The reproducibility of impedance measurements made using the Human-IM system (Dietosystem, Milan-Italy) was assessed in a group of normal and diabetic subjects on the basis of three tests made under the same experimental conditions on three consecutive days. A total of 22 normal subjects, 29 insulin-dependent (IDD) and 6 noninsulin dependent (NIDD) diabetic patients were included in the study. The coefficient of variation between the three successive tests ranged between 0 and 2.7% (normals 1.1 +/- 0.7%, IDD 1.6 +/- 0.7%, NIDD 1.1 +/- 0.4%), thus confirming the good reproducibility of the method in all groups. There was no significant difference between impedance measurement tests in the two group of diabetic patients with regard to TBW, FAT and FFM. NIDD patients differed from normal subjects due to higher FAT levels, whereas there was no significant difference between IDD patients and normal subjects.