Corresponding author: E. Zini, Clinic for Small Animal Internal Medicine, Vetsuisse Faculty, University of Zürich, Winterthurerstr 260, 8057 Zürich, Switzerland; e-mail: email@example.com.
Background: The Guardian REAL-Time is a continuous glucose-monitoring system (CGMS) recently developed to provide instantaneous interstitial glucose concentrations; the system does not require a monitor being fixed to the animal.
Hypothesis: The CGMS provides accurate and reproducible real-time readings of glucose concentration in cats.
Animals: Thirty-two diabetic cats, 2 cats with suspected insulinoma, and 5 healthy cats.
Methods: Prospective, observational study. CGMS accuracy was compared with a reference glucose meter at normal, high, and low blood glucose concentrations using error grid analysis. Reading variability of 2 simultaneously used CGMS was determined in diabetic cats by calculating correlation and percentage of concordance of paired data at different glycemic ranges. The time interval between increasing glycemia and a rise in interstitial fluid glucose measured by the CGMS was assessed in healthy cats receiving glucose IV; the time point of maximal increase in interstitial glucose concentrations was calculated.
Results: The CGMS was 100, 96.1, and 91.0% accurate at normal, high, and low blood glucose concentrations. Measurements deviated from reference by −12.7 ± 70.5 mg/dL at normal, −12.1 ± 141.5 mg/dL at high, and −1.9 ± 40.9 mg/dL at low glucose concentrations. Overall, paired CGMS readings correlated significantly (r= 0.95, P < .0001) and concordance was 95.7%. The median delay after IV administration of glucose to an increase in interstitial glucose was 11.4 minutes (range: 8.8–19.7 minutes).
Conclusions and Clinical Importance: Although some readings substantially deviated from reference values, the CGMS yields reproducible results, is clinically accurate in cats with hyperglycemia and euglycemia, and is slightly less accurate if blood glucose concentrations are low. Rapidly increasing interstitial glucose after a glycemic rise suggests that the CGMS is suitable for real-time measurement under clinical conditions.
Measurement of blood glucose concentrations and the generation of blood glucose curves are integral parts of monitoring diabetes in cats.1,2 Multiple blood samples during a defined period of time are required to obtain glucose curves that allow adjustment of medication. In general, samples are collected every 1–2 hours over 10–12 hours by venipuncture or capillary blood sampling in cats. Repeated handling or restraining the cats can artifactually elevate blood glucose concentrations because cats are particularly sensitive to stress-induced hyperglycemia. Consequently, the interpretation of the glucose curve may be difficult because of the potential confounding influence of stress. Another limitation is that even with numerous blood samplings the glucose nadir or peak may be missed because of the 1–2-hour intervals between measurements.3,4
To assess glucose concentrations more frequently, continuous glucose-monitoring systems (CGMS) have been developed for diabetic humans almost 2 decades ago; their use has recently also been described in other species, including the cat.5–7 The systems measure glucose concentrations in the subcutaneous interstitial fluid, which have been shown to correlate well with that obtained in whole blood.8,9 In humans, the average delay between a change of the glucose concentration in blood versus interstitial fluid was approximately 5 minutes, thus making the CGMS reliable for real-time monitoring. Because of the high performance, the CGMS is increasingly considered an essential device to improve treatment of human diabetes mellitus.8,10,11
A CGMS was used in 16 diabetic cats and there was a good correlation between glucose concentrations measured in the interstitial fluid and in the capillary or peripheral blood.5 In addition, the system unveiled an excessively low glucose nadir in insulin treated diabetic cats, thus allowing treatment to be modified. In 2 other studies, CGMS were used successfully to obtain glucose curves in diabetic cats.6,7 A drawback of all CGMS previously used in cats is that the monitoring device had to be fixed to the animal, and that the recorded data had to be downloaded manually on a computer in order to be analyzed.5–7
The Guardian REAL-Time CGMS belongs to a new generation of CGMS designed for diabetes monitoring. It provides interstitial glucose concentrations in real-time and enables on-screen recording of data over a 24-hour period. This allows diabetic humans to self-assess their glucose concentrations and to recognize potential episodes of hypoglycemia as early as possible. Further, the monitor can be kept separate from the body. These innovations are also expected to simplify the use of the instrument in cats and, in particular, to substantially decrease the stress for the generation of glucose curves. Recently, the novel CGMS was used in 1 diabetic cat and the device was well tolerated and promptly identified an episode of hypoglycemia caused by an insulin overdose.12
The purpose of the present study reported was to further investigate the performance of the Guardian REAL-Time CGMS for use in cats. The clinical and analytical accuracy were studied, with special attention paid to potential sources of error.
Materials and Methods
The study was performed with 39 cats, including 34 client-owned diseased cats, and 5 healthy experimental catsa (Veterinary Office of Zürich, Switzerland, permission nr. 94/2008). Among the client-owned cats, 32 were diagnosed with diabetes mellitus, 5 of which had diabetic ketoacidosis. Two of the client-owned cats had suspected insulinoma based on severe hypoglycemia that poorly responded to treatment and based on a pancreatic mass that was identified by abdominal ultrasonography. The median age of the 34 cats was 12 years (range: 2–17 years). Twenty-three (67.6%) were neutered males and 11 (32.4%) were spayed females. Twenty-eight cats (82.4%) were domestic short- or longhair, 2 were Norwegian forest cats, and 1 each was a Birman, Persian, Ragdoll, or Egyptian Mau cat. The median age of the 5 experimental cats was 8 months (range: 6–11 months), and all were intact male domestic shorthair. Young cats were used because they were available from another experiment. They were clinically healthy based on physical examination and blood work (data not shown).
The Guardian REAL-Timeb system for continuous glucose monitoring consists of a disposable sensor, a transmitter and a pager-sized monitor (Fig 1). The sensor is housed in flexible 1.5 cm tubing with a membrane-covered side window that allows the active electrode to interact with the interstitial fluid. The sensor is placed in the subcutaneous tissue by means of a 22 G needle. Glucose in the interstitial fluid undergoes an electrochemical reaction on the glucose oxidase-containing electrode that generates a small electric current. This will be subsequently converted to glucose concentration (mg/dL). The sensor is connected to a transmitter that transmits data over a maximal distance of 3 m to a pager-sized monitor; the monitor displays data in real-time for up to 24 hours. Data are collected every 10 seconds and a mean value computed every 5 minutes. The sensor can be used for up to 72 hours. The monitor has the capability to record interstitial glucose concentrations for up to 1 month before downloading on a computer for further analysis. Notably, the monitor shows glucose concentrations between 40 and 400 mg/dL; concentrations beyond this range are correctly recorded by the CGMS but need to be downloaded to be analyzed.
After placement of the sensor, the CGMS requires a 2-h initialization period. During this interval interstitial glucose concentrations are not provided. Calibration needs to be performed by measuring blood glucose concentrations with a rapid method such as a portable blood glucose meter. The recorded value is immediately transcribed in the monitor of the CGMS to initiate readings; if blood glucose is <40 or >400 mg/dL calibration needs to be postponed until the concentration reaches 40–400 mg/dL. The system is recalibrated after 6 hours and subsequently at least twice daily.
In the present study, the CGMS sensor was placed in the subcutaneous tissue of the thorax of cats, at the 6th or 7th intercostal space, and about half-way between the spine and sternum. In brief, a 3 × 3 cm skin area was clipped and disinfected. The sensor was injected subcutaneously and was fixed to skin with tape and protected with a soft bandage placed around the chest. For calibration, the capillary blood glucose concentration was measured from the inner pinna with a portable blood glucose meter (PBGM).13 Glucose concentrations were recorded with cats hospitalized in cages; the monitor was fixed to the cage door. The distance between transmitter and monitor was <1.5 m. All procedures were performed in awake cats.
The readings of the CGMS were compared with those of the PBGM AlphaTRAKc used as a reference. The AlphaTRAK was shown to be precise and accurate13; the glucose concentrations measured in the capillary blood did not differ from those measured in whole blood with the reference laboratory hexokinase method.13
Adverse Reactions Using the CGMS
During monitoring of glucose concentrations, attention was paid to possible abnormal behavior because of the bandage that was used to keep the sensor in place. We recorded whether the cat tried to remove the sensor, developed aggressiveness, or showed decreased appetite. When the bandage and sensor were removed, 1 of the investigators (S.M. or F.T.) recorded adverse skin reactions, especially at the site of sensor placement.
Accuracy of the CGMS
To compare glucose concentrations measured with CGMS to the PBGM, paired samples were taken in the hypoglycemic range (<90 mg/dL), in the euglycemic range (90–180 mg/dL), and in the hyperglycemic range (>180 mg/dL). From each cat, 10–15 paired glucose measurements were collected during the 1st 2–3 days of glucose monitoring. At the same time glucose concentrations were measured with the PBGM, and values shown by the monitor of the CGMS were noted. For interstitial glucose concentrations <40 or >400 mg/dL (ie, not shown by the monitor) the exact sampling time was noted and the values were later downloaded from the CGMS.
Variability of Paired CGMS Readings In Vivo
Two CGMS devices were simultaneously used for 24 hours in 5 diabetic cats. The CGMS sensors were placed symmetrically on the left and right side of the thorax. The 2 CGMS were calibrated every 10–12 hours with the same glucose measurement obtained with the PBGM.
Variability of Paired CGMS Readings In Vitro
To verify whether technical characteristics of the CGMS contributed to variability of paired readings, in vitro tests were performed using solutions mimicking the hypoglycemic, euglycemic, and hyperglycemic range. Glucosed was added to saline to yield solutions with glucose concentrations of approximately 50, 90, and 300 mg/dL. Two CGMS devices were simultaneously used for 12 hours in each solution. The solutions were put in a closed box to protect them from light exposure; they were kept at room temperature.
Time Delay of the CGMS
To investigate the time delay between a rapid change in blood glucose concentrations and the rise in interstitial fluid glucose recorded by CGMS, an IV bolus of glucose was administered to 5 healthy cats. The CGMS was placed 6 hours before IV administration of glucose. Glucose concentrations were recorded in the interstitial fluid with the CGMS, starting just before the bolus was given and for 2 hours thereafter. During the same period, glucose concentrations were also measured in the capillary blood with the PBGM (before and 5, 10, 15, 30, 45, 60, 90, and 120 minutes after the bolus).
The test was performed in fasted cats sedated with tiletamine/zolazepame and anesthetized with propofol.f Cats were acclimated to anesthesia for 1 hour. Thereafter the glucosed bolus of 1 g/kg was administered via the femoral vein. The time difference between the points of maximal rise of glucose concentration measured in the blood and in the interstitial fluid was calculated; this calculation had been used to quantify the delay to observe a change in interstitial glucose in humans.14
Accuracy between results of the CGMS and the PBGM was evaluated with statistical methods and using a clinically oriented approach. Using the method of residuals, the differences between results of the CGMS and the reference method were plotted.15 The clinical accuracy of CGMS readings was examined by use of the Clarke error grid analysis, as described in cats.16,17 The grid system assigns CGMS measured values (y-axis) versus actual blood glucose values (reference PBGM, x-axis) to 5 zones (A through E) and is based on the assumption that the clinical goal is to maintain blood glucose concentrations between 70 and 180 mg/dL. Measurements in zones A and B are clinically accurate in that they lead to clinically correct treatment decisions. The CGMS readings in zone A deviate from the reference value by no >20%. The CGMS readings in zone B represent benign errors and deviate from reference values by >20%; however, they would either not lead to a change in treatment, or treatment would not have any harmful effects. Values in zones C, D, and E would lead to relevant treatment errors or failure to initiate treatment. Values in zone C would lead to unnecessary correction or overcorrection of the acceptable glucose concentration and would cause the actual blood glucose concentration to fall below 70 mg/dL or to increase above 180 mg/dL. Zone D represents potentially dangerous errors of failing to detect and to treat actual blood glucose values that are outside the target range, because CGMS readings are within the target range. The CGMS readings in zone E are opposite to the actual blood glucose values, and therapeutic actions would be opposite to those indicated.
Because readings obtained with the CGMS may occasionally fluctuate for 5–15 minutes in cats, the accuracy was also calculated by including only measurement that were preceded by stable interstitial glucose concentrations over 30 minutes. Glucose readings were considered stable if their calculated coefficient of variation was below 10%.
To assess reading variability of 2 CGMS devices used simultaneously in the same cat, the Pearson correlation coefficient and the mean absolute difference between paired sensor glucose values were calculated. Concordance in the simultaneous measurements was determined by calculating the percentage of data that could both be classified in the normal, high, and low glucose range.18 However, because paired readings can be similar but could fall into 2 different glycemic ranges, thus providing discordant results but with negligible clinical relevance, we included an additional criteria to calculate concordance; paired samples had to show a difference of at least 10% to be classified as discordant. The same tests were used to assess variability in vitro.
To quantify the time interval between an increase of glucose concentrations in peripheral blood and interstitial fluid, the time difference was calculated for the maximal increase in the slope of the 2 glucose curves computed with 2nd derivatives.14
Significance was set at P < .05. Data were analyzed with a commercially available software.g
Practical Use of the CGMS
The placement of the sensor and transmitter, and the visualization of the data in real-time on the monitor, were successful and easy to perform in each cat. No adverse skin reactions were observed at the site of sensor placement and none of the cats tried to remove the sensor or showed abnormal behavior in relation to the bandage.
Accuracy of the CGMS
Four hundred and fourty-eight paired samples were taken with the CGMS and the PBGM. Based on the latter, 67 samples were in the hypoglycemic range, 176 in the euglycemic range, and 205 in the hyperglycemic range. The Bland and Altmann difference plots are shown in Figure 2; the mean ± 2 standard deviations (2 SD) difference from reference was −12.7 ± 70.5 mg/dL in the euglycemic range, −12.1 ± 141.5 mg/dL in the hyperglycemic range, and −1.9 ± 40.9 mg/dL in the hypoglycemic range. The percentage of underestimated glucose readings was higher than that of overestimated values at normal (69.9 versus 27.3%), high (54.6 versus 42.9%), and low (56.7 versus 32.8%) blood glucose concentrations; values identical to reference were recorded in 2.8, 2.5, and 10.5% of measurements, respectively. Interstitial glucose concentrations differed by >100 mg/dL from the reference PBGM in 1.7% of cases in the euglycemic range, in 10.2% of cases in the hyperglycemic range and in no case in the hypoglycemic range. In the same ranges, interstitial glucose differed by >50 mg/dL in 17.0, 34.1, and in 1.5% of cases, respectively. In the hypoglycemic range, a deviation from the reference PBGM between 25 and 50 mg/dL was recorded in 22.4% of cases.
Results of the Clarke error grid analysis are reported in Figure 3. In the euglycemic range, the CGMS provided measurements that were in the clinically acceptable zones A and B in 63.1 and 36.9%, respectively, of cases. No reading was in zone C, D, or E. In the hyperglycemic range 74.6% of measurements were in zone A, 21.5% in zone B, and 3.9% in zone D. Measurements in zones C or E were not recorded. In the hypoglycemic range 64.2% of measurements were in zone A, 26.9% in zone B, and 9.0% in zone D. Readings in zones C or E were not recorded.
Similar results were obtained by including only CGMS readings that were preceded by a 30-minute period of stable interstitial glucose concentrations (388 paired samples). In the euglycemic range, mean ± 2 SD difference from reference was −6.7 ± 71.9, −1.9 ± 131.3 mg/dL in the hyperglycemic range, and −2.3 ± 38.8 mg/dL in the hypoglycemic range. Based on error grid analysis, the CGMS provided measurements in the euglycemic range that were in zones A and B in 64.7 and 35.3% of cases, respectively. No reading was in zones C to E. In the hyperglycemic range, 76.4% of measurements were in zone A, 21.6% in zone B, and 2.0% in zone D. Measurements in zones C or E were not recorded. In the hypoglycemic range 63.5% of measurements were in zone A, 26.9% in zone B, and 9.6% in zone D. Readings in zones C or E were not recorded.
Variability of Paired CGMS Readings
The 1,445 paired sensor values correlated significantly (r= 0.95, P < .0001, Fig. 4). The correlation coefficient was lower for readings in the hypoglycemic range (r= 0.37) than in the euglycemic (r= 0.64) or hyperglycemic (r= 0.76) range (Table 1).
Table 1. Correlation coefficients, 95% confidence interval (CI), and significance level of glucose concentrations measured simultaneously by 2 continuous glucose-monitoring systems on the left and right side of the thorax; values were subdivided according to glycemic range.
Pearson's Correlation Coefficient (r)
P < .0001
P < .0001
P < .0001
P < .0001
The mean ± 2 SD difference in interstitial glucose concentrations of all paired measurements was 14.3 ± 90.0 mg/dL; the absolute difference between paired readings was lower in the low (6.1 ± 25.6 mg/dL) than in the normal (9.0 ± 42.5 mg/dL) or high (41.9 ± 102.4 mg/dL) glycemic range. When the interstitial glucose values measured with 1 CGMS were classified as being in the euglycemic, hyperglycemic, or hypoglycemic range, 95.7% of all pairs were concordant (Table 2). Of the remaining nonconcordant pairs, 1 interstitial glucose value was classified as in the euglycemic range and the corresponding in the hyperglycemic range (2.8%), or 1 interstitial glucose value was classified as in the euglycemic range and the other in the hypoglycemic range (1.5%). Pairs with 1 value in the hyperglycemic range and the other in the hypoglycemic range were not observed. When analyzed separately for the respective glycemic ranges, readings were 93.9, 96.1, and 100% concordant at normal, high, and low glucose concentrations, respectively.
Table 2. Concordance of glucose values simultaneously measured with the continuous glucose-monitoring system on the left and right side of the thorax according to glycemic range.
Nonconcordant Values (%)
In vitro, paired sensor values correlated significantly (r= 0.99, P < .0001); the correlation coefficient was slightly lower for paired readings in the low glucose range (r= 0.63) than in the euglycemic (r= 0.73) or the hyperglycemic (r= 0.99) range. The mean ± 2 SD difference of paired readings was 11.6 ± 14.9 mg/dL at normal, 8.0 ± 8.4 mg/dL at high and 9.0 ± 16.6 mg/dL at low glucose concentrations. Concordance was 100% in each case.
Time Delay of the CGMS
In the capillary blood of healthy cats, the maximal glucose concentration was always recorded immediately after the IV injection of glucose bolus; the glucose concentration progressively decreased thereafter. In the interstitial fluid, the measured glucose concentrations gradually increased and reached the maximal concentration after 35.0 minutes; after that, the interstitial glucose concentration gradually decreased. The delay from the time of the IV injection of glucose bolus to the maximal rising slope of the glucose curve in the interstitial fluid was 11.4 minutes (range: 8.8–19.7 minutes).
The Guardian REAL-Time CGMS was successfully used in cats. No adverse skin reactions were observed at the site of sensor placement and all cats tolerated the bandage. The latter was likely because of the fact that the monitor of the new CGMS can be kept separate from the cat, thus substantially reducing discomfort. Different from the previous generation of CGMS,5–7 the new device allows to visualize the interstitial glucose concentrations in real-time and to scroll the recorded values over the last 24 hours. Real-time monitoring performed with the new CGMS may thus be particularly useful with uncooperative cats or whenever the measurement of the blood glucose concentration with PBGMs is impractical. A drawback of the Guardian REAL-Time system is that the system only calibrates if the blood glucose concentrations are between 40 and 400 mg/dL; otherwise the monitoring cannot be started. For this reason, the calibration had to be postponed by 1–3 hours in 6 diabetic cats in the present study to allow insulin therapy to decrease blood glucose below 400 mg/dL.
The results of the error grid analysis showed that the clinical accuracy of the CGMS was satisfactory in the euglycemic and hyperglycemic range, and only slightly less in the hypoglycemic range. No reading was assigned to zone C (ie, leading to correction of acceptable blood glucose concentration) or E (ie, leading to opposite therapeutic actions to those indicated) at any blood glucose concentration. However, 9.0 and 3.9% of readings in the hypoglycemic and hyperglycemic ranges, respectively, were in zone D. Values in zone D can lead to potential dangerous errors by failing to detect or to treat blood glucose concentrations that are outside the reference range. By including only CGMS readings preceded by a 30-minute period of stable interstitial glucose concentrations, the results of the error grid analysis were comparable; this suggests that the clinical accuracy of the CGMS is not affected by short-lasting fluctuations of the interstitial glucose concentration.
To better assess interstitial glucose curves with rapid glucose fluctuations, the recently introduced continuous glucose-error grid analysis has been proposed.19 This method takes into consideration the effect of glucose trends of the interstitial glucose curve, in addition to single readings; it has been developed to improve estimation of clinical accuracy of CGMS in humans. In the present study, we opted not to use the continuous glucose-error grid analysis to avoid excessive stress or risk of anemia, because peripheral or capillary blood samples need to be collected every 15 minutes.
Although the mean differences of interstitial glucose concentrations measured with the CGMS versus reference values were minimal at any blood glucose concentration, the spread of results as shown with Bland and Altmann plots was wide, in particular in the hyperglycemic range; based on 2 SD the difference was up to ± 141.5 mg/dL. However, despite this large variation, clinical accuracy assessed with the error grid analysis remained good because 95.6% of readings were in zone A or B. In contrast, in the hypoglycemic range, the difference between the 2 methods based on 2 SD was up to ± 40.9 mg/dL, and a deviation from reference between 25 and 50 mg/dL was observed in 22.4% of readings. When blood glucose concentrations are low, differences of this magnitude may be important and lead to treatment errors, as indicated by the relatively high percentage of readings falling in zone D (9.0%) in the error grid analysis. Therefore, if CGMS measurements are at the lower end of the normal glucose range or below, it is advisable to periodically assess blood glucose concentrations with another method. In the euglycemic range, difference from reference based on 2 SD was ± 71.9 mg/dL; however, despite the large variation, all readings would have led to clinically correct treatment decisions according to error grid analysis.
The overall variability of simultaneous interstitial glucose readings obtained with 2 CGMS was good, with both a high correlation coefficient (r= 0.95; P < .0001) and a high percentage of concordance (95.7%). Of note, the correlation coefficient in the hypoglycemic range was low (r= 0.37), which may suggest inferior performance of the device at low glucose concentrations. However, the mean and absolute differences between paired readings were relatively small (6.1 ± 25.6 mg/dL) and the concordance was 100%. Thus, we consider the low correlation coefficient of paired readings in the hypoglycemic range of little clinical relevance. In the euglycemic and hyperglycemic range, the variability of paired CGMS readings was very good. The large difference observed at high blood glucose concentrations, as assessed with 2 SD (± 102.4 mg/dL), does not seem to have a major influence on treatment decision because only 3.9% of measurements were recorded as hyperglycemic while actually being in the normal range. Accordingly, readings obtained with different CGMS can be considered comparable at all blood glucose concentrations, similar to what has been described in humans.18 Results achieved in vitro were better than in vivo for both correlation analysis and mean differences. At decreasing glucose concentrations the correlation coefficient was slightly lower than at normal or high glucose concentrations, confirming the slight inferior performance of the device with lowering glucose concentration.
Based on the IV administration of a glucose bolus in healthy cats, the CGMS recorded a maximal rise in interstitial glucose with a median delay of 11.4 minutes, which is close to the delay of 10 minutes previously observed in dogs.20 This suggests that changes in blood glucose concentrations are rapidly detected in the interstitial fluid, thus making the device reliable for real-time monitoring under clinical conditions. Of note, we used healthy young cats in our study. In diabetic cats, which typically are older and often slightly dehydrated, an equilibrium of the glucose concentration between blood and interstitial fluid may possibly be reached slightly later. However, to test this hypothesis glucose would need to be injected in diabetic cats, leading to an increase of plasma osmolarity to a dangerous level. In addition, it is important to note that the glucose bolus was administered under anesthesia. Peripheral capillary pressure may have been decreased, thus reducing the glucose shift from the vascular to the interstitial compartment.
In summary, the novel CGMS provides clinically accurate and reproducible measurements in the euglycemic and hyperglycemic range; the clinical accuracy was slightly less in the hypoglycemic range. In the latter case, or when interstitial glucose readings are at the lower end of the normal range, it may be advisable to assess blood glucose concentration with a reference method to confirm the CGMS results. The short interval between an increase of glycemia and a rise in interstitial glucose makes the CGMS useful for real-time monitoring in cats.