BACKGROUND Left atrial (LA) enlargement has been reported in the obese. However, its prevalence in the healthy obese, clinical correlates and relation to left ventricular (LV) mass and diastolic function have been little investigated. METHODS Thirty-five consecutive, healthy, normotensive obese (body mass index greater than 28, mean +/- SD 34.2 +/- 2.3 kg/m2) and 35 nonobese subjects (body mass index 24.6 +/- 2.3 kg/m2) comparable in age and sex underwent echocardiographic measurements of LA posteroanterior (parasternal view), mediolateral and superoinferior (apical views) dimensions, aortic root diameter, LV mass and Doppler assessment of LV diastolic function. LA enlargement was defined as a posteroanterior dimension greater than 40 mm. A ratio of LA posteroanterior dimension to aortic root diameter greater than 1.4 was used as an index for disproportionate LA enlargement. RESULTS LA enlargement was more frequent in the obese than in the nonobese (37% versus 6%, P<0.0001). Similarity, disproportionate LA enlargement was more frequent in the obese (34% versus 6%, P<0.0001). LA posteroanterior dimension correlated well with body mass index (r=0.52, P<0.0001) and LV mass (r=0.56, P<0.0001), and weakly with blood pressure (r=0.28, P<0.02). There was no significant correlation with LV diastolic function, age or sex. In multivariate analysis (multiple r=0.61, P<0.0001), LA posteroanterior dimension correlated significantly only with mass (P<0.005), and the association with body mass index and blood pressure became nonsignificant. Similar results were obtained when LA posteroanterior dimension was replaced with mediolateral or superoinferior dimensions. CONCLUSIONS LA enlargement is frequent in the normotensive, otherwise healthy obese and correlates well with LV mass. It is not mediated through impairment of LV diastolic function, and likely reflects a physiological adaptation of the heart to the obese state. Further studies are needed to determine whether LA enlargement in the obese is associated with adverse long term outcome.