Background and Aims Homeostatic model assessment (HOMA) is used to determine insulin sensitivity and beta-cell function from fasting glucose and insulin concentrations. The model has recently been updated: to include proinsulin in the radioimmunoassay insulin model and renal glucose losses and the insulin secretion curve has been modified to allow for an increase in insulin secretion in response to hyperglycemia. The aims are (1) to compare estimates of insulin sensitivity (%S) and beta-cell function (%B) derived from the updated HOMA model (HOMA2) with estimates from euglycemic and hyperglycemic clamps. (2) To examine the effect of using 3 plasma samples at 5minute intervals versus using a single sample on intra-subject reproducibility.
Subjects and Methods 30 subjects with diet-controlled type 2 diabetes underwent basal sampling for HOMA followed by a hyperglycemic clamp and hyperinsulinemic euglycemic clamp.
Results 24 male, 8 female, median duration of diabetes 2.54 (IQR: 0.9-3.8)years. Mean (SD): age 61.8 (7.7) years, BMI 29.5 (4.0)kg/m2, HbA1c 6.7 (0.9)%, fasting glucose 7.6 (1.5)mmol/l; adiponectin 24.5 (11.3)ng/ml, M/I 26.8 (12.1)L/kg/min. Geometric mean (SD range): fasting plasma insulin 93.5 (54.2-161.4)pmol/l, HOMA%B 64.7 (41.2-101.7)%, HOMA%S 52.9 (30.7-91.1)%. There was a strong correlation between HOMA2%B and the steady state insulin secretion during the final 30 minutes of the hyperglycemic clamp following logarithmic transformation (r=0.92, p≤0.0001); and between Log HOMA%S and log M/I derived from the euglycemic (r=0.82, p≤0.0001). There was a similar degree of correlation between adiponectin and M/I (r=0.52, p=0.0046) and adiponectin and HOMA-%S (r=0.53, p=0.0032). Data show near perfect correlations between HOMA%B and HOMA%S computed from the mean of three basal samples at 5min intervals and from a single basal sample (r=0.99, p≤0.0001). However the use of a single sample gives intra-subject CVs of 10.3% for HOMA%S and 7.6% for HOMA%B compared to 5.8% and 4.3% respectively when the mean of three samples is used.
Conclusion Estimates derived from the HOMA2 model correlate well with estimates from clamps. Perfect correlations would not be expected since HOMA yields basal estimates whereas clamps give estimates at the maximally stimulated end of the dose response curve. The use of single sample yields similar results as the mean of 3 samples, although the intra-subject CV improves with 3 samples.