Accuracy evaluation of two portable blood glucose meters in feline patients using whole blood samples

ABSTRACT: Using portable blood glucose meters (PBGMs) to measure blood glucose (BG) concentration is a common procedure in veterinary practice. Our objective was to evaluate the analytical and clinical accuracy of a human PBGM (Accu-Chek Performa®), (AC) and a veterinary PBGM (GlucoCalea®), (GC) in feline patients. Central venous blood samples were collected from 48 cats at a Brazilian Veterinary teaching hospital. Two devices from each model were used and compared to a reference method (RM). Analytical accuracy was assessed according to ISO 15197:2013 requirements for human PBGMs. Data were compared using Wilcoxon’s nonparametric test and represented by Bland-Altman plots. Hematocrit’s effect on BG measurements was evaluated by the Spearman correlation coefficient. Clinical accuracy was determined using error grid analysis (EGA). Values of BG were significantly higher in all PBGMs compared to the RM. Although ISO’s analytical accuracy requirements could not be met by any of the devices, AC meters were more accurate than GC meters. All AC measurements - but not GC ones - were within zones A and B of the EGA, meeting ISO requirements for clinical accuracy. Significant hematocrit interference was observed in all devices. Therefore, AC showed greater accuracy compared to GC using feline whole blood samples.


INTRODUCTION
The use of portable blood glucose meters (PBGMs) is common in veterinary clinical practice to determine blood glucose (BG) concentration and to guide decision-making due to their lower blood volume requirements allowing less invasive blood collection and quick results.The PBGMs are especially useful when multiple blood samples must be acquired within a short period, e.g., for obtaining a serial BG curve or diabetic ketoacidosis monitoring (WESS & REUSCH, 2000;COHEN et al., 2009).
Ciência Rural, v.53, n.10, 2023.Moresco et al.Although, BG values obtained by PBGMs are strongly correlated with measurements obtained by a reference method (RM, i.e., automated chemistry analyzers), its analytical accuracy is still questionable (WESS & REUSCH, 2000;JOHNSON et al., 2009).Variables such as hematocrit, blood type (venous versus capillary), and device enzymatic method can all influence acquired values (STEIN & GRECO, 2002).Therefore, clinicians should be aware of possible sources of error when interpreting the results.Thus, PBGM's accuracy evaluation is necessary owing to the continuing launch of new devices on the market.The objective of this study was to evaluate the analytical and clinical accuracy of a human PBGM and a veterinary PBGM using whole blood samples from feline patients.

Patients
The study was conducted at the Veterinary Clinic Hospital, Federal University of Rio Grande do Sul, Brazil.Forty-eight feline patients were randomly allocated among the general practice population during regular appointments in which blood sampling was indicated.Informed consent was obtained from all cat owners to allow glucose measurement in the blood collected for the study.

Blood collection and BG determinations
Blood samples from the jugular vein were collected as part of the diagnostic workup of each cat after minimal physical restraint using a 21G needle connected to a five-milliliter syringe.Blood was immediately fractionated into tubes (Vacutainer, BD, New Jersey, USA) containing K2 ethylenediaminetetraacetic acid (EDTA) for hematocrit (microhematocrit method at 9520 g for 5 minutes) and hematological evaluations (0.5 mL), sodium fluoride EDTA for BG evaluation by the RM (2 mL), and without anticoagulant for any other serum measurements needed for each specific cat (2 mL).Blood samples were immediately handled by the Hospital's Veterinary Clinical Analysis Laboratory (LACVet).Mean BG concentration by the enzymatic colorimetric glucose oxidase method (Labtest Diagnostica, Lagoa Santa, MG) was obtained in an automatic spectrophotometer (CM 200, Wiener Lab Group, Argentina) as the RM in duplicates.
The remaining blood in the syringe was used to assess BG concentration with both PBGM models.Blood glucose concentration was assessed using four devices: two identical human PBGMs (Accu-Chek Performa ® , Roche Diagnostics, Basel, Switzerland; AC1 and AC2), and two identical veterinary PBGMs (Gluco Calea ® , Med Trust, Marz, Austria; GC1 and GC2).All exams were performed in duplicate.Both models evaluate BG concentration by the electrochemical method.For the assessment of low BG values, data were obtained from additional 12 blood samples kept for 12 hours in EDTA tubes at room temperature before analysis by PBGMs and the RM (FOBKER, 2014).

Device technical information
According to manufacturers, human AC devices require a minimum blood volume of 0.6 µL and their BG detection limits are 10 to 600 mg/dL.It operates without interference within the 10-65% hematocrit range.The test strip uses an enzymatic reaction of glucose dehydrogenase.The veterinary GC devices require a minimum blood volume of 0.5 µL and their BG detection limits are 20 to 600 mg/dL.It operates without interference within the 35-55% hematocrit range.The test strip uses an enzymatic reaction of glucose oxidase.

Accuracy
Analytical accuracy was assessed according to the INTERNATIONAL ORGANIZATION FOR STANDARDIZATION (ISO 15197:2013) requirements for human PBGMs (ISO, 2013).For a PBGM to be considered accurate, two conditions must be met: 1) when glucose is <100 mg/dL, 95% of its measurements should not differ by more than 15 mg/dL from the RM value, and 2) when glucose is ≥100 mg/dL, 95% of its measurements should not differ by more than 15% from the RM value.
Consensus error grid analysis (EGA) for insulin-dependent diabetic patients was applied to assess the clinical risk of each measure (i.e., clinical accuracy) (PARKES et al., 2000).Error grid analysis compares the BG values from the RM with the PBGM within five error zones associated with the following risk levels: zone A, clinically accurate; zone B, altered clinical action, but with no or minimal effect on clinical outcome; zone C, altered clinical action likely to affect the clinical outcome; zone D, altered clinical action with considerable medical risk; and zone E, altered clinical action with potentially dangerous consequences.For a PBGM to be considered accurate, ISO 15197:2013 stipulates that 99% of values should lie within zones A and B (ISO, 2013).

DISCUSSION
Neither the human nor the veterinary meter here studied has reached the analytical accuracy parameters required by ISO.Surprisingly, we found that the human PBGM (AC) was more accurate than the veterinary device (GC).The reason for such discrepancy; however, remains unclear.Unlike our findings, some studies have shown that veterinary PBGMs are equally or more accurate than those designed for humans (COHEN et al., 2009;ZINI et al., 2009;KANG et al., 2016).
Despite the availability of veterinary devices, the use of human PBGMs is still common in small animal settings.Given the continuous launch of new devices on the market, assessment of their accuracy in clinical practice and validation for the target species are considered a priority in veterinary medicine (BRITO-CASILLAS et al., 2014;CAPASSO et al., 2019).Few studies have evaluated the accuracy of PBGMs exclusively in cats (WESS & REUSCH, 2000;ZINI et al., 2009;DOBROMYLSKYJ & SPARKES, 2010).The models in this study were chosen because AC had a very good performance in dogs compared to other human PBGMs (BRITO-CASILLAS et al., 2014).Conversely, GC was introduced in the Brazilian market and has been only preliminary evaluated in cats (MALERBA et al., 2018).
In our study, the medians of BG values obtained with both human and veterinary PBGMs were higher than the values obtained by the RM.Several studies have shown that some devices consistently overestimate while others underestimate BG values in small animals (WESS & REUSCH, 2000;JOHNSON et al., 2009;ZINI et al., 2009;BRITO-CASILLAS et al., 2014).Errors in BG measurements may have clinical repercussions or dangerous effects on therapeutic conduct, e.g., iatrogenic hypoglycemia due to insulin overdosing or unnecessary glucose supplementation.In this sense, clinical accuracy assessment of PBGMs should be used in conjunction with analytical accuracy assessment to provide complementary information (WESS & REUSCH, 2000;DOBROMYLSKYJ & SPARKES, 2010;BRITO-CASILLAS et al., 2014;KANG et al., 2016;COSTA et al., 2021).
Recent studies in feline patients evaluating the clinical accuracy of human PBGMs by EGA have obtained all values within zones A and B (DOMORI et al., 2014;MORI et al., 2016), including one study with AC (COSTA et al., 2021).Similarly, our study found 100% of the values for the AC meter within zones A (95%) and B (5%).However, according to EGA, GC presented 15% of the values in zone C, (29% within zone A and 56% within zone B), compromising its clinical accuracy.Therefore, evidence suggested its unsuitability for BG measurement in this species as previously suggested in a research abstract (MALERBA et al., 2018).
Such difference in clinical performance could be partially explained by the interference of the samples' low hematocrit values with BG measurements since GC operates without interference in a much more restricted hematocrit range (35-55%) than AC (10-65%).A wider hematocrit range would be desirable for a veterinary PBGM.The larger the number of erythrocytes in a whole blood sample, the lower the volume of plasma that penetrates the test strip reagent layer, leading to inaccurate results.Thus, hemoconcentration leads to lower BG values, while haemodilution produces higher BG values (RAMLJAK et al., 2013) as observed in our study.However, the correlation between hematocrit and both PBGM analytical accuracy was similar among the four devices studied.Despite this influence, both devices showed acceptable CV% (below 5%) in the different glycemic ranges studied.

CONCLUSION
Accuracy evaluations according to the ISO 15197:2013 criteria before using human or veterinary PBGMs in cats are strongly recommended.Although, none of the devices reached analytical accuracy using feline whole blood samples, Accu-Chek Performa ® is a better option than GlucoCalea ® .The former has shown acceptable clinical accuracy in the cat and can be safely adopted in clinical routine without compromising clinical conduct.The latter, in turn, has produced clinical accuracy errors that could result in mistaken conduct and unnecessary medical risks.Haematocrit interference on both PBGMs' analytical accuracy was documented, showing a negative relationship with BG measurements.