BACKGROUND We investigated the association between echocardiographic (Echo) left ventricular mass (LVM) indexed to body surface area (LVM/BSA) or height(2.7) (LVM/H(2.7)) and electrocardiographic (ECG) variables in 546 Chinese hypertensives. METHODS The study group was stratified by gender and by BMI into obese (BMI > or =28 kg/m(2)), overweight (BMI > or =24 kg/m(2) and BMI <28 kg/m(2)), and healthy weight (BMI <24 kg/m(2)) subgroups. Cornell voltage, Sokolow-Lyon voltage, maximum R amplitude in V(1)-V(6), Gubner-Ungerleider voltage, and the products of these amplitude variables with QRS duration were measured. RESULTS None of the ECG and Echo values showed a statistically significant difference between the obese and overweight subgroup; thus, we used only one cut-off point of BMI at 24 kg/m(2) for stratification of the hypertensive population. Cornell voltage, Cornell product and LVM/BSA were only affected by gender in our study. For hypertensives with BMI > or =24 kg/m(2), Cornell product was correlated with LVM/BSA and LVM/H(2.7) most significantly: correlation coefficients were approximately 0.45 for males and 0.40 for females and the correlation trended to be stronger as LVM/BSA or LVM/H(2.7) increased. However, a few ECG variables showed a weak correlation with LVM/BSA or LVM/H(2.7) in the hypertensives without left ventricular hypertrophy (LVH). A low sensitivity and high specificity of ECG criteria for the detection of LVH were also derived using receiver operating characteristic curves. CONCLUSIONS We conclude that Cornell product and Cornell voltage are the most convenient predictors for LVM/BSA with stratification only by gender. They are also the best parameters for predicting LVH in obese and overweight Chinese hypertensives, whereas estimation of LVM/BSA, LVM/H(2.7) by ECG is inaccurate in Chinese hypertensives without LVH. The cut-off point of BMI=24 kg/m(2) is suitable for stratification of body weight in further studies regarding Chinese hypertensives.