The increasing number of horses diagnosed with insulin resistance (IR) and the suggested link between IR and laminitis has highlighted the need to accurately diagnose IR in clinical practice.
The increasing number of horses diagnosed with insulin resistance (IR) and the suggested link between IR and laminitis has highlighted the need to accurately diagnose IR in clinical practice.
The aim of the study was to evaluate the repeatability of the combined glucose-insulin tolerance test (CGIT) as well as to determine the effect of 2 different breeds and the effect of a stressor on the test results.
Clinically normal horses, 9 Standardbred horses and 9 Icelandic horses.
Prospective clinical nonrandomized trial. The CGIT was performed on all horses on 2 occasions 3 weeks apart. An additional CGIT was performed on four of the Standardbred and four of the Icelandic horses after transportation to a new environment (stressor) the day before testing.
Calculated parameters for the glucose curve of the CGIT had low repeatability, whereas the parameters for the insulin curve had high repeatability. There was an effect of breed (Standarbred versus Icelandic horse) as well as stress on the glucose dynamics, but not on the insulin dynamics of the CGIT.
Repeatability of the glucose dynamics of the CGIT is low. In addition, there appears to be breed differences in the glucose dynamics. It is therefore suggested that diagnosis of IR should not be made solely based on results from the glucose curve of the CGIT.
insulin concentration at 45 minutes
area under the glucose curve
area under the insulin curve
body condition score
combined glucose-insulin tolerance test
cresty neck score
euglycemic hyperinsulinemic clamp
frequently sampled intravenous glucose tolerance test
intraclass correlation coefficient
negative phase glucose clearance rate
positive phase glucose clearance rate
positive phase duration for the glucose curve
duration until the insulin curve reached the concentration 100 mU/L
half-time for insulin
Insulin resistance (IR) is defined as a diminished ability of insulin to stimulate transport of glucose into insulin sensitive tissues. This condition has gained more attention in horses in recent years. IR might be associated with conditions such as laminitis, pituitary pars intermedia dysfunction, endotoxemia, and equine metabolic syndrome. The increasing number of horses diagnosed with IR and the proposed link between IR and laminitis has highlighted the need to accurately diagnose IR in clinical practice.
Several methods to evaluate glucose and insulin dynamics have been described in the horse. Quantitative and specific methods for assessing IR in horses are the euglycemic hyperinsulinemic clamp (EHC) and the frequently sampled intravenous glucose tolerance test (FSIGTT) with minimal model analysis. Insulin sensitivity measured by EHC has higher repeatability and lower day-to-day variability compared with minimal model calculations from the FSIGTT. However, the minimal model provides additional information when compared with the EHC, such as the ability to obtain estimates for insulin-independent glucose utilization and measurement of the sensitivity of pancreatic β cell function. The expense and technical difficulties as well as the time-consuming nature of the quantitative methods for assessing insulin sensitivity limit their use in clinical cases. Instead, nonspecific indicators for IR such as measurement of glucose and insulin in single blood samples are commonly used in clinical settings as screening tests, but dynamic tests are essential to assess insulin sensitivity. The most common dynamic tests used in clinical cases include oral or intravenous glucose tolerance tests or the combined glucose-insulin test (CGIT). Although these tests are dynamic, they do not give specific quantitative measurements of the insulin sensitivity in the horse. The CGIT has been recommended for the diagnosis of IR in equine practice because the test is easy to interpret and provides information about both the glycemic and insulinemic responses. The repeatability of the whole glucose curve of the CGIT and the seasonal changes in calculated parameters from the glucose curve have previously been assessed,[7, 8] but there is no information about the repeatability of the insulin curve. In addition, no information is available on whether breed differences or physiological stress before testing could influence the glucose or insulin dynamics of the CGIT in horses.
The primary objective of this study was thus to determine the repeatability of the CGIT in clinically healthy horses as well as to compare the results of the CGIT between a light horse breed (Standardbred) and a pony-related breed (Icelandic horse). A 2nd objective was to evaluate the effect of a stressor (transportation and change of environment) the day before the CGIT to simulate the situation for horses coming into a clinic for the assessment of IR.
Nine Icelandic horses (2 geldings and 7 mares) and 9 Standardbred horses (2 geldings and 7 mares) were used in the study. One week before the experiment, physical examination, body condition score (BCS), the cresty neck score (CNS), and fasting blood samples for the determination of plasma insulin and glucose were obtained (Table 1). Horses eligible to participate did not have a history of laminitis, did not have clinical signs of disease for the whole study period, and had CNS score <3 and BCS scores <7 as well as plasma insulin <20 mU/L and normoglycemia.
|Group||Age (years)||Body Weight (kg)||Body Condition Scores (1–9)||Cresty Neck Scores (1–5)||Plasma Insulin (mU/L)||Plasma Glucose (mmol/L)|
|Icelandic horses (n = 9)||13.6 ± 2.7||379 ± 37a||5.7 ± 0.8||2.4 ± 0.2a||10.2 ± 4.2||4.3 ± 0.5|
|Standardbreds (n = 9)||13.3 ± 4.6||506 ± 46||5.2 ± 0.4||1.9 ± 0.3||8.3 ± 4.4||4.4 ± 0.6|
|Icelandic horses (n = 4)||14.0 ± 3.3||379 ± 38a||5.9 ± 1.0||2.4 ± 0.3||10.2 ± 4.1||4.3 ± 0.4|
|Standardbreds (n = 4)||13.0 ± 5.7||510 ± 64||5.5 ± 0.6||1.8 ± 0.5||8.8 ± 4.3||4.3 ± 0.6|
The Standardbred horses were sedentary horses owned by the Department of Clinical Sciences, Swedish University of Agricultural Sciences. The Icelandic horses were all privately owned (1 owner) and were used for leisure riding at low intensity level once or twice a week. All horses were housed in box stalls with daily turnout in a paddock and fed grass haylage 3 times daily. Water and salt were provided ad libitum. The experiment started in February 2011.
The study (C338/8) was approved by the Ethical Committee for Animal Experiments, Uppsala, Sweden.
The study was performed as a prospective clinical nonrandomized trial. All procedures for evaluating the repeatability of the CGIT were performed in the horses' home environment. The horses were not subjected to any stressful handling activities and tolerated the placement of catheters and the sampling procedures well. The day before each test the horses were weighed, an area over each jugular vein was clipped, physical examinations were performed, and the BCS and the CNS were determined by one of the authors (JB). The CGIT was performed twice with 3 weeks between tests on each horse. Two weeks after the 2nd CGIT, 4 Standardbred horses and 4 Icelandic horses were available for a 3rd CGIT, to test the effect of a stressor on the results of the CGIT (transportation and new environment). The remainder of the horses (10 horses) were used in teaching or training activities and could not be included in this part of the study. The day before the 3rd CGIT, the 8 available horses were transported from their home environment to the large animal clinic at the Swedish University of Agricultural Sciences in a horse trailer for 60 minutes. The transportation of the horses took place between 12:00 am and 4:00 pm.
For all testing procedures, feed was withheld from 19.00 hours the night before the test, but the horses were allowed free access to water. Between 06.00 and 07.00 hours on the morning of the CGIT, intravenous catheters1 were inserted under local anesthesia2 into the left and right jugular veins for blood sampling and infusion of glucose and insulin, respectively. The CGIT commenced between 08.00 and 08.45 hours and sampling for baseline concentrations (see below) of plasma glucose and insulin occurred during the hour preceding the CGIT.
Blood samples were collected at predetermined times before and during the CGIT (see below) in evacuated tubes3 containing lithium heparin for the analyses of plasma glucose and plasma insulin. All sample tubes were placed on ice for approximately 15 minutes and then centrifuged for 10 minutes at 2,700 × g. Plasma was subsequently harvested and stored at −80°C until analysis.
Blood samples for the determination of baseline concentrations of plasma glucose and plasma insulin were drawn from one of the intravenous catheters at 60, 40, 20, and 2 minutes before glucose administration. The CGIT has previously been described by Eiler et al. Briefly, glucose at a dosage of 150 mg/kg body weight was infused IV as a 50% glucose solution4 over approximately 45 seconds, immediately followed by an intravenous administration of 100 mU/kg body weight regular insulin.5 Blood samples for subsequent analyses of plasma glucose and plasma insulin were collected via the contra lateral jugular catheter at 1, 5, 15, 25, 35, 45, 60, 75, 90, 105, 120, 135, and 150 minutes after completion of glucose administration. The second time the CGIT was performed, an extended sampling period with sampling times at 165, 180, and 195 minutes was added to the experimental protocol.
Plasma glucose concentrations were measured as a single batch using an automated analyzer.6 Plasma insulin concentrations were determined in duplicates using an ELISA7 that has previously been validated for equine insulin. The intra-assay CV varied between 5.5% and 8.9% for each batch of insulin analyses as determined from the duplicate analyses.
The area under the glucose (AUCglu) and insulin (AUCins) curves was calculated by use of the trapezoidal method with commercial software.8 The positive phase duration for the glucose curve (PP-Dglu) was defined as the time from start of the CGIT to the time when the plasma glucose concentration had returned to its baseline concentration. Positive phase glucose clearance rate (PP-Clglu) and negative phase glucose clearance rate (NP-Clglu) were calculated as previously described.[7, 8] The duration until the insulin curve reached the concentration 100 mU/L (Tins) was defined as the time from start of the CGIT to the time when the plasma insulin concentration had reached 100 mU/L. The insulin concentrations followed an exponential decay between 5 and 75 minutes (1577.2e−0.0705x). The half-time for insulin (T1/2-ins) in the time segment 5–75 minutes was calculated by use of commercial software.8
Differences in mean AUCglu, PP-Dglu, PP-Clglu, NP-Cl,glu, Tins, and [Ins]-45, T1/2-ins, AUCins, of repeated test periods or between breeds were compared by use of a t-test, paired t-test or a one-way repeated measures analysis of variance (ANOVA), as appropriate. Logarithmic transformations were used to obtain normally distributed residuals for plasma glucose and plasma insulin concentrations. Means for plasma glucose or insulin concentration during the CGIT were compared across treatments (test periods or breeds) and sampling time by use of a 2-way repeated measures ANOVA. When a significant difference was identified by ANOVA, a Tukey′s posthoc test was performed. All results are expressed as mean ± SD, except for logarithmically transformed data, which were expressed as the geometric mean ± 95% CI on the original scale after back transformation.
The repeatability of the CGIT for an individual horse was determined by the coefficient of variation (CV), repeatability coefficient, and intraclass correlation coefficient (ICC). The repeatability coefficient is the maximal difference that occurs between two measurements on the same horse for 95% of the paired observations. It was calculated according to the formula √2 × 1.96Sw whereas the ICC was calculated as ICC = 1 − Sw2/(Sb2 + Sw2) where Sw2 is the mean square of the residuals (within subjects) and Sb2 is the mean square between subjects obtained from the repeated measures ANOVA table.
Statistical analyses were performed by use of a commercial statistical software.9 For all analyses, values of P < .05 were considered significant.
All horses fulfilled the inclusion criteria. The Icelandic horses had higher CNS (P < .001) and lower body weight (P < .001) than the Standardbred horses. There was no change in BCS (5.5 ± 0.7 versus 5.4 ± 0.6; P = .90) or CNS (2.2 ± 0.3 versus 2.2 ± 0.3; P = .78) from the start to the end of the study for the 18 horses. Body weight for all 18 horses was 442 ± 77 kg (mean ± SD) at the beginning of the study and 442 ± 76 kg at the end of the study (P = .99). All 18 horses had glucose curves with a positive phase (hyperglycemia) above baseline followed by a negative phase (hypoglycemia) below the baseline concentration. The test caused clinical signs of hypoglycemia (tremors, muscle fasciculations, blinking) on 1 occasion in 3 Standardbred horses. In 1 horse, the blood glucose concentration dropped to 1.0 mmol/L so the test was interrupted (at 45 minutes) followed by immediate treatment with 120 mL of 50% glucose4 IV.
Plasma glucose and plasma insulin concentrations during the 2 CGIT for all 18 horses are shown in Figure 1A,B. There was no difference between the 2 test occasions for the mean plasma glucose (P = .23) or mean insulin (P = .13) concentration within each time point. Measures of repeatability at specific time points for the glucose and insulin curves are presented in Table 2. The measurement error for the CGIT curves was not consistent over time. For the glucose curve, the lowest repeatability occurred between 35 and 90 minutes (CV: 24.7–31.9%; ICC: 0.59–0.68). The repeatability coefficients during this time period varied between 1.9 and 2.9 mmol/L while the mean glucose concentration varied between 2.4 and 4.2 mmol/L. The insulin curve, on the other hand, had the highest repeatability during the time period between 5 and 45 minutes (CV: 7.0–22.6%; ICC 0.81–0.91). The measures of repeatability for calculated parameters of the glucose and insulin curve of the CGIT are presented in Table 3. Overall, the best combination of test characteristics, ie, low CV, small repeatability coefficient, and high ICC, was observed for the calculated parameters of the insulin curve, except for [Ins]-45. In contrast, the calculated parameters for the glucose curve showed remarkable higher CV, lower ICC, and higher repeatability coefficients except for AUCglu, which demonstrated lower CV and repeatability coefficient.
|Sampling Time (minutes)||Plasma Glucose||Plasma Insulin|
|CV (%)||ICC||Repeatability Coefficient||CV (%)||ICC||Repeatability Coefficient|
|Variable||Test 1||Test 2||P-Value||CV%||ICC||Repeatability Coefficient|
|PP-Dglu (minutes)||33.9 ± 14.7||40.2 ± 22.1||.265||44.9||0.61||45.6|
|PP-Clglu (mmol/L/min)||0.341 ± 0.160||0.297 ± 0.138||.272||37.3||0.69||0.33|
|NP-Clglu (mmol/L/min)||0.043 ± 0.019||0.039 ± 0.016||.504||37.8||0.60||0.043|
|AUCglu (mmol/L/min)||591.6 ± 93.1||617.1 ± 112.8||.254||14.6||0.64||243.5|
|Tins (minutes)||40.3 ± 7.8||40.7 ± 6.2||.691||7.0||0.91||8.1|
|[Ins]-45 (mU/L)||70.0 ± 40.0||74.3 ± 32.1||.265||22.6||0.91||41.8|
|T1/2-ins (minutes)||9.8 ± 1.6||9.6 ± 1.1||.448||8.2||0.82||1.8|
|AUCins (×103 mU/L/min)||25.2 ± 4.2||25.3 ± 4.0||.933||7.7||0.88||5.5|
The glucose dynamics of the CGIT showed differences between Standardbred and Icelandic horses. The glucose curve showed breed differences in glucose concentrations within the time interval 35 and 75 minutes (P ≤ .001–.022) as well as in the time interval 135 and 150 minutes (P = .006–.007) (Fig 2A). The mean glucose curve for the Icelandic horses was displaced to the right after the peak in glucose concentration compared to the corresponding mean glucose curve for the Standardbred horses. The duration of the negative phase of the mean glucose curve had similar duration between the 2 breeds, but the negative phase continued to 150 minutes in the Icelandic horses. The sampling time after glucose infusion was therefore extended from 150 to 195 minutes in the 2nd test. The right displacement of the glucose curve in the Icelandic horses reflects the difference in the calculated parameters PP-Dglu and PP-Clglu for the glucose curve between the Standardbred and Icelandic horses (Table 4). One Standardbred horse and 6 Icelandic horses out of 9 horses in each breed demonstrated a PP-Dglu > 45 minutes on one of 2 test occasions.
|Variable||Standardbreds (n = 9)||Icelandic Horses (n = 9)||P-Value|
|PP-Dglu (minutes)||26.8 ± 7.3||53.6 ± 24.1a||.006|
|PP-Clglu (mmol/L/min)||0.387 ± 0.107||0.206 ± 0.101a||.002|
|NP-Clglu (mmol/L/min)||0.043 ± 0.014||0.034 ± 0.018||.265|
|AUCglu (mmol/L/min)||781.3 ± 108.5||806.9 ± 130.4||.679|
|Tins (minutes)||38.1 ± 4.3||43.3 ± 7.0||.089|
|[Ins]-45 (mU/L)||59.0 ± 20.8||89.6 ± 35.2a||.039|
|T1/2-ins (minutes)||9.8 ± 1.1||9.3 ± 1.1||.285|
|AUCins (×103 mU/L/min)||24.4 ± 4.3||26.8 ± 3.9||.238|
In contrast, there was no breed difference in the mean insulin curves (Fig 2B) between the 2 breeds (P = .52) or in the calculated insulin parameters Tins, T1/2-ins and AUCins of the insulin curve (Table 4). However, the insulin concentration at 45 minutes was different between the 2 breeds. None of the Standardbred horses but 2 Icelandic horses had plasma insulin concentrations >100 μU/mL at 45 minutes on one of 2 test occasions.
Plasma glucose and plasma insulin concentrations for the 8 horses that completed 3 CGIT are presented in Figure 3A,B. There was an effect on transportation and change in environment on the mean plasma glucose concentration for each time point within the time interval 30–150 minutes (P ≤ .001–.006), but there was no effect on the mean insulin concentrations within any time point (P = .15). The change in environment and transportation had an effect on the calculated variables PP-Dglu and AUCglu from the glucose curve but not on any of the calculated variables from the insulin curve (Table 5). Five horses (3 Standardbreds and 2 Icelandic horses) had PP-Dglu >45 minutes and 1 Icelandic horse had plasma insulin concentration >100 μU/mL at 45 minutes on the CGIT after transportation.
|Variable||Test 1||Test 2||(After Transportation) Test 3||P-Value|
|PP-Dglu (min)||30.5 ± 8.2||40.4 ± 13.1||67.6 ± 29.3a||.005|
|PP-Clglu (mmol/L/min)||0.35 ± 0.18||0.28 ± 0.12||0.18 ± 0.09||.077|
|NP-Clglu (mmol/L/min)||0.044 ± 0.020||0.038 ± 0.013||0.025 ± 0.015||.095|
|AUCglu (mmol/L · min)||559.4 ± 53.7||648.5 ± 53.7||848.8 ± 111.7a||<.001|
|Tins (min)||40.0 ± 7.6||42.4 ± 7.7||41.1 ± 6.1||.474|
|[Ins]-45 (mU/L)||68.4 ± 36.4||84.2 ± 38.9||74.7 ± 26.9||.254|
|T1/2-ins (min)||9,3 ± 1.6||9.2 ± 1.1||8.8 ± 0.9||.402|
|AUCins (×103 mU/L · min)||25.6 ± 4.9||26.8 ± 4.6||25.5 ± 3.5||.477|
This study showed that the calculated parameters for the glucose curve of the CGIT had low repeatability, whereas the calculated parameters for the insulin curve had moderate-to-high repeatability. In addition, there was an effect of breed (Standardbred versus Icelandic horse) as well as of stress on the glucose dynamics but not on the insulin dynamics of the CGIT.
The calculated glucose parameters of the CGIT found in the present study are in accordance with those previously reported for horses without clinical signs of disease.[7, 13, 14] The AUCins reported here are higher compared to those reported for nonobese mares by Frank et al. The present study detected higher plasma insulin concentrations in the time interval 1–25 minutes compared to the aforementioned study. The most likely explanation for this discrepancy in results is that the 2 studies used different assays for measurement of insulin (Mercordia Insulin ELISA versus Coat-A-Count Insulin RIA). Öberg et al showed that results between these assays were in good agreement for samples with equine insulin concentrations <160 μU/mL. On the other hand, analyses of samples from horses with higher insulin concentrations have shown to have poor agreement between different assays. Another important aspect is that different assays not only have variations in their specificity for equine insulin but also for exogenous insulin. Thus, this contributes to the variable results seen between assays for samples that contain a mixture of insulin, eg, samples from the CGIT. Therefore, it is difficult to compare results from CGIT studies using different assays for insulin, at least for calculated parameters using data points with high insulin concentrations such as AUCins and T½-ins.
The glucose curve of the CGIT is more variable in its nature compared to the insulin curve. The low repeatability for plasma glucose concentrations within the time interval used for calculation of the glucose dynamics resulted in low repeatability for PP-Dglu, PP-Clglu, and NP-Clglu. The opposite situation was seen for the insulin curve, which gave moderate-to-high repeatability for the calculated parameters Tins, [ins]-45, T1/2-ins and AUCins. The results with lower repeatability for plasma glucose within the individual sampling times used for calculations of the glucose dynamics for the CGIT are in agreement with a previous study evaluating seasonal changes of the CGIT. The CV in the present study for PP-Dglu, PP-Clglu, and NP-Clglu were higher (37.3–44.9%) compared to the mean CV reported in Funk et al study (21.0–33.3%). The reason for this discrepancy is not obvious, but it is possible that keeping the horses off feed before testing in the present study might have induced a low stress response in some of the horses, thereby causing a lower repeatability in the glucose dynamics of the CGIT.
The situation with a more variable glucose curve increases the risk that the calculated parameters used to define the normal glucose response of the CGIT give false positive results. False positive results, based on the glucose response of the CGIT, have been described in previous studies.[7, 8] In the present study, 7 of 18 horses demonstrated an abnormal glucose curve reflected by PP-Dglu >45 minutes on one of 2 test occasions. This should, however, be interpreted in light of the fact that six of these 7 horses were Icelandic horses. The duration of the hyperglycemic curve during the CGIT was longer in the Icelandic horses compared to the Standardbred horses. The hypoglycemic curve, which follows immediately after the hyperglycemic curve, had the same nadir and duration in the 2 breeds. This difference in glucose dynamics of the CGIT between Standardbred and Icelandic horses had an impact on the calculated parameters PP-Dglu and PP-Clglu but not on NP-Dglu and AUCglu. Therefore, the results of the present study suggest that the Icelandic horses should be defined to have a normal glucose response to a CGIT if PP-Dglu <60 minutes instead of <45 minutes. In contrast to the glucose dynamics, the Icelandic horses had similar insulin dynamics of the CGIT as the Standardbred horses (T, T½-ins, AUCins) except for a higher mean insulin concentration at 45 minutes. Taken together, it appears that in the Icelandic horses the ability of insulin to clear exogenously administered glucose was delayed suggesting that there are differences in insulin sensitivity between the 2 breeds. The results of the present study suggest that the cutoff limits for parameters that are used to define a normal response to the CGIT such as PP-Dglu need to be adjusted for some breeds, most likely for pony-related breeds. The present study evaluates the repeatability of the CGIT, but the accuracy of the test has yet to be determined.
The study underscores the importance of time needed to acclimate after a stressful event before insulin sensitivity is determined. The CGIT was conducted the day after transportation and arrival to the equine hospital, since this is common practice in Sweden. The stressor had an effect on the glucose dynamics of the CGIT but no effect on the insulin dynamics. Transportation of horses has been shown to increase blood or salivary cortisol concentrations up to a few hours after transportation.[16, 17] It is therefore likely that increased cortisol concentrations after transportation had an impact on the glucose dynamics in the present study. A previous study used a 3-day acclimation period to the hospital environment before testing, which appeared to be sufficient time to reduce the effect of the stressors on the results of the CGIT. Only 8 horses were available to test the effect of the transportation and change in environment on the results of the CGIT. The lack of randomization could have had an impact on the results of the CGIT after transportation.
In conclusion, the results of the present study demonstrate that the insulin dynamics to the CGIT is more repeatable than the glucose response. The calculated parameters from the glucose curve, except for AUCglu, had low repeatability, and the use of these parameters as the only means to determine whether the horse is IR must be questioned. Future studies are needed to evaluate the accuracy of the CGIT where the results from the CGIT are compared with quantitative measurements of insulin sensitivity with the FSIGT and minimal model analysis or the EHC. The introduction of a stressor had an effect on the glucose, but not on the insulin dynamics to the CGIT. In addition, comparison of the CGIT between Standardbred and Icelandic horses also identified differences in the glucose dynamics, but not in the insulin dynamics, suggesting that some of the reference limits for the CGIT needs to be adjusted for some breeds. In summary, it is advisable not to only rely on the glucose curve and its calculated parameters to determine if a horse is IR or not but to also assess the insulin dynamics to increase the sensitivity of the test.
The study was funded by internal funding within the Swedish University of Agricultural Sciences. A special thanks to Hans and Eva-Lena Stiernström at Edeby farm for providing the Icelandic horses for this study.
Conflict of Interest Declaration: None of the authors of this article has a financial or personal relationship with other people or organizations that could inappropriately influence or bias the content of the article.
Milacatch, 1.3 × 13 cm, 14 gauge × 5.25 inch; Mila International, Inc, Erlanger, KY
Emla (emulsion with 2.5% lidocain and 2.5% prilocain); AstraZenicaAB, Södertälje, Sweden
BD Vacutainer LH (Lithium Heparin) 10 mL; BD, Belliver Industrial Estate, Plymouth, UK
Glucos Fresenius Kabi (500 mg/mL); Fresenius Kabi, Uppsala, Sweden
Humulin Regular (100 U/mL); Eli Elly Sweden AB, Solna, Sweden
Architect ci8200; Abbott Scandinavia AB Diagnostics, Solna, Sweden
Mercodia Insulin ELISA; Mercodia AB, Uppsala, Sweden
GraphPad Prism, version 6.0 for Windows; GraphPad Software Inc, San Diego, CA
JMP, version 10.0; SAS Institute Inc, Cary, NC