In normal subjects, endogenous glucose production (EGP) is usually assumed to be completely suppressed during euglycaemic clamp studies performed at high insulin levels (> 100 mU L-1). However, this assumption is based on non-steady-state tracer measurements of EGP which are prone to negative errors. We have used purified [6-(3)H]glucose in an optimal tracer infusion protocol to assess the suppression of EGP during 4 h euglycaemic clamps in eight normal men. An insulin infusion rate of 5 mU kg-1 min-1 was chosen to achieve supraphysiological (> 500 mU L-1) plasma insulin concentrations. Using a labelled exogenous glucose infusion, plasma glucose (mean +/- SEM 5.3 +/- 0.1 mmol L-1) and glucose specific activities (mean 100 +/- 3% of basal) were maintained constant from 80 to 240 min. During hyperinsulinaemia, isotopically determined glucose appearance rates (Ra) were greater than glucose infusion rates (GIR) throughout the euglycaemic clamp period (P < 0.001) and EGP (Ra-GIR) was always greater than zero. In seven of the eight subjects studied EGP was partly suppressed but showed a wide variation (EGP 5 to 91% of basal at 80-120 min and 12 to 87% of basal at 200-240 min) while in one subject EGP rose above basal (by 72% at 80-120 min and 49% at 200-240 min). We conclude that EGP is not completely suppressed during euglycaemic clamps at high insulin levels.