Current issues of ACP Journal Club are published in Annals of Internal Medicine


An electrocardiographic algorithm improved sensitivity for detecting left ventricular hypertrophy

ACP J Club. 1994 Jan-Feb;120:18. doi:10.7326/ACPJC-1994-120-1-018

Source Citation

Norman JE Jr, Levy D, Campbell G, Bailey JJ. Improved detection of echocardiographic left ventricular hypertrophy using a new electrocardiographic algorithm. J Am Coll Cardiol. 1993 Jun;21:1680-6.



To devise and test an electrocardiographic (ECG) voltage criterion for detecting left ventricular hypertrophy that is sex specific and adjusts for age and obesity.


Blinded comparison of ECG and echocardiographic recordings from the Framingham Heart Study. Half of the participants were chosen randomly for model construction and the other half, for evaluation.


A white community in Massachusetts.


1468 men and 1883 women who had an echocardiogram and a 12-lead resting ECG at baseline. Exclusion criteria were the Wolff-Parkinson-White syndrome, bundle-branch block, and a history of myocardial infarction.

Description of test and diagnostic standard

Computerized ECGs were done on a 3-channel simultaneous system, and the Cornell ECG voltage was calculated (the sum of voltages for the R wave of lead aVL and the S wave of lead V3). Standard 12-lead configuration and XYZ orthogonal leads were recorded in analog form, were digitized, and were read. A multiple linear regression algorithm adjusted the Cornell voltage criterion for age and body mass index (BMI) for men and women separately. The M-mode echocardiogram measured left ventricular mass (LVM). Left ventricular hypertrophy was defined as height-normalized LVM values that were 2 standard deviations above the mean values for healthy persons (143 g/m for men and 102 g/m for women).

Main outcome measure

Receiver operating characteristic (ROC) curves for adjusted and unadjusted Cornell voltages.

Main results

Adjusting for age and BMI increased the sensitivity at all specificities (Table). For algorithms derived from and applied to the total population at 90% specificity, adjustment improved the sensitivity from 30% to 55% for women and from 33% to 42% for men (Table). At 98% specificity, adjustment improved the sensitivity from 11% to 23% for women and from 11% to 18% for men (Table). Likelihood ratios are shown in the Table. Adjustment also improved sensitivity for moderate and severe hypertrophy. At 95% specificity, sensitivity for moderate hypertrophy increased from 18% to 31% for women and from 15% to 25% for men; for severe hypertrophy, sensitivity increased from 26% to 62% for women and from 27% to 43% for men (Table). Likelihood ratios are shown in the Table.


Adjustment for age and body mass index improved the sensitivity at all levels of specificity for men and women in the diagnosis of left ventricular hypertrophy using an algorithm based on QRS voltage calculations from computerized electrocardiographic recordings.

Source of funding: Not stated.

For article reprint: Dr. J.E. Norman Jr, National Heart, Lung, and Blood Institute, Room 3a08, Federal Building, Bethesda, MD 20892, USA. FAX 301-480-5298.

Table. Test features for adjusted and unadjusted Cornell voltage algorithm for detecting left ventricular hypertrophy

Outcomes Sex Adjustment Sensitivity Specificity +LR -LR
Any left ventricular hypertrophy Women Yes 55% 90% 5.5 0.5
No 30% 90% 3.0 0.8
Yes 41% 95% 8.2 0.6
No 19% 95% 3.8 0.9
Yes 23% 98% 11.5 0
No 11% 98% 5.5 0.9
Men Yes 42% 90% 4.2 0.6
No 33% 90% 3.3 0.7
Yes 28% 95% 5.6 0.8
No 19% 95% 3.8 0.9
Yes 18% 98% 9.0 0.8
No 11% 98% 5.5 0.9
Moderate left ventricular hypertrophy Women Yes 31% 95% 6.2 0.7
No 18% 95% 3.6 0.9
Men Yes 25% 95% 5.0 0.8
No 15% 95% 3.0 0.9
Severe left ventricular hypertrophy Women Yes 62% 95% 12.4 0.4
No 26% 95% 5.2 0.8
Men Yes 43% 95% 8.6 0.6
No 7% 95% 5.4 0.8

*LRs defined in Glossary and calculated from data in article


The study by Norman and colleagues is based on an excellent data set and methods. The population is appropriate; the echocardiograms and ECGs were recorded and measured using excellent modern technology. The statistical analysis is appropriate. This study sets the standard for ECG diagnosis of left ventricular hypertrophy. The simple algorithm the investigators developed and refined should be implemented in all commercial, computerized ECG recorders so that the probability of left ventricular hypertrophy is estimated by them.

It is disappointing, however, that 2 other potential ECG measurements (QRS duration and total QRS voltage) did not improve the correlation of the equation with LVM. The investigators should consider using intrinsicoid deflection, because this might help differentiate left ventricular dilation from increased LVM. I am not aware of data showing that ECG evidence of left ventricular hypertrophy can be reversed by antihypertensive treatment, so it is unclear to me why the authors included this in their summary.

Congratulations to the authors on an excellent study. Their equation should be an internal medicine board question, and this paper should be required reading for any physician interpreting ECGs.

Victor F. Froelicher, MD
Palo Alto Medical CenterPalo Alto, California, USA