Reproducibility and discrimination of different indices of insulin sensitivity and insulin secretion

Insulin sensitivity and insulin secretion can be estimated by multiple indices from fasting blood samples or blood samples obtained during oral glucose tolerance tests. The test-retest reliability of these indices in repeated measurements within the same individuals can strongly vary. We analyzed data of persons without diabetes who underwent two repeated OGTTs. For each measurement pair, we calculated multiple commonly used indices for the assessment of insulin secretion and insulin sensitivity. We then evaluated the coefficient of variation (standard deviation/mean) and discriminant ratio for each index. 89 persons underwent two OGTTs with a median interval of 86 days (IQR 64-249). Among indices of insulin sensitivity derived from fasting blood samples, the revised quantitative insulin sensitivity check index had the smallest coefficient of variation (2.8 ± 2.1%) whereas the C-peptide based homeostasis model assessment 2 had the highest discriminant ratio (1.97 (1.65-2.39)). As for insulin sensitivity indices that are based on OGTT, the oral glucose insulin sensitivity index had the smallest coefficient of variation (6.5 ± 5.1%). The highest discriminant ratio was found for the non-esterified fatty acids-based insulin sensitivity index (NEFA-ISI, 2.70 (2.30-3.22)). For the assessment of insulin secretion from fasting variables, the lowest mean coefficient of variation was found for C-peptide based homeostasis model assessment 2 beta with 10.8 ± 8% and the highest discriminant ratio for the C-peptide / Glucose-Ratio (2.18 (1.84-2.63)). Among indices assessing insulin secretion from an OGTT, the lowest coefficient of variation was found for the ratio of the areas under the C-peptide and glucose curves from 0 to 120 minutes with 11.3 ± 9.7%. The data reveal large differences in the reproducibility and the discrimination capability of different indices that assess insulin sensitivity or insulin secretion. Our findings can aid the selection of an appropriate index in clinical studies.