Novel Approach to the Detection of Left Ventricular Hypertrophy Using Body Mass Index–Corrected Electrocardiographic Voltage Criteria in a Group of African Ancestry

Abstract

Background

Electrocardiographic (ECG) QRS voltages used to generate criteria for left ventricular hypertrophy (LVH) detection are considerably attenuated by obesity. This effect renders the utility of ECG criteria to detect LVH in obese individuals of African ancestry to be of limited value.

Hypothesis

A novel approach to correcting QRS voltages for the attenuating effect of body mass index (BMI) will improve the ability of ECG criteria to detect LVH in a group of African descent.

Methods

Left ventricular mass was determined from echocardiography in 661 randomly selected participants (43.0% obese) of black African ancestry in South Africa.

Results

As compared with Cornell and Sokolow‐Lyon voltage criteria, BMI best correlated with R_aVL, Gubner‐Ungerleider, and Lewis QRS complex voltages, but these relations were noted only in those with BMI <29 kg/m². Correcting R_aVL and Lewis voltages by the difference in the slope of BMI‐voltage relations in those with BMI <29 kg/m² vs those with BMI ≥29 kg/m² showed the greatest performance for LVH detection (uncorrected R_aVL: 0.695 ± 0.025, corrected R_aVL: 0.733 ± 0.022; P < 0.0001), and also increased the sensitivity (uncorrected R_aVL: 30.6%, corrected R_aVL: 42.4%; P < 0.0005) with no significant change in specificity (uncorrected R_aVL: 86.3%, corrected R_aVL: 83.0%; P = 0.28).

Conclusions

We offer a novel approach to correcting ECG voltages for the attenuating effects of obesity in individuals of African ancestry, and this improves the performance and sensitivity for LVH detection.

Introduction

As demonstrated in the general community and in several clinical populations, the identification of left ventricular hypertrophy (LVH) adds to cardiovascular risk prediction.1, 2, 3, 4, 5, 6, 7, 8 Although echocardiography is the reference method for LVH detection, electrocardiography (ECG) is the more widely available and cost‐effective approach. Hence, ECG identification of LVH is recommended by all hypertension (HTN) guidelines for routine risk prediction. However, the value of ECG criteria for the detection of LVH in the obese9, 10, 11, 12, 13, 14, 15 and in those of black African ancestry16, 17 is questionable. With the increasing prevalence of obesity in groups of black African ancestry and the strong relationship between obesity and the prevalence and incidence of HTN,18, 19 the combined moderating effect of obesity and African ancestry on ECG criteria for LVH detection may have an important impact on risk prediction in these groups. Indeed, obesity results in a particularly striking attenuation of QRS voltages and inability of ECG criteria to detect LVH in obese individuals of African ancestry.20 Identifying strategies to enhance the detection of LVH in groups of black African descent with a high prevalence of obesity is therefore of importance.

A possible strategy to improve LVH detection in obesity is to multiply QRS voltages by body mass index (BMI; the BMI‐QRS voltage product).21, 22 This approach improves the overall performance for LVH detection21, 22 and increases the ability of ECG criteria to predict cardiovascular outcomes22 in groups of European ancestry. However, the BMI‐QRS voltage product does not necessarily correct for the attenuating effect of obesity on QRS voltages, but rather employs BMI together with ECG voltages as a determinant of left ventricular mass (LVM). Moreover, whether this approach enhances the ability to detect LVH in obese individuals of black African ancestry is unknown. Hence, in the present study we assessed the impact of BMI‐QRS voltage product and a novel approach to correcting ECG voltages for the attenuating effects of obesity on the performance and sensitivity for LVH detection in a sample of African ancestry with a high prevalence of obesity. To correct for the attenuating effects of BMI on ECG voltages, we employed the following approach: We first identified the extent to which QRS voltages are diminished in those with a BMI greater than or equal to as compared with less than the BMI threshold at which ECG voltages no longer show positive relations with BMI. We subsequently modified the voltages in participants above this threshold by a mathematically identified correction factor.

Methods

The present study was conducted according to the principles outlined in the Declaration of Helsinki. The Committee for Research on Human Subjects of the University of the Witwatersrand approved the protocol (approval numbers: M02‐04‐72, renewed as M07‐04‐69 and M12‐04‐108). Participants gave informed, written consent. The study design has been described previously.23, 24, 25, 26 Briefly, nuclear families of black African descent with siblings age >16 years were randomly recruited from the South West Township of Johannesburg, South Africa. Of the 1191 participants recruited, in a substudy consisting of 678 participants, 661 had all echocardiographic and ECG data required for the present analysis.

Demographic and clinical data were obtained using a standardized questionnaire. Height and weight were measured using conventional approaches, and participants were identified as being overweight if their BMI was ≥25 kg/m² and obese if their BMI was ≥30 kg/m². Blood tests of renal function, liver function, blood glucose, lipid profiles, hematological parameters, and percentage of glycated hemoglobin (Roche Diagnostics, Mannheim, Germany) were performed. Diabetes mellitus or abnormal blood glucose control was defined as the use of insulin or oral hypoglycemic agents or a glycated hemoglobin value >6.1%. Blood pressure measurements were obtained by a trained nurse‐technician using a standard mercury sphygmomanometer and according to guidelines as previously described.24, 25

A standard 12‐lead ECG was recorded at 25 mm/s and 1.0 mV/cm.20, 23, 24 The R‐ and S‐wave amplitudes in all leads were measured to the nearest 0.05 mV (0.5 mm). QRS duration was measured to the nearest 4 ms. Cornell, Sokolow‐Lyon, Gubner‐Ungerleider, and Lewis voltage or time‐voltage criteria for LVH were calculated using standard formulas that incorporate a sex‐specific constant where required.20 As groups of black African descent show differences in R‐ and S‐wave amplitudes as compared with other ethnic groups,27, 28 thresholds of ECG criteria for LVH were determined from the upper 95% confidence intervals (CI) derived in 150 participants without clinically significant disease and normal clinical blood parameters who were normotensive, nondiabetic, and had a BMI <30 kg/m². Neither Framingham nor Romhilt‐Estes criteria were employed, as these rely on the presence of a significant number of individuals with a strain pattern, which was only noted in <0.5% of participants, and the ability to identify the presence of an intrinsicoid deflection duration in V₅ or V₆ ≥ 0.09 seconds, which we were unable to accurately assess due to a lack of appropriate software.

Two‐dimensional targeted M‐mode echocardiography was performed in the long‐axis parasternal view as previously described,25, 26 and variables were analyzed according to the American Society of Echocardiography convention.29 All measurements were recorded and analyzed offline by experienced investigators who were unaware of the clinical data of the participants. Left ventricular mass was determined using a standard formula30 and indexed (LVMI) to height^2.7. An LVMI >51 g/m^2.7 was considered to be increased,31 and this value was confirmed from the upper 95% CI derived in 150 participants without clinically significant disease and normal clinical blood parameters who were normotensive, nondiabetic, and had a BMI <30 kg/m². In these participants the upper 95% CI for LVMI was 51.8 g/m^2.7.

Statistical Analysis

Database management and statistical analyses were performed with SAS software, version 9.4 (SAS Institute Inc., Cary, NC). Continuous data are reported as mean ± SD or mean ± SEM. Unadjusted means and proportions were compared by the large‐sample Z test and the χ² statistic, respectively. The Z statistics were used to compare correlation coefficients. The performance of ECG criteria for LVH detection was determined from the area (AUC) under the receiver operator characteristic (ROC) curve. The sensitivity and specificity for LVH detection was identified from ECG voltage thresholds using standard approaches.

The BMI thresholds above which BMI‐ECG relations were noted to be reduced were determined using iterative (Loess regression) analysis, where the strongest relationship between BMI and voltage criteria identified when including only data below this threshold was noted. For all voltage criteria except Cornell and Sokolow‐Lyon criteria, where a threshold could not be identified, this threshold approximated 29 kg/m². The ECG voltage criteria were subsequently adjusted for the extent of the attenuation of the relationship between BMI and ECG voltages in those with a BMI ≥29 kg/m² as compared with those with a BMI <29 kg/m². To achieve this goal, the slopes (β‐coefficients) of the BMI‐ECG voltage relationships in those with BMI < and ≥29 kg/m² were identified from linear regression analysis. The ECG voltages in those with BMI ≥29 kg/m² were then corrected by the following formula: ([BMI − 29] × [difference in BMI vs ECG voltage relation β‐coefficients of those < as compared with those ≥29 kg/m²]). This approach was only applied to QRS voltage criteria that showed reasonable correlations with BMI in those with a BMI <29 kg/m² and in which differences in the slopes of the relations between BMI and ECG voltage criteria were noted in those with a BMI ≥ as compared with < 29 kg/m². As the presence of a reduced left ventricular ejection fraction (LVEF) may attenuate the performance of ECG criteria for LVH detection, and 3.6% of the sample had an LVEF <50% (none had an LVEF <40%), sensitivity analysis was conducted without these individuals included. However, as this analysis revealed essentially the same findings as when including these participants, these data are not reported.

Results

The demographic and clinical characteristics of the study sample are given in Table 1 (see also Supporting Information, Table 1, in the online version of this article). A comparison of the characteristics of the participants from the community sample with and without echocardiographic or ECG data is also available in Supporting Information, Table 1, in the online version of this article. No differences were noted in the characteristics. In the study sample, 43.0% of participants were obese, and of the 43.1% of participants that were hypertensive, only 34.5% had controlled blood pressure. In the study sample, 21.8% of participants had LVH, and participants with a BMI ≥29 kg/m² had far more LVH than did participants with a BMI <29 kg/m² (Table 1). Importantly, no participants had evidence of a bundle branch block, a clinically significant arrhythmia, hyperkalemia or hypokalemia, a history of or ECG evidence of a prior myocardial infarction (a rare occurrence in this community) or pacemaker insertion, or an LVEF <40%, all of which could influence the ability to detect LVH using ECG criteria.

Table 1.

Characteristics and ECG and Echocardiographic Measures in Participants of African Ancestry With BMI < or ≥ 29 kg/m²

Characteristic	BMI <29 kg/m2 (n = 351)	BMI ≥29 kg/m2 (n = 310)a
Female sex	48.4	84.8
Age, y	37.0 ± 18.2	50.1 ± 14.4
BMI, kg/m2	23.5 ± 3.4	36.1 ± 5.7
HTN	29.9	58.1
DM or HbA1c >6.1%	12.5	39.4
Conventional SBP/DBP, mm Hg	124 ± 21/82 ± 12	134 ± 23/87 ± 13
LVM, g	144.4 ± 46.0	165.5 ± 52.6
LVMI, g/m2.7	38.3 ± 11.9	47.1 ± 14.4
LVMI >51 g/m2.7	12.8	31.9
Voltage criteria, mV
Cornell	1.63 ± 0.86	2.02 ± 0.76
Sokolow‐Lyon	2.90 ± 0.90	2.42 ± 0.79
RaVL	0.29 ± 0.30	0.50 ± 0.40
Gubner‐Ungerleider	0.83 ± 0.51	1.16 ± 0.67
Lewis	−0.04 ± 0.96	0.60 ± 0.97
Time‐voltage criteria, mV × ms
Cornell × QRS duration	141 ± 91	179 ± 84
Sokolow‐Lyon × QRS duration	248 ± 99	213 ± 88
Gub‐Ung × QRS duration	73 ± 55	104 ± 71

Abbreviations: BMI, body mass index; DBP, diastolic blood pressure; DM, diabetes mellitus; ECG, electrocardiographic; Gub‐Ung, Gubner‐Ungeleider; HbA_1c, glycated hemoglobin; HTN, hypertension; LVM, left ventricular mass; LVMI, left ventricular mass index; SBP, systolic blood pressure; SD, standard deviation.

Data are presented as % or mean ± SD. Sokolow‐Lyon = S_V1 + R_V5 or _V6; Cornell = R_aVL + S_V3 (+0.8 mV in women); Gub‐Ung = R_I + S_III; Lewis = (R_I + S_III) − (R_III + S_I).

^a P < 0.0001 vs BMI < 29 kg/m².

The relations between BMI and QRS voltage criteria in those with a BMI < or ≥29 kg/m² are shown in Figure 1 and also in Supporting Information, Table 2, in the online version of this article. The strongest positive relations between BMI and voltage criteria were with R_aVL, Gubner‐Ungerleider, and Lewis criteria, but these relations were only noted in those with a BMI <29 kg/m². In contrast, in those with a BMI ≥29 kg/m², no significant relations were noted. Body mass index showed no relations with Sokolow‐Lyon criteria in those with a BMI <29 kg/m2 and inverse relations in those with a BMI ≥29 kg/m². Body mass index showed weak relations with Cornell voltages, with no differences in the strength (r value; data not shown) or slopes (β‐coefficients; Supporting Information, Table 2, in the online version of this article) in those with a BMI <29 kg/m² as compared with those with a BMI ≥29 kg/m². Correction of R_aVL, Gubner‐Ungerleider, and Lewis voltages for the attenuating effect of BMI on these voltages in those with a BMI ≥29 kg/m² resulted in an improved strength and similar slopes of BMI‐QRS voltage relations in obese individuals as compared with those noted in individuals with a BMI <29 kg/m² (see Figure 1 and also Supporting Information, Table 2, in the online version of this article).

Figure 1.

Impact of adjusting for obesity effects on QRS voltage criteria for LVH detection on the strength (r values) of the relations between BMI and several ECG criteria for LVH in 661 participants (351 with BMI <29 kg/m² and 310 with BMI ≥29 kg/m²) from a community sample of African ancestry. Adjustments for obesity effects in those with BMI ≥29 kg/m² were as given in Table 2. Cornell and Sokolow‐Lyon criteria could not be adjusted for obesity effects on QRS voltages (see text). * P < 0.05, ** P < 0.01 vs relationship in those with a BMI <29 kg/m²; † P < 0.05, †† P < 0.005 vs relationship before correction for obesity effects. Abbreviations: BMI, body mass index; CI, confidence interval; corr, corrected voltage; ECG, electrocardiographic; LVH, left ventricular hypertrophy.

Table 2.

Impact of Adjusting for Obesity Effects on QRS Voltages on the Performance (Area Under the ROC Curve) of ECG Criteria for LVH (LVM Indexed to Height^2.7 > 51 g/m^2.7 [LVMI]) Detection in Body Size–Specific Groups

	AUC ± SEM
LVH vs	BMI < 29 kg/m2, n = 351	BMI ≥ 29 kg/m2, n = 310	BMI ≥ 29 kg/m2, n = 310	Formula for Correctiona if BMI ≥ 29 kg/m2
Adjustment for Obesity Effect→		None	Corrected Voltagea
Cornell voltage	0.678 ± 0.046b	0.554 ± 0.036c	—	—
Sokolow‐Lyon voltage	0.577 ± 0.044	0.489 ± 0.036	—	—
RaVL	0.777 ± 0.041b	0.589 ± 0.034d, e	0.621 ± 0.033b, c, f	RaVL voltage + ([BMI − 29] × 0.017)
Gub‐Ung voltage	0.739 ± 0.039b	0.575 ± 0.035d, e	0.600 ± 0.034b, c, f	Gub‐Ung voltage + ([BMI − 29] × 0.019)
Lewis voltage	0.734 ± 0.043b	0.590 ± 0.034d, e	0.632 ± 0.033b, c, f	Lewis voltage + ([BMI − 29] × 0.060)

Abbreviations: AUC, area under the ROC curve; BMI, body mass index; ECG, electrocardiographic; Gub‐Ung, Gubner‐Ungeleider; LVH, left ventricular hypertrophy; LVM, left ventricular mass; LVMI, left ventricular mass index; ROC, receiver operating characteristic; SEM, standard error of the mean.

^aCorrected using the following formula: voltage + ([BMI − 29] × [difference in BMI vs ECG voltage relation β‐coefficients of those < as compared with those ≥ 29 kg/m²]). Cornell and Sokolow‐Lyon criteria could not be adjusted for obesity effects on QRS voltages (see text).

^b P < 0.0001 for significance of AUC.

^c P < 0.05 vs AUC in those with a BMI <29 kg/m².

^d P < 0.05 for significance of AUC.

^e P < 0.0001 vs AUC in those with a BMI <29 kg/m².

^f P < 0.005 vs uncorrected AUC in those with a BMI ≥29 kg/m².

Table 2 shows the performance (AUC of ROC) of ECG voltage criteria in those with BMI < or ≥29 kg/m². The strongest performance for LVH detection in those with BMI <29 kg/m² was noted for R_aVL, Gubner‐Ungerleider, and Lewis criteria, with Cornell criteria showing an intermediate level of performance and Sokolow‐Lyon criteria showing no significant performance for LVH detection. Although R_aVL, Gubner‐Ungerleider, and Lewis criteria showed a modest level of performance for LVH detection in those with BMI ≥29 kg/m², neither Cornell nor Sokolow‐Lyon criteria showed a significant performance for LVH detection. Importantly, the performance of R_aVL, Gubner‐Ungerleider, and Lewis criteria in those with BMI ≥29 kg/m² was markedly lower than that noted in those with BMI <29 kg/m². However, adjusting R_aVL, Gubner‐Ungerleider, and Lewis voltages for the attenuating effect of BMI on these voltages resulted in an improved performance for LVH detection. Nevertheless, the performance of the BMI‐corrected QRS voltage criteria still showed a reduced performance for LVH detection in those with BMI ≥29 kg/m² as compared with the QRS voltage criteria alone in those with BMI <29 kg/m² (Table 2).

Table 3 shows the performance, sensitivity, and specificity for the detection of LVH in all participants before and after adjusting QRS voltages using the QRS voltage × BMI product or using the approach shown in Table 2 (corrected QRS voltage). Figure 1 shows the ROC curves before and after correcting R_aVL and Lewis voltages using the approach shown in Table 2 (corrected QRS voltage). Of all the criteria for LVH detection, R_aVL and Lewis showed the strongest performance. The performance and sensitivity for LVH detection were significantly improved for R_aVL and Gubner‐Ungerleider criteria when using both the QRS voltage × BMI product and the corrected QRS voltage approach (Table 3, Figure 2). However, the corrected QRS voltage approach, but not the QRS voltage × BMI product, improved the performance for LVH detection for Lewis criteria. The performance and sensitivity for LVH detection were significantly improved for Cornell and Sokolow‐Lyon voltage criteria when using the QRS voltage × BMI product (Table 3). However, the corrected QRS voltage approach for R_aVL and Lewis criteria gave the best overall performance for LVH detection (Table 3, Figure 2). Although the QRS voltage × BMI product for R_aVL, Gubner‐Ungerleider, and Cornell voltage criteria showed the greatest sensitivity for LVH detection, this was at the expense of the greatest decreases in specificity. In contrast, the corrected QRS voltage approach for R_aVL and Lewis criteria showed the next‐highest sensitivity for LVH detection with only modest (Lewis) or no significant (R_aVL) decreases in specificity. At 85% specificity, the corrected QRS voltage approach (32.7% to 38.8%; P < 0.05) and the QRS voltage × BMI product approach (32.7% to 38.8%; P < 0.05) for R_aVL, but not for any of the other ECG criteria, showed a significantly improved sensitivity for LVH detection (Table 3).

Table 3.

Impact of Adjusting for Obesity Effects on QRS Voltages on the Performance (Area Under the ROC Curve), Sensitivity, and Specificity of ECG Criteria for LVH (LVM Indexed to Height^2.7 [LVMI] >51 g/m^2.7) Detection in a Community Sample of African Ancestry (N = 661)

		AUC			Sensitivity			Specificity
Adjustment for Obesity Effect→	None	Voltage‐BMI Product	Corrected Voltagea	None	Voltage‐BMI Product	Corrected Voltagea	None	Voltage‐BMI Product	Corrected Voltagea
LVH vs
Cornell voltage	0.635 ± 0.026	0.706 ± 0.024b	—	13.2	51.4c	—	94.0	78.7c	—
Sokolow‐Lyon voltage	0.513 ± 0.027	0.645 ± 0.025b	—	6.3	33.3c	—	91.7	81.2c	—
RaVL	0.695 ± 0.025	0.718 ± 0.024b	0.733 ± 0.022b, d	30.6	53.5c	42.4c	86.2	78.0e	83.0
Gub‐Ung voltage	0.668 ± 0.025	0.719 ± 0.023b	0.698 ± 0.024b, d	33.3	52.1c	38.9e	87.0	76.6e	84.3
Lewis voltage	0.685 ± 0.024	0.690 ± 0.025	0.724 ± 0.023b, d	34.0	43.1c	49.3c	86.3	82.2e	81.0e

Abbreviations: AUC, area under the ROC curve; BMI, body mass index; ECG, electrocardiographic; Gub‐Ung, Gubner‐Ungeleider; LVH, left ventricular hypertrophy; LVM, left ventricular mass; LVMI, left ventricular mass index; ROC, receiver operating characteristic.

^aCorrected using the approach given in Table 2, using the following formula in those with a BMI ≥ 29 kg/m²: voltage + ([BMI − 29] × [difference in BMI vs ECG voltage relation β‐coefficients of those < as compared with those ≥ 29 kg/m²]). Cornell and Sokolow‐Lyon criteria could not be adjusted for obesity effects on QRS voltages (see text). Voltage thresholds to determine sensitivity and specificity are 3.05 mV for Cornell, 4.00 mV for Sokolow‐Lyon, 0.73 mV for R_aVL, 1.50 mV for Gubner‐Ungerleider, and 1.17 mV for Lewis. All AUC values are significant at P < 0.0001 except Sokolow‐Lyon uncorrected, which shows no significant performance.

^b P < 0.0001 vs AUC for uncorrected data.

^c P < 0.001 vs uncorrected.

^d P < 0.0005 vs voltage‐BMI product.

^e P < 0.05 vs uncorrected.

Figure 2.

Effect of adjustments for obesity effects on QRS voltages on the performance (ROC curves) of R_aVL and Lewis voltages for LVH detection (LVMI >51 g/m^2.7) in 661 participants from a community sample of African ancestry. Adjustments for obesity effects were as follows: (A) voltage × BMI; (B), voltage corrected as described in Table 2. Cornell and Sokolow‐Lyon criteria could not be adjusted for obesity effects on QRS voltages (see text), and Gubner‐Ungerleider voltages were only modestly altered. A comparison of the AUCs is made in Table 2. Abbreviations: AUC, area under the curve; BMI, body mass index; LVH, left ventricular hypertrophy; LVMI, left ventricular mass index; ROC, receiver operator characteristic.

Discussion

The study subjects consisted of a large community‐based sample of African ancestry with a high prevalence of obesity, HTN, and LVH. In those with a BMI ≥29 kg/m², either no significant relationships (R_aVL, Lewis, and Gubner‐Ungerleider), modest relations (Cornell), or inverse relations (Sokolow‐Lyon) between standard ECG voltage criteria and LVMI, and a markedly reduced performance for LVH detection, were noted. Correcting for the attenuating effects of BMI on ECG voltages (by identifying the extent to which QRS voltages are reduced in those with BMI ≥29 kg/m² as compared with those with BMI <29 kg/m² and modifying the voltages in those with a BMI ≥29 kg/m² by a mathematical formula (voltage + [{BMI − 29} × {difference in BMI vs ECG voltage relation β‐coefficients of those < as compared with those ≥29 kg/m²}]) showed significantly improved performance for Lewis, Gubner‐Ungerleider, and R_aVL criteria and an improved sensitivity for LVH detection of R_aVL voltages at 85% specificity. Although the BMI‐QRS voltage product (for Cornell, R_aVL, Gubner‐Ungerleider, and Lewis) increased the performance for LVH detection, the corrected R_aVL voltages showed overall the greatest performance and the most marked increases in sensitivity for only modest reductions in specificity when employing voltage thresholds to detect LVH.

The attenuating effect of obesity on QRS voltage criteria for LVH detection is well‐described.9, 10, 11, 12, 13, 14, 15 This has recently been demonstrated to be in part amended by use of a product of QRS voltage amplitudes and BMI in European populations.21, 22 However, this approach does not correct for the attenuating effect of obesity on QRS voltages, but rather employs BMI together with ECG voltages as a determinant of LVMI. Moreover, whether this approach is appropriate in subjects of African ancestry, where the dampening effects of obesity on QRS voltages are particularly striking,20 is unknown. In keeping with that noted in European populations, we show an improved performance (AUC for ROC) for LVH detection21, 22 using the QRS voltage‐BMI product. We nevertheless demonstrate that the best overall performance for LVH detection is by modifying the QRS voltages for R_aVL and Lewis criteria in those with BMI ≥29 kg/m² by the aforementioned mathematical formula. Moreover, modifying R_aVL voltages in those with BMI ≥29 kg/m² by a mathematical formula improved the sensitivity for LVH detection at 85% specificity, and using thresholds of R_aVL voltages produced the greatest increase in sensitivity at the expense of insignificant decreases in specificity. In contrast, although the use of the R_aVL voltage‐BMI product improved the sensitivity for LVH detection at 85% specificity, using thresholds of R_aVL voltages produced an increased sensitivity at the expense of marked decreases in specificity. This is in apparent contrast to the ability of the QRS voltage‐BMI product to improve the sensitivity to detect LVH in a European sample, with only modest reductions in specificity.21 In the present study, although we are unable to adequately confirm the value of using the QRS voltage‐BMI product in groups of black African descent, we offer a novel and apparently better approach to improving the value of QRS voltages to enhance LVH detection. In this regard, in individuals of African ancestry with BMI ≥29 kg/m², we suggest that QRS voltages are corrected for the attenuating effects of BMI using a mathematical formula specific for R_aVL: R_aVL voltage + ([BMI − 29] × 0.017).

Prior approaches to enhancing the value of QRS voltage criteria in detecting LVH and risk predicting have focused on Cornell voltages, and to some degree Sokolow‐Lyon criteria.21, 22 In the present study, BMI showed no relations with Sokolow‐Lyon voltages in those with BMI <29 kg/m² and inverse relations with Sokolow‐Lyon voltages in those with BMI ≥29 kg/m². Moreover, BMI showed weak relations with Cornell voltages, with no differences in the strength (r value) or slopes (β‐coefficients) of these relations in those with BMI <29 kg/m² as compared with those with BMI ≥29 kg/m². The inverse or weak relations between BMI and Cornell or Sokolow‐Lyon voltages are attributed to marked inverse relations between BMI and R‐ or S‐wave voltages in chest leads in the present community sample.20 These inverse or weak relations prevented us from identifying the extent to which obesity, rather than other factors, attenuates the relationship between BMI and Cornell or Sokolow‐Lyon voltages. Thus, a mathematical equation to adjust Cornell or Sokolow‐Lyon voltage criteria for the attenuating effects of obesity could not be determined. Moreover, the weak relationship between BMI and Cornell voltages, even in those with BMI <29 kg/m², limited the value of using the Cornell voltage‐BMI product to detect LVH. In this regard, this would have been no better than using BMI alone to detect LVH.

Although we show that adjusting QRS voltages in those with BMI ≥29 kg/m² for the attenuating effect of obesity on these voltages generates a significant performance for LVH detection, the performance for LVH detection nevertheless remained lower than that noted in those with BMI <29 kg/m². Whether this is attributed to variations in the impact of the attenuating effect of obesity on QRS voltages between individuals (hence, the assumption that the same equation should apply to all is incorrect), or whether the equation generated requires improvement, is unknown. In this regard, to identify the actual slope of the BMI‐QRS voltage relations, we assumed that QRS voltages for R_aVL, Gubner‐Ungerleider, or Lewis voltages were unaffected by body size or fat mass in those with BMI <29 kg/m² (the threshold value below which the strongest relations between BMI and ECG voltages were noted). Further work is required in a significantly larger database to identify whether an excess adiposity in overweight individuals with an even lower BMI attenuates the impact of BMI on QRS voltages.

Study Limitations

The present study has several limitations. In this regard, we assessed LVH using echocardiography rather than magnetic resonance imaging. As magnetic resonance imaging provides a more accurate evaluation of LVM than echocardiography, it is possible that our ability to detect LVH using QRS voltage criteria or BMI‐adjusted QRS voltage criteria may have been better than that described. Second, more women than men volunteered for the present study; hence, our ability to improve on the detection of LVH using QRS voltage criteria may relate mainly to women and not to men. In this regard, we were not statistically powered to perform sex‐specific analysis. Third, we employed BMI only as an index of excess adiposity. Whether alternative indices of excess adiposity that account for fat distribution provide different information is unknown. Fourth, in the present study, largely because of the low prevalence of an ECG strain pattern (consistent with the prevalence in the general population), we did not assess either the Framingham or Romhilt‐Estes criteria. However, as with Sokolow‐Lyon and Cornell criteria, these criteria rely heavily on voltage criteria in chest leads, which, as previously demonstrated,20 show no positive relations with BMI even in those with BMI <29 kg/m². As we are unable to correct chest lead R‐ or S‐wave voltages for the attenuating effects of obesity, the performance of neither Framingham nor Romhilt‐Estes criteria would have been significantly modified by this approach.

Conclusion

In a community of African ancestry with a high prevalence of obesity, uncontrolled HTN and LVH, we show that correcting R_aVL voltages for the attenuating effects of obesity using a simple formula (R_aVL voltage + [{BMI − 29} × 0.017]) enhances both the performance and sensitivity for LVH detection. The improved performance exceeded that of the QRS voltage (any criteria) × BMI product, which also showed a strikingly decreased specificity despite an improved sensitivity for LVH detection. We therefore offer a novel approach to correcting ECG voltages for the attenuating effects of obesity in individuals of African ancestry, where obesity markedly diminishes the ability of any ECG criteria to detect LVH.

Supporting information

Table S1. Characteristics of study participants with and without echocardiographic data.

Table S2. Impact of adjusting for obesity effects on QRS voltages on the slope (β‐coefficient) of the relationship between body mass index (BMI) and electrocardiographic criteria for left ventricular hypertrophy detection in participants of a community sample of African ancestry.