J. ELECTROCARDIOLOGY 20(3), 1987, 227-232

Electrocardiographic Diagnosis of Left Ventricular Hypertrophy in the Presence of Left

Bundle Branch Block BY RICHARD J. HASKELL, M.D., LEONARD E. GINZTON, M.D. AND MICHAEL M. LAKS, M.D.

S U M M A R Y

The electrocardiographic diagnosis of LVH in the presence of LBBB has previously been difficult. Thirty-seven patients with complete LBBB were identified and had echocardiography performed. Using an accepted echocardiographlc formula, left ventricular mass was calculated. Twenty of the 37 patients (54%) were classified as having severe LVH. Multiple conventional ECG criteria for LVH were then evaluated. No QRS voltage criteria showed any difference between patients with and without LVH (P = NS). There was also no correlation between either QRSaxis or left atrial enlargement and left ventricular mass (P = NS). However, the QRS duration was significantly longer in the patients with LVH (160 + 12 msec) than in the normal patients (148 _+_ 11 msec) (P < 0.001). The sensitivity, specificity, positive predictive value, and accuracy of several voltage criteria and QRS duration were examined. The best voltage criteria had a sensitivity of only 50% and a predictive value of 63%. However, a QRS duration >155 msec had a sensitivity of 60% and a predictive value of 82%. This study demonstrates that the conventional QRS voltage criteria for LVH are not accurate in LBBB. A relationship exists between increasing QRS duration in LBBB and LVH; therefore, the relative probability adjectives: "consider," "possible," and "probable" should be used. ORS duration > 155 msec is predictive of LVH despite the presence of LBBB.

Left ventricular hypertrophy (LVH) has long been an important diagnosis that can be made using a 12-lead electrocardiogram (ECG). However, the presence of left bundle branch block (LBBB) had previously invalidated the accepted criteria for LVH. The important diagnosis of LVH is often missed on physical examination and routine chest roentgenogram. Since the ECG is part of the routine clinical evaluation of most cardiac patients, a sus- picion of LVH based on ECG criteria would be useful. Several studies have attempted to define criteria for LVH in the presence of LBBB in au- topsy patients but have had little success. 1-4 More recently, echocardiographic criteria for the diag- nosis of an increased left ventricular (LV) mass

From the Division of Cardiology, Department of Medicine, Har- bor-UCLA Medical Center, Torrance, CA and Hoag Memorial Hospital, Newport Beach, CA.

Reprint requests to: Michael M. Laks, M.D., Division of Car- diology, Harbor-UCLA Medical Center, 1000 W. Carson, Tor- rance, CA 90509.

have been developed and validated. ~ Using this echocardiographic approach, Kafka and co-workers developed four unrelated ECG parameters for the diagnosis of LVH, which, when used in combina- tion, yield a satisfactory sensitivity and specificity. 7 The purpose of the present study was to develop simpler and more accurate ECG criteria for LVH in the presence of LBBB and to relate the criteria to echocardiographic LV mass. Multiple ECG mea- surements in patients with LBBB were correlated with LV mass in order to develop a criterion for the diagnosis of LVH.

MATERIALS AND METHODS Study Population

All ECGs done at Harbor-UCLA Medical Center are stored on computer disc. Over a one-year period, all ECGs with complete LBBB were revie~ved and 37 patients were identified who met the following criteria: Alive and avail- able for an echocardiogram, LBBB on the ECG on the day of the echocardiogram, and a good quality M-mode echocardiogram could be obtained in which septal and posterior wall edges were easily defined. No cases of first or second degree LBBB were included since these may

227

228 HASKELL ET AL

TABLE ! Electrocardiographic Criteria in Patients with Bundle Branch Block: Non-LVH Group

SV1 + PT QRSd SV1 RV5/V6 RV5/V6 AXIS LVlD PWT IVST LVM

1 131 35 12 47 - 4 6 4.3 0.8 0.7 106 2 133 21 8 29 2 4.8 0.8 0 .7 131 3 140 25 15 40 - 7 5.0 0.8 0.7 142 4 153 30 7 37 -21 4.8 0.8 1.1 184 5 147 32 16 48 11 3.1 1.6 1.4 191 6 147 25 3 28 - 5 7 4.0 1.2 1.2 192 7 148 20 8 28 63 4.0 0.9 1.6 205 8 150 22 8 30 1 5.1 1.0 1.0 220 9 163 22 10 32 4 6.7 0.7 0.7 226

10 147 17 4 21 - 5 3 6.2 0.8 0.8 232 11 164 20 3 23 - 4 6 7.2 0.6 0.7 237 12 133 10 6 16 - 4 2 5.8 1.0 0.8 240 13 137 23 12 35 -41 5.6 0.9 1.0 242 14 154 27 27 54 - 3 5 5.4 1.2 0.8 244 15 153 27 9 36 9 5.2 1.1 1.0 244 16 170 16 9 25 - 6 9 6.2 0.8 0.9 251 17 152 23 7 30 - 2 7 6.3 0.6 1.2 279 Mean 148 23 10 33 -21 5.3 0.9 1.0 210 SD 11.2 6.1 5.8 10.1 33.3 1.1 0.1 0.3 47

Table I, Continued LVH Group

18 164 26 11 37 9 5.2 1.0 1.4 296 19 147 19 4 23 69 5.6 1.2 1.0 297 20 149 36 14 50 95 7.5 1.0 0.5 305 21 158 24 19 43 23 4.6 1.5 1.3 306 22 t65 20 5 25 - 5 6 5.6 1.0 1.3 316 23 149 26 10 36 - 4 6.5 1.0 0.9 317 24 147 22 11 33 - 3 4 6.0 1.2 1.0 335 25 165 22 6 28 - 6 3 6.8 0.7 1.2 344 26 166 4 6 10 0 7.7 0.8 0.8 348 27 154 28 5 33 63 5.4 1.3 1.3 355 28 147 17 20 37 78 5.6 1.4 1.1 356 29 185 24 20 44 - 3 2 6.0 1.3 1.0 356 30 169 22 16 38 - 2 5 6.8 1.1 1.0 392 31 155 25 8 33 85 6.2 1.2 1.2 400 32 168 25 12 37 - 3 0 8.4 1.0 0.7 441 33 147 10 11 21 - 6 2 6.6 0.8 1.6 445 34 162 6 9 15 4 7.7 1.2 0.8 460 35 153 35 16 51 1 6.8 1.3 1.1 469 36 180 11 5 16 151 6.8 1.3 1.2 496 37 176 22 20 42 42 10.5 1.09 0.8 717

Mean 160 21 11 33 16 6.6 1.1 1.1 387 SD 11.6 8.4 5.5 11.4 59.0 1.3 0.1 0.3 100

IVST = interventricular septal thickness (in cm); LVID = left ventricular internal diameter(in cm); LVM = left ventricular mass (in g); PT = patient; PWT = posterior wall thickness (in cm); QRSd = QRS duration (in msec); RV5/V6 = height of the largest R wave in V5 or V6 (in mV); SD = standard deviation; SV1 = depth of the S wave in V1 (in mV).

actually represent cases of LVH rather than bundle branch disease, s Of these 37 patients, 62% were women. The mean age was 60.7 _ 12.0 years (range 22-95 years). The diagnosis of the patients was: 51% with idiopathic or dilated cardiomyopathy, 24 % with valvular disease, 16 %

with ischemic disease, and 5% with hypertensive heart disease. Only one patient was clinically normal.

Nineteen patients, who had an echocardiogram done during a one-month period and who had no known car- diac disease on physical examination, ECG, chest x-ray,

J. ELECTROCARDIOLOGY 20(3), 1987

LVH IN LEFT BUNDLE BRANCH BLOCK 229

TABLE II Sensitivity, Specificity, Positive Predictive Value, and Accuracy of Selective Electrocardiographic Criteria

in the Diagnosis of Left Bundle Branch Block

Criteria Sens Spec Pred Value Accuracy

SVl + RV5/V6 > 3.5 mV RV5/V6 > 2.6 mV SV1 or SV2 > 3.0 mV RV5 or RV6 > 3.0 mV

Mean QRS axis < - 3 0 Mean QRS axis < 0

QRSd > 150 msec QRSd > 155 msec QRSd > 160 msec

50% 64% 63% 57% 0 94 0 16

10 82 40 43 0 100 0 46

30 47 43 38 40 64 42 46

70 65 70 68 60 82 80 70 55 82 79 68

Pred Value = predictive value; QRSd = QRS duration (in msec); RV5 = height of the R wave in V5 (in mY); RV6 = height of the R wave in V6 (in mV); RV5/V6 = height of the largest R wave in V5 or V6 (in mV); Sens = sensitivity; Spec = specificity; SVl = depth of the S wave in V l (in mV); SV2 = depth of the S wave in V2 (in mV).

or echocardiogram were used to define our normal pop- ulation for LV mass. Most of these patients were being evaluated for such non-specific conditions as dizziness, palpitations, or atypical chest pain.

ECG Criteria

All electrocardiograms were recorded using Hewlett- Packard multichannel recorders. LBBB criteria were 1) QRS duration > 120 msec; 2) a predominantly negative complex in V, and a predominantly upright, wide, and slurred R wave in V5 and Vs; 3) mid-conduction delay defined as notching or a plateau in the mid-QRS; 4) ven- tricular activation time (intrinsicoid deflection) > 50 msec after the onset of the QRS. The R and S waves were measured to the nearest 0.1 mV (averaging, if necessary). QRS durations were measured by the computer to the nearest msec. The mean QRS axis in the frontal plane using the area method was also generated to the nearest degree by the computer. In all cases the largest voltage and the longest duration for several complexes were av- eraged.

Echocardiography

All echocardiograms were performed on either a com- mercially available Hewlett-Packard Model 77020A with a 3.5 mHz transducer or an ATL Mark III with a 3.0 mHz transducer. Measurements were taken from the M-mode paper recordings and the LV mass was calculated using the Penn method. ~

Statistical Methods

The unpaired T-test was used to compare differences between normal and LVH patients. Linear regression analysis was used to correlate ECO diagnoses of LVH with the number of patients defined by echocardiogram as having LVH. Specificity was the number of negative ECO diagnosis/number of patients not having LVH echo- cardiographically. Positive predictive value was the num-

ber of true positive diagnoses/true positives plus false positives. The positive predictive value reflects the degree to which a positive test signifies disease. Accuracy was the number of correct diagnoses/number of patients. All values are expressed as mean + standard deviation. A P < 0.05 was considered statistically significant.

RESULTS Th e LV mass calculated in the 19 normal pat ients

was 171 + 55 g (range 85-299 g). The mean plus two s tandard deviations in this normal populat ion using echocardiographic cri ter ia was 281 g. An LV mass >281 g was defined as LVH.

Thi r ty -seven pat ients had sat isfactory 12-lead ECGs and echocardiograms per formed within 24 hours of each other. Using 281 g as an upper l imit of normal, 17 pat ients (46 %) were classified as hav- ing normal LV mass (non-LVH group) and 20 pa- t ients (54%) were classified as having LV H (LVH group) based on echocardiographic measurements (Table I). T h e use of LV mass index (LV mass divided by body surface area) did not change the distr ibution between normal and abnormal LV mass and was, therefore, not used in fur ther analysis. The average LV mass of all the L B B B patients was 387 g • 100 g and was significantly above the average of 171 g • 55 g in our normal populat ion (P < 0.001). T h e magni tude Of the S in VI (SV1), t h e largest R in V5 or VG (RV5/V6), and the sum of SV1 • RV5/V6 are presented in Table I.

Many voltage criteria were compared to assess their abili ty to correct ly diagnose LVH. Some, bu t not all, of these cri teria are shown in Table II and include the most popular and probably the most accurate Sokolow-Lyon c r i t e r i a (SV1 + RV5 or RV6 > 35 mm)241 T h e poin t system of Romhil t -

J. ELECTROCARDIOLOGY 20(3), 1987

230 H A S K E L L ET A L

BOO

+

p<.O01 eO0 is .61

400 4-

>, ~ +

,,'o ,Io ,,o' ,;o ,1o ,.o' ,,o ORS D u r a t i o n (mser

Fig. 1. Correlation between QRS duration and left ven- tricular mass as calculated from an echocardiographic formula in 37 patients with left bundle branch block.

Estes also did not accurately predict LVH? 2 No voltage criteria had significant correlation with LV mass; the Sokolow-Lyon criteria had very poor cor- relation (r = 0.006, P = NS).

The mean frontal plane axis was statistically dif- ferent when comparing non-LVH patients ( - 2 1 __ 33 degrees) with LVH patients (16 _ 59 degrees) (Table I). However, the correlation of LV mass with axis was not significant (r = 0.35, P = NS), reflect- ing the wide scatter amongst the two groups. The sensitivity and specificity of a leftward QRS axis as criteria for LVH were also very low (30% and 47 %, respectively), with a low predictive value (Table II). The presence or absence of left atrial enlargement was also not predictive of LVH.

The QRS duration (QRSd) was significantly dif- ferent when the LBBB patients without an in- creased LV mass (non-LVH) were compared with the LVH patients (P < 0.01). The mean QRSd in the non-LVH patients with LBBB was 148 __ 11 msec and the QRSd in the LVH patients was 160 _ 12 msec (Table I). In addition, there was a signif- icant correlation when the QRSd was compared to the LV mass (r = 0.61, P < 0.0001 in Fig. 1). When the sensitivity and specificity of different QRSd were calculated (Table II), the highest sensitivity, but lowest specificity, occurred at a QRSd of > 150 msec. In contrast, a QRSd of >160 msec created lower sensitivity (still higher than any voltage cri- teria), but a higher specificity. An intermediate QRSd of > 155 msec yielded intermediate sensitiv- ity and specificity, but the highest positive predic- tive value (Table II).

DISCUSSION Left bundle branch block has been an electro-

cardiographic diagnosis that has precluded accu- rate diagnosis of LVH. Previous autopsy studies

have shown that patients with LBBB have much larger hearts than patients without LBBB. z-4 Fifty- four percent of our patients had eehocardiographic LVH and the average LV mass of all of our LBBB patients was considerably larger than normal.

Although there is a need for an accurate ECG diagnosis of LVH despite LBBB, previous voltage criteria for LVH are inaccurate in the presence of LBBB. This is best shown in instances when the same patient who meets ECG criteria for LVH sud- denly develops LBBB. Cokkinos 13 reported in cases of intermittent LBBB that there was a significant decrease in the voltage of most of the leads when the ECG changed from LVH to LVH with LBBB. ZmyslinzkP showed that patients have lower QRS voltages after developing LBBB compared to their ECGs prior to the onset of LBBB. In fact, in that study most patients met criteria for LVH before the LBBB, but then failed to meet voltage criteria for LVH after the LBBB developed.

Despite studies that have shown no or even re- versed correlation between ECG voltage criteria and LVH in LBBB, investigators have continued to at- tempt to develop diagnostic criteria for LVH in the presence of LBBB. Pantr idge" reported that there was a good relationship between increased voltage and cardiac enlargement despite the presence of LBBB on the ECG. However, his assessment of LV size was based on clinical grounds only, and not on pathological or non-invasive data. Scott and Norris z reported that 17 of 29 autopsy cases (59 %) of LVH and LBBB met conventional voltage criteria for LVH. However, all of their LBBB cases were con- sidered to have anatomic LVH, which is much higher than any other autopsy series. This may reflect their possible inclusion of incomplete LBBB, which may represent LVH rather than true LBBB.

The majority of studies have continued to show no clinically useful relationship between ECG volt- age and LVH in the presence of LBBB. In 1971 an autopsy series of 50 patients with LBBB showed that only 14 of the 40 patients (35%) with anatomic LVH could be diagnosed by any ECG criteria as having LVH3 Another autopsy series in 1980 with 43 patients with LBBB showed that only 34-54% of the LVH patients Could be diagnosed using one of several conventional LVH criteria based on volt- age. 3 The largest autopsy series (1982) with 70 LBBB patients concluded that any voltage criteria poorly predicted the presence of LVH, with most sensi- tivities in the 30-50% range.*

Three recent series used the echocardiographic diagnosis of an increased L V mass instead of au- topsy data. NobeP 5 reported very low sensitivities

J. ELECTROCARDIOLOGY 20(3), 1987

LVH IN LEFT BUNDLE BRANCH BLOCK 231

in 28 LBBB patients for all voltage criteria tested (e.g., sensitivity of 37% for SV1 + RV5/V6 > 3.5 mV). In no case was any accuracy greater then 55 %. Left atrial enlargement was their most sensitive and specific predictor of LVH. QRSd was not reported. Kafka v reported that no single voltage criterion was highly sensitive and specific for predicting echo- cardiographic LVH. However, when three different voltage criteria and an axis criterion were used in a "cumulative fashion," there was a higher sensi- tivity (73%) and specificity (66%). The clinical use- fulness of such a complex formula is unclear. They also showed the highest sensitivity of any single ECG criterion (68%) for QRSd > 160 msec for the diagnosis of LVH, although their specificity was low (25 % ). Klein TM used a combination of LV mass, wall thickness, or previous LVH on ECG prior to the development of LBBB to define LVH. They showed that precordial voltage, QRSd, and left atrial en- largement were all highly correlated with the pres- ence of LVH despite the presence of LBBB.

In the present study there was no correlation between any voltage criterion and LV mass. When the sensitivity and specificity of the best voltage criteria are calculated, there is little justification for the use of voltage criteria for the diagnosis of LVH (Table I). The best voltage criteria were those of Sokolow and Lyon, 9 yet the sensitivity was only 50 %. This agrees with previous electrocardiograph- ic beliefs that voltage criteria for LVH are invalid in the face of LBBB.

The concept that QRSd may correlate with LVH has been proposed by others. Selvester ~7 has used a computer simulation of ventricular depolarization based on electrophysiologic and anatomic princi- ples to show that a prolongation of the QRSd be- yond 160 msec is presumptive evidence for LVH in the presence of LBBB. He also showed in this the- oretical model that the severity of the LVH is a function of the QRSd. This computer model has been validated in this present study which showed that LVH is likely to occur with LBBB if the QRSd is greater then 155-160 msec and that the degree of LVH is correlated with the QRSd.

This study revealed that there is a significant correlation between LV mass and QRSd. Since the spectrum of LVH is a continuum, the use of an "artificial" cut-off point to define the presence or absence of LVH is somewhat arbitrary. The process of calculating a sensitivity and specificity requires that there be a definition of "normal" for both the disease being tested and the test results. The re- ciprocal relationship between sensitivity and spec- ificity can be manipulated by varying the definition

of these normal ranges. For example, decreasing the cut-off of a normal LV mass to 215 g, as was done by others, would have increased the specificity of a QRSd of >155 msec to 100%, but decreased the sensitivity to 50%. Conversely, using 281 g as a definition of increased LV mass but decreasing the minimum QRSd from 155 msec to 150 msec would have increased the sensitivity to 70% but decreased the specificity to 65%. Therefore, because these data are continuous, a correlation is probably more meaningful than sensitivity and specificity. A prac- tical translation to clinical diagnosis of continual data as presented is to use the adjectives "consid- er," "possible" and "probable" in front of the di- agnosis of left ventricular hypertrophy to indicate the different degrees of probability. For example, a QRSd of 150-155 msec would mean "consider LVH," a QRSd of 156-160 msec would mean "pos- sible LVH," and a QRSd of > 160 msec would in- dicate "probable LVH." However, "criteria" for di- agnosis are developed from sensi t ivi ty and specificity data. Using a cut-off of 155 msec to di- agnose LVH, these criteria yield a sensitivity of 60% and a specificity of 80%. This compares very favorably with previous studies which showed a best sensitivity and specificity for LVH in patients with- out LBBB to be 56% and 90%, respectively, l~

The majority of the patients in this study had a cardiomyopathy as the primary diagnosis. This may be due to the fact that LBBB may be a marker of disordered conduction often seen in cardiomyopa- thies or ischemic heart disease. ~s The fact that there was a correlation between QRSd and LV mass would seem to bear this out. This study also used a very strict definition of LBBB. This purposely excludes many patients with QRSd of 120 to 140 msec since these patients probably have LVH with the char- acteristic delayed conduction and not LBBB. 8

Conclusion: This study reaffirms that voltage cri- teria are invalid in the face of LBBB. However, QRS duration correlates with LV mass even in the face of LBBB. The severity of LVH is related to the degree of QRS prolongation. LVH should be strong- ly considered in the presence of LBBB if the QRS duration is greater than 155 msec, although there still may be a number Of patients who will not be correctly diagnosed. However, the QRS duration is much better at correctly identifying the presence or absence of LVH than any voltage criteria.

REFERENCES

1. SCOTT, R C ANb "NORRIS, R. J: Electrocardiographic- pathologic correlation study of left ventricular hy-

J. ELECTROCARDIOLOGY 20(3), 1987

232 HASKELL ET AL

pertrophy in the presence of left bundle-branch block. Circulation 20:766, 1959

2. PETERSEN, G V AND TmOFF, G: Left bundle branch block and left ventricular hypertrophy: electrocar- diographic-pathologic correlations. Chest 59:174, 1971

3. ZMYSLINSKI, R W, RICHESON, J F AND AKIVAMA, T: Left ventricular hypertrophy in presence of complete left bundle-branch block. Br Heart J 43:170, 1980

4. HAVELDA, C J, SOHI, G S, FLOWERS, N C AND HORAN, L G: The pathologic correlates of the electrocardio- gram: complete left bundle branch block. Circulation 65:445, 1982

5. DEVEREUX, R B AND REICHEK, N: Echocardiographic determination of left ventricular mass in man: ana- tomic validation of the method. Circulation 55:613, 1977

6. REICHEK, N AND DEVEREUX, R B: Left ventricular hypertrophy: relationship of anatomic, echoeardio- graphic and electrocardiographic findings. Circula- tion 63:1391, 1981

7. KAFKA, H, BURGGRAF, G ~V AND MILLIKEN, J A: Elec- trocardiographic diagnosis of left ventricular hyper- trophy in the presence of left bundle branch block: an echocardiographic study. Am J Cardiol 55:103, 1985

8. SODI-PALLARES, D, ESTANDIA, A, SOBERON, J AND RODRIOUEZ, M I: The left intraventricular potential of the human heart: II. Criteria for diagnosis of in- complete bundle branch block. Am Heart J 40:655, 1950

9. SOKOLOW, M AND LYON, T P: The ventricular com- plex in left ventricular hypertrophy as obtained by unipolar precordial and limb leads. Am Heart J 37: 161, 1949

I0. ROMHILT, D l~V, BOVE, K E, NORRIS, R J, CONYERS, E, CONRADI, S, ROWLANDS, D T AND SCOTT, R C: A critical appraisal of the electrocardiographic criteria

for the diagnosis of left ventricular hypertrophy. Cir- culation 40:185, 1969

II. KOSSMANN, C E, BURCHELL, H B, PRUH'F, R D AND SCOTT, R C: The electrocardiogram in ventricular hypertrophy and bundle-branch block: a panel dis- cussion. Circulation 26:1337, 1962.

12. ROMHILT, D ~,V AND ESTES, E H: A point-score system for ECG diagnosis of left ventricular hypertrophy. Am Heart J 75:752, 1968

13. COKKINOS, D V, DEMOPOULOS, J N, HEIMONAS, E T, MALLIOS, C, PAPAZOGLOU, N AND VORIDES, E M: Elec- trocardiographic criteria of left ventricular hypertro- phy in left bundle-branch block. Br Heart J 40:320, 1978

14. PANTRIDGE, J F: Observations on the electrocardio- gram and the ventricular gradient in complete left bundle branch block. Circulation 3:589, 1951

15. NOBEL, L M, HUMPHREY, S B AND MONAGHAN, G B: Left ventricular hypertrophy in left bundle branch block. J Electrocardiol 17:157, 1984

16. KLEIN, R C, ZAKAUDDIN, V, DEMARXA, A N AND MASON, D T: Electrocardiographic diagnosis of left ventric- ular hypertrophy in the presence of left bundle branch block. Am Heart J 108:502, 1984

17. STARTT/SELVESTER, R H AND SOLOMON, J C: Com- puter simulations of ventricular depolarization, QRS duration and quantitation of wall thickness. In Com- puter ECG Analysis: Towards Standardization. Pro- ceedings of the IFIP-IMIA Working Conference on Computer ECG Analysis, J L Willems, J H Van- Bemmel and C Zyweitz, Eds, Elsevier-North-Hol- land, Amsterdam, 1986, pp 261-266

18. VASSALLO, J A, CASSIDY, D M, MARCHLINSKI, F E, BUXTON, A, I, VAXMAN, H L, DOHERTY, J U AND JOSEPHSON, M E: Endocardial activation of left bun- dle branch block. Circulation 69:914, 1984

J. ELECTROCARDIOLOGY 20(3), 1987