Assessment of the left ventricular mass (LVM) from electrocardiograms may be improved by the addition of clinical variables into a multivariate equation. As the heart-thorax distance may affect the results, its relationships with electrocardiographic and clinical data have been evaluated in a group of 220 subjects (53 +/- 15 years, 126 female, 175 without demonstrated heart disease) who were assessed for echocardiographic LVM and heart-thorax distance. Sokolow, Cornell, and total QRS voltage indexes were obtained. Multiple regression equations with LVM as the dependent variable were fit, with an ECG index, body mass index (BMI), age, and gender as the independent predictors. Each of the 3 ECG indexes, BMI, age, and sex was shown to be independent predictors of LVM, with the ECG and BMI contributing with most of the explanatory power. When added to the model, the distance from the interventricular septum to the precordium (septal-LVD) was not a predictor of LVM, but when BMI was withdrawn, septal-LVD became an independent predictor of LVM (P < .001). This was not observed when septal-LVD was substituted for any other clinical or ECG variable, thus suggesting that septal-LVD accounts for information contained in BMI but not in the remaining variables. In addition, the distance from the center of LV to the precordium (mid-LVD) achieved significance as an independent LVM predictor, although the coefficient of multiple determination (R) practically did not change. Almost identical results are obtained when LVM is indexed for body surface area. Body mass index supplies virtually all the information contained in the heart-thorax distance.