Estimates of in vivo insulin sensitivity (S1) can be derived from minimal model analysis of a frequently sampled intravenous glucose tolerance test (FSIVGTT). Modification of the FSIVGTT by administration of tolbutamide improves the correlation between S1 and more traditional glucose clamp measurements. Recently, a modified FSIVGTT using insulin infusion (instead of tolbutamide) has been described. This modification may be useful for studying subjects with poor insulin secretion. We directly compare the standard FSIVGTT with the insulin modified FSIVGTT in ten normal subjects (age 29 +/- 2 yr, BMI 22.5 +/- 1.2 kg/m2). Each subject received both tests. The order of the studies was randomized and the interval between paired studies was approximately one week. After an overnight fast, glucose (0.3 g/kg) was infused from time 0 to 2 min. For the modified FSIVGTT, insulin (4 mUkg-1.min-1) was infused from time 20 to 25 min. None of the subjects became hypoglycaemic after insulin infusion. Paired t-test analysis did not reveal significant differences between the two protocols for any of the minimal model parameters. The standard FSIVGTT gave estimates for S1 of 5.63 +/- 1.32 x 10(-4) min-1 per microU/ml while the insulin modified FSIVGTT gave estimates for S1 of 5.11 +/- 0.83 x 10(-4) min-1 per microU/ml. However, the fractional standard deviation for S1 was significantly smaller with the insulin modified protocol (3.6 +/- 1.0 vs. 22.2 +/- 9.0, p < 0.037). Thus, under our experimental conditions, insulin infusion during FSIVGTT appears to improve the precision of minimal model estimation of S1 in normal subjects.