Many different assays have been used to determine insulin concentration in horses, the results of which are important in screening for insulin resistance and determining the prognosis in equids with Equine Cushing's Syndrome. While these assays have generally been validated for use in horses, there are minimal data on comparisons between different assays currently in use around the world. The aim of this study was to compare 4 assays (3 radioimmunoassay kits and one ELISA) for measurement of equine serum insulin concentration.
Blood samples were collected as part of an epidemiological study of insulin resistance in 208 ponies. Samples were centrifuged within 3 hours of collection, aliquots of serum taken and frozen at - 800C until analysis. All samples were analysed using the DSL-1600 Insulin RIA kit (n = 208). Aliquots from the same sample were also analysed using the DPC Insulin RIA (n = 137), DiaSorin Insulin RIA (n = 59) and/or the Mercodia ELISA (n = 22) kits. Serum and plasma aliquots from the same horse (n = 17) were also compared using the DPC insulin RIA. All kits had been validated for use in horses for their performance on sensitivity, parallelism, and precision (intra-assay variation). All kits were within the acceptable parameters for recovery (>90%). All kits showed excellent dilutional parallelism except the DSL RIA which tended to overestimate the insulin concentration above a 1:4 dilution. Intra-assay variation was good for the DiaSorin RIA (<5%), Mercodia ELISA (5.3%), and DPC RIA (7.3%), and acceptable for the DSL 1600 RIA (10.9%). Insulin values ranged from 0.3-645 µIU/ml and were not normally distributed. A Spearman rank correlation coefficient (r) and coefficient of determination (R2)were determined between pairs of tests using linear regression. The Wilcoxon signed rank test was used to determine any differences between data sets.
All tests were strongly significantly correlated. Serum and plasma insulin both measured using the DPC RIA had the highest correlation (r = 1.00, R2 = 0.98) and were not significantly different. The different assay kits had a range of r from 0.92-0.96, and R2 from 0.84-0.96. However, despite excellent linear correlation, serum insulin concentrations were not directly comparable between different assays, with significant differences detected between datasets (P<0.01). For example, at an insulin of 30 µIU/ml measured by the DPC RIA, values for the DSL, Mercodia and DiaSorin assays were, respectively, 20, 54 and 72 µIU/ml. Furthermore, at high serum insulin concentrations (>200 µIU/ml), there was a reduction in the linear correlation, especially with the DSL RIA kit. In conclusion, the 4 commonly used, validated assays for measurement of serum insulin concentration were highly correlated but data from different assays should not be directly compared.