William da Costa, MD,a Andrés Ricardo Perez Riera, MD,b,⁎ Francisco de Assis Costa, MD,c
Maria Teresa Nogueira Bombig, MD, PhD,a Ângelo Amato Vicenzo de Paola, MD, PhD,a Antonio Carlos Camargo Carvalho, MD, PhD,a Francisco Helfenstein Fonseca, MD, PhD,a
Bráulio Luna Filho, MD, PhD,a Rui Póvoa, MD, PhDa
aUniversidade Federal de São Paulo (Federal University of São Paulo), Escola Paulista de Medicina (Paulista School of Medicine), São Paulo, SP, Brazil bABC's Medical School, ABC Foundation, Santo André, São Paulo, Brazil
cUniversidade Estadual de Ciências da Saúde de Alagoas (State Health Sciences University of Alagoas), Maceió, Brazil
Received 18 January 2008
Abstract Introduction:Left ventricular hypertrophy (LVH) and obesity are important cardiovascular risk factors. This study evaluates the influence of obesity on the diagnostic performance of the most used electrocardiographic criteria for LVH in hypertensive patients.
Methods:One thousand two hundred four outpatients from the Hypertensive Unit of the Hospital São Paulo, São Paulo, SP, Brazil, were studied. All underwent 12-lead electrocardiogram and echocardiogram. The most known electrocardiographic criteria for LVH were assessed and compared with the left ventricular mass index obtained by echocardiogram in obese and nonobese groups of hypertensive patients.
Results:The population's mean age was 57.4 ± 4.7 years; 351 were men (29.1%) and 853 women (70.8%). Cornell voltage, Cornell duration, Sokolow-Lyon voltage, Romhilt-Estes criteria, and R wave in aVL 11 mm or higher showed a positive correlation with left ventricular mass index (P b .05). Notwithstanding, there were no changes regarding specificity for obese or nonobese characteristics. However, sensitivity had a statistically significant decrease in obese patients in regard to Sokolow-Lyon voltage and Romhilt-Estes criteria and strain pattern (P b .05).
Conclusion:Cornell voltage and Cornell duration criteria, Perugia score, R wave in aVL, and QTc variable had no significant changes in diagnostic sensitivity in the obese patients.
© 2008 Elsevier Inc. All rights reserved.
Keywords: Obesity; Left ventricular hypertrophy; Electrocardiography
Introduction
Left ventricular (LV) mass increase diagnosed by the electrocardiogram (ECG) involves clinical and prognostic consequences. Particularly in hypertensive patients, ECG patterns of LV hypertrophy (LVH) significantly increase the cardiovascular risk.1 The Framingham Heart Study and others consistently showed that changes in QRS voltage and ventricular repolarization are important determinants of cardiovascular morbidity and mortality.2
Obesity, on the other hand, is also an independent cardiovascular risk factor and is usually associated with other comorbidities such as diabetes mellitus, systemic arterial hypertension, and metabolic syndrome.3
The ECG is a low-cost, widely used, noninvasive method with good reproducibility and substantial epidemiologic importance, but unfortunately, it has a low sensitivity as regards LVH. In this context, it is known that several conditions may alter its sensitivity and specificity. Obesity undoubtedly is one the most important and less studied.4-7
This study with a large sample of obese individuals with hypertension and LV mass determined by echocardiogram evaluates the accuracy of the most known electrocardio- graphic criteria for LVH.
Journal of Electrocardiology 41 (2008) 724 – 729
www.jecgonline.com
⁎Corresponding author. 04417-100, São Paulo, SP, Brazil. Tel.: +55 011 5621 2390; fax: +1 011 5625 7278/5506-0398.
E-mail address:riera@uol.com.br
0022-0736/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.jelectrocard.2008.05.010
Patients
This is a prospective study with 1204 consecutive outpatients from the Hypertension Unit of the Cardiovas- cular Division at the São Paulo Hospital—Universidade Federal do Estado de São Paulo (UNIFESP), São Paulo, Brazil. Patients with cardiovalvular disease, acute or chronic coronary artery disease, previous myocardial infarction, Chagas' disease, rhythm disorders, bundle branch blocks or other conduction abnormalities, digitalis use, preexcitation syndromes, poor technical quality of echocardiogram, or any other condition that may potentially distort LV geometry, such as emphysema and pericardial and pleural effusion, were excluded.
Electrocardiogram
All patients underwent 12-lead ECG (record speed, 25 mm/s; standard calibration, 1.0 mV/cm; Dixtal EP3 device, Brazil). The tracings were interpreted in a blind fashion by the certified cardiologists with experience in ECG reading. Neither the cardiologist nor the echocardiographer had knowledge of the interpreted examinations. A 5-fold magnifying glass was used by the ECG observers for better analysis of the tracings.
Only one cardiologist analyzed all ECG tracings. QRS complex axis and duration, R wave amplitude in leads aVL, V5, and V6, and S wave amplitude in V1, V2, and V3were
measured in all ECG. The strain pattern in V5and V6was
also identified. Eight LVH electrocardiographic criteria were assessed separately.
1. Cornell voltage criteria: RaVL + SV320 mm or higher
for women and 28 mm or higher for men.8
2. Cornell duration criteria: (RaVL + SV3) × QRS
duration; for women, add 8 mm, 2440 mm milli- seconds or higher.9
3. Sokolow-Lyon voltage criteria: SV1 + RV5 or V6
35 mm or higher.10
4. Romhilt-Estes score: higher R or S amplitude 30 mm or higher in horizontal plane or 20 mm or higher in frontal plane or strain pattern in lead V5 or V6 (if using
digitalis, this means only one point) or left atrium increase according to Morris index (3 points); electric axis of AQRS (spatial axis of QRS on frontal plane) greater than −30° (2 points); QRS duration 90 milliseconds or higher in V5 or V6, or ventricular
activation time 50 milliseconds or higher in V5or V6
(one point). According to this score, LVH is diagnosed when the sum is 5 points or higher.11
5. Perugia score: LVH is diagnosed by the presence of one or more of the following findings: Cornell criteria, considering the limit 20 mm or higher for women and 24 mm or higher for men, Romhilt-Estes score, and strain pattern.12
6. R wave of aVL 11 mm or higher.13
7. Presence of strain pattern: defined as a convex ST- segment depression with asymmetric T-wave inversion opposed to QRS complex in leads V5or V6.14
8. QTc interval: measured in milliseconds from the beginning of Q wave to the end of T wave (corrected by Bazett's formula: QTc = QT/RR1/2, normal values
from 350 to 440 milliseconds).15
The reproducibility analysis of the method was done at random with 100 tracings from the original sample. With this objective, the amplitudes of R and S waves and QRS complex duration were measured.
Transthoracic echocardiogram
The tests were performed in the Echo-Doppler-Cardio- graphy Department at Universidade Federal do Estado de São Paulo Escola Paulista de Medicina (UNIFESP-EPM) (Doppler echocardiography Philips - ATL HP 1500 device, Philips, Bothell, WA, USA) using 2.0 and 3.5 MHz transducers. The patients were placed in the left lateral decubitus, and the images were obtained from the left parasternal area, between the fourth or fifth intercostal space, evaluating the usual sections for the complete study for M- and bidimensional modes. According to the Penn Conven- tion, the following measurements were performed: LV size in systole and diastole, LV interventricular septum thickness and posterior wall thickness at end diastole, end-diastolic, end-systolic, and diastolic shortening, and ejection fraction by the cube method. The LV mass was calculated by the formula: LV mass = 0.8 × {1.04[(IVSD + LVEDD + LVPWTD)3 − (LVEDD)3]} + 0.6 g,16 where IVSD is the interventricular septum in diastole, LVEDD is the LV end- diastolic diameter, and LVPWTD is the LV posterior wall in diastole. The LV mass was indexed to the body surface area for adjustment of heart weight to patient size variations. Body surface area was calculated by the following formula: BSA = (W − 60) × 0.01 + H, where BSA is the body surface area, in square meters, W is the weight, in kilograms, and H is the height, in meters.17 Body mass index (BMI) was calculated by dividing the weight (kg) by the squared height (m2). Patients with BMI less than 30 kg/m2were considered
nonobese; patients with BMI 30 kg/m2 or higher were
considered obese. Left ventricular hypertrophy diagnosis was confirmed when LV mass index (LVMI) was 89 g/m2or
higher for women and 103 g/m2or higher for men.18 The ECG was recorded after the echocardiograms.
Statistical analysis
Continuous variables were expressed as mean ± SD. Categorical variables were expressed as percentage. Pearson linear correlation coefficient was used to associate LVMI with several electrocardiographic criteria studied in both obese and nonobese patients. The obtained sensitivity,
Table 1
Sample characteristics according to BMI, age, and sex BMI b30 (n = 852) BMI ≥30 (n = 352)
Age Men/women Age Men/women
58.3 ± 11.6 281 (33%)/ 571 (67.0%)
55.3 ± 10.8 70 (19.9%)/ 282 (80.1%)
specificity, and positive and negative predictive values were used to determine the performance of diagnostic criteria for LVH. McNemar test assessed the statistical difference of the several electrocardiographic criteria analyzed between obese and nonobese patients. The reproducibility study was conducted by 3 observers who read the ECG independently. To analyze the ECG variables, we applied the κ test.19This one is a measurement index for a nominal or categorical variable that corrects the agreement observed only by chance. When the κ test has values higher than 0.75, it is interpreted as excellent; when it ranges from 0.75 to 0.40, it is interpreted as good; and when it is lower than 0.40, it is interpreted as poor agreement. For statistical significance, 95% confidence intervals and P b .05 were considered for comparison. Results
We studied 1204 patients, 351 male (29.1%) and 853 female (70.8%), whose mean age was 57.4 ± 4.7 years. In this group, 852 (70.7%) were nonobese (BMI b30 kg/m2) and 352 (29.3%) were obese, that is, with BMI 30 kg/m2or higher (Table 1).
According to Pearson statistics, there was a significant positive correlation between LVMI and the electrographic variables in both nonobese and obese groups. The QTc interval did not show correlation between LVMI in obese and nonobese patients.
Because of statistical limitation, Perugia score and strain pattern were not correlated with LVMI (they are qualitative data) (Table 2).
Regarding specificity values of LVH electrocardiographic criteria, obesity did not cause significant differences. However, regarding the sensitivity value, the Sokolow- Lyon voltage criteria, the Romhilt-Estes point score, and the strain pattern showed a significant decrease. Such decrease was not observed for Cornell voltage and Cornell duration criteria, Perugia score, and R in aVL 11 mm or higher. These findings suggest strongly that the latter criteria are much more reliable to detect LVH in obese hypertensive popula- tion (Table 3). Fig. 1 shows an ECG sample of an obese patient with LVH by echocardiogram who is positive according to Perugia and Cornell voltage and duration criteria but not for Sokolow-Lyon and Romhilt-Estes.
Reproducibility analysis showed the interobserver con- cordance among the 3 observed values ranging from 0.82 to QRS complex duration and 0.98 for S and R wave amplitudes. Both are excellent values for κ statistics. Discussion
Left ventricular hypertrophy is a consistent and independent predictor of high cardiovascular risk without distinction regarding comorbidities, race, and presence or absence of systemic arterial hypertension or coronary artery disease, both in clinical and epidemiologic studies. This is the main reason to screen patients for LVH by a diagnostic method such as ECG, which presents easiness, availability, and large applicability combined with low cost. In obese individuals that are nowadays a growing risk group, the accuracy of ECG for detection of LVH has been reported as of lower precision.20 Population and cohort studies in hypertensive patients showed ECG-diagnosed LVH as an independent factor for increase in the incidence of coronary artery disease, cardiovascular events, and all- cause mortality besides a strong marker for heart failure and stroke.21,22
Echocardiogram is an excellent examination to assess LV mass. This technique constitutes the standard for comparison and validation of LVH electrocardiographic criteria.23 However, LV mass indexation by body surface area (LVMI) underestimates the real heart mass in obese patients, which also can undervalue the significance of this cardio- vascular risk factor. Electrocardiogram, even with low sensitivity, may distinguish these patients and is evidently
Pearson bivariable correlation between LVMI and the several electro- cardiographic variables studied in nonobese and obese groups
Variable Nonobese BMI b30 Obese BMI ≥30 Pearson bivariable correlation P Pearson bivariable correlation P Cornell voltage 0.384 b.05 0.359 b.05 Cornell duration 0.461 b.05 0.392 b.05 Sokolow-Lyon 0.458 b.05 0.285 b.05 Romhilt-Estes 0.499 b.05 0.329 b.05 R wave in aVL 0.338 b.05 0.303 b.05 QTc 0.079 NS 0.082 NS
NS indicates not significant.
Table 3
Sensitivity, specificity, and P values regarding the electrocardiographic variable for LVH study in obese and nonobese patient groups
Variable Sensitivity (CI) (%) Specificity (CI) (%)
Nonobese BMI b30 Obese BMI ≥30 P Nonobese BMI b30 Obese BMI ≥30 P Cornell voltage 18.8 (14.9-23.6) 18.9 (13.2-26.5) .892 96.5 (94.6-97.8) 97.6 (94.7-99.0) .545 Cornell duration 22.4 (18.1-27.4) 21.9 (15.8-29.8) .885 95.4 (93.3-96.9) 97.6 (94.9-99.0) .187 Sokolow-Lyon voltage 17.5 (13.7-22.2) 3.7 ⁎ (1.6-8.6) .006 96.5 (94.6-97.8) 97.6 (94.7-99.0) .545 Romhilt-Estes 20.4 (16.0-25.1) 6.8 ⁎ (3.6-12.5) .021 95.7 (94.1-97.2) 96.3 (92.9-98.1) .967 Perugia 39.6 (34.3-45.2) 27.2 (14.9-34.7) .184 89.9 (87.1-92.2) 89.4 (84.6-92.8) .927 R wave in aVL ≥11 mV 10.3 (7.5-14.3) 11.3 (7.0-17.9) 1.000 96.3 (94.4-97.6) 97.2 (94.1-98.7) .695 Strain pattern 34.7 (29.6-40.2) 18.9 ⁎ (13.2-26.5) .040 92.6 (90.2-94.6) 90.7 (85.7-93.7) .381 QTc 35.6 (30.1-41.6) 37.1 (28.4-46.8) .813 72.7 (68.8-76.2) 69.0 (62.6-74.8) .121
CI indicates confidence interval.
more reproducible than the echocardiogram in this popula- tion mainly during follow-up.
Thus, studies capable of assessing LVH electrocardio- graphic criteria in obese patients are important to determine the real performance of this method in this clinical context, because they, when used in the clinical setting, have a similar accuracy in the general population. The modified Devereux formula16 has a good correlation with the LV mass in necropsy studies (r = 0.90, P b.001). This formula applies to normal geometric conformation ventricles considered to be ellipsoid and within standards, which allow volume extra- polation by the cube formula. Therefore, in dilated hearts, instead, it tends to have a globoid shape. Consequently, LV
mass estimation is much more likely to be calculated wrongly and should be avoided. With this concern, patients with significant LV geometric distortions were excluded.24-25Left ventricular mass index allowed comparison among subjects with different body structures, thus, obtaining values that may reliably identify groups of high risk for cardiovascular and cerebrovascular events.26
The correlation between LVMI and hypertrophy by the electrocardiographic criteria has shown to be regular (Pearson test) in nonobese and obese patients.
In the literature, these indices present variable correla- tions. In the studies of Horton et al,5a 0.686 correlation of the Sokolow-Lyon criterion with LV mass was found, a
Fig. 1. Electrocardiogram of a female obese patient with LVH in the echocardiogram (118 g/m2) presenting only positive Perugia and Cornell voltage and
duration criteria for LVH in the ECG.
Table 4
Sensitivity, specificity, positive and negative predictive values, and confidence interval for both sexes in obese and nonobese patients according to Cornell voltage and duration criteria, Sokolow-Lyon, and Romhilt-Estes criteria
Nonobese (BMI b30) Obese (BMI ≥30)
Women Men Women Men
Cornell voltage Sen (CI) (%) 0.222 (0.171-0.282) 0.115 (0.065-0.194) 0.207 (0.142-0.292) 0.095 (0.027-0.289) Spec (CI) (%) 0.959 (0.933-0.975) 0.978 (0.945-0.992) 0.971 (0.934-0.988) 1.000 (0.921-1.000) PPV (CI) (%) 0.758 (0.638-0.848) 0.733 (0.480-0.891) 0.821 (0.644-0.921) 1.000 (0.342-1.000) NPV (CI) (%) 0.678 (0.637-0.717) 0.679 (0.621-0.733) 0.655 (0.595-0.711) 0.703 (0.582-0.801) Cornell duration Sen (CI) (%) 0.203 (0.154-0.262) 0.198 (0.131-0.289) 0.270 (0.196-0.360) 0.190 (0.077-0.400) Spec (CI) (%) 0.961 (0.936-0.977) 0.967 (0.931-0.985) 0.965 (0.926-0.984) 0.956 (0.852-0.988) PPV (CI) (%) 0.754 (0.629-0.848) 0.760 (0.566-0.885) 0.833 (0.681-0.921) 0.667 (0.300-0.903) NPV (CI) (%) 0.674 (0.632-0.713) 0.698 (0.639-0.751) 0.672 (0.611-0.728) 0.717 (0.592-0.815) Sokolow-Lyon Sen (CI) (%) 0.113 (0.077-0.163) 0.313 (0.229-0.411) 0.045 (0.019-0.101) ⁎ 0.000 (0.000-0.155) ⁎
Spec (CI) (%) 0.978 (0.957-0.989) 0.940 (0.896-0.966) 0.988 (0.959-0.997) 0.956 (0.852-0.988) PPV (CI) (%) 0.750 (0.579-0.867) 0.732 (0.581-0.843) 0.714 (0.359-0.918) 0.000 (0.000-0.658) NPV (CI) (%) 0.654 (0.613-0.693) 0.724 (0.664-0.777) 0.616 (0.557-0.671) 0.672 (0.550-0.774) Romhilt-Estes Sen (CI) (%) 0.118 (0.081-0.168) 0.396 (0.304-0.496) 0.063 (0.031-0.124) ⁎ 0.095 (0.027-0.289) ⁎
Spec (CI) (%) 0.972 (0.950-0.985) 0.929 (0.883-0.958) 0.959 (0.918-0.980) 0.978 (0.884-0.996) PPV (CI) (%) 0.714 (0.549-0.837) 0.745 (0.611-0.845) 0.500 (0.268-0.732) 0.667 (0.208-0.939) NPV (CI) (%) 0.654 (0.613-0.693) 0.747 (0.687-0.799) 0.613 (0.554-0.670) 0.698 (0.576-0.798) Sen indicates sensitivity; Spec, specificity; PPV, positive predictive value; NPV, negative predictive value; CI, confidence interval.
much superior result than that seen in our material (r = 0.458 for the nonobese and 0.285 for the obese). Probably, these different values are reflected in the studied population. In our material, the population consisted exclusively of hyperten- sive patients; patients with valvar diseases are excluded from the sample. When the nonobese and obese were separated according to sex, we observed that specificity does not change with sex; however, for Sokolow-Lyon and Romhilt- Estes criteria, sensitivity presents a significant reduction (P b .05) in obesity both in men and women (Table 4).
The Sokolow-Lyon voltage and Romhilt-Estes criteria showed a greater decrease in the presence of obesity, which also happens when diagnostic sensitivity is analyzed. Sokolow-Lyon voltage, Romhilt-Estes criteria, and the strain pattern presented a significant sensitivity decrease when comparing both studied groups (P b .05). This change in Sokolow-Lyon voltage criteria, which predominantly assess QRS complex amplitude in precordial leads, may be explained in obese patients because thoracic and epicardial fat increases the impedance and lessens the QRS complex voltage.27In regard to Romhilt-Estes point score that takes into account QRS amplitude, obesity also has a significant influence. The strain pattern was less affected, although obesity may underestimate the degree of ST-segment depression, making its diagnosis difficult.
The low performance of Sokolow-Lyon voltage and Romhilt-Estes criteria comparatively to the Cornell voltage and Cornell duration criteria may be attributed to the sample population. It is known that hypertensive patients that constitute the sample of this study have a higher prevalence of concentric than eccentric LVH.28 The literature points out the better performance of Cornell criteria in this setting.29-32
What determines the different performances of the electro- cardiographic criteria is not clear. It is known that there are specific limitations for each particular criterion besides the demographic and clinical aspects of the studied sample.
Cornell voltage criteria, Cornell duration, Perugia, R in aVL, and the QTc interval variable did not show significant changes in sensitivity when compared between nonobese and obese patients.
It is worth to point out that Cornell voltage and Cornell duration criteria are derived from vectocardiographic analysis of LVH abnormalities. As a consequence of the hypertrophic process, the electric forces are more horizon- tally oriented. It means that their projection on aVL lead captures better the inscription of R wave. The same phenomenon is also posteriorly oriented, and its projection on the V3lead makes a better inscription of the S wave. The
V3lead is the precordial intermediate of LV and possibly less
affected by the distance changes between the myocardium and the exploring electrode on the thoracic surface. This should be the main reason for the steadfast performance of Cornell voltage and duration, Perugia, and RaVL criteria in nonobese and obese patients.
QTc interval showed no significant correlation with LVMI in the obese and nonobese patients. In contrast, Peng et al,33 studying hypertensive Chinese patients, found a significant correlation of heart mass with the QTc
interval. Probably, their findings express variability among different populations. It is widely known that ECG sensitivity and specificity for detecting LVH vary accord- ing to factors such as sex, race, and age. The population studied in this article include predominantly white people, few African descendants, and Brazilian mulattoes and is formed exclusively of hypertensive patients in regular treatment. Of course, studies in the field of aortic and mitral valve diseases and hypertrophy cardiomyopathy would be important for the generalization of findings in the obese population.
According to the findings of this study, obesity did not change ECG specificity. However, the ECG sensitivity presented a significant decrease using Sokolow-Lyon voltage, Romhilt-Estes criteria, and strain pattern. On the other hand, Cornell voltage and Cornell duration criteria, Perugia, and R in aVL showed no significant sensitivity decrease in obese patients.
References
1. Levy D, Labib SB, Anderson KM, et al. Determinants of sensitivity and specificity of electrocardiographic criteria for left ventricular hyper- trophy. Framingham Heart Study. Circulation 1990;81:815.
2. Prineas RJ, Rautaharju PM. Independent risk for cardiovascular disease predicted by modified continuous score electrocardiographic criteria for 6-year incidence and regression of left ventricular hypertrophy among clinically disease free men: 16-year follow-up for the multiple risk factor intervention trial. J Electrocardiol 2001;34:91.
3. Wong CY, O'Moore-Sullivan T, Leano R, et al. Alterations of left ventricular myocardial characteristics associated with obesity. Circula- tion 2004;110:3081.
4. Avelar E, Cloward TV, Walker JM, et al. Left ventricular hypertrophy in severe obesity. Hypertension 2007;49:34.
5. Horton JD, Sherber HS, Lakatta EG. Distance correction for precordial electrocardiographic voltage in estimating left ventricular mass: an echocardiographic study. Circulation 1977;55:509.
6. Abergel E, Tase M, Menard J, et al. Influence of obesity on the diagnostic value of electrocardiographic criteria for detecting left ventricular hypertrophy. Am J Cardiol 1996;77:739.
7. Rautaharju PM, Zhoiu SH, Calhoun HP. Ethnic differences in ECG amplitudes in North American white, black, and Hispanic men and women. Effect of obesity and age. J Electrocardiol 1994;27(Suppl):20. 8. Casalle PN, Devereux RB, Alonso DR, et al. Improved sex-specific criteria of left ventricular hypertrophy for clinical and computer interpretation of electrocardiograms: validation with autopsy findings. Circulation 1987;75:565.
9. Okin PM, Roman MJ, Devereux RB, et al. Electrocardiographic identification of increased left ventricular mass by simple voltage- duration products. J Am Coll Cardiol 1995;25:417.
10. Sokolow M, Lyon TP. The ventricular complex in left ventricular hypertrophy as obtained by unipolar precordial and limb leads. Am Heart J 1949;37:161.
11. Romhilt DW, Estes EH. A point-score system for the ECG diagnosis of left ventricular hypertrophy. Am Heart J 1968;75:752.
12. Schillaci G, Verdecchia P, Borgioni C, et al. Improved electrocardio-