44

Comparison of ST Segment/Heart Rate Indexto Standard ST Criteria for Analysis of

Exercise ElectrocardiogramBruce Lachterman, MD, Kenneth G. Lehmann, MD, Robert Detrano, MD, PhD,

Joel Neutel, MD, and Victor F. Froelicher, MD

The objective of our study was to compare the discriminating power of a proposed STsegment/heart rate index with that of a standard method of assessing exercise-induced STsegment depression for diagnosing coronary artery disease. We used a cross-sectional retro-spective analysis of exercise test and coronary angiographic data. The study took place in a1,200-bed Veterans Affairs Medical Center, participants were 328 male patients who hadundergone both a sign and symptom-limited treadmill test and coronary angiography. Thesensitivity of the ST segment/heart rate index was 54% at a cut point of 0.021 mm/(beats/min),corresponding to a specificity of 73%. The standard visual ST segment analysis had a sensitivityof 58%o at this same specificity, which corresponded to an ST segment cut point of 1-mmdepression relative to rest (p=NS). Similarly, for diagnosing three-vessel or left main coronarydisease, no significant difference was found between the sensitivities or the two measurementsat cut points of equivalent specificity. In this consecutive series of patients presenting forroutine clinical testing, the ST segment/heart rate index did not improve the diagnosticaccuracy of the exercise test for identifying the presence or severity of coronary artery diseaserelative to standard visual criteria. (Circulation 1990;82:44-50)

T he exercise-induced depression of the ST seg-ment is an accepted sign for predicting thepresence of coronary artery disease and an

aid for the decision to perform coronary angiography.'Several investigators have shown an improvement inthe diagnostic accuracy of ST depression by using avariety of heart rate-adjustment schemes.2-12 Some ofthese methods are laborious or require computeriza-tion. Some groups, however, have noted superioraccuracy by using a simple method of heart rateadjustment.25,6,8'2 The purpose of this study was toassess this approach for detecting any angiographiccoronary disease as well as identifying left main coro-nary or three-vessel disease.

MethodsPopulation

In our laboratory, 2,044 patients, approximately30% of whom had coronary angiography, underwentroutine clinical exercise testing from 1985 to 1987.

From the Long Beach VA Medical Center, Long Beach,California.

Supported in part by a grant from the HSRD Veterans Affairs.Address for reprints: Victor F. Froelicher, MD, Chief, Cardiol-

ogy (lllC), Long Beach Veterans Administration Medical Center,5901 East Seventh Street, Long Beach, CA 90822.

Received June 26, 1989; revision accepted February 23, 1990.

After excluding women (less than 2% of our popu-lation), patients with previous revascularization pro-cedures, and patients with left bundle branch block,328 patients remained. Most were referred for testingbecause of chest pain syndromes; the remainder weretested for functional capacity evaluation or miscella-neous other reasons.

See p 302Exercise TestThe exercise test was performed by using a stan-

dard progressive protocol. All tests were sign andsymptom-limited maximal tests using recommendedcriteria for termination; fatigue or chest pain werethe reasons for stopping in the majority. The Borgscale of perceived exertion and systolic blood pres-sure was recorded for each stage, and METsachieved were estimated from the treadmill speedand grade.6 The treadmill was stopped abruptly atthe completion of exercise, and the patient wasplaced supine within 1 minute of stopping. Electro-cardiographic recording continued for at least 6minutes into recovery or until electrocardiographicchanges stabilized.Electrocardiographic MeasurementsQ waves were considered diagnostic if they

occurred in two adjacent leads, and were 40 msec in

by guest on July 11, 2018http://circ.ahajournals.org/

Dow

nloaded from

Lachterman et al ST Segment/Heart Rate Index 45

TABLE 1. Clinical Features of Study Population

Patients without Patients with CADFeatures All patients (n) CAD (n) n 1-2 VD 3 VD/LM p*

Patients (n) (%) 328 (100) 80 (24) 248 (76) 183 (56) 65 (20)Age (yr) 59±8.8 57±8.9 60±8.7 59±8.7 63±8.0 0.0004Chest pain description (n) (%)

Typical angina 177 (54) 23 (29) 154 (62) 116 (63) 38 (59) 0.0001Atypical chest pain 77 (23) 32 (40) 45 (18) 34 (19) 11 (17) ...

No chest pain 74 (23) 25 (31) 49 (20) 33 (18) 16 (25) ...History of previous MI (n) (%) 141 (43) 21 (26) 120 (48) 89 (49) 31 (48) <0.0001Medication used (n) (%)

a-blockers 127 (39) 17 (21) 110 (44) 81 (44) 29 (45) <0.0001Digoxin 32 (10) 14 (18) 18 (7) 13 (7) 5 (8) 0.03Calcium antagonists 115 (35) 19 (24) 96 (39) 74 (40) 22 (34) 0.03

Q waves at rest (n) (%) 131 (40) 16 (20) 115 (46) 81 (44) 34 (52) <0.0001Anterior 41 (13) 6 (8) 35 (14) 26 (14) 9 (14) NSInferior 92 (28) 10 (13) 82 (33) 58 (32) 24 (37) <0.0001Lateral 14 (4) 2 (3) 12 (5) 8 (4) 4 (6) NS

Values for age are mean± SD. CAD, coronary artery disease; 1-2 VD, one- or two-vessel coronary artery disease; 3VD/LM, triple-vessel disease, left main coronary artery disease, or both; MI, myocardial infarction; NS, not significant.

*Statistical significance (p) for comparison of patient groups using analysis of variance.

duration and 25% in amplitude of the following Rwave. The PR segment was considered the isoelectricline, and ST measurements were made at the J-junction. The J-junction was used to avoid overlapwith the T wave at high heart rates although we havedemonstrated little change in classification betweenthis point and 70 msec later.13 If the ST segmentlevel at standing rest was above the isoelectric line(early repolarization), the resting baseline ST seg-ment level was considered zero; if it was below theisoelectric line, the negative value was used. DeltaST (AST) was calculated as the resting standingvalue (determined as previously stated, with earlyrepolarization considered zero) subtracted from thevalue at maximal exercise without slope considered.Delta heart rate (AHR) was calculated as standingpreexercise heart rate subtracted from the maximalheart rate value during exercise. All 12 leads wereexamined with the lead of maximum change usedfor analysis. The greatest change usually occurredin the lateral precordial leads (V4, Vs, and V6). Forthe standard interpretation, ST segment changes inrecovery were included; however, calculation of themaximal AST segment value only consideredchanges occurring during exercise. For standardcriteria, the ST segment response was only consid-ered abnormal if it was horizontal or downslopingover the 60 msec after the J-junction.

AngiographyCoronary angiograms were performed in the stan-

dard manner. Stenoses of at least 75% of the diam-eter of an adjacent normal arterial segment wereconsidered significant. The number of vessels dis-eased was calculated as one for the left anteriordescending artery, diagonal artery, or both, one for

the right coronary artery, and one for the circumflexartery, obtuse marginal artery, or both.

Data AnalysisContinuous variables are described as the mean± 1

SD. Two-tailed Student's t tests were used to com-pare means of groups for continuous data, and x2 andMcNemar tests were used to compare uncorrelatedand correlated properties, respectively. Sensitivitiesand specificities are given with their 95% confidenceintervals.

Receiver operating characteristic (ROC) curveswere also computed and compared using the methodof Hanley and McNeil.14 A probability (p) value of0.05 was used as the level of statistical significance.

ResultsTable 1 describes the clinical characteristics of the

study sample grouped by the number of major vesselswith significant coronary stenosis. Of the 328patients, 80 (24%) had no such obstruction, 183(56%) had one- or two-vessel disease, and 65 (20%)had left main or three-vessel disease, or both. Themean age in the sample was 59 years. Slightly morethan half of the patients presented with typicalangina pectoris, and approximately 40% of thepatients had a history of previous myocardial infarc-tion (MI) or had electrocardiograms (ECGs) withdiagnostic Q waves. Medication status during test-ing was usually not altered from that already pre-scribed. Thus, 39% of the patients were taking,B-blockers and 10% were taking digoxin. For thefeatures listed in Table 1, no significant differenceswere found between patients with one- or two-vessel disease and those with three-vessel or leftmain disease; however, the last column of the table

by guest on July 11, 2018http://circ.ahajournals.org/

Dow

nloaded from

46 Circulation Vol 82, No 1, July 1990

TABLE 2. Effect of Q Waves and History of Myocardial Infarction on Prevalence of Angiographic Coronary ArteryDisease

Q waves History of MI

Yes No Yes No(n =131) (n =-197) (n = 141) (n = 187)

Any significant disease (%) 115 (88) 133 (68)* 120 (85) 128 (68)*Three-vessel or left main disease (%) 34 (26) 31 (16)* 31 (22) 34 (18)

MI, myocardial infarction.*p<0.O1.(These values are the percentage of patients with abnormal angiographic findings according to Q wave or history of

MI.)

shows most of these features were significantlydifferent for the subjects with no vessel diseaseversus those patients with any disease.

Myocardial Infarction by History/Q WavesTable 2 demonstrates the relation between diag-

nostic Q waves or previous history of MI and thepresence of any angiographic coronary disease, or

three-vessel or left main disease. Because of silentand non-Q wave MIs and regression of Q waves, theQ wave and MI classifications did not agree completely.The presence of either Q waves or a history of MI,however, similarly increased the probability of anysignificant coronary artery disease and the probabilityof three-vessel or left main disease. Because the prev-alence of any coronary artery disease was at least 85%in the Q wave/MI groups regardless of the exercise testresults, the clinical usefulness of the exercise test inthese groups depends on its ability to distinguish severe

coronary artery disease (three-vessel or left main dis-ease) from nonsevere coronary artery disease.

Exercise Test ResponsesTable 3 describes the exercise test responses of the

study sample grouped by the number of vessels with75% or greater diameter narrowing. The mean max-imal heart rate was significantly different betweengroups and was inversely related to the severity ofcoronary disease. The age-predicted maximal heartrate for this group was approximately 150 beats/minbut the lower mean value achieved is partiallyexplained by clinical indications for stopping the testand the use of }-blockers. The mean maximal Borgscale value of 17 (very hard) is consistent with a nearmaximal effort; it was 18 in those patients withoutangina. The peak systolic blood pressure, maximalaerobic capacity estimated in METs, and the differ-ence in the double product from standing rest to

TABLE 3. Exercise Test Results Grouped by Extent of Coronary Artery Disease

Patients Patients with CADAll patients without

(n) CAD (n) Total (n) 1-2 VD 3 VD/LM p*Patients (n) (%) 328 (100) 80 (24) 248 (76) 183 (56) 65 (20) ...Maximal heart rate (beats/min) 126±22 130±24 124+21 125±22 121±20 0.03Peak systolic pressure (mm Hg) 161±31 165±33 160±30 163±31 151+27 NSWorkload (METs) 6.7±2.9 6.9±3.2 6.7±2.9 7±2.8 6±2.6 0.01Double product difference

[(beats/min)* mm Hg] 10.6±5.2 11.2±5.3 10.4±5.0 11.0±5.1 8.5±0.4 0.001Perceived exertion (Borg scale) 17±2.8 17±2.8 17±2.8 17±2.6 17±3.2 NSWith angina during test (n) (%) 115 (35) 14 (17) 101 (41) 74 (40) 27 (42) <0.0001Abnormal by standard ST criteria (n)(%)>1 mm 168 (51) 22 (28) 146 (59) 98 (54) 48 (74) <0.0001>2 mm 98 (30) 10 (13) 88 (36) 50 (27) 38 (59) <0.0001

With downsloping ST-exercise (n) (%) 69 (21) 9 (11) 60 (24) 33 (18) 27 (42) <0.001With downsloping ST-recovery (n) (%) 116 (35) 9 (11) 107 (43) 66 (36) 41 (63) <0.001AST (mm) 1.2±1.2 0.6±0.9 1.3±1.3 1.1±1.4 1.9±1.5 <0.0001ST/HR index (uVlbeats/min) 2.6±3.3 1.4±2.1 3.0±3.5 2.5±3.3 4.4+3.6 <0.0001Abnormal by ST/HR criteria (n) (%)

>1.6 uV/beats/min 174 (53) 26 (32) 148 (60) 102 (56) 46 (71) <0.001>3.2 uV/beats/min 101 (31) 11 (14) 90 (36) 52 (28) 38 (59) <0.001

CAD, coronary artery disease; 1-2 VD, single- or double-vessel disease; 3VD/LM, three-vessel or left main coronaryartery disease; Double product difference, difference in heart rate-systolic pressure product between baseline andmaximal exercise; AST, change in ST segment depression from baseline to maximal exercise; HR, heart rate.

*Statistical significance (p) for comparison of patient groups using analysis of variance.

by guest on July 11, 2018http://circ.ahajournals.org/

Dow

nloaded from

Lachterman et al ST Segment/Heart Rate Index 47

maximal exercise were significantly lower in thosewith left main or three-vessel disease.

ST Segment Responses and the Prediction of anySignificant Coronary DiseaseST elevation occurred during exercise in 17

patients, all ofwhom had Q waves, a previous MI, orboth, and was accompanied by abnormal depressionin other leads in seven patients. To compare thediscriminating power of the two types of measure-ments, a clinically relevant cut-point value of 1-mmST segment depression was chosen to compare sen-sitivities at the same specificity. The specificity of thestandard criteria of this cut point was 73% (95%confidence interval (CI), 63-82). The sensitivity ofthis cut point was 58% (95% CI, 52-64). The ST/heart rate index required a cut point of 2.1 uV/(beats/min) to achieve the same specificity of 73%.Its sensitivity at this cut point was only 54% (95%confidence limit, 48-60), 4% lower than the standardcriterion (p=NS). This calculation was repeated forthe 197 subjects without significant Q waves on theirresting ECGs. The specificity of 1-mm depression ofthe standardized criterion was identical, that is, 73%(95% confidence limit, 60-81). The sensitivity,however, was higher, that is, 69% (95% confidencelimit, 60-76). For the ST/heart rate index cut pointof 2.3 uV/beats/min corresponding to the samespecificity of 73%, the sensitivity was significantlyless, that is, 56% (95% confidence limit, 47-64)(p=0.01). ROC curves for all subjects and for thosewithout Q wave infarction are displayed in Figures1 and 2. There was no significant differencebetween the areas under these curves.

Identification of Three-Vessel Disease, Left MainDisease, or BothThe sensitivity at equivalent specificity for triple-

vessel or left main disease was also calculated forboth types of measurement. The specificity used inthis instance (that corresponding to 1-mm ST seg-ment depression by the standard measurement) was55% (95% confidence limit, 49-61). The sensitivitiesat this specificity for the two measurements were74% (95% confidence limit, 62-83) and 71% (95%confidence limit, 59-80) for the conventional mea-surement and the ST heart rate index, respectively.McNemar's test showed no significant differencebetween these two sensitivities. The ROC curvesfor these measurements were calculated and aredisplayed in Figure 3.

DiscussionTable 4 is a summary of investigations assessing

adjustment of ST segment depressions for heart rate.The sensitivities and specificities listed in the tableare for the prediction of any significant angiographicobstruction. All the investigations tabulated (with theexception of the present investigation) have shown asignificant increase in accuracy by application ofheart rate adjustment. Results closest to those of the

SenS

ti

ty

0.8

0.6

0.4

0.2

0 0.2 0.4 0.6 0.8 1

False Positive RateFIGURE 1. Plotting showing receiver operating characteristiccurves for comparing ST depression with heart rate-adjustedST depression forpredicting any >75% angiographic obstruc-tion (all 328 patients are included). Area under heart rate-adjusted curve is 0.66±0.03. The area under ST depressioncurve is 0.67±0.03 (p=NS). -u-, ST depression; -+-, heartrate adjusted.

present report (demonstrating only modestly in-creased accuracy) were achieved by Detrano andcolleagues6 in a group of 303 patients at the Cleve-land Clinic. They found an increase in sensitivity of4% at a fixed specificity of 73%, corresponding to anexercise-induced ST segment depression of 1 mm.On the other hand, Kligfield and colleagues5 found avery large increase in sensitivity, from 68% to 91%,with a smaller increase in specificity from 83% to93%. The results of the other investigators fallbetween these two extremes. Methodological differ-ences between the study designs might explain someof these differences. First, some investigators haveused a greater-than-50% rather than a greater-than-75% luminal diameter narrowing to define sig-nificant angiographic disease. A glance at the table,however, reveals that this could hardly explain a largepart of the variability in the resulting sensitivities andspecificities. Furthermore, our recent meta-analysisof the exercise electrocardiographic literatureshowed no relation between the definition of angio-graphic narrowing and the variance in sensitivity andspecificity.15Another possible explanation of the differing

results is the different method of measuring heartrate-adjusted ST segment depression. In Table 4, thenames of authors marked by asterisks indicate theinvestigators who used the same method of calculat-ing heart rate-adjusted ST depression as used in the

by guest on July 11, 2018http://circ.ahajournals.org/

Dow

nloaded from

48 Circulation Vol 82, No 1, July 1990

SenS

tiVity

0.8

0.6

0.4

0.2

0 0.2 0.4 0.6 0.8 1

False Positive RateFIGURE 2. Plotting showing receiver operating characteristiccurves for ST depression and heart rate-adjusted ST depres-sion forpredicting any > 75% angiographic obstruction. Onlythe 197 patients without diagnostic Q waves are included.Areas for ST depression and heart rate-adjusted curves are

0.72±0.04 and 0.71+0.04 (p=NS). -n-, ST depression;-, heart rate adjusted.

current investigation. That is, the difference betweenST segment depression at peak exercise and STsegment depression at rest is divided by the exercise-induced increase in heart rate. Note that Kligfield eta15 also used this method in the report listed in Table4. The reports lacking asterisks indicate the use ofthe ST segment/heart rate maximum-slope methodfirst introduced by Elamin et al.4 This can be done, asElamin did, by using a magnifying glass and carefullyplotting ST segment changes against heart rate, or itcan be done with a computer. Manual performanceof these calculations takes several hours per test andis therefore not applicable to busy exercise labs.Computer algorithms have been prepared,16 whichrapidly perform these calculations. These algorithms,however, are subject to the vagaries and inconsisten-cies that we have seen in other computerized mea-surement strategies.17 Furthermore, Kligfield et a118have shown only an insignificant difference betweenthe results of the ST segment/heart rate slopemethod of Elamin et al and the ST segment/heartrate index method used in their most recent work5and in the present report.

Differences in the application of the conventionalcriterion could certainly affect the differencesbetween the accuracies of the two methods. Specifi-cally, the treatment of ST segment slope has poten-tial importance. In our recent meta-analysis,15 thetreatment of slope was one of the few technical

SenS

tiVity

0 0.2 0.4 0.6 0.8 1

False Positive RateFIGURE 3. Plotting showing receiver operating characteristiccurves for comparing ST depression with heart rate-adjustedST depression for predicting three-vessel or left main disease(all 328patients are included). Area under heart rate-adjustedcurve (-+- ) is 0. 70+0.04. The area under ST depression curve

is 0.70±0.04 (p=NS).

variables that significantly affected the accuracy ofthe exercise-ST segment depressions. In the presentinvestigation, only exercise-induced' depression of theJ-point was measured, and this was assigned thevalue 0 when ST segment slope was upsloping. Theeffect of this on test accuracy can be seen by studyingthe ROC curve for the initial criteria in Figure 2.Note that the curve for the conventional ST segmentdepression has a rather sharp "bump" at the pointcorresponding to 1-mm ST depression (at a false-positive rate of 0.27 and a sensitivity of 0.66). If theslope itself is ignored, this bump will disappear andthe false-positive rate of 1-mm ST segment depres-sion will increase at the same sensitivity. This effect,however, appears small. Furthermore, the otherinvestigators listed in the table had differentapproaches toward slope, which did not negate theincreased accuracy of heart rate-adjusted ST depres-sions. Noteworthy, however, is that Detrano et a16looked at the effect of slope separately. Althoughtheir treatment of slope was slightly different thanours, their method also treated upsloping depressionconservatively. When this was done, no differencewas noted between sensitivities of the conventionalmethod and sensitivities of the heart rate-adjustedST segment depressions at the same specificity of73%. We conclude that our treatment of slopeexplains some of the difference between our resultsand those of the investigators listed in Table 4.

by guest on July 11, 2018http://circ.ahajournals.org/

Dow

nloaded from

Lachterman et al ST Segment/Heart Rate Index 49

TABLE 4. Sensitivities and Specificities Reported in Literature for Heart Rate-Adjusted Versus Nonadjusted ST Depression

Conventional criteria Heart rate-adjusted criteria Only% Diameter Slope patients with

Author (yr) Subjects (n) Specificity Sensitivity Specificity Sensitivity narrowing used cardiac cath

Simoons (1977)*2 138 95 (82/86) 63 (33/52) 93 (80/86) 81 (42/52) 70 No NoElamin (1982)4 206 88 (15/17) 93 (120/129) 100 (138/138) 100 (168/168) 75 No YesDetrano (1986)*6 303 73 (120/164) 65 (90/139) 73 (120/164) 69 (96/139) 50 No YesFinkelhor (1986)7 64 78 (24/27) 62 (23/37) 89 (24/27) 76 (28/37) 75 Yes YesDeckers (1988)*12 345 90 (158/175) 65 (111/170) 90 (158/175) 78 (133/170) 50 No NoKligfield (1989)*5 300 83 (125/150) 68 (102/150) 93 (139/150) 91 (137/150) 50 ... NoSato (1989)1" 142 76 (25/33) 61 (66/109) 97 (29/30) 70 (73/104) 75 Yes YesThis report* 328 73 (58/80) 58 (144/248) 73 (58/80) 54 (134/248) 75 Yes Yes

Sensitivity and specificity values represent percentages, with values in parentheses indicating number of patients. % Diameter narrowing,criterion for significant coronary artery disease; cath, catheter.*ST segment/heart rate adjustment done by dividing the differences in the ST segment level at peak exercise and at rest by the difference

between the peak heart rate and the resting heart rate.

Aside from the technical issues previously dis-cussed, patient population differences might explaindifferences in reported accuracies.19 As Table 4shows, four of the seven previous reports includedonly patients undergoing cardiac catheterization.Three of the reports, including that of Kligfield et al,5included so-called "normal" controls. These are usu-ally chosen as a group of individuals with very lowprobability of coronary artery disease. Because thesesubjects' probability of coronary artery disease is verylow, the authors assumed that an abnormal exercise-ST segment depression is a false-positive and can bediscounted. The use of normal controls usually has apractical as well as a theoretical justification. Thepractical justification is the lack of sufficient numbersof catheterized patients who do not have a significantcoronary artery obstruction. Lacking sufficient num-bers, an investigator cannot have confidence in thecalculated test specificity without turning to anothergroup of "normal subjects." The theoretical justifica-tion is that patients with normal coronary angiogramswill produce an inordinate number of false-positiveresults because of workup or referral bias.20 Weargue to the contrary, that is, if a subject's probabilityof coronary artery disease is so low that, no matterhow abnormal his exercise test result, he is stillassumed not to have disease, he and his test resultsare of no interest to the clinical users of exercisetesting. He is among the "wellest of the well'"2' andshould not be used in comparing various protocolsand exercise variables. In our estimation, it is a lessserious error to use normal angiogram control sub-jects, for whom a more accurate exercise test mighthave precluded the angiogram, than to use controlsubjects who would never be considered for angiog-raphy or exercise testing. It is noteworthy that in thethree reports in Table 4 using normal controls, therewere large differences in the accuracies of the twomethods. We believe that this difference in controlgroup is responsible for some of the differencesbetween our results and those of some investigatorslisted in Table 4. We have reviewed the data ofDeckers et al12 and found that when the patients not

undergoing coronary angiography are excluded, theincrease in accuracy caused by heart rate adjustmentdecreases, thus supporting our contention.Another difference in the analysis relative to pop-

ulation issues between our studies is that Kligfield eta15 included patients with previous MIs but did notconsider the test performance with and without themseparately, as we did. Diagnosis of coronary disease isnot usually an issue in patients with previous MIsbecause they are very likely to have disease.Another methodological factor that could result in

an apparent increase in accuracy is the treatment ofequivocal results when using the standard criterion.We have found that bias resulting from this factorcan significantly affect the reputed accuracy of exer-cise testing.15 An example of such bias is illustrated inthe report of Kligfield et al.5These investigators excluded equivocals (17% of

study group) from the numerators for the calculationof sensitivity and specificity. In an overview of 147publications, we found this treatment of equivocalresults used only one time.22 Such a biased treatmentof equivocals falsely improves both sensitivity andspecificity.23 The test results so assigned and there-fore excluded involved upsloping ST segments. Theseshould be treated as normal' as we have done in thepresent investigation. It can be deduced from thereport of Kligfield et a15 that the specificity of thestandard criterion would have been similar to that ofthe heart rate-adjusted method if these authors hadtreated equivocals in a conventional manner.Another difference between the studies is that

Kligfield et a15 did not consider recovery changes, aswe did. We have found recovery measurementsto increase sensitivity without a decrease in pre-dictive value.24

Limitation of Present StudyMeticulous review of the exercise electrocardio-

graphic literature22 has taught us the difficultiesinherent in studying the factors affecting sensitivityand specificity of the exercise ST segment. Reportscommonly suffer from methodological bias and miss-

by guest on July 11, 2018http://circ.ahajournals.org/

Dow

nloaded from

50 Circulation Vol 82 No 1, July 1990

ing data. Different populations respond differently tothe stress of exercise, and measurement error alsointroduces noise into reported results. In the presentreport, we have tried to avoid bias and includedetailed information concerning our population andmethods. Still, ours, like most similar studies, isretrospective with all the inherent problems. More-over, our population (male veterans with high-disease prevalence) is not representative of the pop-ulation as a whole. We await the publication ofreports on large multicenter databases that can more

clearly explain the factors affecting the accuracy ofvarious methods of analysis.

AcknowledgmentsThanks to Lesley Anderson for typing and to

Margo Hullett for performing the tests.

References1. Members of Task Force on Assessment of Cardiovascular

Procedures and Members of Subcommittee on Exercise Test-ing: Guidelines for exercise testing. A report of the AmericanCollege of Cardiology/American Heart Association task forceon assessment of cardiovascular procedures. JAm Coll Cardiol1986;8:725-738

2. Simoons ML: Optimal measurements for detection of coro-nary artery disease by exercise ECG. Comput Biomed Res1977;10:483-489

3. Hollenberg M, Budge WR, Wisneski JA, Gertz EW: Treadmillscore quantifies ECG response to exercise and improves testaccuracy and reproducibility. Circulation 1980;61:276-285

4. Elamin MS, Boyle R, Kardash MM, Smith DR, Stoker JB,Whitaker W, Mary DASG, Linden RJ: Accurate detection ofcoronary heart disease by new exercise test. Br Heart J1982;48:311-320

5. Kligfield P, Ameisen 0, Okin PM: Heart rate adjustment ofST segment depression for improved detection of coronaryartery disease. Circulation 1989;79:245-255

6. Detrano R, Salcedo E, Passalacqua M, Friis R: Exerciseelectrocardiographic variables: A critical appraisal. JAm CollCardiol 1986;8:836-847

7. Finkelhor RS, Newhouse KE, Vrobel TR, Miron SD, BahlerRC: The ST segment/heart rate slope as a predictor ofcoronary artery disease. Comparison with quantitative thal-lium imaging and conventional ST segment criteria. Am HeartJ 1986;112:296-304

8. Haraphongse M, Kappagoda T, Tymchak W, Rossall RE: Thevalue of sum of ST segment depression in 12 lead electrocar-diogram in relation to change in heart rate during exercise topredict the extent of coronary artery disease. Cardiovasc Med1986;2:64-67

9. Ameisen 0, Kligfield P, Okin M, Miller DH, Borer JS: Effectsof recent and remote infarction on the predictive accuracy ofthe ST segment/heart rate slope. J Am Coll Cardiol 1986;8:267-273

10. Quyyumi AA, Raphael MJ, Wright C, Bealing L, Fox KM:Inability of the ST segment/heart rate slope to predict accu-rately the severity of coronary artery disease. Br Heart J1984;51:395-398

11. Sato I, Keta K, Aihara N, Ohe T, Shimomura K, Hasegawa Y:Improved accuracy of the exercise electrocardiogram in detec-tion of coronary artery and three vessel coronary disease.Chest 1989;94:737-744

12. Deckers JW, Rensing BJ, Tijssen JGP, Vinke RVH, Azar AJ,Simoons ML: A comparison of methods of analyzing exercisetests for diagnosis of coronary artery disease. Br Heart J1989;62:438-444

13. Savvides M, Ahnve S, Bhargava V, Froelicher VF: Computeranalysis of exercise-induced changes in electrocardiographicvariables: Comparison of methods and criteria. Chest 1983;84:699-706

14. Hanley JA, McNeil BJ: A method of comparing the area underreceiver operating characteristic curves derived from the samecases. Radiology 1983;148:839-843

15. Gianrossi R, Detrano R, Mulvihill D, Lehmann K, Dubach P,Colombo A, McArthur D, Froelicher VF: Exercise-inducedST depression in the diagnosis of coronary artery disease: Ameta-analysis. Circulation 1989;80:87-98

16. Okin PM, Kligfield P: Computer-based implementation of theST-segment/heart rate slope. Am J Cardiol 1989;64:926-930

17. Detrano R, Salcedo E, Leatherman J, Day K: Computer-assisted versus unassisted analysis of the exercise electrocar-diogram in patients without myocardial infarction. JAm CoilCardiol 1987;10:794-799

18. Kligfield P, Ameisen 0, Okin PM: Relation of the exerciseST/HR slope to simple heart rate adjustment of ST segmentdepression. J Electrocardiol 1987;20:135-140

19. Philbrick JT, Horwitz RI, Feinstein AR, Langou RA, Chand-ler JP: The limited spectrum of patients studied in exercisetest research. JAMA4 1982;248:2467-2470

20. Philbrick JT, Horwitz RI, Feinstein AR: Methodologic prob-lems of exercise testing for coronary artery disease: Groups,analysis and bias. Am J Cardiol 1980;46:807-812

21. Rozanski J, Diamond G, Forrester J, Berman D, Morris S,Swan HJC: Alternative referent standards for cardiac normal-ity. Ann Intem Med 1984;101:164-171

22. Detrano R, Gianrossi R, Froelicher V: The diagnostic accu-racy of the exercise electrocardiogram: A meta-analysis of 22years of research. Prog Cardiovasc Dis 1989;32:173-206

23. Feinstein A: Clinical Epidemiology. The Architecture of CinicalResearch. Philadelphia, WB Saunders Co, 1985, pp 225-226

24. Lachterman B, Lehmann KG, Abrahamson D, Froelicher VF:Does incidence of "recovery only" ST segment depressionaffect the predictive accuracy of the exercise test? Ann IntemMed 1990;112:11-16

KEY WORDS * exercise testing * coronary angiographyheart rate * ST segment depression

by guest on July 11, 2018http://circ.ahajournals.org/

Dow

nloaded from

B Lachterman, K G Lehmann, R Detrano, J Neutel and V F Froelicherexercise electrocardiogram.

Comparison of ST segment/heart rate index to standard ST criteria for analysis of

Print ISSN: 0009-7322. Online ISSN: 1524-4539 Copyright © 1990 American Heart Association, Inc. All rights reserved.

is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Circulation doi: 10.1161/01.CIR.82.1.44

1990;82:44-50Circulation. 

http://circ.ahajournals.org/content/82/1/44the World Wide Web at:

The online version of this article, along with updated information and services, is located on

  http://circ.ahajournals.org//subscriptions/

is online at: Circulation Information about subscribing to Subscriptions: 

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

  document. Permissions and Rights Question and Answer information about this process is available in the

located, click Request Permissions in the middle column of the Web page under Services. FurtherEditorial Office. Once the online version of the published article for which permission is being requested is

can be obtained via RightsLink, a service of the Copyright Clearance Center, not theCirculationpublished in Requests for permissions to reproduce figures, tables, or portions of articles originallyPermissions:

by guest on July 11, 2018http://circ.ahajournals.org/

Dow

nloaded from