GH stimulation tests are widely used in the diagnosis of GH deficiency (GHD), although they are associated with a high false positive rate. We have examined, therefore, the performance of other tests of the GH axis [urinary GH excretion, serum insulin-like growth factor I(IGF-I), and IGF-binding protein-3 (IGFBP-3) levels] compared with GH stimulation tests in identifying children defined clinically as GH deficient. Group I comprised 60 children (mean age, 10.3 +/- 4.8 yr) whose diagnosis of GHD was based on a medical history indicative of pituitary dysfunction (n = 43) or on the typical phenotypic features and appropriate auxological characteristics of isolated GHD (n = 17). Group II comprised 110 short children (mean age, 9.8 +/- 4 yr) in whom GHD was not suspected, but needed exclusion. The best sensitivity for a single GH test was 85% at a peak GH cut-off level of 10 ng/mL, whereas the best specificity was 92% at 5 ng/mL. The sensitivities of IGF-I, IGFBP-3, and urinary GH, using a cut-off of -2 SD score were poor at 34%, 22%, and 25%, respectively, with specificities of 72%, 92%, and 76% respectively. Only 2 of 21 pubertal children in group I and none of the 27 subjects with radiation-induced GHD had an IGFBP-3 SD score less than -1.5. We devised a scoring system based on the positive predictive value of each test, incorporating data from the GH test and the IGF-I and IGFBP-3 levels. A specificity of 94% could be achieved with a score of 10 or more (maximum 17) (sensitivity 34%). The latter could not be improved above 81% with a score of 5 points or more (specificity, 69%). A high score was, therefore, highly indicative of GHD, but was achieved by few patients. A normal IGFBP-3 level, however, did not exclude GHD, particularly in patients with radiation-induced GHD and those in puberty. A GH test with a peak level more than 10 ng/mL was the most useful single investigation to exclude a diagnosis of GHD.