OBJECTIVE All of the presently used methods for in‐vivo determination of body composition have inherent methodological errors and depend on various assumptions. We have therefore compared several different methods used to measure body fat in adult GH deficiency during GH treatment.DESIGN Comparison of body composition data from a two‐phase trial with an initial placebo‐controlled, double‐blind 6‐month period, followed by open treatment with GH until all patients had received GH for 12 months.PATIENTS Twenty‐five patients with known GH deficiency entered the study. Baseline examinations were complete in 23 patients, and 22 patients (16 males, 6 females) completed all examinations after treatment.MEASUREMENTS Body fat calculated from total body potassium (TBK) by whole‐body40K counting, total body water (TBW) by tritium dilution, total body nitrogen (TBN) by neutron activation, and bioelectric impedance (BIA) measurements were compared to body fat determinations by dual‐energy X‐ray absorptiometry (DEXA) in two‐compartment and multicompartment body composition models.RESULTS At baseline, DEXA fat mass agreed well at group level with measurements based on TBW or TBK alone, in a four‐compartment model based on TBK and TBW, and a multicompartment model based on bone mineral (by DEXA), TBN and TBW. Body fat by BIA agreed less well. After 12 months of GH treatment, body fat decreased by all methods used. This decrease was smaller by DEXA than by the other methods. The four‐compartment model based on TBK and TBW, and TBW alone, showed the best agreement with changes in DEXA fat.CONCLUSION All methods showed a decrease of body fat with GH treatment, but variati