The Electrocardiogram in Morbid ObesityMartin A. Alpert, MD, Boyd E. Terry, MD, Michael V. Cohen, MD,

T. Michael Fan, MD, PhD, Jack A. Painter, MD, and Clara V. Massey, MD

Obesity, particularly morbid obesity, is known toproduce changes in cardiac morphology such as

left ventricular (LV) enlargement, and eccentric LVhypertrophy and right ventricular hypertrophy.1,2

Thus, morbid obesity might also be expected to alterthe electrocardiogram. Prior studies assessing theelectrocardiogram in obesity have identified T-waveabnormalities in the inferior leads and trends towardleftward axis and low QRS voltage as the most com-mon alterations.3–5 These studies assessed a limitednumber of electrocardiographic (ECG) variables inpersons with variable degrees of obesity and did notuniformly exclude hypertensive patients or use a con-trol group. The present study provides a comprehen-sive survey of ECG variables in normotensive mor-bidly obese subjects and normal lean controls, anddetermines the frequency of established ECG abnor-malities in both groups.

• • •Morbid obesity was defined as actual body weight

equal to or more than twice the ideal body weightbased on 1993 Metropolitan Life Insurance tables.6

Morbidly obese patients were selected from thoseundergoing clinical evaluation for bariatric surgery.The control group was selected from a pool of leanhealthy normotensive subjects and was matched forage and gender. Lean body weight was defined asweight within the normal range based on 1993 Met-ropolitan Life Insurance tables.6 Neither the morbidlyobese nor the control groups had clinical, electrocar-diographic, radiographic, or echocardiographic evi-dence of systemic hypertension, coronary artery dis-ease, valvular heart disease, cardiomyopathy (unrelat-ed to obesity), pericardial disease, or congestive heartfailure. Those who were not in sinus rhythm at thetime of the index electrocardiogram, those with serumelectrolyte disturbances, and those receiving drugsthat might affect the electrocardiogram were excludedfrom the study. The study group consisted of 100morbidly obese patients and 100 controls. Patientcharacteristics of morbidly obese subjects and controlsare listed in Table I.

A 12-lead resting ECG was obtained in the supineposition using a standard technique with a Hewlett-Packard 1517 electrocardiograph (Andover, Massa-chusetts) with a filter setting of 100 Hz at speed of 25mm/s. Axes were measured by hand. All electrocar-diograms were interpreted by 1 investigator (MAA)on 3 separate occasions. The ECG interpreter was

blinded to patient identification and clinical data. Theintraobserver variability was,2% for all variables.M-mode and 2-dimensional echocardiograms wereobtained in the left lateral and supine positions usinga Hewlett-Packard Sonos 1000 ultrasonograph with a2.25-MHz transducer in accordance with AmericanSociety of Echocardiography recommendations.7

Echocardiographic measurements were obtained inaccordance with American Society of Echocardiogra-phy criteria.7 Echocardiographic LV mass/height in-dex was calculated using the formulas of Devereuxand Reichek8 and Levy et al.9 Normal ranges for LVmass/height index were,104 g/m in women and 120g/m in men.6

ECG abnormalities assessed are listed and definedin Table II. Cardiac arrhythmias and conduction dis-turbances were diagnosed using previously publishedcriteria.10,11ECG variables were classified as categor-ical or continuous. Categorical ECG variables used inthis study are listed in Table II. They were defined asECG abnormalities for which specific diagnostic cri-teria exist. Continuous ECG variables used in thisstudy are listed in Table III. They were defined asECG measurements of rate, duration, or amplitude, orratios that were not limited by specific diagnosticcriteria.

The chi-square test was applied to categorical datato determine if significant differences existed in thefrequency of ECG abnormalities between morbidlyobese and normal lean patients. The Student’st testwas used to determine if significant differences inmean values for specific continuous variables existedbetween morbidly obese and normal lean patients.Mean values are expressed6 1 SD. A p value,0.05was considered statistically significant.

• • •Table II shows the frequency of categorical vari-

From the Division of Cardiology, University of South Alabama, MedicalCenter, Mobile, Alabama; and the Department of Surgery, University ofMissouri, Columbia, Missouri. Dr. Alpert’s address is: University of SouthAlabama, Suite 10C, Medical Center, 2451 Fillingim Street, Mobile,Alabama 36617. Manuscript received July 28, 1999; revised manu-script received and accepted October 22, 1999.

TABLE I Patient Characteristics

Variables

MorbidlyObese

(n 5 100)

LeanControls

(n 5 100)

Women/men 85/15 85/15Mean age (yrs) 37 6 6 37 6 8Mean body weight (kg) 135 6 6 61 6 4Body mass index (kg/m2) 48 6 2 22 6 2Mean heart rate (beats/min) 79 6 4 78 6 6Systolic blood pressure (mm Hg) 128 6 4 126 6 3No. with increased LV mass

height index64 0

No. with RV free wall thickness$5 mm

2 0

No. with left atrial dimension.4.0 cm

15 0

RV 5 right ventricular.

908 ©2000 by Excerpta Medica, Inc. All rights reserved. 0002-9149/00/$–see front matterThe American Journal of Cardiology Vol. 85 April 1, 2000 PII S0002-9149(99)00894-2

ables in the morbidly obese and control groups. Categor-ical variables that occurred with significantly greaterfrequency in morbidly obese than in lean subjects in-cluded low QRS voltage, multiple ECG signs of LVhypertrophy, T-wave flattening in leads II, III, andaVF 6 V5 or V6, T-wave flattening in leads V5 or V6,T-wave flattening/inversion in leads I and aVL, andmultiple ECG criteria for left atrial abnormality. Allpatients with low voltage had normal LV mass/weightindex. There were no significant differences betweenmorbidly obese patients and controls in the followingvariables: premature complexes (atrial, atrioventricularjunctional, ventricular), sinus arrest or block, atrioven-

tricular block (first-, second-, andthird-degree), atrioventricular dissocia-tion, wandering pacemaker, and allforms of fascicular block.

Table III lists mean values of con-tinuous ECG variables in the mor-bidly obese and lean groups. P-,QRS, and T-wave axes were signifi-cantly more leftward in morbidlyobese than in lean subjects. The P-terminal force was significantly morenegative in morbidly obese than inlean subjects. Mean RaVL and SaVRand mean R/S ratio in lead V1 weresignificantly higher in morbidlyobese than in lean subjects.

• • •This study is the first comprehen-

sive survey of ECG variables andabnormalities in normotensive mor-bidly obese subjects.3–5,12–14Our re-sults confirm and extend several pre-vious observations.

In 1982, Eisenstein and col-leagues5 reported that P-, QRS, andT-wave axes were significantly moreleftward in obese (.50 lbs over-weight) than in normal patients. Ourresults confirm these observations.Although there is a leftward QRSaxis shift in normotensive morbidlyobese subjects, the frequency of left-(or right-) axis deviation is low. Pos-sible reasons for the leftward axisshift in morbidly obese subjects in-clude horizontal displacement of theheart due to abdominal adiposity andthe presence of LV hypertrophy.

Low QRS voltage has been fre-quently listed as an ECG abnormal-ity associated with obesity, but criti-cal evaluation of this variable is lack-ing.3,5 Eisenstein et al5 reportedminimally lower QRS voltage inobese than in normal patients. Al-though low voltage in the limb leadsoccurred with significantly higherfrequency in morbidly obese than inlean subjects in this study, the fre-

quency in the morbidly obese group was low. This isnot surprising in that nearly two thirds of morbidlyobese patients have LV hypertrophy.

ECG criteria for LV hypertrophy occurred withgreater frequency in morbidly obese patients than inlean controls. However, this is somewhat misleading.Given that almost two thirds of morbidly obese pa-tients in this study had echocardiographic LV hyper-trophy, the frequency range (9% to 23%) for variousECG criteria for LV hypertrophy was relatively low.Moreover, mean voltage was significantly higher inobese than in lean subjects for only 2 ECG variablesused to diagnose for LV hypertrophy, both of which

TABLE II Categorical Variables

ECG VariablesMorbidly Obese

(n 5 100)Lean Controls

(n 5 100) p Value

Sinus bradycardia* 1 3 NSSinus tachycardia† 0 0 NSSinus arrhythmia‡ 2 2 NSLow voltage§ 11 1 ,0.001Left-axis deviation\ 4 0 NSRight-axis deviation\ 2 0 NSSigns of LV hypertrophy

Romhilt-Estes $5 points 23 3 ,0.001SV1 1 RV5 or RV6 .35 mm 25 2 ,0.005RV5 or V6 .26 mm 19 2 ,0.001RaVL $11 mm 16 1 ,0.001RaVL $7.5 mm 16 2 ,0.001RaVF .20 mm 14 5 ,0.01(RI–RIII) 1 (SIII–SI) $17 mm 17 1 ,0.001SV1 .24 mm 9 0 ,0.01SaVR .14 mm 15 0 ,0.001OID in V5 or V6 .0.07 s 16 0 ,0.001

Signs of right ventricular hypertrophyR/S in V1 .1 5 0 NSR . in V5 or V6 ,1 5 0 NSRV1 1 SV5 or V6 .5 mm 5 0 NSrSR1 in V1 with R1 $10 mm 2 0 NSqR in V1 4 0 NSSV1 ,2 mm 6 1 NSRV1 .5 mm 0 0OID in V1 0.035–0.055 5 0 NSDecrease in R/S ratio between 2precordial leads to the right oftransition

0 0 NS

Left atrial abnormalitiesP-terminal force V1 $ 20.04 mm z s 18 0 ,0.001P-wave duration (II) .110 ms 24 5 ,0.025P-wave duration (II) .120 ms 17 1 ,0.01Interpeak notch (II) .40 ms 11 0 ,0.05

Right atrial abnormalityP amplitude .2.5 mm in lead II 2 0 NS

Repolarization abnormalitiesT-wave inversion (II, III, aVF) 4 1 NST-wave flattening (II, III, aVF) 58 2 ,0.0001T-wave inversion (V5, V6) 7 1 NST-wave flattening (V5, V6) 29 1 ,0.0001T-wave inversion (II, III, aVF, V5, V6) 7 1 NST-wave flattening (II, III, aVF, V5, V6) 44 1 ,0.0001T-wave flattening/inversion (I, aVL) 18 2 ,0.01ST depression (II, III, aVF, 6 V5, V6) 4 0 NSST depression and/or T-waveinversion in the right precordialleads

2 0 NS

*,60 beats/min; †.60 beats/min; ‡.10% difference between longest and shortest cycle in asequence; §QRS amplitude #5 mm in all standard leads; \axis deviation: left 5 230° to 290°; right 5

191° to 1180°.

BRIEF REPORTS 909

involved standard rather than precordial leads. Thus,LV hypertrophy is underdiagnosed by electrocardiog-raphy in morbidly obese subjects.

We suspect the excessive adipose accumulation inthe subcutaneous tissues of the chest wall (and possi-bly increased epicardial fat) reduces the QRS voltagesignal generated by the left ventricle. This wouldexplain the higher frequency of low QRS voltage inmorbidly obese patients than in lean controls as wellas the relatively low frequency of LV hypertrophydiagnosed with electrocardiography. We have previ-ously demonstrated low sensitivity for various ECGcriteria for LV hypertrophy.12 Others studies5,13,14re-ported a significantly lower mean spatial QRS,Solkolow index voltage, and RV5 amplitude, and ahigher RaVL amplitude in obese than in lean subjects.

T-wave flattening and inversion in the inferolateralleads of obese patients were originally described byJaffee et al in 1938.4 In a later study, Eisenstein et al5

reported T-wave inversion in lead III in 28% of obesesubjects and 21% of normal persons. Flattened Twaves in the inferolateral leads were noted in 49% ofobese patients and no normal subjects. The more de-tailed analysis of repolarization abnormalities in thisstudy indicates that T-wave flattening is much morecommon in the inferior, lateral, and inferolateral leadsof morbidly obese subjects than frank T-wave inver-sion. T-wave flattening in these leads is significantlymore common in morbidly obese subjects than in leancontrols. Although there was a trend toward morefrequent T-wave inversion in the inferolateral leads inmorbidly obese subjects than in controls, these differ-ences were not statistically significant. T-wave abnor-malities in the inferior and lateral leads have previ-ously been attributed to horizontal displacement of theheart. LV hypertrophy may also play a role.

Multiple criteria for left atrialabnormality occurred with signifi-cantly higher frequency in mor-bidly obese than in lean subjects inthis study. Moreover, P-terminalforce was significantly more nega-tive in morbidly obese than in leanpatients. Lavie et al13 previouslyanalyzed P-terminal force in nor-motensive and hypertensive obeseand lean subjects and reported sim-ilar findings. The low frequency ofechocardiographic left atrial en-largement suggests that left atrialhypertension and/or atrial conduc-tion abnormalities may account forthe relatively high frequency ofECG criteria for left atrial enlarge-ment in the morbidly obese.

There were no significant differ-ences in the frequencies of cardiacarrhythmias and conduction distur-bance between obese and lean pa-tients (all of whom had a sinus mech-anism).

In summary, P-, QRS, and T-wave axes aremore leftward in morbidly obese patients than innormal controls. Low QRS voltage, multiple ECGcriteria for LV hypertrophy and left atrial abnor-mality, and T-wave flattening in the inferior andlateral leads occur with significantly greater fre-quency in morbidly obese subjects than in leancontrols.

1. Amad RH, Brennan JC, Alexander JK. The cardiac pathology of chronicexogenous obesity.Circulation 1965;32:740–745. 2. Warnes CA, Roberts WC.The heart in massive (more than 300 pounds or 136 kiligrams) obesity: analysisof 12 patients studied at necropsy.Am J Cardiol1985;54:1087–1091.3. Proger SH. The electrocardiogram in obesity.Arch Intern Med1931;47:64–69.4. Jaffee HD, Corday E, Master AM. Evaluation of the precordial leads of theelectrocardiogram in obesity.Arch Intern Med1938;36:911–917.5. Eisenstein I, Edelstein J, Sarma R, San Marco M, Selvester R. The electro-cardiogram in obesity.J Electrocardiogr1982;15:115–118.6. Metropolitan Life Insurance Tables, Metropolitan Life Insurance Co., NewYork, 1993.7. Sahn, D, DeMaria A. Kisslo J, Weyman A. Recommendations regardingquantitation in M-mode echocardiography.Circulation 1978;58:1072–1083.8. Devereux RB, Reichek N. Echocardiographic determination of left ventricularmass in man: anatomic validation of the method.Circulation 1977;55:613–618.9. Levy D, Savage DB, Ganison RJ, Anderson KM, Kannel WB. Echocardio-graphic criteria for left ventricular hypertrophy. The Framingham Study.Am JCardiol 1987;54:956–960.10. Scheidt S. Diagnosis of cardiac rhythm. In: Scheidt S, ed. Basic Electrocar-diography. West Caldwell, NJ: Ciba-Geigy Pharmaceuticals,1993:21–38.11. Dunn MI, Lipman BS. Cardiac arrhythmias, parts II-V. In: Dunn MI, LipmanBS, eds. Clinical Electrocardiography. 8th ed. Chicago: Year Book MedicalPublishers, 1989:345–443.12. Nath A, Alpert MA, Terry BE, Kelly DL. Sensitivity and specificity ofelectrocardiographic criteria for left and right ventricular hypertrophy in morbidobesity.Am J Cardiol1988;62:126–130.13. Lavie CJ, Amodeo C, Ventura HO, Messerli FH. Left atrial abnormalitiesindicating diastolic dysfunction in cardiopathy of obesity.Chest1987;92:1042–1046.14. Rautaharju PM, Zhou SH, Calhoun HP. Ethnic differences in ECG ampli-tudes in North American white, black, and hispanic men and women. Effect ofobesity and age.J Electrocardiogr1994;127:20–31

TABLE III Continuous Variables

ECG Variables

MorbidlyObese

(n 5 100)Lean Controls

(n 5 100) p Value

Heart rate (beats/min) 79 6 5 78 6 4 NSP-wave axis (°) 33 6 16 46 6 14 ,0.05QRS axis (°) 28 6 14 52 6 10 ,0.01T-wave axis (°) 35 6 17 48 6 18 ,0.05P-wave duration (ms) 108 6 9 103 6 6 NSP-terminal force

(mm z s)20.016 6 0.008 20.0003 6 0.0001 ,0.01

PR interval (s) 0.15 6 0.03 0.14 6 0.02 NSQRS duration (s) 0.06 6 0.02 0.06 6 0.01 NSQTc interval (s) 0.37 6 0.03 0.36 6 0.03 NSSV1 1 RV5 or V6 (mm) 24 6 7 18 6 5 NSRaVL (mm) 7 6 2 4 6 2 ,0.01SaVR (mm) 8 6 2 6 6 1 ,0.05SV1 8 6 5 7 6 3 NS(RI–RIII) 1 (SIII–SI) (mm) 7 6 2 8 6 1 NSOID V1 or V2 (sec) 0.04 6 0.02 0.03 6 0.01 NSR/S in V1 or V2 (max) 0.20 6 0.04 0.12 6 0.03 ,0.01R/S in V5 or V6 (max) 0.10 6 0.03 0.12 6 0.02 NSRV1 1 SV5 (mm) 1 6 2 2 6 1 NSRV1 (mm) 1 6 2 1 6 1 NS

Abbreviation as in Table II.

910 THE AMERICAN JOURNAL OF CARDIOLOGYT VOL. 85 APRIL 1, 2000