Clinical testing of growth hormone (GH) sufficiency is a controversial area in endocrinology. Due to the episodic nature of endogenous GH secretion, diagnosis of GH deficiency has been defined as a failure to achieve normal GH levels in response to at least two stimuli. This testing is associated with significant patient morbidity and cost. We analyzed our experience over a 4-y period to determine whether clinical or biochemical variables could be used to predict the results of a specific GH testing procedure. Of 180 cases analyzed (67% male, mean age 8.89 ± 4.39 y, range neonate-16 y), eight cases had incomplete GH testing results. Of the remaining 172, 19 were GH deficient (GH level <7 ng/mL). Younger age, higher body mass index and a greater degree of bone age delay were characteristic of the GH-deficient population; however, none of these variables alone was of diagnostic utility. Serum IGF-I level was below the normal range for 81% of the GH deficient and 47% of the GH-sufficient children; and was the only single variable that provided a reasonable between-group distinction. Discriminant analysis resulted in development of a new variable, based on IGF-I z scores, chronologic age, degree of bone age delay, and body mass index, which would have allowed exclusion of GH deficiency without provocative testing for 58% of the GH sufficient population, whereas permitting the diagnosis of GH deficiency for all GH-deficient subjects. Our data are dependent on the IGF-I assay method and the clinical definition for GH deficiency; therefore, the calculated predictive values are not applicable to all clinical populations. However, our data provide a new perspective on the integration of IGF-I levels and clinical information in predicting GH sufficiency.