Recent developments in dual-energy X-ray absorptiometry (DXA) have rendered feasible the determination of whole-body composition in small laboratory animals by directly measuring fat, fat-free and mineral bone masses. Our aim was to evaluate this technique by cross-calibrating the DXA method with the carcass chemical analysis in a heterogeneous population of nondiabetic Wistar and diabetic GK rats (21 animals were used for precision error and reproducibility determinations and 26 were used for accuracy studies). We report that this technique is optimized for weights >200 g. The respective CV for lean mass, fat mass and percentage of fat mass determined in short-term or transversal studies was 1.1 [+ or -] 0.1, 3.0 [+ or -] 1.3 and 3.1 [+ or -] 0.4% (mean [+ or -] SD) respectively. Further, this technique is valid for rats weighing from 130 to 200 g by using three successive scans. In longitudinal studies, daily calibrations significantly increased the percentage of fat mass CV to 6.6 [+ or -] 3.3%, but it was significantly decreased to 3.0 [+ or -] 2.7% by the use of triplicate scans. The accuracy for DXA was excellent in reference to the chemical extraction technique ([r.sup.2] = 0.95 for percentage of fat mass, P < 0.0001), using an adjustment factor of 0.75 (limits of agreement between the two methods for percentage of fat mass = -1.72.3%). Mimicry of longitudinal changes in body composition with intraperitoneal injections of saline solution demonstrated a satisfactory detection of body component changes ([less than or equal to]2% of error for each final component analyzed, when increasing total lean mass by 11.8%). We conclude that DXA is appropriate for rat whole-body composition determination, allowing reliable long-term follow-up of individual animals for the first time. KEY WORDS: dual-energy X-ray absorptiometry rats body composition fat content in vivo technique