The use of univariate and multivariate techniques to derive estimates of insulin sensitivity from the insulin modified FSIGT were investigated in 12 Type 2 diabetic subjects aged (mean+/-S.D.) 59+/-9.5 years and BMI 28.1+/-2.2 kg m(-2), who underwent both a FSIGT and an isoglycemic hyperinsulinemic clamp. Reproducibility of the FSIGT was tested in four patients on three separate occasions. FSIGT data were assessed by both univariate and multivariate techniques. The sensitivity index for the FSIGT ranged from 0.162 to 3.292 (mean 1.378) x 10(-4) x l min(-1) mU(-1) for the univariate approach and from 0.163 to 2.727 (mean 1.378) x 10(-4) x l min(-1) mU(-1) for the multivariate method. Mean S(Iclamp) was 44.41 x 10(-4) x l(-2) min(-1) x mU(-1) (range 22.0-77.92). The correlation of the insulin sensitivity indices between the clamp and the FSIGT was 0.51 (P=0.056) for the univariate and 0.67 (P=0.017) for the multivariate analyses. Repeated FSIGTs showed a lower variability for the multivariate than for the standard approach.