All horse owners provided informed written consent. Thirty-five client owned horses and ponies previously diagnosed with ID by referring veterinarians using an oral sugar test  were enrolled in the study. The diagnosis of ID was based on blood samples obtained between 60 and 90 minutes after oral administration of the syrup and analyzed for insulin (Mercodia equine insulin ELISA, Mercodia AB, Uppsala, Sweden). To be eligible to participate in the study the insulin concentration had to be > 90 µIU/mL (insulin concentrations > 45 µIU/mL is considered diagnostic for ID). In addition, 26 clinically healthy horses and ponies owned by the Swedish University of Agricultural Sciences were included in the study. The horses were divided into 3 major groups: warmblood horses (warmbloods and Standardbreds), Icelandic horses and ponies (pony crossbreds, Welsh ponies, Gotland ponies and Shetland ponies). Criteria for inclusion were no ongoing episode of laminitis based on clinical examination and normal plasma ACTH adjusted for the season. All horses in the study were fed a hay or haylage diet supplemented with minerals. Horses were housed in individual box stalls and allowed daily turnout in a dirt or sand paddock. None of the horses had been on grass pasture for at least 2 months before testing.
All horses were acclimatized for at least 48 hours to the environment where sampling was to take place. After acclimatization, horses were sampled for FI and FG at 7 AM on two consecutive days immediately followed by an EHC the second day. Blood sampling for FI and FG as well as the EHC took place after feed withdrawal overnight.
The day before testing an IV catheter (Intranule, 2.0 × 105 mm. Vygon, Ecouen, France) for blood sampling was inserted into one of the jugular veins under local anesthesia (EMLA, AstraZeneca AB, Södertälje, Sweden). Blood samples for FI and FG were collected from the jugular catheter at 7 AM on two consecutive days. Fasting insulin and glucose concentrations were used to calculate proxies. Insulin sensitivity was estimated using the proxies 1/Insulin , RISQI  and QUICKI  whereas the β-cell response was estimated using FI and the proxies MIRG  and HOMA-β .
Proxies were calculated using the following formulae:
Glucose concentrations expressed in the SI-unit mmol/L were used for HOMA-β but were converted into mg/dl before insertion into the formulae for QUICKI and MIRG. Insulin concentration in µIU/mL were used in all calculations.
EHC – euglycemic hyperinsulinemic clamp
A second IV catheter (Intranule, 2.0 × 105 mm. Vygon, Ecouen, France) for infusions was inserted under local anesthesia (EMLA, AstraZeneca AB, Södertälje, Sweden) into the contralateral jugular vein in the afternoon on the day preceding the EHC. Blood samples for determination of FI and FG were drawn from the sampling IV catheter immediately before the start of the EHC at 7 AM. The EHC procedure has previously been described for use in horses [23, 29]. A continuous rate infusion of regular insulin (Humulin Regular, Eli Lilly Sweden AB, Solna, Sweden) was maintained throughout the 180 min clamp procedure at 3 mIU/kg/min. Blood glucose was kept at 5 mmol/L using a variable continuous rate infusion of glucose (Glucose Fresenius Kabi 500 mg/ml, Fresenius Kabi AB, Uppsala, Sweden). Adjustment of the glucose infusion was made based on the results of measurement (Accu-Check Aviva, Roche Diagnostics Scandinavia AB, Bromma, Sweden) of blood glucose concentration performed every 5 minutes. Serial blood samples were obtained every 10 minutes during the clamp for later analyses of plasma glucose to enable calculation of whole body glucose uptake, i.e. metabolic rate of glucose (M index). The steady-state period of the clamp was defined as the last 60 minutes. The M index was defined as the infusion rate of exogenous glucose administered during the steady state after correction of the glucose space [23, 29]. Horses were classified as IR if their M index was < 2.4 mg/kg/min. This cut-off level was based on the normal distribution of the M index (mean and 95 % confidence interval; 3.8 (2.4–5.2) in a group of metabolically healthy control horses (Icelandic horses and Gotland ponies) from a previous study . The lower confidence interval for the M index in this group of horses was used as the cut-off for IR in the present study.
Analysis of blood samples
All blood samples were collected into evacuated tubes (Vacuette 9 ml, Greiner Bio-One GmbH, Kremsmünster, Austria) containing lithium heparin and immediately placed on ice for 5 minutes before centrifugation (10 min, 2700 × g). Plasma was separated, frozen rapidly and stored at -80°C until later analysis of plasma insulin and glucose concentrations. Plasma glucose concentrations were measured enzymatically with an automated clinical chemistry analyser (YSI 2300 Stat Plus Analyzer, YSI Incorporated, Yellow Spring, Ohio). Endogenous concentration of plasma insulin was measured using a commercialised equine-optimised ELISA (Mercodia equine insulin ELISA, Mercodia AB, Uppsala, Sweden) and insulin concentrations were verified with a commercial kit (Mercodia animal insulin control; low, medium and high, Mercodia AB, Uppsala, Sweden) . All analyses were performed in duplicate.
All data were analyzed using a commercially available software program (JMP® Pro, version 15.0.0, SAS Institute Inc, Cary, North Carolina). The age of the horses and their M index were compared between groups of horses using one-way ANOVA. Comparisons between groups were performed by use of the Tukey-Kramer post hoc test. Variables were tested for homogeneity of variance using the Levene’s test. All residuals were analyzed for normality using the Shapiro-Wilk test. Data are presented as mean ± standard deviation (SD).
According to Bergman’s hypothesis insulin secretion is inversely related to IS in a rectangular hyperbolic relationship; y = constant \(\times\) 1/x, where y and x represent β-cell response and IS respectively [5, 9]. This equation can be re-expressed through a log-transformation to a linear model: Ln (y) = constant + β \(\times\) Ln (x), where β is the regression coefficient. If β was close to -1 (if the 95 % CI for β included − 1 but excluded 0) the rectangular hyperbolic relationship was considered to be fulfilled. This linear function was used to describe the relationship between fasting indices for β-cell response (FI, MIRG and HOMA-β) and IS (M index).
The proxies for IS (1/Insulin, RISQI and QUICKI) were initially compared with quantitative measurement of IS (M index) as an independent variable using simple linear regression. The residuals were normally distributed but showed heteroscedasticity and the regression model was therefore changed to weighted linear regression.
Bland-Altman plots of absolute differences between test and retest values against their mean were initially used to assess for systematic bias and uniform data distribution. The absolute difference plots demonstrated heteroscedasticity and data were therefore presented in relative difference Bland-Altman plots where the difference between test and retest values were divided by their means, expressed as percentage, and then plotted against their mean . The relative differences between test and retest values were assessed for normal distribution using the Shapiro-Wilk test. The 95 % CI for the mean relative differences between test and retest values was calculated. If the 95 % CI included 0, no systematic bias was evident. Coefficient of repeatability (CR) was calculated from the SD of the relative differences (absolute value of 1.96 \(\times\) SD) and expressed as percentage . The CVs were calculated using the root mean square method and reported with 95 % confidence interval . The intra-class correlation coefficient (ICC) was computed for all fasting indices using a one-way random effects model ANOVA on log transformed data to ensure normal distribution among residual. Comparison between fasting indices sampled at day 1 and day 2 were performed using a paired t-test on log transformed data to ensure normal distribution among all data. Values of P < 0.05 were considered as significant for all analyses.
Receiver operating characteristic (ROC) curve and Youden’s index analysis were used to determine the optimal cut-off for all proxies and FI with M index as the reference. Sensitivity, specificity and area under the curve (AUC) were calculated for each ROC curve analysis and reported with a 95 % CI.