Assay-dependent variability of serum insulin concentrations: a comparison of eight assays

Although insulin measurement is essential for both clinical and research purposes, there is currently no reference method for insulin assays. The aim of this study was to compare results of serum insulin determined by a number of commercially available assays. We compared eight insulin assays by analyzing 165 serum samples. Assays included two chemiluminescence (Roche and DiaSorin), four ELISA (Tosoh, Mercodia, Monobind, and Diametra), and two IRMA (Izotop and BioSource) methods. Each assay was compared with the mean of all assay methods and Bland-Altman difference plots were used to measure agreement between each assay and overall mean. Least squared perpendicular distance regression analysis (Deming's method) was used to calculate slope and intercept for bias and also for each assay vs. mean of eight assays. Findings showed that the lowest and highest median insulin concentrations varied by a factor of 1.8. Maximum and minimum correlations with mean of assays were observed for Roche (0.992) and BioSource (0.844), respectively. Significant bias was observed in six assays. In pairwise comparisons of different assays, the highest and least mean differences were 7.78 μU/mL and -0.14 μU/mL, respectively. In conclusion, serum insulin measurement with different assays showed a maximum of 1.8-fold difference, a point that should be taken into consideration in the interpretation of circulating insulin levels in both clinical and research fields.