To assess whether we could predict left ventricular (LV) geometric patterns on echocardiography from voltages in standard electrocardiograms (ECG) in patients with essential hypertension, standard 12-lead ECG and echocardiograms were recorded in 106 consecutive, untreated patients (50+/-11 yr old) with essential hypertension. Subjects were assigned to the following four groups based on relative wall thickness (RWT) and LV mass index (LVMI) as determined by echocardiography: a normal geometry group (n = 44), a concentric remodeling group (increased RWT and normal LVMI, n = 10), an eccentric hypertrophy group (increased LVMI and normal RWT, n = 23), and a concentric hypertrophy group (increased RWT and LVMI, n = 29). The following ECG variables were determined: Sokolow-Lyon voltage (SV1 = RV5: SL), Cornell voltage (RaVL + SV3: CN), sum of 12-lead QRS voltage (12-lead sum), and RV6/RV5 ratio (RV6/V5). LVMI correlated with SL, CN, and 12-lead sum, but not with RV6/V5 in the study group as a whole. The concentric hypertrophy group showed increased voltages for all ECG variables except RV6/V5. The concentric remodeling group showed increased voltages for SL and 12-lead sum, but a decreased RV6/V5 ratio. In contrast, the eccentric hypertrophy group had increased voltage only for the 12-lead sum. The combination of SL, RV6/V5, and CN showed modest sensitivity and specificity in the diagnosis of concentric remodeling, concentric hypertrophy, and normal geometry, but not in the diagnosis of eccentric hypertrophy. Conventional ECG criteria can predict LVMI, but not LV geometry in the patients with essential hypertension. The combination of SL, CN, and RV6/V5 is useful in differentiating the four LV geometric patterns seen in essential hypertension.