Hyperinsulinemia during exercise in people with diabetes requiring exogenous insulin is a major clinical problem. The aim of this study was to assess the significance of portal vein versus arterial insulin to hepatic effects of hyperinsulinemia during exercise. Dogs had sampling (artery, portal vein, and hepatic vein) and infusion (vena cava and portal vein) catheters and flow probes (hepatic artery and portal vein) implanted >16 days before a study. Protocols consisted of equilibration (-130 to -30 min), basal (-30 to 0 min), and treadmill exercise (0-150 min) periods. Somatostatin was infused and glucagon and insulin were replaced in the portal vein to achieve basal arterial and portal vein levels at rest and simulated levels during the first 60 min of exercise. From 60 to 150 min of exercise, the simulated insulin infusion was sustained (C; n = 7), modified to selectively create a physiologic increment in arterial insulin (Pe; n = 7), or altered to increase arterial insulin as in Pe but with a concomitant increase in portal insulin (PePo; n = 7). Euglycemic clamps were performed in all studies. Portal and arterial insulin were 15 +/- 2 and 4 +/- 1 micro U/ml (mean +/- SE of all groups), respectively, at t = 60 min in all groups. Insulin levels were unchanged for the remainder of the exercise period in C. Arterial insulin was increased from 3 +/- 1 to 14 +/- 2 micro U/ml, whereas portal insulin did not change in Pe after t = 60 min. Arterial insulin was increased from 3 +/- 1 to 15 +/- 2 micro U/ml, and portal insulin was increased from 16 +/- 3 to 33 +/- 3 micro U/ml in PePo after t = 60 min. Endogenous glucose production (R(a)) rose similarly from basal during the first 60 min of exercise in all groups (mean +/- SE of all groups was from 2.2 +/- 0.1 to 6.8 +/- 0.5 mg. kg(-1). min(-1)). The increase in R(a) was sustained for the remainder of the exercise period in C. R(a) was suppressed by approximately 40%, but only after 60 min of hyperinsulinemia, and by approximately 20% after 90 min of hyperinsulinemia in Pe. In contrast, the addition of portal venous hyperinsulinemia caused approximately 90% suppression of R(a) within 20 min and for the remainder of the experiment in PePo. Measurements of net hepatic glucose output were similar to R(a) responses in all groups. Arterial free fatty acids (FFAs), a stimulus of R(a), were increased to 1,255 +/- 258 micro mol/l in C but were only 459 +/- 67 and 312 +/- 42 micro mol/l in Pe and PePo, respectively, by 150 min of exercise. Thus, during exercise, the exquisite sensitivity of R(a) to hyperinsulinemia is due entirely to portal venous hyperinsulinemia during the first 60 min, after which peripheral hyperinsulinemia may control approximately 20-40%, possibly as a result of inhibition of the exercise-induced increase in FFA.