Abstract
Background
Information is limited about the classification accuracy of electrocardiographic (ECG) criteria for left ventricular hypertrophy (LVH) in the presence of myocardial infarction (MI).
Methods
We evaluated LVH classification accuracy for a set of 16 ECG criteria and some combinations derived from them in 1642 patients (105 with MI) suspected of coronary heart disease with two‐dimensional echocardiography evaluation and a standard 12‐lead ECG recorded at the same time. Patients with left bundle branch block had previously been excluded. Measures of classification accuracy included sensitivity, specificity, likelihood ratios, and positive and negative predictive values.
Results
Diagnostic accuracy varied widely for different LVH criteria. The criteria with the best overall performance had highest sensitivity in the presence of MI and sensitivities of approximately 30% with relatively low specificities ranging from 72% to 78%. However, the classification accuracy for them was similar to that for patients without MI. The prevalence of LVH in patients with MI was higher (56%) than in those with no MI (31%). Classification accuracy of the best single previously published LVH criteria was comparable to that of the best combinations of any three of them.
Conclusions
The classification accuracy of LVH criteria in the presence of MI is comparable to that in patients without MI, in part possibly due to the higher LVH prevalence in the MI group. The presence of a well‐validated computer database facilitates comparative evaluation of ECG‐LVH criteria and derivation of optimal combinations of criteria for any given clinical application.
Keywords: electrocardiography, left ventricular hypertrophy, myocardial infarction, computer‐assisted interpretation
Left ventricular hypertrophy (LVH) increases considerably cardiovascular morbidity and mortality in patients with arterial hypertension as well as in patients with myocardial infarction (MI).1, 2, 3 Electrocardiography (ECG) is the most available noninvasive technique for the diagnosis of LVH, and its use as a first‐line screening tool is re‐commended by current guidelines to detect LVH in all patients with hypertension.4, 5, 6, 7 Several LVH electrocardiographical indices have been developed and in general they all present a fairly low sensitivity (usually <50%) and high specificity (85–90%).8 In patients with MI, the progressive development of left ventricular (LV) remodeling, with LV dysfunction and increased LV mass, leads to congestive heart failure and increases long‐term mortality. Although the ECG can be used to assess infarction size, evaluate patency of infarct‐related artery after reperfusion therapy, and to estimate prognosis, its ability to assess the presence of LVH, one of the landmarks of LV remodeling, has not been systematically studied. Subsequently, little is known about the effect of MI on ECG accuracy in the diagnosis of LVH. To our knowledge the only published work that has directly addressed this issue is a small study by Hanna et al.9 who retrospectively compared 49 patients with anterior wall MI with a historical cohort. In general, the ECG indices used presented both a lower sensitivity and specificity for anterior wall MI group when compared with the historical cohort.
The precise application of the ECG for the diagnosis of LVH is limited by difficulties in measuring manually QRS in the standard 12‐lead ECG. Therefore, computer‐assisted measurement and interpretation of digital ECG could be helpful to reduce inaccuracies and consequently to better assess the diagnostic ability of ECG for detecting LVH.10 Furthermore, computer‐assisted ECG interpretation can facilitate diagnostic algorithm implementation. ELECTROPRES is an internet platform developed in Spain to provide computer‐assisted detection of LVH by ECG which uses the validated HES® software for ECG interpretation.11, 12 In order to improve the diagnostic performance of the platform, several diagnostic algorithms based on the combination of standard LVH ECG criteria have been developed to be applied besides standard ECG criteria of LVH. By using the ELECTROPRES internet platform we have recently demonstrated that the presence of left bundle branch block (LBBB) does not affect the diagnostic accuracy for the detection of LVH.13
The aim of this study is to investigate the diagnostic accuracy of standard ECG diagnostic criteria of LVH and different diagnostic algorithms as applied by ELECTROPES in patients with MI whose digital ECG and two‐dimensional (2D) echocardiography were performed simultaneously.
METHODS
Study Population
From January 2003 to August 2009, a cohort of 2090 patients (60.7% out of 3441 consecutive patients with echocardiogram) had a 2D echocardiogram and ECG performed during the same visit to an outpatient cardiac clinic. Patients younger than 18 years old were excluded. Those with poor quality echocardiogram (215 patients) were also excluded, which left a sample of 1875 patients. Finally, patients with LBBB were excluded, leaving the 2 study groups: documented clinical history of MI (n = 105) and patients with neither LBBB nor MI (n = 1537). Figure 1 shows the steps followed for the inclusion of patients in the study.
Figure 1.
Flow chart depicting the steps for the inclusion of patients in the study. ECG = electrocardiogram; ECHO = echocardiogram; MI = myocardial infarction; LBBB = left bundle branch block.
The research protocol was approved by the Ethical Committee of Hospital Virgen de la Salud of Toledo (Spain).
Electrocardiography
All patients underwent a standard 12‐lead ECG with a MAC 1200 ST electrocardiograph (GE Medical Systems, Pittsburgh, PA). All recordings were digitally stored in a GE Cardiosoft database (v. 6.5, GE Healthcare, Pittsburgh, PA) and exported in XML format to the ELECTROPRES platform for computerized measurement and analysis.
ELECTROPRES is an internet‐based platform aimed at early detection of LVH by electrocardiography that uses the program Hannover ECG System (HES; Corscience, Erlangen, Germany) for ECG measurement and interpretation. The ECG analysis can be performed online although in this study this was done off‐line. This software, approved by FDA, was developed in 1969, has undergone several updates, and has shown a high accuracy in ECG interpretation.14
Besides standard ECG criteria of LVH, ELECTROPRES also uses three diagnostic algorithms of LVH which were considered diagnostic if one or more criteria include in them was met: algorithm A, the European Society of Cardiology hypertension guidelines recommended criteria (LVH by Sokolow‐Lyon voltage or by Cornell voltage‐duration product [VDP]); algorithm B (LVH by Sokolow‐Lyon voltage criterion, Cornell voltage criterion, Sokolow VDP or Cornell VDP); algorithm C (LVH by Sokolow‐Lyon voltage, Cornell voltage, Gubner‐Ungerleider, Lewis index, R‐V6/R‐V5, R wave in aVL, QRS 12 leads, Sokolow duration‐voltage product, Cornell duration‐voltage product, QRS 12 leads duration‐voltage product, Gubner‐Ungerleider duration‐voltage product, or R in aVL duration‐voltage product)15 (Table 1). Part of the electrocardiographic criteria, namely those of Sokolow‐Lyon, Cornell, and Gubner‐Ungerleider, the R in aVL, and the sum of the 12‐lead QRS, were analyzed not only in the voltage, but in the VDP as well.
Table 1.
Electrocardiographic Criteria for Left Ventricular Hypertrophy
| Criteria | Formula | LVH Criteria |
|---|---|---|
| Sokolow‐Lyon Voltage (mV) | S(V1) + max (R(V5) or R(V6) ) | ≥3.5 mV |
| Cornell Voltage (mV) | R(aVL) + S(V3) | ≥2.8 mV (men) |
| ≥2.0 mV (women) | ||
| R6:R5 | R(V6)/R(V5) | >1 |
| RaVL (mV) | R (aVL) | >1.1 mV |
| Gubner‐Ungerleider (mV) | R(I) + S(III) | >2.5 mV |
| Lewis (mV) | (R(I) + S(III)) ‐ (R(III) + S(I)) | >1.7 mV |
| QRS 12 (mV) | R wave + S wave (or Q wave, whichever is higher) in all 12 leads | >19.530 mV (men) |
| >18.499 mV (women) | ||
| HES | Logistic regression equation | |
| VDP Sokolow (ms x mV) | S(V1) + max (R(V5) x R(V6)) x QRS duration | >367.4 mVms (men) |
| >322.4 mVms (women) | ||
| VDP Cornell | Males: R(aVL) + S(V3) x QRS duration | >244 mVms |
| Females: (R(aVL) + S(V3) + 0.6 mV) x QRS duration | ||
| VDP Gubner | Gubner x QRS duration | >207 mVms |
| VDP RaVL | R(aVL) x QRS duration | >103 mVms |
| VDP QRS 12 | QRS area in 12 leads | >2348.8 mVms (men) |
| >1960.7 mVms (women) | ||
| Dalfó | R(aVL) + S(V3) | >1.6 mV (men) |
| >1.4 mV (women) | ||
| Perugia | (a) S(V3) + R(aVL) > 2.4 mV male or > 2.0 mV females, or | Any of those 3 variables |
| (b) Left ventricular strain pattern, or | ||
| (c) Romhilt‐Estes point score ≥5 | ||
| Romhilt‐Estes (points) | >4 or >5 points |
ECG = electrocardiographic; LVH = left ventricular hypertrophy; HES = Hannover ECG System; VDP = voltage‐duration product.
Echocardiography
At the time of the ECG recording all patients underwent a 2D echocardiogram performed with a Vivid 3 or Vivid 4 General Electric echocardiograph with a 2.5 MHz transducer by the same cardiologist (LRP). The images selected were those where visualization was optimal. Measurements were performed from long‐axis parasternal views displayed on the screen over the 2D end‐diastolic image, according to standard technique, assuring that the most perpendicular distance from the different structures were used. LV mass was calculated by the following equation:16, 17
where LVID: end‐diastolic LV internal dimension, IVS: end‐diastolic interventricular septum, LVPW: end‐diastolic LV posterior wall thickness. LVH was considered to be present when LVM index was > 134 g/m2 (men) or > 110 g/m2 (women).18
Statistical Analysis
We constructed a 2 × 2 table with the cutoff points for LVH and for each ECG criterion, calculating the sensitivity, specificity, positive and negative likelihood ratio, positive and negative predictive values, pretest odds and probability of a positive and negative test, and post‐test odds and probability of a positive or negative test using the standard definitions. Chi‐square test was used to test differences between ECG criteria and the ROC (receiver operated characteristic) curves to complete the analysis of the diagnostic accuracy of the different ECG criteria using standard methods.
In the 2‐tailed tests, a P < 0.05 was considered to indicate statistical significance. Calculations were performed with the SPSS v. 17.0 statistical software package (IBM Inc, Armonk, NY).
We assessed intraobserver variability by measuring twice, in a blinded fashion, 3 dimensions in 15 echocardiograms of randomly selected patients (a total of 42 measurements) within a 2‐week interval. Variability was calculated as standard deviation of the differences between the first and second measurements made by the same observer, and expressed as a percentage of the mean value.
RESULTS
Baseline Characteristics
The baseline characteristics of the patients are shown in Table 2. The MI group consisted of 82 men (87.6%) and 13 women (12.4%), with a mean age of 64.9 ± 11.4 years. The most common diagnoses were hypertension (n = 57; 54.3%), diabetes mellitus (n = 29; 27.6%), and hypercholesterolemia (n = 49; 46.7%). The localization of the MI was anterior in 28 patients (26.6%), inferior in 58 patients (55.2%), anterior and inferior in 5 patients (4.8%), and in 14 patients (13.4%) the MI was localized to areas other than the LV or no Q wave was observed on the ECG.
Table 2.
Baseline Characteristics (Demographic, Echocardiographic, and Electrocardiographic) of the Population with Myocardial Infarction and without Myocardial Infarction or Left Bundle Branch Block
| Variable | MI (n = 105) | Without LBBB or MI (n = 1537) | P |
|---|---|---|---|
| Age, years | 64.9 ± 11.4 | 53.6 ± 15.5 | <0.0010 |
| Men, % | 87.6 | 55.2 | <0.0010 |
| Hypertension, % | 54.3 | 51.8 | 0.6867 |
| Hypercholesterolemia % | 46.7 | 26.8 | <0.0010 |
| Diabetes mellitus % | 27.6 | 7.7 | <0.0010 |
| Body weight, kg | 95.5 ± 107.1 | 85.5 ± 67.8 | 0.1498 |
| Height, cm | 168.7 ± 7.3 | 168.8 ± 9.2 | 0.9131 |
| Body mass index | 33.5 ± 37.4 | 30.3 ± 27.2 | 0.2533 |
| LV mass index, g/m2 | 102.5 ± 27 | 82.7 ± 41.4 | <0.0010 |
| IVS end‐diastolic thickness, mm | 11.7 ± 2 | 10.8 ± 2.7 | <0.0010 |
| LV end‐diastolic diameter, mm | 46.8 ± 6.8 | 43 ± 5.7 | <0.0010 |
| Posterior wall end‐diastolic thickness, mm | 10.3 ± 1.6 | 9.9 ± 2.8 | 0.1479 |
| Sokolow‐Lyon voltage, mV | 1.65±0.72 | 1.99±0.73 | <0.0010 |
| Cornell voltage, mV | 1.25 ± 0.66 | 1.13 ± 0.59 | 0.0455 |
| R6:R5 | 0.94 ± 0.41 | 0.85 ± 0.19 | <0.0010 |
| R(aVL), mV | 0.55 ± 0.33 | 0.51 ± 0.34 | 0.2426 |
| Gubner‐Ungerleider, mV | 0.95 ± 0.58 | 1.1 ± 0.58 | 0.0104 |
| Lewis, mV | 0.62 ± 0.74 | 0.71 ± 0.8 | 0.2625 |
| 12‐lead QRS, mV | 11.9 ± 2.8 | 13.4 ± 3.7 | <0.0010 |
| Sokolow VDP, ms, mV | 174.6 ± 88.5 | 193.3 ± 80.7 | 0.0225 |
| Cornell VDP | 145.9 ± 103.4 | 137.2 ± 70 | 0.2344 |
| Gubner VDP | 102.6 ± 69.6 | 107.3 ± 63.3 | 0.4645 |
| R(aVL) VDP | 59.7 ± 41.2 | 50.1 ± 36.6 | 0.0100 |
| 12‐lead QRS VDP | 1283.7 ± 490.9 | 1293.9 ± 438.9 | 0.4984 |
| Dalfó | 1.25 ± 0.66 | 1.13 ± 0.59 | 0.0455 |
| Perugia | 0.30 ± 0.46 | 0.08 ± 0.28 | <0.0010 |
| Romhilt‐Estes, points | 2.42 ± 2.10 | 1.19 (1.32) | <0.0010 |
Unless otherwise indicated, the data are expressed as the mean ± standard deviation. LBBB = left bundle branch block; MI = myocardial infarction; LV = left ventricular; R(aVL) = R wave in lead aVL; IVS = interventricular septal; VDP = voltage‐duration product.
In the group of patients with neither LBBB nor MI there were 849 men (55.2%) and 688 women (44.8%), with a mean age of 53.6 ± 15.5 years. Suspected or known heart disease (n = 160; 79.3%), essential hypertension (n = 810; 51.9%), and arrhythmia (n = 160; 68.7%) were the most frequent clinical diagnoses.
The patients with MI were older, more likely to be male, and had increased LV mass index, IVS end‐diastolic thickness, LV end‐diastolic diameter, and posterior wall end‐diastolic thickness. A relatively low intraobserver variability in the echocardiographic measurements of 3% was observed.
Correlation between ECG Voltages and LV Mass
The correlation coefficients between LV mass index on echocardiography and each of the ECG indices are shown in Figure 2. In the MI group, the Perugia criterion presented the highest correlation with LV mass index (r = 0.313), followed by the Romhilt criterion (r = 0.286) and Cornell VDP criterion (r = 0.255). In the patients who did not have MI or LBBB, the correlation with the LV mass index was highest for R in aVL VDP (r = 0.695) followed by R in aVL (r = 0.664). The Perugia criterion presented the closest correlation values between the MI and no MI or LBBB group (r = 0.313 and r = 0.307; respectively). Greatest differences between the two patient groups were observed in the R in aVL and the R in aVL VDP criterion.
Figure 2.
Correlation coefficients between the left ventricular mass index as measured in the echocardiogram and the electrocardiographic criteria considered in this study in patients with MI and in those who did not have MI or LBBB. LV = left ventricular; ECG = electrocardiographic; MI = myocardial infarction; LBBB = left bundle branch block.
Sensitivity, Specificity, and Likelihood Ratios
The prevalence of LVH as determined by echocardiography with the threshold values considered (LVM index was >134 g/m2 in men or >110 g/m2 in women) was 56.2% in the MI group and 30.7% in the no MI or LBBB group. Tables 3 and 4 show the sensitivity, specificity, predictive values, likelihood ratios, and posttest probabilities of all electrocardiographic indices analyzed for the patients with MI and those with no MI no LBBB, respectively.
Table 3.
Sensitivity (S), Specificity (Sp), Positive Test Accuracy (PTA), Negative Test Accuracy (NTA), Positive Likelihood Ratio (+LR), Negative Likelihood Ratio (–LR), Prevalence of Left Ventricular Hypertrophy (LVH), Negative Posttest Probability (–PPT), and Positive Posttest Probability (+PPT) of All Electrocardiographic Criteria Considered in Patients with Myocardial Infarction (n = 105)
| S (%) ± | Sp (%) ± | PTA (%) ± | NTA (%) ± | Incidence | |||||
|---|---|---|---|---|---|---|---|---|---|
| Criteria | CI 95% | CI 95% | CI 95% | CI 95% | LR + | LR − | of LVH, % | PPT − | PPT + |
| Algorithm A | 15.2 ± 9.2 | 89.1 ± 9.0 | 64.3 ± 25.1 | 45.1 ± 10.2 | 1.4 | 0.95 | 56.2 | 64.3 | 55.0 |
| Algorithm B | 17.0 ± 89.6 | 89.1 ± 9.0 | 66.7 ± 23.9 | 46.6 ± 10.3 | 1.56 | 0.93 | 66.7 | 54.4 | |
| Algorithm C | 57.6 ± 12.6 | 43.5 ± 12.3 | 56.7 ± 12.5 | 44.4 ± 14.5 | 1.02 | 0.97 | 56.7 | 55.6 | |
| Sokolow‐Lyon voltage | 0.0 ± 0.0 | 97.8 ± 4.2 | 0.0 ± 0.0 | 43.3 ± 9.5 | 0.00 | 1.02 | 0.0 | 56.7 | |
| Cornell voltage | 5.1±5.6 | 95.7 ± 5.9 | 60.0 ± 42.9 | 44.0 ± 9.7 | 1.17 | 0.99 | 60.0 | 56.0 | |
| R6:R5 | 27.1 ± 11.3 | 78.3 ± 11.9 | 61.5 ± 18.7 | 45.6 ± 11.1 | 1.25 | 0.93 | 61.5 | 54.4 | |
| RaVL | 3.4 ± 4.6 | 93.5 ± 7.1 | 40.0 ± 42.9 | 43.0 ± 9.7 | 0.52 | 1.53 | 40.0 | 57.0 | |
| Gubner‐Ungerleider | 1.7 ± 3.3 | 97.8 ± 4.2 | 50.0 ± 69.3 | 43.7 ± 9.6 | 0.78 | 1.00 | 50.0 | 56.3 | |
| Lewis | 6.8 ± 6.4 | 91.3 ± 8.1 | 50.0 ± 34.7 | 43.3 ± 9.9 | 0.78 | 1.02 | 50.0 | 56.7 | |
| 12‐lead QRS | 0.0 ± 0.0 | 97.8 ± 4.2 | 0.0 ± 10.0 | 43.3 ± 9.5 | 0.00 | 1.02 | 0.0 | 56.7 | |
| HES | 5.1 ± 5.6 | 97.8 ± 4.2 | 75.0 ± 42.4 | 44.5 ± 9.7 | 2.34 | 0.97 | 75.0 | 55.4 | |
| Sokolow VDP | 3.4 ± 4.6 | 95.7 ± 5.9 | 50.0 ± 49.0 | 43.6 ± 9.7 | 0.78 | 1.01 | 50.0 | 56.4 | |
| Cornell VDP | 15.2 ± 9.2 | 89.1 ± 9.0 | 64.3 ± 25.1 | 45.1 ± 10.2 | 1.40 | 0.95 | 64.3 | 54.9 | |
| Gubner VDP | 8.5 ± 7.1 | 89.1 ± 9.0 | 50.0 ± 31.0 | 43.1 ± 10.1 | 0.78 | 1.03 | 50.0 | 56.8 | |
| R(aVL) VDP | 15.3 ± 9.2 | 87.0 ± 9.7 | 60.0 ± 24.8 | 44.4 ± 10.3 | 0.97 | 1.2 | 60.0 | 55.6 | |
| 12‐lead QRS VDP | 11.9 ± 8.2 | 91.3 ± 8.1 | 63.6 ± 28.4 | 44.7 ± 10.0 | 0.97 | 1.41 | 66.6 | 55.3 | |
| Dalfó | 23.7 ± 10.9 | 71.7 ± 13.0 | 51.9 ± 18.9 | 42.3 ± 10.9 | 0.84 | 1.06 | 51.9 | 57.7 | |
| Perugia | 32.2 ± 11.9 | 71.7 ± 13.0 | 59.4 ± 17.0 | 45.2 ± 11.4 | 1.14 | 0.95 | 59.4 | 54.8 | |
| Romhilt‐Estes > 4 | 10.2 ± 7.7 | 91.3 ± 8.1 | 60.0 ± 30.4 | 44.2 ± 10.1 | 0.98 | 1.19 | 60.0 | 55.8 | |
| Romhilt‐Estes > 5 | 16.9 ± 9.6 | 80.4 ± 11.5 | 52.6 ± 22.4 | 43.1 ± 10.5 | 1.03 | 0.84 | 52.6 | 57.0 |
ECG = electrocardiographic; MI = myocardial infarction.
Table 4.
Sensitivity (S), Specificity (Sp), Positive Predictive Value (PPV), Negative Predictive Value (NPV), Positive Likelihood Ratio (+LR), Negative Likelihood Ratio (–LR), Prevalence of Left Ventricular Hypertrophy (LVH), Negative Posttest Probability (–PPT), and Positive Posttest Probability (+PPT) of All Electrocardiographic Criteria Considered in Patients with No Myocardial Infarction (MI) or Left Bundle Branch Block (LBBB) (n = 1561)
| S (%) | Sp (%) | PPV (%) | NPV (%) | Incidence | |||||
|---|---|---|---|---|---|---|---|---|---|
| Criteria | (±CI 5%) | (±CI 5%) | (±CI 95%) | (±CI 95%) | LR + | LR − | of LVH (%) | PPT + | PPT − |
| Algorithm A | 13.9 ± 3.1 | 95.7 ± 1.2 | 59.3 ± 9.1 | 71.5 ± 2.3 | 3.28 | 0.90 | 30.7 | 59.3 | 28.5 |
| Algorithm B | 14.8 ± 3.2 | 93.8 ± 1.4 | 51.5 ± 8.3 | 71.3 ± 2.4 | 2.39 | 0.91 | 51.5 | 28.7 | |
| Algorithm C | 53.1 ± 4.5 | 70.7 ± 2.7 | 44.6 ± 4.1 | 77.3 ± 2.6 | 1.81 | 0.66 | 44.6 | 22.8 | |
| Sokolow‐Lyon voltage | 3.5 ± 1.6 | 96.2 ± 1.0 | 36.2 ± 13.7 | 69.4 ± 2.3 | 1.28 | 0.99 | 36.2 | 30.6 | |
| Cornell voltage | 6.5 ± 2.2 | 99.3 ± 0.5 | 79.5 ± 12.7 | 70.5 ± 2.3 | 8.73 | 0.94 | 79.5 | 29.5 | |
| R6:R5 | 6.5 ± 2.2 | 95.5 ± 1.2 | 38.8 ± 10.7 | 69.7 ± 2.3 | 1.42 | 0.98 | 38.7 | 30.3 | |
| R(aVL) | 10.0 ± 2.7 | 96.6 ± 1.1 | 56.5 ± 10.5 | 70.7 ± 2.3 | 2.92 | 0.93 | 56.5 | 29.3 | |
| Gubner‐Ungerleider | 5.2 ± 2.0 | 98.6 ± 0.7 | 62.5 ± 15.0 | 70.1 ± 2.3 | 3.75 | 0.96 | 62.5 | 29.9 | |
| Lewis | 19.0 ± 3.5 | 93.4 ± 1.5 | 56.2 ± 7.7 | 72.3 ± 2.4 | 2.89 | 0.87 | 56.2 | 27.8 | |
| 12‐lead QRS | 6.5 ± 2.2 | 96.0 ± 1.2 | 41.9 ± 11.2 | 69.8 ± 2.3 | 1.62 | 0.97 | 41.9 | 30.2 | |
| HES | 9.2 ± 2.9 | 97.2 ± 1.0 | 59.5 ± 11.2 | 70.7 ± 2.3 | 3.30 | 0.93 | 59.5 | 29.3 | |
| Sokolow VDP | 5.2 ± 2.1 | 96.9 ± 1.0 | 42.4 ± 12.6 | 69.7 ± 2.3 | 1.66 | 0.98 | 42.4 | 30.3 | |
| Cornell VDP | 12.3 ± 2.9 | 97.4 ± 1.0 | 67.8 ± 9.8 | 71.4 ± 2.3 | 4.75 | 0.90 | 67.2 | 28.6 | |
| Gubner VDP | 13.3 ± 3.0 | 95.6 ± 1.2 | 57.1 ± 9.2 | 71.3 ± 2.3 | 3.00 | 0.91 | 57.1 | 28.7 | |
| R(aVL) VDP | 15.2 ± 3.2 | 95.1 ± 1.3 | 57.9 ± 8.6 | 71.6 ± 2.3 | 3.10 | 0.89 | 57.9 | 28.4 | |
| 12‐lead QRS VDP | 12.7 ± 3.0 | 92.2 ± 1.6 | 42.1 ± 8.0 | 70.4 ± 2.4 | 1.64 | 0.95 | 42.1 | 29.6 | |
| Dalfó | 37.1 ± 4.3 | 85.4 ± 2.1 | 53.0 ± 5.3 | 75.4 ± 2.4 | 2.54 | 0.74 | 54.0 | 24.7 | |
| Perugia | 16.0 ± 3.3 | 95.1 ± 1.3 | 59.2 ± 8.5 | 71.8 ± 2.3 | 3.27 | 0.88 | 59.2 | 28.2 | |
| Romhilt‐Estes > 4 | 6.0 ± 2.1 | 97.6 ± 0.9 | 52.7 ± 13.2 | 70.1 ± 2.3 | 2.51 | 0.96 | 52.7 | 30.0 | |
| Rohmhilt‐Estes > 5 | 6.3 ± 2.2 | 98.7 ± 0.7 | 68.2 ± 13.8 | 70.3 ± 2.3 | 4.83 | 0.95 | 68.2 | 29.7 |
In general, in patients with MI sensitivities were low or intermediate and ranged from 0% for the Sokolow‐Lyon voltage and 12‐lead QRS voltage to 57.6% for algorithm C. Specificities were intermediate or high, ranging from 43.5% for algorithm C to 97.8% for Sokolow‐Lyon voltage, HES, Gubner‐Ungerleider, and total QRS. Positive likelihood ratios ranged from 0.0 for Sokolow‐Lyon voltage to 2.34 for HES, and negative likelihood ratios ranged from 0.84 for Romhilt‐Estes > 5 to 1.53 for RaVL. For the prevalence of LVH observed in this study, these likelihood ratios result in a negative test accuracy probability (NTA) of LVH that ranges from 0% for Sokolov‐Lyon voltage to 75.0% for HES, and a positive test accuracy (PTA) of LVH that ranges from 54.4 for the algorithm B and RV6:RV5 to 57.7% for Dalfó.
In the no MI or LBBB group the sensitivities were overall low (between 3.5% for the Sokolow‐Lyon voltage and 19% for the Lewis criterion), apart from the Dalfó (37.1%) and the algorithm C (53.1%). Specificities were high in all electrocardiographic indices, ranging from 70.7% for the research algorithm to 99.3% for the Cornell voltage. Positive likelihood ratios ranged from 1.28 for the Sokolov‐Lyon voltage criteria to 8.73 for the Cornell voltage, and negative likelihood ratios ranged from 0.66 for the research algorithm to 0.99 for the Sokolow‐Lyon voltage. For the prevalence of LVH observed in this study, these likelihood ratios give a negative post‐test probability of LVH that ranges from 22.8% for the research algorithm to 30.6% for the Sokolow‐Lyon voltage, and a positive posttest probability of LVH that ranges from 36.2% for the Sokolov‐Lyon voltage criterion to 79.5% for the Cornell voltage.
When comparing sensitivity between the MI and the no MI or LBBB groups, the majority of the electrocardiographic indices showed no significant differences and when analyzing specificity between the same groups no significant differences were found (Table 5).
Table 5.
Comparisons between the Sensitivities and Specificities of the Two Groups of Patients: with Myocardial Infarction (MI) and No MI or Left Bundle Branch Block (LBBB)
| Sensitivity, % | Specificity, % | |||||
|---|---|---|---|---|---|---|
| No LBBB or | No LBBB | |||||
| Criteria | MI (n=1537) | MI (n=105) | P | (n=1537) | MI (n=105) | P |
| Algorithm A | 13.9 ± 3.1 | 15.2 ± 9.2 | 0.5337 | 95.7 ± 1.2 | 89.1 ± 9.0 | 0.8269 |
| Algorithm B | 14.8 ± 3.2 | 17.0 ± 89.6 | 1 | 93.8 ± 1.4 | 89.1 ± 9.0 | 0.8277 |
| Algorithm C | 53.1 ± 4.5 | 57.6 ± 12.6 | 0.0044 | 70.7 ± 2.7 | 43.5 ± 12.3 | 0.0755 |
| Sokolow‐Lyon voltage | 3.5 ± 1.6 | 0.0 ± 0.0 | 0.2391 | 96.2 ± 1.0 | 97.8 ± 4.2 | 1 |
| Cornell voltage | 6.5 ± 2.2 | 5.1 ± 5.6 | 1 | 99.3 ± 0.5 | 95.7 ± 5.9 | 0.9145 |
| R6:R5 | 6.5 ± 2.2 | 27.1 ± 11.3 | <0.0010 | 95.5 ± 1.2 | 78.3 ± 11.9 | 0.4309 |
| R(aVL) | 10.0 ± 2.7 | 3.4 ± 4.6 | 0.1485 | 96.6 ± 1.1 | 93.5 ± 7.1 | 0.9141 |
| Gubner‐Ungerleider | 5.2 ± 2.0 | 1.7 ± 3.3 | 0.3442 | 98.6 ± 0.7 | 97.8 ± 4.2 | 1 |
| Lewis | 19.0 ± 3.5 | 6.8 ± 6.4 | 0.0413 | 93.4 ± 1.5 | 91.3 ± 8.1 | 1 |
| 12‐lead QRS | 6.5 ± 2.2 | 0.0 ± 0.0 | 0.0379 | 96.0 ± 1.2 | 97.8 ± 4.2 | 1 |
| HES | 9.2 ± 2.9 | 5.1 ± 5.6 | 0.4597 | 97.2 ± 1.0 | 97.8 ± 4.2 | 1 |
| Sokolow VDP | 5.2 ± 2.1 | 3.4 ± 4.6 | 0.5429 | 96.9 ± 1.0 | 95.7 ± 5.9 | 1 |
| Cornell VDP | 12.3 ± 2.9 | 15.2 ± 9.2 | 0.7568 | 97.4 ± 1.0 | 89.1 ± 9.0 | 0.7431 |
| Gubner VDP | 13.3 ± 3.0 | 8.5 ± 7.1 | 0.4114 | 95.6 ± 1.2 | 89.1 ± 9.0 | 0.8269 |
| R(aVL) VDP | 15.2 ± 3.2 | 15.3 ± 9.2 | 1 | 95.1 ± 1.3 | 87.0 ± 9.7 | 0.7417 |
| 12‐lead QRS VDP | 12.7 ± 3.0 | 11.9 ± 8.2 | 1 | 92.2 ± 1.6 | 91.3 ± 8.1 | 1 |
| Dalfó | 37.1 ± 4.3 | 23.7 ± 10.9 | 0.1626 | 85.4 ± 2.1 | 71.7 ± 13.0 | 0.4909 |
| Perugia | 16.0 ± 3.3 | 32.2 ± 11.9 | 0.0267 | 95.1 ± 1.3 | 71.7 ± 13.0 | 0.2519 |
| Romhilt‐Estes > 4 | 6.0 ± 2.1 | 10.2 ± 7.7 | 0.2683 | 97.6 ± 0.9 | 91.3 ± 8.1 | 0.8279 |
| Rohmhilt‐Estes > 5 | 6.3 ± 2.2 | 16.9 ± 9.6 | 0.0185 | 98.7 ± 0.7 | 80.4 ± 11.5 | 0.3731 |
HES = Hannover ECG System; R(aVL) = R wave in lead aVL; VDP = voltage‐duration product.
Finally, there were no statistically significant differences in sensitivity or specificity when the voltage and VDP criteria were compared.
DISCUSSION
The results of this study confirm the relatively low sensitivity and high specificity that has previously been reported for the different LVH electrocardiographical indices.8 We have recently demonstrated that the presence of LBBB does not affect the diagnostic accuracy of the ECG for the diagnosis of LVH using the computer‐assisted internet platform ELECTROPRES.13 In this study, using the same platform and the same ECG criteria we have found that in the subgroup of patients with previous MI the sensitivity and specificity were similar to those observed in patients with no MI or LBBB. In patients with MI, RV6:RV5 and Perugia criteria showed a modest but significant increase in sensitivities with a nonsignificant reduction in specificities; hence, these two criteria appear to perform better in patients with MI, whereas no significant differences were observed in the other indices between the two populations studied. Therefore, at least when assessed by a computer‐assisted system, the presence of a previous MI nor LBBB does not seam to hamper the identification of LVH by different ECG criteria. This type of modern ECG evaluation offers several advantages such as fast and accurate blind measurement and simultaneous evaluation of several indices, as shown in this study. Nevertheless, although the combination of several LVH ECG criterion clearly improved sensitivity in our study this was at the expense of reduced specificity and our results therefore do not indicate a clear advantage for these algorithms over the old LVH criterion.
To the best of our knowledge, in patients with previous MI, very little is known with regards to the utility of ECG indices to predict LVH. A small retrospective study by Hanna et al.9 analyzed the accuracy of 5 LVH ECG indices (Cornell, R(aVL) > 1.1, R(I) ≥ 1.5, Sokolow‐Lyon and R(V6):R(V5) criterion) in patients with previous MI. They reported an overall worse sensitivity and specificity for the prediction of LVH in patients with MI, contrary to our findings. This difference may be due to a relatively small patient sample size and the use of a historical cohort as controls in the aforementioned study and our use of a computer‐based diagnostic system, which may have helped to more accurately measure the QRS waves and reduce interobserver variability.
The study also found a significantly larger prevalence of LVH in the MI group (56.2%) than the no MI or LBBB group (30.7%). This seems logical since (a) LVH secondary to essential hypertension is well known to significantly increase cardiovascular morbidity and mortality1, 2, 3 and (b) LVH may be the result of LV remodeling in patients with a previous MI. Furthermore this may, in part, explain the comparable classification accuracy of the ECG indices for LVH in MI patients and patients without MI by reasons of Bayesian considerations.
The electrocardiographical detection of LVH in the context of a past MI may therefore be of prognostic value, since an increase in the LV mass is indicative of LV remodeling which invariable translates into an elevated risk for developing congestive heart failure and worsening the long‐term prognosis. Nevertheless, the prognostic value of the ECG indices for LVH in patients with prior MI needs validation in further studies.
Despite our findings that the diagnostic performance of the LVH electrocardiographic indices is not significantly reduced in patients with previous MI its sensitivity is still low and this will invariably limit its use as a screening tool for LVH. Furthermore, being a more sensitive and specific diagnostic tool, echocardiography is nowadays regarded as the most useful tool for the detection of LVH. Nonetheless, many patients will undergo ECG testing for other reasons and it makes sense to use the information provided by this technique to assess the presence of LVH despite its limitations. In addition and in contrast to echocardiography, the low cost and readily availability of ECG testing constitute an important advantage.
Limitations of our study should be noted. We used 2D echocardiographic measurements, which can be less accurate than M mode echocardiography. Nevertheless, the technique used to measure was aimed to minimize errors in measurement and gave a prevalence of LVH similar to what has previously been reported in comparable samples.19, 20 In addition, 2D echocardiographic measurements are commonly used in clinical practice, which increases its possibility of extrapolation of the results and is more representative of effectiveness of the method. In the study by Hanna et al.9 a reduced sensitivity was found in patients with anterior and lateral wall MI when compared to anterior wall MI only. In our study no distinction was made between the localization of the MI and it is therefore possible that a subgroup analysis could have revealed differences in the diagnostic accuracy depending on the LV region affected by the MI. Also, the relatively limited sample size of patients with MI and the large number of criteria evaluated, including multiple combinations, makes it possible that some of the observed classification accuracies could have been influenced by random chance. Validation of the results in other larger population is therefore warranted. Finally, since all electrocardiographic measurements have been carried out using a computer‐assisted system that reduces measurement error, special care is necessary when a manual technique of measurement is used aimed at reproducing our findings.
In conclusion, our findings indicate that the presence of a previous MI does not limit the diagnostic accuracy of the ECG in the detection of LVH in patients with known or suspected heart disease, at least when a computer‐assisted internet platform is used. This type of diagnostic system constitutes an attractive approach since it allows to accurately and rapidly evaluate multiples LVH electrocardiographic indices in different clinical settings.
This work was supported by an unrestricted grant from Sanofi‐Aventis which also supported the development and implementation of the platform ELECTROPRES.
REFERENCES
- 1. Kannel WB. Prevalence and natural history of electrocardiographic left ventricular hypertrophy. Am J Med 1983;75:4–11. [DOI] [PubMed] [Google Scholar]
- 2. 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. [DOI] [PubMed] [Google Scholar]
- 3. Sundstrom J, Lind L, Arnlov J, et al. Echocardiographic and electrocardiographic diagnoses of left ventricular hypertrophy predict mortality independently of each other in a population of elderly men. Circulation 2001;103:2346–2351. [DOI] [PubMed] [Google Scholar]
- 4. Williams B, Poulter NR, Brown MJ, et al. British Hypertension S. Guidelines for management of hypertension: Report of the fourth working party of the British Hypertension Society, 2004‐BHS iv. J Hum Hypertens 2004;18:139–185. [DOI] [PubMed] [Google Scholar]
- 5. Chobanian AV, Bakris GL, Black HR, et al. National Heart L, Blood Institute Joint National Committee on Prevention DE, Treatment of High Blood P, National High Blood Pressure Education Program Coordinating C. The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment Of High Blood Pressure: The JNC 7 report. JAMA 2003;289:2560–2572. [DOI] [PubMed] [Google Scholar]
- 6. Mancia G, De Backer G, Dominiczak A, et al. Management of arterial hypertension of the European Society of H, European Society of C. 2007 guidelines for the management of arterial hypertension: The task force for the management of arterial hypertension of the European society of hypertension (ESH) and of the European Society Of Cardiology (ESC). J Hypertens 2007;25:1105–1187. [DOI] [PubMed] [Google Scholar]
- 7. Mancia G, Laurent S, Agabiti‐Rosei E, et al. European Society of H. Reappraisal of European Guidelines on Hypertension Management: A European Society of Hypertension task force document. J Hypertens 2009;27:2121–2158. [DOI] [PubMed] [Google Scholar]
- 8. MacFarlane PW. Comprehensive Electrocardiography: Theory and Practice in Health and Disease. Oxford, Pergamon Press, 1988. [Google Scholar]
- 9. Hanna EB, Glancy DL, Oral E. Sensitivity and specificity of frequently used electrocardiographic criteria for left ventricular hypertrophy in patients with anterior wall myocardial infarction. Proc (Bayl Univ Med Cent) 2010;23:15–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Martin‐Rioboo E, Lopez Granados A, Cea Calvo L, et al. Grupo de Hipertension Arterial de la Sociedad Andaluza de HTA, estudio P. [interobserver agreement on electrocardiographic diagnosis of left ventricular hypertrophy in hypertensive patients in andalusia. Prehvia study]. Aten Primaria 2009;41:248–254. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Willems JL, Abreu‐Lima C, Arnaud P, et al. Evaluation of ecg interpretation results obtained by computer and cardiologists. Methods Inf Med 1990;29:308–316. [PubMed] [Google Scholar]
- 12. Willems JL, Abreu‐Lima C, Arnaud P, et al. The diagnostic performance of computer programs for the interpretation of electrocardiograms. N Engl J Med 1991;325:1767–1773. [DOI] [PubMed] [Google Scholar]
- 13. Rodriguez‐Padial L, Rodriguez‐Picon B, Jerez‐Valero M, et al. Diagnostic accuracy of computer‐assisted electrocardiography in the diagnosis of left ventricular hypertrophy in left bundle branch block. Rev Esp Cardiol 2012;65:38–46. [DOI] [PubMed] [Google Scholar]
- 14. Institute for Medical Diagnostics . Validation according to IEC60601–2‐51 of the Hannover ECG system HES®. Version 1.0, 24–05‐2006.: Copyright @ Biosigna GmbH, 2000–2006.
- 15. Surawics B. Chou's Electrocardiography in Clinical Practice. Philadelphia, Saunders Elsevier, 2008. [Google Scholar]
- 16. Devereux RB, Alonso DR, Lutas EM, et al. Echocardiographic assessment of left ventricular hypertrophy: Comparison to necropsy findings. Am J Cardiol 1986;57:450–458. [DOI] [PubMed] [Google Scholar]
- 17. Devereux RB, Reichek N. Echocardiographic determination of left ventricular mass in man. Anatomic validation of the method. Circulation 1977;55:613–618. [DOI] [PubMed] [Google Scholar]
- 18. Koren MJ, Devereux RB, Casale PN, et al. Relation of left ventricular mass and geometry to morbidity and mortality in uncomplicated essential hypertension. Ann Intern Med 1991,114:345–352. [DOI] [PubMed] [Google Scholar]
- 19. Pewsner D, Juni P, Egger M, Battaglia M, et al. Accuracy of electrocardiography in diagnosis of left ventricular hypertrophy in arterial hypertension: Systematic review. BMJ 2007;335:711–719. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Ang DS, Fahey TP, Wright GA, Struthers AD. Development and validation of a clinical score to identify echocardiographic left ventricular hypertrophy in patients with cardio‐vascular disease. Am J Hypertens 2008;21:1011–1017. [DOI] [PubMed] [Google Scholar]