Diagnostic Cardiology Cardiology 1992;81:164-171
Human Performance Laboratory, Division of Cardiovascular Medicine, and Statistical Laboratory, University of California at Davis, Calif.. USA
Key Words Exercise test Coronary artery disease Multivariate models Probability Angina
The Value of Chest Pain during the Exercise Tolerance Test in Predicting Coronary Artery Disease
Abstract The predictive power of 10 common exercise test parameters compared with coronary angiography was studied. Only the exercise electrocardiogram (EXECG), maximal rate pressure product (MAXRPP), and exercise chest pain (EXCP) contrib uted unique predictive information with the emergence of two interactions involving EXCP (EXCP-EXECG and EXCPMAXRPP). In conclusion: (1) EXCP appears to be a more serious finding only in those higher risk individuals with either a positive EXECG or lower MAXRPP; (2) EXCP and its interactions may help discriminate between anginal and nonanginal, exertional chest pain, and (3) the contradictory results found when EXCP was allowed to interact may explain conflicting results in previous multivariate models regarding the predictive significance of EXCP.
Introduction Although exercise electrocardiography (ECG) has been utilized to detect coronary artery disease (CAD) for 50 years, uncertain ties and controversies persist regarding this method [1-3]. Test results are confounded by arbitrary criteria and difficulty in interpreta tion. Because of the limitations of this proce dure in discriminating between normal and
Received: July 7. 1992 Accepted: July 16, 1992
abnormal, based on the ST-segment response alone, numerous studies have been conducted to evaluate aspects of the exercise response other than the traditional ECG criteria (i.e. exertional chest pain) [4-30], Although this approach has led to a more appropriate pro babilistic interpretation of test results instead of a dichotomous, ‘positive-negative’ result based on the presence or absence of ischemictype ST depression, no definitive conclusion
Mark T. Richardson College of Education. Area of Health and Human Performance Studies A.B. Moore Hall. University of Alabama Tuscaloosa, A I, 35487 (USA)
© 1992 S. Kargcr AG. Basel 0008-6312/92/ 0813— 0164S2.75/0
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
Mark T. Richardsona Robert G. Hollya Ezra A. Amsterdamb Michael F. Millerc
regarding the total diagnostic contribution of these additional parameters has been reached. It remains to be determined which variables contribute independent predictive informa tion and, more importantly, how this infor mation is influenced by interactions among variables. When considered simultaneously with the ST-segment response and other com monly measured exercise tolerance test (ETT) variables, the presence of exertional chest pain has been shown to either increase the risk of CAD [7. 23, 29] or add no unique predic tive information [10, 25, 27,28, 30] concern ing the presence or absence of CAD. The purpose of this study was to assess the significance of exertional chest pain and other common ETT variables and their interactions in conjunction with the ST-segment response in predicting CAD.
Treadmill Tests A progressive treatmill ETT utilizing the Bruce protocol [31] was performed. Tests were terminated due to either exhaustion, signs or symptoms of exer tional intolerance, or the patient's request to stop. All ECGs (resting, exercise, and postexcrcisc) were taken using the Mason-Likar modification of the standard twelve-lead placements [32], Twelve-lead ECGs were recorded prior to and immediately following exercise. Leads I, aVF and V5 were recorded every minute dur ing exercise and every' other minute during recovery which lasted until the tracing returned to baseline (ap proximately 6 min). Blood pressure was recorded with the standard sphygmomanometric method before, during each stage of and after exercise. The criterion fora positive test was: ( I) > I mm of exercise-induced ST-segment depression measured 0.08 s from the T point of any morphology (upsloping, horizontal or downsloping) or (2) the development of > I mm STsegment elevation. A negative response w'as one in which the patient achieved > 8 5 % predicted maxi mum heart rate with < I mm ST-segment deviation from the resting tracing. Coronary Angiography
Patient Selection Records of 1,138 consecutive patients who had both selective coronary angiography and ETT (usually within 3 days of each other) were examined retrospec tively. Only the records from male patients were con sidered for this study. These records were further reviewed and patients knowm to have the following conditions or disorders w'hich might influence the interpretation of the ETT were excluded from this study: (1) a history of previous myocardial infarction and/or coronary' artery bypass surgery: (2) cardiac dis orders other than CAD including congenital heart dis ease. cardiomyopathies, pericardial disorders or valvu lar heart disease: (3) ECG evidence of left bundle branch block, left ventricular hypertrophy or pre-exci tation syndromes: (4) drug therapy including digitalis. (3-blockers. calcium channel blockers, antiarrhvthmic agents or psychoactive agents, and (5) failure to reach 85% of age-predicted maximal heart rate during the treadmill test unless ECG evidence of ischemia oc curred prior to test termination. One hundred thirtyfive male subjects met the selection criteria, of whom 122 had complete data (maximal systolic blood pres sure was missing in 13 subjects).
Selective coronary arteriograms were obtained by standard methods. Fifty percent or greater luminal diameter narrow'ing in a major coronary artery was considered a significant lesion. All studies were inter preted by at least two experienced angiographers.
Data Collected The following common ETT parameters were se lected and coded for computer-assisted statistical anal ysis: patient’s age (AGE) in years: resting ECG repolar ization abnormalities (e.g.. T-wavc inversion. ST de pression 235 but 284) beats-mm H g-m in'1-102 levels of maximum rate pressure prod uct (MAXRPP) (n = 122) N O C A D CAD n n
CAD %
O b serv ed SE
10
56
12a b
20
95
5a
M A X RPP
EXCP
Low
absent
8
p resen t
1 22
7
24
p resen t
3
10
77
absent
16
13
45
p resen t
9
3
25
M id H ig h
absent
8a b c 12c Ç a. C
13 a c
racy of a positive test, and predictive accuracy of a negative test were 75, 73. 78, and 70%, respectively. MAXSBP was not recorded for 13 subjects during the exercise tolerance test. Therefore, for any analyses involving this variable (in cluding MAXRPP) the sample was reduced to 122. Of these 13 subjects, 10 (77%) had CAD, and 8 had a positive treadmill test with all 8 having CAD.
It is obvious that the selected variables are highly intercorrelated with each other (ta ble 1) and individually related to CAD (ta ble 2). Further analysis revealed that of these 10 variables, only 3 (EXECG. MAXRPP and EXCP) contributed unique predictive infor mation. although EXCP only contributed when it was allowed to interact with the other 2 variables. While EXCP alone added no unique predictive information regarding the
167
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
a b c D i f f e r e n t f r o m e a c h o t h e r t o a t le a s t p = 0 .0 5 .
Discussion Predictive Value o f E T T Variables This study demonstrated that, of the 10 ETT variables analyzed, exertional chest pain and maximal rate pressure product in con junction with the ST-scgment response ac counted for the total amount of predictive information available in determining the risk for CAD. Although the majority of variables are significantly related to CAD (table 2), be cause of a high degree of colinearity (table I), only EXECG and MAXRPP emerged from the first stepwise logistic regression analysis as independent predictors of CAD. When EXCP was forced into the model, its importance through interactions with other variables be came apparent. Subsequent log-linear analy sis confirmed this conclusion. EXCP became an extremely important predictor of CAD, although only through its interactions with other variables, specifically EXECG (table 3) and MAXRPP (table 4). Perhaps the most striking finding of this study was the extent to which the predictive values of EXECG and MAXRPP were altered by EXCP. As expected, a higher risk of CAD was found in those subjects with a positive rather than a negative EXECG (75 vs. 29%) (table 3) and a low to mid rather than a high MAXRPP (58 vs. 39%) (table 4). The effect of EXCP was most marked in these higher risk groups. When EXECG was positive (table 3), CAD was present in 92% of the patients with EXCP compared to 53% of those without EXCP. Similarly, when MAXRPP was in the low to mid range (table 4), CAD was present
168
in 95 and 77% of the patients, respectively, with EXCP compared to 56 and 24%, respec tively. of those without EXCP. Surprisingly and in contrast to these find ings. EXCP exerted the opposite effect in those groups at lower risk for CAD: In those with a negative EXECG (table 3) or high MAXRPP (table 4). EXCP was associated with a lower, not higher, risk of CAD. When EXECG was negative, CAD was present in none (0%) of those with EXCP but in 34% of those without EXCP. Similarly, when MAXRPP was high. CAD was present in 25% of those with EXCP but in 45% of those with out EXCP. The cause of this contradictory effect of EXCP in high and low risk groups remains speculative but the following possi bilities exist: First, this may be a characteristic specific to our population. This seems un likely since EXCP has not consistently been found to be an important predictive variable in previous multivariate models [10, 25. 27, 28, 30]. In fact, this interactive effect of EXCP, which has not previously been de scribed. may well explain this prior lack of consistency. Secondly, and more likely, this may be due to the nonspecific nature of chest pain. It is likely that if the initial complaint were chest pain (particularly exertional chest pain), then further evaluation might reveal this chest pain to be nonanginal in many cases. The ETT may be an excellent test to further discriminate anginal from nonangi nal, exertional chest pain, in that anginal pain would cluster with higher risk attributes like a positive EXECG and a lower MAXRPP. Conversely, nonanginal chest pain would cluster with lower risk attributes like a nega tive EXECG and a high MAXRPP. Only a dichotomous evaluation of EXCP was avail able for this study. Both a more qualitative as well as qantitative description of exertional chest pain during the ETT would aid greatly in explaining its interactive effects. Moreover,
Richardson/Holly/Amstcrdam/Miller
Chest Pain during the Exercise Tolerance Test
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
presence or absence of CAD, its interac tions with EXECG (EXCP-EXECG) and MAXRPP (EXCP-MAXRPP) significantly altered the value of both EXECG and MAXRPP in predicting CAD (tables 3, 4).
between the EXECG and the presence or absence of CAD. Thus, exclusion of these patients from analyses involving the diagnos tic ability of the ETT to detect CAD is com mon [4,23.28.29]. Implications and Limitations o f the Present Study The present analysis produced a model which reveals important predictive interac tions involving the presence of chest pain dur ing the ETT. With standard multiple regres sion models, these associations would not have been apparent. However, it should be noted that this model does not discriminate between those with or without disease. Rather the model predicts the risk of having CAD (ta bles 3.4). This is a more appropriate probabi listic analysis of exercise test results compared to the conventional dichotomous ‘positivenegative’ approach to identifying CAD. How ever, one would not expect an analysis of 10 stress test variables to account for all the vari ance in a disease as complex as CAD. Consid eration of conventional risk factor data (pre test likelihood) would increase the degree of predictable variance and aid a multivariate analysis of this nature. In addition, CAD itself was artificially measured as a dichotomous variable (>50% occlusion in one or more coronary arteries). A more quantitative mea surement might have increased the predictive value of the 10 stress test variables. Also, it should be stressed that the model produced from the present analysis was generated retro spectively from a relatively small patient sam ple, which was particularly true of a number of the subgroups (tables 3,4). Thus, these con clusions should be tested prospectively on a larger, independent population, paying par ticular attention to the nature of chest pain. Finally, the selected nature of this popula tion should be appreciated: Our subjects were all referred for coronary angiography. Fur
169
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
it should be cautioned that this apparently contradictory effect of EXCP in patients with a negative ECG or high MAXRPP was, in part, generated from subgroups with small numbers. A prospective study with greater numbers that focuses on the specific nature of chest pain is necessary to test this possibility. Furthermore, these findings do not preclude the existence of a relatively small number of patients who exhibit effort angina pectoris without significant ST-T changes [36], Re gardless. in assessing the predictive power of various exercise test variables, it is important to consider their interactive as well as direct effects since EXCP and its interactions mark edly alter the prediction of risk within a given risk category (e.g.. positive EXECG). Finally, the presence of EXCP appears to be a more serious finding only in those higher risk indi viduals with either a positive EXECG or lower MAXRPP. The removal of 13 subjects from the an alysis of the interaction between EXCPMAXRPP and CAD occurrence (table 4) may have biased the data since both the prev alence of CAD (77%) and the predictive ac curacy of a positive test (100%) were greater in this group than in the entire population (54 and 78%. respectively). The effect of this missing data on our conclusions remains un known. Another potential source of bias may have resulted from the exclusion of patients who failed to reach 85% of age-predicted maximal heart rate during the treadmill test without ECG evidence of ischemia. A num ber of these individuals may have been stopped prematurely because of severe chest pain due to CAD. This would tend to distort the present finding that patients with a nega tive exercise ECG and EXCP did not have CAD. This remains speculative, as the data on these individuals are not available. How ever. to have included these patients in the analysis may have distorted the relationship
thermore, many medicated patients and those with pre-existing abnormal resting ECGs (who may have the most severe CAD) were excluded from the study due to the confound ing effects these conditions have on the inter pretation of the exercise ECG.
Conclusions Probabilities for CAD were objectively de termined from routine data gathered during stress testing. The conclusions which can be drawn from this study are as follows: (1) of the 10 stress test variables analyzed, the vari ables EXECG. MAXRPP and EXCP ac
counted for the total amount of predictive information available in determining the risk of CAD: (2) important interaction terms in volving EXCP markedly altered the risk in subgroups of EXECG and MAXRPP: (3) the presence of EXCP appears to be a more seri ous finding only in those higher risk individu als with either a positive EXECG or lower MAXRPP « 2 8 .4 0 0 ): (4) EXCP and its in teractions may help discriminate between an ginal and nonanginal. exertional chest pain, and (5) the interactions with EXCP may ex plain conflicting results in previous multivar iate models regarding the predictive signifi cance of EXCP.
1 Feil H. Sicgal M: Electrocardio graphic changes during attacks of angina pectoris. Am J Med Sci 1928: 175:255-260. 2 Robb GP. Marks HH: Latent coro nary artery disease: Determination of its presence and severity by the exercise electrocardiogram. Am J Cardiol 1964:13:603-618. 3 KoppesG. McKiernan T. Bassan M. Froelichcr VF: Treadmill exercise testing. II. Curr Probl Cardiol 1977: 2:1-45. 4 Berman JL. Wynne J. Cohn PF: A multivariate approach for interpret ing treadmill exercise tests in coro nary artery disease. Circulation 1978:58:505-512. 5 Chaitman BR. Bourassa MG. Wagniart P. Corbara F. Ferguson RJ: Improved efficiency of treadmill ex ercise testing using a multiple lead ECG system and basic hemody namic exercise response. Circula tion 1978:57:71-79. 6 McNecr JF. Margolis JR. Kerry LL. Kisslo JA. Peter RH. Kong Y, Behar VS. Wallace AG. McCants CB. Rosati RA: The role of the exercise test in the evaluation of patients for isch emic heart disease. Circulation 1978:57:64-70.
170
7 Weiner DA. McCabe C ll. Hardy WL. Ryan TJ: Multivariate analysis using exercise data to predict extent of coronary disease (abstract). Cir culation 1978;58(suppl II):200. 8 Blomqvist CG. Gaffney FA. Atkins JM. Nixon JV. Mullins CB. Willerson JT. Moore WE. Schutte J: The exercise ECG and related physiolog ical data as markers of critical coro nary artery lesions. Acta Med Scand 1978:615(suppl): 5 1- 6 1. 9 Bruce RA: Exercise testing for eval uation of ventricular function. N Engl J Med 1977:296:671-675. 10 Ellestad MH. Savitz S. Bergdall D. Tcskc J: The false positive stress test. Multivariate analysis of 215 subjects with hemodynamic, angio graphic. and clinical data. Ant J Car diol 1977;40:681-685. 11 Weiner DA. McCabe C, Hucter DC. Ryan TJ. Hood WB: The predictive value of anginal chest pain as an indicator of coronary1disease during exercise testing. Am Heart J 1978: 96:458-462. 12 Cole JP, Ellestad MH: Correlation of chest pain during treadmill exer cise electrocardiography and coro nary events (abstract). Circulation I976;54(suppl II):206.
13 Band AG. Behar VS. Peter RH. Orgain ES. Kong Y: Graded exercise stress tests in angiographically docu mented coronary artery disease. Cir culation 1974;49:348-356. 14 Faris JV. McHenry PL. Jordan JW. Morris SN: Prevalence and repro ducibility of exercise-induced ven tricular arrhythmias during maxi mal exercise testing in normal men. Am J Cardiol 1976:37:617-622. 15 McHenry PL. Morris SN, Kavalier M. Jordan JW: Comparative study of exercise-induced ventricular ar rhythmias in normal subjects and patients with documented coronary artery disease. Am J Cardiol 1976: 37:609-616. 16 Jelinek MV, Lown B: Exercise stress testing for exposure of cardiac ar rhythmia. Prog Cardiovasc Dis 1974:16:497-522. 17 Zaret BL. Conti CR Jr: Exerciseinduced ventricular irritability: He modynamic and angiographic corre lations (abstract). Am J Cardiol 1972:29:298-299.
Richardson/Holly/Amsterdam/Miller
Chest Pain during the Exercise Tolerance Test
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
References
24 Greenberg PS. Cangiano B. Leamy L. Ellestad MD: Use of the multivar iate approach to enhance the diag nostic accuracy of the treadmill stresstest. J Eleclrocardiol 1980:13: 227-236. 25 Fisher LD. Kennedy JW. Chaitman BR. Ryan TJ. McCabe C, Weiner D. Tristani F. Schloss M. Warner HR: Diagnostic quantification of CASS (coronary artery surgery study) clini cal and exercise test results in deter mining presence and extent of coro nary artery disease, a multivariate approach. Circulation 1981:63:987-
1000. 26 Ellcstad MH, Greenberg PS: Multi variate approach to the treadmill stress test: Prospective study. Car diology 1981 ;68(suppl 2):27—34. 27 Goldman L. Cook EF. Mitchell N. Flatley M, Sherman H, Rosali R, Harrell F. Lee K, Cohn PF: Incre mental value of the exercise test for diagnosing the presence or absence of coronary artery disease. Circula tion 1982:66:945-953. 28 Kansal S. Roitman D. Bradley EL Jr. Sheffield LT: Enhanced evalua tion of treadmill tests by means of scoring based on multivariate analy sis and its clinical application: A study of 608 patients. Am J Cardiol 1983:52:1155-1160. 29 Weintraub WS. Barr-Aldcrfer VA. Scelaus PA. Bodenheimer MM, Madeira SW Jr. Katz Rl. Feldman MS. Agarwai JB. Banka VS. Heifant RH: A sequential approach to the diagnosis of coronary artery disease using multivariate analysis. Am Heart J 1985:109:999-1005.
30 Dctry JMR. Robert A. Luwaert RJ. Rousseau MF, Brasseur LA. Melin JA. Brohel CR: Diagnostic value of computerized exercise testing in men without previous myocardial infarction. A multivariate, compartmental and probabilistic approach. Eur Heart J 1985:6:227-238. 31 Bruce RA. Hornsten TR: Exercise stress testing in evaluation of pa tients with ischemic heart disease. Prog Cardiovasc Dis 1969:11:371390. 32 KoppesG. McKicrnan T. Bassan M. Froelicher VF: Treadmill exercise testing. I. C'urr Probl Cardiol 1977; 2:1-44. 33 Nic NH. Hull CH. Jenkins JG. Stainbrenner K. Bent D ll: Statisti cal Package for the Social Sciences, ed 2. New York. McGraw-Hill. 1975, pp 276-300. 34 Engclman L: Stepwise logistic re gression: in Dixon WJ (ed): BMDP Statistical Software. Berkeley, University of California Press. 1983. pp 330-344. 35 Brown MB: Two-way and multiway frequency tables - measures o f asso ciation and the log-linear model: in Dixon WJ (ed): BMDP Statistical Software. Berkeley. University of California Press. 1983. pp 143-206. 36 De Pasquale G. Pinelli G. Tartagni F. Manini GL. Dondi M: Effort an gina pectoris without electrocardio graphic changes in coronary disease patients: Correlations between scin tigraphic and coronav angiographic findings. Int J Cardiol 1986; 12:243— 253.
171
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
18 Goldschlager N. Cake D, Cohn K: Exercise-induced ventricular ar rhythmias in patients with coronary' artery disease. Their relation to an giographic findings. Am J Cardiol 1973:31:434-440. 19 Hclfant RH. Pine R. Kabde V. Vanka VS: Exercise-related ventric ular premature complexes in coro nary heart disease. Correlations with ischemia and angiographic se verity. Ann Intern Med 1974:80: 589-592. 20 Carvalho A. Amsterdam E. DeMaria A. Mason DT: Réévaluation of exercise electrocardiography in diagnoses of coronary artery’ dis ease: Significance of heart rate at which ST-segmcnt depression oc curs (abstract). Circulation 1976; 54(suppl li):206. 21 Goldschlager N. Selzer A. Cohn K: Treadmill stress tests as indicators of presence and severity of coronary artery disease. Ann Intern Med 1976:85:277-286. 22 Hollenberg M. Zoltick JM. Go M. Yaney SF. Daniels W. Davis RC. Bedynek JL: Comparison of a quan titative treadmill exercise score with standard electrocardiographic crite ria in screening asymptomatic young men for coronary artery dis ease. N Engl J Med 1985:313^600606. 23 Cohn K. Kamm B. Fetcih N. Brand R. Goldschlager N: Use of treadmill score to quantify ischemic response and predict extent of coronary dis ease. Circulation 1979:59:286-296.
Human Performance Laboratory, Division of Cardiovascular Medicine, and Statistical Laboratory, University of California at Davis, Calif.. USA
Key Words Exercise test Coronary artery disease Multivariate models Probability Angina
The Value of Chest Pain during the Exercise Tolerance Test in Predicting Coronary Artery Disease
Abstract The predictive power of 10 common exercise test parameters compared with coronary angiography was studied. Only the exercise electrocardiogram (EXECG), maximal rate pressure product (MAXRPP), and exercise chest pain (EXCP) contrib uted unique predictive information with the emergence of two interactions involving EXCP (EXCP-EXECG and EXCPMAXRPP). In conclusion: (1) EXCP appears to be a more serious finding only in those higher risk individuals with either a positive EXECG or lower MAXRPP; (2) EXCP and its interactions may help discriminate between anginal and nonanginal, exertional chest pain, and (3) the contradictory results found when EXCP was allowed to interact may explain conflicting results in previous multivariate models regarding the predictive significance of EXCP.
Introduction Although exercise electrocardiography (ECG) has been utilized to detect coronary artery disease (CAD) for 50 years, uncertain ties and controversies persist regarding this method [1-3]. Test results are confounded by arbitrary criteria and difficulty in interpreta tion. Because of the limitations of this proce dure in discriminating between normal and
Received: July 7. 1992 Accepted: July 16, 1992
abnormal, based on the ST-segment response alone, numerous studies have been conducted to evaluate aspects of the exercise response other than the traditional ECG criteria (i.e. exertional chest pain) [4-30], Although this approach has led to a more appropriate pro babilistic interpretation of test results instead of a dichotomous, ‘positive-negative’ result based on the presence or absence of ischemictype ST depression, no definitive conclusion
Mark T. Richardson College of Education. Area of Health and Human Performance Studies A.B. Moore Hall. University of Alabama Tuscaloosa, A I, 35487 (USA)
© 1992 S. Kargcr AG. Basel 0008-6312/92/ 0813— 0164S2.75/0
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
Mark T. Richardsona Robert G. Hollya Ezra A. Amsterdamb Michael F. Millerc
regarding the total diagnostic contribution of these additional parameters has been reached. It remains to be determined which variables contribute independent predictive informa tion and, more importantly, how this infor mation is influenced by interactions among variables. When considered simultaneously with the ST-segment response and other com monly measured exercise tolerance test (ETT) variables, the presence of exertional chest pain has been shown to either increase the risk of CAD [7. 23, 29] or add no unique predic tive information [10, 25, 27,28, 30] concern ing the presence or absence of CAD. The purpose of this study was to assess the significance of exertional chest pain and other common ETT variables and their interactions in conjunction with the ST-segment response in predicting CAD.
Treadmill Tests A progressive treatmill ETT utilizing the Bruce protocol [31] was performed. Tests were terminated due to either exhaustion, signs or symptoms of exer tional intolerance, or the patient's request to stop. All ECGs (resting, exercise, and postexcrcisc) were taken using the Mason-Likar modification of the standard twelve-lead placements [32], Twelve-lead ECGs were recorded prior to and immediately following exercise. Leads I, aVF and V5 were recorded every minute dur ing exercise and every' other minute during recovery which lasted until the tracing returned to baseline (ap proximately 6 min). Blood pressure was recorded with the standard sphygmomanometric method before, during each stage of and after exercise. The criterion fora positive test was: ( I) > I mm of exercise-induced ST-segment depression measured 0.08 s from the T point of any morphology (upsloping, horizontal or downsloping) or (2) the development of > I mm STsegment elevation. A negative response w'as one in which the patient achieved > 8 5 % predicted maxi mum heart rate with < I mm ST-segment deviation from the resting tracing. Coronary Angiography
Patient Selection Records of 1,138 consecutive patients who had both selective coronary angiography and ETT (usually within 3 days of each other) were examined retrospec tively. Only the records from male patients were con sidered for this study. These records were further reviewed and patients knowm to have the following conditions or disorders w'hich might influence the interpretation of the ETT were excluded from this study: (1) a history of previous myocardial infarction and/or coronary' artery bypass surgery: (2) cardiac dis orders other than CAD including congenital heart dis ease. cardiomyopathies, pericardial disorders or valvu lar heart disease: (3) ECG evidence of left bundle branch block, left ventricular hypertrophy or pre-exci tation syndromes: (4) drug therapy including digitalis. (3-blockers. calcium channel blockers, antiarrhvthmic agents or psychoactive agents, and (5) failure to reach 85% of age-predicted maximal heart rate during the treadmill test unless ECG evidence of ischemia oc curred prior to test termination. One hundred thirtyfive male subjects met the selection criteria, of whom 122 had complete data (maximal systolic blood pres sure was missing in 13 subjects).
Selective coronary arteriograms were obtained by standard methods. Fifty percent or greater luminal diameter narrow'ing in a major coronary artery was considered a significant lesion. All studies were inter preted by at least two experienced angiographers.
Data Collected The following common ETT parameters were se lected and coded for computer-assisted statistical anal ysis: patient’s age (AGE) in years: resting ECG repolar ization abnormalities (e.g.. T-wavc inversion. ST de pression 235 but 284) beats-mm H g-m in'1-102 levels of maximum rate pressure prod uct (MAXRPP) (n = 122) N O C A D CAD n n
CAD %
O b serv ed SE
10
56
12a b
20
95
5a
M A X RPP
EXCP
Low
absent
8
p resen t
1 22
7
24
p resen t
3
10
77
absent
16
13
45
p resen t
9
3
25
M id H ig h
absent
8a b c 12c Ç a. C
13 a c
racy of a positive test, and predictive accuracy of a negative test were 75, 73. 78, and 70%, respectively. MAXSBP was not recorded for 13 subjects during the exercise tolerance test. Therefore, for any analyses involving this variable (in cluding MAXRPP) the sample was reduced to 122. Of these 13 subjects, 10 (77%) had CAD, and 8 had a positive treadmill test with all 8 having CAD.
It is obvious that the selected variables are highly intercorrelated with each other (ta ble 1) and individually related to CAD (ta ble 2). Further analysis revealed that of these 10 variables, only 3 (EXECG. MAXRPP and EXCP) contributed unique predictive infor mation. although EXCP only contributed when it was allowed to interact with the other 2 variables. While EXCP alone added no unique predictive information regarding the
167
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
a b c D i f f e r e n t f r o m e a c h o t h e r t o a t le a s t p = 0 .0 5 .
Discussion Predictive Value o f E T T Variables This study demonstrated that, of the 10 ETT variables analyzed, exertional chest pain and maximal rate pressure product in con junction with the ST-scgment response ac counted for the total amount of predictive information available in determining the risk for CAD. Although the majority of variables are significantly related to CAD (table 2), be cause of a high degree of colinearity (table I), only EXECG and MAXRPP emerged from the first stepwise logistic regression analysis as independent predictors of CAD. When EXCP was forced into the model, its importance through interactions with other variables be came apparent. Subsequent log-linear analy sis confirmed this conclusion. EXCP became an extremely important predictor of CAD, although only through its interactions with other variables, specifically EXECG (table 3) and MAXRPP (table 4). Perhaps the most striking finding of this study was the extent to which the predictive values of EXECG and MAXRPP were altered by EXCP. As expected, a higher risk of CAD was found in those subjects with a positive rather than a negative EXECG (75 vs. 29%) (table 3) and a low to mid rather than a high MAXRPP (58 vs. 39%) (table 4). The effect of EXCP was most marked in these higher risk groups. When EXECG was positive (table 3), CAD was present in 92% of the patients with EXCP compared to 53% of those without EXCP. Similarly, when MAXRPP was in the low to mid range (table 4), CAD was present
168
in 95 and 77% of the patients, respectively, with EXCP compared to 56 and 24%, respec tively. of those without EXCP. Surprisingly and in contrast to these find ings. EXCP exerted the opposite effect in those groups at lower risk for CAD: In those with a negative EXECG (table 3) or high MAXRPP (table 4). EXCP was associated with a lower, not higher, risk of CAD. When EXECG was negative, CAD was present in none (0%) of those with EXCP but in 34% of those without EXCP. Similarly, when MAXRPP was high. CAD was present in 25% of those with EXCP but in 45% of those with out EXCP. The cause of this contradictory effect of EXCP in high and low risk groups remains speculative but the following possi bilities exist: First, this may be a characteristic specific to our population. This seems un likely since EXCP has not consistently been found to be an important predictive variable in previous multivariate models [10, 25. 27, 28, 30]. In fact, this interactive effect of EXCP, which has not previously been de scribed. may well explain this prior lack of consistency. Secondly, and more likely, this may be due to the nonspecific nature of chest pain. It is likely that if the initial complaint were chest pain (particularly exertional chest pain), then further evaluation might reveal this chest pain to be nonanginal in many cases. The ETT may be an excellent test to further discriminate anginal from nonangi nal, exertional chest pain, in that anginal pain would cluster with higher risk attributes like a positive EXECG and a lower MAXRPP. Conversely, nonanginal chest pain would cluster with lower risk attributes like a nega tive EXECG and a high MAXRPP. Only a dichotomous evaluation of EXCP was avail able for this study. Both a more qualitative as well as qantitative description of exertional chest pain during the ETT would aid greatly in explaining its interactive effects. Moreover,
Richardson/Holly/Amstcrdam/Miller
Chest Pain during the Exercise Tolerance Test
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
presence or absence of CAD, its interac tions with EXECG (EXCP-EXECG) and MAXRPP (EXCP-MAXRPP) significantly altered the value of both EXECG and MAXRPP in predicting CAD (tables 3, 4).
between the EXECG and the presence or absence of CAD. Thus, exclusion of these patients from analyses involving the diagnos tic ability of the ETT to detect CAD is com mon [4,23.28.29]. Implications and Limitations o f the Present Study The present analysis produced a model which reveals important predictive interac tions involving the presence of chest pain dur ing the ETT. With standard multiple regres sion models, these associations would not have been apparent. However, it should be noted that this model does not discriminate between those with or without disease. Rather the model predicts the risk of having CAD (ta bles 3.4). This is a more appropriate probabi listic analysis of exercise test results compared to the conventional dichotomous ‘positivenegative’ approach to identifying CAD. How ever, one would not expect an analysis of 10 stress test variables to account for all the vari ance in a disease as complex as CAD. Consid eration of conventional risk factor data (pre test likelihood) would increase the degree of predictable variance and aid a multivariate analysis of this nature. In addition, CAD itself was artificially measured as a dichotomous variable (>50% occlusion in one or more coronary arteries). A more quantitative mea surement might have increased the predictive value of the 10 stress test variables. Also, it should be stressed that the model produced from the present analysis was generated retro spectively from a relatively small patient sam ple, which was particularly true of a number of the subgroups (tables 3,4). Thus, these con clusions should be tested prospectively on a larger, independent population, paying par ticular attention to the nature of chest pain. Finally, the selected nature of this popula tion should be appreciated: Our subjects were all referred for coronary angiography. Fur
169
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
it should be cautioned that this apparently contradictory effect of EXCP in patients with a negative ECG or high MAXRPP was, in part, generated from subgroups with small numbers. A prospective study with greater numbers that focuses on the specific nature of chest pain is necessary to test this possibility. Furthermore, these findings do not preclude the existence of a relatively small number of patients who exhibit effort angina pectoris without significant ST-T changes [36], Re gardless. in assessing the predictive power of various exercise test variables, it is important to consider their interactive as well as direct effects since EXCP and its interactions mark edly alter the prediction of risk within a given risk category (e.g.. positive EXECG). Finally, the presence of EXCP appears to be a more serious finding only in those higher risk indi viduals with either a positive EXECG or lower MAXRPP. The removal of 13 subjects from the an alysis of the interaction between EXCPMAXRPP and CAD occurrence (table 4) may have biased the data since both the prev alence of CAD (77%) and the predictive ac curacy of a positive test (100%) were greater in this group than in the entire population (54 and 78%. respectively). The effect of this missing data on our conclusions remains un known. Another potential source of bias may have resulted from the exclusion of patients who failed to reach 85% of age-predicted maximal heart rate during the treadmill test without ECG evidence of ischemia. A num ber of these individuals may have been stopped prematurely because of severe chest pain due to CAD. This would tend to distort the present finding that patients with a nega tive exercise ECG and EXCP did not have CAD. This remains speculative, as the data on these individuals are not available. How ever. to have included these patients in the analysis may have distorted the relationship
thermore, many medicated patients and those with pre-existing abnormal resting ECGs (who may have the most severe CAD) were excluded from the study due to the confound ing effects these conditions have on the inter pretation of the exercise ECG.
Conclusions Probabilities for CAD were objectively de termined from routine data gathered during stress testing. The conclusions which can be drawn from this study are as follows: (1) of the 10 stress test variables analyzed, the vari ables EXECG. MAXRPP and EXCP ac
counted for the total amount of predictive information available in determining the risk of CAD: (2) important interaction terms in volving EXCP markedly altered the risk in subgroups of EXECG and MAXRPP: (3) the presence of EXCP appears to be a more seri ous finding only in those higher risk individu als with either a positive EXECG or lower MAXRPP « 2 8 .4 0 0 ): (4) EXCP and its in teractions may help discriminate between an ginal and nonanginal. exertional chest pain, and (5) the interactions with EXCP may ex plain conflicting results in previous multivar iate models regarding the predictive signifi cance of EXCP.
1 Feil H. Sicgal M: Electrocardio graphic changes during attacks of angina pectoris. Am J Med Sci 1928: 175:255-260. 2 Robb GP. Marks HH: Latent coro nary artery disease: Determination of its presence and severity by the exercise electrocardiogram. Am J Cardiol 1964:13:603-618. 3 KoppesG. McKiernan T. Bassan M. Froelichcr VF: Treadmill exercise testing. II. Curr Probl Cardiol 1977: 2:1-45. 4 Berman JL. Wynne J. Cohn PF: A multivariate approach for interpret ing treadmill exercise tests in coro nary artery disease. Circulation 1978:58:505-512. 5 Chaitman BR. Bourassa MG. Wagniart P. Corbara F. Ferguson RJ: Improved efficiency of treadmill ex ercise testing using a multiple lead ECG system and basic hemody namic exercise response. Circula tion 1978:57:71-79. 6 McNecr JF. Margolis JR. Kerry LL. Kisslo JA. Peter RH. Kong Y, Behar VS. Wallace AG. McCants CB. Rosati RA: The role of the exercise test in the evaluation of patients for isch emic heart disease. Circulation 1978:57:64-70.
170
7 Weiner DA. McCabe C ll. Hardy WL. Ryan TJ: Multivariate analysis using exercise data to predict extent of coronary disease (abstract). Cir culation 1978;58(suppl II):200. 8 Blomqvist CG. Gaffney FA. Atkins JM. Nixon JV. Mullins CB. Willerson JT. Moore WE. Schutte J: The exercise ECG and related physiolog ical data as markers of critical coro nary artery lesions. Acta Med Scand 1978:615(suppl): 5 1- 6 1. 9 Bruce RA: Exercise testing for eval uation of ventricular function. N Engl J Med 1977:296:671-675. 10 Ellestad MH. Savitz S. Bergdall D. Tcskc J: The false positive stress test. Multivariate analysis of 215 subjects with hemodynamic, angio graphic. and clinical data. Ant J Car diol 1977;40:681-685. 11 Weiner DA. McCabe C, Hucter DC. Ryan TJ. Hood WB: The predictive value of anginal chest pain as an indicator of coronary1disease during exercise testing. Am Heart J 1978: 96:458-462. 12 Cole JP, Ellestad MH: Correlation of chest pain during treadmill exer cise electrocardiography and coro nary events (abstract). Circulation I976;54(suppl II):206.
13 Band AG. Behar VS. Peter RH. Orgain ES. Kong Y: Graded exercise stress tests in angiographically docu mented coronary artery disease. Cir culation 1974;49:348-356. 14 Faris JV. McHenry PL. Jordan JW. Morris SN: Prevalence and repro ducibility of exercise-induced ven tricular arrhythmias during maxi mal exercise testing in normal men. Am J Cardiol 1976:37:617-622. 15 McHenry PL. Morris SN, Kavalier M. Jordan JW: Comparative study of exercise-induced ventricular ar rhythmias in normal subjects and patients with documented coronary artery disease. Am J Cardiol 1976: 37:609-616. 16 Jelinek MV, Lown B: Exercise stress testing for exposure of cardiac ar rhythmia. Prog Cardiovasc Dis 1974:16:497-522. 17 Zaret BL. Conti CR Jr: Exerciseinduced ventricular irritability: He modynamic and angiographic corre lations (abstract). Am J Cardiol 1972:29:298-299.
Richardson/Holly/Amsterdam/Miller
Chest Pain during the Exercise Tolerance Test
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
References
24 Greenberg PS. Cangiano B. Leamy L. Ellestad MD: Use of the multivar iate approach to enhance the diag nostic accuracy of the treadmill stresstest. J Eleclrocardiol 1980:13: 227-236. 25 Fisher LD. Kennedy JW. Chaitman BR. Ryan TJ. McCabe C, Weiner D. Tristani F. Schloss M. Warner HR: Diagnostic quantification of CASS (coronary artery surgery study) clini cal and exercise test results in deter mining presence and extent of coro nary artery disease, a multivariate approach. Circulation 1981:63:987-
1000. 26 Ellcstad MH, Greenberg PS: Multi variate approach to the treadmill stress test: Prospective study. Car diology 1981 ;68(suppl 2):27—34. 27 Goldman L. Cook EF. Mitchell N. Flatley M, Sherman H, Rosali R, Harrell F. Lee K, Cohn PF: Incre mental value of the exercise test for diagnosing the presence or absence of coronary artery disease. Circula tion 1982:66:945-953. 28 Kansal S. Roitman D. Bradley EL Jr. Sheffield LT: Enhanced evalua tion of treadmill tests by means of scoring based on multivariate analy sis and its clinical application: A study of 608 patients. Am J Cardiol 1983:52:1155-1160. 29 Weintraub WS. Barr-Aldcrfer VA. Scelaus PA. Bodenheimer MM, Madeira SW Jr. Katz Rl. Feldman MS. Agarwai JB. Banka VS. Heifant RH: A sequential approach to the diagnosis of coronary artery disease using multivariate analysis. Am Heart J 1985:109:999-1005.
30 Dctry JMR. Robert A. Luwaert RJ. Rousseau MF, Brasseur LA. Melin JA. Brohel CR: Diagnostic value of computerized exercise testing in men without previous myocardial infarction. A multivariate, compartmental and probabilistic approach. Eur Heart J 1985:6:227-238. 31 Bruce RA. Hornsten TR: Exercise stress testing in evaluation of pa tients with ischemic heart disease. Prog Cardiovasc Dis 1969:11:371390. 32 KoppesG. McKicrnan T. Bassan M. Froelicher VF: Treadmill exercise testing. I. C'urr Probl Cardiol 1977; 2:1-44. 33 Nic NH. Hull CH. Jenkins JG. Stainbrenner K. Bent D ll: Statisti cal Package for the Social Sciences, ed 2. New York. McGraw-Hill. 1975, pp 276-300. 34 Engclman L: Stepwise logistic re gression: in Dixon WJ (ed): BMDP Statistical Software. Berkeley, University of California Press. 1983. pp 330-344. 35 Brown MB: Two-way and multiway frequency tables - measures o f asso ciation and the log-linear model: in Dixon WJ (ed): BMDP Statistical Software. Berkeley. University of California Press. 1983. pp 143-206. 36 De Pasquale G. Pinelli G. Tartagni F. Manini GL. Dondi M: Effort an gina pectoris without electrocardio graphic changes in coronary disease patients: Correlations between scin tigraphic and coronav angiographic findings. Int J Cardiol 1986; 12:243— 253.
171
Downloaded by: Univ.of California Berkeley 128.32.10.230 - 7/25/2018 9:47:15 PM
18 Goldschlager N. Cake D, Cohn K: Exercise-induced ventricular ar rhythmias in patients with coronary' artery disease. Their relation to an giographic findings. Am J Cardiol 1973:31:434-440. 19 Hclfant RH. Pine R. Kabde V. Vanka VS: Exercise-related ventric ular premature complexes in coro nary heart disease. Correlations with ischemia and angiographic se verity. Ann Intern Med 1974:80: 589-592. 20 Carvalho A. Amsterdam E. DeMaria A. Mason DT: Réévaluation of exercise electrocardiography in diagnoses of coronary artery’ dis ease: Significance of heart rate at which ST-segmcnt depression oc curs (abstract). Circulation 1976; 54(suppl li):206. 21 Goldschlager N. Selzer A. Cohn K: Treadmill stress tests as indicators of presence and severity of coronary artery disease. Ann Intern Med 1976:85:277-286. 22 Hollenberg M. Zoltick JM. Go M. Yaney SF. Daniels W. Davis RC. Bedynek JL: Comparison of a quan titative treadmill exercise score with standard electrocardiographic crite ria in screening asymptomatic young men for coronary artery dis ease. N Engl J Med 1985:313^600606. 23 Cohn K. Kamm B. Fetcih N. Brand R. Goldschlager N: Use of treadmill score to quantify ischemic response and predict extent of coronary dis ease. Circulation 1979:59:286-296.
