Optimal Left Ventricular Hypertrophy Classification and Quantification: Insights from Body Surface Potential Maps

Fred Kornreich

Bellingen, Belgium

Correspondence: Fred Kornreich, E-mail: fkornre@vub.ac.be, phone 32-2-360-25-28, fax 32-2-356-99-52

1.  Introduction

Left ventricular hypertrophy (LVH) is associated with increased risk of cardiovascular mortality, particularly when QRS high voltage is associated with ST-T abnormalities. Conventional criteria were designed to detect the presence or absence of LVH, and as well known, their sensitivity is low. In addition, these criteria have a limited utility in quantitative assessment of LVH or echocardiographic left ventricular mass (LVM). Body surface potential maps (BSPM) contain diagnostic information on LVH not present in conventional lead systems (Kornreich et al. 1988). The present report describes results from two different statistical approaches for extracting BSPM information, namely, discriminant analysis for separating normal subjects from patients with LVH and multiple regression analysis for prediction of LVM (Kornreich et al. 1990).

2.  Material and Methods

Discriminant analysis was performed on simultaneously recorded 120-lead BSPM data from 250 normal subjects aged 20 to 30 years and 122 patients with LVH. The criteria for the presence of LVH included cardiac catheterization with coronary arteriography and ventriculography, radionuclide angiography, chest X-rays and cardiac surgery. Echocardiography was available in 72 normal subjects and in 84 patients with LVH. Most patients in the pure LVH group had isolated aortic valve disease (60%); the remaining patients had hypertensive heart disease in the absence of confounding factors. After extracting optimal BSPM features and leads in a subgroup of 173 normal subjects and 122 patients with pure LVH, the model was then tested in the remaining 77 normal subjects and 92 patients with complicated LVH. Subsequent to discriminant function analysis, multiple regression analysis was used to detect optimal set of features and leads for prediction of echocardiographic LVM in 102 men and 54 women.

3.  Results

3.1. Discriminant Analysis

Discriminant analysis model with 6 features from 5 torso sites (Table 1) yielded LVH classification sensitivity of 94% at specificity of 95%. In comparison, a multivariate model for the standard 12 lead ECG correctly classified 86% of pure LVH and 83% of complicated LVH cases at specificity rates of 94% and 93%, respectively.

3.2. Multiple Regression Analysis

Multiple regression analysis for prediction of LVM selected 6 features from 3 torso sites in men and from the same 3 sites plus 2 others in women (Table 1). The model yielded correlation between echocardiographic and electrocardiographic estimates of LV mass of 0.89 and 0.88 in men and in women, with standard errors of the estimate (SEE) 31 g and 22 g, respectively. Regression analysis of the standard 12-lead ECG in the same groups produced a correlation of 0.76 in men and 0.75 in women, with SEE 48g and 35 g, respectively. Using the criteria reported by Casale et al. (Casale et al. 1987) yelded correlations of 0.67 (SEE 38 g) and 0.71 (SEE 51 g) in men and in women, respectively. Regression equations introduced by Rautaharju et al. (Rautaharju et al. 1996) produced correlations of 0.72 (SEE 50 g) and 0.70 (SEE 37 g) in men and in women, respectively.

4.  Discussion

Discriminant analysis of BSPM data and also of the 12-lead ECG produced a rather high level of LVH classification accuracy in comparison with commonly reported relatively poor level of sensitivity by the conventional LVH criteria. The more advanced state of LVH in the patient group of the present study largely explains to high level of performance of the ECG classifiers derived. Optimal features selected (amplitudes from early part of the ST segment and near the peak of the T wave as well as QRS and P wave durations) may also reflect special characteristics of the patient group. Optimal BSPM lead location for LVH classification outside of the locations of the conventional 12-leads explains why BSPM models in general perform better than the conventional leads for LVH classification.

The accuracy of prediction of LVM with multiple regression models derived from BSPM as well as from 12 lead ECG was relatively high. The utility of ECG-LVM as a continuous variable is likely to be valuable particularly in future ECG studies with risk analysis and also in connection with hypertension intervention studies. It is interesting to observe the similarity of the results achieved by using in the present study group earlier regression models derived from the standard 12 lead ECG, with only slightly different performances in LVM prediction (Casale et al. 1987, Rautaharju et al. 1996).

References

Kornreich F, Montague TJ, Rautaharju PM, Kavadias M, Horacek BM. Identification of best electrocardiographic leads for diagnosing left ventricular hypertrophy by statistical analysis of body surface potential maps. American Journal of Cardiology 62(17):1285-1291, 1988.

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 75:565-572, 1987.

Rautaharju PM, Manolio TA, Siscovick D, Zhou SH, Gardin JM, Kronmal R, Furberg CD, Borhani NO, Newman A for the Cardiovascular Health Study Collaborative Research Group. Utility of new electrocardiographic models for left ventricular mass in older adults. Hypertension 28(1):8-15, 1996.