|Home | About | Journals | Submit | Contact Us | Français|
Pewsner et al take no account of age or race when assessing accuracy of electrocardiography (ECG) for diagnosing left ventricular hypertrophy (LVH).1 Ignoring them reduces specificity,2 and results in ?disease of electrocardiographic origin? in screening programmes. In particular, the upper normal limits of QRS voltages in black men are greater than in white men, while the difference between black and white women increases with age.3
We routinely report upper limits of normal QRS voltage for RaVL, RaVL+SV3, and RV5+SV1 on all ECG referrals for hypertension. These upper limits of normal are calculated as the mean plus 2 standard deviations from data in Rautaharju et al.3 A diagnosis of LVH is suggested if any value is greater than the age, sex, and race adjusted upper normal limit. When we compared the Pewsner criteria with our criteria in a recent sample of patients (table(table),), the Pewsner criteria resulted in roughly twice as many diagnoses of LVH using the Sokolow-Lyon and Cornell indexes. Using just one positive criterion for a diagnosis of LVH also increased the number of diagnoses (table). The table shows the value of using RaVL alone.
Requiring only one of three criteria to be positive to diagnose LVH increases the false positive rate, but I know of no data on the effect of combining age, sex, and race adjusted ECG measurements. Theoretically, three independent tests that are each normally distributed?have 2.5% of measurements above the upper limit of normal?will provide a specificity of 92.7%. However, ECG measurements are not independent, so specificity will be higher. Table 2 of Pewsner's paper reports a median specificity for Sokolow-Lyon of 89%, Cornell 96%, Cornell product 85-97%, and Gubner 96%. If these four measurements were combined, the theoretical specificity would be about 75%, or 82% if the Cornell product were omitted. Even allowing for the correlation between the ECG measurements, the resulting specificity would be too low for a screening test.
Using magnetic resonance imaging to screen for LVH is impracticable. The alternative of combining ECG measurements to generate a test with a relatively low sensitivity but a high specificity is a pragmatic one.
Competing interests: None declared.