A novel index of insulin sensitivity, the quick insulin sensitivity check index, termed QUICKI (1/[log (insulin) + log (glucose)]), was recently developed. We examined whether QUICKI accurately reflects changes in insulin sensitivity after exercise training, a perturbation known to improve insulin sensitivity. Sedentary, nondiabetic adults underwent a frequently sampled iv glucose tolerance test before and after 6 months of training. Insulin sensitivity was estimated from the glucose tolerance test using Bergman's minimal model (insulin sensitivity-minimal model), and QUICKI was calculated from basal insulin and glucose. Exercise increased (P = 0.003) insulin sensitivity-minimal model but did not change (P = 0.12) QUICKI. Before and after training, the rank-correlation between QUICKI and insulin sensitivity-minimal model was significant (r = 0.79, P = 0.0005; r = 0.56, P = 0.03, respectively). However, the rank-correlation between fasting insulin alone with insulin sensitivity-minimal model was as good (before training r = -0.77, P = 0.0009; after training r = -0.55, P = 0.03) as that between QUICKI and insulin sensitivity-minimal model. Fasting glucose was not related to insulin sensitivity-minimal model at either time. When difference scores (i.e. after pretraining values) were examined, neither QUICKI nor fasting insulin correlated with insulin sensitivity-minimal model (QUICKI vs. insulin sensitivity-minimal model r = 0.24, P = 0.39; fasting insulin vs. insulin sensitivity-minimal model r = -0.40, P = 0.14). We conclude that fasting insulin is equivalent to fasting insulin plus glucose (i.e. QUICKI) at estimating basal insulin sensitivity in nondiabetic adults. However, QUICKI does not accurately reflect exercise-induced changes in insulin sensitivity within individual subjects.