SEARCH

SEARCH BY CITATION

Keywords:

  • Camelid;
  • Glucometer;
  • Hyperglycemia;
  • Hypoglycemia;
  • Insulin

Abstract

  1. Top of page
  2. Abstract
  3. Acknowledgments
  4. References

Purpose

Hospitalized alpacas are often hyperglycemic requiring frequent blood glucose testing.

Objectives

To compare the performance of 4 brands of glucometers with a laboratory-based analyzer (LCA) over a range of glucose concentrations in alpacas.

Animals

Four healthy male alpacas.

Methods

A 2-treatment cross-over study was utilized. The alpacas were given 0.4 U/kg of regular insulin intravenously and then 500 mg/kg of dextrose intravenously with a 1 week washout period between treatments. Blood samples were collected from 10 minutes before until 6 hours after drug administration. Glucose concentrations were measured in whole blood and plasma samples on 4 glucometers, and serum glucose was measured on an LCA.

Results

Glucometer performance varied depending on whether glucose concentrations were measured in plasma or whole blood. Based on error grid analysis, the Precision Xtra and One Touch Ultra 2 glucometers were clinically acceptable for testing whole blood samples, whereas the Accu-Chek Aviva and Nova StatStrip Xpress glucometers were clinically acceptable for testing plasma samples in comparison with serum glucose concentrations determined by the LCA. All glucometers had systematic and proportional biases that varied based on sample type.

Conclusions and Clinical Importance

Human-based glucometers in alpacas should be used cautiously, particularly at higher blood glucose concentrations. The blood sample type (plasma or whole blood) can alter meter performance when compared with serum glucose concentrations and potentially lead to errors in clinical decisions.

Abbreviations
AC

Accu-Chek Aviva

CV

coefficient of variation

EDTA

ethylenediamine tetra-acetic acid

LCA

laboratory chemistry analyzer

NX

Nova Xpress

OTU

One Touch Ultra 2

POC

point-of-care

PX

Precision Xtra

RBC

red blood cell

RCF

relative centrifugal force

WB

whole blood

Alpacas commonly experience hyperglycemia because of stress and a variety of disease states. Hyperglycemia in sick alpacas can persist for days until the primary medical condition resolves and normal metabolic feedback pathways reassert control. Alpacas have poor glucose tolerance and are relatively insulin resistant, which complicates medical management and the response to exogenous treatments.[1, 2] Hypoglycemia is less commonly observed in alpacas but can occur with a variety of conditions such as sepsis, inadequate nutrition in neonates, or insulin therapy.

Veterinary hospitals that treat alpacas with these glycemic disorders use a variety of point-of-care (POC) glucometers designed for use in humans. Reports evaluating these glucometers in humans and animals have found variations in performance between meters and with disease states, medications, and sample types.[3, 4]

A currently available veterinary glucometer has been evaluated in alpacas but is not currently marketed for use in this species.[5] Alpacas also have unique, elliptical red blood cells (RBC) that interfere with many table-top blood analyzers, but it is unknown whether there is an effect on POC glucometers.[6]

The objective of this study was to compare 4 commercial POC glucometers against the gold standard of a laboratory chemistry analyzer (LCA) over a range of blood glucose concentrations in alpacas. To evaluate whether alpaca RBCs affected the meters, plasma glucose was also evaluated. It was hypothesized that the whole blood glucose results from the POC meters would be clinically acceptable for use, using serum glucose reference intervals, in alpacas across a wide range of glucose concentrations.

Four apparently healthy, 1.5–2-year-old, intact male alpacas from the Colorado State University (CSU) Department of Clinical Sciences research herd were used for this study. The CSU Institutional Animal Care and Use Committee approved all procedures before conducting this research.

For determining coefficient of variation (CV) in normoglycemic and hyperglycemic blood, a whole blood sample (3 mL) was collected by jugular venipuncture from 1 alpaca before and 10 minutes after receiving 500 mg/kg of dextrose via the jugular vein.

For the remainder of the study, intravenous catheters were aseptically placed in the right jugular vein of the alpacas at least 12 hours before the first blood sample collection to allow the animals to acclimate to the catheter, handling, and indoor facilities. In this 2-treatment, cross-over study, each alpaca was administered 0.4 U/kg of regular insulin (week 1) and 500 mg/kg of dextrose (week 2) with a 1-week washout period between each treatment. The insulin and dextrose were administered through the jugular catheter. The dextrose was diluted from 50 to 25% using an equal volume of 0.9% saline and administered over approximately 2–3 minutes. Following drug administration, the catheter was flushed with 4–6 mL of heparinized saline (1 mL of 1 : 1,000 heparin in 250 mL of 0.9% saline). After insulin treatment, if signs of hypoglycemia were observed (head tremors, excessive humming, and skittish behavior) or whole blood glucose was less than 30 mg/dL on any POC glucometer, the alpaca was to be administered a dextrose solution IV (5–12.5% solution) depending on severity of hypoglycemia and signs.

Catheter blood sample collection was performed using a 3-syringe technique. Five milliliters of blood were a collected into a 6-mL syringe containing 1 mL of heparinized saline. Twenty milliliters of blood were then removed from the catheter and tested as described below. The heparinized saline/blood sample was injected back into the catheter, and the catheter flushed with 3–4 mL of heparinized saline. Blood samples were collected at specific time intervals starting 10 minutes before drug administration, and then at 5, 10, 15, 30, 60, 90, 120, 180, 240, and 360 minutes after drug administration. The jugular catheter was removed at the end of the blood collection period. We had previously verified that treatment through the same catheter used for blood collection did not affect glucose concentrations by comparing concurrent blood draws from the catheter and venipuncture.

Following blood collection, the whole blood sample was immediately tested on the 4 POC glucometers. The remainder of the blood was divided between serum, sodium heparin, and K3 EDTA vacutainer tubes.1 To minimize variations because of RBC glucose utilization, plasma and serum glucose concentrations were measured within 1 hour of collection. The sodium heparin and EDTA tubes were centrifuged for 5 minutes at 1,000 relative centrifugal force and plasma glucose measured on the POC glucometers. Serum tubes were allowed to clot, then centrifuged (as above), and serum glucose was measured on the LCA.2

The 4 POC glucometers were calibrated and operated according to the manufacturers’ instructions. The PX and AC meters came with a calibration strip with each box of test strips. The other 2 meters were not calibrated as there was no calibration solution available at the time of the study. The meters and test strips were maintained and operated within a temperature range of 65–75°F (18–24°C). All meters utilized single-use test strips that measured blood glucose through electrochemical technology but used different enzymes to create an electrochemical gradient for measurement. The Accu-Check3 (AC) meter and Precision Xtra4 (PX) meter utilized a glucose dehydrogenase enzyme-based biosensor. The AC manufacturer's range for blood glucose monitoring was 10–600 mg/dL, and the PX manufacturer's range was 20–500 mg/dL. The One Touch Ultra 25 (OTU) meter and Nova Xpress6 (NX) meter measured blood glucose concentration with a glucose oxidase biosensor. The OTU manufacturer's range for blood glucose monitoring was 20–600 mg/dL, and the NX manufacturer's range was 10–600 mg/dL. The glucose ranges were provided with the manufacturers’ test strips.

Statistical analyses included CV, percentage differences [(POC − LCA)/LCA × 100], Bland–Altman difference plots, and error grid plots. Statistical analyses were performed using a commercial statistical software package.7 Only POC glucose concentrations between 20 and 500 mg/dL were included. We established critical decision limits at glucose concentrations of ≤60 and ≥300 mg/dL based on clinical experience for increased likelihood of administering dextrose or insulin for glucose concentrations outside these limits.

The coefficient of variation (CV) was calculated for each POC glucometer using normoglycemic and hyperglycemic blood samples. The glucose concentration was repeatedly measured 10 times on each POC glucometer for each sample.

Bland–Altman plots were evaluated to assess agreement between the POC glucometers and the LCA glucose concentrations using methods for multiple observations per test subject.[7] The plots describe the limits of agreement (LOA) between 2 methods as the mean difference between measurements ±1.96 standard deviations of the differences. Additionally, by visually examining the layout of the data, one can detect systematic biases (mean difference) and proportional biases (positive or negative slope in the data). Acceptance criteria were established based on a 3-tier LOA. Because the effects of a large LOA would be more critical at lower glucose concentrations, we required an LOA of ±10 mg/dL at LCA glucose concentrations ≤60 mg/dL, but allowed an LOA of ±50 mg/dL at LCA glucose concentrations ≥300 mg/dL for acceptable results. An LOA of ±30 mg/dL was used for glucose concentrations between 60 and 300 mg/dL.

Error grid plots were developed to evaluate clinical decision-making if the POC glucometer was used rather than the LCA. We modified the Clarke error grid analysis method and acceptance criteria because our critical limits were different than those used in humans.[8] Ten percent limits were plotted around the perfect correlation line and treatment limits established at ≤60 mg/dL or ≥300 mg/dL. The POC glucometer would be considered acceptable if at least 95% of the POC glucometer readings were within zone A. This zone was defined as the region where the reading for both the POC glucometer and LCA was between 60 and 300 mg/dL, if the POC glucometer reading was <60 mg/dL and the LCA value was also <60 mg/dL, and if the POC glucometer reading was >300 mg/dL when the LCA value was also >300 mg/dL. Zone B was defined as POC glucometer readings that would lead to overtreatment for either hypo- or hyperglycemia. Zone C was defined as POC glucometer readings that would lead to inadequate treatment for either hypo- or hyperglycemia. Zone D was POC glucometer readings that were opposite of the LCA glucose readings, leading to treatment of hypoglycemia rather than hyperglycemia or vice versa.

A total of 88 blood samples were obtained from the 4 alpacas. All of the POC meters except for the NX meter had 1–7 readings that were not included in the statistical calculations because they were above or below the operating range for the meters. Blood glucose concentrations peaked by 5 minutes after dextrose administration. No intravascular hemolysis was detected after administration of the dextrose. After insulin administration, blood glucose concentration began to decrease at the 10-minute collection time and the lowest blood glucose concentration was reached at either the 90- or 120-minute collection time. All animals returned to pretreatment glucose concentrations by 6 hours. Three of the 4 alpacas were administered supplemental dextrose because of signs of hypoglycemia or low whole blood glucose concentrations, but none required a 2nd treatment. No other adverse effects were seen with the insulin or dextrose administrations.

The CVs for all POC meters were larger for the hyperglycemic sample versus the normoglycemic sample. The normoglycemic and hyperglycemic CVs were smaller for the OTU (1.07 and 3.92%) and NX (1.63 and 4.01%) than those for the AC (3.01 and 4.76%) and PX (4.37 and 5.26%).

The percentage of POC glucometer-determined glucose concentrations that were within 5, 10, 15, or 20% of the LCA-determined glucose concentrations is shown in Table 1. Results differed between the POC glucometer depending on whether plasma or whole blood was analyzed. None of the meters had 100% of the samples within 5, 10, or 15% of the LCA results. The NX meter had 100% of the values within 20% of the LCA results but only on the sodium heparin plasma samples. The AC meter had no samples within 20% of the LCA results on the whole blood samples.

Table 1. Percentages of point-of-care (POC) glucometer results that were within 5, 10, 15, and 20% of the laboratory chemistry analyzer glucose results [(POC − LCA)/LCA × 100] and clinical decision results from error grid analyses. Error grid analysis was used to determine the percentage of POC glucometer results that would have potentially led to inappropriate treatment decisions
MeterPOC Results within Range of LCA Results (%)Error Grid Analysis (% Samples per Zone)
5101520Zone AZone BZone CZone D
  1. AC, Accu-Chek Aviva; EDTA, plasma from EDTA tube; HEP, plasma from sodium heparin tube; NX, Nova StatStrip Express; OTU, One Touch Ultra 2; PX, Precision Xtra; WB, whole blood.

AC WB0000798130
NX WB031851796150
OTU WB50758489100000
PX WB2160748499100
AC EDTA631708795050
NX EDTA4072939798020
OTU EDTA4571801740
PX EDTA710121882850
AC HEP24708791100000
NX HEP658797100100000
OTU HEP0245811640
PX HEP581113821440

The degree of agreement between the POC glucometers and the LCA glucose concentrations, as determined using the Bland–Altman difference plots, varied depending on whether the POC glucometers were analyzing whole blood or plasma (Table 2). When whole blood glucose was analyzed, all of the POC glucometers had some degree of proportional bias as blood glucose concentrations increased (Fig 1). The PX meter had the smallest amount of proportional bias that occurred primarily at glucose concentrations <150 mg/dL; this meter met the 3-tier LOA criteria we established. The OTU glucometer also met the 3-tier LOA criteria. The NX glucometer only met the low and high LOA criteria and had more proportional bias and wider overall LOAs than the PX and OTU glucometers. The AC glucometer had a large proportional bias and exceeded the LOA criteria.

Table 2. Bland–Altman difference plot results comparing 4 point-of-care (POC) glucometers to a laboratory chemistry analyzer (LCA). The data were calculated using multiple readings per individual. Samples processed on the POC meters were whole blood, EDTA plasma, and sodium heparin plasma. The LCA glucose results were obtained from serum. The mean difference is the difference between the LCA and the POC meter. Limits of agreement (LOA) represent ±1.96 times the standard deviation of the differences between the LCA and POC glucometer
MethodMean Difference (mg/dL)LOA (mg/dL)
  1. AC, Accu-Chek Aviva; EDTA, plasma from EDTA tube; HEP, plasma from sodium heparin tube; LCA, laboratory chemistry analyzer; OTU, One Touch Ultra 2; PX, Precision Xtra; NX, Nova StatStrip Express; WB, whole blood.

AC WB50.9−12.8 to 114.7
NX WB36.3−10.6 to 83.2
OTU WB−4.0−33.6 to 25.5
PX WB−11.3−38.7 to 16.1
AC EDTA19.5−4.8 to 43.9
NX EDTA12.4−22.5 to 47.3
OTU EDTA−59.6−118.4 to −0.7
PX EDTA−46.1−101.3 to 9.2
AC HEP12.2−9.4 to 33.7
NX HEP−2.8−26.2 to 20.7
OTU HEP−64.5−125.8 to −3.2
PX HEP−50.0−111.0 to 11.0
image

Figure 1. Bland–Altman difference plots using methods for multiple observations per test subject. The plots compare whole blood glucose results from 4 point-of-care (POC) glucometers to serum glucose values obtained from a laboratory chemistry analyzer (LCA). The x-axis is the average glucose concentration from the LCA and POC glucometer, and the y-axis is the difference between the LCA and POC glucose concentrations. The dashed lines indicate the 95% limits of agreement (±1.96 times the standard deviation of the differences), and the solid line indicates the mean difference between both methods. AC, Accu-Chek Aviva; LCA, laboratory chemistry analyzer; OTU, One Touch Ultra 2; PX, Precision Xtra; NX, Nova StatStrip Express.

Download figure to PowerPoint

Results between EDTA and sodium heparin plasma concentrations were similar to each other (plots not shown). When using plasma samples, the AC and NX glucometers compared more favorably with serum glucose concentrations than the OTU and PX glucometers. All 4 meters had some degree of proportional bias, with the NX glucometer having both the smallest systematic and proportional bias when testing sodium heparin plasma glucose. The AC and NX were acceptable according to our LOA criteria for plasma samples, whereas the OTU and PX meters had proportional biases at glucose concentrations <150–200 mg/dL and the LOAs exceeded the criteria.

Error grid analysis charts for whole blood samples are included as representative examples (Fig 2). The error grid analyses (Table 1) showed that the AC and NX glucometers were clinically acceptable when testing heparin and EDTA plasma samples and the OTU and PX glucometers were clinically acceptable when testing whole blood samples in comparison with serum glucose concentrations. None of the glucometers had readings that fell into zone D.

image

Figure 2. Error grid plots comparing whole blood glucose results from 4 point-of-care glucometers and serum glucose results obtained from a laboratory chemistry analyzer (LCA). The solid diagonal line is x = y, and dashed lines represent the ±10% limits of agreement of the x = y line. AC, Accu-Chek Aviva; LCA, laboratory chemistry analyzer; OTUm One Touch Ultra 2; PX, Precision Xtra; NX, Nova StatStrip Express.

Download figure to PowerPoint

This study compared glucose concentrations determined in whole blood and plasma by 4 different POC glucometers with serum glucose concentrations determined by LCA. Many medical conditions such as anemia, hypertriglyceridemia, and hyperproteinemia, as well as a variety of drugs, can affect blood glucose measurements.[4, 9] To establish a baseline performance of the POC glucometers, we selected alpacas that were apparently healthy, not anemic, and not on any medications.

Testing of whole blood and plasma samples was performed to determine whether the elliptical RBCs, hematocrit, or other unknown factors could also contribute to variations in POC glucometer readings. Interestingly, two of the POC glucometers had acceptable results when glucose concentration was measured in whole blood and the other two when using plasma. The reason for this is unknown but is likely related to differences in calibration and sample correction functions between meters.[4] According to owner's manuals or test strip inserts, the manufacturers of the POC meters used in this study applied different calibration methods using whole blood, plasma, or a correction factor. The correction factors adjust for sample type as the distribution of glucose in the blood varies between plasma and whole blood.[4] The glucose distribution ratio has been established in humans and some animals, but has not been documented in camelids.[10] Plasma and whole blood comparisons using the AC glucometer have been performed in alpacas and horses, and results were similar to what we observed in this study.[3, 5]

Other possible sources of variation between POC glucometer and LCA results include operator factors, collection method, test strips, and RBC contact time on the serum and plasma. Two different operators ran samples on the POC glucometers. Neither operator received specific training on the use of any of the POC glucometers, but both followed manufacturer instructions provided with each meter. Operators noticed that the amount of blood applied to the test strip before the OTU meter began to count down could alter the glucose levels. This was noted in the manufacturer's instruction guide for the OTU; however, the degree of the effect was not known. The instructions for the PX and AC glucometers indicated the devices would not start counting down until a sufficient blood sample was applied. Application of plasma rather than whole blood did not appear to alter any of the glucometers’ functionality.

To decrease affects of test strips, all were stored and handled as recommended by the manufacturers. Because of the number of samples collected, different lots of blood glucose test strips were used over the course of the study which could have affected glucose results. It was determined that the different test strip enzyme systems were not a factor because each of the enzyme systems was used within a meter that performed best with whole blood or with plasma.

Glucose utilization by the RBC's in the plasma and serum blood tubes was expected and has been reported to be 5–7% per hour.[11, 12] We did not attempt to quantitate this utilization but instead tried to maintain consistency in timing of the sample processing. However, there would still be minor decreases in glucose concentrations because the serum tubes required time to clot and short delays occurred early in the sample collection because of the number of samples to be run on the LCA.

Coefficients of variation were similar between meters and variation increased as glucose concentrations increased. The increased CV and data scatter observed on the Bland–Altman and error grid plots at the higher glucose concentrations could have been because of the smaller number of samples in this range or design factors in the meters. It is known that there is more variation in POC glucometer readings at higher glucose concentrations; different organizations and testing societies have developed different levels of acceptable performance criteria for glucometers based on higher or lower glucose concentration ranges.[4] The different performance criteria recommended for glucometers used for people range from within ±5% at all glucose concentrations to within 25% at low and 20% at high concentrations when compared with a laboratory method.[4] None of the glucometers met the American Diabetes Association requirement of ±5% for all glucose concentrations on the alpaca samples, which was not surprising because these requirements are the most restrictive of the various organizations.[4] Even so, only the NX meter (testing heparin plasma) had all results within 20% of the LCA results. All of the glucometers had some degree of systematic and proportional bias as demonstrated in the Bland–Altman plots, but this was expected because of the known differences in LCA and glucometer glucose measurement techniques and sample types.

Clinical application for POC glucometer use is obviously easier and quicker when testing whole blood samples rather than using heparin or EDTA plasma samples. Therefore, for testing whole blood, the OTU and PX meters would be more appropriate as these meters were deemed acceptable for this use.

The AC and NX meters were deemed acceptable only when using plasma samples when compared with serum glucose concentrations. Therefore, plasma is recommended for these meters if decisions will be based on serum glucose reference limits and they are not recommended for the use of whole blood. Alternatively, it may be possible to mathematically correct for the systematic and proportional biases seen when using whole blood, to obtain values comparable with serum glucose concentrations. Another option, if one desires to use the AC or NX meters to measure glucose concentration in whole blood, would be to validate independent reference intervals and critical decision limits for that particular glucometer, but the cost of this would likely exceed the cost of buying a different glucometer.

The OTU and PX meters provide comparable glucose concentration results to serum glucose concentrations in healthy alpaca blood. Now that the baseline performance of these glucometers has been shown to be acceptable, further work to evaluate performance in sick, hospitalized alpacas is indicated.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Acknowledgments
  4. References

Funding provided by the Colorado State University Center for Companion Animal Studies, PVM Student Research Grant Program. The authors thank Pam Spicciati for assistance collecting samples and the Clinical Pathology Laboratory staff for processing samples.

Conflict of Interest: Authors disclose no conflict of interest.

Footnotes
  1. 1

    Vacutainer Tubes, Becton Dickinson, Franklin Lakes, NJ

  2. 2

    Roche Hitachi 917 Blood Chemistry Analyzer, Hitachi Ltd Tokyo, Japan

  3. 3

    Accu-Chek Aviva, Roche Diagnostics, Indianapolis, IN

  4. 4

    Precision Xtra, Abbott Laboratories, Abbott Park, IL

  5. 5

    One Touch Ultra 2, Lifescan, Inc. Milpitas, CA

  6. 6

    Nova StatStrip Xpress, Nova Biomedical Corporation, Waltham, MA

  7. 7

    MedCalc Software, Version 11.6, Mariakerke, Belgium

References

  1. Top of page
  2. Abstract
  3. Acknowledgments
  4. References