We assessed the utility of measuring the physiological levels of gonadotropins as a diagnostic tool for pubertal onset in girls. Two methods of gonadotropin measurements were compared: the standard frequent sampling method, in which blood samples were drawn every 20 min, and the multiple integrated sampling method, in which samples were obtained continuously at 30 min intervals by a withdrawal pump. The two methods were examined simultaneously overnight in a group of eight girls at different stages of puberty. The following parameters of both LH and FSH secretion, calculated by PULSAR program, were highly correlated between these methods: area under the curve (AUC), mean levels, mean from smoothed baseline and mean peak height. The diagnostic value of multiple integrated sampling of gonadotropins (performed over 6 h) was then assessed in five prepubertal girls and six girls at early puberty (Tanner stages 2 and 3), in whom peak gonadotropins levels in response to GnRH stimulation test were in the prepubertal range. Several parameters of LH (but none of FSH) were significantly higher (p<0.05) in early pubertal compared to prepubertal girls: AUC (5.61 +/- 2.40 vs 2.39 +/- 1.41), mean levels (0.52 +/- 0.21 vs 0.23 +/- 0.14), smoothed mean level (0.43 +/- 0.18 vs 0.18 +/- 0.11) and peak area (0.27 +/- 0.08 vs 0.11 +/- 0.06). We conclude that the technically simple method of multiple integrated sampling is useful in detecting pubertal transition and is superior to the GnRH-stimulation test. This method can be used in selective cases when the stimulation test yields equivocal results.