Evaluation of Three Human-Use Glucometers for Blood Glucose Measurement in Dogs

Background Glucometers or portable sensors are used to quickly measure blood glucose at low cost. They are used in veterinary practice and by guardians to monitor diseases that require, as in diabetes mellitus. However, not all commercially available glucometers (human and veterinary) are suitable for this purpose. Hypotheses/Objectives. The objective was to evaluate the analytical and clinical precision of three human-use portable glucometers. Animals. This study evaluated 115 samples in three glycemic ranges (hypoglycemia, normoglycemia, and hyperglycemia) from 82 dogs recruited from veterinary services. Methods The portable glucometers are the FreeStyle Freedom Lite®, FreeStyle Optium Neo®, and On Call Plus® models. Glucometer results were compared with the enzymatic colorimetric glucose oxidase laboratory reference method. Using descriptive and comparative statistical analysis, there were correlations between these devices and the standard method, ISO 15197 : 2003 and ISO 15197 : 2013 standards, and error grid analysis. Results Only the Freedom Lite® device observed a statistical difference when compared with the reference method. Despite the underestimated glucose concentrations assessed with humane devices, all three tested herein showed a positive coefficient. However, none of these achieved all ISO guidelines. Conclusion and Clinical Importance. Although there was wide use of portable humane devices for dog glucose measurements on routine, the results are generally inferior when compared to the reference method. The FreeStyle Optium Neo® glucometer obtained the best result and is therefore the best option among the glucometers evaluated; however, for the first attendance on veterinary routine, all three glucometers had a satisfactory glucose measurement until the reference method availability.


Introduction
Glucometers or portable sensors are devices used to measure blood glucose quickly, easily, and at low cost in both humans and animals [1]. Te small sample volume required and the generation of immediate results are important advantages over automated analyzers [2]. Portable glucometers developed for human use are widely adopted in veterinary practice; however, they should be validated for adequate use in animals, whereas such devices are developed for humans [3]. Likewise, glucometers developed for veterinary use may not provide reliable results [3][4][5]. An inadequate device can result in an incorrect hypoglycemia diagnosis and lead to inaccurate glycemic control [6] and may be used for decision-making for adjustments in insulin therapy for diabetic patients [2].
In general, portable glucometers underestimate blood glucose concentrations compared to the standard method [3,7]. Te higher the hyperglycemia, the greater this difference [2]. Furthermore, samples may sufer interference from conditions such as hematocrit concentration (the lower the hematocrit level, the greater the blood glucose reading on the glucometer, and vice versa) [5]. Other interferences are sample volume, time, test strip stability, altitude, hemolysis, blood temperature, humidity, partial oxygen pressure in blood samples, prandial state, and the type of sample used (venous, arterial, or capillary blood) [1,2,8].
Due to the large number of portable glucometers available commercially, studies evaluating such products are essential. Tis supports the use of reliable glucometers to measure blood glucose concentration in dogs [5]. Te effciency evaluation of portable glucometers used to be performed by comparing the glycemic results obtained using glucometers in tests with the values generated by the glucose oxidase or hexokinase reference methods [1,3,9,10].
Tus, the aim of this study was to evaluate the performance of three portable glucometers for human use, in diferent dog's blood glucose ranges (hypoglycemia, normoglycemia, and hyperglycemia) and to compare with the enzymatic colorimetric glucose oxidase method (reference test) in animals with normal hematocrit concentration.

Animal Inclusion Criteria.
Te experiment was performed between November 2018 and April 2019, and the animals were recruited from the "Universidade Metropolitana de Santos"-Veterinary Service and other regional veterinary clinics located in Santos City, Sao Paulo, Brazil. All owners agreed and signed the authorization for sample use. Were included dogs over six months of age, both sexes, diagnosed with diabetes or another endocrine pathology or yet healthy and who were not hospitalized. Corporation) (coulometry assay; glucose oxidase measurement method). In other words, each sample was measured with the same glucometer three diferent times after the sample collection due to the literature hypothesis that time after blood sampling could interfere with results. All measurements were performed up to four hours after the blood sample.

Glucometers and
Blood samples were collected by venipuncture and were distributed as follows: tubes containing disodium ethylenediaminetetraacetic acid (EDTA) with added fuoride; tubes with EDTA; gel bottles with clot retractors and approximately 0.5 mL of residual whole blood. Te glucometers were divided into three groups (three units of each model): group one (FreeStyle Freedom Lite ® ), group two (FreeStyle Optium Neo ® ), and group three (On Call Plus ® ).
All laboratory analyses were performed on the same day. Blood samples from the tube containing EDTA were used to obtain the hematocrit level by centrifuging whole blood in a microhematocrit centrifuge (CELM ® model LS 3 PLUS), and reading was performed using a hematocrit card. Te tubes containing fuoride EDTA and clot retracting gel were centrifuged to obtain plasma (up to four hours after the blood sample) and serum, respectively. Te fuoridated plasma aliquot was used to measure plasma glucose concentration, and the serum sample was used to analyze the serum concentration of triglycerides and cholesterol. Plasma glucose concentrations and triglycerides and cholesterol measurements were performed using the GOD-trinder (glucose oxidase) and colorimetric (Enzymatic Trinder) methods, respectively. Te readings were performed using commercial kits (Labtest ® ) and an automated biochemical analyzer (Labmax Plenno, Labtest ® ). Measurement of blood glucose was performed using the enzymatic colorimetric glucose oxidase reference method.

Blood Sample Exclusion Criteria.
As recommended by the manufacturer of the automated biochemical analyzer, triglyceride concentrations ≥ 1,000 mg/dL may sufer signifcant interferences in laboratory blood glucose testing. Still, according to the information provided by the glucometer manufacturers, samples with cholesterol concentrations ≥ 500 mg/dL and triglycerides ≥ 3,000 mg/dL may be afected by the results. Terefore, samples that presented triglycerides ≥ 1,000 mg/dL, cholesterol ≥ 500 mg/dL, and hematocrit outside the 37 to 55% range were excluded. Glucometer results of blood glucose concentration that obtained on-quantitative readings (LOW or HI) were also excluded from this work, in addition to laboratory samples containing hemolysis, jaundice, or insufcient blood.

Statistical Analysis and Sampling.
Te sample size was determined using the ISO 15197 : 2013 guidelines, and this document determines that 100 samples should be collected and divided into diferent concentrations of glucose.
Descriptive analysis was performed with an estimate of the mean, median, minimum, maximum, interquartile range, and standard deviation of the quantitative variables. Te fnal data were tested for normal distribution using the Shapiro-Wilk test, and the nonparametric approach was used (Shapiro-Wilk p value <0.001 for the three models). Te correlation between glucometers and laboratory exams was evaluated using Spearman's correlation coefcient and the statistical diference between these using the Mann-Whitney U test. For better visualization of the data, scatterplots and boxplots were produced, comparing the glucometers and the laboratory exam. All tests were considered signifcant when p < 0.05, and these analyses were performed using SPSS 21.0 [11].
Accuracy between the glucometers against the reference method was assessed based on ISO 15197 : 2003 [12] and ISO 15197 : 2013 [13] standards. According to the ISO 15197 : 2003 [12] guidelines, 95% of results may vary up to 15 mg/dL in blood glucose concentration up to 74 mg/dL and have a maximum variation range of 20% for blood glucose ≥75 mg/dL. However, (stricter) ISO 15197 : 2013 [13] guidelines establish that 95% of the results vary up to 15 mg/ dL in blood glucose concentration up to 99 mg/dL and have a maximum variation range of 15% for blood glucose concentration ≥100 mg/dL.
In addition, error grids with the glucometers and the glucose oxidase laboratory reference method were produced [14]. Te error grid is divided into zones that represent the risk due to an incorrect measurement: zone A represents errors without clinical efects (analytical precision); zone B represents values that deviate more than 20% from the reference values, justifying changes in treatment, but without signifcant repercussions for the patient; zone C, values that can induce unnecessary treatment; zone D, the danger of serious errors in treatment; zone E, errors that can induce clinical conduct with dangerous consequences [10]. Error grid analysis is widely used in human medicine and categorizes measurements based on therapeutic consequences [1].
It was performed a Bland-Altman analysis to determine whether there was an agreement between tested glucometers and the reference method. Such an analysis estimates the diferences between measurements and the mean diference, standard deviation, and concordance mean (CI 95%). Also, was used a Passing Bablok regression to evaluate the systematic diference between the standard method and the glucometers. Statistical signifcance was determined as 5%, and analysis was performed in R environment 4.0.4 (R Core Team, 2021).
To defne a glucometer as precise, the ISO 15197 : 2013 [13] standard requires that 99% of the results be in zones A and B [7]. Te error grid method was created in 1987 [15] and modifed in 2000 [10,14]. Tis analysis was performed in the R environment [16] using the "ega" package [17].

Results
Repeated blood glucose concentrations measured between glucometer models using the same blood sample compared with the reference method showed no statistically signifcant diference ( Figure 1). Te three assessed glucometers underestimated the blood glucose concentration compared with the reference method. Tere was a smaller diference in the results obtained using the Optium Neo ® model. A glycemic average of 158 mg/dL was observed using the reference method, and the averages with glucometers were Optium Neo ® (151 mg/ dL); On Call Plus ® (145 mg/dL), and Freedom Lite ® (111 mg/dL) ( Table 2).
Te three assessed glucometers obtained a good correlation when compared with the reference method. Portable device results had a positive coefcient, showing that the values moved in the same direction in relation to the reference method. However, graph analysis indicated that samples higher than 200 mg/dL had more signifcant dispersion between the glucometer results in relation to the reference method ( Figure 2). Also, the results obtained with the Freedom Lite ® model had more signifcant dispersion, especially in the higher blood glucose range.
Te concordance results were presented in Table 3

Discussion
Most portable glucometer blood glucose results are generally inferior to the results measured by the reference methods, and the higher the hyperglycemia, the greater the diference [3,6]. In our study, all three glucometers underestimated the results of blood glucose concentration in relation to the reference method as well previously described in the literature [3,7]. Analyzing the scatterplots, it was also observed that the higher the hyperglycemia, the greater the dispersion of the results in comparison with the reference method. Te mean value for blood glucose concentration obtained by the reference method was 158 mg/dL. Te FreeStyle Optium Neo ® model showed less variation compared to the reference method (151 mg/dL), followed by the On Call Plus ® model (145 mg/dL) and fnally by Freedom Lite ® (111 mg/ dL). Te three evaluated glucometers obtained a good correlation when compared with the reference method. Results from portable devices had a positive coefcient,  showing that the values moved in the same direction compared to the reference method but does not attend the expected accuracy as previously discussed [7].
A study evaluated the FreeStyle Freedom Lite ® model according to the ISO 15197 : 2013 guidelines [7] and reached 10% of results in the desired ranges. Regardless, our study showed that 7% of the results were in the established ranges.
Regarding the error grid analysis, a study showed that Freedom Lite ® glucometer results were 94% in zones A and B [7], and our results using the same model had 100% of values in zones A and B but with 80.87% of the analyses in zone B.
Until the conclusion of our study, there was no previous publication regarding the use of FreeStyle Optium Neo ® and On Call Plus ® models for dogs. However, previous models of the Optium product line have already been evaluated by some authors. On the Medisense Optium ® model evaluation, a study analyzed the error grid and found that 78% of blood glucose concentration results were in zone A, 21% in zone B, and 1% in zone C (99% in zones A-B) [10].
Another study evaluated the Optium Xceed ® model according to the ISO 15197 : 2013 guidelines and also performed an error grid analysis but failed to meet the established guidelines [7]. In such a study, 45% of blood glucose concentrations were in the established ranges, and when compared with our results using the Optium Neo ® model, we obtained 92.2% of the results in the desired glycemic ranges. Regarding the error grid analysis, the literature shows that 99% of the results were in zones A and B with the Optium Xceed ® glucometer [7], and using the Medisense Optium ® model, another study observed 99% of the values in zones A and B [10]. In our study, when assessing the Optium Neo ® model, 100% of the results were in zones A and B which do not difer from that previously described.  To defne a glucometer as precise, the ISO 15197 : 2013 [13] standard recommends that 99% of the obtained results be in zones A and B [7]. All results obtained herein were in the A and B ranges (error grid). However, as the results in zone B deviate more than 20% in relation to the reference method, the authors recommend the use of glucometers that showed better performances in zone A, Optium Neo ® (97, 39%) and On Call Plus ® (86,09%).
Because of the type of sample used (venous, capillary, or arterial), there may be variations in the glucometer results [1,2,8], and diferent samples types could have been used. However, our study does not evaluate that since the main objective was to use only whole blood samples in comparison with the reference method. No glucometer inconsistency was identifed during the calibrations, and the calibration results agreed with the manufacturer's instructions. A limitation of our study was the absence of reproducibility and repeatability analysis which measures the glucometer consistency in repeated tests. Since the Bland-Altman analysis was performed to determine the agreement between each glucometer and the reference method, we observed a signifcant bias, mainly for Freedom Lite. Ten, another analysis was performed to explore such bias. Te Passing Bablok results showed agreement between the glucometers and the reference method. Tese analysis discrepancies could have diferent hypotheses. One of them should be glucose measurement values >300 mg/dL where the discrepancy was more important when compared to <300 mg/dL (     Optium Neo ® had a satisfactory glucose measurement in dogs and can be implemented in veterinary routines for easy and quick glucose evaluation purposes.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.