Ability of standard ECG parameters to detect the body surface isopotential abnormalities of pacing induced myocardial ischemia in the dog

J. ELECTROCARDIOLOGY 18 (1), 1985, 77-86

Ability of Standard ECG Parameters to Detect the Body Surface Isopotential Abnormalities of Pacing Induced Myocardial Ischemia in the Dog By DAVID M. MIRVIS, M.D. SUMMARY The ability of parameters derived from standard, scalar ECG leads to predict abnormal body surface isopotential distributions was tested in 25 dogs. Atrial pacing to rates of 170 beats per minute or greater, two weeks after implantation of an ameroid constrictor on the lef~t circumflex coronary artery, resulted in abnormal isopotential patterns due to subendocardial ischemia. A total of 96 scalar ECG variables (potential and slope measurements at 20, 40, 60 and 80 msec into the ST-segment in each of 12 leads) were computed. Ability of each variable and potential-slope pairs to distinguish normal from ischemic map forms was tested using discriminant function analysis and simple classification procedures. Results demonstrated that accuracy of prediction was leaddependent, and time-dependent, with different sensitivities, specificities and boundary values in different leads and at different instants during the ST-segment. These new concepts m a y have direct relevance to the selection of clinical exercise ECG diagnostic criteria.

per minute or greater results in major spatial STsegment abnormalities indicative of myocardial ischemia 4. These consist of the appearance of negative voltages over the inferior anterior, left lateral and posterior torso, replacing the positive voltages observed normally and at slower pacing rates. Such a distinct pattern change permits a clear, binary classification of isopotential records as being normal or ischemic. Application of surface mapping procedures to clinical exercise testing has been reported 5-7 but is limited by technical considerations, e.g., computer processing, electrode numbers, etc. This study was therefore undertaken to determine which features, if any, in the standard twelve ECG lead records are predictive of the characteristic isopotential pattern described above. These data may then be used to evaluate the relative accuracy of specific diagnostic criteria.

Electrocardiographic changes during exercise are commonly used, although imperfect, diagnostic criteria for clinical ischemic heart disease. Alterations in ST-segment voltages and slope are Lhe major features examined. Thus, ST-segment depression beyond 100 to 200 microvolts (uV) with a fiat or negative slope are typically used diagnostic criteria. Recently, we described an animal m o d e l in which these abnormalities could be experimental[y produced t. In it, coronary obstruction is produced by a surgically implanted ameroid constricLor, tachycardia is caused by rapid atrial pacing, and the resultant ST-segment abnormalities are studied using isopotential body surface mapping ~echniques. Two weeks after ameroid placement, when significant but subtotal coronary arterial abstruction occurs 2.3, pacing to rates of 170 beats From the Division of Cardiovascular Diseases, Department of Medicine, University of Tennessee Center for the ttealth Sciences, Memphis, TN. This project was supported by research grants from the National Institutes of Health, Bethesda, Maryland 1ttL20597. HL00560). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "'advertisement" in accordance with 18 U.S.C. w1734 solely to indicate this fact. Reprint requests to: David M. Mirvis, M.D., Divisionof Cardiovascular Diseases, 956 Court Avenue, Room 2F18, Memphis, TN 38163.

MATERIALS AND METHODS Twenty-five adult mongrel dogs, weighing 14 to 22 kg and in good health, were studied. A left thoracotomy was performed under sterile conditions after induction of anesthesia with sodium pentothal (20 mg/kg) and a halothane-nitrous oxide-oxygen mixture. A quadripolar plaque electrode was sutured to the left atrial appendage, and a 2.77 mm internal diameter ameroid constrictor was placed around the proximal left circumflex coronary artery. The pericardium was loosely sutured, the pacemaker wires tracked to the

77

78

MIRVIS

IN

A.SR

§

*

*

§

156,-42UV D.!70/MIN

_

_

* *

239,-74UV

E.!9o/M!N

-

- _

389,-109UV

!

_ _

-

-

-

245,-142UV

g24,-92UV F.210/N![N

_

_

_

:

_

_

-

_

366,-167UV -/~40007/15-23

Fig. 1. Body surface isopotential distributions from one dog, 40 msec into the ST-segment, two weeks after implantation of an ameroid constrictor on the left circumflex coronary artery. Markings are detailed in the text. Maps were constructed from voltages registered during sinus rhythm (Panel A, SR) and during atrial pacing at rates of 130 (Panel B), 150 (Panel C), 170 (Panel D), 190 (Panel E) and 210 (Panel F) beats per minute.

nape of the neck and exteriorized, the chest wall closed in layers, and air evacuated under a water seal. Acepromazine {10 rag} was given as an analgesic, and antibiotics were administered for two to three days. All dogs were well and free of infection within 48 hours. Electrocardiographic Recordings. Electrocardiograms were recorded 14 days after surgery at a time when significant vascular obstruction by the ameroid device had occurred 2.3. Before each recording session, animals were sedated with 1 to 2 cc Innovar (20 mg fentanyl and 0.4 mg droperidol per cc) and sodium pentoba~bital (7 mg/kg). The dogs were then intubated and respired on 100% oxygen. Eighty-four chloridized silver electrodes were secured to the shaven chest in 14 strips of 6 electrodes each. These extended from the level of the clavicles to below the inferior rib margins, with twothirds located on the anterior torso half. Additional electrodes were placed on the extremities to derive a Wilson central terminal voltage and to record the standard bipolar and unipolar limb leads. Dogs were then placed in a stable upright posture in a support sling. Electrocardiographic signals were amplified by a bank of 87 low noise, differential (Wilson central terminal vs. torso grid electrode potential} capacitorcoupled amplifiers. Gains (1 to 16 thousand) and offsets {-4.5 to +4.5 millivolts) were individually set under computer control so that the output filled the input range of the analog-to-digital convertor. Signals were then converted to digital form at a rate of 500 samples per channel per second.

Analog signals were observed on a multiple channel oscilloscope to assure baseline stability and low noise levels. Experimental Protocol. Fourteen seconds of ECG data were recorded during spontaneous sinus rhythm at rates of 56 to 84 beats per minute. Subsequently, atrial pacing was initiated at the lowestrate giving uniform capture, usually 90 beats per minute, with bipolar, 2 msec pulses 25% above diastolic threshold. The rate was increased by 20 beats-per minute every three minutes until a peak rate of 250 beats per minute was reached. If second-degree atrioventricular (AV) block occurred, intravenous atropine (1-2 mg) was given; this uniformly restored one-to-one AV conduction. Intraventrieular conduction block has not occurred in our experience 1. Electrocardiographic signals from all electrodes were simultaneouslyregistered during the last 14 seconds of pacing at each rate. Control Group. A second set of 45 dogs was evaluated to determine the normal values for the ECG parameters to be described. These were studied once in an unoperated condition, during spontaneous sinus rhythm, and using the same preparative and recording procedures described above. Data Analysis. ECG waveforms in each lead with similar morphologies, as determined by an automated autocorrelation routine 1, were averaged. Onsets and offsets of the P-wave, QRS complex and ST-T interval were manually selected from root-mean-square potential plots. A 10-20 msec portion of the terminal PR-

J. ELECTROCARDIOLOGY 18 (1), 1985

E C G A N D S U R F A C E M A P S IN I S G H E M I A

interval was selected as a zero potential baseline, and linear baseline drift corrected. lsopotential distributions were then constructed at 2 :nsec intervals during the ST-T interval. Contour lines were drawn using a linear-bilinear interpolation routine ]t zero and at plus and minus 10, 20, 40, 60, 100, 200 tnd 400 microvolt levels. Serial maps from each data set were examined and "ategorized as being normal or abnormal. Examples ilustrating the criteria used are presented in Fig. 1. Data from one dog, two weeks after ameroid implanta;ion are shown. In each isopotential map, plus and ninus signs mark electrode locations, with the sign :orresponding to the polarity of the sensed voltage. I'he center of each frame, marked by the dotted ver:ical line, is along the sternum with the left and right mrders representing the right and left paravertebral ines, respectively. Contour lines connect points at .'qual potential relative to the Wilson central terminal; .he zero voltage lines are overdrawn for emphasis. ~lagnitudes of the maximum and minimum voltages of .'ach distribution are listed below the pattern. During sinus rhythm at a rate of 84 beats per minute, .he distribution 40 msec into the ST-segment (Panel A) s dominated by anterior positive voltages (maximum = 156 ~V) and posterior negative ones (minimum = -42/~V). This pattern with positive potentials covering .he anterior caudal chest was observed at all instants luring the initial 80 msec of the ST-segment in all :ases during sinus rhythm. Subsequent panels show the patterns at successively dgher rates. At rates of 130, 150 and 170 beats per ninute (Panels B, C and D, repseetively), the overall

79

features are like those at rest (Panel A); although intensity of the extrema increased with increasing rate, the spatial form was unchanged. This pattern of increasing magnitude was seen in all normal, control dogs at all rates and in dogs with ameroid constrictors at lower rates 1. At rates of 190 beats per minute and higher, isopotential patterns demonstrated new, abnormal negative voltages over the anterior, caudal torso (Panels E and F). These potentials, corresponding to ST-segment depression in unipolar ECG leads from the affected areas, represented subendocardial myocardial ischemia 4 induced by the tachycardia. Maps from all data sets with features like those in Panels A through D were classified as "normal" and those as in Panels E and F were termed "ischemic". This classification was made during viewing of maps presented in random order and without knowledge of heart rate. Potentials at points 20, 40, 60 and 80 msec into the ST-segment in the three standard bipolar limb leads, the three augmented unipolar limb leads and from six unipolar left precordial electrodes, approximating the six V-lead sites, were tabulated for each data set. Also computed were the slopes of the ST-segment in each lead from 10 msec to 20, 40, 60 and 80 msec into that interval. Thus, a total of 96 ECG parameters (four potentials and four slopes in each of 12 leads) were compiled. Mean values for each ECG variable were then determined in each map group ("normal" and "ischemic"), and differences between the means tested by analysis of variance s'9. If the means were significantly different

TABLE I ~,bility of single ECG variables, 40 Msec into ST-segment, to detect abnormal isopotential map patterns

Variable I,P* I,S* II,P II,S III,P aVR,P aVL, P aVF,P V1,S V3,P V4,P V4,S V5,P V5,S V6,P V6,S

Normal** Mean, STD Error 40.50 -5.34 75.34 21.17 33.81 17.42 3.35 54.57 38.56 162.39 136.55 46.11 117.63 39.76 82.54 28.78

Boundary * * Value

F*

Sens ( % )

Spec (%)

Accur (%}

3.49 -10.O7 -74.49 -15.83 -79.20 -38.98 41.33+ -76.84 67.62+ 50.47 -22.22 21.80 -19.76 15.72 -11.82 - 4.50

61.19 9.31 109.59 10.86 64.00 64.38 17.27 90.11 10.48 35.89 94.34 8.73 86.32 8.14 57.29 9.30

86.0 86.4 100.O 75.9. 88.9 88.9 65.6 100.O 64.9 82.4 94.9 70.6 97.4 75.0 97.5 76.0

82.7 63.0 86.0 63.6 76.3 76.3 60.3 81.7 62.1 70.5 83.9 63.9 84.2 67.8 87.3 61.4

84.2 68.4 91.6 67.6 81.1 81.1 62.1 88.4 63.2 74.7 88.4 66.3 89.5 70.5 91.6 67.4

_+ 11.14 +_ 2.72 + 19.25 + 4.68 -- 14.47 • 7.38 _+ 8.75 + 16.O7 + 3.94 • 14.88 • 22.28 • 5.99 + 20.99 • 5.60 + 17.44 • 4.51

*P = potential; S = slope; F = F-statistic comparing means of t w o groups. *Values in uV for potentials, uV/30 msec for slope. -Values for ischemic group greater than for normal group; in all other cases, normal values greater than for abnormals.

ELECTROCARDIOLOGY

18 (1), 1985

80

MIRVIS

A.170/MIN _

~

_

_ v ~

-

POT,SLOPE 76.7,-4.1

_

I~ ~ ~ ~ . .

109.5,45.8 318,-110UV

B.210/MIN ",',, \ l b:.,"~t,il .....-"_"~_,r_ _1 . . . . .

- ~ /.

.

,11, . ~ ~ , . _ ~ t ~ ~ .

130.5,48.6

,4 V4-4-44

-67.3,-38.9

II ~

431.5,26.5"

378,-345UV III~ :~40010

~

-167.8,65.9

Fig. 2. Examples of isopotential maps, (left) drawn as in Fig. 1, ECG waveforms (center) from standard leads I, I I and III, and values (right) for potential (POT) and slope 40 msec into the ST-segment. In Panel A, examples are during atrial pacing at a rate of 170 beats per minute and in Panel B from the same animal during pacing at 210 beats per minute. Two seconds of ECG data are plotted, with the scale figure corresponding to 500 msec. Arrows in the center column identify stimulus artefacts and the dashed lines mark computer generated lmselines after correction for linear drift. The asterisk identifies a criterion resulting in misclassification of the data set.

(p<0.01), a discriminant function analysis 9J~ was performed to identify the boundary point best separating the normal and ischemic groups. Next, potential and slope at each time point and in each electrode were entered into a disciminant function analysis 9j~ The statistical significance of the difference in the values of the computed discriminant functions for both single and two variable cases in the normal and ischemic data sets were determined by Wilks lambda statistic. If significant (p<0.01), the boundary points using one or both parameters were computed. Next, a simple classification procedure was used to allocate each map to a normal or isehemic population based upon the boundary points computed using each one or two ECG parameters. True positive (map originally determined to be ischemic and allocated to ischemic population by classification procedure), true negative (map originally determined to be normal and allocated to normal population), false positive {map originally determined to be normal but allocated to ischemic population) and false negative {map originally

determined to be ischemic hut allocated to normal population) rates were computed and sensitivity, specificity and overall accuracy calculated using standard formulae n.

RESULTS All animals survived the experimental period and those evaluated postoperatively were successfully paced to rates of 230 beats/minute or higher. ST-segment patterns in the experimental group were normal during sinus rhythm two weeks after surgery; abnormal or "ischemic" patterns, such as described above, developed in all animals at rates of 170-190 beats/minute. In the twenty-five dogs with ameroid constrictors, a total of 230 data sets or cases were analyzed. In each, m a p s c o n s t r u c t e d from p o t e n t i a l s sensed 20, 40, 60 a n d 80 msec into the STs e g m e n t were classified similarly, i.e., either all four were " n o r m a l " or all were " i s c h e m i c " . One

J. E L E C T R O C A R D I O L O G Y

18 (1), 1 9 8 5

ECG AND SURFACE MAPS IN ISCHEMIA

RATE

LEAD II

!POT.,SL OPE

84

150

81

i18.5,11.9

j,,Ij

,JJj rirr r rrr

190

24.1,20.8

i-41.1,9.0

ig. 3. ECG waveforms and variables from the same experiment as displayed in Fig. 1. On the left, 4 second records 'om lead II are shown at rates of 84, 150 and 190 beats per minute. Values of slope and potential, 40 msec into the T-segment, are tabulated to the right. Markings are as in Fig. 2.

undred and six were classified as abnormal or ;chemic; the remaining 124 cases Were normal. ingle Variable Analysis. The 96 scalar ECG ariables at each instant were each analyzed as escribed previsously. Results using voltages or lopes computed 40 msec into the ST-segment are abulated in Table I. Only those variables for ,hich the means of the two groups (normal and ~chemic) were significantly different are listed. Five concepts are apparent. First, only 16 }7%) of 24 parameters at this time point were Latistically different in the two groups, 40 msec ~to ST-segment. For points 20, 60 and 80 msec 1to ST-segment, percentages were 58%, 75%, nd 42%, respectively. Second, in some leads only otentials or only slopes were significantly dif.~rent in the normal and ischemic groups. For exmple, in lead III, potentials 40 msec into the STegment were significantly greater in normal ban in abnormal groups {37.98 4- 11.44 gV vs. 196.38 _ 27.68 ~V), but the slopes of the STegment in the groups were not different. In this ;ad, potentials at all four times studied were

ELECTROCARDIOLOGY 18 (1), 1985

significant discriminators; slopes at none of these instants were. In lead V1, slopes in ischemic group were greater (90.81 _ 13.99/~V/30 msec} than in normal cases (44.42 _+ 4.50/~V/30 msec), whereas potentials in the two groups were not significantly different. Third, boundary points varied widely among leads. For example, potential values separating the two groups were 3.49/~V in lead I and -79.20 /~V in lead III. Fourth, leads varied in the accuracy of separation resulting from the computed boundary points. Sensitivities ranged from 64.9% in lead V, to 100% in leads II, and aVF; specificities varied from 61.4% in V6 to 86% in lead II. Ranges were similar for potential and slope parameters. Fifth, in some leads, parameters were significantly different in the two groups at certain times during the ST-segment but not at others. For example, in lead aVL, potentials 20, 40 and 60 msec into ST-segment were significantly different in the two groups, whereas t h a t 80 msec into the interval was not. In lead V1, voltages 60 and 80

82

MIRVIS

TABLE II Accuracies (%) of concordant and discordant patterns of potentials and slopes, 4 0 msec into STsegment, in predicting isopotential map patterns. Slope, potential pattern Pot nl* Pot n l * * Pot abn** Lead Slope nl Slope abn Slope nl I II aVF V4 V5 V6

86.2 89.4 82.8 86.2 87.3 89.4

33.1 30.9 22.6 25.4 25.4 25.4

78.3 84.9 84.9 81.4 74.8 86.0

Pot abn** Slope abn 92.2 94.7 96.4 96.4 100.0 98.6

*Calculated as no. normal maps/total no. with pattern. as no. abnormal maps/total no. with pattern.

* *Calculated

msec into ST-segment were discriminators, whereas those 20 and 40 msec into the interval were not. Examples of isopotential maps and ECG waveforms illustrating these notions are presented in Figs. 2 and 3. Iri Fig. 2 (Panel A), an isopotential distribution recorded 40 msec into the STsegment at a paced rate of 170 beats per minute is typical of a "normal" pattern; anterior positivity (maximum = 318 ~V) and posterosuperior negativity (minimum = -110 ~V) predominate. E C G waveforms from leads I, II and I I I demonstrate fiat ST-segments with minimal positive potentials. Quantitated values for potential and slope (40 msec values), as tabulated in the right column, reflect these features. All quantitated values are within the "normal" range as deter-

mined by single-variable discriminant function analysis (Table I). Thus, ECG analysis would result in a true negative classification of this case. Patterns and values at a paced rate of 210 beats per minute are presented in Panel B. The isopotential pattern is typical of an "ischemic" one, with inferior and left lateral negativity (minimum = -345 gV). ECG waveforms now show ST-segment depression in leads II and III. Values for voltages and slope, 40 msec into the ST-segment, are tabulated on the right. Levels of potential in all three leads are less than the boundary values tabulated in Table I; hence, this case w o u l d h a v e been c o r r e c t l y classified as "ischemic" using these criteria. Similarly, the value for slope in Lead I is within the "ischemic" range. However, that in lead II is greater than the boundary point of -15.83 ,V/30 msec (Table I). This case would have been erroneously classified as normal using this one criterion. The ECG waveforms and values in Fig. 3 are from the same experiment as used in Fig. 1. In the top row, a four second segment of lead II during sinus rhythm at a rate of 84 beats per minute is shown. The isopotential maps, as shown in Fig. 1A, was "normal". E C G waveforms showed flat isoelectric ST-segments (center); potential and slope, 40 msec into the ST-segment were within the normal range (Table I). Similarly, at a rate 150 beats per minute (middle row), ST-segments were flat and isoelectric, and q u a n t i t a t e d variables were " n o r m a l " . The isopotential distribution, as shown in Fig. 1C was likewise "normal". Thus, these cases were correctly classified by standard ECG parameters as belonging to the "normal" population.

TABLE III Ability of potential and slope, 4 0 msec into ST-segment to predict abnormal isopotential map pattern Electrode

Cp

Cs

Boundary Value

Sens (%)

Spec (%)

Accur (%)

I II III aVR aVL aVF V1 V2 V3 V4 V5 V6

0.99 0.87 0.82 0.83 0.76 0.85 0.23 0.15 0.85 0.84 0.82 0.85

-0.03 -0.47 -0.56 -0.55 -0.64 - 0.52 0.97 -0.98 -0.51 -0.53 - 0.56 -0.52

3.99 - 57.80 - 64.37 -31.83 35.45 - 60.71 95.06 - 30.58 19. 81 -30.44 - 25.17 - 7.67

86.0 100.0 100.0 100.0 72.2 100.0 64.7 64.4 86.1 97.4 95.2 95.2

82.7 87.5 79.0 79.0 66.1 83.1 60.7 66.0 74.6 85.7 88.7 88.7

84.2 92.6 86.3 86.3 72.2 89.5 62.1 65.3 78.9 90.5 91.6 91.6

Cp = coefficient for potential;

Cs = coefficient for slope.

J. E L E C T R O C A R D I O L O G Y

18 (1), 1 9 8 5

ECG AND SURFACE MAPS IN ISCHEMIA

At a rate of 190 beats per minute, the isopoten~1pattern (Fig. 1E) was abnormal, with inferior gativity typical of "ischemic" responses. ECG weforms (Fig. 3, bottom row) showed only ght degrees of ST-segment deviation (left). mntitated values for slope and for potential ght) were greater than the boundary values reputed by discriminant function analysis ~ble I). Thus, these standard ECG criteria were ormal" even though the isopotential pattern Ls markedly abnormal. As a result, this case told have been erroneously classified as a norfl, i.e., a false negative. [n six leads, both the potential and the slope 40 ~ec into the ST-segment were significantly dif9ent in the two groups. These leads were I, II, F , V4, V5 and V6. Data from these leads were alyzed to determine the relative accuracies of ncordant (both slope and potential were either rmal or abnormal) and discordant (either potenl or slope was abnormal but not both) patterns the two variables. Accuracies for the concornt pattern with both values being normal were reputed as the ratio of the number of such cases th normal isopotential maps to the total mber of cases with the normal, normal pattern. the other three cases, accuracy was the ratio of number of such cases with abnormal maps to total number of cases with the potential, slope ttern under consideration. In all cases, potenIs and slopes were considered normal or abnord based upon computed boundary points ~ble I). Fwo results are apparent from data tabulated Table II. First, concordant patterns more acrately predicted normal or abnormal isopoteni maps than did either discordant form. Values ' concordant patterns ranged from 0.82 to 1.00, Lereas those for discordant patterns varied ,m 0.22 to 0.86. Second, if only one variable was normal, accuracy of predicting an ischemic ~p was greater if it was potential rather than ,pe t h a t was abnormal. For example, 84% of Lps were abnormal if the potential in lead aVF .s abnormal, but only 22% were abnormal if the ,pe in t h a t lead was abnormal. ~o-Variable Analysis. As detailed above, both tential and slope in each lead at each time point re entered into a discriminant function analy9,1o. The result was a discriminant function uation of the form z = Cp • Potential + Cs • Slope Lere z = discriminant score, Cp = coefficient 9potential term and Cs = coefficient for slope

"LECTROCARDIOLOGY 18 (1), 1985

83

term. The boundary point was the value of z t h a t best separated the two groups. Results, 40 msec into ST-segment, are tabulated in Table III. Equations computed at each time point in each lead were statistically (p < 0.01) distinguish between the normal and ischemic groups. The adequacy of the result, as defined by either sensitivity, specificity or accuracy, did vary widely from one lead to another. For example, specificity varied from 66.1% in aVL to 88.7% in leads V5 and V6. Also noted is the wide range of relative magnitudes of the potential and the slope coefficients. In lead I, for example, the Cp approached u n i t y while Cs was near zero. A nearly opposite pattern was observed for leads V1 and V2. DISCUSSION This effort was intended to evaluate the relative abilities of variables derived from the standard ECG to detect abnormal torso potential patterns. We are keenly aware of the limitations as well as the advantages of the methods used to do so. The animal model closely approximates the clinical situation. As shown by, for example, Hill et al. 2, a 50% reduction in anterograde coronary flow with a concomitant increase in retrograde, collateral perfusion may be expected by the fourteenth postoperative day. Although resting flow thus remains normal, coronary reserve is abolished 3 so t h a t the stress of increased heart rate results in significant transmural redistribution of flow, with subendocardial ischemia 2. This, in turn, results in epicardial and body surface STsegment depression 4 of a form similar to t h a t observed during exercise stress tests in patients with coronary atherosclerosis 1. The major advantage of this model is t h a t it permits control of lesion location and heart rate t h a t is not possible in clinical situations. A second benefit, particularly valuable in this study, is the clear-cut differences in normal and abnormal spatia! patterns {Fig. 1). This permitted use of discriminant function analysis and simple classification procedures to explore the utility of the continuous variables of the standard scalar ECG. Limitations are however significant, First, it is an animal model. Differences in mechanism of obstruction, volume conductor form, and coronary physiology may all limit extrapolation to clinical conditions. Second, and most relevant here, measurements were limited to body surface potentials. We measured neither myocardial metabolic variables t h a t may be more sensitive detectors of ischemia

84

MIRVIS

than are electrical measures nor epimyocardial electrical ~voltages that may be abnormal in the absence of the body surafce potential deviations. Thus, the problem considered is limited to the evaluation of the ability of parameters derived from the scalar ECG to identify any body surface abnormality detectable by isopotential mapping. If such indices cannot detect map abnormalities, then they cannot reasonably be expected to detect ischemia that does not cause isopotential abnormalities. This argument is of course premised upon the superiority, of total body surface isopotential mapping techniques over standard scalar elect r o c a r d i o g r a p h y . T h a t this is so has been repeatedly demonstrated 12:6 and is based upon two concepts. First, the placement of as m a n y as 150 electrodes on anterior as well as posterior torso regions permits recording of diagnostically important information not projected to the usually sampled precordial or limb sites 12. Thus, the routine leads are but one subset of larger electrode arrays. Second, the display form of isopotential maps permits evaluation of spatial as well as intensity factors that determine electrocardiographic waveforms. Both have been shown to be important in identification of multiple, simultaneously active cardiac events 13, in diagnosis of ventricular enlargemenO 4, in defining the mechanism of terminal QRS notches 15, in localizing accessory AV nodal bypass tracts 16 and in other experimental and clinical applications. One last methodologic limitation is the s t u d y of only one lesion location. We previously showed that the spatial forms of ischemic changes caused by tachycardia vary with the location of the coronary arterial obstruction 17. Thus, the exact correlations of isopotential patterns would probably also vary with lesion location. However, the general conclusions t o b e derived from the data would likely persist. These concerns n o t w i t h s t a n d i n g , the d a t a detailed above relate to the use of scalar ECG p a r a m e t e r s to d e t e c t i s c h e m i c e l e c t r o c a r diographic abnormalities. First, there are ECG parameters that do correlate with isopotential distributions. For example, as shown in Tables I and III, either potential or slope plus potential in lead II were over 90% accurate. The leads with the greatest a c c u r a c y were either caudal-rostal bipolar (e.g., II) or left precordial unipolar (e.g., V4-6) ones. This is consistent with the inferior, left lateral position of the spatial abnormality (Fig. 1). Thus, such criteria may be acceptable substitutes for surface mapping. This would be of

importance clinically where full torso recording is impractical or technically difficult, and experimentally where use of open-chest models would prevent registration of total surface voltages. Second, the differences in the accuracies of the various measurements may relate to problems in clinical exercise electrocardiography. Numerous proposals for diagnostic criteria have been suggested. These include ST-segments depression beyond a fixed amount, and ST-segment slopes less than a fixed amount is as well as criteria based upon changes in other portions of the PQRST complex 19. Other electrocardiographic variables include the number of leads recorded and the lead system to be used 2~ Current data suggest the following considerations in choosing such criteria. First, diagnostic accuracy is lead-dependent. Both accuracy as well as values separating normal from abnormal differed widely from one lead to another. Thus, formulation of diagnostic measurements should be specific for individual leads. For example, STdepression of -50 ~V, 40 msec into ST-segment in lead V4, V5 or V6 would have been abnormal but in lead III would have been normal in our study (Table I). Values m a y vary also with location of lesion or lesions. Second, accuracy in any given lead will be timedependent. As noted above, in lead aVL, for example, potentials 20, 40 and 60 msec into the STsegment were significantly different in normal and ischemic groups, but voltages 80 msec into the ST-segment were not. Third, use of multiple parameters, e.g., potential and slope, rather than a single one, improves diagnostic accuracy (Tables I and III). Although we did not test the case, it is also likely that evaluation of parameters derived from more than one lead would also improve performance. When two criteria in the same lead were individually evaluated, agreement in classification by both measurements resulted in higher accuracy than if conclusions differed. Last, if the two did differ as to conclusion, measurements of potentials were s i g n i f i c a n t l y m o r e a c c u r a t e p r e d i c t o r s of ischemia than were those of ST-segment slope. Each of these considerations may have direct application in designing or evaluating diagnostic criteria for exercise stress tests. Although the particular quantitative values presented probably have little relevance to the clinical situation, the conclusions m a y be widely applicable and should be further tested. J. ELECTROCARDIOLOGY 18 (1), 1985

ECG AND SURFACE MAPS IN ISCHEMIA

REFERENCES MIRvls, D hl AN'I Goml~:Y, R L: Electrocardiographic effects of myocardial ischemia induced by atrial pacing in dogs with coronary stenosis. I. Repolaraization changes with progressive left circumblex coronary artery narrowing. J Am Coil Cardiol I: 10.()0, 1983 Ilii.i., I{ C, Ki.I-:i.";MAN, I, 1l, Tii.i.l-:il, li!ir H, CiilTv,'ooi), W 1{, l{l-'.',.llll-'il'i', J C ,',,Nil Glil-:l-:Nl-'li:i.I), J C: Myocardial Iflood flow and function during gradual coronary occlusion in awake dog. Am J Physiol 244:tt60, I!)83 S('iil:.l:.i., K W, GilAN(;I-:Ii, I1 J, niitiliY, D A ANI) KF.i.i.Ell, F %%': Mechanisms of collateral development and hemodynamics of gradual coronary occlusion. Basic Res Cardiol 69:338, 1974 Guv'r(ix. t{ A. McCi.ENA'rilAN, J H, NEWMAN, G E hNI) MIcllAm.ls, L I,: Significance of suliendocardial S-T segment elevation caused by coronary stenosis in the dog. Am J Cardiol 40:373, 1977 M ii.i.i:ll, tlV T, Si'ACii, ]%,IS ANII lIVAiiill-:N, l{ B: Total body surface potential mapping during exercise. QRS-T-Wave changes in normal young adults. Circulation 62:632, "1980 Fox, K M, SEI.wv~, A, OAKi.i-:Y, D ANn SHEH.INt;FORI), J P: Relation between the precordial projection of S-T segment changes after exercise and coronary angiographic findings. Am J Cardiol 44:1068, 1979 YANowrrz, F G, VINCENT, G M, I,tlx, R L, MI-:RCiiANT, 1%,,I,GREEN, 1,, S AND AIIli.DSKOV, J A: Application of body surface mapping to exercise testing: S-T80 isoarea maps in patients with coronary artery disease. Am J Cardiol 50:1109, 1982 SOKAI., R R ANI) Rtilll.}', F J: Biometry. W. H. Freeman, San Francisco, 1969, i ) 175 DixoN, W J AND BROWX, M B: Biomedical Computer Programs. University of California Press, Berkeley, 1977 COOI.FY, I,V %VANI) LiiIINFS, P R: Multivariate Data Analysis. New York, Wiley, 1971

ELECTROCARDIOLOGY 18 (1), 1985

85

11. JtmNSON, J C, HORAN, L G AND FI.OWEliS, N C: Diagnostic accuracy of the electrocardiogram. Cardiovasc Clinics 8:25, 1977 12. SPACm M S, BAllli, R C, BI.UMENSClIEIN, S O, BOINEAU, J P: Clinical implications of isopotential surface maps. Ann Int Med 69:919, 1968 13. l~h*wm, D M, KEH.ER, F W, IDEKmi, R E, Cox, J V',r, ZF:TTERt;EN, D G AND DOWDIE, R F: Values and limitations of isopotential mapping techniques in the detection and localization of multiple discrete epicardial events. J Electrocardiol 10:347, 1977 l'l. BI.UMI-:NSCIIF:IN,S D, SpACm M S, BOINI-:AU, J P , BARR, R C, GAI.I,IE, T M AND WAI.I.ACE, A G:

15.

16.

17.

18.

19.

20.

Genesis of body surface potentials in varying types of right ventricular hypertrophy. Circulation 38:917, 1968 DEAMBROiIGI, L, TACCARI)I, B ANn IIACCI, E: Body surface maps of heart potentials: Tentative localization of pre-excited areas in forty-two WolffParkinson-White patients. Circulation 54:251, 1976 MIRvls, D M: Electrogenesis of terminal QRS notches in normal subjects. J Electrocardiol 16:113, 1983 hlll~vls, D M: Differential electrocardiographic effects of myocardial ischemia induced by atrial pacing in dogs with varying location of coronary stenosis. Circulation, in press CiiAITMAN, B R, BOUIiASSA, M G, WA(;NIAI{T, P, CORBARA, F AND FERt;USON, R J: Improved efficiency of treadmill exercise testing using a multiple lead ECG system and basic hemodynamic exercise program. Circulation 57:71, 1978 Gtn.DscHi.At;EIt, N, SELZER, A ANI) COLIN, K:Treadmill stress tests as indicators of presence and severity of coronary artery disease. Ann Int Med 85:177, 1976 Bi)N(IRiS, P E, GRI':ENBI':R(;, P S, CIIRISTIS(IN, G W, CASTEi.I.FNI.:T, ]~i J ANI) EI.I.ESTAt), M H: Evaluation of R wave amplitude changes versus STsegment depression in stress testing. Circulation 57:904, 1978