Malnutrition in patients with severe respiratory insufficiency can lead to severe complications, justifying the use of objective nutritional assessment techniques, such as bioelectrical impedance analysis (BIA), which is an easy, noninvasive method of measuring body composition. The purpose of this study was to develop, and validate against dual-energy X-ray absorptiometry (DXA), a BIA formula to predict fat-free mass (FFM) specific for patients with chronic severe respiratory insufficiency. Seventy-five ambulatory patients (15 females and 60 males) with severe chronic respiratory insufficiency (obstructive and restrictive) aged 63.6+/-19.2 yrs (mean+/-SD), in a stable pulmonary and cardiac condition for > or = 2 months, were measured simultaneously with BIA and DXA. Patients younger than 45 yrs of age and with a body mass index > or = 32 kg x m(-2) were excluded. The best-fitting multiple regression equation to predict FFM = -6.06 +/- (height x 0.283) +/- (weight x 0.207) - (resistance x 0.024) +/- (sex (males=1, females=0) x 4.036), gave a correlation coefficient of r=0.952, slope+/-SEM 0.902+/-0.034, standard error of the estimate 1.670, and p<0.0001. The mean difference for FFM was 0.2+/-2.3 kg (mean+/-SD) and percentage fat mass was -0.7+/-3.8%. These results suggest that the bioelectrical impedance analysis formula specific to patients with severe respiratory insufficiency give a better correlation and smaller mean differences than 12 different bioelectrical impedance analysis formulae described in the medical literature. A prediction equation, validated against dual-energy X-ray absorptiometry and based on subjects with similar clinical characteristics, is more applicable to the patients with respiratory insufficiency than a formula developed for healthy subjects.