There is a large body of evidence that the electrocardiogram (ECG) is insensitive in the recognition of left ventricular hypertrophy (LVH), in comparison with the echocardiogram; however, its specificity is high. In this study we further analyzed the performance of the ECG in detecting LVH in 200 consecutive patients (124 men and 76 women, mean age 50.9 years) with mild to moderate essential hypertension, using echocardiographically determined left ventricular mass (LVM) as the standard for comparison. To test the hypothesis that, owing to the high number of true positive findings, the ECG may still be useful for clinical purposes by selecting subsets of hypertensives with higher degrees of LVH, we compared the mean values of LVM index corresponding to either positive (true positive) or negative (false negative) electrocardiographic signs of LVH. In this study 69 patients (34.5%) had echocardiographic LVH, as defined by a LVM index exceeding 125 g/m2 for men and 112 g/m2 for women. Almost all criteria demonstrated high levels of specificity (> or = 89%). In the whole group the Lewis index ((RI - RIII)+(SIII - SI) > or = 17 mm) showed a slight superiority in diagnosing LVH (sensitivity = 43%) in comparison to the remaining criteria; the confidence intervals estimate of sensitivities confirmed such diagnostic superiority only with respect to those criteria with a sensitivity < or = 17%. However, the use of McNemar's test to compare sensitivities of all electrocardiographic criteria at matched specificities (> or = 95%) did not show significant differences (P < .05).(ABSTRACT TRUNCATED AT 250 WORDS)