Prior studies have demonstrated that ST-segment elevation (STE) in electrocardiographic (ECG) lead aVR in the setting of acute coronary syndromes may indicate the presence of severe stenosis or occlusion of the left main coronary artery (LMCA) or proximal left anterior descending coronary artery (LAD) (1- 10). Data regarding the significance of STE in lead aVR in the setting of exercise treadmill testing (ETT), however, are limited. Indeed, current practice guidelines indicate that it should be disregarded in the interpretation of exercise ECG findings (11), notwithstanding reports suggesting that STE in lead aVR may be predictive of hemodynamically significant LMCA or very proximal LAD stenosis in this setting (12- 14). Accordingly, in the present study, we tested the hypothesis that ETT-induced STE in lead aVR is a useful indicator of hemodynamically significant LMCA or ostial LAD stenosis.

A database search of patients undergoing cardiac catheterization at our institution between January 2008 and July 2009 was performed to identify 3 groups of patients: 1) those with significant LMCA or ostial LAD disease (≥50% luminal diameter reduction) in addition to any other coronary lesions (≥70% luminal diameter reduction for mid and distal LAD segments); 2) those with coronary artery disease (CAD) (≥70% luminal diameter reduction) but not LMCA or ostial LD disease; and 3) those free of significant CAD (LMCA and all 3 major vessels free of any stenosis ≥50%). Ostial LAD stenosis included any lesion proximal to the first septal perforator. All patients had undergone ETT according to the standard Bruce protocol with (n = 378) or without (n = 76) myocardial perfusion imaging (MPI; rest/stress 99mTc-methoxyisobutylisonitrile [MIBI]) ≤6 months before the clinically indicated cardiac catheterization. LMCA (n = 38) or ostial LAD stenosis (n = 42) was present in 75 patients (5 patients had both). The remainder had either CAD of varying severity that did not involve the LMCA or ostial LAD (n = 276) or no significant CAD (n = 103). All were consecutive cases in that they were included in unbiased fashion in the order in which they were encountered in the database search, provided they met study entry criteria.

Patients with acute coronary syndromes or prior coronary artery bypass grafting were excluded. Patients who had undergone pharmacological stress in conjunction with MPI also were excluded, as were those whose baseline ECG studies indicated left bundle branch block, intraventricular conduction delays ≥ 120 ms, left ventricular hypertrophy with marked strain pattern (down-sloping ST-segment depression [STD] ≥1 mm with biphasic or inverted T waves), or marked anterior T-wave inversions (the Wellens pattern) concerning for active ischemia or possible LMCA or proximal LAD stenosis. Leads V₁ to V₃ were not used for ischemia assessment in patients with right bundle branch block, although inferior leads and leads V₄ to V₆ were if the baseline was isoelectric. No patient had STE in lead aVR on rest electrocardiography. Mild STD (<1 mm) in 1 or more leads on rest electrocardiography was not cause for exclusion.

Pertinent clinical details of the study population are provided in (Table 1). The study received institutional review board approval.

SPECT MPI using 99mTc-MIBI (n = 370) or thallium (n = 8) was performed according to standard 1-day protocols and image acquisition guidelines (15). A dual-head Siemens gamma camera (E-CAM or C-CAM; Siemens Medical Systems, Erlangen, Germany) equipped with a low-energy, high-resolution collimator (32 views per camera head in a 64 × 64 matrix) was used for image acquisition. The patient performed treadmill exercise according to the standard Bruce protocol (11). One minute before completion of symptom-limited exercise (test end point as defined in current practice guidelines [11]), radiotracer was injected intravenously. Gated images were obtained 30 to 45 min later in the same fashion as noted previously.

Rest and stress myocardial perfusion images were analyzed objectively, in a quantitative fashion, using a standard 17-segment model (15) and a commercially available SPECT image analysis program (4DM-SPECT; Invia Medical Imaging Solutions, Ann Arbor, MI). Analysis was performed blinded to the results of coronary angiography. The program computed all MPI parameters used in logistic regression analyses. The parameters used were left ventricular ejection fraction (LVEF) (after stress), sum stress score, sum difference score, transient ischemic dilation (TID) ratio, percent ischemia in the LAD, left circumflex coronary artery (LCx), and right coronary artery zones, sum ischemia in the LAD and LCx zones, and total ischemia (all zones).

The following should be noted regarding the use of the image analysis program. All data were inspected for patient motion and extracardiac “hot spot” activity in the vicinity of the myocardium. Motion artifact had been addressed at the time of image acquisition, by reimaging the patient or using motion correction software or both. The analysis program is interactive and so permits the operator to constrain the area around the myocardium to avoid the inclusion of extracardiac hot spot activity in the bull's-eye plot. Thus, it permits avoidance of normalization artifact that could spuriously affect all MPI parameters generated by the program. Inability to eliminate a hot spot occurred in rare cases and was cause for elimination of the patient from the final database. Accordingly, all image datasets were technically satisfactory for analysis, albeit with operator and technologist assistance as noted.

All image data had been read previously for clinical purposes by experienced nuclear cardiologists and radiologists (consensus of 2 observers for all myocardial perfusion images) who, by study design, could not have known the cardiac catheterization data, which were obtained after the SPECT MPI studies. Defect severity was subjectively graded as 1 = mild, 2 = moderate, or 3 = severe. Visual assessment of ischemia (segments with full or partial resolution of stress MPI defects in the rest images) was available for all SPECT MPI studies.

The number of ischemic segments determined by visual analysis in combined LAD and LCx zones, because LMCA and ostial LAD stenosis detection was the objective of the study, as a predictor of LMCA or ostial LAD stenosis was tested by receiver-operating characteristic (ROC) curve analysis in the same fashion as the computer-determined parameters. Areas under the curves (AUCs) were compared statistically to determine how well visual analysis performed compared with computer-determined parameters.

Coronary arteriograms were analyzed visually by an experienced, invasive cardiologist blinded to ECG, ETT, and SPECT imaging data. Luminal diameter reduction ≥ 70% defined hemodynamically significant CAD in the right coronary artery, LCx, and LAD (mid and distal segments). The LMCA and ostial LAD (including the segment proximal to the first septal perforator) were considered to have significant stenosis with luminal diameter reductions ≥ 50%.

Patients in whom the LMCA or ostial LAD and all other major coronary vessels had no stenosis ≥ 50% were considered free of significant CAD.

The baseline and peak stress electrocardiograms of each patient were interpreted without knowledge of the SPECT MPI or coronary angiographic data. The configuration of the ST segment was noted (horizontal, down-sloping, or up-sloping) and the degree of displacement from the isoelectric line measured at 80 ms from the J point.

SPECT MPI variables were noted previously. Other parameters chosen for testing as univariate predictors of LMCA or ostial LAD stenosis were selected on the basis of a combination of physiological principles, clinical knowledge, and reference to the published research (16). The parameters tested are listed in the following. It should be noted that traditional CAD risk factors (age, sex, dyslipidemia, hypertension, smoking, diabetes, and family history of CAD) were included in univariate analysis.

In addition to lead aVR, lead aVL, the inferior leads (II, III, and aVF), and pre-cordial leads V₁ and V₄ to V₆ were evaluated, and the amount of STD or STE at 80 ms after the J point both at rest and during peak stress was tested in univariate regression analysis. In lead aVR, only horizontal or up-sloping STE was considered indicative of ischemia. J-point elevation only with down-sloping ST segment was not.

Total exercise time, reason for stopping the test (including the presence or absence of angina), beta-blocker or nitrate therapy, maximal rate-pressure product (RPP), metabolic equivalents, Duke score, and percent maximal predicted heart rate (MPHR) attained all were considered in univariate analysis.

Continuous variables are expressed as mean ± SD. Paired and unpaired Student t tests and chi-square tests were used as indicated. ROC curves for the detection of significant LMCA or ostial LAD stenosis were obtained for selected, statistically significant (after Bonferroni's correction) ETT, ECG, and MPI parameters identified in univariate logistic regression analysis. AUCs were computed by fitting a logistic regression model and were compared using a nonparametric method that took into account the correlated nature of the data (17). The method avoids strong assumptions on the distribution of the data. A sample size of 454 patients provided 80% power to detect a minimal AUC difference of 0.1 between 2 ROC curves using a 2-sided test with 5% type I error (18).

A stepwise multivariate regression analysis was also performed using all parameters that attained statistical significance in univariate analysis. The final multivariate model included only those univariate parameters that made statistically significant independent contributions to the prediction of LMCA or ostial LAD stenosis. SAS version 9.2 (SAS Institute Inc., Cary, NC) was used for statistical analysis. Bayesian analysis was performed, as described by others (19). P values <0.05 were considered statistically significant.

The mean age of the population was 62 ± 11 years (range 23 to 87 years), and 366 were men (81%) (Table 1). Significant LMCA (n = 38) or ostial LAD stenosis (n = 42) was present in 75 of 454 patients (17%; 5 patients had both). The patients with LMCA or ostial LAD stenosis were similar to the CAD group as a whole in terms of age (mean 65 ± 10 years) and sex distribution (62 of 75 [83%] were men). The incidence of single-vessel disease (1VD), double-vessel disease (2VD), and triple-vessel disease (3VD) without LMCA or ostial LAD involvement was 15%, 25%, and 9%, respectively, of patients with CAD. Isolated LMCA or ostial LAD stenosis was present in 34 of the patients with CAD (10%) and 7% of all patients. No significant CAD was present in 103 patients (23% of all patients studied).

Per study entry criteria, no patients had any of the following findings on baseline electrocardiography: left bundle branch block, left ventricular hypertrophy with strain, >1-mm STD in multiple lead groups, and deep T-wave inversions in anterior precordial leads (the Wellens sign). No patient had >0.5-mm STE in lead aVR at rest. Right bundle branch block was present in 21 patients (5%). However, none had >1-mm STD in inferior or V₄ to V₆ lead groups. Leads V₁ to V₃ were considered uninterpretable for ischemia in the face of right bundle branch block. Accordingly, all patients, per the entry criteria, had baseline electrocardiograms that were interpretable for ischemia.

None of the traditional risk factors for ischemic heart disease were statistically significant predictors of LMCA or ostial LAD stenosis per se (Table 2).

Univariate logistic regression analysis demonstrated that the best ETT or ECG predictor of significant LMCA or ostial LAD stenosis was stress-induced STE in ECG lead aVR (p < 0.0001, AUC: 0.82). Other ECG leads were significant univariate predictors (Table 2) but were not significant in multivariate analysis. ETT variables, save for Duke score, which was inversely correlated with LMCA or ostial LAD stenosis, failed to emerge as univariate predictors. Duke score, however, was not an independent predictor in multivariate analysis.

MPI variables predictive of significant LMCA or ostial LAD stenosis in univariate analysis are listed in (Table 2) and included TID (p < 0.001, AUC: 0.65), reversible LAD ischemia (p < 0.0001, AUC: 0.64), and post-stress LVEF (p < 0.03, AUC: 0.60). Sum stress score failed to emerge as a significant univariate predictor of LMCA or ostial LAD stenosis, although sum difference score, combined LAD and LCx ischemia, and total ischemia did (Table 2). The AUC for combined LAD and LCx ischemia determined by computer analysis (0.59) did not differ significantly from that determined by qualitative visual analysis (0.57) (p = 0.76) although the latter also was a significant univariate predictor (Table 2).

Although the AUC for STE in lead aVR was greater than that for reversible LAD ischemia and LVEF (both p values <0.0001), LAD ischemia and LVEF were included in the final model, because each made a statistically significant independent contribution to prediction of the outcome variable. Other significant univariate MPI predictors were tested but not included in the final multivariate model, because they failed to make statistically significant independent contributions to prediction of the outcome.

Univariate logistic regression analysis was performed of patients with 1VD, 2VD, and 3VD without LMCA or ostial LAD stenosis to determine the correlation of each with STE in lead aVR. There was no significant correlation between 1VD, 2VD, or 3VD without LMCA or ostial LAD stenosis and STE in lead aVR. In contrast, STE in lead aVR was a significant (p < 0.0001, AUC: 0.76) positively correlated predictor of selected cases of pure, isolated LMCA or ostial LAD stenosis (n = 34). Because leads V₅ and V₆ are reciprocal to lead aVR, STD in either of these leads as a univariate predictor of LMCA or ostial LAD stenosis was determined. Although both were correlated strongly (lead V₅: p < 0.0001, AUC: 0.69; lead V₆: p < 0.0001, AUC: 0.70) (Table 2), the AUC for each was significantly less than that for STE in lead aVR (0.82) (both p values <0.001). Finally, although stress-induced STE in lead V₁ was also a significant predictor of LMCA or ostial LAD stenosis (p < 0.0001, AUC: 0.59) it too was significantly less than that for STE in lead aVR (p < 0.0001 for AUC comparison) and did not add independent information in multivariate analysis (Table 2).

Using a cut-point of 1-mm STE in lead aVR, chi-square tests were used to compare the prediction of LMCA or ostial LAD stenosis with 1VD, 2VD, and 3VD without LMCA or ostial LAD stenosis. Thus, 76% of patients with LMCA or ostial LAD stenosis had 1-mm STE in lead aVR with stress, compared with 17%, 27%, and 39%, respectively, with 1VD, 2VD, and 3VD without LMCA or ostial LAD stenosis (chi-square = 42, 38, and 12 respectively; all p values < 0.005 vs. LMCA or ostial LAD; df = 1 for all chi-square tests). Only 8 patients (8%) without significant CAD had 1-mm STE in lead aVR with stress (chi-square = 84, df = 1, p < 0.005 vs. LMCA or ostial LAD).

The optimal cut-point for the amount of stress-induced STE in lead aVR was defined by the greatest sum of sensitivity and specificity for the detection of significant LMCA or ostial LAD stenosis. Cut-points of 0.5-, 1.0-, 1.5-, and 2.0-mm horizontal STE were tested. The greatest sensitivity-specificity sum (1.56) was reached at 1.0-mm STE. Thus, sensitivity was 75%, specificity was 81%, positive predictive accuracy was 44%, negative predictive accuracy was 94%, and overall predictive accuracy was 80%. A Bayesian plot of post-test versus pre-test probability of significant LMCA or ostial LAD stenosis on the basis of 1-mm STE in lead aVR is shown in (Figure 1). It demonstrates, as anticipated, that the test performed best in the intermediate pre-test probability range and for the current population raised the pre-test probability almost 3-fold from 17% to 45% after the test.

This study tested the hypothesis that ETT-induced STE in lead aVR is a useful indicator of hemodynamically significant LMCA or ostial LAD stenosis. The data obtained support this hypothesis. Sensitivity, specificity, and overall predictive accuracy all were approximately 75% to 80%, with high negative predictive accuracy (94%), but only modest positive predictive accuracy (44%). Bayesian analysis of the data (Figure 1), however, demonstrated that the finding of ≥1-mm horizontal STE in lead aVR almost tripled the post-test odds of finding significant LMCA or ostial LAD stenosis and so should not be ignored in the interpretation of electrocardiograms obtained with ETT (Bruce protocol).

The relationship between SPECT MPI findings and STE in lead aVR as predictors of significant LMCA or ostial LAD stenosis is noteworthy. First, as anticipated, several previously reported SPECT MPI findings (e.g., TID [(20)], percent reversible LAD ischemia [21]) were significant univariate predictors. However, by ROC analysis, these variables were not as strong as STE in lead aVR (Table 2). Nonetheless, both LVEF (after stress) and percent reversible LAD ischemia added independent predictive power to STE in lead aVR as an indicator of LMCA or ostial LAD stenosis. Because true rest LVEFs were not obtained, we cannot know if the post-stress LVEF reflects a decline from rest, as reported by others (22), in cases of reversible, particularly LAD, ischemia, although this seems likely and is consistent with the fact that TID was also a significant univariate but not multivariate predictor.

Because lead aVR monitors the left ventricular cavity (viewing from apex to base [23]), in theory, it should be sensitive to proximal, anterior septal ischemia, which will appear as an injury current (i.e., elevation of the baseline) when viewed from the left ventricular cavity (24). Accordingly, while correlated with classical SPECT MPI findings of LMCA or ostial LAD stenosis such as TID and extensive LAD ischemia (Figure 2), it will also detect ischemia related to LMCA or ostial LAD stenosis in the absence of such findings (Figure 3), a heretofore underappreciated observation and likely accounting for its stronger predictive value with respect to classical SPECT MPI findings.

Several additional issues regarding SPECT MPI as used in the present study should be considered. First, the computer program used for quantitative analysis of the images has been previously validated and is widely used (25- 26). The images were all technically satisfactory for analysis, because careful attention was paid in the acquisition phase to avoidance or correction of motion artifact and in the processing phase to the exclusion of intense subdiaphragmatic activity, which could interfere with proper normalization of myocardial activity and thereby generate spurious polar plots. The fact that only treadmill exercise tests were used and that patients on average reached 84% MPHR and RPPs of 22,000 mm Hg/min and completed the first 2 stages of the Bruce protocol plus 36 s of the third stage indicates in general that they had adequate stress to detect hemodynamically significant CAD, especially LMCA or ostial LAD, and that image quality was much less likely to be compromised by subdiaphragmatic activity often seen with adenosine stress.

Furthermore, it is standard clinical practice in cases in which patients fail to reach 85% MPHR and the electrocardiogram does not show evidence of ischemia to interpret the ECG findings as nondiagnostic because of failure to reach 85% MPHR. American Society of Nuclear Cardiology guidelines (27) also advise reading SPECT myocardial perfusion images in light of relevant clinical and stress test information and modifying the interpretation accordingly, especially in equivocal cases. Thus, it is recognized that studies that appear normal at submaximal levels of stress may be falsely negative for CAD. Accordingly, in the present study, any patients whose exercise levels may have been submaximal as judged by failure to reach 85% MPHR certainly would not, as a result, have advantaged stress electrocardiography over the MPI study for the detection of LMCA or ostial LAD disease. Indeed, the reverse is likely the case, because SPECT MPI in general is more sensitive for the detection of CAD than stress electrocardiography alone (87% [(28)] vs. 67% [11], respectively), although stress electrocardiography generally does not use STE in lead aVR as an indicator of CAD (i.e., positive test results).

Moreover, because LMCA or ostial LAD disease often causes signs and symptoms of ischemia, which may be severe, at low work load, it is unsurprising that various exercise parameters (e.g., metabolic equivalents, exercise time, percent MPHR) failed to emerge as univariate predictors. This is so because low levels of these parameters also may be associated with beta-blocker therapy, poor physical condition of the patient, or inadequate effort in the absence of CAD or with only mild or moderate disease. Thus, patients with very severe disease and those with mild or moderate or even no disease may have similar, limited exercise capacity, but for very different reasons. The Duke score accounts for many of these factors and so was a univariate predictor (inverse relation as expected) of LMCA or ostial LAD stenosis. Furthermore, the predictive power of SPECT MPI for the detection of LMCA or ostial LAD stenosis may be confounded by the precise morphology of the lesion, which will have an important effect on its hemodynamic severity (29). Thus, at similar work loads, the scan may show evidence of severe ischemia in a patient with a functionally very severe lesion (Figure 2) or no ischemia (Figure 3) despite more than adequate work load in a patient with hemodynamically less severe disease but still sufficient to cause endocardial ischemia, which would be manifest only electrically, particularly in intracavitary lead aVR (Figure 3).

Considerable attention was paid to the possibility that STE in lead aVR was not so much an indicator of LMCA or ostial LAD stenosis per se but more simply a reflection of multivessel CAD. To this end, we examined the univariate predictive value of STE in lead aVR for 2VD and 3VD absent LMCA or ostial LAD stenosis involvement. Neither was found to be a significant univariate correlate. In contrast, STE in lead aVR remained a highly significant (p < .0001) predictor of LMCA or ostial LAD stenosis in a subset of patients (n = 34) without other CAD. Furthermore, 76% of patients with significant LMCA or ostial LAD stenosis demonstrated ≥1-mm horizontal STE in lead aVR in comparison with only 27% and 39% of patients with 2VD and 3VD, respectively, without LMCA or ostial LAD stenosis (chi-square = 38 and 12, respectively, both p values <0.005, df = 1 vs. LMCA or ostial LAD disease). Finally, while STD in leads V₅ and V₆, which are reciprocal to lead aVR, is a common ECG finding in response to ETT in patients with multivessel CAD (24) and was a significant univariate predictor of LMCA or ostial LAD stenosis, the AUC for both parameters was significantly less than that for STE in lead aVR (p < 0.001) and failed to provide additional independent predictive power in multivariate analysis. The same was true of STE in lead V₁. Important ETT-related variables such as beta-blocker therapy, percent MPHR, and metabolic equivalents were not univariate predictors of LMCA or ostial LAD stenosis. The Duke score correlated inversely in univariate but not multivariate analysis.

Comparison of the results of the present study with those of prior investigations in this area is difficult. One large study (n = 557 patients) classified ECG data on the basis of a finding of ≥0.05-mm STE in lead aVR and used pharmacological stress, primarily adenosine, in 78% of patients (12). Accordingly, the methods and diagnostic criteria used differ importantly from those used in the present study. Nonetheless, these investigators (12) did observe that STE in lead aVR was more predictive of anterior wall defect on SPECT 99mTc-MIBI scans than 1-mm STD in lead V₅. However, only 21% of their patients underwent heart catheterization, and of these, the incidence of LMCA or ostial LAD stenosis was not given, although they did report a trend (p = 0.06) for patients with STE in lead aVR (8 of 37) to have 2VD with LAD involvement (≥70% luminal diameter reduction) compared with those who did not (4 of 77) (12). Furthermore, because adenosine stress generally induces regional myocardial blood flow disparities between normal and stenotic vascular perfusion territories but not ischemia, while ETT is capable of doing both, it would be anticipated that STE in lead aVR will be more predictive of LMCA or ostial LAD stenosis with exercise, as opposed to adenosine stress.

Another smaller study (n = 106) retrospectively investigated a highly selected population, those with Duke scores ≤−11, and compared those with and without STE ≥ 1 mm in lead aVR for the incidence of LMCA stenosis (≥50% luminal diameter reduction) (13). The investigators observed that STE in lead aVR alone had sensitivity of 93% and specificity of 49% for the detection of LMCA stenosis. ROC analysis was not performed, other potential predictors were not considered, and all patients had associated 3VD. The later observation (100% with 3VD) may well account for the apparent high sensitivity and low specificity of the finding in this patient population. While the investigators (13) also reported that associated STE in lead V₁ improved the specificity of the diagnosis for the LMCA to 82% with only a small decrease in sensitivity to 86%, the independent contribution of concomitant 3VD could not be assessed.

A clinical review of LMCA stenosis in the 1980s (1) reported on 120 consecutive patients with LMCA stenosis, of whom 65% had unstable angina and 89 underwent ETT, whose results were abnormal in 99% of those tested. The investigators observed that the most common ECG finding on ETT was “ST-segment depression of 2 mm or more in leads V₄, V₅, and V₆, and ST segment elevation in leads V₁ and aVR” (1). Concomitant CAD was not discussed in relationship to the ECG findings on ETT, which apparently was performed in a sizable percent of patients who presented with what was thought to be unstable angina. Accordingly, the patient population was not comparable with those reported in the present study, because the prior investigation included only patients with LMCA stenosis, 65% of whom presented with unstable angina and 68% with post–myocardial infarction angina.

Although all patients meeting the entry criteria over the time period specified were studied, the investigation nevertheless was retrospective in nature and so shares all the limitations of such studies. While patients with more severe MPI reversible defects or markedly positive ECG responses to ETT undoubtedly were more likely to be referred for cardiac catheterization, the fact that current guidelines for interpretation of ECG findings on ETT recommend disregarding STE in lead aVR (and lead V₁) (11), if anything, would suggest that such patients would be less rather than more likely to be referred for catheterization on the basis of the ECG findings alone. Finally, although the coronary angiograms were interpreted visually, with all the known limitations of so doing (30), this method conforms to everyday, “real-world” clinical practice, and so the correlations made would have the precision of routine clinical practice.

Although often disregarded in the interpretation of ECG findings on ETT, lead aVR, which monitors the left ventricular cavity, contains important diagnostic information. Data obtained in the present study indicate that stress-induced horizontal STE ≥1 mm in this lead substantially increases the post-test probability of significant LMCA or ostial LAD stenosis. Furthermore, as demonstrated by (Figure 3) and univariate analysis, STE in lead aVR may detect LMCA or ostial LAD stenosis when SPECT MPI does not, notwithstanding the fact that in multivariate analysis, SPECT MPI parameters, post-stress LVEF, and percent reversible LAD ischemia were statistically significant predictors of LMCA or ostial LAD stenosis.

The authors wish to acknowledge the expert assistance of all members of the Massachusetts General Hospital nuclear cardiology laboratory in the performance of ETT and SPECT MPI.