The short insulin tolerance test (ITT) is both a simple and valid method of quantifying insulin sensitivity although arterialization of samples and the risk of hypoglycaemia remain as potential difficulties. We examined the safety and reproducibility of using venous sampling with insulin doses of 0.1 U kg-1 and 0.05 U kg-1 in healthy subjects. Whole blood glucose concentrations were measured contemporaneously and the rate of plasma glucose decline (mmol l-1 min-1) for each test was estimated from unlogged venous plasma glucose concentrations measured at 1 min intervals. The mean rates of plasma glucose decline for the 0.1 U kg-1 and 0.05 U kg-1 insulin doses were 0.26 mmol l-1 min-1 (n = 11, range = 0.17-0.41, intrasubject coefficient of variation (CV) = 9.4%) and 0.25 mmol l-1 min-1 (n = 6, range 0.19-0.46, intrasubject CV = 15.9%), respectively. Reversal of significant hypoglycaemia was necessary in one subject before 15 min post-insulin. We found that: (1) venous sampling provides a reproducible measure of glucose uptake after insulin, (2) contemporaneous bedside glucose sampling identifies those at risk of significant hypoglycaemia during the ITT, and (3) the 0.1 U kg-1 dose response is more reproducible and no less safe than the half dose response. We conclude that the current ITT protocol would be made safer and simpler with the above modifications although further studies comparing venous with arterialized sampling are needed.