Download Electrocardiographic criteria of left ventricular hypertrophy in general

Eur J Epidemiol
DOI 10.1007/s10654-008-9234-6
CARDIOVASCULAR DISEASE
Electrocardiographic criteria of left ventricular hypertrophy in
general population
Edoardo Casiglia Æ Laura Schiavon Æ Valérie Tikhonoff Æ Anna Bascelli Æ
Bortolo Martini Æ Alberto Mazza Æ Sandro Caffi Æ Daniele D0 Este Æ
Francesco Bagato Æ Monica Bolzon Æ Federica Guidotti Æ Hilda Haxhi Nasto Æ
Mario Saugo Æ Francesco Guglielmi Æ Achille C. Pessina
Received: 11 August 2006 / Accepted: 19 February 2008
Springer Science+Business Media B.V. 2008
Abstract The question on whether the electrocardiographic criteria are reliable for detection of left ventricular
hypertrophy (LVH) and play a role in predicting outcome
is open. Answer can only proceed from population-based
studies over unselected people followed up for years. In
this study, 1,699 subjects from general population underwent echocardiogram and standard electrocardiogram
(ECG) codified for LVH with Minnesota code and with
other five methods. Other items were also recorded and
used as covariables. Left ventricular mass index (LVMI)
was 127.6 ± 44.9 g m-2 in men and 120.8 ± 41.2 g m-2
in women, and correlated directly with age in both genders.
Prevalence of echocardiographic LVH was 36.6% in men
and 53.4% in women. LVMI correlated directly with the
E. Casiglia (&) L. Schiavon V. Tikhonoff A. Bascelli F. Bagato M. Bolzon F. Guidotti H. H. Nasto A. C. Pessina
Department of Clinical and Experimental Medicine, University
of Padova, Via Giustiniani No. 2, 35128 Padova, Italy
e-mail: [email protected]
B. Martini
Department of Cardiology, General Hospital of Schio, Schio,
Italy
A. Mazza
Department of Internal Medicine, General Hospital of Rovigo,
Rovigo, Italy
S. Caffi M. Saugo
Department of Cardiology, General Direction and Data
Elaboration Centre, General Hospital of Thiene, Thiene, Italy
D. D0 Este
Department of Cardiology, Hospital of Mirano, Mirano, Italy
F. Guglielmi
Department of Cardiology, Hospital of Thiene, Thiene, Italy
Sokolow–Lyon score in both genders at any age, with the
Romhilt–Estes, Cornell and RaVL scores in all subjects but
elderly men, and with the Lewis score in men and women
aged B69 years. Sensitivity and the predictive value of
electrocardiographic tests, as well as the prevalence of
LVH diagnosed with electrocardiographic criteria, were
always low. Specificity was high for all the tests, and in
particular for the Cornell index. Only when diagnosed with
echocardiogram or with the Sokolow–Lyon criterion, LVH
was an independent predictor of mortality. We conclude
that electrocardiographic tests cannot be used as a surrogate of echocardiogram in detecting LVH in the general
population because their positive predictive value (PPV) is
unacceptably low. On the contrary, they could replace
echocardiography in the follow up and for prediction of
outcome, when LVH has previously been correctly diagnosed with other methods.
Keywords Mortality Population Epidemiology Electrocardiogram Echocardiogram
Abbreviations
CASTEL
CArdiovascular STudy in the ELderly
CI
Confidence interval
ECG
Electrocardiogram
EDIVS
End-diastolic inter-ventricular septum
FN
False negative
FP
False positive
LEOGRA
Last Evidences of Genetic risk factors in the
Aged
LVEDD
Left ventricular end-diastolic diameter
LVEDPWT Left ventricular posterior wall thickness
LVH
Left ventricular hypertrophy
LVM
Left ventricular mass
LVMI
Left ventricular mass index
123
E. Casiglia et al.
NPV
PPV
RR
VN
VP
Negative predictive value
Positive predictive value
Relative risk
Very negative
Very positive
Introduction
Although newer diagnostic tools are available, the electrocardiogram (ECG) remains the most common mean for
evaluating cardiac disease. In particular, several ECG criteria have been proposed for the detection of left ventricular
hypertrophy (LVH) both in clinical practice and in epidemiological studies. Nevertheless, doubts have been raised
about their predictive value, particularly in the elderly,
where cardiac fibrosis is highly represented [1, 2]. Left
ventricular mass (LVM) tends to increase with age, mainly
due to increase in electrically-inactive fibrous tissue [3–5].
Furthermore, in the elderly the ECG abnormalities that are
commonly attributed to LVH often depend on conduction
defects rather than on increase of muscular tissue [6],
making the ECG diagnosis of LVH less precise. ECG tests
of LVH have particularly been accused of having low
sensitivity, leading particularly in the elderly to underestimation of LVH, to errors in detecting LVH regression in
clinical trials, and to inclusion of a great number of subjects
in erroneous percentiles in epidemiological studies [7].
This problem is still open. In fact, the only way to
clarify whether or not ECG criteria are reliable in diagnosing LVH in the elderly is to test them against a
echocardiography in a population-based frame, but only a
very limited number of epidemiological studies were specifically dedicated to this question in the elderly.
The present analysis is aimed at evaluating (1) the
prevalence of LVH diagnosed with echocardiography and
with different ECG criteria in elderly versus younger
subjects from the general population, (2) the reliability of
the ECG criteria in epidemiological and clinical settings,
and (3) their role, if any, in predicting outcome. For these
purposes, a large general population was employed [8–14].
Methods
General protocol of the study
The here described analysis was performed in the frame of
the CArdiovascular STudy in the Elderly (CASTEL) and of
the Last Evidences Of Genetic Risk factors in the Aged
(LEOGRA) [8–14] studies. The aim of these studies was to
evaluate the prevalence of hypertension, the cardiovascular
123
risk pattern and the feasibility of a program for hypertension control in general population.
The CASTEL included 3,282 subjects and the LEOGRA
1,457 subjects aged 65–95 years from the northern Italian
towns of Castelfranco Veneto, Chioggia, Torrebelvicino
and Valli del Pasubio, all at few kilometers from each
other. For the aims of the present analysis the subjects were
cumulated together, being the two studies identical in
protocol and procedures. The response rate was 73%.
The characteristics of the subjects who gave informed
consent and were enrolled did not differ significantly from
those of subjects who did not participate. When subjects
had more than one ECG, only the best-quality one was
considered. Subjects with artificially paced rhythms and
pre-excitation syndrome were excluded; 1,298 subjects had
good-quality ECG recording and echocardiogram, gave
informed consent and entered the here described analysis.
The general characteristics of subjects included in the
analysis did not differ significantly from those of subjects
who were not included.
As control cohort, 200 men and 201 women aged 18–
64 years representative subjects representative of the
younger population of the same geographical area, identified from the register’s office files, were studied with the
same procedures used for elderly subjects.
The analysis was therefore performed on a total of 1,699
subjects, whose general characteristics are shown in Table 1.
Clinical and instrumental measurements
Anthropometrics
Body weight (in kg) was measured while fasting with a
mechanical device with the subjects wearing underwear,
body height (in m) was measured without shoes. Body
mass index was calculated (in kg m-2) as weight/squared
height ratio.
Blood pressure
Lying arterial blood pressure was measured by trained
doctors with a mercury manometer. Eight measurements
were taken: 3 at 15-min intervals at baseline, 3 at 15-min
interval a month later, and 2 at 15-min interval after
another month. To minimize the alert reaction and white
coat effects, if any, the average of the last two measurements was taken into account for the analysis of data. Any
terminal digit preference was also avoided.
Electrocardiogram
Standard 12-lead ECG was codified with Minnesota code
[15] and with other five methods (the Sokolow and Lyon
criterion [16], the Romhilt and Estes point score system
Electrocardiographic criteria of LVH
Table 1 General characteristics of the study population
Whole population (n = 1,699)
Men (n = 707)
Age (years)
65.8 ± 14.3 (65.1–66.5)
64.1 ± 14.5 (63.1–65,2)
66.9 ± 14.0 (66.1–67.8)*
Body mass index (kg m-2)
26.6 ± .4.3 (26.4–26.8)
26.7 ± 3.7 (26.4–27.0)
26.6 ± 4.6 (26.3–26.9)
Total cholesterol (mg dl-1)
222.6 ± 44.3 (220.5–224.7)
215.5 ± 41.6 (212.4–218.6)
227.7 ± 45.5 (224.8–230.5)*
-1
Women (n = 992)
Triglycerides (mg dl )
123.3 ± 79.3 (119.5–127.1)
130.4 ± 92.1 (123.6–137.3)
118.3 ± 68.3 (114.0–122.5)*
HDL-C (mg dl-1)
49.8 ± 15.2 (49.0–50.5)
46.2 ± 14.4 (45.1–47.2)
52.3 ± 15.2 (51.4–53.3)*
Glycemia (mg dl-1)
105.7 ± 24.7 (104.5–106.9)
107.4 ± 24.6 (105.6–109.3)
104.5 ± 24.7 (102.9–106.0)*
Creatininemia (mg dl-1)
0.91 ± 0.22 (0.90–0.92)
0.99 ± 0.21 (0.97–1.00)
0.85 ± 0.21 (0.84–0.86)*
Uric acid (mg dl-1)
5.17 ± 1.41 (5.11–5.24)
5.68 ± 1.34 (5.58–5.78)
4.82 ± 1.36 (4.73–4.90)*
Cigarettes (die-1)
1.39 ± 4.20 (1.19–1.59)
2.24 ± 5.53 (1.83–2.65)
0.79 ± 2.77 (0.62–0.96)*
VC (% theoretical)
80.6 ± 19.8 (79.6–81.5)
79.6 ± 19.6 (78.1–81.0)
81.3 ± 20.0 (80.1–82.6)*
FEV1 (% theoretical)
Na (mEq l-1)
81.2 ± 21.6 (80.2–82.2)
141.5 ± 2.5 (141.4–141.7)
82.0 ± 21.1 (80.4–83.6)
141.2 ± 2.5 (141.0–141.4)
80.6 ± 22.0 (79.2–82.0)
141.8 ± 2.4 (141.6–141.9)*
K (mEq l-1)
4.22 ± 0.44 (4.20–4.24)
4.25 ± 0.44 (4.22–4.29)
4.19 ± 0.44 (4.17–4.22)*
Hematocrit (%)
41.9 ± 3.4 (41.7–42.0)
43.7 ± 2.0 (43.4–43.9)
30.6 ± 3.1 (40.4–40.8)*
Systolic blood pressure (mmHg)
160.5 ± 25.6 (159.2–161.7)
158.7 ± 23.8 (156.9–160.4)
161.7 ± 26.7 (160.1–163.4)*
Diastolic blood pressure (mmHg)
89.9 ± 10.8 (89.3–90.4)
90.4 ± 10.4 (89.6–91.2)
89.5 ± 11.1 (88.8–90.2)*
Heart rate (bpm)
71.2 ± 10.1 (70.7–71.7)
69.2 ± 10.2 (68.5–70.0)
72.6 ± 9.7 (72.0–73.2)*
Alcohol (ml 9 die-1)
36.5 ± 34.4 (34.8–38.1)
54.1 ± 43.4 (50.9–57.4)
23.9 ± 17.3 (22.8–24.9)*
Diabetes (%)
237/1699 (13.9)
102/707 (14.4)
135/992 (13.6)
Respiratory insufficiency (%)
847/1699 (49.8)
391/707 (66.3)
456/992 (46.0)*
Mean ± standard deviation (95% CIs) are shown
* p \ 0.001 men versus women
HDL-C, high-density-lipoprotein serum cholesterol; VC, vital capacity; FEV1, forced expiratory volume in 1 s; Na and K, serum sodium and
potassium
[17], the Lewis index [18], the R in aVL (RaVL) voltage
criterion [19, 20] and the Cornell score [21, 22]) by an
expert who did not know the aim and design of the study.
The cut-off values are summarized in Table 2. According
to Casale [21, 22], the Cornell cut-off was different in men
and women.
More than 30 ECG criteria of LVH have been proposed,
but those mentioned above are the most commonly used in
clinical practice and epidemiology.
Echocardiogram
Although magnetic resonance imaging has been suggested
to be the best way to check for LVH in limited numbers of
patients and volunteers [23], practical and economical
considerations make obviously impossible the use of this
technique in large-scale population-based studies including
a great number of subjects. Echocardiographic diagnosis of
LVH has notoriously good sensitivity and specificity when
compared to post-mortem weighing of heart [24–27], represents the method of choice for the evaluation of LVH
in vivo [28] and has been approved [28] for validation of
ECG criteria. M-mode 2D-guided echocardiogram (Megas
device, Esaote, Firenze, Italia) was therefore chosen as a
gold standard. For the purpose of the present study, left
ventricular end-diastolic diameter (LVEDD), end-diastolic
inter-ventricular septum (EDIVS) and left ventricular
posterior wall thickness (LVEDPWT) were measured
according to the American Society of Echocardiography
and Penn convention [29]. Recordings were analyzed
automatically during the exam using the inner software,
and simultaneously printed and then analyzed separately by
an independent operator who did not know the aim and
design of the study. Average of the two measurements was
used for analysis of data.
LVM was calculated from [29]:
LVM ðgÞ ¼ 0:832 ½ðEDIVS þ LVEDD
þ LVEDPWTÞ3 ðLVIDÞ3 þ 0:6
and indexed [Left ventricular mass index (LVMI), g m-2]
for body surface area (m2) calculated from [30]
71.84 9 weight0.425
9 heigth0.725. As suggested by Hamkg
mond et al. [31], echocardiographic diagnosis of LVH was
based on LVMI [134 g m-2 in men or [110 g m-2 in
women. Other methods of diagnosis and cut-off values
123
E. Casiglia et al.
Table 2 Electrocardiographic criteria of left ventricular hypertrophy
Method of detection
Criteria
Cut-off values for LVH
Sokolow and Lyon score system
S amplitude in V1 + R amplitude in V5 or V6
C35 mm
Romhilt and Estes points system
Amplitude criteria (highest R or lowest S in limb leads
C20 mm, or S in V1 or V2 C30 mm, or R in V5 or V6
C30 mm): 3 points; typical ST-T: 3 points; left atrial
enlargement (P wave C40 ms and terminal
negativity C1 mm in V1): 3 points; left axis deviation: 2
points; QRS C90 ms: 1 point; intrinsecoid
deflection C50 ms in V5 or V6: 1 point. The score must
be reduced by 1 point in the presence of digitalic
treatment
C 4 points
Lewis score system
Net positivity in I + net negativity in III
C17 mm
Cornell index
S amplitude in V1 + R amplitude in aVL
[25 mm in men or [22 mm
in women
Minnesota code
Code 3.1: R amplitude [26 mm in V5 or V6; or [20 mm
in any I, II, II or aVF; or [12 mm in aVL. Code 3.3: R
amplitude 15 B 20 mm in I, or R amplitude in V5 or
V6 + S amplitude in V1 [35 mm
code 3.1 or 3.3
R in aVL system
R amplitude in aVL
C11 mm
were proposed [28, 32], but those just mentioned have been
preferred because largely adopted and validated.
considering as prevalence in each age and sex group that
given by:
Reliability of ECG tests
prevalence ¼ ðtrue positives
þ false negativesÞ=numerosity.
To test reliability of ECG criteria of LVH, sensitivity,
specificity, and the predictive value of positive and negative results, were employed.
Sensitivity [33] was calculated from two-way tables in
relation to the gold standard from the algorithm:
sensitivity ð%Þ ¼ ½true positives=ðtrue positives
þ false negativesÞ 100
and specificity [32] from the algorithm:
specificity ð%Þ ¼ ½true negatives=ðtrue negatives
þ false positivesÞ 100:
According to Bayes’s approach [33, 34], the predictive
values of a positive (PPV) or negative (NPV) ECG test was
defined as:
PPVð%Þ ¼½ðprevalence sensitivityÞ=
fðprevalence sensitivityÞ
þ ½ð1 prevalenceÞ ð1 specificityÞg
100
NPVð%Þ ¼½ð1 prevalenceÞ specificity=
f½ð1 prevalenceÞ specificity
þ ½prevalence ð1 sensitivity)]g
100
123
Mortality was drawn from the National Institute of
Statistics forms and double-checked through the analysis of
medical files and asking the general practitioners.
Statistics
Comparison between means was made with analysis of
variance and the Tukey’s post-hoc test, that of frequencies
with the v2 test. Linear correlations were evaluated with the
Pearson coefficient and the Bonferroni’s conservative
correction. Adjustment for confounders such as age, gender
and body mass index were made when necessary in multiple regressions. Logistic regression was used to evaluate
the relative weight of covariables on the dichotomic
dependent variable LVH. Analysis of mortality was performed by the Mantel–Haenszel procedure after generating
Kaplan–Meier survival curves. Relative risk (RR) of LVH
as diagnosed with the different criteria was calculated,
together with 95% confidence intervals (CI), from Cox
multivariate analysis including as covariables body mass
index, smoking, alcohol, serum lipids, diabetes and—when
proper-age and gender.
Statistical analysis was performed first on the entire
population and then repeated after stratification for age and
gender; for age stratification the value of median (69 years
inclusive) was chosen as cut off.
Electrocardiographic criteria of LVH
Results
General characteristics of the population
Women were older than men and had higher values of
serum total and HDL cholesterol, sodium, systolic blood
pressure and heart rate; men had higher values of diastolic
blood pressure, serum triglycerides, glucose, uric acid and
potassium, higher hematocrit, and higher prevalence of
smoking, drinking, diabetes and respiratory disease
(Table 1).
Echocardiographic LVH
In the whole population, LVMI was 123.6 ± 42.9 g m-2,
i.e. 127.6 ± 44.9 (CI 124.2–130.9) g m-2 in men and
120.8 ± 41.2 (CI 118.2–123.5) g m-2 in women
(p = 0.002), and correlated directly with age both in men
(r = 0.45, p = 0.0001) and women (r = 0.38, p = 0.0001).
After stratification, LVMI was 110.6 ± 38.1 (CI 108.0–
113.2) g m-2 until 69 years and 137.4 ± 43.4 (CI 134.4–
140.4) g m-2 after this age (p \ 0.0001), with lower values
in women (B69 years: 105.4 ± 35.3, CI 102.1–108.6 vs.
g m-2; [69 years: 134.6 ± 41.2, CI 131.0–138.2 g m-2,
p \ 0.0001) than in men (B69 years: 116.6 ± 40.2, CI
112.6–120.6 g m-2; p \ 0.0001 vs. women having the same
age; [69 years: 142.1 ± 46.6, CI 136.7–147.5 g m-2,
p = 0.02 vs. women having the same age).
Up to 69 years of age, LVMI directly correlated with
systolic blood pressure both in men (r = 0.41, p \ 0.001)
and women (r = 0.52, p \ 0.005) even after correction for
age and body mass index, while no correlation was present
after 69 years of age, when LVMI was only predicted by
body mass index (Fig. 1).
Fig. 1 Correlation adjusted for
body mass index of systolic
blood pressure with
echocardiographic left
ventricular mass index (LVMI)
and with the values of the
electrocardiographic tests for
LVH based on numeric scores.
RaVL is the amplitude of the R
wave in aVL lead. See Table 2
for score definition
123
E. Casiglia et al.
Prevalence of echocardiographic LVH was 44.4% in the
whole population, i.e. 36.6% in men (25.8% in those
B69 years and 50.4% in those [69 years, p \ 0.0001) and
53.4% in women (38% in those B69 years, and 66.9% in
those [69 years, p \ 0.0001) (p \ 0.0001 men vs. women
for any age class).
In multivariate logistic regression analysis adjusted for
confounders, echocardiographic LVH was predicted by
hypertension (RR 3.80, CI 2.70–5.34), also including as
covariables age[69 years (RR 2.27, CI 1.83–2.82), female
gender (RR 2.07, CI 1.67–2.55) and body mass index
C25 kg m-2 (RR 1.25, CI 1.00–1.56). After stratification,
hypertension predicted echocardiographic LVH both in
men (RR 3.06, CI 1.56–5.98) and women (RR 6.90, CI
3.98–11.90) until 69 years, but not after this age.
Electrocardiographic LVH
The Sokolow–Lyon score correlated with systolic blood
pressure in both genders at any age; the other scores are
summarized in Fig. 1. LVMI correlated directly with the
Sokolow–Lyon score in both genders at any age, with the
Romhilt–Estes, Cornell and RaVL scores in all subjects but
elderly men, and with the Lewis score in men and women
aged B69 years.
Sensitivity of the ECG tests of LVH is summarized in
Table 3. The predictive value of a positive ECG test was
always low (from 5.9% to 25.8%), and the prevalence of
LVH diagnosed with the different ECG tests was always
falsely low when compared to the gold standard (Fig. 2). In
Table 3, the PPV and NPV of ECG tests are summarized.
Before the age of 69 years, arterial hypertension predicted
LVH diagnosed with the Sokolow–Lyon criterion in
women (RR 4.10, CI 1.20–14.0), with the Romhilt–Estes
score both in men (RR 2.82, CI 1.34-5-96) and women (RR
10.30, CI 2.97–35.40), with the Lewis criterion both in men
(RR 16.4, CI 2.20–12.2) and women (RR 7.56, CI 2.29–
25.0), with Minnesota code both in men (RR 4.53, CI 1.74–
11.80) and women (RR 3.23, CI 1.09–9.53), and with the
RaVL criterion both in men (RR 9.67, CI 1.28–73.1) and
women (RR 13.8, CI 1.84–104.00). No prediction was
possible after the age of 69 years.
Only the Cornell criterion was able to put in evidence
the higher prevalence of LVH in women (6.6%) than in
men (2.8%, p \ 0.001 vs. women), a difference that was
shown by the gold-standard. Nevertheless, due to low
Table 3 Sensitivity, specificity, positive and NPVs of the five electrocardiographic tests of left ventricular hypertrophy
Gender
Age
Men
(n = 707)
B69
(n = 396)
[69 (n = 311)
Women
(n = 992)
B69
(n = 461)
[69 (n = 531)
Sokolow–
Lyon score
Romhilt–
Estes score
Sensitivity
(%)
17.0
22.5
Specificity
(%)
90.8
86.1
PPV (%)
16.5
NPV (%)
84.5
Sensitivity
(%)
Specificity
(%)
PPV (%)
NPV (%)
Cornell
index
Lewis
index
Minnesota
code
R in aVL
5.9
16.0
23.5
8.0
99.3
92.1
90.1
94.2
21.6
5.9
15.4
21.5
8.2
79.3
95.1
85.5
79.4
92.8
16.7
25.5
3.8
10.9
17.2
7.1
82.5
77.3
96.1
86.4
92.9
92.2
17.5
25.8
4.0
11.7
16.3
7.4
83.5
75.2
97.0
89.3
84.6
93.6
Sensitivity
(%)
10.3
15.4
10.3
13.1
9.1
5.2
Specificity
(%)
97.6
96.2
97.6
93.7
96.2
63.4
PPV (%)
10.0
14.4
10.0
12.9
9.1
7.9
NPV (%)
90.9
86.5
91.0
88.1
91.9
93.6
Sensitivity
(%)
12.7
14.6
10.1
16.7
12.7
7.9
Specificity
(%)
93.7
86.9
97.2
89.7
92.6
93.1
PPV (%)
12.5
15.1
9.9
16.3
12.6
8.2
NPV (%)
88.5
86.0
91.1
84.6
88.4
92.8
TP, true positives; TN, true negatives; FP, false positives; FN, false negatives; PPV, positive predictive value; NPV, negative predictive value
123
Electrocardiographic criteria of LVH
Fig. 2 Prevalence (black bars)
of LVH as detected with the
Sokolow and Lyon score
system, the Romhilt and Estes
point system, the Lewis score,
the Cornell method and the
Minnesota code as compared
with echocardiographic LVH.
See Table 2 for score
definitions. Prevalence of
echocardiographic LVH (open
bars) is also shown for
comparison
sensitivity, this criterion was able to detect only a very
limited fraction of LVH in both genders.
Prognostic role of LVH
Only the LVH diagnosed with echocardiogram and that
diagnosed with the Sokolow–Lyon criterion were independent predictors of overall mortality (Fig. 4), with a RR
of 1.36 (CI 1.01–1.83) for the former and 1.69 (CI 1.17–
2.43) for the latter. Age [69 years, diabetes and female
gender acted in both cases as predictive covariables. When
the two above mentioned criteria (echocardiography and
Sokolow–Lyon) were both present (what happened in 6.2%
of population), RR was 2.10 (CI 1.23–3.56). On the other
hand, the presence of the echocardiographic criterion
alone, being negative the Sokolow–Lyon, increased the
risk insignificantly.
Discussion
Are ECG tests suitable for screening of LVH in the adult
population? The answer to this question is clearly negative.
As shown in Fig. 3, prevalence of LVH as detected with
ECG criteria was always lower than that detected with the
echocardiographic gold standard in both genders and in
both age classes below or above 69 years. When detected
with echocardiogram, prevalence of LVH was very high;
when detected with ECG, it was low. This is not surprising,
as echocardiogram detects the whole LVM (myocytes + fibrous matrix) while ECG indirectly extrapolates
from electric vectors only the electrically-active mass
represented by left ventricular vital myocytes [6]. Echocardiogram and ECG therefore give information about
different aspects of cardiac anatomy. In this context, it is
interesting to observe that echocardiographic LVMI was
strictly age-dependent and did not correlate with systolic
blood pressure in older subjects, where the fibrous component is particularly relevant. On the contrary, the
Sokolow–Lyon score correlated with systolic blood pressure in both genders at any age, and the other numeric ECG
scores correlated with systolic blood pressure in all subjects but elderly men (Fig. 1), thus demonstrating the more
direct link of ECG scores with systolic work. The correlation between ECG indices and LVMI was practically
limited to subjects B69 years; after this age, a correlation
was only detectable in women for the Sokolow and Lyon,
the Lewis and the Cornell scores (Fig. 1).
It must be underlined that the high prevalence of LVH in
the population (perfectly in line with the 28–71% found in
123
E. Casiglia et al.
Fig. 3 Values of the
electrocardiographic tests for
LVH based on numeric scores,
plotted against
echocardiographic LVMI.
Horizontal lines represent the
cut-off values for
echocardiographic LVH,
vertical lines the cut-off values
of electrocardiographic LVH
with the different criteria. TP
are the true positives, FP the
false positives, TN the true
negatives and FN the false
negative versus the
echocardiographic gold
standard; s indicates subjects
aged B69 years, • those aged
[69 years
123
Electrocardiographic criteria of LVH
Fig. 4 Cumulative survival
during the 5 years of follow-up.
Echo: echocardiographic LVH,
SL: Sokolow–Lyon criterion
many other large-scale studies) [35–38] could have contributed to keep low the sensitivity of the ECG criteria.
A first conclusion of this population-based study is that
the ECG tests cannot be used as a surrogate of echocardiogram in detecting LVH in a population, in a sample of
persons and even in clinical practice, because they give
information that are different from those given by echocardiogram and above all because PPV is unacceptably low
(never over 27%). Diagnosing LVH with ECG therefore
means mistaking by defect in a lot of cases. This limitation,
due to high frequency of false negatives (FNs), is particularly evident for the Cornell criterion (that in our
experience detected less than 6% of true hypertrophy
among men) and for the RaVL criterion (that in both gender
did not reach 8%). These two criteria can therefore lead to
disownment of [90% of cases of LVH, erroneously labelleing them as normotrophic despite a LVMI over the
recommended limits. Also with the other ECG criteria, the
ability of detecting LVH was always inadequate, largely
under the canonical 50% representing the PPV of the
launch of a coin. The attitude of any tests to screen subjects
with a trait is low when the trait prevalence is also low [34,
39]. This is also why in our experience the PPV of some
ECG tests was greater in the cohort having higher prevalence of LVH (that of elderly women), however remaining,
largely under a reasonable limit for epidemiological or
clinical use.
The reason why so many FNs are found with ECG
probably depends on the fibrous tissue accompanying left
ventricular myocytes that reduces surface potentials, particularly in limb leads and especially in the elderly.
Coexistent myocardial ischemia (more frequent in old age,
when ECG test of LVH are just less sensitive) may also
interfere with ventricular conduction, further reducing
surface potentials.
These results are in agreement with those of other
Authors. In the experience of Weber [40], like in our survey, ECG criteria were positive in \20% of subjects with
LVMI [200 g m-2, and in that of Dollar and Roberts [41]
only in 40% of subjects whose hearts weighted [500 g.
They are, on the contrary, in disagreement with those of
other Authors who occasionally found a surprisingly high
sensitivity of voltage criteria [16–22, 24, 42, 43]. These
discrepancies are most likely attributable to the choice of
different gold standards, such as post-mortem examination
or chest X-ray, as well as ti lack of utilization of samples
representative of general population.
When used in population screening or in the clinical
milieu, all the ECG tests therefore appear to be inadequate;
worse, they may furnish erroneous information leading in
the first case to underevaluating the real prevalence of LVH
in the population, in the second case to labelling subjects
with LVH as normotrophic, and in both cases to including
subjects in wrong categories. It would therefore be wrong
to leave out of consideration echocardiography in population screening or for detecting hypertrophic subjects in
clinical practice [44].
On the contrary, all the ECG tests used in the present
study had high specificity. The Cornell was the criterion
with the highest specificity (never under 96%) and highest
negative predictive value (NPV; never under 91%), and
also the only one able to detect the higher prevalence of
LVH in women put in evidence by the gold standard. At
variance, the Lewis and the RaVL criteria did not recognize
any difference between genders, and the Sokolow–Lyon
criterion, the Romhilt–Estes score and the Minnesota code
even detected higher prevalence in men than in women.
Practically all the subjects labelled as hypertrophic by the
Cornell criterion were really hypertrophic at the gold
standard, an agreement that was higher in men aged B69%
123
E. Casiglia et al.
and lower—but anyhow very high—in those who were
[69 years old. Diagnosing LVH through the Cornell criterion implied a risk of including less than 9% of
normotrophics, an error that came down to 3% in elderly
men.
Having high specificity and high NPV, and correlating
well with LVMI, the ECG tests (particularly the Sokolow–
Lyon, the Romhilt–Estes and the Lewis) could legitimately
replace echocardiography provided that LVH has previously been correctly diagnosed with other methods (for
instance in regression trials) [45–47].
Apart from echocardiography, only the Sokolow–Lyon
method—despite its low sensitivity—was able to predict
mortality (Fig. 4). This finding is in agreement with the
results obtained in patients by Richardson et al. [48] and in
general population by Sundström et al. [49]. Not only this,
but also in our survey the excess mortality observed in
subjects with echocardiographic LVH was limited to those
who also had positive Sokolow–Lyon criterion; in these
subjects, the risk of dying within 4 years was double than
in normotrophic, while in those having nothing more than a
positive echocardiogram the risk was only insignificantly
increased. This finding underlines that electrocardiographic
and echocardiographic LVH are not identical conditions
[50] and that the ECG items can provide independent
information to fully assess the risk of death.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
References
1. Varagic J, Susic D, Frohlich E. Heart, aging and hypertension.
Cardiology 2001;16:336–41.
2. Morales MA, Ferdeghini EM, Piacenti M, Dattolo P, Distante A,
Maggiore Q. Age dependency of myocardial structure: a quantitative two-dimensional echocardiographic study in a normal
population. Echocardiography 2000;17:201–8.
3. Fuster V, Danielson MA, Robb RA. Quantization of left ventricular myocardial fiber hypertrophy and interstitial tissue in
human hearts with chronically increased volume and pressure
overload. Circulation 1977;55:504–8.
4. Keller K, Wanger KC, Goepfrich M, Stegaru B, Buss J, Heene
DI. Morphological quantification and differentiation of left ventricular hypertrophy in hypertrophic cardiomyopathy and
hypertensive disease. A two dimensional echocardiographic
study. Eur Heart J 1990;11:65–74.
5. Strauer BE. Structural and functional adaptation of the chronically overloaded heart in arterial hypertension. Am Heart J
1987;114:948–57.
6. Piccolo E, Raviele A, Delise P, Dainese F, Pascotto P, Totaro G,
et al. The role of left ventricular conduction in the electrogenesis
of left ventricular hypertrophy. Circulation 1979;59:1044–55.
7. Casiglia E, Maniati G, Daskalakis C, Colangeli G, Tramontin P,
Ginocchio G, et al. Left ventricular hypertrophy in the elderly:
unreliability of electrocardiographic criteria in 477 subjects ages
65 years or more. The CArdiovascular STudy in the ELderly
(CASTEL). Cardiology 1996;87:429–35.
8. Casiglia E, Spolaore P, Mormino P, Maschio O, Colangeli G,
Celegon L, et al. The CASTEL project (Cardiovascular Study in
123
21.
22.
23.
24.
25.
26.
27.
28.
the Elderly): protocol, study design, and preliminary result of the
initial survey. Cardiologia 1991;36:569–76.
Casiglia E, Mazza A, Tikhonoff V, Pavei A, Privato G, Schenal
N, et al. Weak effect of hypertension and other classic risk factors
in the elderly who have paid their toll. J Hum Hypertens
2002;16:21–31.
Casiglia E, Palatini P. CArdiovascular risk factors in the elderly. J
Hum Hypertens 1998;12:575–81.
Casiglia E, Tikhonoff V, Mazza A, Pessina AC. Systolic and
pulse hypertension. Aging Health 2005;1:85–94.
Casiglia E, Basso G, Guglielmi F, Martini B, Mazza A, Tikhonoff
V, et al. German origin clusters for high cardiovascular risk in an
Italian enclave. Int Heart J 2005;46:489–500.
Casiglia E, Tikhonoff V, Mazza A, Rynkiewicz A, Limon J, Caffi
S, et al. C-344T polymorphism of the aldosterone synthase gene
and blood pressure in the elderly: a population-based study. J
Hypertens 2005;23:1991–6.
Casiglia E, Saugo M, Schiavon L, Tikhonoff V, Rigoni G, Basso
G, et al. Reduction of cardiovascular risk and mortality. A population-based study. Adv Ther 2006;23:905–20.
Rose GA, Blackburn H, Gillum RF, Prineas RJ. Cardiovascular
survey methods, Annes 1: classification of the electrocardiogram
for population studies. 2nd ed. Geneva: WHO; 1982.
Sokolow JA, Lyon TP. The ventricular hypertrophy as obtained
by unipolar or precordial and limbs lead. Am Heart J
1949;37:161–85.
Romhilt DW, Estes EH. A point-score system for the ECG
diagnosis of left ventricular hypertrophy. Am Heart J
1968;75:752–8.
Lewis T. Observation upon ventricular hypertrophy with especial
reference to preponderance of one or other chamber. Heart
1914;5:367–72.
Schack JA, Rosemann RH, Katz LN. The aV limb leads in the
diagnosis of ventricular strain. Am Heart J 1950;30:697–705.
Kilty SE, Lepeschkin E. Effect of body build on the QRS voltage
of the electrocardiogram in normal men. Its significance in the
diagnosis of left ventricular hypertrophy. Circulation
1965;31:77–84.
Casale PN, Devereux RB, Kligfield P, Eisenberg RR, Miller DH,
Chaudhary BS, et al. Electrocardiographic detection of left ventricular hypertrophy: development and prospective validation of
improved criteria. J Am Coll Cardiol 1985;6:572–80.
Casale PN, Devereux RB, Alonso DR, Campo E, Klingfield 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.
Alfakih K, Walters K, Jones T, Ridgway J, Hall AS, Sivananthan
M. New gender-specific partition values for ECG criteria of left
ventricular hypertrophy. Recalibration against cardiac MRI.
Hypertension 2004;44:175–9.
Bemmett DH, Evans DW. Correlation of left ventricular mass
determined with echocardiographic voltage measurements. Br
Heart J 1974;36:981–7.
Liebson PR, Devereux RB, Horan MJ. Echocardiography in the
measurement of left ventricular wall mass. Hypertension
1987;9(Suppl 2):2–5.
Crow SR, Hannan P, Grandits G, Liebson P. Is the echocardiogram an appropriate ECG validity standard for the detection and
change in left ventricular size? J Electrocardiol 1996;29(Suppl):248–55.
Devereux RB, Alonso DR, Lutas EM, Gottlieb GJ, Campo E,
Sachs I, et al. Echocardiographic assessment of left ventricular
hypertrophy: comparison to necropsy findings. Am J Cardiol
1986;63:237–40.
Clémenty J, Bergère P, Bricaud H. Electrocardiography and
vectorcardiography in the evaluation of left ventricular
Electrocardiographic criteria of LVH
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
hypertrophy due to pressor overload. Eur Heart J 1982;3(Suppl
A):37–47.
Devereux RV. Evaluation of cardiac structure and function by
echography and other non-invasive techniques. In Laragh JH,
Brenner BM editors. Hypertension: pathophysiology, diagnosis
and management. New York: Raven Press; 1990. p. 1479–90.
Du Bois D, Du Bois F. A formula to estimate the approximate
surface area if height and weight are known. Arch Intern Med
1916;17:863–71.
Hammond IW, Devereux RB, Alderman MH, Lutas EM, Spitzer
MC, Crowley JS, et al. The prevalence and correlates of echocardiographic left ventricular hypertrophy among employed
patients with uncomplicated hypertension. J Am Coll Cardiol
1986;7:613–8.
Fragola PV, Autore C, Magni G, Albertini M, Pierangeli L,
Ruscitti G, et al.. Limitations of the electrocardiographic diagnosis of left ventricular hypertrophy: the influence of left anterior
hemiblock and right bundle branch block. Int J Cardiol
1992;34:41–8.
Greener PF, Mayewski RJ, Mushlim AI, Greenland P. Selection
and interpretation of diagnostics tests and procedures. Ann Intern
Med 1981;94:553–600.
Rifkin RD, Hood WB. Bayesian analysis of electrocardiographic
exercise stress testing. N Engl J Med 1977;297:681–6.
Lieb W, Mayer B, Stritzke J, Doering A, Hense H-W, Loewel H,
et al. Association of low-grade urinary albumin excretion with
left ventricular hypertrophy in the general population The
MONICA/KORA Augsburg Echocardiographic Substudy.
Nephrol Dial Transplant 2006;21:2780–7.
De Simone G, Kizer JR, Chinali M, Roman MJ, Bella JN, Best
LG, Lee ET, Devereux RB, Strong Heart Study Investigators.
Normalization for body size and population-attributable risk of
left ventricular hypertrophy. Am J Hypertens 2005;18:191–6.
Dawson A, Morris AD, Struthers AD. The epidemiology of left
ventricular hypertrophy in type 2 diabetes mellitus. Diabetologia
2005;48:1971–9.
Ferrara LA, Vaccaro O, Cardoni O, Mancini M, Zanchetti A.
Arterial hypertension increases left ventricular mass: role of hight
blood pressure control. J Hum Hypertens 2004;18:637–42.
Selzer A. The Bayes theorem and clinical electrocardiography.
Am Heart J 1981;101:360–3.
40. Weber JR. Left ventricular hypertrophy: its prime importance as a
controllable risk factor. Am Heart J 1988;116:272–9.
41. Dollar AL, Roberts WC. Usefulness of total 12-lead QRS voltage
compared with other criteria for determining left hypertrophy in
hypertrophic cardiomyopathy: analysis of 54 patients studied at
necropsy. Am J Med 1989;87:377–81.
42. Chou TC, Scott RC, Booth RW, McWhorter HB. Specificity of
the current electrocardiographic criteria in the diagnosis of left
ventricular hypertrophy. Am Heart J 1960;60:371–7.
43. Romhilt DW, Bove KE, Norris RJ, Conyers E, Conradi S,
Rowlands DT, et al. A critical appraisal of electrocardiographic
criteria for the diagnosis of left ventricular hypertrophy. Circulation 1969;75:752–8.
44. De Simone G, Schillaci G, Palmieri V, Devereux RB. Should all
patients with hypertension have echocardiography? J Hum Hypertens 2000;14:417–21.
45. Zhou SH, Rautaharju PM, Prineas R, Neaton J, Crow R, Calhoun
H, et al. Improved ECG Models for estimation of left ventricular
hypertrophy progression and regression incidence by redefinition
of criteria for a significant change in left ventricular hypertrophy
status. J Electrocardiol 1993;26(Suppl):108–13.
46. Cuspidi C, Michey I, Meani S, Severgnini B, Fusi V, Salerno M,
et al. Trends in hypertension control and left ventricular hypertrophy over three years. Ital Heart J 2002;3:514–9.
47. Okin PM, Devereux RB, Liu JE, Oikarinen L, Jern S, Kjeldsen
SE, et al. Regression of electrocardiographic left ventricular
hypertrophy predicts regression of echocardiographic left ventricular mass: the LIFE study. J Hum Hypertens 2004;18:403–9.
48. Richardson K, Engel G, Yamazaki T, Chun S, Froelicher VF.
Electrocardiographic damage scores and cardiovascular mortality. Am Heart J 2005;149:458–63.
49. Sundström J, Lind L, Ärnlöv J, Zethelius B, Andrén B, Lithell
HO. Echocardiographic and electrocardiographic diagnoses of
left ventricular hypertrophy predict mortality independently of
each other in a population of elderly men. Circulation
2001;103:2346–51.
50. Kohsaka S, Sciacca RR, Sugioka K, Sacco RL, Homma S, Di
Tullio MR. Additional impact of electrocardiographic over
echocardiographic diagnosis of left ventricular hypertrophy for
predicting the risk of ischemic stroke. Am Heart J 2005;149:
181–6.
123