General trends were investigated in the age-evolution of electrocardiogram (ECG) patterns using ECG data of North American white, black, and Hispanic men and women aged 25-74 years, derived from the Second National Health and Nutrition Examination Survey and the Hispanic Health and Nutrition Examination Survey. The data were stratified by race, sex, age, and obesity. There were substantial racial differences in ECG amplitudes. In general, ECG amplitudes and amplitude combinations used in left ventricular hypertrophy (LVH) criteria were larger in blacks than in whites. RV5 was smaller in all age groups of Hispanic women and in younger men, and RaVL was larger in Hispanics than in whites. There was, however, no significant difference in RaVL + SV3 amplitude between Hispanics and whites except in the youngest age group. The QRS frontal plane axis increased in men and women in all three racial groups by about 8 degrees per decade of age. Overweight was associated with 14 degrees more horizontal axis. RaVL and RaVL + SV3 amplitudes increased systematically with age and overweight, but there was an opposite tendency in RV5 amplitude. Multivariate logistic regression analyses indicated that age and the QRS axis had a relatively minor influence on LVH likelihood. Being black was associated with a more than fourfold excess of LVH by the Minnesota code criteria and a nearly threefold excess of LVH by the Cornell voltage criteria. Being overweight did not influence the odds ratio for LVH by the Cornell voltage criteria except in Hispanics, but the odds ratio for LVH was reduced by about 25% according to the Minnesota code criteria. The odds ratio of having LVH differed substantially for men and women by the Minnesota code and Cornell voltage criteria. Whereas there was a more than twofold excess likelihood of LVH in men by the Minnesota code criteria, there was a more than threefold excess of LVH likelihood in women by the Cornell voltage criteria when age, obesity, QRS axis, and race were simultaneously included as covariates in the logistic regression model.