Prognostic Indicators from Radionuclide Angiography in Medically Treated Patients with Coronary Artery Disease

DAVID B. PRYOR, MD, FRANK E. HARRELL, Jr., PhD, KERRY L. LEE, PhD,

ROBERT A. ROSATI, MD, R. EDWARD COLEMAN, MD, FREDERICK R. COBB, MD,

ROBERT M. CALIFF, MD, and ROBERT H. JONES, MD

The purpose of this investigation was to determine which variables obtained when performing radio- nuclide angiography predict subsequent survival or total events (cardiovascular death or nonfatal myocardial infarction) in stable patients with symptomatic coronary artery disease (CAD). Uni- variable and multivariable analyses of 6 variables, including ejection fraction (EF) at rest and exercise, change in EF with exercise, development of ische- mic chest pain or electrocardiographic changes, left ventricular (LV) wall motion abnormalities and ex- ercise time were examined in 366 patients followed up to 4.5 years. Univariate analyses revealed that the exercise EF was the variable most closely as-

sociated with future events (p <O.Ol), followed by EF at rest, wall motion abnormalities and exercise time. Multivariable analyses revealed that once the exercise EF was known, no other radionuclide variables contributed independent information about the likelihood of future events. Multivariable anal- yses also revealed that the exercise EF describes much of the prognostic information of coronary anatomy. Our ftndings suggest that the radionuclide angiogram is useful in predicting future events in patients with stable CAD, although examination in conjunction with other clinical descriptors will be necessary to further quantify this contribution.

(Am J Cardiol 1984;53:18-22)

The value of rest and exercise radionuclide angiography (RNA) in predicing outcome in patients with stable coronary artery disease (CAD) is unknown. Previous studies in patients with significant CAD have shown that left ventricular (LV) function is a primary deter- minant of prognosis.1-3 A close correlation exists be- tween measurements of LV function obtained with cardiac catheterization and with RNA. Therefore, RNA would be expected to be helpful in predicitng future cardiovascular events in patients with CAD. This in- vestigation determines which variables obtained with

From the Departments of Medicine, Community and Family Medicine, and Surgery, Duke University Medical Center, Durham, North Carolina. This work was supported by Research Grant HS- 17670 from the Na- tional Heart, Lung, and Blood Institute, Bethesda, Maryland; Research Grants HS-03834 and HS-04873 from the National Center for Health Services Research, OASH, Hyattsville, Maryland: Training Grant LM- 07003 and Grants LM-03373 and LM-00042 from the National Library of Medicine, Bethesda, Maryland; and grants from the Prudential In- surance Company of America, Newark, New Jersey; the Kaiser Family Foundation, Palo Alto, California, and the Andrew W. Mellon Foundation, New York, New York. Manuscript received May 31, 1983; revised manuscript received September 12, 1983, accepted September 22, 1983.

Address for reprints: David B. Pryor, MD, Box 3531, Duke University Medical Center, Durham, North Carolina 27710.

RNA are associated with future cardiovascular deaths or nonfatal myocardial infarction (MI).

Methods

Our patient population and information system have been described in detail.ld4 This study includes 386 consecutive (except as will be noted), medically treated (without previous cardiac surgery) symptomatic patients who underwent cardiac catheterization between January 1977 and May 1982. Only patients with significant CAD (275% luminal diameter nar- rowing of a major coronary artery5) who had RNA within 3 months of cardiac catheterization were included. In our population, patients studied by RNA demonstrate charac- teristics similar to those of the total population undergoing catheterization except for a lower incidence of class IV angina, progressive angina and a history of congestive heart failure.” Patients with coexistent valvular heart disease including moderate or severe mitral regurgitation were excluded. Baseline characteristics (clinical and catheterization data) of the study population are shown in Table I.

Selection of medical therapy was made by the patient, usually on recommendation of the patient’s physician. Criteria used to recommend therapy varied with individual clinicians but included consideration of the patient’s pain pattern, coronary anatomy, ventricular function, suitability for oper- ation and general condition. Medical therapy was individu-

January 1, 1984 THE AMERICAN JOURNAL OF CARDIOLOGY Volume 53 19

TABLE I Population Characteristics (388 Patients)

%

Age (median) 52 yr Sex (male) 81 History of

Class 3 or 4 congestive failure Previous MI 5: Peripheral vascular disease Progressive chest pain 357 Typical angina

ECG Q waves 5: Cardiomegaly (chest x-ray) 9 Coronary anatomy

l-vessel 2-vessel z; 3-vessel 30 Left main 1

Left ventricular contraction Normal Mild or moderate abnormalities :: Diffusely abnormal 9

l All values are percentages except where noted. Ml = myocardial infarction.

alized for each patient but usually consisted of nitrates, p blockers and, more recently, calcium antagonists in a combi- nation and dosage adequate to relieve angina without inducing intolerable side effects.

The methods of performing RNA have been described.6-g Briefly, studies at rest and peak exercise were performed in the erect position with the subject seated on a bicycle er- gometer (Fitron, Luminex Inc.). For both rest and exercise studies, 15 mCi of technetium-99m pertechnetate were in- jected as a bolus, and counts were obtained at 25-ms intervals for a 30-second period using a multicrystal gamma camera (System Seventy-Seven, Baird Inc.). Exercise was begun at a work load of 200 kpm/min and increased by lOO-kpm in- crements every minute while the blood pressure and ECG were monitored. Exercise was continued until one of the fol- lowing end points: 85% of the maximal predicted heart rate was achieved; pain or electrocardiographic changes suggestive of ischemia occurred; or other symptoms precluding further exercise developed such as fatigue and dyspnea. All RNA data were processed using software developed at this institu- tion.6-g

Follow-up methods including verification of the detection of nonfatal MI have been described.‘-:$ Prospective follow-up was obtained at 6 months, 1 year and annually thereafter. The outcome events of interest included cardiovascular death (including death that occurred suddenly or was due to MI, congestive heart failure or cerebrovascular events) and total cardiac events (cardiovascular death or nonfatal MI). Fol- low-up was more than 99% complete.

RNA variables included in this study were rest and exercise ejection fraction (EF), difference between exercise and rest EF, whether the patient reached an ischemic end point (de- velopment with exercise of angina1 chest pain or significant electrocardiographic changes [ 10.1 mV ST-segment dis- placement from control with an ST-segment slope that was horizontal or directed away from the baseline]), wall motion index and exercise time. No wall motion abnormalities at rest or exercise received a score of 0, resting or exercise only wall motion abnormalities received a score of 1 and new exercise abnormalities in addition to resting abnormalities received a score of 2. Wall motion assessment was derived from the consensus of at least 3 experienced observers without knowledge of the patient’s clinical history or coronary angi- ographic results. The 6 RNA variables used in the analyses were selected beforehand from 30 variables or variable com-

binations obtained at the time of the RNA study as those most likely to be associated with future cardiovascular events. Before any analyses were performed, the 30 variables were divided into 4 groups: (1) ventricular function measures-rest EF, exercise EF, difference between exercise EF and rest EF, average of exercise and resting EF, product of exercise and rest EF, an EF ratio (exercise EF - rest EF/rest EF) and an EF index (exercise EF X the exercise heart rate); (2) wall motion measures-resting wall motion abnormalities, exercise wall motion abnormalities, exercise induce or worsened wall mo- tion abnormalities and the wall motion index just described; (3) “ischemia” measures-exercise electrocardiographic changes, exercise-induced angina, exercise to an ischemic end point (development of either angina or electrocardiographic changes), an ischemic index (coded as 0 for neither exercise- induced angina nor electrocardiographic changes, 1 for either and 2 for both); and (4) hemodynamic or exercise stress vari- ables-heart rate at rest and during exercise, diastolic blood pressure at rest and during exercise, systolic blood pressure at rest and during exercise, the change in heart rate with ex- ercise, a heart rate ratio (the exercise heart rate - resting heart rate/resting heart rate), exercise double product, resting double product, double product ratio, whether or not the target heart rate was reached, exercise time, maximal work load achieved, and whether an adequate exercise end point was achieved (target heart rate or ischemic end point reached). To reduce the likelihood of finding spurious associations (from multiple variable comparisons), we included a priori in the multivariable analyses between 6 and 8 of the variables, in- cluding exercise and rest EF, difference between exercise and rest EF, exercise time or heart rate (whichever was stronger univariately) and the single best wall motion and ischemia measure. An additional ventricular function measure and hemodynamic or exercise stress variable would also be in- cluded if a significantly stronger univariate correlation was found. Preliminary univariable analyses supported the se- lection of the foregoing 6 variables described.

Breslow’s formulationlo of the Cox proportional hazards model” was used to determine the univariable and multi- variable associations between the RNA variables and the outcome events of interest. The dependent variable in these analyses was the time to an event rather than simply whether or not an event occurred within a fixed time period during follow-up. In the multivariable analyses, a stepwise strategy was used that was similar to that described in determining independent associations between variables obtained at car- diac catheterization and prognostic outcomes.lJ For the 33 patients who subsequently underwent surgery (median time to surgery 8.5 months), follow-up time was censored at the time of surgery. These patients tended to be younger but were otherwise similar to those in the study population. Multi- variable analyses identified variables contributing indepen- dent information about the likelihood of future events.

Because of the limited number of events that have occurred to date, we did not feel that analyses that described the added value of RNA testing to all other information known about each patient (clinical descriptors and descriptors from the cardiac catheterization) could be performed. To determine the added value of a test, it is necessary to include all infor- mation known about the patient before the test.‘” Because this requires many descriptors, in the present analyses this would have resulted in including nearly as many covariants as the number of outcome events. However, we did examine whether the anatomic information obtained at catheterization con- tributed significant information beyond that known from the RNA evaluation. For these analyses, we repeated the Cox models described herein, but included as 1 additional covar- iate a summary descriptor of the patient’s anatomy. The de-

20 RADIONUCLIDE ANGlOGRAPHIC PROGNOSTIC VARIABLES

TABLE II Univariable Associations* Between Radionuclide Angiographic Variables and Future Events

Outcome

Variable Cardiovascular Total Cardiac

Death Events

Rest EF Exercise EF EF difference (between exercise

and rest) lschemic end point Wall motion abnormalities Exercise time

:: :8” 2.64 3.35

0.00 0.19 20 12

3.85 7.23

l Chi-square statistics (x2 >3.84, p <0.05). EF = ejection fraction.

scriptor used was one developed from 2,262 consecutive, medically treated, symptomatic patients catheterized for suspected CAD between 1969 and 1983 and is a weighted score of the maximal percent stenosis in the left main, left anterior descending, left circumflex and right coronary arteries. The scores were developed using Cox regression analyses that quantitated the importance of the maximal percent narrowing in each vessel to subsequent survival or infarct-free sur- vival.

Results

Among the 386 patients, 27 cardiovascular deaths and 45 total cardiac events occurred during the follow-up period (median duration 2 years; 25 and 75 percentiles 0.5 and 3 years, respectively). Seven of the 27 deaths and 22 of the 45 events ocurred in patients with a rest EF 250.

Univariable associations between RNA variables and the occurrence of future events are shown in Table II. Univariable predictors of future events included rest and exercise EF, the presence of wall motion abnor- malities and exercise time. The development of an ischemic end point and the difference between exercise and rest EF were not associated with future events.

Multivariable analyses revealed that only the exercise EF contributed independent information about the

TABLE Ill Importance* of Anatomy Adjusting for Measures of Ventricular Function

Outcome

Analysis Survival Total Events

Anatomy alone Adjusted for rest EF

26.7 (<O.OOl)

Adjusted for exercise EF 11.6 (<O.OOl)

19.1 (<O.OOl) 8.8 (0.003)

4.3 (0.04) 3.3 (0.07)

* Associations shown are Cox model chi squares (p value). EF = ejection fraction.

likelihood of future events. No other RNA variables contributed information about the likelihood of future events once the exercise EF was considered. Conversely, the exercise EF still contributed significant prognostic information after accounting for the other RNA vari- ables (p = 0.0001 for survival and p = 0.0011 for total events).

The ability of the exercise EF to identify patients at high risk is illustrated in Figures 1 and 2. In Figure 1, Kaplan-Meier survival and total cardiac event-free (Fig. 1B) rates are shown as a function of follow-up time. In Figure 2, the 2-year event rates as a function of exercise EF (rounded to the nearest 10) are shown. Patients with a reduced exercise EF are more likely to have future cardiac events.

In Figure 3, total cardiac event-free rates are shown. Each graph is for a different subgroup of the population based on the exercise EF. Within each subgroup the effect of the difference between rest and exercise EF is illustrated. Once the exercise EF is known, total event-free rates are similar regardless of whether the exercise EF was greater or less than the resting EF. The median EF difference (exercise - rest) for patients with an exercise EF 150 was 6 for patients whose exercise EF was greater or equal to the rest EF and -9 for patients whose exercise EF was less than the resting EF. Corre- sponding values for patients with an exercise EF of 35 to 49 were 4 and -13, and for patients whose exercise EF was <35, median values were 4 and -10, respec- tively.

FIGURE 1. Kaplan-Meier survival (A) and total cardiac event-free (B) rates for groups of patients stratified by exercise ejection fraction (EF). The number of patients free of events and entering each year for each subgroup is shown under the abscissa.

January 1, 1984 THE AMERICAN JOURNAL OF CARDIOLOGY Volume 53 21

The results of the Cox model analyses that included as a covariate the anatomic information obtained at cardiac catheterization are shown in Table III. The strength of the association between anatomy and the subsequent outcome is described by the Cox model chi squares. Anatomy considered by itself (univariately) is strongly correlated with subsequent outcome. After adjusting for the importance of resting ventricular function, its importance declines, although the associ- ation is still strong. However, when the effect of exercise EF is considered, the importance of knowing anatomy, although still significant, markedly declines.

Discussion

This study demonstrates that information obtained with RNA is associated with future events for medically treated patients with stable CAD. The best predictor is the exercise EF. No other RNA variables that were considered improved the predictive value of RNA once the exercise EF was known. The exercise EF was a better predictor of prognosis than resting ventricular function, anatomy, exercise electrocardiographic changes or exercise time.

Ventricular function in patients with CAD may im- prove, be unchanged or worsen with exercise. For any patient with a given rest EF, the exercise EF describes the change in ventricular function that occurs with ex- ercise. For example, for 2 patients with a rest EF of 55%, a decrease with exercise to 40% is associated with a worse prognosis than an increase with exercise to 60%. The exercise EF also reflects the difference in prognosis in patients with a different rest EF who have similar responses to exercise. A patient whose rest EF of 30% increases with exercise to 35% has the same prognosis as a patient whose rest EF of 50% decreases to 35%, but

IOO-

u-l 90-

: > (v 80-

k

z p: 70- Y + Z

‘;r 60- w

0 >

50 -

T

d

L 1 1 I , L I 20 30 40 50 60 70 80

EXERCISE EF

f IGURE 2. Two-year survival and total cardiac event-free rates as a function of the exercise ejection fraction (EF) rounded to the nearest 10. Numbers in parentheses are the number of patients within each exercise EF subgroup. (Five patients with an exercise EF <15 or >85 are not included.)

a poorer prognosis than a patient whose rest EF of 50% increases with exercise to 55%. Neither the rest EF nor the change in EF with exercise alone adequately de- scribes the prognosis for medically treated patients with CAD. The exercise EF, by combining the information from the resting EF and the change in EF that occurs with exercise, summarizes the risk of future events for medically treated patients with CAD.

The ability of the Cox model analyses to detect rela- tions is dependent, in part, on the number of events that occur during follow-up. With time and as more events occur, other variables may be identified that contribute prognostic information.

The purpose of this investigation was to determine which RNA variables were prognostically important in medically treated patients with CAD. Although patients who had normal coronary arteries or insignificant dis- ease were excluded, similar analyses that included these patients (n = 627) were substantially unchanged, which suggests that the exercise EF is also useful in estimating the prognosis for patients who have not undergone cardiac catheterization.

Analyses including a description of the anatomic se- verity of disease found at catheterization were included to determine if knowledge of the coronary anatomy contributed prognostic information beyond that known from RNA testing. The results show that the anatomic description of the severity of disease still contributes significant prognostic information even after adjusting for the exercise EF. The clinical importance of this ad- ditional information cannot be determined from the present study because although the associations are only marginally significant and much less than the exercise EF, only 5 patients with significant left main disease were included in the study. By examining the relative importance of anatomy by itself and after adjusting for the rest and exercise EF, it appears that much of that

FIGURE 3. Kaplan-Meier total cardiac event-free rates for patients with an exercise ejection fraction (EF) 150 (A), an exercise EF of 35 to 49 (B) and an exercise EF (35 (C). Patients within each exercise EF group have been divided into those whose execise EF at least equalled the rest EF (straight line) and those whose EF decreased with exercise (broken line). The number of patients free of events and entering each year for each subgroup is shown under the abscissa.

22 RADIONUCLIDE ANGlOGRAPHIC PROGNOSTIC VARIABLES

prognostic information from the patient’s anatomy is described by the exercise EF. These findings are in ac- cord with those of Brown et a1,13 who found that ad- justing for the number of defects with exercise thallium imaging reduced the importance of the number of dis- eased vessels in describing prognosis for patients without a previous infarction.

This study should be considered preliminary because no attempt was made to determine the information the RNA contributed beyond that known from the clinical characteristics of the patients. An insufficient number of events has occurred to date to formally analyze the added value of RNA.12 Previous analyseG3 have shown that ventricular function at rest and anatomy are the strongest predictors of cardiovascular death. In this analysis, an exercise measure of ventricular function was a stronger predictor of future events than a resting measure and also describes much of the prognostic in- formation from anatomy. These findings suggest that RNA is useful in identifying patients with CAD who are at greater risk of experiencing future events. Only 1 patient with an exercise EF >45 died within 2 years of the study. This subgroup of patients represented two thirds of the study population. If these findings replicate in other patient samples, RNA may be superior to clinical variables (such as congestive heart failure and history of previous MI) or descriptors of anatomy or resting ventricular function in predicting out- come.1,3,14,15

Acknowledgmenti We acknowledge the assistance of Cristy Vollmar, Nancy Terry and Madelyn Farrell in pre- paring this manuscript.

References

1. Harris PJ, Harrell FE Jr, Lee KL, Behar VS, Rosa6 RA. Survival in medically treated coronary artery disease. Circulation 1979;60: 1259-1269.

2. Harris PJ, Lee KL, Harrell FE Jr, Behar VS, Rosati RA. Outcome in medi- cally treated coronary artery disease. Circulation 1980;62:7 18-726.

3. Harris PJ, Harrell FE Jr, Lee KL. Rosali RA. Nonfatal mvocardial infarction in medically treated patients with coronary artery disease. Am J Cardiol 1980:46:937-942.

4. Rosaii RA, McNeer JF, Starmer EF, Mittler BS, Morris JJ, Wallace AG. A new information system for medical practice. Arch Intern Med 1975; 135:1017-1024.

5. Harris PJ, Behar VS, Conley MJ, Harrell FE, Lee KL, Peter RH, Kong Y, Rosati RA. The prognostic significance of 50% coronary stenosis in medically treated patients with coronary artery disease. Circulation 1980;62:240-248.

6. Campos CT, Chu HW, D’Agostino HJ Jr, Jones RH. Comparison of rest and exercise radionuclide angiography and exercise tread’mill testing for diaanosis of anatomicallv extensive coronarv arterv disease. Circulation 1983;6:1204-1210. .

I ,

7. Jones RH, McEwan P, Newman GE, Port S, Rerych SK, Scholz PM, Upton MT, Peter CA, Austin EH, Leong K, Gibbons RJ, Cobb FR, Coleman RE, Sablston DC. The accuracy of diaanosis of coronarv arterv disease bv ra- dionuclide measurements bf left vjentricular function during rest and &xer- cise. Circulation 1981;64:586-601.

6. Rerych SK, Scholz PM, Newman GE, Sabiston DC Jr, Jones RH. Cardiac funciion at rest and during exercise in normals and in patients with coronary heart disease. Evaluation by radionuclide angiocardiography. Ann Surg 1978;187:449-464.

9. Upton MT, Rerych SK, Newman GE, Bounous EP Jr, Jones RH. The re- producibility of radionuclide angiocardiographic measurements of left ventricular function in normal subiects at rest and during exercise. Circu- lation 1980;62:126-132.

10. Breslow N. Covariance analysis of censored survival data. Biometrics 1979;30:89-99.

11. Cox DR. Regression models and life tables (with discussion), J Roy Statist Sot (series B) 1972;34:187-202.

12. Harrell FE, Califf RM, Pryor DB, Lee KL, Rosali RA. Evaluating the yield of medical tests. JAMA 1982;247:2543-2546.

13. Brown KA, Boucher CA, Okada RD, Gulney TE, Newell JB, Strauss HA, Pohost GM. Proanostic value of exercise thallium-201 imaoina in oatients presenting for &aluation of chest pain. J Am Coil CardiG 1>8$1:994- 1001.

14. Callfl RM, Bounous P, Harrell FE, McCants B, Lee KL, McKlnnis RA, Rosati RA. The prognosis in the presence of coronary artery disease. In: Braunwald E, Mock MB, Watson J. eds. Congestive Heart Failure: Current Research and Clinical Applications. New York: Grune & Stratton, 1983:31-40.

15. Califf RM, Tomabechi Y, Lee KL, Philfips H, Pryor DB, Harrell FE, Harris PJ, Peter RH, Behar VS, Kong Y, Rosatl RA. Outcome in one-vessel cor- onary artery disease. Circulation 1983;87:283-290.