To determine the role of IGF-binding proteins in mediating the direct effects of recombinant human IGF-I on insulin requirements in type 1(insulin-dependent) diabetes mellitus, overnight changes in IGF-I, IGF-II, and IGF-binding protein-1, -2, and -3, collected under euglycemic conditions, were compared in nine subjects after double blind, randomized, sc administration of recombinant human IGF-I (40 μg/kg) or placebo at 1800 h. On both nights a somatostatin analog infusion (300 ng/kg·h) suppressed endogenous GH production, and three timed discrete GH pulses (total, 0.029 IU/kg·night) ensured identical GH levels. After recombinant human IGF-I administration, IGF-I levels and the IGF-I/IGF-binding protein-3 ratio increased [mean ± SEM: IGF-I, 401 ± 22 ng/ml; placebo, 256 ± 20 ng/ml (P = 0.0002); IGF-I, 0.108 ± 0.006; placebo, 0.074 ± 0.004 (P = 0.0003), respectively], and insulin requirements decreased (IGF-I, 0.12 ± 0.03; placebo, 0.23 ± 0.03 U/kg·min; P = 0.008). The normal within-individual inverse relationships between insulin and IGF-binding protein-1 levels were observed (lag time 2 h: r = -0.34; P < 0.01). Yet despite reduced free insulin levels (8.5 ± 1.5; placebo, 12.2 ± 1.2 mU/liter; P = 0.03), IGF-binding protein-1 levels were reduced after recombinant human IGF-I administration (53.7 ± 6.8; placebo, 82.2 ± 11.8 ng/ml; P = 0.008). The largest reductions in free insulin levels after recombinant human IGF-I and thus putative improvement in insulin sensitivity occurred in subjects with the smallest increase in the plasma IGF-I/IGF-binding protein-3 ratio (r = 0,7; P = 0.03). Taken together, these data are consistent with the hypothesis that transcapillary movement of IGF-I (perhaps mediated by IGF-binding protein-1), out of the circulation facilitates altered insulin sensitivity. These data have important implications for risk-benefit assessment of recombinant human IGF-I therapy in type 1 diabetes mellitus.