GH therapy variably reduces the height deficit of short children born small for gestational age (SGA) but is associated with hyperinsulinemia. Intermittent, higher-dose GH regimens may be alternatives to continuous, lower-dose treatment. We examined whether the growth response to GH therapy is related to pre-treatment indices of endogenous somatotropic activity, and studied the reversibility of the insulin resistant state induced by GH. 13 non-GH deficient short SGA children were randomised to high-dose GH (100 mcg/kg/day) by daily sc injection (n = 9), or no GH treatment (n = 4) for 2 years (2/4 controls subsequently received GH treatment). Overnight GH profiles were performed at baseline; intravenous glucose tolerance tests were performed at baseline, yearly on GH treatment, and 3 months post-GH treatment. Fasting glucose, insulin and proinsulin levels were measured, insulin sensitivity estimated using Bergman's minimal model, and glucose tolerance calculated from rate of glucose disappearance. In all GH-treated children, gain in height SDS (mean gain yr 1 = +1.2 SDS) was inversely related to baseline peak overnight GH (r = -0.88, n = 10, p = 0.0008), IFG-I (r = -0.74, n = 11, p = 0.009), and fasting insulin levels (r = -0.71, n = 11, p = 0.014). GH treatment increased fasting glucose (means: baseline vs. yr 2: 3.7 vs. 4.4 mmol/l, p = 0.005), insulin (3.8 vs. 13.9 mU/l, p = 0.0002), and proinsulin levels (1.7 vs. 4.5 pmol/l, p = 0.004), and decreased insulin sensitivity (26.9 vs. 4.0 per min/mU/1 x 10(4), p = 0.002). Glucose tolerance initially decreased (baseline: 2.62 min(-1); yr 1: 2.18, p = 0.02; yr 2: 2.39, p = 0.12). However, by 3 months post-GH treatment significant improvements were seen in fasting insulin (post-GH: 5.2 mU/1, p = 0.0003 vs. yr 2), proinsulin (1.7 pmol/l, p = 0.002), and insulin sensitivity (17.6 per min/mU/1 x 10(4), p = 0.0001). Post-GH treatment, fasting glucose levels (4.1 mmol/l, p = 0.04) and glucose tolerance (2.49 min(-1), p = 0.4) were similar to baseline, and the slight increase in fasting insulin levels (5.2 mU/1, p = 0.04) was similar to that observed in non-GH treated children over the 2 yr study period (baseline vs. 2 years: 3.9 vs. 5.9 mU/1, n = 4). In conclusion, in this study of 13 short non-GH-deficient SGA children, high-dose GH therapy induced growth responses that were associated with reversible decreases in insulin sensitivity, and that were predicted by pretreatment markers of endogenous, but not stimulated, somatotropic activity.