Author + information
- Received October 6, 1997
- Revision received March 13, 1998
- Accepted April 23, 1998
- Published online August 1, 1998.
- Rachel L Murkofsky, MDa,
- George Dangas, MDa,
- Joseph A Diamond, MDa,
- Davendra Mehta, MD, PhD, FACCa,
- Abraham Schaffer, MD∗,a and
- John A Ambrose, MD, FACCa,* ()
- ↵*Address for correspondence: Dr. John A. Ambrose, Cardiovascular Institute (Box 1030), The Mount Sinai Hospital, One Gustave L. Levy Place, New York, New York 10029
Objective. We sought to determine whether a prolonged QRS interval duration is associated with decreased left ventricular (LV) systolic function.
Background. The 12-lead electrocardiogram (ECG) is a routine test for suspected cardiac disease. Although several scoring systems have been devised to estimate LV systolic function, no studies have examined the direct relationship between QRS duration alone and LV systolic function.
Methods. We analyzed the standard 12-lead surface ECG of 270 consecutive patients, referred for radionuclide ventriculography. Patients (n = 44) with bundle branch blocks, atrial flutter or fibrillation, pacemaker rhythm, recent myocardial infarction or bypass surgery, and patients on antiarrhythmic drugs were excluded. In the remaining patients (n = 226), we correlated the QRS duration on standard resting ECG, and the resting LV ejection fraction (EF), end-systolic and end-diastolic counts (ESC and EDC, respectively; LV volume indices), as obtained by radionuclide angiography. We used a multivariate analysis to identify independent predictors of reduced ventricular function entering QRS duration, the previously described R-wave score and clinical variables in our model.
Results. The QRS duration in the abnormal EF group was significantly longer than in the normal EF group (0.102 vs. 0.091 s, p < 0.0001). A QRS duration >0.10 s was highly specific (83.6%), but modestly sensitive (43.8%), for the prediction of abnormal EF. Furthermore, an abnormal EF was predicted with incrementally increased specificity (83.6% to 99.3%) and a corresponding decrease in sensitivity (43.8% to 13.8%) for each 0.01-s increase in the definition of prolonged QRS (from >0.10 to >0.12 s). Accordingly, the positive likelihood ratio for the prediction of decreased LV function was increased from 2.67 to 19.7 as the definition of prolonged QRS duration was increased from >0.10 to >0.12 s. In the multivariate analysis, a prolonged QRS duration and a low R-wave score were the only independent predictors of decreased LV systolic function.
Conclusions. Prolonged QRS duration (>0.10 s) obtained from a standard resting 12-lead ECG is a specific, but relatively insensitive indicator of decreased LV systolic function. Further prolongation of the QRS had a higher specificity for decreased LV EF and a higher positive likelihood ratio for predicting abnormal LV EF.
The 12-lead electrocardiogram (ECG) is the most readily available noninvasive test for the detection of cardiac disease. Prior to the technologic revolution in noninvasive cardiac imaging of the 1970s and 1980s, the ECG and chest x-ray were the tests ordinarily ordered for cardiac diagnosis. Today, the utility of the ECG has been overshadowed by the ability of echocardiography and nuclear cardiology to evaluate possible cardiac dysfunction. However, several studies have shown that a normal 12-lead ECG in patients is a relatively sensitive and specific marker for normal left ventricular (LV) function (1–4). In patients with LV dysfunction and prior myocardial infarction, various electrocardiographic scoring systems have been developed to estimate LV function (5–8), but their utility has been limited (9,10). It would be clinically useful and cost effective if the standard 12-lead ECG could be used to predict LV dysfunction.
Anecdotally, we observed that a prolonged QRS duration in the absence of a typical bundle branch block configuration was present in patients with decreased LV function. To systematically study this association, we correlated QRS duration by computer measurements, which were manually confirmed on the 12-lead ECG, with resting LV ejection fraction (EF), end-systolic and end-diastolic counts (ESC and EDC, respectively; LV volume indices), as obtained by radionuclide ventriculography.
The records of 270 consecutive patients, who were referred to the Nuclear Cardiology Laboratory of the Mount Sinai Hospital, New York, NY, from July 1, 1994 through June 30, 1996 for radionuclide exercise ventriculography, were reviewed. At the time of testing, a standard resting 12-lead ECG was obtained. Data collected included the baseline patient characteristics, resting 12-lead ECGs, end-systolic counts (ESC), and end-diastolic counts (EDC), and LV EFs at rest. A history of ischemic heart disease was defined as documented coronary artery disease (CAD) by cardiac catherization, a history of previous myocardial infarction, or typical angina pectoris. All radionuclide data had been previously calculated at the time of testing and therefore analyzed independently of the ECG.
Patients with known causes of a prolonged QRS duration were excluded—patients on antiarrhythmic drugs (n = 2), with a history of recent myocardial infarction (n = 1), with typical left or right bundle branch block (n = 22), or pacemaker rhythm (n = 6), as well as patients with insufficient ECG or radionuclide data (n = 16 including 11 patients with atrial fibrillation or flutter). Some patients met more than one exclusion criteria. Of the original 270 patients, 226 were included in the final analysis.
Standard resting 12-lead ECGs were obtained at the time of radionuclide exercise ventriculography at 25 mm/s and 1 mV/cm standardization with a Case 12/15 Marquette Stress System (Marquette Electronics Inc., Milwaukee, WI). ECGs were interpreted using standard criteria (11). The widest QRS duration on each ECG was manually measured after magnification to 140% by one of the investigators (R.L.M.), who was blinded to the nuclear data. The method of QRS measurement was validated by the same investigator in 37 sample ECGs: compared with computer measurements of QRS duration, the correlation coefficient (r) was 0.95, with a SE of 0.06, p < 0.0001 (Fig. 1).
We also analyzed all ECGs according to previously published electrocardiographic criteria for the prediction of LV dysfunction (5–8). Because our study population had a low incidence of electrocardiographically evident Q-wave myocardial infarctions (11.6%), we did not use scoring systems that had been proven useful in the determination of infarct size and estimation of LV function shortly after a myocardial infarction (5–7). We used the R-wave scoring system, which was described as predictive of LV EF in an outpatient population, regardless of the presence of a recent infarction (8); the sum of the R-waves (in mV) in leads aVL, aVF, and v1–v6was obtained from each ECG tracing.
Gated equilibrium radionuclide ventriculography was performed at rest after in vivo red blood cell labeling with 25 to 30 mCi of Tc99mpertechnetate. Images were obtained using a 409 MA Elscint gamma-camera (Elscint Ltd., Haifa, Israel) equipped with a parallel-hole general-all-purpose collimator (APC-3, Elscint Ltd.). The energy window (20%) was peaked at 140-keV. Images at rest were obtained in three standard views: anterior, left lateral and left anterior oblique. Data were acquired by electrocardiographic gating with 16 frames per R-R interval. The data were spatially and temporally smoothed. Global LV EF values were calculated from the left anterior oblique image using standardized semiautomated software (Technicare, Milwaukee, WI), by an experienced operator. In our Nuclear Cardiology Laboratory, reproducibility, with a 95% confidence interval, has been documented at 96.0%. The lower limit of normal LV EF using this software has previously been defined as 45%; therefore, EF < 45% was defined as abnormal. End-systolic and end-diastolic frames were automatically determined and the peak counts from these frames were recorded (LV ESC and LV EDC, respectively). These peak counts were used as estimates of LV end-systolic and end-diastolic volumes.
Statistical comparisons were performed utilizing the two-tailed Fisher’s exact test for categorical variables, and the Student’s two-tailed t-test for continuous variables; a p < 0.05 was considered significant. Three separate analyses were performed for the prediction of an abnormal EF (<45%), based on three definitions of prolonged QRS durations: >0.10, >0.11, or >0.12 s. In addition, the sensitivity and specificity of ECG criteria for predicting abnormal EF values were determined; the positive likelihood ratio was calculated as sensitivity/(1-specificity). Linear logistic regression analysis was used to assess correlations between continuous variables; the correlation coefficient r was calculated.
The R-wave score was correlated with the presence of an abnormal EF, the QRS duration, the LV ESC, the LV EDC and the LV EF. We used the R-wave score <4 mV as a nominal criterion for the prediction of decreased LV EF, as defined previously (8). We conducted a multivariable stepwise logistic regression analysis to determine the independent correlates of an abnormal EF (outcome variable); age, gender, hypertension, diabetes, presence of coronary disease, the nominal R-wave score criterion and the QRS duration criteria (each one individually) were included in our multivariable model. Because R-wave score, QRS duration, and LV EF are continuous variables, we also entered them as such in the above multivariable model.
The entire study population (n = 226) was divided according to the presence of a normal (≥45%) or abnormal (<45%) resting LV EF, as previously defined. The patient characteristics of each group are shown in Table 1. The patients with an abnormal LV EF were older (mean age 58 vs. 50 years), had a higher incidence of male gender and a previous history of ischemic heart disease compared to patients with normal EF. Examples of ECGs from three of the patients are shown in Figure 2, with their measured QRS duration.
Pertinent ECG and radionuclide data of the two groups are shown in Table 2. The QRS duration in patients with an abnormal EF was significantly longer compared with that in patients with normal EF (p < 0.0001). A prolonged QRS duration defined as a QRS >0.10 s, was detected in 59 (26.1%) ECG recordings: of these 59 patients, 35 (43.8%) had an abnormal EF and 24 (16.4%) had a normal EF (p < 0.001). When prolonged QRS was defined as QRS > 0.11 s or QRS > 0.12 s, there were still statistically significant differences in the incidence of an abnormal LV EF among patients with or without prolonged QRS.
In addition to EF, a prolonged QRS duration was associated with significantly increased LV ESCs and LV EDCs, utilizing all three definitions of an abnormal QRS duration (Table 3). There were also direct, moderate correlations (r = 0.54, p < 0.0001) of QRS duration with LV ESC, LV EDC, and LV EF in the entire study population (n = 226): r = 0.54, r = 0.54, r = −0.36, respectively (p < 0.001 for all).
The sensitivity and specificity of a prolonged QRS duration for predicting an abnormal EF are shown in Table 3. For each successive 0.01-s increase in the definition of prolonged QRS duration (from >0.10 to >0.12 s) there was an increase in specificity from 83.6% to 99.3%, with a corresponding decrease in sensitivity from 43.8% to 13.8%. The positive likelihood ratio of the QRS duration for predicting decreased LV function increased from 2.67 to 19.7 as the definition of prolonged QRS duration was increased from >0.10 to >0.12 s. Furthermore, as QRS duration increased from 0.10 to 0.12 s, mean EF decreased from 40.3% to 30.9% (p < 0.05).
We also analyzed the effects of prolonging the QRS duration on LV EF within the subgroup of 80 patients with abnormal LV EFs (Table 4). Patients with a prolonged QRS duration (>0.10 s) had lower LV EFs compared to patients with a normal QRS duration. These differences were present for all three definitions of a prolonged QRS duration but did not reach statistical significance.
R-wave score and QRS duration criteria
Patients with an abnormal EF had a significantly lower R-wave score compared to patients with an EF ≥ 45%; 5.9 ± 0.3 vs. 7.7 ± 0.3 mV, p < 0.001. The R-wave score correlated weakly with the value of the LV EF (r = 0.30, p < 0.001) but did not correlate significantly with the LV ESCs, the LV EDCs or the QRS duration. The R-wave score was not significantly different among patients either with a QRS duration ≤0.12 vs. > 0.12 s, a QRS duration ≤ 0.11 vs. >0.11 s or a QRS ≤ 0.10 vs. >0.10 s.
An R-wave score <4 mV had a 33% sensitivity and a 95% specificity for the prediction of decreased LV EF (positive likelihood ratio = 6.6, negative likelihood ratio = 1.4, p < 0.001). There was no significant correlation between the nominal R-wave criterion and the nominal QRS duration criteria.
Combination of the criteria for prediction of an abnormal EF yielded the following results: 92% of patients with both an R-wave score <4 mV and a QRS duration >0.10 s had a decreased LV EF, whereas only 22% of patients with an R-wave score ≥4 mV and a QRS duration ≤0.10 s had decreased LV EF; 53% of patients with either criterion alone had a decreased LV EF (p < 0.001). Accordingly, the combined criterion of R-wave score <4 mV and QRS duration >0.10 s had a 15.3% sensitivity, a 99% specificity, a positive likelihood ratio of 15, and a negative likelihood ratio of 1.2 for the prediction of a decreased LV EF (p < 0.001).
In the multivariate analysis, we evaluated the independent predictors of an abnormal EF (<45%). Our model evaluated the R-wave criterion and each of the QRS duration criteria, and controlled for age, gender and the presence of diabetes, hypertension or coronary heart disease. Only the ECG criteria were independent predictors of an abnormal EF: Both the R-wave <4 mV criterion (odds ratio 7.3, 95% confidence interval 2.9 to 20.3) and the QRS duration >0.10-s criterion (odds ratio 2.6, 95% confidence interval 1.3 to 5.3) were independent predictors of a decreased EF (p < 0.01 for both). In the multivariable model, which evaluated R-wave score and QRS duration as predictors of LV EF entering these three parameters as continuous variables, results were qualitatively similar to the above model: age, gender and the other clinical variables were not significantly associated with LV EF, whereas both the QRS duration and the R-wave score independently correlated with LV EF (p < 0.0001 for both).
In this study a prolongation of the QRS duration (>0.10 s) on a standard 12-lead ECG was associated with lower LV EF and larger end-systolic and end-diastolic volumes, as determined by radionuclide gated blood pool imaging. The high specificity of the 12-lead ECG for the prediction of abnormal LV systolic function suggests that in patients with QRS duration, >0.10 s, there is a high likelihood that the resting LV EF is abnormal. The resting 12-lead ECG is widely available for all patients with suspected or proven cardiac disease. Although valuable information to diagnose the rhythm and presence or absence of acute or remote myocardial infarction and left ventricular hypertrophy is contained in the resting ECG, the utility of the QRS duration as a marker of LV systolic function has been overlooked. Previous studies indicated that a normal 12-lead resting ECG is associated with normal LV function in 92% to 95% of cases (1–4). QRS scores incorporating several ECG variables have been devised by several investigators. Palmeri et al. (5)and Wagner et al. (6)used a 29-point system based on the duration of Q and R waves and on the ratios of R-to-Q amplitude and R-to-S amplitude. Palmeri et al. (5)showed certain QRS scores to be proportional to the severity of wall–motion abnormalities (determined by radionuclide gated blood pool scanning) and to have inverse correlations with measured LV EF. Roubin et al. (7)used the same scoring system to show that a QRS score of ≥7 had a specificity of 97% and a sensitivity of 59% for predicting an LV EF of <45%. Askenazi et al. (8)demonstrated that the sum of the R-waves in leads aVL, aVF, and v1to v6correlated with the LV EF, and an R-wave sum of <4 mV was the best predictor of a decreased EF. However, the utility of such scoring systems has been questioned. Fioretti et al. (9)showed that QRS score of Wagner et al. (6)was of little value in estimating LV EF. Young et al. (10)found that the correlation between the modified QRS score of Wagner et al. (6)and LV EF to be only fair, and the sum of R-wave voltage criterion of Askenazi et al. (8)to correlate poorly with LV EF.
In the present study, we assessed the direct relationship between QRS duration and resting LV EF. The determination of QRS duration was highly predictive of resting LV function in this cohort of stable patients referred to a nuclear cardiology laboratory for assessment of LV systolic function. A prolonged QRS (>0.10 s), was highly specific, but relatively insensitive, for predicting LV dysfunction. Thus, although a QRS >0.10 s was highly associated with an abnormal resting LV EF, a normal QRS duration ≤0.10 s did not reliably exclude abnormal LV EF. Furthermore, within the subgroup of patients with abnormal EFs (n = 80; Table 4), patients with a more prolonged QRS (>10 s) had a trend toward worse LVEF. Although no additional statistically significant differences were found when the other two, more strict, definitions of QRS prolongation were used in this subgroup, the sample size was small and may have obscured significant differences.
Comparison with other ECG scoring systems
We compared our QRS duration criteria for the prediction of a decreased EF to the previously described R-wave scoring criterion (8). As only a small number (11.6%) of ECGs had evidence of a Q-wave myocardial infarction, we did not use other described ECG scoring systems for the prediction of EF in postinfarction patients (5–7). In a multivariate analysis we determined that both a prolonged QRS duration (>0.10 s), and a low R-wave score (<4 mV) were independent predictors of an abnormal EF. Additionally, the two criteria did not correlate with each other, suggesting that they offer complementary information in the ECG-based prediction of a decreased EF. The combination of the two criteria (R-wave <4 mV and QRS duration >0.10 s) increased the positive likelihood ratio for the prediction of decreased LV EF compared with either criterion alone. Thus, the two criteria should be applied routinely in the prediction of LV EF based on the resting surface ECG tracing. Patients who have not suffered an anterior myocardial infarction should probably be evaluated with the QRS criterion first, whereas postinfarction patients should initially be evaluated with the R-wave criterion.
LV systolic dysfunction is known to occur with excessive widening of the QRS complex associated with left bundle branch block, ventricular paced rhythm, and ventricular premature beats (12–15). LV EF has not previously been directly correlated with an increase in QRS duration of sinus beats that conduct via the normal His–Purkinje system. It has been suggested that in the presence of myocardial disease, dilatation of the intracellular T-tubular system and the presence of fibrillar material within the lumen of the T-tubules might interfere with the conduction of the impulse from the cell surface into the depth of the cell via the T-tubular system (16). Altered microanatomy related to ischemic or nonischemic myopathy can also result in impaired sodium conductance, and thereby decreased conduction velocity (dv/dtmax), resulting in altered intraventricular conduction velocity. It is not known whether these results are due to long-term mechanical LV overload or another unknown initiating factor.
Diffuse interstitial collagen accumulation might also affect cell-to-cell communication by desmosomes and lead to a prolonged QRS duration. Although the orientation of myocardial fibrils is relatively unchanged in cardiomyopathy, it is possible that conditions governing anisotropic conduction are altered in such a way that longitudinal conduction predominates, but relatively slower transverse propagation from endocardial to epicardial surfaces might still become exaggerated because of diffusely impaired electrical continuity between adjacent myofibrillar bundles (16). Furthermore, longer impulse path in the dilated heart with increased LV mass would be expected to result in a longer total ventricular activation time and wider QRS. It still remains unclear, however, why in the presence of a decreased LV EF and increased LV volumes, some patients still have a normal QRS duration.
Several limitations to our study should be considered when evaluating the clinical implications of our findings. This patient population consisted mostly of stable outpatients and a few stable inpatients without acute cardiac decompensation. Thus, these data cannot necessarily be applied to more acutely presenting patients in other than routine evaluation. In addition, although bundle branch blocks were excluded, patients with left anterior hemiblock were not excluded. However, only 5.3% of cases in our study had left anterior hemiblock, and exclusion of these patients did not alter the results of the study.
In conclusion, these data suggest that the presence of a nonspecific prolonged QRS duration (>0.10 s) on a standard resting 12-lead ECG in the absence of typical features of bundle branch block is indicative of decreased resting LV systolic function. The mechanism for such a finding needs to be better elucidated so as to determine why certain patients with diminished LV function may still maintain a normal QRS duration. Further studies are needed to assess the utility of this finding in patients with acute cardiac decompensation. Larger studies may also be warranted to assess the association, if any, between the severity of LV dysfunction and QRS duration. In the meantime, the finding of a prolonged QRS on a patient’s ECG can serve as a bedside clue to the presence of decreased LV systolic function.
We thank Sylvan Wallenstein, PhD, Department of Biomathematical Sciences for expert assistance in statistical analysis, and Perwaiz Meraj, BS, for his indispensable help in manuscript preparation.
↵∗ Deceased, 1997.
- coronary artery disease
- end-diastolic counts
- ejection fraction
- end-systolic counts
- left ventricle
- Received October 6, 1997.
- Revision received March 13, 1998.
- Accepted April 23, 1998.
- American College of Cardiology
- Swartz M.H,
- Pichard A.D,
- Meller J,
- Teichholz L.E,
- Herman M.V
- Christian T.F,
- Chareonthaitawee P,
- Miller T.D,
- Hodge D.O,
- Gibbons R.J
- Wagner G.S,
- Freye C.J,
- Palmeri S.T,
- et al.
- Roubin G.S,
- Shen W.F,
- Kelly D.T,
- Harris P.J
- Fioretti P,
- Brower R.W,
- Lazzeroni E,
- et al.
- Young S.G,
- Abouantoun S,
- Savvides M,
- Madsen E.B,
- Froelicher V
- Chou T.-C
- Park R.C,
- Little W.C,
- O’Rourke R.A
- Moulton K.P,
- Medcalf T,
- Lazzara R
- Ferrans V.J,
- Massumi R.A,
- Shugoll G.I,
- Ali N,
- Roberts W.C