124
JACC Vol. 27, No. 1
January 1996:124-31
Electrocardiographic Identification of Left Ventricular Hypertrophy:
Test Performance in Relation to Definition of Hypertrophy and
Presence of Obesity
P E T E R M. OKIN, MD, FACC, M A R Y J. R O M A N , MD, FACC,
R I C H A R D B. D E V E R E U X , MD, FACC, P A U L K L I G F I E L D , MD, F A C C
New York, New York
Objectives. This study sought to assess a test performance of the
electrocardiogram (ECG) in relation to 1) varying definitions of
left ventricular hypertrophy based on different methods of adjusting left ventricular mass for body size, and 2) the presence or
absence of obesity.
Background. Although left ventricular mass is most commonly
indexed for body surface area or height when defining left
ventricular hypertrophy, recent work suggests that normalization
for height to the power of 2.7 (height z'7) may decrease variability
among normal subjects and correctly identify the impact of
obesity on hypertrophy.
Methods. The product of Cornell voltage and QRS duration
(Cornell product) and Framingham-adjusted Cornell voltage were
determined from 12-lead ECGs in 212 patients. Left ventricular
hypertrophy was defined on the basis of left ventricular mass
indexed to body surface area, height and height 2"7.
Results. Using partitions with matched specificity of 95%, the
sensitivity of ECG criteria varied with the definition of hypertrophy, ranging from 39% to 52% for the Cornell product and from
24% to 33% for adjusted Cornell voltage. When left ventricular
mass was indexed to body surface area or to height 2'7, the 52% and
39% sensitivities of the Cornell product were significantly greater
than the 24% (p < 0.001) and 29% (p < 0.05) sensitivities of
adjusted Cornell voltage, with a similar trend when left ventricular mass was indexed to height (43% vs. 33%, p = 0.10).
Comparison of receiver operating characteristic curves confirmed
the superior overall performance of the Cornell product relative to
adjusted Cornell voltage for hypertrophy defined by body surface
area and height 2"7 and demonstrated greater reproducibility of
overall performance, as measured by the coefficient of variability,
for the Cornell product (1.7%) than for adjusted Cornell voltage
(5.8%). Sensitivity of adjusted Cornell voltage was significantly
greater in obese than in nonobese subjects (50% to 59% vs. 18% to
24%, p < 0.01), but the Cornell product had only minimally higher
sensitivity in nonobese than in obese subjects (40% to 54% vs. 32%
to 44%, p = NS).
Conclusions. The ability of ECG criteria to detect left ventricular hypertrophy differs depending on the method of indexing left
ventricular mass for body size and with the presence or absence of
obesity. Further, the Cornell product provides the best combination of overall accuracy and low variability of performance
between definitions of hypertrophy. These findings have important implications for the clinical and epidemiologic use of 12-lead
ECG criteria for the detection of left ventricular hypertrophy.
(J Am Coil Cardiol 1996;27:124-31)
The presence of increased left ventricular mass detected by
echocardiography has been associated with increased risk of
cardiovascular morbidity and mortality among members of the
general population (1,2) and in patients with hypertension
(3,4) or coronary artery disease (5), making accurate and
cost-effective detection of increased left ventricular mass a
clinical priority (6). Although standard 12-lead electrocardiographic (ECG) evidence of left ventricular hypertrophy is also
associated with an increased risk of adverse cardiovascular
outcome (7-10), accurate, widespread utilization of the ECG
for the detection of hypertrophy has been limited by the poor
sensitivity at acceptable levels of specificity of current ECG
criteria (11-20). However, new approaches to analysis of the
12-lead ECG based on adjustment of QRS amplitude for age
and obesity (18) or on an approximation of the area under the
QRS complex (19,20) may increase test accuracy.
Although left ventricular mass is most commonly indexed to
body surface area or height when defining left ventricular
hypertrophy (1-3,21-24), recent work suggests that normalization for height to the 2.7 power (height 27) decreases variability
and reduces the impact of obesity on determination of hypertrophy (24) and may improve the prediction of adverse prognosis (25). Because quantitative measurements of left ventricular mass at autopsy or by echocardiography are the standards
against which performance of the ECG for detection of
hypertrophy is measured, it is apparent that accuracy of the
ECG for detection of left ventricular hypertrophy may vary
From the Division of Cardiology, Department of Medicine, The New York
Hospital-Cornell Medical Center, New York, New York.
Manuscript received March 29, 1995; revised manuscript received July 11,
1995, accepted August 10, 1995.
Address for correspondence: Dr. Peter M. Okin, The New York HospitalCornell Medical Center, 525 East 68th Street, New York, New York 10021.
©1996 by the American College of Cardiology
0735-1097/96/$15.00
0735-1097(95)00421-1
JACC Vol. 27, No. 1
January 1996:124-31
OKIN ET AL.
E L E C T R O C A R D I O G R A P H I C IDENTIFICATION O F H Y P E R T R O P H Y
both according to the ECG criteria used and how left ventricular mass is indexed to body size in the definition of hypertrophy. Therefore, the present study was designed to test the
performance and the variability of performance of ECG
criteria in relation to varying definitions of left ventricular
hypertrophy based on different methods of indexing left ventricular mass for body size and to assess the effect of obesity on
ECG sensitivity.
Methods
Study group. Cases were selected from review of the
autopsy log of the Hospital of the University of Pennsylvania
and of The New York Hospital-Cornell Medical Center, as
previously described (19). Complete clinical data and 12-lead
ECGs of adequate technical quality in a nonpaced rhythm
obtained within a mean of 16 days of death were available for
a total of 212 patients (113 men, 99 women; mean [_+SD] age
60 _+ 16 years). Women with a left-sided mastectomy were
excluded because the decreased distance from the heart to the
chest surface has been shown to increase ECG voltage out of
proportion to left ventricular mass (26,27). Forty-one subjects
(19%) had a body mass index exceeding partition values used
to recognize overweight status (28,29) (27.8 kg/m2 in men,
27.3 kg/m2 in women) and were considered overweight to
mildly obese (24).
Electrocardiography, Standard 12-lead ECGs were recorded at 25 mm/s and 1-mV/cm standardization with equipment having frequency response characteristics in accordance
with American Heart Association recommendations (30).
Tracings were coded and interpreted by a single investigator
who had no knowledge of clinical or autopsy findings. QRS
duration was measured to the nearest millisecond and QRS
amplitudes to the nearest microvolt on digitized ECGs (n =
117); analog ECGs were measured to the nearest 10 ms and
nearest 0.5 mm (50/zV), respectively, with use of a magnifying
graticle (n = 95).
Two widely used simple voltage criteria for detection of left
ventricular hypertrophy were examined: 1) the Sokolow-Lyon
voltage (sum of the amplitude of the S wave in lead V 1 and the
R wave in lead V5 or V6) (31), and 2) the gender-specific
Cornell voltage (sum of the R wave in lead aVL and the
amplitude of the S wave in lead V3 adjusted by the addition of
8 mm [0.8 mV] in women) (12,19). Several variations of the
Cornell ECG criteria were also examined: 1) the Cornell
voltage-duration product (gender-specific Cornell voltage
times QRS duration) (19,20); 2) the Cornell multivariate
regression score (9), a weighted score that incorporates QRS
duration, QRS voltages, repolarization changes and P wave
abnormalities; and 3) the gender-specific adjusted Cornell
voltage (18), adjusted for age and body mass index, a measure
of obesity, according to regression equations derived in a
cohort of the Framingham study (18). Two additional complex
ECG criteria that incorporate QRS duration, QRS amplitudes,
repolarization abnormalities and P wave findings were also
examined: 1) the Romhilt-Estes point score (32), and 2) the
125
gender-specific Novacode regression equations of Rautaharju
et al. (17).
Left ventricular mass and determination of hypertrophy.
Left ventricular mass was measured by the chamber partition
method (33) and was indexed to body surface area, height and
heightz7. When left ventricular mass was indexed to body
surface area, left ventricular hypertrophy was defined as left
ventricular mass index >118 g/m2 in men and >104 g/m2 in
women, which approximates the upper 97th percentile of
normal left ventricular mass index in a subset of 39 autopsy
studies from patients with neither intrinsic disease nor hemodynamic load affecting the left ventricle (12). Left ventricular
mass/height >143 g/m in men or >102 g/m in women was
considered to represent hypertrophy (18,21), as was left ventricular mass/height 27 > 5 0 g/m2"7 in men or >47 g/m27 in
women (24).
Data analysis and statistical methods. The strength of the
relations between ECG criteria and left ventricular mass
indexed according to each method was assessed by Pearson
correlation coefficients. Differences in coefficients of correlation were compared statistically by two-tailed tests after applying the Fisher Z transformation. Definitions of test sensitivity
and specificity conform to standard use (34). Test sensitivity of
the Cornell product for the identification of left ventricular
hypertrophy according to each method of indexing left ventricular mass for body size was compared with the other ECG
criteria at partitions with matched specificities of 95% using
the McNemar modification of the chi-square method for
paired proportions. Because sensitivity and specificity of a test
are dependent on the partition value chosen for test positivity,
overall test accuracy was compared by receiver operating
characteristic (ROC) curve analysis, which compares test performance over a wide range of possible partition values to
identify differences in test performance between methods
independent of empirically derived criteria (35). The ROC
curves were compared statistically by means of a two-tailed
univariate Z test of the difference between the areas under two
performance curves (36).
Variability of measurements of correlation and of the area
under performance curves as a function of how left ventricular
mass was indexed was expressed as the coefficient of variability,
in percent (i.e., the standard deviation divided by mean
measurement multiplied by 100). For all comparisons, p < 0.05
was required for rejection of the null hypothesis.
Results
Study group. The prevalence of left ventricular hypertrophy in the overall cohort varied significantly according to the
method of indexing left ventricular mass for body size. The
prevalence of hypertrophy was highest (43%) when left ventricular mass was adjusted by body surface area, similar (42%)
when mass was indexed to height2"7 and significantly lower
(38%) when mass was indexed to height (p < 0.05 for each
comparison).
Characteristics of the study subjects in relation to the
126
OKIN ET AL.
ELECTROCARDIOGRAPHIC IDENTIFICATION OF HYPERTROPHY
JACC Vol. 27, No. 1
January 1996:124-31
Table 1. Demographic Variables in Relation to Presence or Absence of Left Ventricular HypertrophyAccording to Method of Indexing
Ventricular Mass for Body Size
LV Mass Indexed to BSA
Age (yr)
M/F (no.)
Height (cm)
Weight (kg)
BSA (m 2)
BMI (kg/m2)
Obesity (%)
LV Mass Indexed to Height
LV Mass Indexed to Height 27
No LVH
(n = 120)
p Value
LVH
(n = 92)
No LVH
(n = 131)
p Value
LVH
(n 81)
No LVH
(n = 122)
p Value
LVH
(n = 90)
58 _+ 16
65/55
167 _+ i0
67 _+ 17
1.75 _+ 0.23
23.8 _+ 5.7
19
0.004
NS
NS
NS
NS
NS
NS
64 _+ 14
48/44
166 + 12
65 _+ 15
1.71 _+ 0.22
23.4 _+ 5.0
20
59 + 16
79/52
168 + 11
64 +_ 16
1.72 + 0.23
22.8 _+ 5.0
15
0.03
0.01
NS
NS
NS
0.008
0.03
63 + 14
34/47
166 _+ 11
68 _+ 17
1.75 _+ 0.23
24.8 _+ 5.8
27
58 _+ 16
69/53
169 + 10
65 _+ 16
1.74 _+ 0.23
22.8 _+ 5.1
15
0.02
NS
0.01
NS
NS
0.02
0.05
63 _+ 15
44/46
165 _+ 11
67 _+ 16
1.72 _+ 0.22
24.6 + 5.6
26
Data presented are mean value _+ SD, unless otherwise indicated. BMI = body mass index; BSA - body surface area; F = female; LV - left ventricular; LVH =
left ventricular hypertrophy; M = male.
presence or absence of left ventricular hypertrophy for each
definition of hypertrophy are presented in Table 1. Independent of how left ventricular mass was indexed for body size,
subjects with hypertrophy were older than those without
hypertrophy but were similar in weight and body surface area.
Gender distribution was similar in subjects with and without
hypertrophy when left ventricular mass was indexed to body
surface area or height27, but there was a greater proportion of
women in the group with hypertrophy when mass was indexed
to height. When left ventricular mass was indexed to body
surface area, there was no significant difference in body mass
index or in the prevalence of mild obesity between subjects
with and without hypertrophy. In contrast, subjects with left
ventricular hypertrophy defined by left ventricular mass indexed to height or height 27 had a higher body mass index and
a greater prevalence of obesity than subjects without hypertrophy.
Linear correlations of ECG criteria. The linear correlation
of ECG criteria with indexed left ventricular mass according to
each method of indexation and the variability of correlation as
a function of the method of normalization are shown in Table
2. All ECG criteria exhibited modest correlations with indexed
left ventricular mass, with the Cornell product and Cornell
multivariate score having the highest correlation coefficients
and the Novacode score having the lowest coefficients of
correlation for each method of indexing mass. Most of the
ECG criteria exhibited relatively similar levels of linear correlation for each method of indexed left ventricular mass. As a
consequence, variability of correlation as a function of the
method of indexation was generally low and was lowest for the
Cornell multivariate score (1.2%) and Cornell product (2.9%).
In contrast, linear correlation of adjusted Cornell voltage was
more highly dependent on the method of indexing left ventricular mass, with a much higher coefficient of variability (23.0%)
than the other ECG criteria.
The greater variability of correlation of adjusted Cornell
voltage compared with the other ECG criteria reflects the
variable relation of these ECG criteria to unindexed left
ventricular mass, height, weight and body surface area (Table
3). Similar to findings with indexed left ventricular mass, the
Cornell product and multivariate score exhibited the strongest
and the Novacode the weakest positive correlations with
unindexed mass. In addition, there were striking differences in
the strength and direction of linear correlation of ECG criteria
with height, weight and body surface area. Adjusted Cornell
voltage exhibited a modest but significant negative linear
correlation with height and stronger positive correlations with
weight than with ventricular mass. Simple Cornell voltage had
a similar inverse relation to height but had only weak and
negative correlations with weight and body surface area. In
Table 2. Linear Correlations and Variability of Correlations of Electrocardiographic Criteria for Left Ventricular Mass Indexed to Body
Surface Area, Height and Height 27
Correlation Coefficient
Criteria
LV Mass Indexed to
Body Surface Area
LV Mass Indexed to
Height
LV Mass Indexed to
Height 27
Sokolow-Lyon voltage
Cornell voltage
Adjusted Cornell voltage
Cornell product
Romhilt-Estes point score
Novacode score
Cornell multivariate score
0.428*
0.478t
0.3025
0.576
0.452t
0.3445
0.598
0.391"
0.444t
0.414t
0.546
0.439t
0.3225
0.584
0.381 *
0.493
0.485
0.555
0.424t
0.3195
0.592
Correlation Coefficient
(mean +_ SD)
0.400
0.471
0.400
0.559
0.438
0.328
0.592
_+ 0.025
_+ 0.025
_+ 0.092
_+ 0.016
_+ 0.014
_+ 0.014
_+ 0.007
*p < 0.01, tp < 0.05, 5p < 0.001 versus Cornell product, All correlation coefficients significant at p < 0.0001. LV = left ventricular.
Coefficient of
Variability
(%)
6.3
5.3
23.0
2.9
3.2
4.3
1.2
JACC Vol. 27, No. 1
January 1996:124-31
OKIN ET AL.
ELECTROCARDIOGRAPHIC IDENTIFICATION OF HYPERTROPHY
127
Table 3. Linear Correlationsof ElectrocardiographicCriteria for Left VentricularMass and Measures
of BodySize
Correlation Coefficient (p value)
Criteria
Sokolow-Lyon voltage
Cornell voltage
Adjusted Cornell voltage
Cornell product
Romhilt-Estes point score
Novacode score
Cornell multivariate score
LV Mass
Height
Weight
Body Surface Area
0.379*
(<0.0001)
0.399*
(<0.001)
0.361"
(<0.001)
0.519
(<0.001)
0.429*
(<0.001)
0.311 $
(<0.001)
0.557
(<0.001)
-0.029
(0.672)
0.193
(0.005)
-0.218
(0.001)
-0.059
(0.391)
0.033
(0.638)
-0.066
(0.342)
-0.057
(0.407)
-0.132
(0.055)
0.127
(0.065)
0.478t
(<0.001)
-0.081
(0.242)
0.021
(0.767)
-0.111
(0.109)
-0.034
(0.619)
-0.108
(0.116)
-0.157
(0.022)
0.300:~
(<0.001)
-0.066
(0.337)
0.013
(0.852)
-0.094
(0.171)
0.026
(0.707)
*p < 0.05, tp < 0.001, Sp < 0.01 versus Cornell product. LV = left ventricular.
contrast, none of the other ECG criteria had statistically
significant linear correlations with height, weight or body
surface area.
Performance of ECG criteria in detection of left ventricular
hypertrophy. The sensitivity and overall performance of ECG
criteria for the identification of left ventricular hypertrophy in
relation to the method of indexing left ventricular mass are
examined in Tables 4 and 5. Test performance of simple and
adjusted Cornell voltage, relative to the Cornell product,
varied with the definition of hypertrophy. When hypertrophy
was defined by left ventricular mass/body surface area, the 52%
sensitivity of the Cornell product was significantly greater than
the 45% sensitivity of simple Cornell voltage and than the 24%
sensitivity of adjusted Cornell voltage. For hypertrophy defined by left ventricular mass/height, the 43% sensitivity of the
Cornell product was significantly greater than the 35% sensitivity of Cornell voltage with a trend toward increased accuracy
compared with the 33% sensitivity of adjusted Cornell voltage
Table 4. Sensitivityof ElectrocardiographicCriteria for Left
Ventricular Hypertrophyin Relation to Method of Indexationof
Left VentricularMass Using Partitions With Matched Specificities
of 95%
Sensitivity (%) for LVH Defined by LV Mass
Indexed to Measures of Body Size
Criteria
Body Surface Area
Height
Height z7
Sokolow-Lyon voltage
Cornell voltage
Adjusted Cornell voltage
Cornell product
Romhilt-Estes point score
Novacode score
Cornell multivariate score
30*
452
24*
52
27*
23*
46
22*
355
33
43
23*
17'
37
22t
36
29~
39
23?
217
38
*p < 0.001, tp < 0.01, Sp < 0.05 versus Cornell product. LV - left
ventricular; LVH = left ventricular hypertrophy.
(p = 0.10). When left ventricular mass was indexed to
height2"7, the 39% sensitivity of the Cornell product was
significantly greater than the 29% sensitivity of adjusted Cornell voltage but only slightly higher than the 36% sensitivity of
simple Cornell voltage. Comparison of ROC curve areas
(Table 5) confirmed the superior overall performance of the
Cornell product relative to adjusted Cornell voltage for left
ventricular hypertrophy defined by normalization to body
surface area and height27 and further demonstrated that
overall performance of the Cornell product was greater than
simple Cornell voltage for each definition of hypertrophy. In
contrast, test performance of Sokolow-Lyon voltage and the
multivariate scores relative to the Cornell product did not vary
significantly with a changing definition of hypertrophy. Sensitivity and overall performance of the Cornell product were
significantly higher than that found for Sokolow-gyon voltage,
Romhilt-Estes point score and Novacode score and were
similar to that of the Cornell multivariate score, independent
of how left ventricular mass was indexed for body size (Tables
4 and 5).
Variability of overall performance of ECG criteria in
relation to the definition of left ventricular hypertrophy is
examined in Table 5. Overall performance of simple Cornell
voltage and the Cornell product varied only slightly with the
definition of hypertrophy, producing the lowest coefficients of
variability (1.0% and 1.7%). In contrast, performance of the
Romhilt-Estes point score and adjusted Cornell voltage were
more highly dependent on the method of indexing left ventricular mass, with coefficients of variability of 5.6% and 5.8%,
respectively.
Effect of obesity on performance of ECG criteria. The
relation of the sensitivity of ECG criteria for the identification
of left ventricular hypertrophy to the presence or absence of
obesity according to the different methods of indexing left
ventricular mass is shown in Table 6. Independent of how
128
OKIN ET AL.
ELECTROCARDIOGRAPHIC IDENTIFICATION OF HYPERTROPHY
JACC Vol. 27, No. 1
January. 1996:124-31
Table 5. Overall Performance and Variability of Performance Defined by Receiver Operating Characteristic Curve Areas of
Electrocardiographic Criteria for Left Ventricular Hypertrophy in Relation to Method of Indexation of Left Ventricular Mass
ROC Curve Areas for LVH Defined by LV Mass Indexed
to Measures of Body Size
Coefficient of
Criteria
Body Surface
Area
Height
Height27
ROC Curve Area
(mean ± SD)
Variability
(%)
Sokolow-Lyon voltage
Cornell voltage
Adjusted Cornell voltage
Cornell product
Romhilt-Estes point score
Novacode score
Cornell multivariate score
0.725*
0.775*
0.698~:
0.821
0.777t
0.704*
0.842
0.686*
0.774?
0.779
0.799
0.714'
0.679*
0.807
0.701'
0.760?
0.765t
0.794
0.700*
0.653:]:
0.799
0.704 ± 0.016
0.770 + 0.008
0.747 + 0.043
0.805 + 0.014
0.730 ± 0.041
0.679 _+ 0.026
0.816 ± 0.023
2.3
1.0
5.8
1.7
5.6
3.8
2.8
*p < 0.01, ?p < 0.05, Sp < 0.001 versus Cornell product. ROC = receiver operating characteristic; other abbreviations as Table 4.
hypertrophy was defined, Cornell voltage adjusted for age and
body mass index demonstrated significantly higher sensitivities
in obese than in nonobese subjects. In contrast, there were
trends toward higher sensitivities in the nonobese subjects for
the remaining ECG criteria that did not achieve statistical
significance, except for the higher sensitivity of Sokolow-Lyon
voltage in nonobese subjects when left ventricular mass was
indexed to body surface area. Among nonobese subjects, the
40% to 54% sensitivities of the Cornell product were significantly greater than the 18% to 24% sensitivities of adjusted
Cornell voltage for each definition of hypertrophy (each p <
0.001). In contrast, among obese subjects, sensitivity of adjusted Cornell voltage was significantly greater than that of the
Cornell product when left ventricular mass was indexed to
height, with trends toward increased sensitivity when hypertrophy was defined according to indexation to body surface area
or height 27. The strong relation of sensitivity of adjusted
Cornell voltage to body habitus is further accented by its
significantly lower sensitivity in markedly underweight subjects.
In contrast to the increased sensitivity of adjusted Cornell
voltage in obese subjects, sensitivity of adjusted Cornell voltage
in subjects below the fifth percentile of gender-specific body
mass index partitions (29) ranged from only 0% to 13% for
each definition of left ventricular hypertrophy, significantly
lower than the 60% to 72% sensitivities of the Cornell product
(p < 0.05 to p < 0.01). Furthermore, sensitivity of the Cornell
product was significantly greater than sensitivity of SokolowLyon voltage, Romhilt-Estes point score and Novacode score,
independent of the presence or absence of obesity for each
definition of hypertrophy.
Discussion
The present findings demonstrate that performance of
ECG criteria for the identification of left ventricular hypertrophy varies according to the method of normalizing left ventricular mass for body size and with the presence or absence of
obesity. These data further demonstrate that the Cornell
product provides the best combination of overall accuracy and
low variability of performance between definitions of hypertrophy relative to simple and adjusted Cornell voltage and to
complex, multivariate ECG scores. These findings have important implications for the use of 12-lead ECG criteria for the
detection of left ventricular hypertrophy in clinical and epidemiologic studies.
Electrocardiographic detection of left ventricular hypertrophy: relation to definition of hypertrophy. Although numerous ECG criteria have been reported and compared for the
Table 6. Sensitivity of Electrocardiographic Criteria for Left Ventricular Hypertrophy in Relation to Presence or Absence of Obesity
According to Method of Indexation of Left Ventricular Mass Using Partitions With Matched Specificities of 95%
Sensitivity (%) for LVH Defined by LV Mass Indexed to Measures of Body Size
Body Surface Area
Height
Height 27
Criteria
Obese
(n = 18)
Nonobese
(n = 74)
Obese
(n = 22)
Nonobese
(n = 59)
Obese
(n - 23)
Nonobese
(n = 67)
Sokolow-Lyon voltage
Cornell voltage
Adjusted Cornell voltage
Cornell product
Romhilt-Estes point score
Novacode score
Cornell multivariate score
11
39
50
44
22
11
44
35*
46
18t
54
28
26
46
9
27
59
32
18
5
27
27
37
24~
47
25
22
41
9
30
52
35
17
9
35
27
37
21?
40
25
25
39
*p < 0.05, tp < 0.01 versus obese subjects. Abbreviations as in Table 4.
JACC Vol. 27, No. 1
January 1996:124-31
OKIN ET AL.
E L E C T R O C A R D I O G R A P H I C IDENTIFICATION OF H Y P E R T R O P H Y
detection of left ventricular hypertrophy (11-20,31), various
reference standards for hypertrophy and differing methods of
normalizing left ventricular mass for body size have been
utilized in the derivation and testing of these criteria. Cornell
voltage criteria and the Cornell multivariate criteria were
derived in relation to left ventricular mass indexed to body
surface area using both echocardiographic (11) and autopsy
(12) left ventricular mass determinations, as was the more
recently derived Cornell product (19,20). In comparison, adjusted Cornell voltage criteria were derived using a multivariate regression equation in which echocardiographic left ventricular mass adjusted for height was the dependent variable
(18). The Romhilt-Estes point score was initially tested against
autopsy left ventricular mass adjusted for height (32), and the
Novacode multivariate score was derived using radiographically determined left ventricular size as the dependent variable
(17). As a result, test performance might be expected to vary
when these criteria are applied for the detection of left
ventricular hypertrophy defined in different ways.
To our knowledge, this is the first study to critically examine
the relation of test performance of ECG criteria for the
detection of hypertrophy to the method of normalizing left
ventricular mass for body size. These data illustrate that test
performance of these criteria varies importantly with the
definition of left ventricular hypertrophy and that, in general,
test sensitivity and overall performance tend to be greatest
when hypertrophy is defined according to the technique used
in the original derivation of the criteria. Thus, Cornell voltage,
the Cornell product and the Cornell multivariate score had
their highest accuracy when left ventricular mass was indexed
to body surface area, whereas adjusted Cornell voltage performed best when tested against left ventricular mass/height.
As a consequence, relative performance of the Cornell product
and adjusted Cornell voltage for the identification of left
ventricular hypertrophy also varied as a function of how left
ventricular mass was corrected for body size, with significantly
greater performance of the Cornell product when hypertrophy
was defined by left ventricular mass indexed to body surface
area or height 27 but only a trend toward increased accuracy
when mass was indexed to height. Varying definitions of left
ventricular hypertrophy may in part account for differences in
the relative performance of adjusted and unadjusted Cornell
voltages found in previous analyses (18,37). Independent of
hypertrophy definition, Sokolow-Lyon voltage, Romhilt-Estes
point score and Novacode score had significantly poorer
sensitivity and overall performance for the identification of left
ventricular hypertrophy, limiting their clinical utility.
Reproducibility of performance for the detection of left
ventricular hypertrophy and of linear correlation with indexed
left ventricular mass varied importantly among the different
ECG criteria as a function of the method of indexing mass for
body size. Among the criteria with the better overall accuracy,
the greater variability of adjusted Cornell voltage in comparison to simple Cornell voltage criteria, the Cornell product and
the Cornell multivariate score most likely reflects the stronger
relation of adjusted Cornell voltage to measures of body size.
129
The inverse relation of adjusted Cornell voltage with height
and positive correlations of it with body surface area and,
especially, weight are not surprising in light of the inclusion of
body mass index in the regression equation used to calculate
adjusted Cornell voltage (18). In contrast, the Cornell product
exhibited no significant correlations with height, weight or
body surface area.
Relation of test performance to obesity. Obesity has been
demonstrated to be associated with the presence of left
ventricular hypertrophy (38,39) and, conversely, with decreased sensitivity of the ECG for hypertrophy (15,18,40) due
to the attenuating effects of obesity on QRS amplitudes
(15,18). Although adjusted Cornell voltage criteria were developed in part to improve ECG performance by taking the
relation of Cornell voltage to body mass index into account
(18), test performance of these criteria in relation to obesity
was not examined. The current study demonstrates that,
independent of the varying abilities of the different methods of
indexing left ventricular mass for body size to correctly adjust
for the effects of obesity on left ventricular hypertrophy (24),
adjusted Cornell voltage appears to overcorrect for obesity,
with significantly greater sensitivities in obese subjects and
lower sensitivities in underweight subjects. In contrast, most
other ECG criteria examined had a slightly lower sensitivity
and the Novacode score exceptionally low sensitivity in obese
subjects. The relation of test performance to obesity was
similar among men and women. The increased sensitivity of
adjusted Cornell voltage among obese subjects clearly reflects
the greater correlation with height and weight of these criteria
(Table 3), which were derived to take advantage of the known
relation of left ventricular hypertrophy to obesity (18,38,39).
These observations suggest that the relative performance of
adjusted and unadjusted Cornell criteria will vary with the
prevalence of obesity in the study group and provide a
potential explanation for the greater overall performance of
adjusted Cornell voltage in the Framingham cohort (18), in
which body mass index was substantially higher than in the
current study.
Limitations of the study. Although the current study examined three different methods of indexing left ventricular
mass for body size, the clinically most useful method remains
to be determined (24,25,41). In addition to the methods of
normalizing left ventricular mass used in the current study,
indexing left ventricular mass to height to the 2.0 power (41) or
to body surface area to the 1.5 power (24) have also been
proposed. However, similar differences in test sensitivity and
similar degrees of variability of performance were obtained if
ECG criteria were also examined in relation to definitions of
hypertrophy based on these alternative methods of normalizing left ventricular mass. In addition, the current study determined left ventricular mass at autopsy, making it possible that
ECG findings were distorted by metabolic consequences of the
terminal illness. Although there is a strong correlation between
echocardiographic and autopsy left ventricular mass (42), and
ECG variables have been similarly related to necropsy and
echocardiographic left ventricular mass in previous studies
130
OKIN ET AL.
ELECTROCARDIOGRAPHIC IDENTIFICATION OF HYPERTROPHY
(11,12), the present findings should be confirmed in patients,
using echocardiographically determined left ventricular mass.
The lower overall performance of Sokolow-Lyon criteria
and the Romhilt-Estes point score relative to the performance
of Cornell voltage criteria in the present study could in part be
due to a higher prevalence of concentric than eccentric hypertrophy in the study group. Previous studies utilizing echocardiographic left ventricular mass determinations have demonstrated that performance of Cornell criteria is lower in patients
with eccentric hypertrophy (20,37,43) and higher in those with
a greater prevalence of concentric hypertrophy (11,12). However, the determination of left ventricular mass at autopsy-when postmortem contracture variably changes wall thickness
and chamber dimensions from those present during life
(42,44)--precludes analysis in the present study of the relation
of test performance to the geometric patterns of hypertrophy
in these patients.
Clinical implications. Left ventricular mass indexed to
body surface area and to height have both predicted adverse
cardiovascular outcome in a variety of patient groups (1-4,45),
and further improvement in risk stratification may be obtained
by indexing left ventricular mass by measures of body size to
the allometrically correct power (25). The current data argue
that until the optimal method of indexing echocardiographic
left ventricular mass for body size can be determined, the
Cornell product provides a combination of the best test
accuracy relative to the other criteria tested combined with low
variability of performance as a function of hypertrophy definition. Additional evaluation of other newly proposed ECG
criteria, such as the time-voltage integral of the QRS complex
(46), according to hypertrophy definition, will be necessary to
define more precisely their diagnostic value in relation both to
refinement of echocardiographic criteria for left ventricular
hypertrophy and to the presence or absence of obesity. However, the present data suggest that either the Cornell product
or Cornell multivariate score provides more stable sensitivity
for left ventricular hypertrophy, independent of the presence
or absence of obesity, than adjusted Cornell voltage.
References
1. Levy D, Garrison RJ, Savage DD, Kannel WB, Castelli WP. Left ventricular
mass and incidence of coronary heart disease in an elderly cohort. Ann
Intern Med 1989;110:101-7.
2. Levy D, Garrison RJ, Savage DD, Kannel WB, Castelli WP. Prognostic
implications of echocardiographically determined left ventricular mass in the
Framingham Heart Study. N Engl J Med 1990;322:1561-6.
3. Casale PN, Devereux RB, Milner M, et al. Value of echocardiographic
measurement of left ventricular mass in predicting cardiovascular morbid
events in hypertensive men. Ann Intern Med 1986;105:173-8.
4. Koren MJ, Devereux RB, Casale PN, Savage DD, Laragh JH. Relation of
left ventricular mass and geometry to morbidity and mortality in men and
women with uncomplicated essential hypertension. Ann Intern Med 1991;
114:345-52.
5. Ghali JK, Liao Y, Simmons B, Castaner A, Cao G, Cooper RS. The
prognostic role of left ventricular hypertrophy in patients with or without
coronary" disease. Ann Intern Med 1992;117:831-6.
6. Devereux RB, Casale PN, Wallerson DC, et al. Cost-effectiveness of
echocardiography and electrocardiography for detection of left ventricular
JACC Vol. 27. No. 1
January 1996:124-31
hypertrophy in patients with systemic hypertension. Hypertension 1987;9
Suppl 2:69-76.
7. Kannel WB, Gordon T, Offutt D. Left ventricular hypertrophy by electrocardiogram: prevalence, incidence, and mortality in the Framingham Study.
Ann Intern Med 1969;71:89-105.
8. Kannel WB, Gordon T, Castelli WP, Margolis JR. Electrocardiographic left
ventricular hypertrophy and risk of coronary heart disease: the Framingham
Study. Ann Intern Med 1970;72:813-22.
9. MacMahon S, Collins G, Rautaharju P, et al. Electrocardiographic left
ventricular hypertrophy and effects of antihypertensive drug therapy in
hypertensive participants in the Multiple Risk Factor Intervention Trial. Am
J Cardiol 1989;63:202-10.
10. Sullivan JM, Vander Zwaag RV, el-Zeky F, Ramanathan KB, Mirvis DM.
Left ventricular hypertrophy: effect on survival. J Am Coil Cardiol 1993;22:
508-13.
11. Casale PN, Devereux RB, Kligfield P, et al. Electrocardiographic detection
of left ventricular hypertrophy: development and prospective validation of
improved criteria. J Am Coil Cardiol 1985;6:572-80.
12. Casale PN, Devereux RB, Alonso DR, Campo E, Kligfield P. Improved
sex-specific criteria of left ventricular hypertrophy for clinical and computer
interpretation of electrocardiograms: validation with autopsy findings. Circulation 1987;75:565-72.
13. Reichek N, Devereux RB. Left ventricular hypertrophy: relation of anatomic, echocardiographic and electrocardiographic findings. Circulation
1981;63:1391-8.
14. DeveretLx RB, Casale PN, Eisenberg RR, Miller DH, Kligfield P. Electrocardiographic detection of left ventrieular hypertrophy using echocardiographic determination of left ventricular mass as the reference standard.
Comparison of standard criteria, computer diagnosis and physician interpretation. J Am Co0 Cardiol 1983;2:82-7.
15. Levy D, Labib SB, Anderson KM, Christiansen JC, Kannel WB, Castelfi WP.
Determinants of sensitivity and specificity of electrocardiographic criteria for
left ventricular hypertrophy. Circulation 1990;81:815-20.
16. Romhilt DW, Bove KE, Norris RJ, et al. A critical appraisal of the
electrocardiographic criteria for the diagnosis of left ventricular hypertrophy. Circulation 1969;40:185-95.
17. Rautaharju PM, La Croix AZ, Savage DD, et al. Electrocardiographic
estimate of left ventricular mass versus radiographic cardiac size and the risk
of cardiovascular disease mortality in the epidemiologic follow-up of the first
National Health and Nutrition Examination Survey. Am J Cardiol 1988;62:
59-66.
18. Norman JE, Levy D, Campbell G, Bailey JJ. Improved detection of
echocardiographic left ventricular hypertrophy using a new electrocardiographic algorithm. J Am Coll Cardiol 1993;21:1680-6.
19. Molloy TJ, Okin PM, Devereux RB, Kligfield P. Electrocardiographic
detection of left ventricular hypertrophy by the simple QRS voltage-duration
product. J Am Coll Cardiol 1992;20:1180-6.
20. Okin PM, Roman MJ, Devereux RB, Kligfield P. Electrocardiographic
identification of left ventricular mass by simple voltage-duration products.
J Am Coil Cardiol 1995;25:417-23.
21. Levy D, Savage DD, Garrison R J, Anderson KM, Kannel WB, Castelli WP.
Echocardiographic criteria for left ventricular hypertrophy: the Framingham
study. Am J Cardiol 1987;59:956-60.
22. Devereux RB, Lutas EM, Casale PN, et al. Standardization of M-mode
echocardiographic left ventricular anatomic measurements. J Am Coil
Cardiol 1984;4:1222-30.
23. Hammond IW, Devereux RB, Reis G, et al. The prevalence and correlates
of echocardiographic left ventricular hypertrophy among employed patients
with uncomplicated hypertension. J Am Coil Cardiol 1986;7:639-50.
24. de Simone G, Daniels SR, Devereux RB, et al. Left ventricular mass and
body size in normotensive children and adults: assessment of allometric
relations and impact of overweight. J Am Coll Cardiol 1992;20:1251-60.
25. de Simone G, Devereux RB, Daniels SR, Koren MJ, Meyer RA, Laragh JH.
Effect of growth on variability of left ventricular mass: assessment of
allometric signals in adults and children and of their capacity to predict
cardiovascular risk. J Am Coil Cardiol 1995;25:1056-62.
26. Lamonte CS, Freiman AH. The electrocardiogram after mastectomy. Circulation 1965;32:746-54.
27. Horton JD, Sheuber HS, Lakatta EG. Distance correction for precordial
electrocardiographic voltage in estimating left ventricular mass. Circulation
1977;55:509-12.
28. National Institute of Health Consensus Development Panel on the Health
JACC Vol. 27, No. 1
January 1996:124-31
29.
30.
31.
32.
33.
34.
35.
36.
37.
OKIN ET AL.
ELECTROCARDIOGRAPHIC IDENTIFICATION OF HYPERTROPHY
Implications of Obesity. Health implications of obesity. Ann Intern Med
1985;103:1074-7.
Najjar MF, Rowland M. Anthropometric reference data and prevalence of
overweight United States, 1976-80. Hyattsville (MD): National Center for
Health Statistics; 1987. Vital & Health Statistics; Series 11, No. 238. DHHS
Publication No. (PHS) 87-1688.
Pipberger HV, Arzbaecher RC, Berson AS, et al. Recommendations for
standardization of leads and specifications for instruments in electrocardiography and vectorcardiography. Report of the Committee on Electrocardiography, American Heart Association. Circulation 1975;AHA Supp[:I-131.
Sokolow M, Lyon TO. The ventricular complex in left ventricular hypertrophy as obtained by unipolar precordial and limb leads. Am Heart J
1949;37:161- 86.
Romhilt DW, Estes EH. A point score system for the ECG diagnosis of left
ventricular hypertrophy. Am Heart J 1968;75:752-8.
Bove KE, Rowlands DT, Scott RC. Observations on the assessment of
cardiac hypertrophy utilizing a chamber partition technique. Circulation
1966;33:558 -68.
Vecchio TJ. Predictive value of a single diagnostic test in unselected
populations. N Engl J Med 1966;274:1171-3.
Hanley JA, McNeil BJ. The meaning and use of the area under a receiver
operating characteristic (ROC) curve. Radiology 1982;143:29-36.
Metz CE, Wang P, Kronman HB. A new approach for testing the significance of differences between ROC curves measured from correlated data.
In: Deconick F, editor. Information Processing in Medical Imaging. The
Hague: Martinus Nijnhoff, 1984:432-45.
Okin PM, Roman MJ, Devereux RB, Kligfield P. Gender differences and the
electrocardiogram in left ventricular hypertrophy. Hypertension 1995;25:
242-9.
131
38. Levy D, Anderson KM, Savage DD, Kannel WB, Christiansen JC, Castelli
WP. Echocardiographically detected left ventricular hypertrophy: prevalence
and risk factors. Ann Intern Med 1988;108:7--13.
39. de Simone G, Devereux RB, Roman MJ, Alderman MH, Laragh JH.
Relation of obesity and gender to left ventricular hypertrophy in normotensive and hypertensive adults. Hypertension 1994;23:600-6.
40. Devereux RB, Phillips MC, Casale PN, Eisenberg RR, Kligfield P. Geometric determinants of echocardiographic left ventricular hypertrophy. Circulation 1983;67:907-11.
41. Lauer MS, Anderson KM, Larson MG, Levy D. A new method for indexing
left ventricular mass for differences in body size. Am J Cardiol 1994;74:48791.
42. Devereux RB, Alonso DR, Lutas EM, et al. Echocardiographic assessment
of left ventricular hypertrophy: comparison to necropsy findings. Am J
Cardiol 1986;57:450-8.
43. Roman MJ, Kligfield P, Devereux RB, et al. Geometric and functional
correlates of electrocardiographic repolarization and voltage in aortic regurgitation. J Am Coil Cardiol 1987;9:500-8.
44. Maron BJ, Henry WL, Roberts WC, Epstein SE. Comparison of ecbocardiographic and necropsy measurements of ventricular wall thickness in
patients with and without disproportionate septal thickening. Circulation
1977;55:341-6.
45. Bolognese L, Dellavesa P, Rossi L, Sarasso G, Sante Bongo A, Scianaro
MC. Prognostic value of left ventricular mass in uncomplicated acute
myocardial infarction and one-vessel coronary artery disease. Am J
Cardiol 1994;73:1-5.
46. Okin PM, Roman MJ, Devereux RB, Borer JS, Kligfield P. Electrocardiographic diagnosis of left ventricular hypertrophy by the time-voltage integral
of the QRS. J Am Coil Cardiol 1994;23:133-40.