Evaluation of growth hormone (GH) secretion using pharmacological GH stimulation tests (GHSTs) remains current practice, although the reliability of GHSTs has been questioned and many pitfalls have been pointed out. We have analysed all the 6,373 GHSTs which led to the initiation of GH therapy in 3,233 children treated in France from 1973 to 1989. Eleven different pharmacological tests were used, and 62 out of the 66 theoretical pairs of tests were used at least once. The most frequent combination of tests was used in 12.7% of patients. Reliability of GH peak measured by comparing the results of two tests in the same patient was poor, as measured by intraclass correlation coefficients (all under 0.8). Multivariate analysis identified several parameters positively or negatively associated with peak plasma GH. We believe that several of these factors (i.e. weight standard deviation score (SDS), genetic target height SDS and nature of the agent) identify biases in the diagnosis of GH deficiency (GHD). In addition, we re-evaluated GH secretion in 208 young adults formerly treated with GH for childhood onset GHD. Peak plasma was superior or equal to 10 ng/ml in 81% of patients with former idiopathic GHD. We conclude that the current use of GHSTs as well as the criteria for idiopathic childhood GHD should be questioned.