- Email: [email protected]
Comparison of the diagnostic accuracy of different methods of measurement of heart rate-adjusted ST-segment depression during exercise testing for identification of coronary artery disease
Corn arison of the Diagnostic Accuracy of Di iI erent Methods of Measurement of Heart Rate-Adjusted ST-Segment Depression During Exercise Testing for Identification of Coronary Artery Disease Bernard
Daniel Herpin, MD, JBr6me Ferrandis, MD, Patrick Borderon, MD, Gaudeau, MD, Stephanie Ragot, Sandrine Gigbn, and Jean Demqnge,
MD
Within a population of 160 consecutive symptomatic patients who all had undergone catheterization (80 with 21 stenosis >50%), we compared the accuracy of different computerized measurements of the exerciseinduced changes in ST-segment: (1) the standard criterion (20.1 mV flat/downsloping ST depression or 20.15 mV upsloping depression, both 60 ms after the J point); (2) heart rate (HR)-ad’usted ST-segment depression (ST/HR index measur eci at 0, 20, 40, 60, and 80 ms from the J point); (3) the HR-adjusted ST integral (ST/HR integral measured from 0 to 40 ms and from 40 to 80 ms after the J point). None of the ST/HR indexes or integrals were found to have a significantly greater sensitivity than the standard criterion. On the contrary, all ST/HR indexes and integrals showed a higher specificity (0.78 to 0.89) than did the standard criterion
(0.65); moreover, the earlier the measurement within the repolarization phase, the better the overall accuracy: 0.71 for the standard criterion, 0.83 (pcO.OOl), 0.80 (p
omputer-based processing of ST-segment changes C during exercise has allowed some accurate measurements to be obtained from original data. This article
darone; likewise, all patients with unstable angina, patent hypertrophic cardiomyopathy, or valvular disease were excluded, as well as those with left bundle branch block or atrial fibrillation. However, patients with systemic hypertension, left ventricular hypertrophy, and/or left ventricular repolarization abnormalities (whatever their origin) were not excluded. Exercise test: An exercise electrocardiogram was obtained while the patients were pedaling an electronically braked cycle ergometer. The workload was increased by 30 W every 3 minutes until clinical symptoms (dyspnea, exhaustion, chest pain, and so forth) or significant electrocardiographic modifications occurred. Patients who both did not achieve a maximal HR 285% of the theoretical maximal HR (220 - age) and did not exhibit a pathologic standard electrocardiographic criterion were excluded from the present analysis. Computer-calculated ST-segment amplitudes were obtained in leads VF, V,, and V, at rest (both in recumbent and sitting positions) after each minute of exercise, at peak exercise, and after each minute of recovery. The most negative value obtained for each patient was taken into account for analysis. Exercise tests were evaluated using a software specifically developed by 2 team members (SG, JD) for calculations of the standard electrocardiographic criterion and the ST/HR indexes, according to the recommendations of the American Heart Association.lo,u Measurements were based on arithmetic averaging of the previous 1.5seconds of normal complexes; sampling rate was 500 Hz, and time discrepancy was
focuses on heart rate (HR)-adjusted ST-segment depression (ST/HR index) and ST/HR integral. Conflicting data may be found in published reports1-9 regarding both the calculation and the diagnostic value of these measures for the identification of coronary artery disease. Accordingly, the present study was designed to compare, within a representative population of consecutive symptomatic patients, the accuracy of the standard criterion with that of the ST/HR index measured at 0, 20, 40, 60, or 80 ms from the J point, and the ST/HR integral measured from 0 to 40 ms and from 40 to 80 ms after the J point.
METHODS Study group: One hundred sixty consecutive patients entered the study (113 men and 47 women; mean age + SD 60 i 9 years; range 34 to 76). All patients were referred to our clinic for evaluation of chest pain; they performed a symptom-limited exercise test and subsequently underwent coronary angiography within a median interval of 1.5 days. None of them had a prior myocardial infarction or were taking digitalis or amioFrom the Service Cardiologie B, Centre HospitabUniversitaire, and Unite Genie Physiologique, Faculte des Sciences, Poitiers, France. Manuscript received June 2, 1995; revised manuscript received and accepted August 2S, 1995. Address or reprints: Daniel Herpin, MD, Service Cardiologie B, Centre HospitaloUniversitaire, 8602 1 Poitiers Cedex, France.
CORONARY
ARTERY
DISEASE/HEART
RATE-ADJUSTED
ST-SEGMENT
DEPRESSION
1147
ms after the J point, because we found this mode of calculation to provide a Group 1 Group 2 Group 3 higher accuracy than did c&ulation (n = 48) (n ‘= 80) p Value Characteristics (I-I = 32) made 80 ms after the J point (data not yet published). 45 (56) 0.0003 Men 41 (85) 27 (84) 61 i: 10 60 * 6 59~88. NS ST/HR indexes were evaluated by be (yd Systemic 17 (35) 12 (37) 31 (39) NS dividing the difference between endhypertension exercise and resting ST-segment deviLeft ventricular 16 (33) 21 (26) NS 8 (251 ation by the exercise-induced change hypertrophy in HR. Baseline and exercise ST-segRepolarization 22 (46) 14 (44) 28 (35) NS abnormalities ment elevations were not normalized Exercise duration (min) 8*3 924 1023 NS to the zero line. Measurements were Peak heart rate (beats/min] 125 zt 18 136 * 19* 152 f 15tt <0.0001 obtained to the nearest 10 pV, every Peak systolic blood 171 i 27 187*27§ 187 ~t2811 0.005 20 ms interval between the J point and pressure (mm Hg) Maximal workload 90 f 34 100 * 41 108 f 3911 0.029 80 ms after the J point (inclusively), achieved (W) with the end of the PR segment as a *Group 2 versus group 1: p
I Patient
Characteristics
and
Exercise
Test Results
L
1.0
0.9 0.8 0.7
FIGURE 1. Receiver-operating characteristic curves of the heart rate (HR)-adjusted STment (ST/HR) indexes cal& gted at 20,60, and 80 ms =? from the J point in 160 con+utive symptomatic patients who had undergone coronary brkty catheterization (sensitivity and specificity of the standard iriterion are shown by the i@ersecfing lines; xaxis is 1 ‘ginus specificity).
0.6 0.5 0.4 0.3 0.2 0.1 0.0
0.0
1148
THE AMERICAN
0.1
0.2
JOURNAL
0.3
0.4 0.5 l-specificity
OF CARDIOLOGY@
0.6
VOL.
76
0.7
DECEMBER
0.8
0.9
1, 1995
1.0
were used. In case of a discrepancy, a third reading was done. The degree of stenosis was defined as the greatest reduction in percentage of lumen diameter in any view compared with the nearest normal segment. Coronary artery disease was considered significant when the degree of obstruction was evaluated at 250%. Left main artery stenosis was scored as a 2-vessel disease. Accordingly, we found 32 patients with l-vessel disease, 26 patients with 2-vessel disease, and 22 patients with 3vessel disease; 13 patients had left main coronary artery disease. Statistical methods: Comparison of test sensitivity and specificity of the ST/HR indexes with those of the standard criterion was performed using McNemar’s modification of the chi-square method for paired proportions; the p value was calculated from the binomial distribution when necessary. Test accuracy of each ST/HR index was also compared using receiver-operating characteristic (ROC) curve analysis. ROC curves were generated using StatView Software (512 TM, Calabasas, California) and compared statistically by means of a univariate Z score test of the difference between the areas under 2 curves.13T14Between-group comparisons were performed using l-way analysis of variance for continuous variables and the &i-square test for discrete variables. Data are given as mean + SD. For all comparisons, a p value co.05 was considered significant.
RESULTS Group characteristics: Table I lists the clinical characteristics of the study groups as well as the exercise test results. Of note, group 3 patients achieved a significantly higher peak HR than did both group 1 and 2 patients; the area under the ROC curve of peak HR was as high as 0.81. With a partition value of 85% of the theoretical maximal HR, peak HR yielded a specificity of 0.88 and a sensitivity of 0.59. Effect of measurement ment/heart rate indexes:
timing on accuracy
of ST-seg-
Table II lists the respective accuracies of the 5 ST/HR indexes according to the cutoff values obtained from examination of the ROC curves. The ROC curves of the measurements performed 20,60, and 80 ms after the J point are shown in Figure 1; the ROC curves of the ST/HR indexes calculated 0 and 40 ms after the J point are not shown because they were quite similar to that of the ST/I-IR index measured at 20 ms. The areas under the curves are displayed in Table III. Consistently, there were no significant differences between the measurements obtained at 0, 20, or 40 ms after the J point, which clearly appeared to be more accurate than the measurements performed at 60 or 80 ms. The improvement in diagnostic accuracy was obviously linked to a better specificity compared with the standard criterion. At a matched sensitivity of 0.76 (which was that of the standard criterion), the specificity of the 5 ST/I-IR indexes was 0.89, 0.86, 0.79, 0.71, and 0.59, respectively. At a matched specificity of 0.65, the sensitivity of the 5 ST/HR indexes was 0.83,0.83,0.83,0.79, and 0.74, respectively. Effect of measurement interval on accuracy of the STsegment/heart rate integral: Table II lists the accuracies
of ST/HR integrals calculated from 0 to 40 ms and from CORONARY
ARTERY DISEASE/HEART
RATE-ADJUSTED
ST-SEGMENT
DEPRESSION
1149
TABLE III ST/HR
Receiver-Operating Integral
Characteristic
Standard Criterion
Variable
ST/HR
0.75
Area under the curve 95% Confidence interval
Curve
0
o.a5*t
0.68-0.83
Areas
ST/HR
of the Standard
20
ST/HR
0.87$§
0.79-0.91
0.81-0.92
*p ~0.01 versus standard criterion; tp co.04 versus rate Abbreviations and explanations as in Table II.
80.
Criterion
40
and
ST/HR
0.86*t
0.82
0.80-0.92
0.75-0.88
ST-segment/heart
rate 80;
40 to 80 ms after the J point. Figure 2 displays the ROC curves of these parameters and areas under the ROC curves are listed in Table III. The integral calculated in the early phase of repolarization was significantly more accurate and more specific than the standard criterion. Severity of coronary obstruction: All ST/FIR indexes and integrals accurately distinguished between the patients with 2- or 3-vessel disease and those with l-vessel disease, whereas the standard criterion could only distinguish patients with normal angiograms from those with coronary obstruction.
DISCUSSION The major findings of our study were: (1) ST/HR measurements yielded both a better specificity and a better accuracy than did the standard criterion. (2) The earlier the measurement within the repolarization phase, the greater the overall accuracy of both ST/HR indexes and ST/HR integrals. There are some points of contention within the large body of literature devoted to this topical issue. Of course, the diagnostic value of all indexes to be considered will be strongly influenced by 4 main factors: (1) exercise test procedure, (2) definition of significant coronary stenosis, (3) timing and method of measurement of ST-segment changes, and (4) prevalence of
fp ~0.004
Different
60
Measurements
ST/HR
80
of ST/HR
ST/HR Integral O-40
0.78 0.71-0.85
versus
standard
criterion;
Index
ST/HR Integral 40-80
0.86*t
0.82
0.80-0.92 sp eO.004
and
0.75-0.88 versus
ST-segment/heart
coronary artery disease and associated pathology witbin the patient groups selected for a given study. Exercise test protocol: Some procedural differences may be found among the different investigators3: treadmill testing was performed according either to the Bruce protocol’,* or to the gradual Cornell protocol.5,6,9 Our protocol, using an exercise bicycle, was more like the latter than the former. These procedural differences could result in different maximal HRs achieved, a variable whose discriminating value has been clearly emphasized both in the present study and some others.15 Definition of significant coronary stenosis: Either a 50%1*5,6,9or a 70%2,7 cutoff value for coronary narrowing was used in the literature. In our study group, a second grouping using a 70%, rather than 50%, cutoff value resulted in a slight decrease in the specificities of standard criterion and ST/FIR integral at 0 to 40 ms to 0.64 (57 of 89) and 0.85 (76 of 89), respectively, whereas their sensitivities increased to’O.80 (57 of 71) and 0.79 (56 of 71), respectively. Measurement of ST-segment changes: The standard criterion itself was not uniformly defined in the literature: the measurement was performed either at 60 ms2,5,7 or at 80 ms1y9 after the J point. The performance of the standard criterion seemed to be unaffected by the time
1.0 0.9 0.8 0.7 0.6 >I -5
0.5 0.4 0.3
.r I ri
0.2 0.1 0.0
FIGURE 2. Receiveroperatin characteristic curves o 3 the heart rate (HR)-ad'~~kd STsegment (ST I HR) integrab in 160 consecutive symptomatic patients who had undergone coronary artery catheterization p$~~~pJ+f$ criterion are shown by infersechg lines; xaxis is 1 minus specificity).
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
l-specificity
1150
THE AMERICAN
JOURNAL
OF CARDIOLOGY@
VOL.
76
DECEMBER
1, 1995
0.9
1.0
chosen for measurement if the ST-segment was flat or downsloping, 3,5 but the measurement of the depressed ST-segment was found to be more accurate at 60 ms when it was upsloping. Regarding the upsloping depression, it was either not taken into account,2T5 or had to achieve different levels (120 I.LV [Morise and Duva19] or 150 p,V [Okin et a16]) for being considered as pathologic. In our experience, the exclusion of upsloping depressions resulted in an increase in the specificity of the standard criterion to 0.71, along with a decrease in its sensitivity to 0.65. Regarding the ST/HR index, baseline ST elevation was not neglected in our study, whereas it was normalized to zero baseline by most investigators2,5-7; interestingly, Okin et all6 showed that incorporation of resting ST elevation did not alter the performance of HR-adjusted criteria for the identification of coronary obstruction. Patient
&Ction:
It is Worth
noting
the
%%
PreVa-
lence of coronary artery disease in our consecutively selected patients, a rate highly appropriate for evaluating the diagnostic capabilities of a test.17 In addition, all patients were referred for testing because of symptoms of coronary artery disease, with some of them having associated abnormalities such as hypertension or electrocardiographic evidence of left ventricular hypertrophy, or both. These characteristics account for the specially low specificity (0.65) yielded by the standard criterion in our study; after exclusion of the patients with left ventricular hypertrophy, specificity increased to 0.74, a figure close to that reported in Gianrossi’s meta-analysis.is Our population was undoubtedly more representative than those analyzed in other studies, where patients with coronary obstruction often presented with severe lesions or previous myocardial infarction,2,6,9 and where normal subjects were free of symptoms and did not undergo catheterization for ethical reasons.5,6,9 Moreover, patients with left ventricular hypertrophy or baseline ST-T changes, or both, were excluded by some investigators2,9 and not by others. L5A Our conclusion is that the divergent results obtained by the different inves-
CORONARY
tigators may be explained by differences in population selection criteria; the originality of our study is related to the consecutive inclusion of representative patients in whom the diagnosis of coronary artery disease was questionable. 1. Detrano R, Salcedo E, Passalacqua M, Friis R. Exercise electrocardiographic variables: a critical appraisal. JAm Coil Cardiol 1986;8:836-847. 2. Lachterman B, Lehmann KG, Detrano R. Neutel J, Froelicher VF. Comparison of ST segment/heart rate index to standard criteria for analysis of exercise electrocardiogram. Circulalion 1990;82:44-50. ‘3. Sheffield LT. Upsloping ST segments. Easy to measure, hard to agree upon. Circulation 1g91.84,42~428.
,
4. Bobbio M, Detrano R. A lesson from the controversy about heart rate adjustment of ST segment depression. Circulation 1991;84: 1410-1413. 5. Okin PM, Bergman G, Kligtield P. Effect of ST measurement point on performance of standard and heart rate-adjusted ST segment criteria for the identification of coronary artery disease. Circulation 1991;84:57-66. 6. Okin PM, Bergman G, Kligfield P. Measurement variables for optimal performance of the ST integral. JAm Coil Car&2 1993;22:168-174. 7. Ribisl PM, Liu J, MOuSa I, Herbert WG, Miranda CP, Froning JN, Froelicher VF. Comparison of computer ST criteria for diagnosis of severe coronary artery disease. Am J Cardiol 1993;71:546-551. 8. Rodriguez M, Froning J, Froelicher VF. ST 0 or ST 60? Am Heart .J 1993;126:
752-754.
9. Morise P, Duval RD. Diagnostic accuracy of heart rate-adjusted ST segments compared with standard ST-segment criteria. Am J Cardiol 1995;75:118-121. 10. American Heart Association Committee on Electrocardiography. Recommendations for standardization of leads and of specifications for instruments in ECGNCG. Circulation 1975;52: 1 l-25. 11. Bhargava V, Watanabe K, Froelicher VF. Progress in computer analysis of the exercise electrocardiogram. Am J Cardiol 1981;47:1143-1151. 12. Ellestad MH. Stress Testing. Principles and Practice. 3rd ed. Philadelphia: FA
Davis, 1986:234-236. 13. Hanley JA, McNeil BJ. The meaning
and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143:29-36. 14. Hanley JA, McNeil BJ. A method of comparing the areas under receiver opersting curves derived from the same cases. Radiology 1983;148:839-843. 15. Detry JMR, Lnwaert RA, Rousseau MF, Brasseur LA, Melin JA, Brohet CR, Denwael Barchy L, Fester R, Vawbuutseler RJ. Diagnostic value of computerized exercise testing in men without previous myocardial infarction. A multivariate cornpartmental and probabilistic approach. Eur Heart J 1985;63227-238. 16. Okin PM, Chen J, Kligfield P. Effect of baseline ST segment elevation on test performance of standard and heart rate-adjusted ST segment depression criteria. Am Heart J 1990;119:128&1286. 17. Philbrick JT, Horwitz RI, Feinstein AR. Methodologic problems of exercise testing for coronary artery disease: groups, analysis and bias. Am J Cardiol1980;46:
807-812. 18. Giaorossi
R, Detrano R, Mulvihill D, Lehman K, Dubach P, Colombo A, McArthur D. Fro&her V. Exercise-induced ST depression in the diagnosis of coronary tiery disease. A meta-analysis. Circulation 1989;80:87-98.
ARTERY DISEASE/HEART
RATE-ADJUSTED
ST-SEGMENT
DEPRESSION
1151
Daniel Herpin, MD, JBr6me Ferrandis, MD, Patrick Borderon, MD, Gaudeau, MD, Stephanie Ragot, Sandrine Gigbn, and Jean Demqnge,
MD
Within a population of 160 consecutive symptomatic patients who all had undergone catheterization (80 with 21 stenosis >50%), we compared the accuracy of different computerized measurements of the exerciseinduced changes in ST-segment: (1) the standard criterion (20.1 mV flat/downsloping ST depression or 20.15 mV upsloping depression, both 60 ms after the J point); (2) heart rate (HR)-ad’usted ST-segment depression (ST/HR index measur eci at 0, 20, 40, 60, and 80 ms from the J point); (3) the HR-adjusted ST integral (ST/HR integral measured from 0 to 40 ms and from 40 to 80 ms after the J point). None of the ST/HR indexes or integrals were found to have a significantly greater sensitivity than the standard criterion. On the contrary, all ST/HR indexes and integrals showed a higher specificity (0.78 to 0.89) than did the standard criterion
(0.65); moreover, the earlier the measurement within the repolarization phase, the better the overall accuracy: 0.71 for the standard criterion, 0.83 (pcO.OOl), 0.80 (p
omputer-based processing of ST-segment changes C during exercise has allowed some accurate measurements to be obtained from original data. This article
darone; likewise, all patients with unstable angina, patent hypertrophic cardiomyopathy, or valvular disease were excluded, as well as those with left bundle branch block or atrial fibrillation. However, patients with systemic hypertension, left ventricular hypertrophy, and/or left ventricular repolarization abnormalities (whatever their origin) were not excluded. Exercise test: An exercise electrocardiogram was obtained while the patients were pedaling an electronically braked cycle ergometer. The workload was increased by 30 W every 3 minutes until clinical symptoms (dyspnea, exhaustion, chest pain, and so forth) or significant electrocardiographic modifications occurred. Patients who both did not achieve a maximal HR 285% of the theoretical maximal HR (220 - age) and did not exhibit a pathologic standard electrocardiographic criterion were excluded from the present analysis. Computer-calculated ST-segment amplitudes were obtained in leads VF, V,, and V, at rest (both in recumbent and sitting positions) after each minute of exercise, at peak exercise, and after each minute of recovery. The most negative value obtained for each patient was taken into account for analysis. Exercise tests were evaluated using a software specifically developed by 2 team members (SG, JD) for calculations of the standard electrocardiographic criterion and the ST/HR indexes, according to the recommendations of the American Heart Association.lo,u Measurements were based on arithmetic averaging of the previous 1.5seconds of normal complexes; sampling rate was 500 Hz, and time discrepancy was
focuses on heart rate (HR)-adjusted ST-segment depression (ST/HR index) and ST/HR integral. Conflicting data may be found in published reports1-9 regarding both the calculation and the diagnostic value of these measures for the identification of coronary artery disease. Accordingly, the present study was designed to compare, within a representative population of consecutive symptomatic patients, the accuracy of the standard criterion with that of the ST/HR index measured at 0, 20, 40, 60, or 80 ms from the J point, and the ST/HR integral measured from 0 to 40 ms and from 40 to 80 ms after the J point.
METHODS Study group: One hundred sixty consecutive patients entered the study (113 men and 47 women; mean age + SD 60 i 9 years; range 34 to 76). All patients were referred to our clinic for evaluation of chest pain; they performed a symptom-limited exercise test and subsequently underwent coronary angiography within a median interval of 1.5 days. None of them had a prior myocardial infarction or were taking digitalis or amioFrom the Service Cardiologie B, Centre HospitabUniversitaire, and Unite Genie Physiologique, Faculte des Sciences, Poitiers, France. Manuscript received June 2, 1995; revised manuscript received and accepted August 2S, 1995. Address or reprints: Daniel Herpin, MD, Service Cardiologie B, Centre HospitaloUniversitaire, 8602 1 Poitiers Cedex, France.
CORONARY
ARTERY
DISEASE/HEART
RATE-ADJUSTED
ST-SEGMENT
DEPRESSION
1147
ms after the J point, because we found this mode of calculation to provide a Group 1 Group 2 Group 3 higher accuracy than did c&ulation (n = 48) (n ‘= 80) p Value Characteristics (I-I = 32) made 80 ms after the J point (data not yet published). 45 (56) 0.0003 Men 41 (85) 27 (84) 61 i: 10 60 * 6 59~88. NS ST/HR indexes were evaluated by be (yd Systemic 17 (35) 12 (37) 31 (39) NS dividing the difference between endhypertension exercise and resting ST-segment deviLeft ventricular 16 (33) 21 (26) NS 8 (251 ation by the exercise-induced change hypertrophy in HR. Baseline and exercise ST-segRepolarization 22 (46) 14 (44) 28 (35) NS abnormalities ment elevations were not normalized Exercise duration (min) 8*3 924 1023 NS to the zero line. Measurements were Peak heart rate (beats/min] 125 zt 18 136 * 19* 152 f 15tt <0.0001 obtained to the nearest 10 pV, every Peak systolic blood 171 i 27 187*27§ 187 ~t2811 0.005 20 ms interval between the J point and pressure (mm Hg) Maximal workload 90 f 34 100 * 41 108 f 3911 0.029 80 ms after the J point (inclusively), achieved (W) with the end of the PR segment as a *Group 2 versus group 1: p
I Patient
Characteristics
and
Exercise
Test Results
L
1.0
0.9 0.8 0.7
FIGURE 1. Receiver-operating characteristic curves of the heart rate (HR)-adjusted STment (ST/HR) indexes cal& gted at 20,60, and 80 ms =? from the J point in 160 con+utive symptomatic patients who had undergone coronary brkty catheterization (sensitivity and specificity of the standard iriterion are shown by the i@ersecfing lines; xaxis is 1 ‘ginus specificity).
0.6 0.5 0.4 0.3 0.2 0.1 0.0
0.0
1148
THE AMERICAN
0.1
0.2
JOURNAL
0.3
0.4 0.5 l-specificity
OF CARDIOLOGY@
0.6
VOL.
76
0.7
DECEMBER
0.8
0.9
1, 1995
1.0
were used. In case of a discrepancy, a third reading was done. The degree of stenosis was defined as the greatest reduction in percentage of lumen diameter in any view compared with the nearest normal segment. Coronary artery disease was considered significant when the degree of obstruction was evaluated at 250%. Left main artery stenosis was scored as a 2-vessel disease. Accordingly, we found 32 patients with l-vessel disease, 26 patients with 2-vessel disease, and 22 patients with 3vessel disease; 13 patients had left main coronary artery disease. Statistical methods: Comparison of test sensitivity and specificity of the ST/HR indexes with those of the standard criterion was performed using McNemar’s modification of the chi-square method for paired proportions; the p value was calculated from the binomial distribution when necessary. Test accuracy of each ST/HR index was also compared using receiver-operating characteristic (ROC) curve analysis. ROC curves were generated using StatView Software (512 TM, Calabasas, California) and compared statistically by means of a univariate Z score test of the difference between the areas under 2 curves.13T14Between-group comparisons were performed using l-way analysis of variance for continuous variables and the &i-square test for discrete variables. Data are given as mean + SD. For all comparisons, a p value co.05 was considered significant.
RESULTS Group characteristics: Table I lists the clinical characteristics of the study groups as well as the exercise test results. Of note, group 3 patients achieved a significantly higher peak HR than did both group 1 and 2 patients; the area under the ROC curve of peak HR was as high as 0.81. With a partition value of 85% of the theoretical maximal HR, peak HR yielded a specificity of 0.88 and a sensitivity of 0.59. Effect of measurement ment/heart rate indexes:
timing on accuracy
of ST-seg-
Table II lists the respective accuracies of the 5 ST/HR indexes according to the cutoff values obtained from examination of the ROC curves. The ROC curves of the measurements performed 20,60, and 80 ms after the J point are shown in Figure 1; the ROC curves of the ST/HR indexes calculated 0 and 40 ms after the J point are not shown because they were quite similar to that of the ST/I-IR index measured at 20 ms. The areas under the curves are displayed in Table III. Consistently, there were no significant differences between the measurements obtained at 0, 20, or 40 ms after the J point, which clearly appeared to be more accurate than the measurements performed at 60 or 80 ms. The improvement in diagnostic accuracy was obviously linked to a better specificity compared with the standard criterion. At a matched sensitivity of 0.76 (which was that of the standard criterion), the specificity of the 5 ST/I-IR indexes was 0.89, 0.86, 0.79, 0.71, and 0.59, respectively. At a matched specificity of 0.65, the sensitivity of the 5 ST/HR indexes was 0.83,0.83,0.83,0.79, and 0.74, respectively. Effect of measurement interval on accuracy of the STsegment/heart rate integral: Table II lists the accuracies
of ST/HR integrals calculated from 0 to 40 ms and from CORONARY
ARTERY DISEASE/HEART
RATE-ADJUSTED
ST-SEGMENT
DEPRESSION
1149
TABLE III ST/HR
Receiver-Operating Integral
Characteristic
Standard Criterion
Variable
ST/HR
0.75
Area under the curve 95% Confidence interval
Curve
0
o.a5*t
0.68-0.83
Areas
ST/HR
of the Standard
20
ST/HR
0.87$§
0.79-0.91
0.81-0.92
*p ~0.01 versus standard criterion; tp co.04 versus rate Abbreviations and explanations as in Table II.
80.
Criterion
40
and
ST/HR
0.86*t
0.82
0.80-0.92
0.75-0.88
ST-segment/heart
rate 80;
40 to 80 ms after the J point. Figure 2 displays the ROC curves of these parameters and areas under the ROC curves are listed in Table III. The integral calculated in the early phase of repolarization was significantly more accurate and more specific than the standard criterion. Severity of coronary obstruction: All ST/FIR indexes and integrals accurately distinguished between the patients with 2- or 3-vessel disease and those with l-vessel disease, whereas the standard criterion could only distinguish patients with normal angiograms from those with coronary obstruction.
DISCUSSION The major findings of our study were: (1) ST/HR measurements yielded both a better specificity and a better accuracy than did the standard criterion. (2) The earlier the measurement within the repolarization phase, the greater the overall accuracy of both ST/HR indexes and ST/HR integrals. There are some points of contention within the large body of literature devoted to this topical issue. Of course, the diagnostic value of all indexes to be considered will be strongly influenced by 4 main factors: (1) exercise test procedure, (2) definition of significant coronary stenosis, (3) timing and method of measurement of ST-segment changes, and (4) prevalence of
fp ~0.004
Different
60
Measurements
ST/HR
80
of ST/HR
ST/HR Integral O-40
0.78 0.71-0.85
versus
standard
criterion;
Index
ST/HR Integral 40-80
0.86*t
0.82
0.80-0.92 sp eO.004
and
0.75-0.88 versus
ST-segment/heart
coronary artery disease and associated pathology witbin the patient groups selected for a given study. Exercise test protocol: Some procedural differences may be found among the different investigators3: treadmill testing was performed according either to the Bruce protocol’,* or to the gradual Cornell protocol.5,6,9 Our protocol, using an exercise bicycle, was more like the latter than the former. These procedural differences could result in different maximal HRs achieved, a variable whose discriminating value has been clearly emphasized both in the present study and some others.15 Definition of significant coronary stenosis: Either a 50%1*5,6,9or a 70%2,7 cutoff value for coronary narrowing was used in the literature. In our study group, a second grouping using a 70%, rather than 50%, cutoff value resulted in a slight decrease in the specificities of standard criterion and ST/FIR integral at 0 to 40 ms to 0.64 (57 of 89) and 0.85 (76 of 89), respectively, whereas their sensitivities increased to’O.80 (57 of 71) and 0.79 (56 of 71), respectively. Measurement of ST-segment changes: The standard criterion itself was not uniformly defined in the literature: the measurement was performed either at 60 ms2,5,7 or at 80 ms1y9 after the J point. The performance of the standard criterion seemed to be unaffected by the time
1.0 0.9 0.8 0.7 0.6 >I -5
0.5 0.4 0.3
.r I ri
0.2 0.1 0.0
FIGURE 2. Receiveroperatin characteristic curves o 3 the heart rate (HR)-ad'~~kd STsegment (ST I HR) integrab in 160 consecutive symptomatic patients who had undergone coronary artery catheterization p$~~~pJ+f$ criterion are shown by infersechg lines; xaxis is 1 minus specificity).
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
l-specificity
1150
THE AMERICAN
JOURNAL
OF CARDIOLOGY@
VOL.
76
DECEMBER
1, 1995
0.9
1.0
chosen for measurement if the ST-segment was flat or downsloping, 3,5 but the measurement of the depressed ST-segment was found to be more accurate at 60 ms when it was upsloping. Regarding the upsloping depression, it was either not taken into account,2T5 or had to achieve different levels (120 I.LV [Morise and Duva19] or 150 p,V [Okin et a16]) for being considered as pathologic. In our experience, the exclusion of upsloping depressions resulted in an increase in the specificity of the standard criterion to 0.71, along with a decrease in its sensitivity to 0.65. Regarding the ST/HR index, baseline ST elevation was not neglected in our study, whereas it was normalized to zero baseline by most investigators2,5-7; interestingly, Okin et all6 showed that incorporation of resting ST elevation did not alter the performance of HR-adjusted criteria for the identification of coronary obstruction. Patient
&Ction:
It is Worth
noting
the
%%
PreVa-
lence of coronary artery disease in our consecutively selected patients, a rate highly appropriate for evaluating the diagnostic capabilities of a test.17 In addition, all patients were referred for testing because of symptoms of coronary artery disease, with some of them having associated abnormalities such as hypertension or electrocardiographic evidence of left ventricular hypertrophy, or both. These characteristics account for the specially low specificity (0.65) yielded by the standard criterion in our study; after exclusion of the patients with left ventricular hypertrophy, specificity increased to 0.74, a figure close to that reported in Gianrossi’s meta-analysis.is Our population was undoubtedly more representative than those analyzed in other studies, where patients with coronary obstruction often presented with severe lesions or previous myocardial infarction,2,6,9 and where normal subjects were free of symptoms and did not undergo catheterization for ethical reasons.5,6,9 Moreover, patients with left ventricular hypertrophy or baseline ST-T changes, or both, were excluded by some investigators2,9 and not by others. L5A Our conclusion is that the divergent results obtained by the different inves-
CORONARY
tigators may be explained by differences in population selection criteria; the originality of our study is related to the consecutive inclusion of representative patients in whom the diagnosis of coronary artery disease was questionable. 1. Detrano R, Salcedo E, Passalacqua M, Friis R. Exercise electrocardiographic variables: a critical appraisal. JAm Coil Cardiol 1986;8:836-847. 2. Lachterman B, Lehmann KG, Detrano R. Neutel J, Froelicher VF. Comparison of ST segment/heart rate index to standard criteria for analysis of exercise electrocardiogram. Circulalion 1990;82:44-50. ‘3. Sheffield LT. Upsloping ST segments. Easy to measure, hard to agree upon. Circulation 1g91.84,42~428.
,
4. Bobbio M, Detrano R. A lesson from the controversy about heart rate adjustment of ST segment depression. Circulation 1991;84: 1410-1413. 5. Okin PM, Bergman G, Kligtield P. Effect of ST measurement point on performance of standard and heart rate-adjusted ST segment criteria for the identification of coronary artery disease. Circulation 1991;84:57-66. 6. Okin PM, Bergman G, Kligfield P. Measurement variables for optimal performance of the ST integral. JAm Coil Car&2 1993;22:168-174. 7. Ribisl PM, Liu J, MOuSa I, Herbert WG, Miranda CP, Froning JN, Froelicher VF. Comparison of computer ST criteria for diagnosis of severe coronary artery disease. Am J Cardiol 1993;71:546-551. 8. Rodriguez M, Froning J, Froelicher VF. ST 0 or ST 60? Am Heart .J 1993;126:
752-754.
9. Morise P, Duval RD. Diagnostic accuracy of heart rate-adjusted ST segments compared with standard ST-segment criteria. Am J Cardiol 1995;75:118-121. 10. American Heart Association Committee on Electrocardiography. Recommendations for standardization of leads and of specifications for instruments in ECGNCG. Circulation 1975;52: 1 l-25. 11. Bhargava V, Watanabe K, Froelicher VF. Progress in computer analysis of the exercise electrocardiogram. Am J Cardiol 1981;47:1143-1151. 12. Ellestad MH. Stress Testing. Principles and Practice. 3rd ed. Philadelphia: FA
Davis, 1986:234-236. 13. Hanley JA, McNeil BJ. The meaning
and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143:29-36. 14. Hanley JA, McNeil BJ. A method of comparing the areas under receiver opersting curves derived from the same cases. Radiology 1983;148:839-843. 15. Detry JMR, Lnwaert RA, Rousseau MF, Brasseur LA, Melin JA, Brohet CR, Denwael Barchy L, Fester R, Vawbuutseler RJ. Diagnostic value of computerized exercise testing in men without previous myocardial infarction. A multivariate cornpartmental and probabilistic approach. Eur Heart J 1985;63227-238. 16. Okin PM, Chen J, Kligfield P. Effect of baseline ST segment elevation on test performance of standard and heart rate-adjusted ST segment depression criteria. Am Heart J 1990;119:128&1286. 17. Philbrick JT, Horwitz RI, Feinstein AR. Methodologic problems of exercise testing for coronary artery disease: groups, analysis and bias. Am J Cardiol1980;46:
807-812. 18. Giaorossi
R, Detrano R, Mulvihill D, Lehman K, Dubach P, Colombo A, McArthur D. Fro&her V. Exercise-induced ST depression in the diagnosis of coronary tiery disease. A meta-analysis. Circulation 1989;80:87-98.
ARTERY DISEASE/HEART
RATE-ADJUSTED
ST-SEGMENT
DEPRESSION
1151