Evaluation of a continuous glucose monitoring system in cats with diabetes mellitus

Abstract

A continuous glucose monitoring system (CGMS) was evaluated in 14 cats with naturally occurring diabetes mellitus. The device measures interstitial fluid glucose continuously, by means of a sensor placed in the subcutaneous tissue. All cats tolerated the device well and a trace was obtained on 15/16 occasions. There was good correlation between the CGMS values and blood glucose concentration measured using a glucometer (r=0.932, P<0.01). Limitations to the use of the CGMS are its working glucose range of 2.2–22.2 mmol/l (40–400 mg/dl) and the need for calibration with a blood glucose measurement at least every 12 h. When compared to a traditional blood glucose curve, the CGMS is minimally invasive, reduces the number of venepunctures necessary to assess the kinetics of insulin therapy in a patient and provides a truly continuous glucose curve.

Continuous glucose monitoring systems (CGMS) are used in human diabetic patients to monitor interstitial fluid glucose (Bode et al 1999). This enables accurate adjustment of insulin dosing and dietary regimes in order to improve glycaemic control (Bode et al 1999). In feline diabetic patients stability is usually assessed using a combination of clinical signs (Goossens et al 1998, Briggs et al 2000, Nelson 2000), blood and urine glucose analysis (Goossens et al 1998, Nelson 2000, Schaer 2001), serum fructosamine measurement (Crenshaw et al 1996, Elliot et al 1999, Nelson 2000) and blood glucose curves (Miller 1995, Nelson 2000). Fructosamine measurement has become popular, as it only requires a single blood sample. There are limitations, however, as fructosamine may be lower in cats with reduced serum or plasma total protein (Reusch and Haberer 2001) and in those with hyperthyroidism (Reusch and Tomsa 1999). Also, it is occasionally elevated in cats with stress hyperglycaemia (Crenshaw et al 1996). In the general practice setting, it is often difficult to obtain blood samples overnight, therefore generation of 24-h glucose curves is not always possible. In addition stress-induced hyperglycaemia, which is more common in cats than dogs or humans (Rand 1999), complicates interpretation of blood glucose concentrations (Crenshaw et al 1996). Recently, lancet techniques have been adapted for use in cats allowing owners to measure their pet's blood glucose at home (Wess and Reusch 2000b, Casella et al 2002), however, this is reliant on an owner's willingness and availability to perform the technique. Even with frequent sampling, the glucose nadir or peak could fall between two sample times. Therefore, a minimally invasive method of monitoring glucose concentrations more frequently could improve the management of diabetic cats. Reported median survival times for cats with naturally occurring diabetes mellitus range from 17 to 29 months (Kraus et al 1997, Goossens et al 1998) and Kraus et al (1997) reported that one-third of the cats in their study died from causes related to their diabetes. Therefore, a method that would improve glycaemic control could potentially lead to improved survival.

The Medtronic MiniMed CGMS (Fig 1) consists of a disposable sensor, which is placed subcutaneously, that is connected via a cable to a pager-sized monitor that stores the data from the sensor. The sensor consists of a small flexible electrode containing glucose oxidase; glucose in the interstitial fluid undergoes an electrochemical reaction on the electrode causing a current to be produced. This current is sampled by the monitor every 10 s and a mean value recorded every 5 min. The sensor can be left in place for up to 72 h. The data from the monitor are downloaded on to a computer for analysis at regular intervals using a dedicated docking station, at this point the electrical signal is converted to a glucose concentration. The computer software is based on a Windows Excel spreadsheet. Gross et al (2000) showed that when human diabetic patients used both the CGMS and a glucometer there was good agreement between the results. Davison et al (2003) evaluated the MiniMed CGMS in diabetic dogs and found good agreement when compared to a conventional blood glucose curve. There has been one report of the successful use of the MiniMed CGMS in five cats, only two of these were diabetic (Wiedmeyer et al 2003).

Fig 1.

MiniMed continuous glucose monitoring system.

The aim of this study was to evaluate the MiniMed CGMS in diabetic cats. Our hypothesis was that it would produce continuous glucose data, with the need for fewer blood glucose samples than a conventional glucose curve and therefore reduced stress for the patient.

Materials and methods

Fourteen diabetic cats were recruited for the study and a second evaluation was performed in two cats, therefore, 16 CGMS traces were performed. All cats were at a point in their therapy that necessitated a glucose curve for assessment of their diabetic stability. Patients were only recruited if they were sufficiently stable to have blood glucose concentrations that were likely to remain within the working range of the CGMS for most of the day, ie, between 2.2 and 22.2 mmol/l (40–400 mg/dl). The patients were seen at the Queen's Veterinary School Hospital, University of Cambridge (n=7), Wey Referrals, Woking, Surrey (n=4), the Royal Veterinary College, University of London (n=4) and Phoenix Veterinary Group, Sandhurst, Berkshire (n=1). The CGMS was performed in addition to routine management and the original clinician retained primary care for the case and made any therapeutic decisions.

Cats were admitted to the hospital and allowed a few hours to acclimatise to their surroundings. Prior to attaching the CGMS, blood glucose concentration was measured using the Glucometer Esprit 2 (Bayer) to ensure that it was within the range of the sensor. This glucometer has been evaluated for use in cats (Wess and Reusch 2000a). A small patch of hair was clipped on the side of the thorax, caudal to the olecranon (Fig 2a) and the sensor placed subcutaneously using the insertion device (Fig 2b). The sensor was attached to the monitor using the cable (Fig 2c) and taped in place (Fig 2d). The monitor was then attached to a harness (Fig 2e). The device was left in situ for up to 72 h and the data downloaded on to a laptop computer at least once daily (Fig 2f). During this time, blood glucose concentrations were measured at 4-h intervals in most cases, although the exact frequency was dependent on the cat's tolerance of blood sampling. The first blood sample was always taken 1 h after sensor placement as the sensor takes 1 h to ‘initialise’ or develop a stable current, it then requires the input of a blood glucose concentration for calibration. Further blood glucose concentrations were entered for calibration three times daily. The manufacturers recommend that a minimum of three blood glucose concentrations are entered in any 24-h period and that a gap of no longer than 12 h occurs between entries. The remaining three values were not used for calibration, but used to calculate the correlation between blood and interstitial fluid glucose concentrations. These additional samples were also taken to ensure there would be some useful information for the patient even if the CGMS did not generate a trace. In order to verify that the glucometer used was accurate, selected blood samples (n=15) were tested concurrently on the glucometer and at the clinical pathology laboratory, Queen's Veterinary School Hospital, using the Beckman Synchron CX5-CE wet chemistry analyser. Serum fructosamine was assayed by Diagnostic Services, Royal Veterinary College, using the Synermed IR200 machine with ABX Diagnostics Fructosamine 300/600 reagents (reference range 219–375 μmol/l).

Fig 2.

(a) Shaved patch illustrating the ideal position for sensor placement, caudal to the elbow, where normal locomotion will not dislodge the sensor. (b) Placement of the sensor using the needle introducer and insertion device. The sensor can also be manually ‘injected’ into the subcutaneous tissue. (c) The sensor is kept in place with its adhesive patch and attached to the cable that transmits currents to the monitor. (d) The sensor taped in place with a full circle of elastoplast. In this photograph, the sensor is being initialised via the monitor. (e) The MiniMed CGMS in position on a patient. The redundant cable is wrapped loosely around the monitor and a small section of cable left exposed to allow freedom of movement. (f) Patient wearing the CGMS whilst the information is downloaded via the docking station on to the computer.

Any changes to the management of a case were based on a combination of history, clinical signs, blood glucose concentrations and serum fructosamine. The clinician in charge of the case made the alteration and, therefore, there was some variability in the exact criteria used to make therapeutic decisions. In general, a nadir blood glucose of <3.0 mmol/l (54.0 mg/dl) or serum fructosamine of <300 μmol/l was used as an indication to decrease the insulin dose. If the nadir blood glucose was >12.0 mmol/l (216 mg/dl) or fructosamine was >500 μmol/l this was considered a reason to increase the dose or frequency of insulin administration.

Statistical analysis

The Spearman Rank correlation coefficient was calculated, using SPSS v10.0 for Windows, to compare interstitial fluid glucose concentrations to blood glucose concentrations (values outside the working range of the CGMS were excluded). This correlation was performed for all blood glucose samples taken and also using only those glucometer values that were not entered into the CGMS for calibration. This was also used to compare blood glucose concentrations measured using the glucometer versus the diagnostic laboratory.

The CGMS software reports intra-subject correlations between the CGMS and glucometer values used to calibrate the monitor and also mean absolute difference between the two. The intra-use mean absolute difference and median correlation were calculated for all cats in the study, to allow comparison of the results with a human study.

Results

The device was well tolerated by all 14 cats (Fig 3) and a trace obtained on 15/16 occasions. There was no evidence of irritation at the site of sensor placement in any cat on any occasion. Table 1 summarises the data. A total of 605 h of CGMS data were recorded. Figure 4 shows a typical trace where there was good correlation between interstitial fluid glucose and blood glucose. There was a high degree of correlation between blood and interstitial fluid glucose concentrations (Fig 5a, r=0.932, P<0.01, n=102) and between glucometer and laboratory glucose (r=0.979, P<0.01, n=15).

Fig 3.

A cat settled into his kennel wearing the CGMS, he is displaying normal kneading behaviour, and does not appear distressed.

Table 1.

Data obtained for all patients in the study

Case no	Reason for entering study	Current treatment	Time CGMS attached (h)	Blood glucose range (mmol/l)	Sensor glucose range (mmol/l)	Serum fructosamine (reference range 219–375 μmol/l)	Action taken
1a	Check stability	Porcine lente 4.5 iu bid	11.5	3.5–13.5	3.4–17.4	247	No change
1b	Check stability second trace	Porcine lente 4.5 iu bid	24	2.8–15.8	2.2–16.2	165	↓ Dose
2	Assess glipizide therapy	2.5 mg bid glipizide	26	8.9–15.3	4.9–12.3	500	No change
3	Assess nadir and duration	Porcine lente 3 iu bid	24	8.5–29.8	Too high	–	↑ Dose
4a	Check stability	Bovine lente 2 iu bid	28	1.6–11.1	2.2–13.4	380	No change
4b	Check post-DM resolution second trace	Bovine PZI 2 iu sid	53	3.1–12.8	2.2–13.2	275	No change
5	Hypoglycaemia	Bovine PZI 1 iu bid	24	11.0–13.4	5.2–15.9	324	No change
6	Assess nadir and duration	Bovine lente 4.5 iu sid	47	16.7–32.2	12.8–22.2	536	↑ Frequency bid
7	Hypoglycaemia	Bovine PZI 2 iu sid	48	0.6–4.8	2.2–4.5	194	↓ Dose
8	Check stability	Bovine lente 4 iu bid	29	3.1–21.2	2.6–22.2	356	No change
9	Check stability	Bovine PZI 3 iu sid	18	2.2–2.2	2.2–2.7	298	↓ Dose
10	Unstable, anorexic	Bovine PZI 3 iu bid	55	13.2–18.2	4.3–20.6	472	No change
11	Assess stability	Bovine PZI sid with recent changes	54	2.8–21.9	2.2–22.2	683	Change back to PZI sid
12	Unstable	Porcine lente bid	46	0.8–32.8	2.2–22.2	435	Change to PZI bid
13	Hyperglycaemia	Bovine lente 3 iu bid	70	0.7–19.9	2.2–22.2	513	No change
14	Unstable	Bovine lente 3 iu bid	48	4.7–Hi	4.6–22.2	405	No change

CGMS=continuous glucose monitoring system.

PZI=protamine zinc insulin.

DM=diabetes mellitus.

N/A=not applicable.

Fig 4.

MiniMed CGMS trace (—) from cat 1b illustrating fluctuations in the interstitial fluid glucose concentration. Blood glucose concentrations are shown superimposed ().

Fig 5.

(a) Correlation between blood glucose concentrations measured on the glucometer and the interstitial fluid glucose measured on the MiniMed CGMS. (b) Bland–Altmann Plot to show the difference in interstitial fluid versus blood glucose concentrations, plotted against the mean of the two values.

The difference in glucose concentration measured using either the CGMS or glucometer was plotted against the mean of these two measurements (Fig 5b). The intra-use mean absolute error (%) was 9.66+/−6.8, n=15 cases and the median correlation for the glucometer values entered into the CGMS and the interstitial fluid glucose was r=0.885, n=9 cases; however, this excludes any patient with a very stable blood glucose where the variation over 24 h was <5.6 mmol/l (100.9 mg/dl) as the software did not report a correlation in this situation.

Several problems were encountered. The first patient removed the sensor after 12 h; this problem was circumvented in subsequent cases by covering the sensor with elastoplast in a complete circle around the cat's body (Fig 2d). Two cats kinked the sensors during recording and required placement of a second sensor. Although this resulted in a short gap in recording while the new sensor initialised, it did not affect the majority of data obtained.

The working range of the sensor can be a limiting factor to the use of the device and affected the data in 7/16 traces. Cats 7 and 9 had prolonged periods of hypoglycaemia (<2.2 mmol/l or 40 mg/dl) (illustrated in Fig 6) and cats 6, 11 and 14 had prolonged hyperglycaemia (>22.2 mmol/l or 400 mg/dl) all of which were outside the range of the device. Case 12 experienced both hypo- and hyperglycaemia. Case 3 had a blood glucose of 15.7 mmol/l (283 ng/dl) when the sensor was inserted and appeared externally calm, however, after placement her blood glucose increased and remained above 22.2 mmol/l for 6 h, by this time it was not possible to calibrate the sensor and a trace could not be generated.

Fig 6.

MiniMed CGMS trace from cat 7 illustrating the lower limit of the sensor's working range. Meter value refers to a blood glucose concentration measured on the glucometer and not entered into the CGMS for calibration. Paired meter value refers to blood glucose concentrations measured on the glucometer and entered into the CGMS for calibration.

In cats 4b, 11 and 12 which wore the CGMS for more than 24 h, it was apparent that despite identical daily routines there was marked variation in the CGMS curves for consecutive days (Fig 7).

Fig 7.

MiniMed CGMS trace from cat 4b illustrating the variation in curves recorded from the same cat on sequential days, whilst maintaining an identical therapeutic regime.

Discussion

The device was well tolerated in all cats; they behaved normally; eating, drinking and moving around freely in their kennels without apparent discomfort. The sensor position caudal to the elbow was selected to avoid patient interference during normal movement and although two cats kinked their sensors, there does not seem to be a more suitable site on such small patients. In humans, occasional subcutaneous reactions have been seen (MiniMed Trainer Manual). This was not a problem in any of the cats in the current study, including the two that had repeat tests, but larger numbers are required to determine whether this is a problem. The cat that developed hyperglycaemia within 1 h of sensor placement (cat 3) might have been stressed by the device and harness, alternatively hyperglycaemia may have occurred anyway due to hospitalisation or handling and could have affected a conventional blood glucose curve. It is often the case that cats with stress hyperglycaemia do not appear overtly stressed (Sparkes 1999, Nelson 2002).

There are several individual cats listed in Table 1 whose management warrants discussion. Case 1a had low normal serum fructosamine but reasonable blood and interstitial fluid glucose concentrations so no change was made but close monitoring continued; when the nadirs and fructosamine fell further (case 1b) the dose was decreased. Cat 4 had transient diabetes which resolved and recurred between traces a and b and for convenience once daily insulin was instituted when the disease recurred. In case 4b, the CGMS detected a low nadir; yet fructosamine was within the normal range, many clinicians would decrease the dose of insulin with serum fructosamine <300 μmol/l but in this case the cat was clinically well and owned by a veterinary nurse, so close monitoring was possible. In case 8 the low nadirs detected by the CGMS would suggest that a decrease in insulin dose was indicated, as fructosamine was normal and high blood and interstitial fluid glucose levels were also recorded the dose was not changed. In case 11 low blood and interstitial fluid glucose measurements were detected yet fructosamine was very high, this is because the cat had been completely off insulin for 2 weeks before referral. In case 12 although there was one low blood glucose of 0.8 mmol/l three further nadirs were recorded at 3.8, 6.0 and 18.8 mmol/l and fructosamine was acceptable, the final decision to change the type of insulin was due to clinician preference. In case 13 no change was made to the management despite very low blood and interstitial fluid glucoses, this was because an insulin infusion was given during recording and caused the low levels.

As direct measurements of interstitial fluid glucose are not readily available (Rebrin and Steil 2000), it is difficult to investigate the accuracy of the MiniMed CGMS. Therefore, most studies have focused on comparisons between interstitial sensor glucose values and blood glucose concentrations. There were initial concerns in human medicine that subcutaneous interstitial fluid glucose might not accurately reflect the blood glucose (Rebrin and Steil 2000) but most studies have allayed these concerns showing good agreement between the two (Rebrin et al 1999, Rebrin and Steil 2000, Monsod et al 2002, Caplin et al 2003). In 1999, Rebrin et al investigated the use of interstitial fluid glucose versus blood glucose in dogs during both experimental hypo- and hyperglycaemia and found sensor glucose accurately predicted plasma glucose (r>0.9) and that a sensor delay of 5–10 min was present with and without changes in insulin.

In the current study there was good correlation between the interstitial fluid glucose values measured by the MiniMed CGMS and blood glucose concentrations measured on the glucometer (r=0.932). As blood glucose values must be entered into the CGMS device for calibration purposes, these will influence the interstitial fluid glucose concentration reported by the CGMS analysis software. Excluding blood glucose values that had been used for calibration, there was still a good correlation between interstitial fluid glucose and blood glucose values (r=0.862; n=34). Both our values compare favourably to similar studies using the MiniMed CGMS in human diabetic patients (r=0.92; Bode et al 1999) and diabetic dogs (r=0.815; Davison et al 2003). In a previous study using the MiniMed CGMS in five cats, the correlation between interstitial fluid glucose and blood glucose concentrations (r=0.974) was better than in the current study, (Wiedmeyer et al 2003). However, in that study the correlation was calculated from only 16 blood samples that had been taken for calibration purposes. Additionally, only two of the five cats were diabetic and, therefore, the fluctuations in blood glucose concentration in the non-diabetic animals would not have been as marked as that seen in the current cohort of diabetic cats, which may have influenced the results. One further comparison with previous work is the intra-use absolute error (%), or the difference between the blood and interstitial fluid glucose concentrations, for the group, which was 9.66±6.8, this compares favourably with the Bode et al (1999) paper in humans where the value was 19.1±9.0.

In canine diabetic patients, it was reported that discrepancies occurred between interstitial fluid and blood glucose values after feeding, suggesting that diabetic dogs demonstrate a post-prandial hyperglycaemia which is not reflected in the interstitial fluid (Davison et al 2003). In the current study of feline diabetics, no such discrepancy was seen, although there were relatively few blood glucose measurements taken during the post-prandial period. These results are consistent with those from a previous study in which no post-prandial hyperglycaemia was seen in diabetic cats (Martin and Rand 1999).

The range of the CGMS is a limitation to its use, although the device can detect hypo- or hyperglycaemia, it cannot determine the absolute value when it is below 2.2 mmol/l (40 ng/dl) or above 22.2 mmol/l (400 ng/dl). This is most problematic early in stabilisation when patients are still hyperglycaemic or if stress hyperglycaemia develops. This restricts use of the CGMS for the investigation of insulin resistance, which would otherwise be an obvious application.

It is necessary to ensure that a gap of no longer than 12 h occurs between calibration entries. This poses a practical problem for use of the monitor overnight in fractious cats that require two people to perform the blood sampling. The antithesis to this is that cats can be left 12 h undisturbed between samples whilst obtaining a glucose curve, which should ensure as natural a trace as possible and would be advantageous in a general practice setting.

The cost of the equipment is likely to be prohibitive for general practice use. There is the initial outlay for the monitor, downloading station and software which is currently around £2390 ($4270). The sensors have a limited shelf life (maximum of 6 months), and can only be purchased in packs of four at a cost of £276 ($493).

In the current study cats were hospitalised and kept confined in a kennel to reduce the possibility of the sensor dislodging or the wire between the sensor and monitor becoming loose and entangled. Wiedmeyer et al (2003) reported no problems when cats were sent home overnight whilst wearing the monitor. This is an appealing option that should minimise stress hyperglycaemia and enable evaluation of cats in their normal environment. However, blood glucose concentrations must be entered at least every 12 h, which currently limits the time spent at home and travel to the surgery to calibrate the CGMS could lead to stress hyperglycaemia. Human patients using the CGMS enter blood glucose concentrations themselves and it might be possible to teach some owners of diabetic cats to do this using the ear prick technique (Wess and Reusch 2000b, Casella et al 2002).

This study has highlighted several advantages of the CGMS. It did, as hoped, generate a continuous curve. In case 1b, illustrated in Fig 4, intermittent blood glucose measurements would have missed the second nadir but this was detected by the CGMS showing that the cat was at risk of hypoglycaemia, fructosamine was also indicative of hypoglycaemia in this cat, interestingly there were no clinical signs. Also case 9 appeared well controlled clinically but the CGMS profile showed that this cat was hypoglycaemic for most of the day. This is similar to the situation in humans where the CGMS has identified previously undetected hypoglycaemia (Cheyne and Kerr 2002). This also supports work by Whitley et al (1997) that reported that cats might have documented hypoglycaemia without overt clinical signs. All the curves produced in this study were achieved with fewer blood samples than a standard glucose curve, which requires at least a 2-h venepuncture. Therefore, the cats were handled and restrained less and could be left undisturbed for longer.

The sensor can record for up to 72 h, enabling recordings to be made on sequential days. This should allow the most accurate clinical decision-making. In dogs, glucose curves can vary from one day to the next (Fleeman and Rand 2003); the current study has shown that this is also true in cats (Fig 7), suggesting that clinical judgements should not be made based exclusively on a single day's trace. In addition, in one human study, patients were asked to wear two sensors simultaneously and differences of >50% were found in 7% of measurements (Metzger et al 2002). Therefore, therapeutic changes should be based on trends and not single sensor values.

Overall this study has shown that the MiniMed CGMS can be used successfully in cats. The patients tolerate the device well and behave normally. The data generated correlate well with conventional blood glucose concentrations measured on a glucometer. Whilst use of the CGMS involves the cat wearing the monitor, it provides more information, ie, a continuous glucose curve and reduces the need for repeated venepuncture.

Based on this study, we are able to make several recommendations for veterinary surgeons wishing to use the device in their feline diabetic patients. Only cats that are likely to have blood glucose within the range of the sensor, ie, 2.2–22.2 mmol/l (40–400 ng/dl) should be selected. Cats should be allowed a few hours to acclimatise to their surroundings before sensor insertion to minimise stress hyperglycaemia. A minimum of three blood glucose values per 24 h (ie, each calendar day) should be entered into the device for optimal accuracy and the interval between calibrations must not exceed 12 h. The device should be left in place for at least two interinsulin injection periods before the data are used for clinical decision-making and results should always be interpreted in the light of clinical signs.