Special Lecture Series Biosensors and Instrumentation

Transcription

1 !1 Special Lecture Series Biosensors and Instrumentation Lecture 11 - An Implantable Drug Delivery Microsystem This lecture covers the development of a drug delivery device in the Institute for Integrated Micro and Nano Systems with microfabrication at the SMC. This was funded by the Japanese company Senju Pharmaceuticals who specialise in drugs to treat diseases of the eye. What we mean by drug delivery and controlled release is being able to release controlled doses of drugs from an implanted device near to the site of disease. Localisation of release means that you can use smaller doses and reduce negative side effects that can arise from systemic dosing. The obvious example here is chemotherapy in cancer treatment. You can also use very concentrated drugs for the same reason though there are obvious issues with toxicity effects. Either way, controlled drug release is a very active field. There are two main ways of achieving delivery to the point of disease, either with passive implants that slowly dissolve within the body or an active drug delivery implant where the doses are released depending on outside signals. Slide 4 shows two examples, on the left some passive nanoparticles and on the right an implantable electronic device. This system is from the company MicroChips Inc. who are commercialising technology developed at MIT which uses a similar release method to the IMNS drug delivery device. Drug Release Into the Eye (DRIE) The application we were concerned with was an implant to deliver drugs to the eye to control glaucoma. Senju were particularly interested in developing drugs which affect the intraocular pressure as the Japanese population are particularly prone to developing a certain type of glaucoma. The whole system would be implanted into the vitreal or posterior cavity of the affected eye and the drugs released there to have a direct effect on the intra-ocular pressure. Slides 6-8 show a simple block diagram of the proposed drug delivery system. The first part is the delivery device consisting of a chip with an array of individually addressable cavities. Ideally for their application they wanted one dose (one cavity) a day for at least a year so we worked towards having 400 cavities per device. The second part is the control chip which determines when a cavity should be opened and may determine which to open next when powered up. It will also generate the signals required to control the drug release process. Power is obviously very important. This could come from a battery but this could seriously increase the size of the implanted device. That s a particular problem inside the eye and so near-field inductive coupling would be used to transfer power wirelessly from an external transmitter basestation. The lack of wires is obviously another advantage for an implanted device as they would require a break in the skin or in the white or sclera of the eye, which would represent a possible source of infection. Drug Delivery Technology The drug release technology used in the device was originally developed by researchers at MIT, who went on to found MicroChips Inc. and design the large implantable drug delivery system you saw in the introduction. Slide 9 shows a cross section through one drug reservoir in the device, which is a bulk micromachined cavity in a silicon chip that is sealed at one end with a gold electrode. By applying the correct voltage to the gold cap, when it is submerged

2 !2 in an aqueous solution containing chloride ions, the surface can be corroded away. Most fluids in the human body will contain chloride ions. The gold will form a soluble chloride salt when it corrodes. Once the cap has been dissolved away the reservoir is open and the contents can diffuse out. The speed at which this happens will obviously depend on the surrounding environment. In our devices the cavities were filled with a liquid drug but it is also possible to contain the drug in a soluble solid matrix. An alternative drug release scheme was also considered which used a type of polymer microsphere available commercially under the name Expancel. These are a soft polymer shell containing a hydrocarbon liquid with a relatively low boiling point. When heated the liquid turns to gas and the spheres irreversibly expand in volume by 40 times or more. The presentation had a little movie of the spheres expanding which isn t available here. Slides show the concept for the Expancel based drug delivery device with a similar microreservoir chip containing the drugs, which is bonded on top of another wafer which has microheater elements (resistors) with a layer of PDMS (polydimethyl siloxane or silicone) which has some of the Expancel beads mixed in with it to make an ExPDMS actuator. When heated this expands to fill the cavity and force the drug out of the cavity. The photo (on slide 16) shows a prototype system that has been cracked open to show the ExPDMS that has expanded into the shape of the cavities on the upper chip. In the end it was considered that this technique required more power than the gold dissolution and so it was not used in the development of the prototype device. Designing and Fabricating the Devices There are a number of important issues that needed to be considered when designing the device. Firstly, the method used to fabricate the silicon drug reservoir cavities determines how many you can fit onto a chip of a certain size and the volume of drug that can be stored. Secondly the gold membrane is only a few hundred nm thick and is therefor very fragile. Protecting this during processing and packaging is very important. Thirdly, methods needed to be developed for filling and sealing liquid drugs into the cavities and the packaging of electrical connections for insertion into liquid environments needed to be considered. Finally we needed to determine the best signals to apply to the electrodes to ensure fast, reliable drug delivery actuation. Reservoir Fabrication: The method used to fabricate the drug reservoirs has a significant impact on their size. Hopefully you ve seen the crystal plane dependent wet etch process in silicon which uses chemicals like KOH or TMAH. This can be used to make a large cavity with a small opening on one side (for the gold cap) and a large opening for filling on the other side of the wafer. The alternative is to use a deep silicon etch process like the Bosch DRIE process, which could produce small, densely packed cavities. Using 3 (75mm) diameter silicon wafers with a thickness of ~385μm it is possible to make cavities with the dimensions shown in slide 21. The obvious problem with our design specification of 400 cavities on a chip is the large area on the back side of the chip required with this type of etch process. Slide 22 shows an SEM of cavities fabricated at the SMC using a TMAH wet etch. The volume of each cavity will be around 40 nl. With the deep reactive ion etch process it s possible to make very closely spaced cavities with a similar sized opening on the front as the wet etched holes. This would allow the 400 cavities to be fabricated in a chip that is under 5mm across, which is the specification we had for the final system. The negative side to this is that each cavity only has a volume of around 1 nl so the drug used will have to be highly concentrated.

3 !3 Fabrication Process to Protect the Gold Membrane: Slides show some of the steps of fabricating the drug delivery prototype with wet etched cavities. The steps before slide 24 include the growth of passivation layers of silicon dioxide and silicon nitride and the patterning, on the back of a double side polished wafer, of the openings for the wet etch process. Following that, the gold electrodes are fabricated using the lift off process where the gold is deposited through a patterned layer of photoresist which is subsequently removed to leave the gold pattern. The next step (slide 25) is to protect the gold with a passivation layer of the polymer Parylene which is deposited with a vacuum process. This will protect the electrodes during processing and, once patterned, will provide a passivation layer to define the areas open to the solution in the final device. Now the drug reservoirs can be etched using a wet process with TMAH (slide 26). This isn t as fast or as selective as KOH but it is considered to be compatible with CMOS electronics which might be an advantage in later devices. The etch can take many hours to go through the full thickness of a wafer but the slower etch is more controllable. The wet etch leaves a thin oxide/nitride membrane underneath the gold electrodes. This needs to be removed before the cavity can be filled and so a reactive ion etch process is used to clear the gold membrane (slide 27). The final step, before filling of the cavity, is to pattern and etch (with an oxygen plasma) the Parylene passivation to leave nothing but the thin gold membrane electrode covering the reservoir (slide 28). Slide 29 shows a microscope image of one electrode pair on a prototype chip at the end of processing. You can just about see a bulge in the gold membrane, indicating some compressive stress in the metal. This definitely shows that the dielectric membrane has been removed from the back of the gold electrode. The electrodes are marked anode and cathode rather than WE and CE because when developing the system we knew that the gold cap would be dissolved with a positive voltage on the WE. Prototype Chip Design: The size of the prototype chip is 3mm square and there are three layers in the design. On the front side (slide 30) are the the gold metal layout of the electrodes and the interconnect wires used to connect them up (in purple). At the left and right edges of the chip are 14 contacts used to make wire bond connections between the chip and the packaging that connects it to the outside world. There is one connection for each of the anode gold caps and two separate counter electrode (cathode) connections, one for the top 6 electrode pairs and one for the lower 6. These are required to allow the connections to be made in a single level of metal. You might just be able to see the second part of the layout (in orange) on the top surface which are the openings in the top level of passivation that defines the areas of gold that are open to the solution. The back side of the chip (slide 32) has this layout which defines the openings in the passivation that is used as a mask for the wet etch of the cavities. The actual holes will be a little larger than this after etching as there will be significant undercutting of the passivation layer due to misalignment to the crystal planes of the silicon and relatively poor anisotropic selectivity with TMAH compared to KOH. Filling and Packaging: I didn t go into the filling of the drug reservoirs too deeply in this lectures as it would take a while to describe all of the different methods that were considered. Filling small cavities like this with liquid and sealing them without introducing air bubbles is quite challenging and other techniques, such as incorporating the drug in a soluble solid matrix, may be more useful in the future. In the end we worked out a method using waterproof adhesive tape which was fairly successful and prototype chips were filled with a fluorescent dye. The next step was to package the chips, and for in-vitro testing this was done using 24 pin dual-in-line packages. The chips are glued into the package and gold wire bonds used to

4 !4 connect it to the legs of the package. Then a thick UV curing epoxy was used to cover the wire bonds to protect them from the electrolyte and a small cut section of plastic syringe is added to make a reservoir with a similar volume to the vitreal cavity of the eye. Prototype chips were also prepared for in-vivo testing using rabbits that would be undertaken by researchers at Senju in Japan. The chip is connected to a small PCB which will be implanted into the animal s eye. This is connected to the outside world with a bundle of very thin teflon coated platinum wires which are wrapped up in a PTFE tape. This is quite fragile and so after a short length these are soldered to an intermediate PCB which connects to a more substantial cable. The end of this connects to the instrumentation used to power these wired devices which we ll see later. Slide 36 shows the active end of the in vivo device, a PCB around 6mm long which is coated with the UV curing epoxy for protection of the wiring and to make it more biocompatible for short term implantation use. The devices used in the experiments in Japan were a little more rounded than this image suggests. The idea for the final wireless device with 400 doses was that it would be no larger than this system, basically a cube 5x5x5mm! Electrical Testing: In vitro experiments looking at the dissolution of gold used a standard phosphate buffered saline solution designed to have similar ionic concentrations to fluid conditions inside the body. The most important component of the electrolyte for this effect is the presence of chloride ions. When the correct voltage is applied to the electrode the surface will be oxidised to form a soluble gold complex. If a DC voltage is applied this tends to self passivate blocking further reactions, so cycling the voltage was found to be better for reliable release. The results in slide 37 are from cyclic voltammetry experiments on a gold working electrode with a large external platinum counter electrode and a saturated calomel reference electrode (SCE). The applied voltage is a triangular signal with a voltage change rate of about 5V/s and this is shown here. The resulting current drops with each cycle as the electrode area is reduced as it is dissolved. The cyclic voltammetry plot (slide 37) shows a small peak in the current at around 1.1V where the oxidation is occurring. Slide 38 shows results from reference [1] for gold electrode cyclic voltammetry taken from an implanted device (rat model) using a platinum (pseudo) reference electrode which accounts for the shift in the voltage for the corrosion peak compared to in vitro results vs. SCE. This graph shows that at voltages higher than the corrosion level the electrode can cause hydroliysis of water rather accounting for the high current. This can also passivate the surface preventing dissolution and requiring the voltage to be cycled to a level below the corrosion peak to de-passivate. This is a reduction of the surface that gives a negative current peak. After this the voltage can be ramped up again to dissolve more gold. The researchers found little effect on the corrosion peak height or position from long term (over 40 days) implantation. The initial experiments used a proper saturated calomel reference electrode but the final device would operate in a two electrode system without a proper reference. The cap is the working electrode and the surrounding electrode is a combined counter and reference. However, as it s made of gold this is not a proper reference electrode. The voltages required to open the cavity will vary depending on ionic concentrations and the ph of the electrolyte. There is likely to be a significant offset between the CV results for the gold corrosion peak voltage and the signals required to operate this device without a proper RE. Slide 40 shows some initial results with the on-chip electrodes where a 1 Hz square wave is applied to open the cavity within 30 seconds. Further experiments showed that a 500 Hz signal with a peak to peak voltage of 3 V and a duty cycle of 80% would open the cavities

5 !5 within seconds. The voltage required is significantly higher than the CV results for the corrosion peak vs. the saturated calomel reference or even the in-vivo results from the literature which used a platinum pseudo reference. Slide 44 shows the result of releasing fluorescein into a phosphate buffered saline electrolyte which strongly resembles the conditions in the eye. A 3 V signal at 500 Hz is enough to open the cavity within 5 seconds and uses no more than 90 μw of power. We ll see in lecture 12 that this can easily be supplied over a wireless link. Electrical Control System: A self contained switch box was built to make it easy for the Senju scientists to use our devices without having to work with signal generators etc. It allows any one of the twelve electrodes to be selected using the dial while a simple push button activates the output and sends the correct square wave signal to the selected electrode. An LED lights to show that the output is live. SMART Drug Delivery Back near the start of this lecture it was stated that the specification required a dose to be administered every day for a year and we settled on a total of 400 individual doses for the final wireless system. If we made it in the same way as the prototype this would require 400 wires to connect up the cavities plus at least one more for the on-chip counter electrodes. This is completely unrealistic to produce a reliable wire bonded chip like this. Adding some electronics on the drug delivery chip can provide row and column addressing. Then we need at least 42 connections: 20 for row selection; 20 for column selection; 1 to supply the drive power to open the selected cavity; and one for the counter electrodes. Further reduction in the number of off-chip connections can be achieved by using some demultiplexing to turn an address of the selected cavity into the activation of one row and one column. This could reduce it to around 6 connections though the exact number will depend on the scheme used to address the electrodes. A further complication that was considered was a method to detect the opening of the cavities. This would be required in an implanted system without on-board memory to store the address of the next cavity to be opened. Instead, when powered up the system would test each cavity in the array before opening the first un-opened cavity. In the scheme considered in slide 50, the opening in the passivation over the cap electrode extends to either side and there are two connections to the anode. When the cavity is opened the connection between the two anode contacts will be broken and can be measured as an increase in the resistance. Slide 50 shows the layout of one pixel of a planned system with row and column selection. It requires two pairs of MOS transistors gated by a row line and a column line. When both are selected the two anode connections can be used to test if the cavity is opened or to dissolve the cap and release the dose. Slide 52 shows the full layout for a 20x20 array and suggests how a single cell or single drug reservoir can be activated by selecting one row and one column. Finally I just want to discuss the concept for an autonomous drug delivery system which uses an on-board sensor to determine whether or not a dose of drug should be released when it is powered up. In the ocular drug delivery system designed for glaucoma treatment this might be a pressure sensor to measure the intra-ocular pressure and decide if drug release is required. For full autonomy we would also need to remove the need for an external power source and add on-board generation. This is an exciting area for research at the moment with energy scavenging systems which generate power from movement or things like bio-fuel

6 !6 cells being considered. There are even systems which use a small amount of radioactive material to generate power! References [1] G. Voskerician et al.; In vivo Inflammatory and Wound Healing Effects of Gold Electrode Voltammetry for MEMS Micro-Reservoir Drug Delivery Device, IEEE Transactions on Biomedical Engineering, vol. 51, no. 4, pp , 2004.