Analytical problems in detecting rapid insulin secretory pulses in normal humans.

The present study was undertaken to define the optimal experimental and analytical conditions necessary to reproducibly detect, in the systemic blood, small-amplitude rapid oscillations (period 8-15 min) of insulin and C-peptide. Samples for insulin, C-peptide, and glucose were drawn at 2-min intervals for 2 h from six normal subjects during constant glucose infusion and from five of the same subjects under basal conditions. To reduce measurement error, insulin and C-peptide levels were measured 16 times at each time point. Three algorithms for pulse analysis (ULTRA, Cluster, PulseFit) were used to identify significant pulses, whereas autocorrelation and spectral analysis were used to identify potential regular periodic components in the data. In three of the five subjects studied under basal conditions, regular rapid oscillations could be consistently detected by autocorrelation when the analysis was based on eight replicates but not on duplicate series. In the remaining two basal studies and in all studies during glucose infusion, the majority of profiles did not have a significant periodic component. However, formal pulse analysis demonstrated that the number of pulses was similar during glucose infusion and basal conditions. Reproducibility was enhanced by increasing the number of replicates used in the analysis. We conclude that in the analysis of small-amplitude rapid insulin and C-peptide oscillations, the sensitivity and specificity of the analysis is likely to be enhanced by performing multiple estimations at each time point and by using a minimum of two contrasting analytical approaches for pulse detection, which incorporate a method evaluating periodicity in conjunction with a pulse detection program designed to evaluate each individual oscillation separately.