States have symptomatic coronary artery disease [6], identification of patients with multivessel disease by cardiac catheterization would be logistically burdensome. A more feasible approach is to first select patients likely to have multivessel disease by noninvasive tests, and then p e r f o r m c o r o n a r y a n g i o g r a p h y to g u i d e revascularization. Exercise thallium-201 imaging is, arguably, one of the more reliable noninvasive methods for the detection of coronary artery disease. Its sensitivity, using quantitative analysis, is close to 90% and its specificity is close to 80% [7], with most of the falsenegative results occurring in patients with less severe disease [8]. Because exercise electrocardiography is less sensitive than thallium-201 imaging [913], its additional value in the diagnosis of coronary artery disease, compared with thallium-201 imaging alone, is questionable. Whereas some studies have demonstrated that the results of exercise electrocardiography are additive to those of thallium201 imaging, the specific manner in which this is 8o has not been defined [12,14]. Furthermore, alMarch 1991
The American Journal of Medicine Volume 90
345
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL Exercise E l e c t r o c a r d i o g r a p h y
TABLE I Characteristics of the Two Patient Cohorts
Variable
UVA (n = 383)
MGH (n = 325)
Age Males No CAD or SVD
58 • 10 326 (85%) 165 (43%)
53 4- 10 272 (84%) 131 (40%)
Multivessel CAD
218 (57%)
194 (60%)
p Value 0.001 0.60 0.49
UVA= Universityof Virginia; MGH = MassachusettsGeneralHospital:CAD= coronar artery disease;SVD= single-vesseldisease.
though clinical variables, such as age, gender, and type of chest pain, have also been shown to add marginally to the imaging data for predicting the presence or absence of coronary artery disease [12,15], their value in predicting multivessel disease is unknown. Since multivessel disease is one end of the spectrum of coronary artery disease rather than a separate disease entity, it stands to reason that the same factors predictive of any coronary artery disease would be predictive of multivessel disease. We hypothesized, therefore, that the noninvasive diagnosis of multivessel disease could be optimized by combining thallium-201 imaging variables with those of exercise electrocardiography and clinical history. We evaluated this hypothesis in a population of 383 patients referred to the University of Virginia Medical Center for evaluation of chest pain. We derived a model to predict the probability of multivessel disease based on this patient cohort. We then tested the validity of this model by applying it to a similar but unrelated population of 325 patients evaluated at the Massachusetts~ General Hospital. PATIENTS
AND METHODS
The patients and methods from both the University of Virginia and Massachusetts General Hospital have been described in detail in previous reports [8,16]. A limited summary of patient characteristics and methods is provided here.
All patients underwent symptom-limited exercise testing terminated for fatigue, claudication, angina, dyspnea, hypotension, or ventricular tachycardia [8,16]. The electrocardiogram, heart rate, and blood pressure were recorded every minute during baseline, exercise, and the 5-minute recovery period. All test results were classified as either positive or negative. When the baseline electrocardiogram was normal, a test was considered positive if it had 1 mm of flat or downsloping ST depression, or 2 mm of upsloping ST depression measured 0.08 msec after the J point [17]. In the presence of baseline ST segment depression, but in the absence of left bundle branch block, left ventricular hypertrophy, or digitalis use, the test result was considered positive if there was an additional 2 mm of ST depression. Any other test result was considered negative. Nondiagnostic test results due to left bundle branch block, left ventricular hypertrophy, or digitalis use were interpreted as negative.
Exercise Thallium-201 Imaging At peak exercise, 1.5 to 2.0 mCi of thallium-201 was injected intravenously and the patient was encouraged to exercise for another 30 to 60 seconds. Images were obtained in the anterior and 45 ° and 70 ° left anterior oblique views. In the University of Virginia cohort, quantitative analysis was performed in seven segments in the anterior and 45 ° left anterior oblique views at the time when these data were collected [11]. Although the quantitative analysis at the Massachusetts General Hospital was performed in 15 segments from all three views [8], for this study only the seven segments corresponding to those quantitated at the University of Virginia were analyzed. The quantitative lung/heart ratio of thallium-201 was not analyzed since it was not routinely performed in the patients studied at the University of Virginia when data for this study were collected. Since the purpose of this study was to develop a model for the detection of multivessel disease; rather than prognosis, only initial defects were analyzed; redistribution was not included in the analysis.
Patient Population Both cohorts were comprised of patients referred for evaluation of chest pain who underwent both exercise thallium-201 imaging and cardiac catheterization within 3 months of each other between 1978 and 1981. Although the patients from the Massachusetts General Hospital were younger, there were no significant differences between the two cohorts in terms of age and the incidence of multivessel disease (Table I).
346
Cardiac Catheterization
The coronary angiographic data were analyzed in a blinded manner. They were analyzed by one experienced observer at the University of Virginia by means of calipers and a ruler. At the Massachusetts General Hospital, the data were subjectively analyzed by three observers and a consensus opinion was obtained. Multivessel disease was defined as stenosis (50% or greater luminal diameter narrow-
March 1991 The American Journal of Medicine Volume 90
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL
ing) of at least two of the three main arteries or their major branches [18]. The major branch of the left anterior descending artery was the diagonal artery. When present, an intermedius artery was included as a branch of the left anterior descending artery. The major branch of the left circumflex artery was the obtuse marginal artery and the major branch of the right coronary artery was the posterior descending artery. In left dominant systems, the posterior descending artery was considered a major branch of the left circumflex artery. Patients with coronary artery disease involving less than two major arteries or their major branches or showing no significant stenosis were coded as not having multivessel disease. Left main stenosis was considered equivalent to multivessel disease. Statistical Analysis
Logistic regression models were used to help evaluate the performance of findings from exercise electrocardiography, exercise thallium-201 imaging, and clinical risk factors in predicting the presence or absence of multivessel disease [19]. Tests of significance were based on likelihood ratio statistics [20]. All tests of significance were two-sided. Estimates of odds ratios were derived from the regression coefficients (also based on maximum likelihood calculations) by comparing meaningful groups (the relative odds of multivessel disease in men versus women for instance). Ninety-five percent confidence intervals for the odds ratios were also included. To evaluate the model's fit, observed rates of multivessel disease were compared with predicted rates. Predicted rates were calculated by using a logistic model to estimate the probability of multivessel disease for each patient. These rates were calculated from this model by ranking the individual probabilities. The ranked probabilities were then grouped into quintiles where patients who fell in the lowest quintile were predicted to have the lowest rate of multivessel disease, whereas patients in the highest quintile were predicted to have the highest rate. Within each quintile, the predicted rate of multivessel disease was determined by averaging the individual probabilities. The predicted values were then compared with the actual rates of multivessel disease using a chi-square goodness-offit test. To further evaluate the model's performance, the logistic model derived from the University of Virginia data was used to predict multivessel disease in the patients from the Massachusetts General Hospital. As with the University of Virginia patients, those from the Massachusetts General Hospital
TABLE II Description of Patients from the University of Virginia Cohort Based upon the Number of Diseased Vessels Variable Males >58 years Typical angina >_2 segments with thallium-201 defects ST depression on EKG Exercise-induced VA Abnormal BP response* <5 METSworkload Exercise-induced angina
NoCAD(%) (n = 93)
SVD(%) (n = 72)
MVD(%) (n = 218)
68 (73) 37 (40) 39 (42) 14 (15)
63 (88) 32 (44) 50 (69) 42 (65)
195 (89) 141 (65) 160 (73) 163 (74)
21 (23) 26 (28) 12 (13) 13 (14) 13 (14)
28 (39) 38 (53) 15 (21) 17 (24) 13 (18)
135 (62) 98 (45) 48 (22) 64 (29) 50 (23)
BP = bloodpressure:CAD= coronaryartery disease:EY,G = electrocardiogram:MVO= multivesseldisease:SVD= single-vesseldisease:VA = ventriculararrhythmias. ° Definedas fall or failureto riseof systolicbloodpressureduringexercise.
were also ranked and grouped into quintiles based on patient-specific risk factors. The predicted rates of multivessel disease in each quintile were then compared with the actual rates using a chi-square goodness-of-fit test. RESULTS Univariate Analysis
The distribution of coronary artery disease in relation to clinical, exercise electrocardiographic, and exercise thallium-201 variables in the University of Virginia population is displayed in Table II. Except for exercise-induced ventricular arrhythmias, there is a progressive rise in the frequency of occurrence of every variable from no coronary artery disease, to single-vessel coronary artery disease, to multivessel disease. Nevertheless, the frequency of exercise-induced ventricular arrhythmias is higher in patients with coronary artery disease than in those without this condition. Patients over 58 years in age constitute fewer than half of those with no coronary artery disease and single-vessel coronary artery disease, but account for the majority of those with multivessel disease. Patients with more than one thallium defect are uncommon in the no coronary artery disease group, b u t make up approximately three fourths of the multivessel disease group. The occurrence of ST depression on the exercise electrocardiogram increases steadily among the groups. Approximately one fifth of the patients without coronary artery disease have ST depression compared with two fifths and three fifths with singlevessel and multivessel disease, respectively. The odds ratios (with their 95% confidence intervals) derived from univariate logistic analysis are shown in T a b l e III. Women have half the odds of multivessel disease relative to men (p <0.01), while
March 1991 The American Journal of Medicine Volume 90
347
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL
TABLE III
Logistic Regression Analysis with Multivessel Disease Predicted by Each Risk Factor
Risk Factor Gender Age (years) Type of chest pain METS achieved Abnormal BP response ST depression on EKG Exercise-induced angina Thallium-201 defects Ventricular arrhythmias
Comparison
Odds Ratio
Female versus male 20-year difference Typical versus atypical 5 MET difference Yes versus no Yes versus no Yes versus no Two-segment difference Yes versus no
0.5 = 2.3 t 2.41 0.St 1.5 3.91 1.6 3.Or 1.3
UnivariateAnalysis 95% Confidence Interval 0.3-0.8 1.5-3.5 1.5-3.6 0.4-0.8 0.9-2.5 2.5-5.9 0.9-2.7 2.2-4.1 0.9-2.0
Odds Ratio
Multivariate Analysis 95% Confidence Interval
0.6 1.8~ 1.5 0.6 0.9 3.5I I. 1 2.51 0.8
0.3-1.1 1.1-2.4 0.9-2.6 0.3-1.1 O.5-1.7 2.1-5.6 0.6-2.1 1.8-3.6 0.5-1.3
ender = l-female. O-male:type of chest pain = l-typical. O-nottypical: abnormalBP (blood pressure)responseto exercise = l-yes, O-no:ST depression on the EKG(electrocardioram) during exercise = 1-yes. O-no:exercise-inducedangina = l-yes; O-no:exercise-inducedventricular arrhythmias = l-yes. O-no. = Women had significantly less multivesseldisease(p
older pptients have double the odds of this condition compared with those 20 years younger (p <0.001). Patients with typical angina pectoris have twice the odds of multivessel disease in contrast to those with atypical angina (p <0.001). A 5 MET difference in work load attained during exercise doubles the odds of multivessel disease for those who performed less work (p <0.05). The occurrence of ST depression on exercise is associated with a nearly fourfold excess of the odds of multivessel disease as compared with its absence (p <0.001). The odds of having multivessel disease triples in the presence of two segmental defects on initial thallium-201 images (p <0.001). The occurrence of angina, ventricular arrhythmias, or an abnormal blood pressure response during exercise fails to raise the odds of multivessel disease in a significant manner. Multivariate Analysis
The results of the multivariate analysis, with the presence or absence of multivessel disease as the dependent variable, are also shown in Table III. After controlling for all variables, the occurrence of initial thallium-201 defects and exercise-induced ST depression on the electrocardiogram remain significantly and strongly associated with multivessel disease (p <0.001). The odds of multivessel disease increases 2.5 times for patients with two thallium201 defects compared with those having one or no defects. In the presence of ST depression, the odds of multivessel disease more than triples. Age is the only other covariate that remains a significant predictor of multivessel disease (p <0.05) on multivariate analysis, although its association with multivessel d i s e a s e is w e a k e r t h a n t h a t of i n i t i a l thallium-201 defects and ST depression. The joint effects of age, ST depression, and the number of initial thallium-201 defects are depicted
348
in F i g u r e 1. The observed rate of multivessel disease rises as the number of initial thallium-201 defects increases, regardless of age and the presence of ST depression. The lowest rate of multivessel disease, however, is seen in patients younger than 58 years who do not exhibit initial thallium-201 defects or ST depression. In this patient subset, the occurrence of multivessel disease is unlikely. The highest rate for multivessel disease is noted in patients older than 58 years who exhibit multiple initial thallium-201 defects and ST depression. In this subset of patients, the occurrence of multivessel disease is highly likely. Model to Predict Multivessel Disease
Based on the results in Table III, age, ST depression, and the number of initial thallium-201 defects were chosen to estimate the probability of multivessel disease in a logistic regression model. The resultant coefficients and intercept yielded the following expression for the estimated probability of multivessel disease:
March 1991 The American Journal of Medicine Volume 90
p(MVD) -- 1/(1 + exp[3.09 - 0.032*age - 1.28"ST - 0.507*defects]) where M~VD -- multivessel disease, age = age in years, ST = one for an abnormal ST-segment response during exercise and zero otherwise, and defects -- number of defects (range: zero to seven) on initial thallium-201 images. To examine the model's goodness of fit, the observed and predicted rates of multivessel disease in the University of Virginia cohort were compared. As illustrated in F i g u r e 2, there were no differences between the observed and predicted rates of multivessel disease in each quintile (p = 0.97). This finding is not unexpected since the model was derived from the identical population used to examine
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL
Figure 1. The additive value of including information regarding age (either less than 58 years or 58 years or older) and presence or absence of exercise-induced ST depression on the electrocardiogram (x-axis) to each level of abnormality on initial thallium-201 imaging (no defect, one defect, more than one defect) (z-axis) on the detection of c o r o n a r y a r t e r y disease (y-axis). See text for details,
/
/
/
~= =
100
L
80 6o
~_ ~ ~
" >1
4o
I
20
)EFECT
0 >
58
yrs
<
~. ST d e p
58
yrs
+ ST dep
Observed
.¢ uJ e~
Figure 2. The rate of multivessel disease in each quintile as predicted by the model derived using logistic regression analysis from the U n i v e r s i t y of Virginia p o p u l a t i o n and the actual rate of multivessel disease as noted on coronary angiography in the same patients. See text for details.
u) u)
DEFECT
DEFECT
> -
58
yrl - S T dep
yrs
<
ST d e p
~
58
Predicted
100 ]
80 60 40
m
I-
=i ~e
20 0 1
2
it. To examine the model's utility, it was applied to the unrelated patient cohort studied at the Massachusetts General Hospital. The observed and predicted rates of multivessel disease in this population are illustrated in F i g u r e 3 by quintile. Similar to the data in Figure 2, there were no differences between the observed and predicted rates of multivessel disease in this patient population in each quintile (p = 0.91). COMMENTS
In the current study, we hypothesized that despite the superior ability of exercise thallium-201 imaging to diagnose coronary artery disease, clinical and exercise electrocardiographic variables would provide additional independent diagnostic information for the detection of multivessel disease. Through logistic regression analysis, we demonstrated this to be true and developed a model to predict multivessel disease using selected thallium201, exercise electrocardiographic, and clinical vari-
3 4 QUINTILE OF RISK
5
ables. The model was found not only to be accurate in the population from which it was derived, but it was also valid when applied to an unrelated, but largely similar, patient population. Value of Nonimaging Variables in Predicting Multivessel Disease
Since exercise electrocardiography is less sensitive than thallium-201 imaging [9-13], its additional value in determining the presence or absence of coronary artery disease compared with thallium201 imaging alone is not clear. For example, if the thallium-201 test is positive and the exercise electrocardiogram is negative, one usually assumes the latter to be a false-negative result. Similarly, if the thallium-201 test is negative and the exercise electrocardiogram is positive, one might consider the latter to represent a false-positive result. What does it mean, however, when both tests provide concordant information? Our results indicate that when both tests are negative in a patient under 58 years of
March 1991 The American Journal of Medicine Volume 90
349
PREDICTION OF MULTIVESSELDISEASEFROMTHE EXERCISETHALLIUM-201STRESSTEST / POLLOCKET AL
Observed
< uJ u)
~
Predicted
lOO 80
,,-I,
60
~ r~
4o
~
0
2o 1
2
3 QUINTILE
4 OF RISK
age, multivessel disease is unlikely. Similarly, when both tests are positive in such an individual, multivessel disease is highly likely. In the latter situation, the probability of multivessel disease is high even if multiple thaUium-201 abnormalities are not seen. This information is of obvious clinical utility. The results from logistic regression analysis demonstrate that age, number of initial thallium-201 defects, and the presence of exercise-induced ST depression are independent predictors of multivessel disease. When these factors are used jointly, it is possible to enhance the prediction of multivessel disease. Since age, ST depression, and number of thallium-201 defects are known to be useful in predicting the presence or absence of coronary artery disease [12,14], their predictive value for the occurrence of multivessel disease is not unexpec~d. The importance of these variables as predictors of multivessel disease can also be regarded from a bayesian point of view where age provides a pretest probability that is modified by the results of exercise electrocardiography and thallium-201 imaging [14,21]. The finding that only age, ST depression, and number of initial thallium-201 defects are independent predictors of multivessel disease should not be construed as meaning that the other variables are not useful. From univariate analysis, both gender and type of chest pain (typical or not) were, as would be expected [21], also significant predictors of multivessel disease. Multivessel disease was also significantly associated, on univariate analysis, with exercise-induced ventricular arrhythmias and poor exercise performance, findings that have been previously shown to be associated with more severe disease [22-25]. In the current study, however, all
350
Figure 3. The predicted rate of multivessel disease in each quintile of the Massachusetts General Hospital population based upon the model derived from the University of Virginia population and the observed rate of multivessel disease in the same population as noted on coronary angiography. See text for details.
these factors were not significant univariate predictors of multivessel disease because they were superseded by the overriding influence of age, ST depression, and number of initial thallium-201 defects. Value of Other Thallium-201 Variables in Predicting Multivessel Disease
In this analysis, we did not include redistribution on delayed thallium-201 images or increased lung uptake of thallium-201 to determine the probability of multivessel disease. Redistribution on the delayed thallium-201 images reflects reversible ischemia and myocardial viability [7]. Although useful in assessing the risk of future ischemic events [16,26-29], it probably does not have a role in assessing the extent of coronary artery disease. Abnormal lung uptake of thallium-201 has been shown to be predictive of both multivessel disease and future events [28-30]. Quantitative analysis of lung thaUium-201 uptake, however, was not routinely performed at the University of Virginia during the period when data for this study were acquired, and is, therefore, not included in the analysis. It is likely that had-this variable been included in the analysis, it w o u l d h a v e p r o v i d e d u s e f u l a d d i t i o n a l information. Comparison with Previous Studies
That multivessel disease is best predicted by including clinical and exercise electrocardiographic variables with thallium-201 imaging data is in general agreement with previous reports. Maddahi et al [31] calculated changes in sensitivity and specificity for the identification of three-vessel and left main disease by combining exercise electrocardiographic and exercise thallium-201 variables. They
March1991 The American Journal of Medicine Volume 90
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL
found that the optimal identification of high-risk patients was obtained when quantitative thallium201 analysis was combined with the blood pressure and ST-segment response during exercise. McCarthy and colleagues [32] performed stepwise discriminant analysis to predict multivessel disease. Their analysis yielded a complex equation comprising nine clinical, electrocardiographic, and thaUiurn-201 variables. They, however, interpreted thallium-201 images using Fourier transformation, a method that is not widely practiced. Hung and coworkers [33] performed stepwise logistic regression analysis to predict multivessel disease. Similar to our results, they found the number of abnormal thallium-201 segments and ST depression to be useful predictors of multivessel disease. Unlike our results, however, they found that the type of chest pain (typical or not), exercise duration, and a history of hypertension were independent predictors of the severity of coronary artery disease. Their population, however, did not include women and was considerably younger, which m~y account for the differences in the results between their study and ours. The unique aspect of our study is that logistic regression derived from one sizeable population of 383 patients was tested in another unrelated population of 325 patients. Correct prediction for the test population is evidence that the prediction of the probability of multivessel disease is robust and may be extended to other similar populations. Limitations of This Study Both the University of Virginia and the Massachusetts General Hospital cohorts include select populations due to the tertiary nature of these hospitals and to the study requirement that the patients undergo cardiac catheterization. This is evident in the high rate of multivessel disease even in the lowest risk group without thallium-201 defects, ST depression, or older age. Although the multivariate analysis did not indicate gender as an important independent variable, the populations studied were dominantly male, and the prediction could be considered somewhat biased toward the male patient. Conclusions We hypothesized that despite the outstanding ability of thallium-201 imaging to diagnose coronary artery disease, clinical and electrocardiographic variables would provide additional independent diagnostic information. Our results substantiate this hypothesis. In addition, we have developed a statistical model to predict multivessel
disease. This model was found to be valid in another unrelated population. Our results imply that this approach may be useful in identifying patients with probable multivessel disease who could then be referred for cardiac catheterization with a view toward possible revascularization.
REFERENCES 1. The Veterans Administration Coronary Artery Bypass Surgery Cooperative Study Group. Eleven-year survival in the Veterans Administration randomized trial of coronary bypass surgery for stable angina. N Engl J Med 1984; 11: 13339. 2. European Coronary Surgery Study Group. Long term results of prospective randomized studyof coronary artery bypass surgery in stable angina pectoris. Lancet 1982: 2: 1173-80. 3. Reeves TJ, Oberman A, Jones WB, Sheffield LT. Natural history of angina pectoris. Am J Cardiol 1974; 33: 423-30. 4. Humphries JO, Kuller L, Ross RS, Freisinger GC, Page EE. Natural history of ischemic heart diseasein relation to angiographic findings. Circulation 1974; 49: 489-97. 5. BrushkeAV, Proudfit WL, SonesF. Progress studies of 490 consecutive nonsurgical cases of coronary disease followed for 5-9 years. II. Ventriculographic and other correlations. Circulation 1973; 47: 1154-63. 6. 1989 Heart facts. Dallas: American Heart Association, 1988. 7. Kaul S. A look at 15 years of planar thallium-201 imaging. Am Heart J 1989; 118: 581-601. 8. Kaul S, Boucher CA, Newell JB, et al. Determination of the quantitative thallium imaging variables that optimize detection of coronary artery disease. J Am Coll Cardiol 1986; 7: 527-37. 9. Kaul S, Keiss MC, Liu P, et al. Comparison of exercise electrocardiography and quantitative thallium imaging for one-vessel coronary artery disease. Am J Cardiol 1985; 56: 257-61. 10. EsquivelL, Pollock SG, Belier GA, Gibson RS, Watson DD, KaulS. Effect of the degree of effort on the sensitivity of the exercise thallium-201 stress test in symptomatic coronary artery disease. Am J Cardiol 1989: 63: 160-5. 11. Berger B, Watson DD, Taylor GJ, et al. Quantitative thallium-201 exercise scintigraphy for detection of coronary artery disease. J Nucl Med 1981; 22: 585-93. 12. Kaul S, Newell JB, Chesler DA, et aL Value of computer analysis of exercise thallium imaging in the noninvasivedetection of coronary artery disease. JAMA 1986; 255: 508-I I. 13.0kada RD, Boucher CA, Strauss HW, Pohost GM. Exercise radionuclide imaging approaches to coronary artery disease. Am J Cardio11980; 46:1188204. 14. Pollock SG, Watson DD, Gibson RS, Belier GA, Watson DD, Kaul S. A simplifled approach to evaluating multiple test outcomes and multiple disease states in relation to the exercise thallium-201 stress test in suspected coronary artery disease. Am J Cardiol 1989; 64: 466-70. 15. Gitler B, Fishbach M, Steingart RM. Use of electrocardiography and thallium exercise testing in clinical practice. J Am Coll Cardiol 1984; 3: 262-71. 16. Kaul S, Lilly DR, GaschoJA, et al. Prognostic utility of the exercise thallium201 test in ambulatory patients with chest pain: comparisonwith cardiac catheterization. Circulation 1988; 77: 745-58. 17. Sheffield LT. Exercise stress testing. In: Braunwald E, ed. Heart disease: a text book of cardiovascular medicine. Philadelphia:WB Saunders, 1988: 22341. 18. AustenWG, EdwardsSE, FryeRL, et aL American HeartAssociation Committee Report. A reporting system on patients evaluated for coronary artery disease. Circulation 1975; 51: 7-40. 19. Walker SH, Duncan DB. Estimation of the probability of an event as a function of several independent variables. Biometrica 1967; 54: 167-79. 20. Mood AM, Graybill FA. Introduction to the theory of statistics. New York: McGraw-Hill, 1963. 21. Diamond GA, Forrester JS. Analysis of probability as an aid in the clinical diagnosis of coronary artery disease. N Engl J Med 1979; 300: 1350-8. 22. Udell JA, Ellested MH. Predictive implications of ventricular premature con-
March 1991
The American Journal of Medicine
Volume 90
351
PREDICTION OF MULTIVESSEL DISEASE FROM THE EXERCISE THALLIUM-201 STRESS TEST / POLLOCK ET AL tractions associated with treadmill stress testing. Circulation 1977; 56: 985-9. 23. Ellestad MH. Stress testing: principles and practice, 3rd ed. Philadelphia: FA Davis, 1983: 209. 24. Marieb MA, Belier GA, Gibson RS, Lerman BB, Kaul S. Clinical relevance of exercise-induced ventricular arrhythmias in suspected coronary artery disease. Am J Cardiol 1990; 66: 172-8. 25. Califf RM, McKinnis RA, McNeer JF, eta/. Prognostic value of ventricular arrhythmias associated with treadmill exercise testing in patients studied with cardiac catheterization for suspected ischemic heart disease. J Am Coil Cardiol 1983; 2: 1060-7. 26. Brown KA, Boucher LA, Okada RD, eta/. Prognostic value of exercise thallium-20] imaging in patients presenting for evaluation of chest pain. J Am Coil Cardiol 1984; t: 999-1001. 27. Ladenheim MC, Pollock SH, Rozanski A, eta/. Extent and severity of myocardial hypoperfusion as predictors of prognosis in patients with suspected coronary artery disease. J Am Coil Cardiol 1986; 7: 467-71. 28. Kaul S, Finkelstein DM, Homma S, Leavitt M, Okada RD, Boucher CA. Superiority of quantitative exercise thallium-201 variables in determining long-term
352
March 1991 The American Journal of Medicine Volume 90
prognosis in ambulatory patients with chest pain: a comparison with cardiac catheterization. J Am Coil Cardiol 1988; 12: 25-34. 29. Gill JB, Ruddy TD, Newell JB, Finkelstein DM, Strauss HW, Boucher CA. Prognostic importance of thallium uptake by the lungs during exercise in coronary artery disease. N Engl J Med 1987; 317: 1485-9. 30. Homma S, Kaul S, Boucher CA. Correlates of lung/heart ratio of thallium201 in coronary artery disease. J Nucl Med 1987; 28: 1531-5. 31. Maddahi J, Abdulla A, Garcia EV, Swan HJC, Berman DS. Noninvasive identification of left main and triple vessel coronary artery disease: improved accuracy using quantitative analysis of regional myocardial stress distribution and washout of thallium-201. J Am Coil Cardiol 1986; 7: 53-60. 32. McCarthy DM, Sciacca RR, Blood DK, Canon P. Discriminate function analysis using thallium-20l scintigraphy and exercise stress test variables to predict the presence and extent of coronary artery disease. Am J Cardiol 1982; 49: 1917-26. 33. Hung J, Chaitman BR, Lain J, et al. A logistic regression analysis of multiple noninvasive tests for the prediction of the presence and extent of coronary artery disease in men. Am Heart J 1985; 110: 460-9.