Abstract Background Proxies are mathematical calculations based on fasting glucose and/or insulin concentrations developed to allow prediction of insulin sensitivity (IS) and β-cell response. These proxies have not been evaluated in horses with insulin dysregulation. The first objective of this study was to evaluate how fasting insulin (FI) and proxies for IS (1/Insulin, reciprocal of the square root of insulin (RISQI) and the quantitative insulin sensitivity check index (QUICKI)) and β-cell response (the modified insulin-to-glucose ratio (MIRG) and the homeostatic model assessment of β-cell function (HOMA-β)) were correlated to measures of IS (M index) using the euglycemic hyperinsulinemic clamp (EHC) in horses with insulin resistance (IR) and normal IS. A second objective was to evaluate the repeatability of FI and proxies in horses based on sampling on consecutive days. The last objective was to investigate the most appropriate cut-off value for the proxies and FI. Results Thirty-four horses were categorized as IR and 26 as IS based on the M index. The proxies and FI had coefficients of variation (CVs) ≤ 25.3 % and very good reliability (intraclass correlation coefficients ≥ 0.89). All proxies and FI were good predictors of the M index (r = 0.76–0.85; P 9.5 µIU/mL for FI. Conclusions All proxies and FI provided repeatable estimates of horses’ IS. However, there is no advantage of using proxies instead of FI to estimate IR in the horse. Due to the heteroscedasticity of the data, proxies and FI in general are more suitable for epidemiological studies and larger clinical studies than as a diagnostic tool for measurement of IR in individual horses.