We have recently reported that during infusion of commercially available [6-3H]glucose, a radioactive nonglucose contaminant may accumulate in plasma causing errors in the measurement of glucose turnover. To determine whether purification of this tracer by HPLC (high-performance liquid chromatography) before infusion would eliminate the contaminant in plasma and remove the underestimation of glucose turnover reported during hyperinsulinemia, four normal subjects each underwent two 5-h euglycemic clamps during infusion of insulin (1 mU.kg-1.min-1). Glucose turnover was measured with either commercially available [6-3H]glucose or with HPLC-purified [6-3H]glucose. HPLC analysis of samples from the clamps done with commercially available [6-3H]glucose showed that 9.7% of the infused tracer and 26% of the "plasma glucose 3H radioactivity" were contaminants. In contrast, no contaminant was observed in the plasma during infusion of HPLC-purified [6-3H]glucose. During the last hour of the clamp, mean glucose turnover using commercially available [6-3H]glucose was less (P less than 0.01) than the mean glucose infusion rate (7.6 +/- 0.3 vs. 10.5 +/- 0.3 mg.kg-1.min-1) yielding apparent "negative" (P less than 0.001) hepatic glucose release. In contrast, when HPLC-purified [6-3H]glucose was employed, glucose turnover equaled the glucose infusion rate (10.4 +/- 0.9 vs. 10.2 +/- 0.9 mg.kg-1.min-1) and hepatic glucose release was no longer negative. We conclude that removal of a tritiated nonglucose contaminant in [6-3H]glucose by HPLC yields correct estimations of glucose turnover at steady state.