Insulin is secreted as a series of punctuated secretory bursts superimposed on variable basal insulin release. The contribution of these secretory bursts to overall insulin secretion has been estimated on the basis of peripheral vein sampling in humans to encompass > or =75% of overall insulin release. A similar contribution of the pulsatile mode of release was inferred in a canine model by use of portal vein sampling. The primary regulation of insulin secretion is through perturbation of the mass and frequency of these secretory bursts. The mode of delivery of insulin into the circulation seems important for insulin action; therefore, physiological conditions that alter the pattern of insulin release may affect insulin action through this mechanism. Transhepatic intraportal shunt in humans may provide access to portal vein samples, thus potentially improving the sensitivity of detecting and quantitating the frequency, mass, and amplitude of secretory bursts along with basal release and the regularity of these variables. To establish the insulin-secretory mechanism in nondiabetic humans by the use of portal vein sampling, we here assessed the mass, frequency, amplitude, and overall contribution of pulsatile insulin secretion by deconvolution analysis of portal vein insulin profiles. We find that, in nondiabetic humans fasted overnight, the portal vein insulin concentration oscillates at a periodicity of 4.1 +/- 0.2 min/pulse and with secretory peak amplitudes averaging 660% of basal (interpulse) release. The frequency was confirmed by spectral and autocorrelation analyses. The punctuated insulin-secretory bursts partially overlap and are responsible for the majority (70 +/- 4%) of insulin release. After ingestion of a mixed meal, the insulin release was increased through amplification of the secretory burst mass (507 +/- 104 vs. 1,343 +/- 211 pmol x l(-1) x min(-1), P < 0.001), whereas frequency (4.4 +/- 0.2 vs. 4.3 +/- 0.2, P = 0.86) and basal secretion (62 +/- 14 vs. 91 +/- 22 pmol x l(-1) x min(-1), P = 0.33) were unaffected. One subject with diabetes and cirrhosis had a similar insulin-secretory pattern, whereas a subject with insulin-dependent diabetes mellitus and minimal insulin release had preserved pulsatile release. A single subject was entrained to show agreement between entrained frequency and portal vein insulin oscillations. We conclude that insulin release in the human portal vein occurs at a mean periodicity of 4.4 +/- 0.2 min with a high signal-to-noise ratio (pulse amplitude 660% of basal). The impact of noise on the detected high frequency cannot be excluded.