Despite the low sensitivity of the electrocardiogram (ECG) in detecting left ventricular hypertrophy (LVH), ECG-LVH is known to be a strong predictor of cardiovascular risk. Understanding reasons for the discrepancies in detection of LVH by ECG versus imaging could help improve the diagnostic ability of ECG. We examined factors associated with false-positive and false-negative ECG-LVH, using cardiac magnetic resonance imaging (MRI) as the gold standard. We also compared the prognostic significance of ECG-LVH and MRI-LVH as predictors of cardiovascular events. This analysis included 4,748 participants (mean age 61.9 years, 53.5% females, 61.7% nonwhites). Logistic regression with stepwise selection was used to identify factors associated with false-positive (n = 208) and false-negative (n = 387), compared with true-positive (n = 208) and true-negative (n = 4,041) ECG-LVH, respectively. A false-negative ECG-LVH status was associated with increased odds of Hispanic race/ethnicity, current smoking, hypertension, increased systolic blood pressure, prolongation of QRS duration, and higher body mass index and with lower odds of increased ejection fraction (model-generalized R 2 = 0.20). A false-positive ECG-LVH status was associated with lower odds of black race, Hispanic race/ethnicity, minor ST-T abnormalities, increased systolic blood pressure, and presence of any major electrocardiographic abnormalities (model-generalized R 2 = 0.29). Both ECG-LVH and MRI-LVH were associated with an increased risk of cardiovascular disease events (hazard ratio 1.51, 95% confidence interval 1.03 to 2.20 and hazard ratio 1.81, 95% confidence interval 1.33 to 2.46, respectively). In conclusion, discrepancy in LVH detection by ECG and MRI can be relatively improved by considering certain participant characteristics. Discrepancy in diagnostic performance, yet agreement on predictive ability, suggests that LVH by ECG and LVH by imaging are likely to be two distinct but somehow related phenotypes.
The current diagnosis of left ventricular hypertrophy (LVH) by electrocardiogram (ECG) is based on finding electrocardiographic criteria that agree with increased left ventricular mass (LVM) as detected by imaging. However, it has been consistently reported that the magnitude of agreement is rather low. As a result, a significant proportion of cases with true anatomic LVH are misclassified using ECG-LVH criteria. Despite this limitation, it has been repeatedly reported that ECG-LVH provides independent information on the cardiovascular risk even after adjusting for LVM by imaging. Understanding possible reasons for the frequent discrepancy between common ECG-LVH criteria and increased LVM by imaging would help understanding the genesis of electrocardiographic changes that occur as a consequence of increased LVM. This information might possibly help in refining the current ECG-LVH criteria for the purpose of improved predictive ability and for detection of increased LVM. The primary aim of this study was to identify factors associated with false-positive and false-negative ECG-LVH, using cardiac magnetic resonance imaging (MRI) as the gold standard, in the Multi-Ethnic Study of Atherosclerosis (MESA). As secondary aim, we sought to examine the prognostic significance of false-positive and false-negative ECG-LVH as predictors of fatal and nonfatal cardiovascular events.
Methods
MESA is a prospective longitudinal study aimed to explore the prevalence, correlates, and progression of subclinical cardiovascular disease (CVD) in a population-based multiethnic cohort. The description of the MESA study is provided elsewhere. Briefly, from July 2000 to August 2002, a total of 6,814 men and women aged 45 to 84 years and free of clinically apparent CVD were recruited from 6 US communities: Baltimore City and Baltimore County, Maryland; Chicago, Illinois; Forsyth County, North Carolina; Los Angeles County, California; Northern Manhattan and the Bronx, New York; and St. Paul, Minnesota. For the purpose of these analyses, all MESA participants with good quality baseline electrocardiogram and cardiac MRI data were considered. Of those, we excluded participants with major ventricular conduction defect including those with complete bundle branch blocks or QRS duration ≥120 ms. After all exclusions, 4,748 participants remained and were included in the analysis.
The MESA cardiac MRI protocol, image analysis, and inter- and intra-reader reproducibility have been previously reported. Briefly, base to apex short-axis fast gradient echo images (slice thickness 6 mm, slice gap 4 mm, field of view 360 to 400 mm, matrix 256 × 160, flip angle 20°, echo time 3 to 5 ms, repetition time 8 to 10 ms) were acquired using 1.5-T cardiac MRI scanners. The reproducibility of this protocol was assessed on 79 participants with a technical measurement error of 6% and an intraclass correlation coefficient of 0.98.
LVM was measured as the sum of the myocardial area (the difference between endocardial and epicardial contours) times slice thickness plus image gap in the end-diastolic phase multiplied by the specific gravity of the myocardium (1.05 g/ml). Observed LVM was then determined from MRI in all MESA participants. Individual LVM was predicted using the following allometric height and weight indexation equations previously derived from a separate reference MESA subpopulation of 822 men and women (47% Caucasians, 22% Chinese, 18% African-American, 13% Hispanics) without LVH risk factors: predicted LVM (pLVM) = 8.17 × height (in meters) 0.561 × weight (in kilograms) 0.608 for men and pLVM = 6.82 × height (in meters) 0.561 × weight (in kilograms) 0.608 for women.
The ninety-fifth percentile cut-off value of (observed LVM/pLVM) was calculated as 1.31. This cut-off point defines participants with observed LVM >1.31 times of that predicted on the basis of height, weight, and gender had LVM greater than 95% of the reference population as constituting MRI-LVH for the purposes of this study.
Standard 12-lead electrocardiograms were digitally acquired using a Marquette MAC 1200 electrocardiograph (Marquette Electronics, Milwaukee, Wisconsin) at 10 mm/mV calibration and a speed of 25 mm/s. All electrocardiograms were centrally read and coded at the Epidemiological Cardiology Research Center (EPICARE), Wake Forest School of Medicine, Winston-Salem, North Carolina. Electrocardiograms were visually inspected for technical errors and inadequate quality before being automatically processed with the GE Marquette 12-SL program 2001 version (GE Marquette, Milwaukee, Wisconsin). Electrocardiographic abnormalities were classified as minor and major abnormalities using the Minnesota ECG Classification.
ECG-LVH was defined based on the following traditional ECG-LVH criteria which were calculated from the automatically measured electrocardiographic waveforms: Sokolow-Lyon voltage (SV 1 + RV 5 /V 6 ≥ 3.5 mV and/or RaVL ≥ 1.1 mV) and/or gender-specific Cornell voltage (SV 3 + RaVL > 2.8 mV for men and >2.0 mV for women). ECG-LVH was defined in this analysis as presence of positive electrocardiographic criteria by either the Cornell voltage or Sokolow-Lyon criterion.
CVD events were adjudicated by an independent adjudication committee. A detailed description of the adjudication process has already been published. In this analysis, we used a composite outcome of fatal and nonfatal CVD events. These events included myocardial infarction, resuscitated cardiac arrest, definite angina, probable angina (if followed by revascularization), stroke, transient ischemic attack, percutaneous transluminal coronary angioplasty, coronary stent, coronary atherectomy, coronary bypass graft, coronary or other revascularization, congestive heart failure, peripheral vascular disease, death from coronary heart disease, death from stroke, death from other atherosclerotic conditions or other CVD death.
Three seated blood pressure measurements were taken 5 minutes apart using an automated device (Dinamap Pro 100; Critikon, Milwaukee, WI). The mean of the last 2 measurements was considered for analysis. Hypertension was defined as systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or history of intake of blood pressure–lowering drugs. Trained technicians measured height, weight, and waist circumference following a standardized protocol. Diabetes was defined as current use of glucose-lowering medications, fasting glucose ≥126 mg/dl, or nonfasting glucose ≥200 mg/dl. Use of medication, current smoking, ethanol intake, income, and education were ascertained from standardized questionnaires.
Based on the baseline LVH status by MRI (gold standard) and ECG, participants were classified as having false-positive, false-negative, true-positive, or true-negative ECG-LVH. Baseline characteristics of the study participants were then examined and compared across these categories of ECG-LVH using analysis of variance for the continuous variables and chi-square tests for the categorical variables.
Binary logistic regression models were used to identify baseline characteristics that are associated with false-negative and false-positive, compared with true-negative and true-positive ECG-LVH, respectively. These include age, gender, race/ethnicity, education, income, hypertension, systolic blood pressure, diastolic blood pressure, use of blood pressure–lowering drugs, total cholesterol, HDL cholesterol, family history of CVD, statin use, smoking status, left ventricular ejection fraction, LVM by MRI, QRS duration, QRS axis, ST-T abnormalities, and minor/major electrocardiographic abnormalities. We first conducted bivariate analyses. Then, separate final models for predicting false negative and false positive were selected using a stepwise selection procedure.
CVD event rates and 95% confidence intervals (CIs) were calculated by ECG-LVH status. Cumulative incidence was estimated using Kaplan-Meir method and compared using log-rank test across ECG-LVH status. Cox proportional hazard (CPH) models were used to examine the association between baseline ECG-LVH status (false negative, false positive, true positive, true negative [reference group]) with incident CVD events. Three CPH models were created: model 1, unadjusted; model 2, adjusted for sociodemographic characteristics [age, gender, race/ethnicity, and socioeconomic status]; and model 3, adjusted for model 1 plus clinical and anthropometric variables [body mass index, systolic blood pressure, blood pressure–lowering drugs, diabetes, total cholesterol, lipid-lowering drugs, and smoking status]. Similar CPH models were used to examine the association of ECG-LVH (present vs absent regardless of true or false) and MRI-LVH with incident CVD events. Statistical analyses were performed using SAS statistical software (version 9.3; Cary, North Carolina).
Results
These analyses included 4,748 participants (age 61.9 ± 10.1 years, 53.5% women, 38.3% whites, 13.4% American Chinese, 25.8% African-American, 22.5% Hispanic). MRI- and ECG-LVH were present in 10.5% (n = 499) and 6.7% (n = 320) of the participants, respectively. About 2.4% (n = 112) of the participants had LVH by both MRI and ECG (i.e., true-positive ECG-LVH), and 85.1% (n = 4041) did not have LVH by either method (i.e., true-negative ECG-LVH). The remaining 12.5% (n = 595) of the participants had either MRI-LVH but no ECG-LVH (n = 387) (i.e., false-negative ECG-LVH) or the opposite (n = 208) (i.e., false-positive ECG-LVH).
Table 1 outlines the characteristics of the study population by ECG-LVH status (true negative, false negative, false positive, true positive) using MRI-LVH as the gold standard.
| Mean ±SD or n (% ) Variable | True negative (n=4041) | False negative (n=387) | False positive (n=208) | True positive (n=112) | p value |
|---|---|---|---|---|---|
| Age (years) | 61.6± 10.0 | 62.6± 10.3 | 63.7± 9.9 | 65.7± 9.6 | <0.001 |
| Women | 2118(52.4) | 235( 60.7) | 121(58.2) | 66(58.9) | 0.004 |
| White | 1643(40.7) | 125(32.3) | 39( 18.8) | 10(8.9) | <0.001 |
| Americans Chinese | 549(13.6) | 18(4.7) | 57(27.4) | 12(10.7) | |
| African American | 953(23.6) | 130( 33.6) | 80(38.5) | 61(54.5) | |
| Hispanic | 896(22.2) | 114(29.5) | 32(15.4) | 29(25.9) | |
| Income | <0.001 | ||||
| <$20K | 866(22.1) | 88(23.7) | 61(31.1) | 38(36.2) | |
| $20-49K | 1399(35.6) | 165(44.4) | 72( 36.7) | 46(43.8) | |
| >$50K | 1662(42.3) | 119(32.0) | 63(32.1) | 21(20) | |
| Education | 0.002 | ||||
| <HS | 646(16.0) | 72(18.7) | 46(22.12) | 32( 28.83) | |
| HS-College | 2595(64.4) | 247(64.16) | 119( 57.21) | 59(53.15) | |
| >College | 791(19.6) | 66( 17.14) | 43( 20.67) | 20(18.02) | |
| Body mass index (kg/m 2 ) | 27.6± 4.89 | 29.5± 5.59 | 26.68± 4.66 | 28.22± 4.5 | <0.001 |
| Hypertension | 1511 (37.4) | 261 (67.4) | 128(61.5) | 93 (83.0) | <0.001 |
| Systolic blood pressure (mmHg) | 122.93±19.62 | 138.07± 23.3 | 133.31, 23.98 | 151.49, 26.92 | <0.001 |
| Diastolic blood pressure (mmHg) | 71.2± 9.9 | 75.42± 11.43 | 73.23± 10.91 | 79.52± 12.67 | <0.001 |
| Use of BP lowering drugs | 1277(31.6) | 195(50.5) | 111(53.4) | 68(60.7) | <0.001 |
| Total Cholesterol (mg/dL) | 194.42± 35.12 | 194.04± 36.19 | 195.94±36.54 | 195.12± 35.36 | 0.93 |
| HDL-Cholesterol (mg/dL) | 51.29± 15.05 | 51.53±15.27 | 51.84±12.46 | 51.99±15.65 | 0.91 |
| Statin use | 584 (14.5) | 55(14.3) | 28(13.5) | 16(14.3) | 0.98 |
| Diabetes Mellitus | 418(10.4) | 68(17.6) | 24( 11.6) | 27( 24.1) | <0.001 |
| Family history of CVD | 1605(42.2) | 166( 46.1) | 78( 39.2) | 42(41.6) | 0.40 |
| Smoking status | <0.001 | ||||
| Never | 2095(52.0) | 171(44.4) | 124(59.6) | 60(54.1) | |
| Former | 1458(36.2) | 133(34.6) | 64(30.8) | 29(26.1) | |
| Current | 479(11.9) | 81( 21.0) | 20(9.6) | 22(19.8) | |
| LV ejection fraction (%) | 69.43± 6.9 | 66.94± 8.6 | 70.0± 7.4 | 66.7± 10.8 | <0.001 |
| LV mass (gm) | 74.2± 12.7 | 101.8, 14.7 | 78.18, 12.2 | 108.1, 18.2 | <0.001 |
| QRS duration (ms) | 90.7±9.5 | 94.3± 9.7 | 95.1± 9.5 | 95.2±10.1 | <0.001 |
| Abnormal QRS axis | 182( 4.5) | 20( 5.17) | 15( 7.2) | 8(7.1) | 0.18 |
| Any ST/T Abnormalities | 445( 11.0) | 97(25.1) | 56(26.9) | 58(51.8) | <0.001 |
| Major ST/T Abnormalities | 116( 2.9) | 19(4.9) | 20( 9.6) | 32(28.6) | <0.001 |
| Minor ST/T Abnormalities | 392( 9.7) | 94(24.3) | 48( 23.1) | 50(44.6) | <0.001 |
| Any major ECG abnormality | 258( 6.4) | 42(10.9) | 30( 14.4) | 40(35.7) | <0.001 |
In bivariate analyses, several factors were associated with false-positive and false-negative ECG-LVH compared with true ECG-LVH (negative and positive) status ( Table 2 ). Using stepwise selection, Hispanic race/ethnicity, current smoking, hypertension, increased systolic blood pressure, prolongation of QRS duration, higher body mass index, and presence of hypertension were associated with increased odds, whereas increased ejection fraction was associated with lower odds of false-negative ECG-LVH, compared with true-negative ECG-LVH status (model-generalized R 2 = 0.198). In contrast, black race, Hispanic race/ethnicity, minor electrocardiographic ST-T abnormalities, increased systolic blood pressure, and presence of any major electrocardiographic abnormalities were associated with lower odds of false-positive ECG-LVH, compared with true-positive ECG-LVH status (model-generalized R 2 = 0.286; Table 3 ).
| Variable | False negative | False positive | ||
|---|---|---|---|---|
| Odds ratio (95% CI) | p-value | Odds ratio (95% CI) | p-value | |
| Age (year) | 1.01 (1.00, 1.02) | 0.054 | 0.98 (0.96, 1.00) | 0.080 |
| Sex (men vs. women) | 0.71 (0.58, 0.88) | 0.002 | 1.03(0.65, 1.64) | 0.896 |
| Race/Ethnicity | <0.001 | 0.001 | ||
| White | (Reference) | (Reference) | ||
| Chinese-Americans | 0.43 (0.26, 0.71) | 0.001 | 1.22 (0.48, 3.10) | 0.679 |
| Hispanics | 1.67 (1.28, 2.18) | <0.001 | 0.28 (0.12, 0.67) | 0.004 |
| Blacks | 1.79 (1.39, 2.32) | <0.001 | 0.34 (0.16, 0.73) | 0.006 |
| Body mass index (kg/m 2 ) | 1.07 (1.05, 1.09) | <0.001 | 0.93 (0.89, 0.98) | 0.006 |
| Income | <0.001 | 0.085 | ||
| <$20K) | (Reference) | (Reference) | ||
| >$20k<$50K | 1.16 (0.88, 1.52) | 0.284 | 0.98 (0.56, 1.69) | 0.928 |
| >$50K | 0.70 (0.53, 0.94) | 0.017 | 1.87(0.99, 3.54) | 0.055 |
| Education | 0.265 | 0.407 | ||
| High School | (Reference) | (Reference) | ||
| College | 0.85 (0.65, 1.13) | 0.263 | 1.40 (0.81, 2.43) | 0.226 |
| >College | 0.75 (0.53, 1.06) | 0.105 | 1.50 (0.75, 3.00) | 0.257 |
| Hypertension | 3.47 (2.78, 4.33) | <0.001 | 0.33 (0.19, 0.58) | 0.001 |
| Systolic blood pressure (mmHg) | 1.03 (1.03, 1.04) | <0.001 | 0.97 (0.96, 0.98) | <.001 |
| Diastolic blood pressure ( mmHg) | 1.04 (1.03, 1.05) | <0.001 | 0.95 (0.93, 0.97) | <.001 |
| Blood pressure lowering drugs | 2.21 (1.79, 2.72) | <0.001 | 0.74 (0.46, 1.18) | 0.207 |
| Total cholesterol ( mg/dL) | 1.00 (1.00, 1.00) | 0.841 | 1.00 (0.99, 1.01) | 0.846 |
| HDL-cholesterol (mg/dL) | 1.00 (0.99, 1.01) | 0.759 | 1.00 (0.98, 1.02) | 0.925 |
| Statin use | 0.98 (0.73, 1.33) | 0.911 | 0.93 (0.48, 1.81) | 0.838 |
| Diabetes Mellitus | 1.85 (1.40, 2.45) | <0.001 | 0.41 (0.23, 0.76) | 0.004 |
| Family history of CVD | 1.17 (0.94, 1.46) | 0.147 | 0.91 (0.56, 1.47) | 0.690 |
| Smoking status | <0.001 | 0.041 | ||
| Never | (Reference) | (Reference) | ||
| Current | 2.07 (1.56, 2.75) | <0.001 | 0.44 (0.22, 0.87) | 0.018 |
| Past | 1.12 (0.88, 1.42) | 0.356 | 1.07 (0.62, 1.83) | 0.810 |
| LV ejection fraction (%) | 0.96 (0.94, 0.97) | <0.001 | 1.04 (1.01, 1.07) | 0.002 |
| MRI- LV mass (gm) | 1.16 (1.15, 1.18) | <0.001 | 0.86 (0.83, 0.89) | <.001 |
| QRS duration (ms) | 1.04 (1.03, 1.05) | <0.001 | 1.00 (0.98, 1.02) | 0.911 |
| Abnormal QRS axis | 1.16 (0.72, 1.86) | 0.550 | 1.01 (0.41, 2.46) | 0.982 |
| Any ST/T abnormalities | 2.70 (2.11, 3.47) | <0.001 | 0.34 (0.21, 0.55 | <.001 |
| Minor ST/T abnormalities | 2.99 (2.32, 3.85) | <0.001 | 0.37 (0.23, 0.61) | <.001 |
| Major ST/T abnormalities | 1.75 (1.06, 2.87) | 0.028 | 0.27 (0.14, 0.49) | <.001 |
| Any major ECG abnormalities | 1.79 (1.27, 2.52) | 0.001 | 0.30 (0.18, 0.52) | <.001 |
| Variable | False negative ECG-LVH | False positive ECG-LVH | ||||
|---|---|---|---|---|---|---|
| Odds Ratio (95% CI) | p-value | Model R 2 | Odds Ratio (95% CI) | p-value | Model R 2 | |
| Age (year) | 0.99 (0.98, 1.00) | 0.096 | 0.198 | 1.01 (0.98, 1.04) | 0.594 | 0.286 |
| Sex (men vs. women) | 0.39 (0.30, 0.51) | <.0001 | 0.87 (0.51, 1.50) | 0.6214 | ||
| Race/ethnicity | 0.003 | 0.0009 | ||||
| Whites | Reference | — | Reference | — | ||
| Chinese Americans | 0.72 (0.42, 1.23) | 0.230 | 1.18 (0.43, 3.27) | 0.7458 | ||
| Blacks | 1.32 (0.99, 1.75) | 0.055 | 0.36 (0.15, 0.84) | 0.0174 | ||
| Hispanics | 1.56 (1.17, 2.08) | 0.003 | 0.27 (0.10, 0.68) | 0.0056 | ||
| Body mass index (Kg/m 2 ) | 1.02 (1.00, 1.05) | 0.038 | —– | — | ||
| Smoking Status | <0.001 | —– | —- | |||
| Never | Reference | — | ||||
| Current | 2.03 (1.48, 2.79) | <0.001 | ||||
| Past | 1.10 (0.85, 1.43) | 0.469 | ||||
| Hypertension | 1.86 (1.38, 2.50) | <0.001 | —– | — | ||
| Minor ST/T abnormalities | 1.88 (1.41, 2.51) | <0.001 | 0.49 (0.27, 0.88) | 0.017 | ||
| Left ventricular ejection fraction (%) | 0.95 (0.93, 0.96) | <0.001 | ||||
| QRS duration (ms) | 1.05 (1.04, 1.06) | <0.001 | ||||
| Systolic blood pressure (mmHg) | 1.02 (1.02, 1.03) | <0.001 | 0.97 (0.96, 0.98) | <0.001 | ||
| Any major ECG abnormality | —– | — | 0.39 (0.20, 0.75) | 0.005 | ||
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
