The study was approved by the institutional ethics committee of the University of Veterinary Medicine, Vienna (04/12/97/2013; date of approval 17 December 2013) according to the Good Scientific Practice guidelines and the national authority according to § 26 of Law for Animal Experiments [Tierversuchsgesetz 2012 – (TVG 2012); BMWF GZ 68.205/0007-II/3b/2014] as well as by the Slovakian Regional Veterinary Food Administration (428/2014) and was conducted between March and April 2014. A written informed consent was obtained from the farm owner to collect samples and to publish the study results. The farm involved in the study adheres to a high standard of veterinary care based on standard operating procedures.
Animals
A commercial dairy farm in Slovakia with approximately 2700 Holstein-Friesian cows and additional youngstock was chosen as study site. The dairy herd was kept in free stall barns with rubber mats on concrete floors. Before calving the cows were moved in a free stall barn with straw beddings. A total mixed ration was fed twice per day and pushed up frequently. The average energy corrected milk yield (based on 4.0 % butterfat and 3.4 % protein) was 9165 kg in 2013.
A sample size calculation (type I error α = 0.05, type II error β = 0.2) was performed to detect a maximum irrelevant difference in the glucose concentration of 3 mg/dL between the methods evaluated in the study, resulting in 182 animals needed. To compensate for potential data losses due to necessary exclusions because of pre-analytical or analytical problems 240 animals were enrolled in the study. Animals of all lactations within 2 weeks ante-partum up to 4 weeks post-partum were eligible for enrollment.
Sampling procedures
The sampling procedures are illustrated in Fig. 1. To obtain capillary blood, the skin of the exterior vulva was dry-cleaned and punctured with one of three different available lancets, in random order. The lancets used in the study [Microtainer Contact-Activated Lancet (MT, Becton-Dickinson), SafetyLancets special (SL, MED TRUST Handelsges.m.b.H.), MiniCollect Safety Lancets (MC, Greiner Bio-One International AG)] have a penetration depth of 2 mm with differing blade widths between 0.8 (SL) to 1.5 mm (MT and MC). If the volume of the obtained blood drop was insufficient for determination of the glucose concentration, another puncture was made. The quantity of punctures was recorded on a pre-assembled data capture form. Three different electronic hand-held meters [FreeStyle Precision, (FSP, Abbott GmbH & Co. KG), GlucoMen LX Plus, (GLX, A. Menarini GmbH), WellionVet GLUCO CALEA, (WGC, MED TRUST Handelsges.m.b.H.)] and the associated electrochemical test strips [FreeStyle Precision blood glucose, (Abbott Diabetes Care), GlucoMen LX Sensor, (A. Menarini GmbH), WellionVet GLUCO CALEA test strips, (MED TRUST Handelsges.m.b.H.)] were used in the study to determine the blood glucose concentration. The WGC was already validated for use in dogs, cats and horses by using preprogrammed species-specific chips, offered with each batch of test strips. Because of best possible match in terms of red blood cell parameters, the chip initially designed for use in cats was chosen for measurements in this study.
Operating principle of the hand-held devices and laboratory analyses
The test principle of all 3 hand held devices is based on an amperometric biosensor technology. The meter systems consist of two components: The meter itself processes the data by applying integrated algorithms and presents the results on an integrated display. The associated biosensor test strips are designated to analyze a small amount of a blood sample (i.e. for the GLX: 0.3 μL; WGC: 0.5 μL; FSP: 0.6 μL) by initiating further enzymatic reactions.
After the application of the blood on the sensor, an electrochemical reaction starts using either glucose dehydrogenase (FSP) or glucose oxidase (GLX and WGC) as catalysator. The enzymes are oxidoreductases and oxidize glucose to gluconolactone. By this, electrons from the glucose are transferred to the oxidized form of a mediator, thereby converting it to the reduced form. By re-oxidation of the mediator electrons are transferred to the electrode surface. The resulting current is monitored by the meter and directly proportional to the glucose concentration in the blood sample. Using meter specific software algorithms, the current is converted into a measure of the glucose concentration which is displayed as digital value.
The sensors of the 3 hand-held devices were directly dipped onto the surface of the capillary blood drop. After approximately 5 s each, the blood glucose concentrations were presented on the display of the devices.
Besides the manufacturer’s intended and certified use of testing capillary blood, the devices were used “off-label” for measuring the glucose concentration in venous and/or arterial blood. For this, an additional blood sample was drawn from a coccygeal vessel using a blood-collection tube system (Vacuette, Greiner Bio-One GmbH) consisting of a Sodium Fluorid vacuum tube (Vacuette, FX Sodium Fluoride, Greiner Bio-One GmbH) and a 20-gauge needle (Vacuette 0.9 × 38 mm, Greiner Bio-One GmbH). Blood obtained from the coccygeal vessel was tested as well with all three devices as previously described. Approximately 2 h after collection, the coccygeal blood samples were centrifuged at 2200 × g, at a temperature of 18 °C for 5 min. (Eppendorf Centrifuge 5804, Eppendorf AG). Supernatant plasma was split into two aliquots in microtubes of 2 ml each (Microtube, Sarstedt), as reference and back-up sample and were stored at −18 °C until further analyses. The glucose concentration of this plasma sample was analyzed at the Central Clinical Pathology Unit (CCPU) of the University of Veterinary Medicine, Vienna and was considered as the reference value (i.e. the criterion standard) in the present study. Plasma was analyzed with an automated wet chemistry analyzer Cobas 6000/c501 (Roche Diagnostics GmbH, Vienna, Austria) using a colorimetric hexokinase method. Hexokinase catalyzes the phosphorylation of glucose to glucose-6-phosphate and adenosine-diphosphate by ATP. This reaction is coupled with an NADP colorimetric indicator system.
To evaluate the intra-assay variability of the analyses performed at the CCPU, a subset of 20 aliquots of plasma, obtained from a blood sample of one cow, were randomly placed among the samples obtained from the study animals. Furthermore, intra- and inter-assay coefficients of variations (CV) were calculated for each specific hand-held device. For this, three blood samples with different glucose concentrations based on FSP measurements with low (30 mg/dL), medium (56 mg/dL) and high (70 mgl/dL) glucose concentrations were tested ten times with one device (intra-assay) and additionally with ten different devices of the same type (inter-assay).
Statistical analyses
The data were analyzed using SPSS statistics for Windows (Version 20.0; IBM Deutschland GmbH) and BiAS for Windows (Version 10.06; Epsilon-Verlag). Data were tested for normal distribution using the Kolmogorov-Smirnov-Test. For each tested hand-held device, descriptive statistics were estimated for the glucose concentrations analyzed in capillary and coccygeal blood. Additionally, the Pearson correlation coefficients were calculated between the reference and the glucose concentrations measured with each specific device in capillary and coccygeal blood. The level of significance for all statistical tests was set at α = 0.05.
A Passing-Bablok regression [21] was performed to compare the glucose concentration in the reference sample with the concentrations measured with the three hand-held devices in capillary and coccygeal blood. For this, the slopes and the intercepts of the regression lines and their 95 % confidence intervals were determined. The intercept (a) reflects the constant differences and the slope (b) the proportional differences between the two methods. If the confidence interval for the intercept includes the value 0, no constant difference between the two methods exists. If the interval for the slope includes the value 1, no proportional difference occurs. If neither a constant nor a proportional difference could be observed, both methods can be used interchangeably [22]. In addition, the agreement between the reference and the hand-held meters were graphically depicted using the method as reported by Bland and Altman [23].
Based on the glucose concentrations analyzed at the CCPU, samples were classified as hypo- (<40 mg/dL), normo- (40–60 mg/dL) or hyperglycemic (>60 mg/dL). According to these classifications sensitivities (Se) and specificities (Sp) for each hand-held device to detect hypo- and hyperglycemia were calculated. To determine optimized thresholds to identify hypo- and hyperglycemia using the hand-held meters, Receiver Operating Characteristics (ROC) analyses were performed. The closer the resulting graph of the ROC analysis is to the left upper angle of the coordinate system, the greater is the accuracy of the test [24]. The resulting area under the ROC curve (AUC) is a measure of the discriminatory power of a test to identify animals as normoglycemic and hypo- or hyperglycemic, respectively. An AUC of 1 represents a perfect test; an AUC of 0.5 and below represents a worthless test. ROC analyses were based on maximizing the Youden-Index (J), calculated as (Se + Sp - 1). The J reflects the number of all correctly identified outcomes and can range from 0 to 1. A value close to1 indicates a valuable diagnostic test, whereas a value close to 0 implies a worthless test [25].