We assessed the reliability of multiple electrocardiographic variables for detecting left ventricular hypertrophy in 100 patients (aged 23 to 92 years, mean age 39 +/- 14) with complete left bundle branch block and different underlying cardiac diseases. Left ventricular hypertrophy, defined as an echocardiographically evaluated left ventricular mass greater than 241 g, was present in 66 of the 100 patients. The electrocardiographic parameters with the highest sensitivity were both the Cornell voltage criteria (RaVL + SV3 greater than 28 mm in men and greater than 20 mm in women) and the combination of criteria proposed by Kafka (any of these four indexes: RaVL greater than or equal to 11 mm, QRS axis -40 degrees or less, SV1 + RV5 or RV6 greater than or equal to 40 mm, SV2 greater than or equal to 30 mm and SV3 greater than or equal to 25 mm), with a sensitivity of 77%. Both criteria had a very low specificity (32 and 35%, respectively). The high specificities (greater than or equal to 88%) of several electrocardiographic criteria were accompanied by ineffective low sensitivities (less than 35%). Moreover, the cumulative parameters of Kafka and Cornell voltage criteria achieved a sensitivity of 84 and 89%, respectively, in hypertensive patients and in those with valvar diseases. None of the electrocardiographic indexes tested showed a significant difference in sensitivity when applied in categories of patients with left ventricular hypertrophy and different left ventricular geometry (cavity dilation or concentric hypertrophy). These data indicate that both conventional and recently proposed electrocardiographic criteria for left ventricular hypertrophy in the presence of left bundle branch block poorly recognize an augmented left ventricular mass.