The objective was to compare measures from dual-energy X-ray absorptiometry (DXA), bioelectrical impedance analysis (BIA) and anthropometry with a reference four-compartment model to estimate fat mass (FM) and fat-free mass (FFM) changes in overweight and obese women after a weight-loss programme. Forty-eight women (age 39.8 ± 5.8 years; weight 79·2 ± 11·8 kg; BMI 30·7 ± 3·6 kg/m2) were studied in an out-patient weight-loss programme, before and after the 16-month intervention. Women attended weekly meetings for the first 4 months, followed by monthly meetings from 4 to 12 months. Body composition variables were measured by the following techniques: DXA, anthropometry (waist circumference-based model; Antrform), BIA using Tanita (TBF-310) and Omron (BF300) and a reference four-compartment model. Body weight decreased significantly ( − 3·3 (sd 3·1) kg) across the intervention. At baseline and after the intervention, FM, percentage FM and FFM assessed by Antrform, Tanita, BF300 and DXA differed significantly from the reference method (P ≤ 0·001), with the exception of FFM assessed by Tanita (baseline P = 0·071 and after P = 0·007). DXA significantly overestimated the change in FM and percentage FM across weight loss ( − 4·5 v. − 3·3 kg; P < 0·001 and − 3·7 v. − 2·0 %; P < 0·001, respectively), while Antrform underestimated FM and percentage FM ( − 2·8 v. − 3·3 kg; P = 0·043 and − 1·1 v. − 2·0 %; P = 0·013) compared with the four-compartment model. Tanita and BF300 did not differ (P>0·05) from the reference model in any body composition variables. We conclude that these methods are widely used in clinical settings, but should not be applied interchangeably to detect changes in body composition. Furthermore, the several clinical methods were not accurate enough for tracking body composition changes in overweight and obese premenopausal women after a weight-loss programme.