Operating characteristics of the male hypothalamo-pituitary-gonadal Axis: Pulsatile release of testosterone and follicle-stimulating hormone and their temporal coupling with luteinizing hormone

To appraise the physiological pattern(s) of episodic testosterone and FSH release in man, we withdrew blood samples at 10-min intervals for 24–36 h in a total of 15 normal men. We subjected the resulting FSH (15 men) and testosterone (5 men) time series to 3 statistically based and mathematically independent procedures for detecting hormone pulsatility, viz. Cluster analysis, the Detect program, and Fourier transformation. The Cluster technique disclosed discrete testosterone and FSH peaks occurring at mean (±sem) interpulse intervals of 112 ± 14 and 85 ± 3.4 min, respectively. These values were not significantly different from the mean LH interpulse interval of 95 ± 11 min. The average durations of the testosterone and FSH pulsations were 90 ± 11 and 59 ± 3 min, respectively. The mean testosterone pulse amplitude reached a maximal value of 910 ± 92 ng/dL (31.5 ± 3.2 nmol/L), which represented a mean increase of 242 ± 26 ng/dL (8.4 ± 0.9 nmol/L) above the preceding nadir. FSH pulses had a maximum of 7.2 ± 0.3 IU/L, and an incremental amplitude of 1.3 ± 0.1 IU/L. An independent pulse detection procedure, Detect, yielded a testosterone pulse frequency of 12.3 ± 0.8 pulses/day [P = NS vs. Cluster program (13 ± 1.9 pulses/day)]. The Cluster and Detect estimates of FSH pulse frequency were also similar, viz. 16 ± 1.9 and 16 ± 0.6 pulses/day. Further analysis by Fourier transformation revealed significant circadian periodicities for serum testosterone, FSH, and LH, which had mean nyctohemeral amplitudes of 185 ng/dL (6.4 nmol/L), 0.38 IU/L, and 1.3 IU/L, respectively. Cross-correlation analyses disclosed significantly positive uncorrected cross-correlations between LH and testosterone that were maximal at a testosterone lag of 60 min (range, 50–70 min). To eliminate high intrinsic autocorrelations within the testosterone and LH time series, stepwise autoregressive fitting was employed. The resulting partial cross-correlation matrices indicated that LH concentrations at any given instant were significantly positively correlated to testosterone concentrations lagged by 10 and 20 min. Similarly, contemporaneous LH and FSH concentrations were significantly positively correlated (r = 0.40–0.89; P < 0.001). Moreover, autoregressive modeling disclosed significantly positive partial cross-correlations between LH and FSH at a FSH lag of 10 min. In summary, we have identified significant pulsatile as well as circadian (24-h) patterns of testosterone and FSH release in normal men. Moreover, cross-correlation and stepwise autoregressive modeling of simultaneous LH and testosterone or simultaneous LH and FSH time series revealed close temporal coupling between LH and testosterone concentrations, such that testosterone and FSH increases occur coincidently with LH increases or lag LH increases by only 10-20 min. (J Clin EndocrinolMetab 65: 929, 1987).