H* Journal of The Electrochemical Society, *** * H6-H11 2005
****-****/****/*** * /H6/6/$7.00 The Electrochemical Society, Inc.
Design and Testing of an Impedance-Based Sensor
for Monitoring Drug Delivery
Audrey M. Johnson,a Donald R. Sadoway,b,* Michael J. Cima,b
and Robert Langera,z
a
Department of Chemical Engineering and bDepartment of Materials Science and Engineering,
Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA
A new impedance-based sensor to monitor drug delivery from an implantable microelectromechanical systems MEMS device
has been fabricated and tested. The sensor consists of two electrodes on opposing sides of a pyramidal drug reservoir. The
dissolution of the drug and advance of solution into the reservoir cause the impedance to change over time. A 100 times scale
model of the sensor was constructed to examine the effects of electrode geometry on solution resistance. An equivalent circuit was
formulated to interpret the impedance signal in terms of the resistance and double-layer capacitance of the solution in the reservoir.
The circuit was validated by impedance measurements on reservoirs lled with phosphate-buffered saline solutions of varying
concentrations. The sensor was then used to monitor the dissolution of the model drug mannitol from the drug delivery MEMS
device. The measured solution resistance and double-layer capacitance are related to the rate of transport of drug from the device,
making this sensor a potential instrument for noninvasive monitoring of drug transport from the implant in vivo. Experimental
results agree closely with the expected values of capacitance, resistance, and dissolution time calculated from physical parameters.
2004 The Electrochemical Society. DOI: 10.1149/1.1824045 All rights reserved.
Manuscript submitted January 26, 2004; revised manuscript received June 11, 2004. Available electronically November 22, 2004.
imaging;17,18 however these techniques are limited in their resolu-
A sensor for characterizing the transport of a drug from an im-
tion, accuracy, and/or detection limits.19-21 Measurements of blood
plant in vivo, in real time and noninvasively, would have the poten-
tial to yield new information about the body s response to implants and urine drug levels also give some feedback on the operation of a
and the impact on drug or analyte transport. Understanding in vivo drug delivery device. However, many transport barriers exist be-
transport is critical for a number of applications including active tween the implant region and the point of measurement, and the
drug delivery devices and long-term implantable sensors. Most cur- complex pharmacokinetics of many drugs further complicates the
interpretation of these measurements.22-24
rent methods for measuring drug transport from implants are inva-
sive and destructive.1-4 We have designed and tested a novel sensor A drug delivery MEMS device has been developed previously in
that allows the monitoring of drug dissolution from an implantable our laboratory which allows the release of multiple drugs in a pul-
satile fashion.25,26 It consists of an array of microreservoirs etched
microelectromechanical systems MEMS device in real time with-
out disturbing the medium through which the drug is diffusing. into a silicon substrate, each capped by a thin gold membrane. Upon
Furthermore, the physical disposition of the sensor is such that application of a 1 V potential in the presence of chloride ion, the
without recalibration it can continue to give meaningful data even gold membrane oxidizes to form soluble gold chloride, thinning the
under changing conditions in the surrounding tissue, e.g., implant membrane until it fails. The drug within the reservoir then dissolves
encapsulation. into the surroundings as the aqueous solution advances into the
Implants are often walled off from the rest of the body by an reservoir.
avascular brous capsule of tissue that inhibits transport of sub- This paper presents a modi cation of the drug delivery MEMS
stances such as oxygen or nutrients between the body and the device in which two electrodes within each drug reservoir serve as a
implant.5 The formation of a brous capsule around implanted de- novel drug release sensor, as depicted in Fig. 1. Each reservoir has
vices has particularly important implications for drug delivery and the shape of a square pyramid due to the anisotropic etch used to
sensing applications. Current drug delivery implants are usually de- create them, and the electrodes cover two opposing sides of this
signed for a slow, steady release of a drug over time. In such a case, pyramid. Two-electrode impedance spectroscopy is used to measure
the transport pro le of drug is pseudo-steady state, so that while the the electrical characteristics of the reservoir. As the drug dissolves,
brous capsule affects drug release, the device is still able to the change in the electrical signature of the system allows real time
function.6,7 However, in the case of active devices, where quick monitoring of the rate of transport of drug away from the device.
pulsatile or more complex release pro les are desired, the isolation The utility of the sensor was demonstrated by in vitro measurements
of the implant causes a time lag due to the nite amount of time interpreted by means of an equivalent circuit.
required for diffusion of the drug through the capsule.8,9 This time
delay is also a key factor in the performance of in vivo sensors.10,11 Experimental
Even if sensor components are made resistant to the immune re- Chemicals. D-mannitol and potassium chloride Sigma, St.
sponse, the sensor must be continually recalibrated to account for Louis, MO were used as received. Phosphate buffered saline, PBS,
the changing conditions in the tissue surrounding it.12
consisting of 0.137 M NaCl, 0.001 M KH2 PO4, 0.01 M Na2 HPO4,
Most methods for determining the extent of implant encapsula-
and 0.0027 M KCl at pH 7.4 when diluted to 1 time concentration
tion and the drug dissolution pro le in the vicinity of an implant
Roche, Indianapolis, IN was purchased as a 10 times solution and
involve invasive methods that result in the destruction of the sample.
diluted to the appropriate concentration. Solutions were made with
Each data point requires a different implant and animal, contributing
deionized water passed through a MilliQ Millipore, Billerica, MA
to a high variability in measurements. Nondestructive methods have
system, resistivity above 17.8 M cm. Wires and bond pads were
signi cant limitations. Those involving direct observation of drug
insulated with EP42HT medical grade epoxy MasterBond, Hacken-
are restricted to unique systems such as the translucent rabbit ear13
sack, NJ, cured for 24 h at room temperature.
or rat dorsal skin clamped in a glass window for viewing.14 Phar-
macokinetic studies also involve monitoring of drug distributions in Microfabrication of devices. Drug delivery MEMS devices
vivo through such methods as microdialysis15,16 and nuclear were fabricated in the Microsystems Technology Laboratory at MIT.
Silicon wafers were 4 in., double-side polished, 300 m thick, 1-10
cm, p-doped 100 silicon WaferNet, San Jose, CA . A 100 nm
thermal oxide layer was grown on both surfaces and covered with a
* Electrochemical Society Active Member.
150 nm low pressure chemical vapor deposition LPCVD nitride
z
E-mail: abqq6a@r.postjobfree.com
Journal of The Electrochemical Society, 152 1 H6-H11 2005 H7
shadow mask was aligned over the device wafer Karl Suss MA4
Aligner, Suss MicroTec Inc., Waterbury Center, VT and 300 nm
gold over a 10 nm titanium adhesion layer was deposited via elec-
tron beam evaporation to form the impedance electrodes. A 650 nm
thick oxide layer was deposited by plasma deposition and patterned
over the front sides of the wafers to insulate the trigger electrodes
from the environment. Wafers were then diced into 5 5 mm de-
vices on an automatic dicing saw 2060 blade, DAD-2H/6T dicing
saw, Disco Hi-Tec America, Inc., Santa Clara, CA .
Scale model of device reservoirs. A 100 times scale model of a
square pyramidal drug delivery device reservoir was fabricated from
acrylic by the MIT machine shop. The pyramid was of base length
4.8 cm, height 3.0 cm, and base angle 54.4 . One side of the pyra-
mid was removable to provide access to the pyramid interior. Gold
foil electrodes 25 m thick, 99.9 %, Sigma were cut to cover two
opposing interior sides of the pyramid with tabs extending from the
top of the model for electrical connections.
PBS solution conductivity measurements. The conductivities of
PBS solutions were measured using a two-electrode conductivity
cell Industrial Instruments, Inc., Cedar Grove, NJ fully immersed
in the solution to be measured. The impedance spectrum was re-
corded using a Solartron 1255B frequency response analyzer with a
SI1287 electrochemical interface. The impedance was measured for
frequencies between 102 and 106 Hz, suf cient to reveal the critical
point or notch in the Cole-Cole plot. The critical point was indepen-
dent of the magnitude of the ac excitation voltage. The cell constant
was determined using standard solutions of 0.01 D,c 0.1 D, and 1.0
D potassium chloride. All solutions were kept at room temperature,
23.5 C, and the conductivities of the standard solutions at this tem-
28
perature were interpolated from values given by Wu and Koch. The
measured value of the cell constant was then used to determine the conductivity of 0.01,
0.1, 0.5, and 1.0 times PBS solutions.
Impedance of reservoir scale model. The two-electrode imped-
ance of the 100 times scale model of a device reservoir was mea-
sured for frequencies between 10 and 106 Hz. An initial aliquot of
1.0 times PBS was added to the model to make contact between the
Figure 1. The drug delivery MEMS device: a cutaway perspective view of foil electrodes and the impedance measurement was repeated. The
the prototype, with electrodes on the top surface for drug release and elec- impedance measurement and addition of PBS was repeated in incre-
trodes on the bottom surface and in the reservoirs for drug release monitor- ments of 1 to 2 mL solution until the model was entirely full. The
ing, dimensions 5 5 0.3 mm; b view of one device reservoir showing experiment was repeated using 0.5 times PBS and 0.1 times PBS
electrode con guration to scale; c idealized representation of drug release solutions.
from a reservoir.
Impedance of microdevice reservoirs. A drug delivery MEMS
device was packaged with open reservoirs for measurement of im-
pedance submersed in PBS solutions. The impedance electrodes
layer to act as an etch stop standard batch diffusion furnaces, were connected by traces to bond pads along the edges of the de-
Thermco 10K, La Porte, IN . The back sides of the wafers were vice. Insulated gold wires were epoxied MasterBond EP42HT
patterned with an array of 480 m squares using positive photoresist alongside the device and gold wire bonds, diameter 25 m, were
OCG825-20, Arch Chemicals, Norwalk, CT, which were plasma made between the bond pads and the wires. The wires, bonds, pads,
etched through the nitride and oxide layers Plasmaquest Series II and traces were covered with epoxy to insulate them. The two-
Reactor model 145, MKS Instruments, Andover, MA . Wafers were electrode impedance of each of four reservoirs was measured for
immersed in 5.17 M potassium hydroxide at 85 C until the silicon frequencies between 10 and 106 Hz. The device was then submersed
had been etched through the entire thickness of the wafers to form in PBS solutions of 1.0, 0.5, and 0.1 times concentration. The solu-
an array of pyramidal reservoirs capped by nitride/oxide mem- tions were observed to entirely ll the reservoirs, and the impedance
branes. The front sides of the wafers were then patterned with elec- measurement across each reservoir was repeated in each solution.
trodes for the triggering of drug release from reservoirs using image-
Impedance during drug release. The reservoirs of a drug deliv-
reversal resist AZ5214 E, Clariant, Somerville, NJ . Electron beam
ery MEMS device were completely lled with 35-45 g solid man-
evaporation model VES 2550, Temescal Semiconductor Products,
nitol per reservoir by a melt process. First, powdered mannitol was
Fair eld, CA was used to deposit 300 nm gold sandwiched between
spread over the reservoirs of a template device and heated, melting
two 10 nm titanium adhesion layers, and the image-reversal resist
it into the reservoirs. Excess mannitol was removed with a razor
pattern was lifted off in acetone. The remaining nitride and oxide
blade, and the pyramid-shaped mannitol pieces remaining were re-
were removed from the back sides of the wafers by plasma etching,
moved from the reservoirs. These presized mannitol pieces were
and a conformal 100 nm oxide layer was deposited by plasma depo-
then placed into the reservoirs of the drug delivery MEMS device,
sition at 80 C Plasmaquest Series II, model 145 . A shadow mask
wafer27 was prepared separately by deep reactive ion etching heated to melt them in place, and weighed AD-4 Autobalance,
DRIE etching a wafer with the pattern of the gold impedance
electrodes to be deposited inside the device reservoirs DRIE Mul- c
The demal D is a concentration unit used in connection with the electrical con-
tiplex ICP, Surface Technology Systems, Portsmouth, NH . The ductivity of aqueous solutions.28
H8 Journal of The Electrochemical Society, 152 1 H6-H11 2005
Table I. Conductivity of PBS solutions.
PBS concentration Calculated conductivity Measured conductivity
at 25 C S cm 1 at 23.5 C S cm 1
times
2 2
1.0 1.59 10 1.57 10
3 3
0.5 8.37 10 8.32 10
3 3
0.1 1.81 10 1.79 10
4 4
0.01 1.90 10 1.88 10
Perkin-Elmer, Boston, MA . The device reservoirs were then sealed
with a small piece of a glass coverslip and epoxy, and gold wires
were epoxied beside the bond pads. Wire bonds were made between
the impedance electrodes of each reservoir and the wires, then insu-
lated with epoxy. The two-electrode impedance of each reservoir
was measured at frequencies between 103 and 106 Hz. The device
was then submersed in 1.0 times PBS solution, and the impedance
measurement was repeated. After each reservoir was opened, the
reservoir impedance was measured every 3 min until the impedance
did not change with time. At this point it was assumed that the
dissolution of the mannitol was complete.
Results and Discussion
Figure 3. The dependence of the measured solution resistance in the
PBS solution conductivity. The measured values of the electri- square pyramidal model on height of the solution in the model and PBS
cal conductivity of solutions of 1.0, 0.5, 0.1, and 0.01 times PBS at concentration.
room temperature, 23.5 C, are given in Table I. For comparison, the
expected conductivity at 25 C was estimated by summing the molar
conductivities of the ionic species present in each PBS solution. Cole plot for such a system is a vertical line with a real intercept
Molar conductivities of ionic species were interpolated from re- equal to the solution resistance. In physical systems it is generally
ported values KCl, NaCl, except for 1.0 times PBS, which was observed that a constant phase element CPE ts the data better
extrapolated from the nearest reported value or extrapolated from than a simple capacitor. This is also true in our case, as the Cole-
Cole plot has a nite, positive slope. However, although it is gener-
the limiting molar conductivity (Na2 HPO4, KH2 PO4 ) using Kohl-
rausch s law.29,30 Standard KCl solutions gave a value of 1.00 ally assumed that some sort of dispersion of physical properties
0.01 cm 1 for the cell constant. gives rise to CPE behavior, the physics behind this are not well
understood,32 and we found it unnecessary to resort to the inclusion
Impedance of 100 times scale model. The scale model of the of a CPE in order to obtain a reasonably good t to the data.
pyramidal reservoir was constructed to evaluate the effects of the The resistance of a square pyramid can be calculated by starting
unusual electrode geometry on the impedance. The expected equiva- with the formula for conductance of a solid
lent circuit and impedance for the system, an aqueous solution with
no Faradaic processes occurring,31 is shown in Fig. 2. The Cole- Y A/ L
where A is the cross-sectional area, is the resistivity, and L is the
length. By taking a differential horizontal cross section, we can in-
tegrate the conductivities of each slice from x 0 to x H, where
H is the height of the pyramid. In this case, A and L are both
functions of x and of the base angle of the pyramid. For a square
pyramid, at any height, x, the width of the electrode at the edge and
the distance between the electrodes are the same function of x, L ( x ).
This means that the differential area is given by
dA L x dx
Substituting into the equation for Y gives the formula
dY L x dx/ L x
The L ( x ) cancel out, and the equation is easily integrated from zero
to H, yielding
Y H/, R /H
From this formula we see that for the special case of a square
pyramid, the resistance scales with the inverse of the height of the
pyramid. Therefore, as the pyramid model is lled with solution, the
resistance should scale with the inverse of the height of liquid in the
model. This matches the experimental ndings as shown in Fig. 3.
In an ideal case the slope of the plot should be equal to the solution
Figure 2. Equivalent circuit and Cole-Cole plot of experimental data and
resistivity, which is the inverse of the conductivity measurements in
best t to circuit for 100 times scale model of a square pyramidal drug
Table I. For our system the linear least squares slope for each solu-
reservoir containing 10 mL of 1.0 times PBS.
Journal of The Electrochemical Society, 152 1 H6-H11 2005 H9
Table II. Solution resistance and double-layer capacitance of mi-
croreservoirs lled with PBS.
Predicted Measured Measured
solution solution double layer
PBS concentration resistance resistance capacitance
times k k nF
1.0 2.13 1.79 0.12 3.54 0.79
0.5 4.01 3.68 0.44 2.50 0.62
0.1 18.6 16.0 4.7 1.9 1.6
This is likely due to fringing at the edges of the electric eld, which
is more signi cant in smaller systems. Taken together, these obser-
vations demonstrate the validity of the equivalent circuit of Fig. 4,
giving us a good understanding of the physical system.
Monitoring drug release by impedance measurements. A goal
of this work was to demonstrate that the measurement of impedance
would allow us to monitor the rate of drug release from a MEMS
device reservoir in real time. A typical example of the impedance
measurement during release of mannitol, a model drug, from a de-
vice reservoir is shown in Fig. 5a and b, as a succession of Cole-
Cole and Bode plots. The impedance before opening the reservoir is
Figure 4. Equivalent circuit and Cole-Cole plot of experimental data and similar to that observed for reservoirs full of air, which is expected
best t to circuit for a drug delivery MEMS device reservoir lled with 0.1 because mannitol is a semicrystalline solid with negligible conduc-
times PBS. Similar results were obtained for solutions of higher ionic
tivity, and there is no electrolyte present for conduction. After open-
strength and for different device reservoirs.
ing the reservoir, the impedance changes gradually from the single
semicircle characteristic of the RC parallel circuit to the overlapping
double semicircle characteristic of the circuit in Fig. 4. The solution
tion differs from the resistivity by a factor of 0.95 0.01, which resistance drops as the saline solution provides an ever larger low
gives us an effective cell constant for this particular experimental resistance path between the electrodes, and the double-layer capaci-
setup of 0.95/H, very close to the theoretical cell constant of 1/H tance increases as the area of the electrodes in contact with solution
derived above. The experimental data agree very well with the pre- increases. Successive data ts to this circuit give the values of so-
dicted equivalent circuit and expected values for solution resistance. lution resistance and double-layer capacitance vs. time shown in Fig.
In addition, the double-layer capacitance for an aqueous solution is 5c. This corresponds to the dissolution of the mannitol and the ad-
expected to be on the order of 10 F/cm2, and the best t capaci- vance of the PBS solvent front into the reservoir. The dissolution is
tance for our data agrees with this estimate. largely complete after 90 mins. Further experiments with radio-
labeled mannitol indicate that the gradual change in sensor output
Impedance of drug delivery MEMS reservoirs. The resistance
parallels the rate of release measured by scintillation counting of the
of a square pyramid calculated above was then compared to the
release medium.
resistance measured in the drug delivery device reservoir. For the
For comparison, a rst-order estimate of the expected time for
microsystem the equivalent circuit is not as simple, as shown in Fig.
dissolution can be made by considering the dissolution and diffusion
4. The parallel resistance and capacitance arise from a parallel cur-
of a solid in one dimension.33,34 A mass balance gives the following
rent path through the silicon substrate in which the reservoir is
equation for C ( x, t )
etched. This path is due to leakage current from the electrodes
through the silicon oxide layer beneath into the silicon, and takes the 2
C / x 2;
C/ t D C x,0 0
form of a parallel RC circuit, a so-called leaky capacitor. The mag-
nitude of the leak current is determined by the connection between C 0,t C sat ; C,t 0
the wire bonds and the bond pads. This varies from reservoir to
reservoir due to the inconsistency of the wire bonding process. For where C sat is the solubility of the compound in solution and D is the
any particular reservoir the resistance and capacitance of this path diffusivity. This can be solved for the ux at the interface as a
were found to be independent of reservoir contents, as expected. The function of time35
resistance remained constant within 3% during all experiments and
1/2
Nx D dC / dx C sat D / t
the capacitance was constant within 10%. x0 x0
The Cole-Cole plot of this circuit is two overlapping semicircles,
which corresponds well to the experimental data, as depicted in the A mass balance at the interface between solid and liquid can be
representative data t given in Fig. 4. The solution resistance de- written as
pends on the ionic strength of the solution and determines the posi-
A D dC / dx x 0 dt A C s C sat d z
tion of the in ection point between the semicircles, while the silicon
resistance depends on the device reservoir and determines the over-
where A is the area exposed to solution, C s is the density of the solid
all width of the curve. Some depression of the data from the ideal t
was observed, but again, a reasonably good t was obtained without compound, and dz is the differential distance traveled by the inter-
the substitution of a CPE for the capacitive elements. face in the time dt . Note that the equation for C ( x, t ) is only valid if
The average solution resistance and double-layer capacitance of we make the assumption that the motion of the interface is slow
the device reservoirs are given in Table II. The double-layer capaci- compared to the formation of the concentration pro le pseu-
tance is of the correct order of magnitude, a factor of 104 smaller dosteady state . If we substitute the expression for the ux at the
than that observed for the 100 times scale model. The measured interface into the interfacial mass balance, we obtain an expression
solution resistance is lower than that calculated using the formula for the distance traveled by the interface the depth of penetration of
solvent into the reservoir as a function of time
for the resistance of a square pyramid by a factor of 0.86 0.02.
H10 Journal of The Electrochemical Society, 152 1 H6-H11 2005
Figure 5. Impedance of a MEMS de-
vice reservoir during release of manni-
tol into 1.0 times PBS. a Series of
Cole-Cole plots over time, b series
of Bode plots over time, and c series
of best ts of solution resistance and
double-layer capacitance over time.
allow repeated measurements of drug release from the same implant
1/2
zt 2 C sat / C s C sat Dt/
at multiple time points, without the need for explantation or disrup-
tion of the tissue environment.
Conclusion
Substituting appropriate values for the physical parameters ( C sat
0.18 g/cm3, C s 1.5 g/cm3, D 10 5 cm2/s, z 300 m, The drug delivery MEMS device with an impedance-based sen-
we obtain an estimated time for dissolution of 80 min, which sor successfully monitored the release of a drug from a device res-
agrees quite closely with the experimentally observed dissolution ervoir. The system was represented by a simple equivalent circuit
time. whose elements correspond to the physical characteristics of the
The impedance-based sensor was used to monitor dissolution of system. The solution resistance and double-layer capacitance were
mannitol in vitro and the relationship between sensor output and the measured as functions of time during release of the drug. These two
physical system was characterized by an appropriate equivalent cir- parameters are functions of the degree of penetration of solution into
cuit. Future experiments will evaluate the relationship between the the reservoir during drug release, which can be related to the rate of
sensor output and the rate of transport of the drug through various transport of the drug from the device. The sensor has the potential to
transport barriers. In addition, it would be useful to compare sensor be used as a novel method for noninvasive monitoring of drug trans-
output to traditional methods of evaluating drug release, such as port from the implant in vivo.
radio-label or uorophore detection. Finally, a critical test will be to
Acknowledgments
use the sensor to monitor drug release during long-term implantation
in vivo, evaluating the impact of brous capsule development on We thank Dr. D. DeLongchamp and Professor M. Zahn for dis-
drug transport. The nondestructive nature of the measurement will cussions of equivalent circuit analysis, and Dr. R. Shawgo and Dr. L.
Journal of The Electrochemical Society, 152 1 H6-H11 2005 H11
Arana for help with packaging and microfabrication aspects of the 16. M. Muller, Adv. Drug Delivery Rev., 45, 255 2000 .
17. K. A. Salem, A. Szymanski-Exner, R. S. Lazebnik, M. S. Breen, J. Gao, and D. L.
project. This work was supported by U.S. NIH grant no. 1 R24
Wilson, IEEE Trans. Med. Imaging, 21, 1310 2002 .
AI47739-01 and graduate fellowships from the U.S. NSF and
18. A. Szymanski-Exner, N. T. Stowe, K. Salem, R. Lazebnik, J. R. Haaga, D. L.
Merck. Wilson, and J. Gao, J. Pharm. Sci., 92, 289 2003 .
19. W. Wolf, Adv. Drug Delivery Rev., 41, 1 2000 .
Massachusetts Institute of Technology assisted in meeting the publication
20. R. E. Port and W. Wolf, Invest. New Drugs, 21, 157 2003 .
costs of this article.
21. A. J. Fischman, N. M. Alpert, and R. H. Rubin, Clin. Pharmacokinet., 41, 581
2002 .
References
22. B. J. Gudzinowicz, B. T. Younkin, Jr., and M. J. Gudzinowicz, Drug Dynamics for
1. J. F. Strasser, L. K. Fun, S. Eller, S. A. Grossman, and W. M. Saltzman, J. Phar- Analytical, Clinical, and Biological Chemists, Marcel Dekker, New York 1984 .
macol. Exp. Ther., 275, 1647 1995 . 23. X. Li and W. K. Chan, Adv. Drug Deliv. Rev., 39, 81 1999 .
2. P. Kortesuo, M. Ahola, S. Karlsson, I. Kangasniemi, A. Yli-Urpo, and J. Kiesvaara, 24. V. T. DeVita, C. Denham, J. D. Davidson, and V. T. Oliverio, Clin. Pharmacol.
Biomaterials, 21, 193 2000 .
Ther., 8, 566 1967 .
3. A. A. Sharkawy, B. Klitzman, G. A. Truskey, and W. M. Reichert, J. Biomed.
25. J. T. Santini, Jr., M. J. Cima, and R. Langer, Nature (London), 397, 335 1999 .
Mater. Res., 37, 401 1997 .
26. G. Voskerician, M. S. Shive, R. S. Shawgo, H. von Recum, J. M. Anderson, M. J.
4. R. C. Wood, E. L. LeCluyse, and J. A. Fix, Biomaterials, 16, 957 1995 .
Cima, and R. Langer, Biomaterials, 24, 1959 2003 .
5. J. Bodziony, Res. Exp. Med., 192, 305 1992 .
27. G. J. Burger, E. J. T. Smulders, J. W. Berenschot, T. S. J. Lammerink, J. H. J.
6. B. D. Ratner, J. Controlled Release, 78, 211 2002 .
Fluitman, and S. Imai, Sens. Actuators, A, 54, 669 1996 .
7. J. M. Anderson, H. Niven, J. Pelagalli, L. S. Olanoff, and R. D. Jones, J. Biomed.
Mater. Res., 15, 889 1981 . 28. Y. C. Wu and W. F. Koch, J. Solution Chem., 20, 391 1991 .
8. P. S. Leppert, L. Cammack, R. Cargill, L. Coffman, M. Cortese, K. Engle, C. 29. P. Atkins, Physical Chemistry, p. 834, W. H. Freeman & Co., New York 1994 .
Krupco, and J. A. Fix, J. Biomed. Mater. Res., 28, 713 1994 .
30. P. Vanysek, in CRC Handbook of Chemistry and Physics, 73rd ed., D. R. Lide,
9. E. Fournier, C. Passirani, C. N. Montero-Menei, and J. P. Benoit, Biomaterials, 24, Chief Editor, pp. 5-110, CRC Press, Boca Raton, FL 1992 .
3311 2003 . 31. A. J. Bard and L. R. Faulkner, Electrochemical Methods, p. 376, Wiley, New York
10. M. C. Frost and M. E. Meyerhoff, Curr. Opin. Chem. Biol., 6, 633 2002 . 1980 .
11. M. Gerritsen, J. A. Jansen, A. Kros, R. J. M. Nolte, and J. A. Lutterman, J. Invest.
32. Z. Lukacs, J. Electroanal. Chem., 432, 79 1997 .
Surg., 11, 163 1998 .
33. A. Van Hook, Crystallization Theory and Practice, p. 130, Reinhold, New York
12. C. Choleau, J. C. Klein, G. Reach, B. Aussedat, V. Demaria-Pesce, G. S. Wilson, R.
1961 .
Gifford, and W. K. Ward, Biosens. Bioelectron., 17, 647 2002 .
34. J. Christoffersen and M. R. Christoffersen, in The Experimental Determination of
13. G. R. Martin and R. K. Jain, Cancer Res., 54, 5670 1994 .
Solubilities, G. T. Hefter and R. P. T. Tomkins, Editors, p. 77, Wiley, New York
14. K. Erickson, R. D. Braun, D. Yu, J. Lanzen, D. Wilson, D. M. Brizel, T. W.
2003 .
Secomb, J. E. Biaglow, and M. W. Dewhirst, Cancer Res., 63, 4705 2003 .
35. W. M. Deen, Analysis of Transport Phenomena, p. 91, Oxford University Press,
15. K. E. Garrison, S. A. Pasas, J. D. Cooper, and M. I. Davies, Eur. J. Pharm. Sci., 17,
New York 1998 .
1 2002 .
© 2004 The Electrochemical Society. DOI: 10.1149/1.1824045 All rights reserved.