The objective of the present study was to compare anthropometry with bioelectrical impedance (BIA) in relation to densitometry (dual-energy X-ray absorptiometry; DEXA) as methods of nutritional assessment and body composition in out-patients with chronic pulmonary obstructive disease (COPD). We conducted a cross-sectional clinical study with sixty-one patients with COPD (forty-two men and nineteen women), mean age of 66·5 (sd 7·9) years and forced expiratory volume in 1s of 1·3 (sd 0·6) litres (52·2 (sd 19·8) % predicted), referred to the Pulmonary Rehabilitation Center. The patients were evaluated regarding nutrition status and body composition as determined by anthropometry, BIA and DEXA. In the results, 34·4% showed mild obstruction, 31·2%, moderate and 34·4%, severe obstruction. According to the BMI (mean 24·5 (sd 4·5) kg/m2), 45·9% of the patients exhibited normal weight, while 27·9% were underweight and 26·2% were obese. Related to fat-free mass (FFM), anthropometry and BIA compared with DEXA presented high correlations (r 0·96 and 0·95 respectively; p<0·001) and high reliability between the methods (α 0·98; p<0·001). Agreement analysis between the methods shows that anthropometry overestimates (0·62 (sd of the difference 2·89) kg) while BIA underestimates FFM (0·61 (sd of the difference 2·82) kg) compared with DEXA. We concluded that according to the nutritional diagnosis, half of our population of patients with COPD showed normal weight, while the other half comprised equal parts obese and underweight patients. Body composition estimated by BIA and anthropometry presented good reliability and correlation with DEXA; the three methods presented satisfactory clinical accuracy despite the great disparity of the limits of agreement.