Sandeep A. Kamath, MD; Januario de P. Meo Neto, MD; Russell M. Canham, MD; Fatema Uddin; Kathleen H. Toto, MSN, ACNP; Lauren L. Nelson, MS, FNP-C; Patricia A. Kaiser, RN; James A. de Lemos, MD; Mark H. Drazner, MD, MSc 

Am Heart J.  2006;152(2):355-361.  ?2006 Mosby, Inc.

Abstract and Introduction

Abstract

Background: The prognostic implications of low QRS voltage on the electrocardiogram (ECG) in heart failure (HF) are not well characterized.
Methods: We manually measured and summed the QRS voltage in all 12 leads of the ECG (∑QRS) in two cohorts: (1) 415 patients with a low left ventricular ejection fraction followed up in a HF clinic ("clinic cohort") and (2) 100 subjects with advanced HF who had an ECG within 1 year preceding cardiac transplantation ("pretransplant cohort"). Low voltage was defined as the lowest quartile of the clinic cohort (∑QRS <12 mV) and its prevalence was compared in the two cohorts. The associations of low voltage with 1-year outcomes were assessed in the clinic cohort.
Results: In the clinic cohort, the frequency of low voltage was higher in New York Heart Association class 4 versus class 1-3 patients (34% vs 22% respectively, P = .04). The frequency of low voltage in the pretransplant cohort (47%) was twice that of the clinic cohort (24%, P < .001). After 1 year of follow-up in the clinic cohort, low ECG voltage was associated with a higher rate of death (14% vs 5%, P = .008) and the composite end point of death or HF hospitalization (35% vs 20%, P = .004). These associations persisted in multivariable analyses adjusting for important confounders.
Conclusions: Low ECG voltage is a marker of the severity of HF and is a risk factor for adverse outcomes in patients with systolic HF at 1 year.

Introduction

The 12-lead electrocardiogram (ECG) is a widely available tool that has prognostic value in heart failure (HF) in the presence of left ventricular systolic dysfunction. For example, left bundle branch block^[1] and prolonged QRS duration^[2] have been shown to be independent predictors of mortality in HF. However, few studies have addressed whether ECG voltage likewise offers prognostic value in HF.

Although increased ECG voltage as seen in left ventricular hypertrophy (LVH) is associated with increased mortality and risk for development of CHF in the general population,^[3] there are limited data about whether ECG LVH conveys prognostic information once a diagnosis of HF is made.^[4] Furthermore, others suggest that low ECG voltage may be associated with clinical deterioration in HF.^[5,6] We therefore undertook this study to determine whether ECG voltage conveys prognostic information in HF with a reduced left ventricular ejection fraction in the modern era, recognizing that either increased voltage (LVH) or decreased voltage could be risk markers for adverse outcomes in this setting.

Methods

Patient Populations

This is a retrospective study. From a consecutive series of patients (N = 551) evaluated in the Parkland Memorial Hospital CHF clinic^[7,8] ("clinic cohort") between June 1998 and June 2002, we included all patients (n = 415) who had a qualitatively reduced left ventricular ejection at the time of clinic referral and had an ECG within 6 months of their initial clinic visit. Patients with a paced rhythm (n = 11) were excluded.

A second cohort was gathered from the UT Southwestern cardiac transplant program ("pretransplant cohort") as a basis for comparison. We reasoned that if low voltage was an adverse prognostic marker, it would be more common in the pretransplant population as compared with the clinic cohort. Starting with the most recently transplanted patient in our program, we selected 100 consecutive patients in reverse chronological order who had a nonpaced rhythm on an ECG within 12 months before transplant. The ECG chosen for analysis was the most recent one before the date of transplantation. A total of 19 patients in the advanced HF cohort underwent left ventricular assist device implantation before transplantation. In these patients, the ECG chosen for analysis was the most recent one before left ventricular assist device implantation.

To further characterize the relationship between ECG voltage and ventricular size and weight, a third cohort of patients (n = 50) was collected retrospectively from patients of the UT Southwestern HF service who were referred for cardiac magnetic resonance imaging (MRI) between June 2003 and June 2005 and were found to have a cardiac MRI LVEF of <50%.

The study protocol was approved by the Institutional Review Board of University of Texas Southwestern Medical Center. Informed consent was waived owing to the retrospective nature of the study.

Analysis of ECGs

Electrocardiographic information was abstracted manually from each ECG. Heart rate, axis, and intervals were determined by the ECG computer and were recorded as displayed on the ECG tracing. All voltages were measured manually and approximated to the next largest 0.1-mV (1 mm) increment. Total voltages in each lead were calculated by summing the maximum deflection of the QRS complex above and below the baseline. Total limb voltage was obtained by summing the total voltages in each of the 6 limb leads. Total precordial voltage was obtained by summing the total voltages in each of the 6 precordial leads. Total ECG voltage was defined as the sum of the total limb voltage and the total precordial voltage. Low voltage in this study was defined as the lowest quartile of the clinic cohort (<12 mV). We also evaluated traditional low voltage criteria defined as either voltage in each precordial lead <1 mV or voltage in each limb lead <0.5 mV.^[9] Left ventricular hypertrophy was characterized using the Sokoloff-Lyon and Cornell criteria.^[10]

To assess reliability of ECG voltage measurements and dichotomous classification as low voltage or not, 100 ECGs were interpreted by two independent readers. Fifty ECGs were selected from the clinic cohort and 50 from the pretransplant cohort. In each group, one half had low voltage (range 7-10.5 mV) and one half did not have low voltage (range 14-20 mV). The intraclass correlation coefficient for total voltage (0.98) and the ? statistic (0.98) for low voltage dichotomous classification were excellent.

Clinical Characteristics and Outcomes

In each cohort, clinical information was obtained by detailed chart review and computerized database searches. Echocardiographic studies were included if performed within 6 months of the ECG. In the clinic cohort, we determined the association between ECG voltage and all-cause mortality as well as the composite end point of death and hospitalization for HF at 1 year. Hospitalizations for HF were determined from the Parkland Hospital database.^[7] Clinical outcomes could not be assessed in the pretransplant cohort as individuals had been selected after they underwent cardiac transplant. Information regarding the explanted heart (pretransplant cohort) was obtained from the clinical pathology report including right and left ventricular wall thickness, the latter reported as either single or multiple values (anterior, posterior, lateral walls). We averaged all available left ventricular wall thickness measurements.

Statistical Analysis

Continuous variables were expressed as mean ? SD and compared by t tests assuming unequal variance where appropriate or by Wilcoxon rank sum test. Categorical variables were compared by Fisher exact test. Multivariable logistic regression models were constructed to determine the independent association of low voltage with 1 year outcomes. SAS (SAS Institute, Cary, NC) was used for all analyses. A P value of <.05 was considered significant.

Results

Distribution and Correlation of ECG Voltages

Total ECG voltages in the clinic cohort ranged from 4.6 to 45 mV with a median (25th, 75th percentiles) of 15 (12, 20) mV. In the pretransplant cohort, ECG voltages ranged from 3.4 to 27 mV with a median (25th, 75th percentiles) of 12 (9.5, 16) mV. In the clinic cohort, there were significant correlations between total ECG voltage and total limb voltage (r = 0.76), total ECG voltage and total precordial voltage (r = 0.93), and total limb and total precordial voltages (r = 0.48) (P < .001 for each). Nearly identical correlations were seen in the pretransplant cohort. Body mass index (BMI) was inversely associated with total ECG voltage (r = −0.19, P = .001).

Association of Clinical Characteristics and ECG Voltages in the Clinic Cohort

In the clinic cohort, low voltage was associated with a greater prevalence of ischemic cardiomyopathy and q waves on the ECG as well as lower hemoglobin and left ventricular end-diastolic dimension ( Table I ). Body mass index did not differ significantly between the two groups. In 100 patients with low voltage, only 20 had a pericardial effusion: 18 small, 1 moderate, and 1 large. The total voltage in those with versus without these pericardial effusions was the same (9.5 ? 2 vs 9.2 ? 1.9 mV, P = .6).

Association Between Low Voltage and Severity of HF in the Clinic Cohort

In the clinic cohort, total voltage was lower in New York Heart Association (NYHA) class 4 patients (14 ? 5 mV) as compared with patients with NYHA class 1-3 (16 ? 6 mV, P = .006) and 34% of NYHA class 4 patients met the criteria for low voltage as compared with 22% of the remainder of the patients (P = .04). In contrast, there was no association of ECG LVH with NYHA class 4 status (Sokoloff-Lyon criteria: 16% in NYHA class 4 vs 25% NYHA class 1-3, P = .11; P = .7 for Cornell voltage LVH).

Prevalence of Low Voltage in the Pretransplant Cohort

As compared with the clinic cohort, patients in the pretransplant cohort were more often white, male, younger, had an ischemic cardiomyopathy and lower BMI ( Table II ). Of 76 pretransplant patients with an echocardiogram within 6 months of the ECG, all had ejection fraction <50%. Total voltage, precordial voltage, and limb voltage were lower ( Table II ) and the frequency of low voltage was nearly twice as high in the pretransplant HF cohort as compared with the clinic cohort (Figure 1). Low voltage remained more common in the pretransplant cohort in analyses stratified by etiology of cardiomyopathy (ischemic or nonischemic, Figure 1), sex, and race ( Table III ), although the comparison did not reach statistical significance in the African American patients possibly due to the small size of the African American pretransplant group.

Figure 1. 

Frequency of low ECG voltage in patients referred to a HF clinic or those with advanced HF before cardiac transplantation, overall, and when stratified by ischemic or nonischemic etiology of cardiomyopathy. Low voltage is defined as sum of QRS voltage in all 12 leads <12 mV.

     

Association of Low Voltage With Pathological Findings in the Pretransplant Cohort

In comparing those with versus without low ECG voltage in the pretransplant cohort, there was no difference in overall cardiac mass (507 ? 121 vs 540 ? 127 g, P = .3) or right ventricular wall thickness (1 ? 0.4 vs 1 ? 0.3 cm, P = .6, respectively) as assessed by pathological measurement (n = 76). However, low voltage was associated with reduced left ventricular wall thickness (n = 71) (1.5 ? 0.5 vs 1.8 ? 0.4 cm, P = .01).

Association of Low Voltage With Cardiac Dimensions as Assessed by Cardiac MRI

Among 50 patients referred for a cardiac MRI, 15 had low voltage and 35 did not have low voltage. A comparison of these two groups is shown ( Table IV ). In particular, the left ventricular mass indexed to body surface area was strikingly lower in those with ECG low voltage (see Figure 2). There were 10 small pericardial effusions in these 50 patients. There was no association of these small pericardial effusion with low ECG voltage (P = .7).

Figure 2. 

Comparison of left ventricular mass indexed to body surface area in those with or without low ECG voltage. Shown are the mean (25th, 75th) percentiles for both groups. Low voltage is total ECG voltage <12 mV.

     

Association of Low Voltage With Outcomes in the Clinic Cohort

Associations of ECG parameters and clinical outcomes are shown ( Table V ). Total limb voltage was lower in those who died, and total ECG, precordial, and limb voltages were lower in patients who met the composite end point. When considered as a dichotomous trait, low voltage was associated with death and death or HF hospitalization at 1 year, but LVH was not (Figure 3). The majority of the excess risk for adverse outcomes appeared to be confined to the lowest quartile of total ECG voltage (1 year mortality in quartiles 1 through 4: 18%, 5%, 9%, 7%, respectively). Increased QRS duration and atrial fibrillation were also associated with adverse outcomes.

Figure 3. 

Association of low voltage (total ECG voltage <12 mV), LVH, or neither with clinical outcomes at 1 year in patients referred to a HF clinic. LVH represents aggregate of subjects meeting either Sokoloff-Lyon or Cornell LVH criteria.

     

In multivariable analysis adjusting for baseline systolic blood pressure, NYHA functional class, serum creatinine, hemoglobin, and therapy with ß-blockers in the clinic, low voltage as defined remained associated with an increased risk of death (adjusted odds ratio 3.2, 95% CI 1.3-7.7, P < .05) and the composite end point of death or HF hospitalization (adjusted odds ratio 2, 95% CI 1.02-4, P < .05) after 1 year. These associations persisted in additional models when diabetes, ischemic etiology, or BMI were included as covariates.

Comparison With Traditional Low Voltage Criteria

In comparing the criteria proposed herein (∑QRS <12) to the traditional criteria for low voltage, 33 clinic patients were classified as having low voltage by both criteria and 309 patients were classified as not having low voltage by both criteria. There were 73 discordant classifications: 5 subjects were classified by traditional criteria but not by our criteria whereas 68 subjects were classified as having low voltage by our criteria but not by traditional criteria.

Patients classified as having low voltage by ∑QRS <12 but not by standard low voltage criteria had an increased risk of death (18% vs 7%, P = .008) or the composite end point of death or hospitalization (34% vs 22, P = .04) as compared with those without low voltage. There was no significant difference in the frequency of these outcomes when comparing individuals with ∑QRS <12 to those who also met the classic criteria for low voltage (P ≥ .3 for both outcomes).

Other Combinations of ECG Leads

The sum of voltages in two select leads (eg, V₁ and V₅ or leads II and V₅) did correlate with 12-lead voltage (correlation coefficients 0.89 and 0.84, respectively, P < .001 for both). Low voltage as defined by these combinations of two leads (lowest quartile of clinic cohort: <2 mV in V₁ and V₅ and <1.6 mV in II and V₅) had moderate agreement for low voltage as defined by all 12 leads (? = 0.65 [95% CI 0.56-0.74] and ? = 0.62 [0.53-0.71], respectively) although low voltage as defined by these alternative criteria was not independently associated with outcome. A value below 3.7 mV (lowest quartile of clinic cohort) for the sum of voltages in the 6 limb leads was associated with an increased risk of death (odds ratio 3.3, P = .01) and a trend for the composite end point of death or HF hospitalization (odds ratio 1.7, P = .13) in multivariable regression using the same covariates described above.

Discussion

We propose new criteria for defining low ECG voltage (∑QRS <12 mV) in patients with HF due to systolic dysfunction. Using these criteria in a contemporary cohort of HF patients, low voltage was a marker of disease severity, being more common in NYHA class 4 than class 1-3 patients. The prevalence of low voltage was further increased in patients with advanced HF as defined by referral for cardiac transplantation. Further, low ECG voltage was associated with an increased risk of adverse outcomes including overall mortality at 1 year. In contrast, ECG LVH was not associated with HF symptom severity or adverse outcomes.

In an early, small study reported in 1971 before modern therapy for HF, an increase in QRS voltage was observed in 12 subjects with HF who had clinical improvement, whereas a fall in voltage occurred in 3 subjects with clinical deterioration.^[5] Shortly thereafter, low ECG voltage as defined by traditional criteria^[9] was found to be associated with a poor prognosis in the setting of acute myocardial infarction.^[11,12] Low limb voltage associated with an increase in the ratio of transverse to frontal plane QRS voltage has been reported in patients with decompensated HF.^[13] To our knowledge, there has been little further investigation of the prognostic utility of low ECG voltage in HF.

The criterion used for low voltage in this study (<12 mV), chosen because it represented the lowest quartile of our clinic cohort, is less stringent than the standard definition of low voltage^[9] with 68 (16%) of 415 subjects being classified as having low voltage by our criteria but not by the classic criteria. We believe our criteria to be appropriate for the following reasons. First, patients with HF and systolic dysfunction are likely to have different distribution of ECG voltages as compared with individuals without cardiac disease. Thus, the threshold value needed to define abnormally low ECG voltage in the two populations is likely to be different. Second, in our study, subjects who had ∑QRS <12 but not the classic criteria for low voltage were still at increased risk of adverse outcomes. Third, individuals who had ∑QRS <12 but did not meet the classic criteria for low voltage had comparable outcomes with subjects who also met the classic criteria for low voltage.

In contrast to the finding that low voltage is associated with an adverse prognosis, increased QRS voltage as manifest by ECG LVH (Sokoloff-Lyon voltage criteria)^[4] and increased echocardiographic left ventricular mass^[14] have been shown to be adverse prognostic factors. However, in the Multicenter Unsustained Tachycardia Trial, ECG LVH was associated with a 40% increased risk of arrhythmic death but not total mortality after ∼3.3 years.^[15] In the present study, we found no association between ECG LVH and mortality or the composite end point of death or HF hospitalization. Given that low ECG voltage was more common in NYHA class 4 patients from the clinic cohort and in patients from the pretransplant cohort, we hypothesize that ECG voltage may decrease as end-stage HF approaches, thereby obscuring the prognostic value of ECG LVH that has been well demonstrated in other clinical settings.^[3,16]

The basis for lower ECG voltage in the setting of advanced HF is uncertain. Obesity and chronic obstructive pulmonary disease are unlikely explanations given that patients with severe obesity and pulmonary disease are routinely excluded as transplant candidates yet the frequency of low voltage was higher in that cohort ( Table II ). Furthermore, increased BMI (which is associated with lower ECG voltage) has been found to be associated with improved outcomes in HF.^[17] Pericardial effusions are an unlikely explanation of the low voltage because they were relatively infrequent (20% of patients with low voltage), most were small, and were not associated with ECG voltage. Infiltrative processes such as amyloid are also unlikely because these processes were not seen on the explanted heart at the time of transplantation. One previously proposed theory is that biventricular hypertrophy decreases surface ECG voltage as left and right ventricular electromotive forces cancel each other out.^[5] However, we were unable to demonstrate an increase in right ventricular wall thickness in subjects with low voltage upon pathological examination of the heart. Another hypothesis is that large or multiple myocardial infarctions may result in lower voltage due to electrical silence of the infarcted areas.^[11] In our study, we did find a strong association between ischemic etiology of cardiomyopathy and lower voltage. However, we also found that low voltage was an adverse prognostic factor independent of ischemic etiology in multivariable analysis. Increased intracardiac blood volume in the setting of elevated left-sided filling pressures^[18] or peripheral edema^[6] may be associated with both advanced HF and decreased ECG voltage. Finally, a lower hemoglobin has been postulated to lead to a lower ECG voltage due to influences on blood conductivity.^[6] In our study, although there was an association of lower hemoglobin with low ECG voltage ( Table I ), the prognostic value of low voltage persisted in multivariable analyses that adjusted for hemoglobin.

Our data suggest that a reduced ratio of left ventricular mass to body size may be a more proximal cause of the low ECG voltage than the mechanisms proposed above. Patients with low voltage in the clinic cohort had smaller left ventricular end-diastolic dimensions on echocardiography. Some parameters of wall thickness (pathological reports in pretransplant group and posterior wall thickness by cardiac MRI) were also smaller in those with versus without low voltage. Finally, there was a strikingly low left ventricular mass indexed to body surface area in those with low voltage (Figure 2). To our knowledge, this would be the first cardiac trait that identifies patients with systolic HF who are at an increased risk of adverse outcomes when they have a smaller rather than larger ratio of left ventricular mass to body size.

Limitations

This study has the limitation of being a retrospective analysis. Our primary study population was treated in an urban public hospital and has unique sociodemographics. Patients did not undergo routine coronary angiography, which may lead to misclassification of etiology. B-type natriuretic peptide was not routinely measured and thus could not be incorporated into multivariable analysis. Echocardiograms, cardiac MRI, and pathological assessment of the explanted heart were obtained for clinical care without standardization. Only short-term clinical outcome data were available. Hospitalization data were available only from Parkland Memorial Hospital, although most of its patients use this hospital almost exclusively (for financial reasons). Furthermore, because there would be no expectation of differential hospitalization usage by ECG voltage status, any missing hospitalization data would bias our results toward the null. The increased prevalence of low voltage in the transplant cohort may represent survivor bias, although the increased risk of mortality in the clinic cohort argues against this possibility. Likewise, the higher prevalence of low voltage in the pretransplant versus clinic cohorts may be secondary to differences in ethnic composition, although the prevalence of low voltage among non?African American pretransplant patients remained increased in analysis stratified by race ( Table III ).

Conclusion

Low ECG voltage is a marker of severity of HF in patients with systolic dysfunction in the modern era, being more common in those with NYHA class 4 symptoms and those undergoing cardiac transplantation. Furthermore, low ECG voltage is a risk factor for mortality and the composite end point of death or hospitalization at 1 year in patients followed up in a HF clinic.