As my husband can attest, I complain often that, all of its wonderful qualities aside, my continuous glucose monitor’s sensor is a horrid pain to insert. The incremental data provided by the Minimed sensor I wear is well worth the pain of the semiweekly insertion, but still, I can’t help but wonder when a better solution will come along.
Well, in the July 28, 2010 issue of Science Translational Medicine, Dr. David Gough published results from a series of studies that promise a better method indeed: an implanted continuous glucose sensor with a life-expectancy of more than a year. Such a conveniently long life would be hugely beneficial for diabetics of all types, especially given how helpful continuous glucose monitoring is for diabetes management, and how crucial it is for a successful artificial pancreas system.
The notion of a fully implanted glucose sensor is not a new one; in fact, Dr. Gough first became interested in glucose sensing when he heard about it as an undergraduate in the sixties, after reading about the promise of implanted biosensors that were being studied. Implanting foreign sensors in the body, however, is not a simple process, and so more than forty years later, the problem is still not solved.
There are numerous difficulties surrounding fully implanted sensors. Some are technical: How would an implanted device communicate with the outside world? How can you design a device that is small enough to cause minimal damage and yet large enough to have a battery that can maintain power for an extended period? But many of the technical problems have been solved since the idea of glucose biosensors was first approached. Given the wide array of gadgets and devices that are small, wireless, and long-lasting, the challenges nowadays are less about batteries and transmission, and more about the complications that biology introduces: How do you include enough of an enzyme such that the reaction being measured can continue for long periods of time? How do you prevent cells in the body from interrupting and changing the observed reaction? How do you insert and seal in the sensor so that minimal damage to surrounding tissue is caused? How do you implant something in the body without having the body “wall off” the device, making it impossible to access the blood vessels and tissues that are to be monitored?
This problem of the body “walling off” an implanted device has proved a difficult one to overcome, especially in the field of glucose measurement. When a sensor is implanted in the body, the invasion does not go unnoticed, and the body remodels the surrounding tissue in reaction to the implanted object. This means that the local environment of a sensor will change after the initial implantation as the body resizes cells and moves blood vessels around the device, making the tissue around the sensor less sensitive and less exposed to the potentially harmful object.
The body’s encapsulation of the sensor is such a problem because most existing subcutaneous sensors, Minimed and DexCom included, monitor glucose continuously by taking advantage of chemical process that depends on both oxygen and glucose:
glucose + O2 + H2O –> gluconic acid + H2O2
In other words, glucose, oxygen, and water, when exposed to an enzyme that catalyzes (speeds up) the reaction, are recomposed into gluconic acid (a molecule similar to glucose, but with some oxygen removed) and hydrogen peroxide.
This means that you can figure how much glucose is circulating in a certain location by inserting a set of enzymes coupled with chemical electrodes and measuring the current being passed through the system as oxygen is consumed and hydrogen peroxide is produced. The current created by the transferring of electrons through the system can be used to determine the amount of hydrogen peroxide, and the amount of hydrogen peroxide can then be directly related by the chemical formula above to the amount of glucose.
When a sensor is first inserted, it begins to monitor the current of hydrogen peroxide as glucose and oxygen react in the presence of the enzymes. The first point of calibration relates an absolute measurement of the current to the blood glucose value relayed by a finger-stick measurement. The ratio between the amps of electric current generated by the reaction to the blood glucose value can be assumed to be mostly stable for a period; so if the sensor measures an increased current, it can assume there is an increased amount of hydrogen peroxide, with a proportionally increased amount of glucose, that can then be reported as a proportionally increased blood glucose value.
This reliance on the amount of hydrogen peroxide proves to be a major difficulty with implanted sensors. If the body reconfigures the tissue around the sensor, or begins to “biofoul” the sensor, attacking the foreign object it has detected in the body, then the amount of hydrogen peroxide being produced drastically decreases, and the electrical current decays. Traditional sensors can only assume that the amount of generated hydrogen peroxide has decreased because the amount of glucose has decreased, and so there is a smaller reaction occurring.
Many ways of handling this walling off and biofouling of the sensor have been attempted, but thus far the only commercially viable glucose sensors are short-lived, lasting three to seven days each. In order to circumvent the problem of encapsulation, the wearer is required to replace and rotate the sensors before the body has a chance to reconfigure the tissue around the insertion site. Additionally, frequent calibrations are required to try to keep the stored ratio of current-to-blood glucose accurate as it adjusts to the body’s changes.
Companies like Minimed and DexCom have succeeded with this model, but the short-term sensors are an imperfect solution; frequent finger-stick calibration is required, the sensor must remain visible to the outside world, patients tend to dislike the pain and hassle of sensor replacement, and the process of inserting the sensor often agitates the surrounding tissue, causing a variability and volatility that would not be present with a long-term sensor.
Gough has been working to solve the problems inherent in subcutaneous glucose sensors for several decades, beginning with his thesis work at the University of Utah, and continuing in post-doctoral research at the Joslin Clinic at Harvard. He came to the Bioengineering department of the University of California, San Diego (UCSD) in 1976, and has been researching and developing glucose-monitoring biosensors since then. For the first twenty years of that time, Gough faced an uphill battle. There was minimal funding for type 1 diabetes research to begin with, and his proposals were often met with the sentiment that, while a continuous means of monitoring glucose would be useful, there would likely be a cure by the time he developed the device, so what was the point? In the mid-nineties, though, when the cures were still coming up dry, Gough noted a sea-change in the field, as people realized that until a cure came, we really needed better management and treatment solutions.
In 1998, Gough founded GlySens, Inc. with a graduate of UCSD’s Bioengineering program, Joseph Lucisano, in order to bring his years of research and innovation in the field of glucose sensing to manufacture, and ultimately to market. The recently announced successful pig trials of the implanted sensor, though in many senses only the first step to a viable product for diabetics, has been a long time coming, and makes use of years of Gough and his team’s study of and novel insights into the creation of effective biosensors.
So what makes the GlySens sensor different? How can it overcome, even in pigs, the difficulties that have defeated implanted sensors before it?
The GlySens sensor includes several key innovations that, taken together, give the sensor the ability to reliably and accurately monitor glucose levels for at least a year of implanted time. To begin with, though the sensor relies on the same chemical reaction used in existing continuous glucose sensors in which oxygen and glucose are converted into hydrogen peroxide and gluconic acid, that is only the first half of the GlySens story. Whereas existing subcutaneous sensors measure the amount of hydrogen peroxide that is produced, the GlySens sensor focuses on the amount of oxygen that is consumed. So, the chemical reaction used for existing subcutaneous sensors:
glucose + O2 + H2O –> gluconic acid + H2O2
is paired with its reverse reaction, in which the hydrogen peroxide that was produced, in the presence of another enzyme, breaks down into oxygen and water:
H2O2 –> ½ O2 + H2O
The two ends of the process, then, have the net effect of:
glucose + ½ O2 –> gluconic acid
In other words, glucose, in the presence of oxygen and the appropriate catalysts, yields gluconic acid. So, two enzymes are employed to catalyze each side of the pair– glucose with oxygen, and then hydrogen peroxide back into oxygen and water. Any oxygen that is left over can then be measured.
Now, if the GlySens sensor were parroting the existing subcutaneous sensors, but just using oxygen instead of hydrogen peroxide, there would have to be a finger-stick measurement that would provide a ratio for leftover oxygen to blood glucose. The sensor would then be calibrated, and the ratio would hold for as long as the body didn’t modify the surrounding tissue or attack the sensor too much. The problem, clearly, is that the sensor would be facing the same requirements of the current short-term sensors– frequent calibration and a short lifespan before the measured current drops off.
The real novelty of the GlySens sensor, therefore, is the manner in which it works around these problems: the GlySens sensor does not just rely on absolute oxygen levels measured in the glucose-dependent reaction. Instead, each glucose-monitoring sensor in the system is paired with a separate sensor that monitors oxygen levels alone, independent of the reaction with glucose. That way, the GlySens system can keep track of the base level of oxygen in the surrounding tissue, and does not have to assume that less leftover oxygen necessarily means there was more glucose to react with. The change in the base levels of oxygen can be subtracted out of the change in levels of oxygen found in the glucose-dependent reaction, and only the change in oxygen levels that is unique to the glucose-dependent reaction is used to determine how much the levels of glucose have changed.
So, as the body begins to wall off the implanted device, both the glucose-dependent and the oxygen-reference sensors see a drop in oxygen levels, and the drop observed by each is rightfully attributed to change in the tissue composition rather than change in glucose levels. According to the studies conducted by Gough, the body sufficiently encapsulates the device in about two weeks, after which point oxygen reaches more-or-less stable, non-zero levels. Even after the stabilization of the surrounding tissue, though, the differential, rather than absolute, measurement of oxygen levels proves useful, as variability in oxygen levels caused by body temperature, exercise, sleep, and so on will be reflected in both sensors and therefore will not be falsely attributed to changes in glucose levels.
Within a single GlySens sensor, then, there are actually eight sensing systems– four pairs of glucose-dependent and oxygen-reference sensors, each containing the necessary enzymes, all arranged on an alumina disc. The disc is housed inside a cylinder, 3.4 cm (1.3 in) wide and 1.5 cm (0.6 in) tall. The titanium housing also contains a small battery which will last at least one year, and a wireless transmitter capable of transferring measured data to an external monitoring device every two minutes.
The viability and longevity of the sensor array is aided further by the particular enzyme being used. The enzyme electrodes used to catalyze the reaction are far from standard-issue, and have been specially designed (and patented) by Gough and Lucisano. The enzyme is covered in a semipermeable membrane, that keeps out unwanted molecules circulating in the body which might disturb the reaction of interest. This is especially helpful in preventing corruption of the sensor by potential reactants like acetaminophen (Tylenol), which is known to cause inaccuracies in current continuous glucose monitors. Additionally, only a minimal amount of oxygen can diffuse through the membrane at a time, meaning the reactants are not used up too quickly, and the electrode system can persist without consumption or dissipation for up to a year.
So we have a miniature hockey puck– a container made of biocompatible titanium– packed with membrane-covered enzymes and hermetically sealed. The device is then implanted subcutaneously; in humans, it will likely be placed in the chest or abdomen in an an outpatient procedure. For the studies completed thus far, though, the sensor units were surgically inserted into Yucatan minipigs within the subdermal fat layers, with the sensor surface facing inwards. Thirty sensor units were inserted into six pigs to test the surgical procedure and the body’s response to the implanted device; for the eighteen month period during which the implants were monitored and the surrounding tissue was sampled, no adverse reactions like infection, migration, or rejection were observed.
After the initial set of surgical tests, two pigs were each implanted with two sensor units each. The sensor acceptance and output was then monitored in each of these nondiabetic pigs, for about a year in the first pig and 16 days in the second. While the sensors were implanted, researchers monitored the reported glucose values and compared them to periodic blood glucose measurements taken from central venous samples.
The first thing the researchers found while continuously monitoring the glucose levels of nondiabetic pigs is that pigs have ridiculously stable blood sugar levels. According to Gough, the pigs gorged themselves, exercised, slept– all with almost no variation in glucose levels at all. In order to see how the sensors would work with glucose excursions in nondiabetic pigs, the researchers actually had to administer intravascular glucose tolerance tests– that is, give the pigs straight shots of sugar into the bloodstream. These tolerance tests ensured that sensor accuracy could be tested in both the normal, unchanging state, and in states of rapid glucose fluctuation.
After establishing the viability of the sensor in the nondiabetic pigs, Gough and his team wanted to observe the performance of the devices in the scenario more relevant to the end-goal of glucose monitoring for diabetics. So, they made the two pigs diabetic by giving them streptozotocin, a naturally occurring chemical that selectively destroys beta cells. As you might expect, this gave the researchers quite a bit of glucose fluctuation to monitor. The pigs were kept alive with insulin injections and occasional glucose infusions for six months, during which time Gough was able to collect data from both the implanted sensors and venous blood measurements.
After six months with the diabetic pigs, the trial was ended– but not because of sensor malfunction or failure. The trials were concluded as soon as enough data was collected to prove the concept and show that the sensors worked well in the pigs. After that point, according to Gough, it was hard to justify enormous amount of effort and the drain on resources required just to keep the trial running. (Not surprisingly, diabetic pigs are not very good at self-management, and insulin and glucose treatments were required at all hours just to keep the pigs alive.)
With data from the pigs both in nondiabetic and diabetic states, the glucose measurements from the GlySens sensor could be compared to the blood glucose values taken simultaneously. The sensors, as it turns out, fared well, even in situations of rapid glucose fluctuation. Throughout the duration of implantation in the diabetic pigs, no sensor measurements were far enough off the blood-based measurements to be classified on the Consensus Error Grid as having potentially dangerous clinical consequences; only 0.3% would be classified as presenting medical risk; and only 3.6% would be classified as likely to effect clinical outcome. The remaining 96.1% would be comfortably categorized as having no effect on clinical action or no effect on clinical outcome.
These accuracy measures improve, too, when the analysis takes into account the expected delay between the measurements of the sensor and the blood-based measurements. As with existing continuous glucose sensors that are inserted into the subcutaneous or interstitial tissues, there is a delay in the observed measurements that is caused by the amount of time glucose takes to diffuse outward from the blood vessels into the surrounding tissues. For the GlySens unit, there was an average of 11.8 minutes of delay when glucose levels were rising, and an average of 6.5 minutes of delay when glucose levels were falling. This is comparable to the short-term sensors currently available, and also potentially can be improved with the use of predictive algorithms proposed by Gough.
When taking into account the delay, then, the picture improves even more: a full 99% of sensor measurements would be categorized as having no effect on clinical action or on clinical outcome, and the remaining 1% would be likely to effect clinical outcome, but not to the point of medical risk or danger. (As points of comparison, even without adjusting for delay this is better than the existing short-term sensors, but less accurate according to the Consensus Error Grid than the finger-stick meters like the OneTouch Ultra.)
Accuracy isn’t the only measure of success or failure with an implanted device, though. Fortunately, Gough reports promising results along other metrics as well. One prominent result of the construction of the sensor array, with paired glucose-dependent and oxygen-only sensors, is that the absolute levels of glucose in the tissue can be tracked, not just the ratio of produced current to finger-stick measurements. In other words, calibration and comparison to finger-stick measurements becomes secondary. The first calibration point in the pig studies wasn’t performed until a full 186 days in, and after that, there was only one calibration every ten days– a time period that was chosen not out of necessity, but somewhat arbitrarily, so that the researchers could begin to establish some regular procedure. For those continuous glucose monitor wearers like me who grumble at the fact that multiple daily finger-stick calibrations are still required with the currently available sensors, the promise of such infrequent calibrations is a measure of success indeed.
Further, the device was shown to survive with charge and, importantly, consistent and reliable data for even longer than a year in a number of cases. (One sensor was still plugging along after 520 days!) The enzymes and oxygen required for the chemical reaction were not depleted. Oxygen levels dropped off as expected due to encapsulation of the device by the body, but through the dual-sensor system, the relevant glucose levels could still be monitored accurately for the length of the implantation.
When researchers analyzed tissue samples from around the sensor, they found no evidence of irritation or abnormal tissue remodeling. The tissue around the titanium housing was reconfigured, as expected, but the design of the sensor membranes minimizes the release of irritants like electrical current or hydrogen peroxide. The means of implanting the device by blunt dissection, in which the natural separation of the tissue, rather than cut layers, is used to create a pocket, reduces the potential for damage to blood vessels and the tissue itself.
This all sounds promising; but the study, of course, was just an early trial. If there’s one thing the FDA has decided, it’s that pigs are pigs and humans are different– so what will it take to get an implantable sensor, accurate and FDA approved, for human diabetics like me? The next step for GlySens will be clinical trials in human subjects. GlySens has submitted to the FDA proposals for a pilot study, in which a dozen or so participants will have the device implanted. If the pilot study proves successful, a pivotal study will follow, with a much larger body of participants and multiple locations.
In order to get approval for a one-year implantable device, each study must show at least a full year of successful use; so, at the minimum, final FDA approval for sale is a few years out. Even showing success in initial human trials, though, will be a big boon for GlySens as a company. Thus far, the development of the sensor has been funded mostly by grants from the National Institutes of Health (NIH) and the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), with very little private investment. Despite the fact that Gough himself is confident that the human trials will return similarly promising results to the pig trials (pig subcutaneous tissue, after all, is very similar to human subcutaneous tissue), investors and potential acquirers wait for human data before getting too excited.
It will be at least several years, then, before diabetics like me might get the option of implanting a GlySens sensor; until that point, Gough and his team will continue building and running trials between GlySens and UCSD, and I will continue using the commercially available sensors, semiweekly pain of insertion notwithstanding. Even so, as a direct beneficiary of the research Gough and his team are doing, I am pleased to see the promising results coming from these initial animal studies.