Performance of LH pulse-detection algorithms at rapid rates of venous sampling in humans

Abstract

To assess the influence of the sampling rate on the quantitative characterization of pulsatile luteinizing hormone (LH) release, we withdrew blood at 4-min intervals for 8 h in five men and at 1-min intervals for 2 h in six other men. For comparative purposes, significant LH pulses were enumerated by three independent, computerized pulse-detection algorithms currently available. Our results indicate that, although the absolute number of LH pulses detected was influenced by the particular algorithm used and the estimate of intra-assay variance, all three analyses yielded increased pulse-frequency estimates at more intensive rates of venous sampling. Moreover, using a fourth, modified pulse-detection algorithm intended to maximize recognition of true-positive LH pulses while minimizing both false-positive and false-negative pulses, we observed that venous sampling at 4- and 1-min intervals exposed 4- and 12-fold more LH pulses, respectively, than could be discerned at conventional sampling rates. At rapid rates of venous sampling, the pattern of LH pulses comprised high-frequency, low-amplitude LH pulsations superimposed on lower-frequency LH peaks. This pattern suggests that the pituitary gland is responsive to high rates of intermittent neural stimulation. Moreover, these observed profiles are consistent with rapid initial rates of LH disappearance and/or distribution that we could demonstrate after intravenous bolus injections of purified LH in hypogonadotropic volunteers. In conclusion, we have compared results from three different pulse-detection algorithms at various rates of venous sampling and demonstrated a critical influence of sampling rate on apparent LH pulse frequency in humans.