This work was completed at Colorado State University Veterinary Teaching Hospital and Equine Reproduction Laboratory. The study was supported by Abbott Animal Health. The content has not been presented at a meeting.
Corresponding author: Eileen Hackett, DVM, MS, DACVS, DACVECC, Department of Clinical Sciences, Colorado State University, #1678, 300 West Drake Road, Fort Collins, CO 80523; e-mail: firstname.lastname@example.org.
Background: Glucose assessment and regulation are important factors in the treatment of hospitalized horses and foals.
Hypothesis/Objectives: The purpose of this study was to compare glucose measurement by a veterinary glucometer, adjusted by code for use in horses and foals, to a reference chemistry analyzer. It was hypothesized that the veterinary glucometer and reference analyzer would yield similar results and that interpretation of glucose values obtained from a veterinary glucometer would result in clinically appropriate decisions.
Animals: Fifty blood samples from adult horses and 50 blood samples from neonatal foals admitted to the Colorado State University Veterinary Hospital or Equine Reproduction Laboratory for evaluation.
Methods: Glucose concentrations from fresh whole blood samples were evaluated in duplicate with a veterinary glucometer and these values were compared with those obtained with a reference plasma chemistry analyzer. The accuracy of glucometer measurement was evaluated with a Clarke error grid.
Results: The veterinary glucometer accurately measured whole blood glucose concentrations in both horses and foals when compared with a reference plasma chemistry analyzer. Nearly 97% of the glucometer values obtained in this study would have resulted in appropriate clinical decisions based on the Clarke error grid analysis.
Conclusions and Clinical Importance: The veterinary glucometer evaluated has potential utility for point-of-care whole blood glucose evaluation in both horses and foals.
Hypoglycemia, hyperglycemia, and glucose variability are associated with increased morbidity and mortality in human critical care medicine.1 Similar patterns of glucose regulation associated with critical illness are present in equine veterinary medicine. In a recent report, more than two thirds of critically ill neonatal foals were hyperglycemic or hypoglycemic at the time of hospital admission.2 In this same report, extreme hyper- and hypoglycemia were associated with decreased survival.2 Glucose regulation also has been studied in adult equine emergency admissions. Hyperglycemia is a common glucose alteration in adult horses with gastrointestinal disease that is associated with a decrease in survival.3,a Accurate methods of glucose determination are important to provide veterinarians with a means to recognize hyper- and hypoglycemia in horses and foals.
In addition to accuracy, rapid monitoring of glucose is of paramount importance in critical care medicine.4 Point-of-care analyzers allow immediate patient-side clinical decision making, frequent monitoring, and decreases in both cost and time of analysis when compared with standard laboratory methods.5 Additional advantages include use of minimal blood volumes and ease of operation.6 Point-of-care glucometers are an ideally suited method to incorporate into aggressive glucose regulation protocols.7
Most glucometers available on the market and those tested previously in horses were designed for human use. Glucometers designed for human use assume a constant stable relationship between plasma and whole blood and equilibration between plasma and erythrocyte glucose concentrations.8–10 These assumptions and relationships do not apply universally to veterinary species and underlie the inherent inaccuracy of human glucometers in veterinary medicine. For example, in humans, glucose distribution is 50% within erythrocytes and 50% within plasma.9 In dogs, however, glucose distribution is 12.5% within erythrocytes and 87.5% within plasma, and in cats, glucose distribution is 7% within erythrocytes and 93% within plasma.9 Veterinary glucometers tailored to each species potentially are more accurate and require less preparation time than human glucometer counterparts.11 The veterinary glucometer evaluated requires a small volume of whole blood for analysis and gives a rapid result. The objective of this study was to compare glucose measurements from a veterinary point-of-care glucometerb with a reference plasma chemistry analyzer. It was hypothesized that the veterinary glucometer would have adequate accuracy and precision to be a useful adjunct to laboratory analysis in an equine hospital setting.
Materials and Methods
The Institutional Animal Care and Use Committee approved all procedures before conducting this research. Fifty samples from 48 horses >1 year of age and 50 samples from 36 foals <1 month of age admitted to Colorado State University Veterinary Hospital or Equine Reproduction Laboratory were collected over a 75-day period after informed client consent. Twenty-five of the adult horse and foal samples were from patients admitted for elective procedures and considered to be healthy and 25 samples in each age group were emergency admissions and considered to be critically ill. Case number and presenting complaint determined by the admitting clinician (board-certified in a veterinary specialty) were recorded. A jugular venous whole blood sample was obtained with a syringe and 20 G needle, and 0.3 μL immediately was tested with the veterinary glucometerb according to the manufacturer's instructions. The glucometer evaluated utilized glucose dehydrogenase enzyme and coulometric method principles. The linear range of this glucometer was 20–500 mg/dL within the hematocrit range of 15–65%. Two successive replications of glucometer measurement were conducted on each sample with a test time of approximately 15 seconds per sample. The glucometer was calibrated each day using the accompanying calibration solutionc and stored on site. The veterinary glucometer's internal algorithm associated with code 5 was used for testing based on pilot data derived from 50 horse and 50 foal paired glucose measurements. All glucometer testing was performed with strips from the same lot number.d PCV was measured and recorded. The remainder of the whole blood sample then was submitted in an evacuated glass tube containing lithium heparin to the Colorado State University Clinical Laboratory for plasma chemistry analysis.e The reference analyzer utilized hexokinase enzyme and photometric method principles. The linear range of this instrument was 2–750 mg/dL extended by automatic dilution. The plasma chemistry analyzer underwent automatic calibration dependent on start of run, preset intervals, and new lot or bottle of reagent. The reference range of glucose concentration for this reference analyzer in horses was 70–135 mg/dL.
Statistical analysis was performed on the data collected by commercial software.f Continuous data were tested for normality by a Shapiro-Wilk test and, because many of the variables were not normally distributed, results were reported as median and interquartile range (IQR). A Wilcoxon signed-rank test was used to compare glucose measurements between healthy and critically ill categories within each age group. Appropriateness of the glucometer algorithm associated with code 5 for use in the equine species was determined based on bias calculations, Clarke's error grid analysis, Lin's concordance correlation, and Pearson's correlation. Bias was calculated in mg/dL for glucose values ≤70 mg/dL and percentage difference from reference for glucose values >70 mg/dL. Percent bias was calculated using the following formula: % Bias = (average of glucometer values − reference glucose value)/ reference glucose value × 100%.6 The Clarke error grid places reference glucose values on the x-axis and glucometer glucose values on the y-axis.12 Zones drawn on the grid indicate appropriateness of clinical decision based on glucometer test results compared with results obtained from the reference chemistry analyzer.13 The standard target range for Clarke's error grid analysis of blood glucose concentrations of 70–180 mg/dL, identical to that used to evaluate glucometer accuracy in humans, was used.12 In addition, because a consensus target range for glucose is not yet available in horses, a Clarke error grid analysis modified for use in horses with the strictest possible target range equal to that of the reference chemistry analyzer reference range (70–135 mg/dL) was evaluated. Zone A contains glucose values that differ from the reference analyzer by <20% or are <70 mg/dL when the reference analyzer also measures a value of <70 mg/dL.12 Zone B contains glucose values differing from the reference analyzer by >20% that would result in benign or no treatment decisions.12 Values falling in zones A and B are associated with clinically correct or benign decisions, whereas values falling in zones C, D, and E are associated with errors in clinical decision because of under- or overestimation of hyperglycemia and hypoglycemia, or failure to recognize euglycemia.7 Based on current recommendations, a point-of-care glucometer is considered acceptable if at least 95% of observations fall into zone A, <5% fall into zone B, and 0% fall into zones C, D, and E.14 Lin's concordance correlation coefficient, evaluating correlation between glucometer values and reference chemistry glucose values, was reported with 95% confidence intervals (95% CI). Pearson's correlation analysis was reported as r values. Precision between paired glucometer test values was evaluated by calculating coefficient of variation (CV). CV was calculated using the following formula: CV = Standard deviation /Average glucometer value × 100. Samples were omitted from calculations if the PCV or reference glucose were outside the range of the glucometer device.
Adult horses had a median age of 11 years [IQR: 5–15 years]. Presenting complaints for horses considered healthy on admission included nonseptic, noncatastrophic limb injury (11), lameness examination (8), dental examination (3), pre-operative cryptorchidectomy or ovariectomy (2), and head shaking (1). Presenting complaints for horses admitted for emergency evaluation and considered critically ill included colic (15), severe local sepsis with concurrent signs of systemic illness (3), retained placenta (2), placentitis (1), colitis (1), laminitis (1), influenza-associated acute respiratory distress syndrome (1), and acute blood loss (1). Foals had a median age of 4 days [IQR: 1–6 days]. Presenting complaints for foals considered healthy on admission included general health examination (20), companion to mare (3), and angular limb deformity (2). Presenting complaints for foals admitted for emergency evaluation and considered critically ill included prematurity (7), diarrhea (7), sepsis (6), colic (2), failure of passive transfer (1), neonatal iso-erythrolysis (1), and respiratory disease (1).
Reference analyzer values for glucose in adult horses are shown in Figure 1. Healthy adult horses had a median glucose concentration of 111 mg/dL [IQR: 103–117 mg/dL] and critically ill horses 114 mg/dL [IQR: 104–134 mg/dL]. Glucose measurements were not statistically different between healthy and critically ill adult horses (P= .1745). Reference analyzer values for glucose concentration in foals are shown in Figure 2. Healthy foals had a median glucose of 154 mg/dL [IQR: 143–166 mg/dL] and critically ill foals 145 mg/dL [IQR: 122–166 mg/dL]. Glucose measurements were not statistically different between healthy and critically ill foals (P= .2471). Median PCV for adult horses was 38% [IQR: 35–40%]. Median PCV for foals was 38% [IQR: 34–42%]. The calibration solution was within the range indicated on the glucometer test strip vial on each day of the study. Two samples were omitted from analysis for blood concentrations out of range of the glucometer device: 1 sample from a critically ill adult horse with a PCV of 66% and a reference glucose concentration of 253 mg/dL for which the glucometer read 139 and 141 mg/dL, and 1 sample from a critically ill foal with a reference glucose concentration of 12 mg/dL for which the glucometer read ‘LO’ for both replicates. Median absolute bias of the glucometer to the reference chemistry analyzer was 6.1% [IQR: 2.8–11.1%] in adult horses and 5.0% [IQR: 1.4–8.9%] in foals. The standard Clarke error grid is shown in Figure 3. When glucometer measurements from adult horses and foals were combined, 96.94% of values were within zone A and 100% of values were within zones A and B. When the Clarke error grid was modified using a target range representing normal equine blood glucose concentrations, zone distribution of glucose concentrations did not differ from Clarke error grid analysis utilizing the standard range (Fig 4). Lin's concordance correlation coefficients were 0.947 [95% CI: 0.927–0.967, P < .001] for horse and foal combined glucose measurements, 0.886 [95% CI: 0.829–0.944, P < .001] for adult horses, and 0.953 [95% CI: 0.927–0.979, P < .001] for foals. Pearson's r was 0.961 (P < .001) for horse and foal combined glucose measurements, 0.928 (P < .001) for adult horses, and 0.960 (P < .001) for foals. Median CV for glucometer measurements was 1.3% [IQR: 0.6–2.3%] in adult horses overall, 1.3% [IQR: 0.6–2.1%] in healthy adults, 1.4% [IQR: 0.5–2.6%] for critically ill adults, 1.3% [IQR: 0.5–2.5%] for foals overall, 0.9% [IQR: 0.0–2.3%] for healthy foals, and 1.4% [IQR: 1.0–3.2%] for critically ill foals.
This is the 1st report in horses in which a point-of-care glucometer utilizing fresh whole blood showed good agreement with a reference plasma chemistry analyzer. In a previous report in neonatal foals, a point-of-care glucometer designed for human use consistently underestimated blood glucose relative to the reference chemistry analyzer and these differences were considered clinically important.15 In the same report, a point-of-care blood gas instrument consistently overestimated blood glucose.15 Overestimation of blood glucose by point-of-care blood gas instruments also has been reported in human critical care studies.7 In adult horses admitted on emergency basis, whole blood glucose measurements by a point-of-care glucometer designed for human use were poorly correlated with the laboratory standard.11 However, when plasma was separated and tested with the glucometer there was good agreement.11 Improvement of accuracy with equine plasma glucose measurement with a glucometer designed for human whole blood is likely due in part to the difference in distribution of glucose between plasma and erythrocytes in humans and horses. Separation of plasma before testing introduces additional time and error into analysis, counteracting advantages of point-of-care technology. Whereas laboratory plasma glucose measurement is considered a standard for the industry, maintenance of a point-of-care glucometer of known accuracy negates the need to incorporate known errors in glucometer values into clinical protocol.
This study was designed to evaluate clinical patients in a hospital setting. Studies performed in a hospital setting are more informative than those conducted in a controlled laboratory setting because they are more relevant to the point-of-care environment and field conditions.6 Studies of point-of-care devices should be carried out on the target population to gain the most clinically relevant information.16 The majority of bias detected in the use of point-of-care glucometers in a clinical setting has been attributed to operator error.17 In the present study, all glucometer measurements in horses and foals were performed by a single operator (E.H.), which may have diminished bias because of operator error. However, the operator followed all the manufacturer's instructions and did not have specialized training with the glucometer device or clinical laboratory practices.
Point-of-care glucometers are limited by known interference of specific critical care variables such as oxygen pressure, hematocrit, pH, temperature, and common drug concentrations including mannitol, dopamine, and ascorbic acid.18,19 The glucometer evaluated in this study utilized glucose dehydrogenase enzyme and coulometric methods. Glucose dehydrogenase enzyme glucometers had lower glucose concentration differences compared with a reference instrument than glucometers utilizing the glucose oxidase enzyme when interference because of oxygen pressure, hematocrit, and pH were assessed.4,20,21 Glucometers utilizing coulometric methods outperform those utilizing amperometric methods because they evaluate total charge generated instead of steady state current, allowing submicroliter volume sampling and minimizing errors because of hematocrit and temperature affecting only the rate of reaction and not total charge.22 Altitude and variable concentrations of unconjugated bilirubin, creatinine, triglycerides, dopamine, and ephedrine did not aberrantly affect human glucose concentrations in a report investigating a glucometer utilizing similar technology to the veterinary glucometer used in this report.22 Whereas the results achieved with this veterinary glucometer when applied to equine whole blood were promising, evaluation of a single sample with a PCV above the high end of the hematocrit range of 65% underestimated glucose relative to the reference analyzer. Until further study is completed investigating the effect of PCV on glucometer performance in horses, caution should be exercised when interpreting glucometer readings in horses with concurrent hemoconcentration. Other authors evaluating glucose measurement of whole blood samples in horses have not detected an effect of PCV on measurement accuracy.11
The veterinary glucometer evaluated performed well in both horses and foals. Accuracy and precision were within acceptable limits for a point-of-care glucometer device. Equine hospitals should follow the manufacturer's instructions with regard to glucometer operation and calibration to obtain consistent and comparable results to clinical laboratory analysis.
aHassel DM, Hill AE, Rorabeck RA. The effect of hyperglycemia on survival in 300 horses with acute gastrointestinal disease: A retrospective study, 2003–2005. Ninth International Equine Colic Research Symposium, June 15–18, 2008, Liverpool, UK (abstract #26)
bAlphaTRAK blood glucose monitoring system meter, serial number VBGH 326-32966, Abbott Animal Health, Abbott Laboratories, North Chicago, IL
cAlphaTRAK blood glucose monitoring system control solution lot #7F2K01, Abbott Animal Health, Abbott Laboratories
dAlphaTRAK test strips lot #087113, Abbott Animal Health, Abbott Laboratories
eHitachi 917 Blood Chemistry System, Hitachi Ltd, Tokyo, Japan
fSTATA 10.1, StataCorp, College Station, TX
The authors thank Abbott Animal Health for providing the veterinary glucometer, test strips, and funding for laboratory analysis and Elizabeth M. Cozzi, PhD, for assistance with data analysis.