Biosensors & Bioektmnics
6
(1991) 3 l-36
In-vivo Behaviour of Hypodermically Implanted Microfabricated Glucose Sensors M. Koudelka”, F. Rohner-Jeanrenaud”, J. Terrettaz”, E. Bobbioni-Harschb, N. F. de Rooij” & B. Jeanrenaud” “Institute of Microtechnology. University of Neuchdtel, 2 rue A. Breguet. 2000 Neuchltel. Switzerland Qboratories of Metabolic Research. University and Faculty of Medicine, Geneva. Switzerland (Received 18 December
1989: revised version received 24 April 1990; accepted
1 May 1990)
The in-vivo behaviour of microfabricated GOD (glucose oxidase)/ HzOz glucose sensor implanted subcutaneously in normal anaesthetized rats has been studied. The sensor consists of a planar, three-electrode microcell, an enzyme membrane (glucose oxidase and bovine serum albumin cross-linked with glutaraldehyde) and an outer diffusion limiting polyurethane membrane. The sensor behaviour during hyperglycaemic (13.8 mM and Il.2 mM). euglycaemic (7.8 mM) and hypoglycaemic (3.5 mM) plateau levels was determined. The values of the in-vivo sensitivity (0.64 + O-05 nA/mM) and background current (I.25 + 0.4 nA) were determined using a two-point calibration method and then used to calculate apparent subcutaneous glucose concentrations. The results show the presence of a good correlation between all the plasma glucose levels (G) and the apparent subcutaneous tissue concentrations (G’), with G’ = 0997 *G - 0066. r = 0.9782. Abstrack
Keywords: glucose sensor, amperometric glucose, in-vivo experiments.
INTRODUCTION Recent advances in the in-vitro performances of potentially implantable amperometric enzymebased electrodes made it possible to progress toward a more systematic evaluation of their in-vivo behaviour. Most of the in-vivo studies published have described a short-term glucose monitoring performed with needle-type electrodes implanted subcutaneously in animals (Shichiri et al., 1983; Claremont eruf., 1986; Itoeral., 1986; Fischereral., 1987; Matthews ef al., 1988; Pfeiffer & Kemer, 1988; Ertefai & Gough, 1989; Rebrin et&, 1989) or Biosensom& Bioelecttmics
enzyme
electrode.
subcutaneous
men (Shichiri et al., 1986; Matthews et al., 1988; Pickup et al., 1989). The sensor response, i.e. the current, was compared to the blood glucose concentration measured conventionally. The results of these studies validated the subcutaneous tissue as a usable glucose sensing site, although some controversy concerning the exact relationships between blood and tissue glycaemia still remains (as evidenced by the published subcutaneous glucose values which vary between 20 and 85% of the plasma glucose levels; see references cited above). Nevertheless, it has ‘now been shown that for slow variations of glycaemia the subcutaneous glucose concentrations are
31 09%5663/91/$03.500 1991 Elsevier Science Publishers Ltd. England. Printed in Great Britain
32
A4 Koudelka F. Rohner-Jeanrenaud, J. Terrettaz, E. Bobbioni-Harsch, N. F. de Rooij, B. Jeanrenaud
approximately those of the blood (Fischer et al., 1987; Jansson ef al., 1988; Rebrin er al., 1989). Based on these considerations, a two-point in-vivo calibration has been recently proposed (Velho et al., 1988). It has also been shown that the sensors, at the present stage of development, fulfil some of the physiological requirements for intracorporal glucose monitoring, e.g. adequate linear range and response time. However, their sensitivity has been shown to decline with time and a drift of the baseline, most probably due to changes in the conditions at the implantation site, has been reported. As a consequence of this time evolution of the in-vivo performances of the sensors, an in-vivo periodical calibration of the sensors is mandatory. This in-vivo calibration is all the more needed because the sensitivity assessed in vitro may not represent that determined in vivo. We have recently described the realization and the in-vitro characteristics of a miniaturized, microfabricated, planar, enzyme-based glucose sensor (Gemet et al., 1989; Koudelka et al., 1989). This paper describes the in-vivo behaviour and calibration of these sensors, assessed using hyperglycaemia, euglycaemia and hypoglycaemia plateau levels. For this purpose, the sensors were implanted subcutaneously in normal rats, with their sensitive part being in contact with the hypodermis. The sensor responses were compared to conventionally measured plasma glucose levels. The in-vivo sensor sensitivity was calculated according to Velho et al. (1988) and then used to establish the relationship between plasma and apparent subcutaneous glucose concentrations.
MATERIAL
AND METHODS
Chemicals and reagents
All chemicals were obtained from Fluka (Fluka Chemie AG, Buchs, Switzerland) and were of analytical grade quality. A modified Ringer solution was used and contained 115 mM NaCI, 2.5 mM KCI, 5 IIIM K2HP04 and 0.5 IIIM KH2P04, pH 7.5. Glucose oxidase (GOD, EC 1.1.3.4 from Aspergirrus niger, 250 IU/mg) and bovine serum albumin (BSA, fraction V) were obtained from Calbiochem (Lucerne, Switzerland). Polyurethane was a gift of the Japan Erastran Inc. (Tokyo, Japan). The sensor was encapsulated with Araldite
(Ciba-Geigy, Basel, Switzerland) epoxy resin. Insulin was obtained from Novo (Nova Industri A/S, Bagsvaerd, Denmark). The glucose solutions were purchased from Vifor (Vifor SA, Geneva, Switzerland). Equipment
The amperometric measurements were performed using either an IBM EC 225 voltametric analyzer or an in-house-built potentiostat developed for the particular use. The working potential for Hz02 detection was O-7 V with respect to an Ag/AgCl/Clreference electrode. In-vitro calibrations were made in a cell thermostated at 37°C. Sensor fabrication
Described elsewhere in detail (Gemet er al., 1989; Koudelka et al., 1989), the sensor realization consisted basically of three steps. (1) The electrochemical cell (dimensions of 0.8 mm X 3 mm X 0.38 mm) was realized on an Si/SiOz/A1203 substrate and consisted of three thin-film electrodes - two Pt and one Ag/AgCl. Metallic layers were electron-gun evaporated and electrode geometries patterned using the lift-off technique. The Ag layer was partially chloridated in a 0.25 M FeC13 solution. The geometrical area of the Pt working electrode was 0.1 mm2. Following the sensor mounting on printed circuit boards, bonding and epoxy encapsulation, the final ready-to-use form was 2 mm wide and 2 cm long. (2) Chemical co-cross-linking of GOD (50 mg/ ml) and BSA (80 mg/litre) by glutaraldehyde was used to form, in situ, the enzymatic membrane. (3) The outer polyurethane membrane was deposited by dip-coating from 4% polyurethane in a l/9 (v/v) mixture of dimethyl-formamide and tetrahydrofuran. In-vivo experiments
Adult male Sprague Dawley-derived rats (SIVZ, Tierspital, Ziirich. Switzerland) weighing between 300 and 350 g were used. They were anaesthetized in the fed state (pentobarbital, 80 mg/kg, intraperitoneally). Two silastic catheters were inserted via the right jugular vein to reach, respectively, the right auricle (blood samplings for glycaemic
33
In-vivo behaviour of glucose sensors
measurements),
or the vena cava inferior (glucose
Calculation of data
or insulin infusion).
Following a small skin incision, the sensors were inserted subcutaneously by blunt tunnelling in the interscapular region, with the electrodes facing the hypodermis. They were maintained in this position using sutures. After implantation, the sensor current in the basal state (i.e. exposed to the rat’s own glycaemia) was recorded for 90 min. During the last 30 min, blood samples were taken at 5-min intervals and plasma glucose concentrations were assessed using a Beckman glucose analyzer (Beckman Instruments, Palo Alto, CA, USA). Following the 90-min equilibration period, either glucose, or insulin infusions (Precidor pumps, Infors AG, Basle, Switzerland), were started to obtain hyperor hypoglycaemic plateau levels, respectively (infusion of 14.9 * 1.5 mg/kg/min or 25.0 f l-4 mg/kg/min glucose; or of 30 mU/min insulin). Blood samples continued to be collected at 5-min intervals during 60 min. The glucose-or insulin - infusions were then stopped, while continuing blood sampling for 30 min. A typical experiment lasted 3 h and the rat’s blood volume was compensated by administering the rat’s own blood following sampling. At the end of the experiments, the sensors were removed from the rat, rinsed with water and their in-vitro sensitivity checked and found to be practically identical before and after the implantation. During the experiments, the rats were ventilated with a OJCOz (95/5 v/v) mixture and warmed at 37°C using a heating blanket with a rectal probe. The measured subcutaneous temperature at the implantation site was 37.2 f OWY’C (n = 80). using a digital thermometer (Bioscience, Sheerness, Kent, UK). The sensors were prepared at least 3 days (maximum 2 weeks) before their use in vim. They were stored in a phosphate buffer (pH 7.4) at 4°C. Their in-vitro behaviour was tested and only devices having stable (i.e. less than 10% variation in the sensitivity) and comparable characteristics were used. Based on these criteria 80% of the original sensor number were selected. The data presented were obtained with 23 sensors (one sensor per rat). A few of the sensors were tested twice. Also relatively few devices (about 10%) did not work satisfactorily when implanted and were not considered in this study. In most cases this was due to local complications (bleeding) or faulty fixation of the sensor onto the hypodermis.
The sensitivity and background current of each individual sensor were determined as being the slope and the intercept, respectively, of the line constructed by plotting the current versus the glycaemia, using the mean current and mean glucose values during 30 min of baseline (from -30 to 0 min) and 30 min of either hyper- or hypoglycaemic plateau (from 30 to 60 min). Once obtained, the sensitivity of the respective sensors allowed the calculation of the apparent subcutaneous glycaemia using the formula G’ = the apparent G’ = (i - iO)/s, where subcutaneous glycaemia (ml@; i = the sensor current (nA); is = the in-vivo background current (nA); s = the sensitivity (nAjmM). This procedure has been validated elsewhere (Velho et al., 1988). All the calculations were performed using a computer program. Values presented are means of n experiments f SEM. The Student’s twotailed f test for paired data was used to compare the in-vitro and in-vivo background currents and sensitivities. RESULTS Table 1 shows the most important characteristics of the sensors. Their performance characteristics, together with their reasonable stability during at least two weeks (with one in-vitro calibration curve per day), satisfy some of the criteria of an TABLE 1 Main in-vitro Characteristics of the Glucose Sensors Size (width, length. height) Ready-to-use device Enzyme membrane thickness Polyurethane membrane thickness Linear range Sensitivity Background current (at 0 mM) Temperature coefficient 95% response time at 20 mm01 n = 23
0.7 mm X 3 mm X O-38 mm 2mmX2cm 20pm
3flm 20 mM at pOZ > 20 mm Hg 2.1 + 015 nA/mM 1.4f 0.1 nA 5%/T 180f4s
at 20 mM
34
M. Koudelka F. Rohner-Jeanrenaud, J Terrettaz E. Bobbioni-Harsch, h? E de Rooij, B. Jeanrenaud
implantable device for short-term glucose monitoring. Furthermore, the sensor response is pH- (tested in the range of 6-8) and stirindependent. Using the actual glycaemias and the corresponding values of the sensor current, the in-vivo sensitivity of the sensors was calculated and found to be 064 f 0.05 nA/mM (n = 23). Figure 1 shows the apparent subcutaneous glycaemia and the plasma glucose concentrations, during a glycaemia plateau maintained at 13.8 + 0.2 mM (n = 9). As can be seen, blood glucose levels increased from 7.8 f 0.1 mM in the basal state to 13.8 + O-2 mM at the plateau level. It can be seen that the sensor responded with less than 5 min delay and that it mirrored adequately the glycaemia changes, including the descending part of the curve, when the glucose infusion was stopped. As illustrated by Figs 2 and 3, similar behaviours of the sensors were also observed for other glycaemia plateaus performed at 11.2 AZO-2 mM (n = IO), or at 3.5 f 0.1 mM (n = 9). During the establishment of the lowest glycaemic plateau (3.5 mM) there was a transient the actual glycaemia discrepancy between measurements and the apparent subcutaneous a discrepancy that completely glycaemia, disappeared at subsequent time intervals. Taking the results of all the experiments mentioned above, Fig. 4 shows the global plot of the apparent subcutaneous glycaemia versus the plasma glucose levels. As can be seen, there is a good correlation between the two parameters (intercept = -0.066, slope = 0997, correlation
-30
:
:
-20-10
:
0
:
10
:
20
:
30
:
40
:
50
:
60
:
70
:
60
I
90
time (min)
Fig. 2. Apparent hypodermic (subcutaneous) glycaemia (closed symbols), before, after. and during a glycaemia plateau find at Il.2 f 02 m.w compared to the actual glycaemia measurements (open symbols), made on plasma samples by conventional method. Mean of 10 animals +I SEM.
2-I -30
:
-20-10
:
:
0
:
10
:
20
:
30
:
40
:
50
:
60
:
70
:
60
3
time (min)
Fig. 3. Apparent hypodermic (subcutaneous) glycaemia (closed symbols) before, afier and during a glycaemia plateau fued at 3.5 f 0 I mM via an insulin infusion. compared to the actual glycaemia measurements (open symbols). made on plasma samples by conventional method. Mean of 9 animals f SEM.
coefficient = O-9782). It should be noted that a similar correlation was obtained when the sensors were calibrated using only 10 min (instead of 30) of basal and hyper- or hypoglycaemic plateaus (data not shown).
6-4-2-0, : -30-20-10
0-l
:
: 0
: 10
: 20
: 30
: 40
: 50
: 60
: 70
: 60
90
time (min)
Fig. 1. Apparent hypodermic (subcutaneous) (closed symbols), before, after, and during a plateau fired at 13.8 + 02 mM. compared to glycaemia measurements (open symbols), made samples by conventional method. Mean of f SEM.
glycaemia glycaemia the actual on plasma 9 animals
DISCUSSION Previous studies have shown that enzymatic glucose sensors had suitable performance characteristics for short-term in-vivo glucose obtained with monitoring. The results subcutaneously implanted sensors provided
In-vivo behaviour of glucose sensors
w.nous glucose (mmol/l)
Fig. 4. Correlation between estimated apparent (G’) hypodermic (subcutaneous) glycaemia and concomitant plasma glucose levels (G) measured for various glycaemia plateaus (75 pointsfor 23 sensors, G’ = 0 997 *G - 0 066. r = 0 9782).
proof of the existence (supported also by some other studies) of a direct relationship between blood and tissue glucose concentrations. In spite of the lack of knowledge of the actual tissular the glucose concentration, subcutaneous implantation site was thus indirectly validated (see Introduction for references). As evidenced by the data published so far (see Introduction for references), the relationship between blood and tissue glucose concentrations - determined on the assumption that the in-vitro and in-vivo sensitivities are identical - is variable. This usually results in a systematic underestimation of the tissue glucose concentrations. These findings, which could not be entirely attributed to the performance characteristics of the sensor, nor explained by the physiology of microcirculation, made evident the need for the in-vivo sensor calibration. A two-point calibration method was used, based on the equilibration between blood and tissue glucose levels for slow variations of blood glucose concentration. This method was chosen since, in addition to obtaining the value of the in-vivo sensitivity of the sensors, it allows that of the in-vivo background current to be estimated. Each experiment consisted of two plateaus of glycaemia, the first corresponding to the basal state (the rat’s own glycaemia) and the second achieved by using either glucose infusion (hyperglycaemia plateaus) or insulin administration. Due to the lack of data on the minimum time necessary to reach an equilibration between blood and subcutaneous tissue (glucose levels),
35
the plateau levels were usually maintained for about 30 min. Subsequently, the glycaemia was allowed to decrease or increase spontaneously in order to return to basal values. The actual curves of plasma glucose levels and apparent subcutaneous glucose concentrations were closely parallel, prior, during and following the glucose clamps, (Figs l-3). However, it should be noted that although the apparent values of glycaemia were close to the actual ones there was an underestimation of the apparent glycaemia at the end of the experiments (Figs l-3). This can be attributed to the fact that all sensors had a drift in their current that could already be observed at the beginning of the experiments. The reasons underlying such a phenomenon are as yet undetermined, but should be investigated. Note, however, that a good correlation was observed between actual venous glycaemia measurements and the apparent subcutaneous glucose concentrations (Fig. 4), i.e. the sensor response was linear between 3.5 and 13.8 IIIM. The last but very important characteristic of the sensor is the background current, i.e. the current in the absence of glucose. Not very much is known, in general, about the values of background currents in vim, except that they are expected to be quite different from those measured in vitro (e.g. presence of electrochemically active interferences). The twopoint calibration method, although rather impractical to perform, allows the value of the sensor background current to be determined. Thus, by plotting the current as a function of glycaemia and extrapolating the straight line to zero glycaemia a reasonable estimation of the background currents can be obtained 1.25 f O-4 nA, n = 23. For comparison, the in-vitro background current was 1.4 f O-1 nA, n = 23,~ < 0.8, NS. On the contrary, the in-vivo and in-vitro sensitivities of the sensors were statistically significantly different (064 f 005 nA/ ITIM,n = 23 ~~2.1 f 0.15 nA/mM,n = 23,respectively, p < 0.01). A more extensive study is, of course, needed to ascertain the reproducibility and time stability of the in-vivo background current values so that they can be compensated for initially. Thus, a more convenient one-point calibration might possibly be used in the future. The combined data show that microfabricated glucose sensors respond adequately when tested in vivo, at least for the timing chosen in the present study. The duration of the glucose sensors’
36
M. Koudelka, F. Rohner-Jeanrenaud, J Tewettaz# E. Bobbioni-Harsch, N. F. de Rooij, B. Jeanrenaud
performances investigated.
in
vivo
is
currently
being
ACKNOWLEDGEMENTS This work was supported by Grant No. 4.884.0.85.18 (National Programme, Bern, Switzerland); by grants-in-aid from the Fondation Lord Michelham of Hellingly, Geneva, Switzerland; from Nestle SA (Vevey, Switzerland), and the Committee for the Promotion of Applied Scientific Research (Bern, Switzerland). The authors would like to thank Mr S. Gernet for the realization of the transducers. The expert technical help of MS S. Pochon and secretarial help of MS F. Touabi are greatly acknowledged. The authors would like to express their sincere thanks to Dr G. Reach and Dr D. R. Thevenot for very useful scientific discussions.
REFERENCES Claremont, D. J.. Sambrook. I. E.. Penton. C. & Pickup. J. C. (1986). Subcutaneous implantation of a ferrocene-mediated glucose sensor in pigs, Diabetologia, 29, 8 17-2 I. Ertefai, S. & Gough. D. A. (1989). Physiological preparation for studying the response of subcutaneously implanted glucose and oxygen sensors, J. Biomed. Engng, 11,362-8. Fischer, U., Et-de, R., Abel, P., Rebrin. K. Brunstein, E.. Hahn von Dorsche, H. & Freyse. E. J. (1987). Assessment of subcutaneous glucose concentration: validation of the wick technique as a reference for implanted electrochemical sensors in normal and diabetic dogs, Diabetologia, 30,940-5. Gemet. S., Koudelka, M. & de Rooij, N. S. (1989). Fabrication and characterization of a planar
electrochemical cell and its application as a glucose sensor, Sensors & Actuators 18, 59-70. Ito, K, Ikeda, S.. Asai, K.. Naruse. H., Ohkura. K, Ichihashi, H., Kamei. H. & Kondo, T. (1986). Development of subcutaneous-type glucose sensors for implantable or portable artificial pancreas, ACS Symp. Ser., 309,373-82. Jansson, P. A., Fowelin. U., Smith, U. L&nnroth, P. (1988). Characterization by microdialysis of intercellular glucose level in subcutaneous tissue in humans, Amer. J Physiol.. 255, E218-E220. Koudelka, M., Gemet. S. & de Rooij, N. S. (1989). Planar amperometric enzyme-based glucose microelectrode, Sensors & Actuators, 18, 157-65. Matthews, D. R., Bown, E., Beck, T. W.. Plotkin, E.. Lock, L.. Gosden. E. & Wickham. M. (1988). An amperometric needle-type glucose sensor tested in rats and man, Diabet. Med., 5, 248-52. Pfeiffer, E. F. & Kemer, W. (Eds) (1988). Implantable glucose sensors - the state of the art. Hormone & Metabol. Res.. Suppl. Ser., 20. Pickup, J. C.. Shaw, G. W. &Claremont, D. J. (1989). In vivo molecular sensing in diabetes mellitus: an implantable glucose sensor with direct electron transfer. Diabetologia, 32, 2 13-7. Rebrin. K, Fischer, U., Woedtke. TV.. Abel, P. & Brunstein, E. (1989). Automated feedback control of subcutaneous glucose concentration in diabetic dogs. Diabetologia. 32, 573-6. Shichiri, M., Kawamori, R.. Aymasaki. Y.,Nomura. M.. Hakui, N. & Abe, H. (1983). Glycaemic control in pancreatectomized dogs with a wearable artificial endocrine pancreas, Diabetologia, 24, 179-84. Shichiri. M.. Asakawa. N.,Yamasaki.Y., Kawamori, R. & Abe, H. (1986). Telemetry glucose monitoring device with needle-type glucose sensor: useful tool for blood glucose monitoring in diabetic individuals. Diabetes Care, 9, 298-301. Velho. G.. Frogue. Ph., Thtvenot, D. R. & Reach. G. (1988). In vivo calibration of a subcutaneous glucose sensor for determination of subcutaneous glucose kinetics. Diabetes, Nutrition & Metabolism, 3,227-33.
6
(1991) 3 l-36
In-vivo Behaviour of Hypodermically Implanted Microfabricated Glucose Sensors M. Koudelka”, F. Rohner-Jeanrenaud”, J. Terrettaz”, E. Bobbioni-Harschb, N. F. de Rooij” & B. Jeanrenaud” “Institute of Microtechnology. University of Neuchdtel, 2 rue A. Breguet. 2000 Neuchltel. Switzerland Qboratories of Metabolic Research. University and Faculty of Medicine, Geneva. Switzerland (Received 18 December
1989: revised version received 24 April 1990; accepted
1 May 1990)
The in-vivo behaviour of microfabricated GOD (glucose oxidase)/ HzOz glucose sensor implanted subcutaneously in normal anaesthetized rats has been studied. The sensor consists of a planar, three-electrode microcell, an enzyme membrane (glucose oxidase and bovine serum albumin cross-linked with glutaraldehyde) and an outer diffusion limiting polyurethane membrane. The sensor behaviour during hyperglycaemic (13.8 mM and Il.2 mM). euglycaemic (7.8 mM) and hypoglycaemic (3.5 mM) plateau levels was determined. The values of the in-vivo sensitivity (0.64 + O-05 nA/mM) and background current (I.25 + 0.4 nA) were determined using a two-point calibration method and then used to calculate apparent subcutaneous glucose concentrations. The results show the presence of a good correlation between all the plasma glucose levels (G) and the apparent subcutaneous tissue concentrations (G’), with G’ = 0997 *G - 0066. r = 0.9782. Abstrack
Keywords: glucose sensor, amperometric glucose, in-vivo experiments.
INTRODUCTION Recent advances in the in-vitro performances of potentially implantable amperometric enzymebased electrodes made it possible to progress toward a more systematic evaluation of their in-vivo behaviour. Most of the in-vivo studies published have described a short-term glucose monitoring performed with needle-type electrodes implanted subcutaneously in animals (Shichiri et al., 1983; Claremont eruf., 1986; Itoeral., 1986; Fischereral., 1987; Matthews ef al., 1988; Pfeiffer & Kemer, 1988; Ertefai & Gough, 1989; Rebrin et&, 1989) or Biosensom& Bioelecttmics
enzyme
electrode.
subcutaneous
men (Shichiri et al., 1986; Matthews et al., 1988; Pickup et al., 1989). The sensor response, i.e. the current, was compared to the blood glucose concentration measured conventionally. The results of these studies validated the subcutaneous tissue as a usable glucose sensing site, although some controversy concerning the exact relationships between blood and tissue glycaemia still remains (as evidenced by the published subcutaneous glucose values which vary between 20 and 85% of the plasma glucose levels; see references cited above). Nevertheless, it has ‘now been shown that for slow variations of glycaemia the subcutaneous glucose concentrations are
31 09%5663/91/$03.500 1991 Elsevier Science Publishers Ltd. England. Printed in Great Britain
32
A4 Koudelka F. Rohner-Jeanrenaud, J. Terrettaz, E. Bobbioni-Harsch, N. F. de Rooij, B. Jeanrenaud
approximately those of the blood (Fischer et al., 1987; Jansson ef al., 1988; Rebrin er al., 1989). Based on these considerations, a two-point in-vivo calibration has been recently proposed (Velho et al., 1988). It has also been shown that the sensors, at the present stage of development, fulfil some of the physiological requirements for intracorporal glucose monitoring, e.g. adequate linear range and response time. However, their sensitivity has been shown to decline with time and a drift of the baseline, most probably due to changes in the conditions at the implantation site, has been reported. As a consequence of this time evolution of the in-vivo performances of the sensors, an in-vivo periodical calibration of the sensors is mandatory. This in-vivo calibration is all the more needed because the sensitivity assessed in vitro may not represent that determined in vivo. We have recently described the realization and the in-vitro characteristics of a miniaturized, microfabricated, planar, enzyme-based glucose sensor (Gemet et al., 1989; Koudelka et al., 1989). This paper describes the in-vivo behaviour and calibration of these sensors, assessed using hyperglycaemia, euglycaemia and hypoglycaemia plateau levels. For this purpose, the sensors were implanted subcutaneously in normal rats, with their sensitive part being in contact with the hypodermis. The sensor responses were compared to conventionally measured plasma glucose levels. The in-vivo sensor sensitivity was calculated according to Velho et al. (1988) and then used to establish the relationship between plasma and apparent subcutaneous glucose concentrations.
MATERIAL
AND METHODS
Chemicals and reagents
All chemicals were obtained from Fluka (Fluka Chemie AG, Buchs, Switzerland) and were of analytical grade quality. A modified Ringer solution was used and contained 115 mM NaCI, 2.5 mM KCI, 5 IIIM K2HP04 and 0.5 IIIM KH2P04, pH 7.5. Glucose oxidase (GOD, EC 1.1.3.4 from Aspergirrus niger, 250 IU/mg) and bovine serum albumin (BSA, fraction V) were obtained from Calbiochem (Lucerne, Switzerland). Polyurethane was a gift of the Japan Erastran Inc. (Tokyo, Japan). The sensor was encapsulated with Araldite
(Ciba-Geigy, Basel, Switzerland) epoxy resin. Insulin was obtained from Novo (Nova Industri A/S, Bagsvaerd, Denmark). The glucose solutions were purchased from Vifor (Vifor SA, Geneva, Switzerland). Equipment
The amperometric measurements were performed using either an IBM EC 225 voltametric analyzer or an in-house-built potentiostat developed for the particular use. The working potential for Hz02 detection was O-7 V with respect to an Ag/AgCl/Clreference electrode. In-vitro calibrations were made in a cell thermostated at 37°C. Sensor fabrication
Described elsewhere in detail (Gemet er al., 1989; Koudelka et al., 1989), the sensor realization consisted basically of three steps. (1) The electrochemical cell (dimensions of 0.8 mm X 3 mm X 0.38 mm) was realized on an Si/SiOz/A1203 substrate and consisted of three thin-film electrodes - two Pt and one Ag/AgCl. Metallic layers were electron-gun evaporated and electrode geometries patterned using the lift-off technique. The Ag layer was partially chloridated in a 0.25 M FeC13 solution. The geometrical area of the Pt working electrode was 0.1 mm2. Following the sensor mounting on printed circuit boards, bonding and epoxy encapsulation, the final ready-to-use form was 2 mm wide and 2 cm long. (2) Chemical co-cross-linking of GOD (50 mg/ ml) and BSA (80 mg/litre) by glutaraldehyde was used to form, in situ, the enzymatic membrane. (3) The outer polyurethane membrane was deposited by dip-coating from 4% polyurethane in a l/9 (v/v) mixture of dimethyl-formamide and tetrahydrofuran. In-vivo experiments
Adult male Sprague Dawley-derived rats (SIVZ, Tierspital, Ziirich. Switzerland) weighing between 300 and 350 g were used. They were anaesthetized in the fed state (pentobarbital, 80 mg/kg, intraperitoneally). Two silastic catheters were inserted via the right jugular vein to reach, respectively, the right auricle (blood samplings for glycaemic
33
In-vivo behaviour of glucose sensors
measurements),
or the vena cava inferior (glucose
Calculation of data
or insulin infusion).
Following a small skin incision, the sensors were inserted subcutaneously by blunt tunnelling in the interscapular region, with the electrodes facing the hypodermis. They were maintained in this position using sutures. After implantation, the sensor current in the basal state (i.e. exposed to the rat’s own glycaemia) was recorded for 90 min. During the last 30 min, blood samples were taken at 5-min intervals and plasma glucose concentrations were assessed using a Beckman glucose analyzer (Beckman Instruments, Palo Alto, CA, USA). Following the 90-min equilibration period, either glucose, or insulin infusions (Precidor pumps, Infors AG, Basle, Switzerland), were started to obtain hyperor hypoglycaemic plateau levels, respectively (infusion of 14.9 * 1.5 mg/kg/min or 25.0 f l-4 mg/kg/min glucose; or of 30 mU/min insulin). Blood samples continued to be collected at 5-min intervals during 60 min. The glucose-or insulin - infusions were then stopped, while continuing blood sampling for 30 min. A typical experiment lasted 3 h and the rat’s blood volume was compensated by administering the rat’s own blood following sampling. At the end of the experiments, the sensors were removed from the rat, rinsed with water and their in-vitro sensitivity checked and found to be practically identical before and after the implantation. During the experiments, the rats were ventilated with a OJCOz (95/5 v/v) mixture and warmed at 37°C using a heating blanket with a rectal probe. The measured subcutaneous temperature at the implantation site was 37.2 f OWY’C (n = 80). using a digital thermometer (Bioscience, Sheerness, Kent, UK). The sensors were prepared at least 3 days (maximum 2 weeks) before their use in vim. They were stored in a phosphate buffer (pH 7.4) at 4°C. Their in-vitro behaviour was tested and only devices having stable (i.e. less than 10% variation in the sensitivity) and comparable characteristics were used. Based on these criteria 80% of the original sensor number were selected. The data presented were obtained with 23 sensors (one sensor per rat). A few of the sensors were tested twice. Also relatively few devices (about 10%) did not work satisfactorily when implanted and were not considered in this study. In most cases this was due to local complications (bleeding) or faulty fixation of the sensor onto the hypodermis.
The sensitivity and background current of each individual sensor were determined as being the slope and the intercept, respectively, of the line constructed by plotting the current versus the glycaemia, using the mean current and mean glucose values during 30 min of baseline (from -30 to 0 min) and 30 min of either hyper- or hypoglycaemic plateau (from 30 to 60 min). Once obtained, the sensitivity of the respective sensors allowed the calculation of the apparent subcutaneous glycaemia using the formula G’ = the apparent G’ = (i - iO)/s, where subcutaneous glycaemia (ml@; i = the sensor current (nA); is = the in-vivo background current (nA); s = the sensitivity (nAjmM). This procedure has been validated elsewhere (Velho et al., 1988). All the calculations were performed using a computer program. Values presented are means of n experiments f SEM. The Student’s twotailed f test for paired data was used to compare the in-vitro and in-vivo background currents and sensitivities. RESULTS Table 1 shows the most important characteristics of the sensors. Their performance characteristics, together with their reasonable stability during at least two weeks (with one in-vitro calibration curve per day), satisfy some of the criteria of an TABLE 1 Main in-vitro Characteristics of the Glucose Sensors Size (width, length. height) Ready-to-use device Enzyme membrane thickness Polyurethane membrane thickness Linear range Sensitivity Background current (at 0 mM) Temperature coefficient 95% response time at 20 mm01 n = 23
0.7 mm X 3 mm X O-38 mm 2mmX2cm 20pm
3flm 20 mM at pOZ > 20 mm Hg 2.1 + 015 nA/mM 1.4f 0.1 nA 5%/T 180f4s
at 20 mM
34
M. Koudelka F. Rohner-Jeanrenaud, J Terrettaz E. Bobbioni-Harsch, h? E de Rooij, B. Jeanrenaud
implantable device for short-term glucose monitoring. Furthermore, the sensor response is pH- (tested in the range of 6-8) and stirindependent. Using the actual glycaemias and the corresponding values of the sensor current, the in-vivo sensitivity of the sensors was calculated and found to be 064 f 0.05 nA/mM (n = 23). Figure 1 shows the apparent subcutaneous glycaemia and the plasma glucose concentrations, during a glycaemia plateau maintained at 13.8 + 0.2 mM (n = 9). As can be seen, blood glucose levels increased from 7.8 f 0.1 mM in the basal state to 13.8 + O-2 mM at the plateau level. It can be seen that the sensor responded with less than 5 min delay and that it mirrored adequately the glycaemia changes, including the descending part of the curve, when the glucose infusion was stopped. As illustrated by Figs 2 and 3, similar behaviours of the sensors were also observed for other glycaemia plateaus performed at 11.2 AZO-2 mM (n = IO), or at 3.5 f 0.1 mM (n = 9). During the establishment of the lowest glycaemic plateau (3.5 mM) there was a transient the actual glycaemia discrepancy between measurements and the apparent subcutaneous a discrepancy that completely glycaemia, disappeared at subsequent time intervals. Taking the results of all the experiments mentioned above, Fig. 4 shows the global plot of the apparent subcutaneous glycaemia versus the plasma glucose levels. As can be seen, there is a good correlation between the two parameters (intercept = -0.066, slope = 0997, correlation
-30
:
:
-20-10
:
0
:
10
:
20
:
30
:
40
:
50
:
60
:
70
:
60
I
90
time (min)
Fig. 2. Apparent hypodermic (subcutaneous) glycaemia (closed symbols), before, after. and during a glycaemia plateau find at Il.2 f 02 m.w compared to the actual glycaemia measurements (open symbols), made on plasma samples by conventional method. Mean of 10 animals +I SEM.
2-I -30
:
-20-10
:
:
0
:
10
:
20
:
30
:
40
:
50
:
60
:
70
:
60
3
time (min)
Fig. 3. Apparent hypodermic (subcutaneous) glycaemia (closed symbols) before, afier and during a glycaemia plateau fued at 3.5 f 0 I mM via an insulin infusion. compared to the actual glycaemia measurements (open symbols). made on plasma samples by conventional method. Mean of 9 animals f SEM.
coefficient = O-9782). It should be noted that a similar correlation was obtained when the sensors were calibrated using only 10 min (instead of 30) of basal and hyper- or hypoglycaemic plateaus (data not shown).
6-4-2-0, : -30-20-10
0-l
:
: 0
: 10
: 20
: 30
: 40
: 50
: 60
: 70
: 60
90
time (min)
Fig. 1. Apparent hypodermic (subcutaneous) (closed symbols), before, after, and during a plateau fired at 13.8 + 02 mM. compared to glycaemia measurements (open symbols), made samples by conventional method. Mean of f SEM.
glycaemia glycaemia the actual on plasma 9 animals
DISCUSSION Previous studies have shown that enzymatic glucose sensors had suitable performance characteristics for short-term in-vivo glucose obtained with monitoring. The results subcutaneously implanted sensors provided
In-vivo behaviour of glucose sensors
w.nous glucose (mmol/l)
Fig. 4. Correlation between estimated apparent (G’) hypodermic (subcutaneous) glycaemia and concomitant plasma glucose levels (G) measured for various glycaemia plateaus (75 pointsfor 23 sensors, G’ = 0 997 *G - 0 066. r = 0 9782).
proof of the existence (supported also by some other studies) of a direct relationship between blood and tissue glucose concentrations. In spite of the lack of knowledge of the actual tissular the glucose concentration, subcutaneous implantation site was thus indirectly validated (see Introduction for references). As evidenced by the data published so far (see Introduction for references), the relationship between blood and tissue glucose concentrations - determined on the assumption that the in-vitro and in-vivo sensitivities are identical - is variable. This usually results in a systematic underestimation of the tissue glucose concentrations. These findings, which could not be entirely attributed to the performance characteristics of the sensor, nor explained by the physiology of microcirculation, made evident the need for the in-vivo sensor calibration. A two-point calibration method was used, based on the equilibration between blood and tissue glucose levels for slow variations of blood glucose concentration. This method was chosen since, in addition to obtaining the value of the in-vivo sensitivity of the sensors, it allows that of the in-vivo background current to be estimated. Each experiment consisted of two plateaus of glycaemia, the first corresponding to the basal state (the rat’s own glycaemia) and the second achieved by using either glucose infusion (hyperglycaemia plateaus) or insulin administration. Due to the lack of data on the minimum time necessary to reach an equilibration between blood and subcutaneous tissue (glucose levels),
35
the plateau levels were usually maintained for about 30 min. Subsequently, the glycaemia was allowed to decrease or increase spontaneously in order to return to basal values. The actual curves of plasma glucose levels and apparent subcutaneous glucose concentrations were closely parallel, prior, during and following the glucose clamps, (Figs l-3). However, it should be noted that although the apparent values of glycaemia were close to the actual ones there was an underestimation of the apparent glycaemia at the end of the experiments (Figs l-3). This can be attributed to the fact that all sensors had a drift in their current that could already be observed at the beginning of the experiments. The reasons underlying such a phenomenon are as yet undetermined, but should be investigated. Note, however, that a good correlation was observed between actual venous glycaemia measurements and the apparent subcutaneous glucose concentrations (Fig. 4), i.e. the sensor response was linear between 3.5 and 13.8 IIIM. The last but very important characteristic of the sensor is the background current, i.e. the current in the absence of glucose. Not very much is known, in general, about the values of background currents in vim, except that they are expected to be quite different from those measured in vitro (e.g. presence of electrochemically active interferences). The twopoint calibration method, although rather impractical to perform, allows the value of the sensor background current to be determined. Thus, by plotting the current as a function of glycaemia and extrapolating the straight line to zero glycaemia a reasonable estimation of the background currents can be obtained 1.25 f O-4 nA, n = 23. For comparison, the in-vitro background current was 1.4 f O-1 nA, n = 23,~ < 0.8, NS. On the contrary, the in-vivo and in-vitro sensitivities of the sensors were statistically significantly different (064 f 005 nA/ ITIM,n = 23 ~~2.1 f 0.15 nA/mM,n = 23,respectively, p < 0.01). A more extensive study is, of course, needed to ascertain the reproducibility and time stability of the in-vivo background current values so that they can be compensated for initially. Thus, a more convenient one-point calibration might possibly be used in the future. The combined data show that microfabricated glucose sensors respond adequately when tested in vivo, at least for the timing chosen in the present study. The duration of the glucose sensors’
36
M. Koudelka, F. Rohner-Jeanrenaud, J Tewettaz# E. Bobbioni-Harsch, N. F. de Rooij, B. Jeanrenaud
performances investigated.
in
vivo
is
currently
being
ACKNOWLEDGEMENTS This work was supported by Grant No. 4.884.0.85.18 (National Programme, Bern, Switzerland); by grants-in-aid from the Fondation Lord Michelham of Hellingly, Geneva, Switzerland; from Nestle SA (Vevey, Switzerland), and the Committee for the Promotion of Applied Scientific Research (Bern, Switzerland). The authors would like to thank Mr S. Gernet for the realization of the transducers. The expert technical help of MS S. Pochon and secretarial help of MS F. Touabi are greatly acknowledged. The authors would like to express their sincere thanks to Dr G. Reach and Dr D. R. Thevenot for very useful scientific discussions.
REFERENCES Claremont, D. J.. Sambrook. I. E.. Penton. C. & Pickup. J. C. (1986). Subcutaneous implantation of a ferrocene-mediated glucose sensor in pigs, Diabetologia, 29, 8 17-2 I. Ertefai, S. & Gough. D. A. (1989). Physiological preparation for studying the response of subcutaneously implanted glucose and oxygen sensors, J. Biomed. Engng, 11,362-8. Fischer, U., Et-de, R., Abel, P., Rebrin. K. Brunstein, E.. Hahn von Dorsche, H. & Freyse. E. J. (1987). Assessment of subcutaneous glucose concentration: validation of the wick technique as a reference for implanted electrochemical sensors in normal and diabetic dogs, Diabetologia, 30,940-5. Gemet. S., Koudelka, M. & de Rooij, N. S. (1989). Fabrication and characterization of a planar
electrochemical cell and its application as a glucose sensor, Sensors & Actuators 18, 59-70. Ito, K, Ikeda, S.. Asai, K.. Naruse. H., Ohkura. K, Ichihashi, H., Kamei. H. & Kondo, T. (1986). Development of subcutaneous-type glucose sensors for implantable or portable artificial pancreas, ACS Symp. Ser., 309,373-82. Jansson, P. A., Fowelin. U., Smith, U. L&nnroth, P. (1988). Characterization by microdialysis of intercellular glucose level in subcutaneous tissue in humans, Amer. J Physiol.. 255, E218-E220. Koudelka, M., Gemet. S. & de Rooij, N. S. (1989). Planar amperometric enzyme-based glucose microelectrode, Sensors & Actuators, 18, 157-65. Matthews, D. R., Bown, E., Beck, T. W.. Plotkin, E.. Lock, L.. Gosden. E. & Wickham. M. (1988). An amperometric needle-type glucose sensor tested in rats and man, Diabet. Med., 5, 248-52. Pfeiffer, E. F. & Kemer, W. (Eds) (1988). Implantable glucose sensors - the state of the art. Hormone & Metabol. Res.. Suppl. Ser., 20. Pickup, J. C.. Shaw, G. W. &Claremont, D. J. (1989). In vivo molecular sensing in diabetes mellitus: an implantable glucose sensor with direct electron transfer. Diabetologia, 32, 2 13-7. Rebrin. K, Fischer, U., Woedtke. TV.. Abel, P. & Brunstein, E. (1989). Automated feedback control of subcutaneous glucose concentration in diabetic dogs. Diabetologia. 32, 573-6. Shichiri, M., Kawamori, R.. Aymasaki. Y.,Nomura. M.. Hakui, N. & Abe, H. (1983). Glycaemic control in pancreatectomized dogs with a wearable artificial endocrine pancreas, Diabetologia, 24, 179-84. Shichiri. M.. Asakawa. N.,Yamasaki.Y., Kawamori, R. & Abe, H. (1986). Telemetry glucose monitoring device with needle-type glucose sensor: useful tool for blood glucose monitoring in diabetic individuals. Diabetes Care, 9, 298-301. Velho. G.. Frogue. Ph., Thtvenot, D. R. & Reach. G. (1988). In vivo calibration of a subcutaneous glucose sensor for determination of subcutaneous glucose kinetics. Diabetes, Nutrition & Metabolism, 3,227-33.
