For the third year running, I had the privilege to attend the Medtronic Diabetes Advocate Forum.* Many in attendance have begun to share their opinions and recollections, and our very own Catherine Price has a great overview of the event.
One of the most interesting changes this year was that for the first time, the long-promised Medtronic 530G pump, with low-glucose suspend and the new-and-improved Enlite continuous glucose monitoring (CGM) sensor (Enlite sensor), has finally been approved and is available in the US. Thus, Medtronic was able to give us some details about the new pump and sensor system, and forecasts for future systems moved forward one notch.
I was particularly interested in the Enlite sensor, as I had used its predecessor, the Sof-sensor, for two years before switching to the Dexcom G4 sensor. For a long time, I complained at a high pitch (here and here) about the Sof-sensor, and have found the Dexcom G4 to be worlds better. Thus, I was curious to hear how the Enlite sensor, promised by Medtronic to be a vastly improved system, actually stacked up in their eyes.
The numbers presented surely show that Enlite is an improvement over the previous Medtronic sensor. Whereas the Sof-sensor needle was a 22 gauge behemoth, the Enlite sensor has been slimmed down to a 27 gauge needle for insertion. Further, the needle now fully contains the sensor itself, reducing the drag that was caused by the fact that the Sof-sensor’s needle was U-shaped and only partly covered the sensor. The needle is now shorter, too– reduced from a whopping 17.5 mm to 10.5 mm.
The improvement of the insertion needle is key, but the sensor itself is smaller too; it is 8.5 mm long instead of 14 mm, with the total volume decreased by 69%. The reduced size plus the fact that the sensor is less stiff means that the post-insertion wear is much more comfortable for the user.
The basic strategy for converting the raw sensor readings to blood glucose equivalents is the same: the Enlite sensor still uses “multi-point calibration,” in which the current blood glucose value is calculated based on the last four calibration data points, weighted by time. However, a number of device improvements make the estimate more accurate than that of the Sof-sensor. Both the glucose limiting membrane, which limits the amount of glucose that can reach the sensor at a given time, and the formulation of the glucose oxidase, which catalyzes the reaction that allows the sensor to measure glucose, were redesigned to improve accuracy, consistency, and sensor life.
Cleverly, Medtronic also reordered the electrodes on the Enlite sensor. The sensor contains several electrodes that together allow the device to measure the amount of glucose that is reacting with oxygen. One of the electrodes is a reference electrode, which monitors whether the sensor is still “wet,” meaning inserted in tissue. In the Sof-sensor, the reference electrode was close to the beginning of the sensor. Thus, if the sensor shifted even a little bit outside of the skin, the reference electrode would raise an alarm, claiming the whole sensor had been removed from the body. In the Enlite, on the other hand, the reference electrode is at the end of the sensor, deep in the tissue, and the device can therefore tolerate a little bit more jostling or retraction from the skin without panicking.
Now, all these are indeed improvements, designed to reduce the pain, bleeding, skin trauma, and long healing time associated with the Sof-sensor. However, the real question is: how does the Enlite compare to the Dexcom G4? In terms of accuracy, Medtronic measures the Mean Absolute Relative Difference (MARD). MARD is the average percent difference between the sensor’s estimation of blood glucose and a finger-stick blood glucose value. The MARD for the Sof-sensor was about 19%, and the Enlite sensor is a great improvement at about 13%. This MARD puts the Enlite right on par with the Dexcom G4, which also reports a MARD of about 13%.
Perhaps more telling, several of the invited advocates in the room used the Enlite, and reported liking it (after some initial adjustment to the insertion techniques, which differ from that of the Sof-sensor). That said, several others had tried the Enlite, and decided on the Dexcom G4, finding it to be easier to insert and still more accurate.
One cause for the discrepancies is user experience might be the fact that the Enlite sensors have greater sensor-to-sensor variability than the Dexcom G4 sensors. In other words, like the girl with the curl on her forehead, the best Enlite sensors are very good, but the worst sensors are horrid. Medtronic itself identified the variability in sensor performance as one of the biggest challenges to overcome with the Enlite. and thus it is not a surprise that the consistency of the Dexcom G4 sensors relative to the Medtronic sensors is one of the points that Dexcom representatives are often quick to bring up. When asked what the difference between the two manufacturing processes might be, however, the Medtronic representatives declined to speculate.
Based on the MARD and patient reports, let’s assume for a minute that Medtronic has indeed drawn alongside Dexcom in the sensor race; will they also pull ahead? Can Medtronic create the next best sensor?
Greg Meehan, Vice President of the Continuous Glucose Monitoring group and Medtronic, shared a little about what the future generations of Medtronic sensors would look like. Notably, the goal for Medtronic is an integrated system in which the glucose sensor would feed forward into an insulin pump, and a control system would be able to decide based on sensor readings how much insulin to deliver. Such a system is crucial in progressing towards a fully closed-loop artificial pancreas, but requires that the glucose sensor be accurate and reliable well beyond the current standard.
In order to achieve such accuracy and reliability, Medtronic intends to incorporate several glucose sensors into a single device, such that the glucose monitoring will be redundant, and one bad or failed sensor will not lead to faulty decision making on the part of the control algorithms. This redundancy will be achieved in two ways– first, the existing glucose sensors, which work by measuring the current generated as glucose oxidase (GOx) catalyzes the reaction of glucose in the interstitial fluid with oxygen, will be multiplied. That is, instead of one GOx-based sensor on a single inserted device, there will be four independent sensors all contained within a single filament. Even though the quadruple-sensor will be inserted at a single location, Medtronic has found that each individual sensor unit can maintain a distinct signal, and that if one sensor fails or hits a bad pocket of tissue, the others are usually unaffected. Thus, four separate glucose signals can be monitored, yielding a much more reliable system overall. When one sensor reports inaccurate readings, the system will be able to determine from the other three not to rely on the one bad egg, and if the variation between all four readings is too high, the system will know something has gone wrong and the unit needs to be replaced.
However, even with the increased reliability of a quadruple-sensor, a GOx glucose monitor alone might not be reliable enough serve as the basis for an artificial pancreas. So, Medtronic is also working on an orthogonally redundant monitoring system– that is, a redundant sensor that measures something different than the GOx sensors. Consider, for example, that acetaminophen (Tylenol) could render all four of the quadruple-sensors inaccurate; a reliable artificial pancreas would therefore require a second sensor that didn’t rely on the same reaction as the GOx sensors. To this end, Medtronic acquired the Denmark-based Precisense in 2009. Precisense, now under the purview of Medtronic, is developing an optical glucose sensor. Instead of measuring current generated from a glucose-oxidase-based reaction, the optical sensor would have a fiber-optic cable inserted into the tissue. The sensor would expose glucose in the tissue to a fluorescently-labeled receptor, such that the more glucose that were present, the more fluorescent signal that would be generated. The fiber-optic cable would relay the emitted fluorescent signal to a microsensor that could then calculate the amount of glucose present.
The Precisense optical sensor, therefore, would serve as an independent measure of glucose in the body that could be compared back to the readings from the GOx sensors. Even if any one of the sensors could only achieve a MARD of 13%, Medtronic hopes that the combination of all five will allow cherry-picking the best results at each moment, resulting in a much improved overall accuracy.
This system is under development at Medtronic; a version of the multi-GOx-sensor is currently being used in a continuous glucose monitor Metronic manufactures for hospital settings, but translating that into a wearable, patient-managed device will take time. With only a promise of performance, I wonder– do I believe that Medtronic will be able to innovate in the sensor space, and really engineer a glucose sensor that outperform competitors’ sensors? If Dexcom were making such a claim, I would drink the Kool-Aid. However, with Medtronic, I am inclined to say, “I’ll see it when I believe it.” Though the redundancy promised with the next generation sensors sounds impressive, I have yet to see Medtronic prove its mettle in sensor design and implementation. My hesitation, admittedly, is intensely personal: I wore the Sof-sensor for more than two years, and, given the poor performance of that CGM compared to the Dexcom I now wear, I am once burned, twice shy.
That said, given that more accurate sensors are indeed key to progress with the artificial pancreas, I look forward to Medtronic providing pudding full of proof.
*Disclaimer: Medtronic paid for guests’ transportation, hotel and food, but did not pay for any subsequent coverage of the event. All views are my own.
Karmel Allison is science editor of ASweetLife. She writes the blog Where is My Robot Pancreas?.
Follow Karmel on Twitter (@karmel_a)