Accuracy of a flash glucose monitoring system in dogs with diabetic ketoacidosis

Abstract Background A factory‐calibrated flash glucose monitoring system (FGMS; FreeStyle Libre) recently was evaluated in dogs with uncomplicated diabetes mellitus. It is not known if this system is reliable during diabetic ketoacidosis (DKA). Objectives To assess the performance of the FGMS in dogs with DKA and to determine the effect of severity of ketosis and acidosis, lactate concentration, body condition score (BCS), and time wearing the sensor on the accuracy of the device. Animals Fourteen client‐owned dogs with DKA. Methods The interstitial glucose (IG) measurements were compared with blood glucose (BG) measurements obtained using a validated portable glucometer. The influence of changes in metabolic variables (β‐hydroxybutyrate, pH, bicarbonate, and lactate) and the effect of BCS and time wearing on sensor performance were evaluated. Accuracy was determined by fulfillment of ISO15197:2013 criteria. Results Metabolic variables, BCS, and time wearing were not associated with the accuracy of the sensor. Good agreement between IG measurements and BG was obtained both before and after DKA resolution (r = .88 and r = .93, respectively). Analytical accuracy was not achieved, whereas clinical accuracy was demonstrated with 100% and 99.6% of results in zones A + B of the Parkes consensus error grid analysis before and after DKA resolution, respectively. Conclusions and Clinical Importance Changes in metabolic variables, BCS, and time wearing do not seem to affect agreement between IG and BG. Despite not fulfilling the ISO requirements, the FGMS provides clinically accurate estimates of BG in dogs with DKA.

metabolic variables. The syndrome is characterized by a biochemical triad of hyperglycemia, ketosis, and acidosis. [1][2][3][4][5] Treatment of DKA involves IV fluid resuscitation, correction of acid-base and electrolyte derangements, insulin administration, as well as identification and treatment of any concurrent illness. 5 Insulin treatment aims to support cellular glucose uptake, decrease hepatic glucose production, interrupt the process of ketogenesis, and promote ketone metabolism and clearance. 6,7 Frequent glucose monitoring is necessary during treatment as a result of insulin administration, glucose supplementation, and compromised homeostatic mechanisms that are characteristic of ketoacidotic patients. Currently, hospitalized ketoacidotic patients usually are monitored by measuring blood glucose (BG) concentration using a portable blood glucose meter (PBGM). The main limitation of this device is the need for frequent phlebotomies that can lead to iatrogenic anemia (an important cause for increased transfusion requirement and longer duration of hospitalization, especially in small breed dogs and cats), 8,9 or alternatively placement of a second or central catheter for blood sampling, increasing the risk of catheter-related complications, including infection and phlebitis. [10][11][12] Moreover, such BG monitoring methods allow only intermittent assessment of BG concentration (usually every 1-2 hours), limiting the amount of information available on which to base treatment decisions. Finally, these methods can increase patient stress, owner expense, and workload of nursing staff and clinicians. For these reasons, research is being directed toward less invasive methods to monitor BG concentrations continuously in patients with DKA.
In the past 2 decades, there has been growing interest in devices measuring interstitial glucose (IG) concentration, which has been shown to reflect circulating BG concentrations. Several studies have evaluated their accuracy in humans has well as horses, cows, dogs, cats, rats, and rabbits. [13][14][15][16][17][18][19][20][21] The first-generation systems offered only retrospective analysis of glucose concentrations after disconnecting the sensor and uploading the data (continuous glucose monitoring system, CGMS), whereas second-generation instruments measured and displayed the data immediately, allowing direct intervention (real-time CGMS). 22 However, the need for blood collection was not eliminated completely, because these monitoring systems must be calibrated 2 to 3 times per day, requiring BG measurement using capillary or venous blood sampling. 22

| Data collection
In dogs admitted during working hours, the FGMS was applied as soon as the diagnosis of DKA was confirmed; in dogs admitted out-of-hours, application was postponed until the next morning. The sensor was placed on a clipped and clean area of the dorsal part of the neck, and adherence to the skin was further ensured by additional tape (Pic Solution Soffix Stretch, Pikdare Srl, Como, Italy) and bandage (Vetrap, 3M Italia Srl, Milano, Italy) applied around the neck (Figure 1). 21 The sensor has a 1-hour period of initialization. The detection limits of the sensor are between 20 and 500 mg/dL; when the IG concentration is ≤20 mg/dL and ≥500 mg/dL, the reader shows "LO" and "HI," respectively. The IG measurements were compared with BG concentrations obtained within 10 seconds by a PBGM (Optium Xceed, Abbott, UK), validated for use in dogs. 25 Venous or capillary BG concentrations were measured every 1-2 hours from admission to the resolution of DKA, and then less frequently, at the clinician's discretion according to the patient's condition, until discharge.
Body condition score (BCS) was recorded at admission using a previously described 9-point scoring system (World Small Animal Veterinary Association Global Nutrition Committee, BCS chart). The patient's metabolic status (pH and bicarbonate) and lactate concentration (marker of tissue perfusion) were assessed by blood gas analysis, performed using a blood gas analyzer (ABL 800 Flex, Radiometer Medical ApS, Brønshøj, Denmark), every 8-12 hours until DKA was resolved and then 24 hours later to confirm that the patient's metabolic balance had been maintained. The degree of ketosis was quantified every 4 hours by measuring blood BHB using the same PBGM (using ketone test strips), previously validated for dogs. 24 At the end of wearing period, which coincided with discharge, the sites of sensor application on all dogs were judged subjectively for the presence of erythema or other adverse events by the same clinician (C.C.).

| Statistical analysis
The influence of changes in metabolic variables (β-hydroxybutyrate, pH, bicarbonate, and lactate) and the effect of BCS and time wearing on sensor performance were evaluated to investigate whether specific patient variables influenced the accuracy of the device during the reso-

| Accuracy of the FGMS
Analytical and clinical accuracy during DKA and after its resolution was evaluated by comparing the results of the PBGM measurements and those obtained using the FGMS, using the ISO 15197:2013 criteria (BSI Standards Publication, in vitro diagnostic test system-Requirements for BG monitoring system for self-testing in managing diabetes mellitus; EN ISO 15197:2013).
Analytical accuracy was determined by calculating the mean absolute relative difference (MARD), median absolute relative difference (mARD), mean relative difference (MRD), and mean absolute difference (MAD). All these are measures of the average difference between sensor and reference results. Mean absolute relative difference and mARD measure the size but not the direction (higher or lower) of the differences compared with the reference (absolute) as a percentage of the reference value (relative). Mean absolute difference is similar, but just reports the size of the difference (it is not reported as a percentage), and is commonly used to assess accuracy at low BG concentrations (< 100 mg/dL).
Mean relative difference measures the size and direction of the difference compared with the reference as a percentage of the reference value. 26 Mean absolute relative difference traditionally has been used to assess the accuracy of CGMSs, representing it as a single numeric value. 27 Mean absolute relative difference or mARD should be <14%; a value >18% is considered to represent poor accuracy. 28 Second, analytical accuracy was estimated based on ISO 15197:2013 criteria, which state that at least 95% of results must be F I G U R E 1 FreeStyle Libre is composed of the reader (A) and the sensor (B), which is placed on the dorsal part of the neck of the dog (C), secured by an additional tape (D) and a bandage applied around the neck (E). The sensor has to be scanned by the reader, which instantaneously shows the interstitial glucose value (F). The reader shows "HI" and "LO" when the interstitial glucose concentration is ≥500 mg/dL and ≤20 mg/dL, respectively within ±15 mg/dL of the BG concentration for BG concentration <100 mg/dL and within ±15% of the BG concentration for BG concentration ≥100 mg/dL. Clinical accuracy was evaluated using Parkes consensus error grid analysis (EGA) for type 1 DM, which categorizes errors in BG measurement in terms of clinical risk. 29 In this analysis, a scatter plot is generated of the estimated BG concentrations (in our case, IG measurements obtained by the FGMS, y-axis) versus measured BG con-  F I G U R E 2 Parkes consensus error grid analysis (EGA) representation with the percentage of values within different zones before diabetic ketoacidosis (DKA) resolution (A) and after DKA resolution (B). The reference glucose values (blood glucose obtained by a portable glucometer), on the x-axis, are plotted against the interstitial glucose measurements obtained by the flash glucose monitoring system, on the y-axis. The different zones designate the magnitude of risk: no effect on clinical action (zone A), altered clinical action-little or no effect on the clinical outcome (zone B), altered clinical action-likely to affect the clinical outcome (zone C), altered clinical action-could have a significant medical risk (zone D), and altered clinical action-could have dangerous consequences (zone E   suggesting that in some patients the device was more accurate than in others.

| DISCUSSION
Our study is the first to evaluate the clinical accuracy and perfor- with DKA found only a weak association between hydration and the accuracy of the measurements, with the device being more accurate in F I G U R E 3 Inter-patient variability (D = dog). Each patient is represented on the x-axis with a box and whisker plot. The y-axis represents the relative difference defined as IG − BG. BG, blood glucose; IG, interstitial glucose more hydrated patients. 39 Results of this study suggest that this device is a clinically useful tool for monitoring IG concentration in critically ill patients, but it has a number of disadvantages, including the initial cost of the device, the cost of the sensor, and the need to obtain blood samples for calibrations every 8-12 hours. 39 The most important limitation is that glucose measurements are only available retrospectively, after downloading the data onto a personal computer, thereby limiting their clinical usefulness in the management of hospitalized patients. 39  Another limitation is that the skin and SC adipose tissue thickness at the site of application of the sensor were not evaluated, making it impossible to assess their influence on the accuracy of the FGMS. In human medicine, tissue glucose concentration nadirs in muscle have been reported to be delayed in time and lower in magnitude relative to glucose concentrations in adipose tissue and blood, especially during insulin-induced hypoglycemia. 46,47 Decreased thickness of the SC adipose tissue layer may result in closer sensor proximity to the underlying muscle tissue and, consequently, in inaccurate glucose concentration results.
Other limitations of our study include the use of a single sensor for each dog, such that the precision of the FGMS was not investigated. In veterinary medicine, 2 studies investigated the effect of sensor location on performance of the Guardian REAL-Time CGMS. 41,42 In cats, preliminary results suggest that dorsal neck placement may be superior to lateral chest wall and lateral knee fold placement. 41 In dogs, IG concentrations obtained by the CGMS at the lateral thorax site had the best correlation with BG concentrations compared to lateral neck, lumbar, and abdomen sites. 42 In our study, the sensor was placed at a single body site (the dorsal part of the neck, an area not particularly subject to traction and trauma, especially in animals in lateral recumbency), not allowing evaluation of the application site as a variable that might influence the accuracy of the device.
In conclusion, although the ISO 15197:2013 requirements were not fulfilled, the novel FGMS provides clinically accurate estimates of BG concentration compared with PBGM and represents a useful device to monitor BG concentration in critically ill hospitalized dogs with DKA. Acid-base status, BHB and lactate concentrations, BCS, and time wearing did not influence the accuracy of the sensor, making it suitable not only for stable diabetic dogs but also for dogs with DKA.

CONFLICT OF INTEREST DECLARATION
Authors declare no conflict of interest.

OFF-LABEL ANTIMICROBIAL DECLARATION
Authors declare no off-label use of antimicrobials.

INSTITUTIONAL ANIMAL CARE AND USE COMMITTEE (IACUC) OR OTHER APPROVAL DECLARATION
The Scientific Ethics Committee of the University of Bologna (Italy) approved this study.