Good evening from just south of our home city of San Francisco, CA and the 2013 Diabetes Technology Meeting. Day #1’s “pre-meeting” agenda, the focus of this report, included four workshops touching on CGM, BGM, and insulin delivery.
Two of the most notable talks came on the CGM front from Abbott and Medtronic. Dr. Udo Hoss (Abbott Diabetes Care, Alameda, CA) shared new clinical data (n=62) on Abbott’s 14-day sensor (first introduced at EASD 2013 under the name “Flash Glucose Monitoring”). The 14-day sensor’s MARD was 13.9% using the FreeStyle Navigator algorithm, comparable to a MARD of 14.0% using a simulated factory calibration. The results were less accurate than the 8.5% MARD shown at EASD (n=12), though as we understand it, the hardware and algorithms were different in the two studies. Overall, it was good to see this larger study and that factory calibration may be possible. Meanwhile, Dr. Fran Kaufman provided a most valuable update on the company’s slew of CGM products aimed at reducing hypoglycemia – she shared new clinical data on the company’s redundant glucose oxidase system, which achieved an 11% MARD with 2-4 calibrations per week. Notably, the picture of the system also showed a Bluetooth LE transmitter that could connect with a smartphone, a handheld, and a Bluetooth LE pump. We also saw the first feasibility data from an EU study of Medtronic’s Connected Care device – like Dexcom Share, this has the potential to dramatically change diabetes care from a caregiver perspective.
On the insulin delivery side, Dr. Stuart Weinzimer provided an excellent overview of what he sees as the four major trends in pump therapy: 1) integration and communication; 2) new pumps; 3) improving insulin delivery PK/PD; and 4) application of pumps to broader populations and medications. Dr. Barry Ginsberg (Diabetes Technology Consultants, Wyckoff, NJ) followed with an outstanding overview of patch pumps, noting his belief that patch pumps are likely to replace current pumps. The audience appeared to agree with him.
In BGM, Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA) introduced the new surveillance error grid, intended to better characterize the implications of inaccurate blood glucose readings. The grid uses an intensive methodology, with each blood glucose value from 0 mg/dl to 600 mg/dl assigned a unique risk score based on a survey of 206 HCPs. The final product is a tie-dye grid that will help with post-market surveillance. Dr. Courtney Lias (FDA, Silver Spring, MD) followed his presentation, emphasizing many of the same points from her presentation at DTS’ September 9 meeting – she made clear how active industry will need to be to make the post-market surveillance initiatives work and also reinforced the complexity involved in pushing the movement forward.
Other talks from the day included a valuable discussion from Dr. Lutz Heinemann on why insulin infusion sets are the “Achilles heel” of pump therapy, Dr. Allen King (Diabetes Care Center, Salinas, CA) on how basal insulin needs could be lowered in treat-to-target trials, and a whole session on the cost-effectiveness of diabetes technology. There were numerous great Q&A exchanges and audience response questions on all manner of topics – which hormone to add first to the closed loop after insulin, whether smartphones should be used to control pumps, whether patch pumps will eventually replace durable pumps, and much more.
- Executive Highlights
- Detailed Discussion and Commentary
- Workshop: Using Self-Monitoring of Blood Glucose and Continuous Glucose Monitors to Improve Outcomes
- Performance of Sensors Based on Wired Enzyme Technology over a 14-Day Wear Period
- Panel Discussion
- Sensors: The Future and the Use to Mitigate hypoglycemia
- Benefits of Continuous Glucose Monitoring
- Pattern Analysis to Interpret Continuous Glucose Monitoring
- Benefits of Structured Testing
- Using Blood Glucose Data for Treat-to-Target Therapy
- Panel Discussion
- Workshop: Clinical Applications of Pump and Insulin Delivery Systems
- Workshop: Performance of Blood Glucose Monitors
- Workshop: Diabetes Technology Saves Money and Lives
- The Current Economic Environment for New Diabetes Technology
- Self-Monitoring of Blood Glucose: Impact of Accuracy on Clinical and Economic Outcomes
- Panel Discussion
- Excess Mortality in Type 1 Diabetes: Can Technology Make a Difference?
- Cost-Effectiveness of Glucose Monitoring in People with Diabetes
- Panel Discussion
- Workshop: Using Self-Monitoring of Blood Glucose and Continuous Glucose Monitors to Improve Outcomes
Detailed Discussion and Commentary
Workshop: Using Self-Monitoring of Blood Glucose and Continuous Glucose Monitors to Improve Outcomes
Performance of Sensors Based on Wired Enzyme Technology over a 14-Day Wear Period
Udo Hoss, PhD (Director Sensor Chemistry, Abbott Diabetes Care, Alameda, CA)
Dr. Udo Hoss presented data on Abbott’s research efforts to make a 14-day CGM sensor based on the FreeStyle Navigator wired enzyme technology (the term “Flash Glucose Monitoring” was not used like it was at EASD, though this is the sensor technology underlying that system). Most notable was new data from an outpatient clinical study of 62 patients (seven patients excluded) – data was compared to capillary blood glucose values (n=10,008, translating to ~12 per patient per day). The average sensor wear duration was 12.5 days, 78% of sensors lasted 14 days, and there was no significant change in sensor stability over the two-week period. The rest of Dr. Hoss’ presentation focused on calibration – the million-dollar question when it comes to accuracy. The 14-day sensor’s MARD was 13.9% using the five-day FreeStyle Navigator algorithm (extended by adding one calibration every two days; 10 calibrations per 14 days), comparable to a MARD of 14.0% using a simulated factory calibration applied over the whole 14-day sensor wear period. (We would note that this accuracy was lower than the 8.5% MARD [vs. YSI] reported in a 12-patient study at EASD. This was to be expected, given the comparison to SMBG in this study and the larger number of patients included. Additionally, we understand that the hardware [electronics, transmitter form factor, and inserter] as well as the software [data processing algorithm, including calibration] were different in the two studies.) The FreeStyle Navigator calibration scheme achieved 84% points in Zone A of the Consensus Error Grid, comparable to 83% with simulated factory calibration. Dr. Hoss concluded that 14-day sensing is feasible and that factory calibration “is an option demonstrating comparable accuracy to fingerstick calibration.” We look forward to seeing these results confirmed in a study using real-world, prospective factory calibration. A big key on this front will be manufacturing consistency once sensors are manufactured at scale, something that is hard to tease out in early feasibility studies.
- Dr. Hoss began his talk with an important question, “What is limiting sensor wear time?” He succinctly summarized the answers – 1) stability of sensor chemistry (this can be tested evaluated in vitro); 2) sensor/tissue interaction (biofouling, foreign body response; this requires clinical study); and 3) adhesive (“not the focus of this presentation/study”).
- The sensor technology is a modified version of the original FreeStyle Navigator wired enzyme chemistry. Abbott has reduced the sensor width to 0.3 mm and optimized the sensing chemistry for 14 days of use (“It’s not the same chemistry as has been used in the Navigator system”). Dr. Hoss further noted in Q&A, “We picked a good polymer so it did not induce a foreign body response.”
- This clinical study included 62 patients, 50 with type 1 diabetes and 12 with type 2 diabetes. Seven patients were excluded from the study – three withdrew, one had the receiver out of range for three days (very real world indeed!), two patients had less than 10 pairable capillary BGs, and one patient “had low and not glucose correlated sensor signal” (it was somewhat unclear what the latter meant). One sensor was inserted per patient in the upper arm. Patients were required to take at least eight capillary blood glucose values per day (we assume the meter was the FreeStyle Lite, but it was not specified). The total number of capillary BG reference values was 10,008 (implying ~12 per day per patient). Inspection of the insertion site during and at end of the study showed no signs of unexpected skin reactions.
- The signal stability of the sensor was also evaluated in vitro (n=32 sensors). Sensors were exposed to a constant 17 mM glucose at 33 degrees Celsius over 14 days (a point of contention during Q&A). Pre-drift sensitivity was 0.78 nA/mM, comparable to post-drift sensitivity of 0.76 nA/mM – in other words, a 3% drift over the 14-day in vitro test period.
- This study appears in the September 2013 issue of the Journal of Diabetes Science and Technology. The simulated factory calibration discussed in Dr. Hoss’ talk appears to be a post-hoc analysis that is not included in the study (based on the abstract).
Dr. Saleh Adi (UCSF, San Francisco, CA): In the in vitro studies, you had 14 days at a constant glucose concentration and a constant temperature. If you varied the glucose concentration every hour along with the temperature, would you get the same results?
Dr. Udo Hoss (Abbott Diabetes Care, Alameda, CA): We did not do that. We kept the glucose concentration constant during the 14-day exposure. I don’t believe it would make a difference, but don’t know.
Dr. Ken Ward (Pacific Diabetes, Portland, OR): Typically sensors have greater sensitivity in vitro, and in vivo, biofouling makes it worse. But it looked like you had higher sensitivity in vivo?
Dr. Hoss: The conditions in vitro and in vivo do not match. The oxygen concentration is different.
Dr. Ward: At the very beginning of sensing, you have a run-in period. How long do you wait before you start recording glucose data? Is there inaccuracy in the early period?
Dr. Hoss: We start calculating after one hour. The Navigator calibration algorithm has one calibration. As early as one hour after insertion, and then depending on the data, there might be time when no data is reported during the first ten hours. In the case with one calibration factor, it was reported after one hour all the way through.
Dr. Ward: Did you find a higher ARD after one hour? Or was accuracy similar very early vs. late in the sensor wear?
Dr. Hoss: We did not go into the analysis per hour. Per day it was no different.
Dr. Jeffrey Joseph (Thomas Jefferson University, Philadelphia, PA): Interesting data – congrats. I want to ask about the sensitivity of the sensor after it’s placed in the subcutaneous tissue. You usually see an acute drop in sensitivity and variable sensitivity over time. With this sensor, do you believe the improvement is due to sensor chemistry or work with novel membranes? What are the changes in sensor characteristics?
Dr. Hoss: It’s impossible to predict how stable a sensor is in vivo. Obviously, we followed the literature rules with this sensor – we picked a good polymer so it did not induce the foreign body response.
Dr. Joseph: But did the advance come from the combination of chemistry and the membrane? We’ve always stated that it’s the biofouling and protein deposition that was the problem. Perhaps the chemistry is more of an issue?
Dr. Hoss: The chemistry used in these sensors has been optimized and stabilized for 14-day use. It’s not the same chemistry as has been used in the Navigator system.
Q: Is each glucose oxidase enzyme linked to osmium?
Dr. Hoss: No. It’s not linked to osmium. It’s linked to a polymer matrix, but not directly linked.
Q: How much hydrogen peroxide is produced?
Dr. Hoss: It’s determined by how much oxygen is around. And there’s very little oxygen.
Q: So you’re capturing all the electrons with the osmium?
Dr. Hoss: Yes.
Mr. John Walsh (Advanced Metabolic Care and Research, Escondido, CA)): Chemistry is quite important for the Navigator, as well as the adhesive. With using duct tape as an adhesive, I’ve worn the navigator for 43 days with good accuracy. I haven’t been able to do that with other sensors.
Q: The sensor is not in the blood. If glucose is changing, how do you do the calibration?
Dr. Hoss: Everyone using blood to calibrate a subcutaneous sensor is struggling with that difference. One way is to restrict calibrations when glucose is changing rapidly. Another is moving to factory calibration.
Comment: I use Dexcom. I find that if my glucose doesn’t go over 200, the sensor lasts longer than if it goes high more frequently. Apparently, the current increases sensor use at high levels. By keeping things normal, I can go three weeks with the Dexcom.
Dr. Darrell Wilson (Stanford University, Stanford, CA): So a slow ramp up on insulin titration is better?
Dr. Allen King (Diabetes Care Center, Salinas, CA): Start low, go slow.
Dr. Wilson: Given the postprandial contribution to overall average, what about titrating to A1c or fructosamine – using a more integrated measure instead of just the fasting.
Dr. King: I think the fasting is still needed. I don’t use A1c, since it’s an average. The average time of the dinner meal is 8 pm in patients. In reality, most of the type 2 diabetes patients are eating late at night. They’re eating not much for breakfast, not much for lunch, and then a late dinner that extends until 10 pm at night.
Sensors: The Future and the Use to Mitigate hypoglycemia
Francine Kaufman, MD (Medtronic Diabetes, Northridge, CA)
Dr. Francine Kaufman provided a great overview of all things CGM happening at Medtronic, covering threshold suspend, combined CGM and insulin infusion (the “Duo” set), redundant sensing (glucose oxidase alone and glucose oxidase-optical sensing), and connectivity (“Connected Care” remote monitoring). She shared new clinical data on all these fronts – this included new CGM data for the company’s redundant glucose oxidase system, which demonstrated a 11% MARD with 2-4 calibrations per week. Notably, the picture of that system also showed a Bluetooth LE transmitter that could connect with a smartphone, a handheld, and a Bluetooth LE pump. Dr. Kaufman’s final slide summed up all these products into a single staged goal – “Preventing Hypoglycemia.” The company certainly has a robust and encouraging pipeline and we hope that the recent approval of the MiniMed 530G will accelerate it forward here in the US.
- Dr. Kaufman reviewed the ASPIRE in-home study of threshold suspend. As a reminder, this study appears on page 45-47 of our ADA 2013 report and was published in the NEJM in June. In a sub-analysis, Medtronic found that the threshold suspend feature in ASPIRE in-home significantly reduced (p<0.001) nocturnal hypoglycemia independent of when the last bolus was taken (<2 hrs, 2-4 hrs, >4 hrs) – the greatest reduction in area under the curve occurred when the last bolus was within two hours of a nocturnal hypoglycemia event. Another sub-analysis found that those using threshold suspend had significantly less (p<0.001) hypoglycemia the following morning.
- Dr. Kaufman also highlighted data mined from CareLink to further bolster the ASPIRE in-home study findings. The analysis look at those using the Veo pump from May 1, 2011 to September 15, 2013 – a total of 1.3 million low glucose suspend events and 161,809 two-hour suspend events (12.5%). The reduction in hypoglycemia and A1c data was right on par with that observed in ASPIRE in-home. We took particular note of the benefits of the more accurate Enlite – since the product’s introduction, the reduction in hypoglycemia due to low glucose suspend increased by almost 50%.
- Dr. Kaufman showed a data from a “recent regulatory trial” of the combination Duo sensor and infusion set – Medtronic is looking at a launch in Europe. The product uses a single inserter device and houses both the insulin catheter (Sure-T steel needle) and Enlite CGM sensor under the same adhesive (a minimum of 7-8 mm apart). MARD was 12.5% in this study, much better than the 19% MARD shared in poster #970 at ADA 2013 (see page 77-79 here). Dr. Kaufman emphasized in Q&A that the CGM data is quite stable and does not appear affected by nearby insulin infusion. Patients will undoubtedly be a huge fan of this product, though a challenge will be marrying the discordant six-day Enlite sensor wear time with the two or three-day infusion set wear time. We expect the most viable option is to restrict use of the CGM to three days.
- Medtronic has a redundant glucose oxidase sensor system in development, which could feature communication with a variety of devices. Dr. Kaufman showed a picture of the system that we have not seen previously – the system centers around a glucose sensor with redundant glucose sensing elements (the picture seemed to show three sensors in an array-type arrangement; MARD of 11% with 2-4 calibrations per week) and a transmitter Bluetooth LE transmitter with diagnostic. From there, connections were drawn to a handheld with an integrated meter, a Bluetooth LE pump, and even a patient’s mobile device. This was exciting to see from a device connectivity perspective and perhaps a clear response to Dexcom’s development of the Gen 5 mobile platform. The picture of the “handheld” also implied that this system could be targeted at patients not on pumps.
- Medtronic is also working on an orthogonally redundant CGM (glucose oxidase and optical) in partnership with the Helmsley Charitable Trust and JDRF. Dr. Kaufman noted that data would be presented later during DTM. We last saw new information on this at ATTD 2013 from Dr. Anu Bansal – at the time, miniaturization of the external portion of the sensor was a priority. The electrochemical and optical sensors appeared generally accurate in in-vitro, rat, and dog studies, though the optical sensor was more susceptible to noise.
- Dr. Kaufman showed new data from the first study in Europe of Medtronic’s Connected Care remote monitoring device (“rapid feedback to care partners to help reduce hypoglycemia”). This device will send pump and CGM data to the cloud and smartphones, along with text message alerts to caregivers. The average number of daily number of mobile views was 11. Remote connectivity was present 92% of the time, or 22 hours per day. The average number of daily alarm message texts per user was 5.8, with an average delay of just five minutes. High glucose text alerts were the most common (n=692), followed by low glucose text alerts (n=521), low suspend alerts (n=317), meter blood glucose now alerts (n=39), and predicted low glucose alerts (n=26). The largest number of low glucose suspend and low glucose alerts came from 9 pm – 6am.
Benefits of Continuous Glucose Monitoring
David A. Price, MD (Dexcom, San Diego, CA)
Although he was somewhat preaching to the choir, the enthusiastic Dr. Price made a compelling, commonsense-based case that real-time CGM is beneficial for almost anyone that uses insulin. Underlying his presentation was the theme that both pumps and CGM are simply tools. Ultimately, it’s the patient and the physician that control the quality of glucose control, but there’s no doubt that better tools help. Making decisions about insulin dosing is much easier if direction and rate of change of glucose control is known, which argues for the use of CGM even before pumps. As CGM technology improves the sensor consistency, accuracy, reliability and ease of use, patient trust and usage will also improve.
- Myth #1: An insulin pump is a therapy. Insulin is a therapy, and pumps are just tools for delivering it. The quality of the control is really up to the patient, not the tool, although pumps provide many additional benefits over MDI.
- In fact, it’s better to start patients on a CGM even before pumps, since an understanding of the direction and rate of change of glucose is extremely valuable in improving control. Nonetheless, pumps and CGM work very well together. Whether the data gets sent to the pump or to a receiver, or to a third party (via technology such as Dexcom Share) is up to the patients. Dr. Price also suggested that if patients don’t wake for alerts, then they would be candidates for pumps that use CGM with a threshold suspend.
- Studies of CGM users and ex-users show that people who are going to continue use of CGM trust the data and their ability to use it. Dr. Price emphasized the nature of trust, and the acronym ‘CARE’ (consistency across sensors, accuracy, reliability and ease of use) which drove trust. The studies also mentioned that compared to CGM ex-users, the patients who continued with CGM were additionally confident in their ability to understand and take action on the data, and confident that CGM would improve their quality of life. Dr. Price also presented MARD distributions for the different generation of Dexcom CGM devices, showing how accuracy and consistency improved across the generations, resulting in improved patient trust.
- Myth #2: CGM is a highly technical tool only for people who have late stage diabetes. Since real-time CGM can impact so many aspects of diabetes management (including dose, timing, dealing with meal composition, correction dose, basal adjustments etc.), it makes sense for almost everyone who takes insulin, regardless of the duration and intensity of therapy. In short, CGM informs, motivates and protects.
Pattern Analysis to Interpret Continuous Glucose Monitoring
Frank Schwartz, MD, FACE, (Ohio University, Athens, OH)
Dr. Schwarz’s team has been looking at automated methods for analyzing pump and CGM data (from Medtronic’s CareLink data management system). Their approach has been to use mathematical tools that incorporate the skills of physicians. At this conference, Dr. Schwartz presented three approaches for automated learning from pump and CGM data – interpreting historical glucose data and suggesting improvements, capturing glycemic variability using a new metric based on a physician’s perspective, and employing a tool for predicting glucose values up to an hour ahead.
- Dr. Schwartz presented three approaches for automated learning from pump and CGM data.
- 4DSS (4 Diabetes Support System) is a case-based system that detects glucose control problems and provides suggestions via text. Data is obtained from the CareLink data management system. Patients enter life event notes on their phones, which are also used to improve the system. Problems like ‘premeal hyperglycemia’ or ‘nocturnal hypoglycemia’ are detected across multiple weeks of data and categorized.
- CPGM (Consensus Perceived Glycemic Variability) is a score (from 1-4) that captures glycemic variability. It’s been ‘trained’ via a manual physician assessment of variability from CGM results. The new metric automatically incorporates the physician’s ‘gestalt’ and appears to outperform MAGE, but needs clinical validation.
- SVR (Support Vector Regression) is a tool used to predict glucose values 30 minutes of 60 minutes ahead. The SVR approach is used in financial forecasting, but Dr. Schwartz’ algorithm also includes life-event data to make its predictions. The current algorithm outperforms skilled humans. The hope is that it will be used in closed and open loop situations.
Benefits of Structured Testing
Timothy Bailey, MD (AMCR Institute, Escondido, California)
Dr. Timothy Bailey provided an overview of structured blood glucose testing, noting that it “is essential for all patients with diabetes.” However, he explained that the data is the murkiest for patients with type 2 diabetes not on insulin, with a handful of studies showing both benefits (e.g., STeP, ROSES, St. Carlos) and no effect. Dr. Bailey attributed the conflicting findings to four major study design factors: frequency, timing, use of the data by physicians and patients (very key that it’s used to titrate therapy), and adherence. Notably, he was skeptical of the recent AACE/Endocrine Society recommendation to avoid routine SMBG in type 2 patients not on insulin – “I’m afraid this message is going to be used in ways we may not agree with.” We completely agree that this “Choosing Wisely” recommendation could be used as an excuse for patients not to test at all, as opposed to using a more tactical, actionable, and effective approach with structured testing. Fortunately, it seems that many key opinion leaders agree that the official statement was off the mark. We hope that it will change as more solid evidence emerges about the benefits of structured testing in non-insulin-using patients with type 2 diabetes.
- Dr. Bailey had a novel side-by-side comparison slide to illustrate how (simplistic) structured testing, although easier to interpret by providers, may not address the needs of all patients with diabetes. He highlighted that many patients have stopped charting data, perhaps due to a perception that the data are not reviewed by their physician. He also emphasized that people have complicated lives that do not easily fit into our current (mostly paper) logbooks. Dr. Bailey feels that each patient (depending on type of diabetes and medication) has different testing needs.
- Developing new software tools to process unstructured data (which requires less patient effort to collect) may extend the benefits of structured testers to all people with diabetes. For infrequent testers, software could also suggest optimal test timing and frequency.
Ease of Collection
Ease of Analysis
Using Blood Glucose Data for Treat-to-Target Therapy
Allen King, MD (Diabetes Care Center, Salinas, CA)
Treat to target trials yield a typical on-target dose of 0.6 units/kg of basal insulin, yet two CGM-based studies showed that targets can be reached at much lower doses (0.25 units/kg). Dr. King and his team investigated this finding and were able to show that by moving the evening meal back from 8pm to 6pm, a lower dose of basal insulin was possible – it seems likely that basal insulin is being used to treat the evening meal or that un-noticed hypoglycemia is occurring at night. In another experiment, a higher starting dose of insulin led to increased insulin resistance, yielding a higher final dose that was double of the control group. Finally, Dr. King observed that if a patient has an intrinsically higher variability of fasting plasma glucose (FPG), then the patient is more likely to end up with a higher target dose of insulin. The implications seem to be to use more CGM or test during the night, switch to earlier meals at least occasionally, increase dose gradually (‘start low and go slow’), and understand the variability of FPG to avoid unwarranted dose escalation.
- Across many treat to target (TTT) trials, there seems to be a tight linear correlation between the ending FPG and basal insulin dose. An analysis of multiple TTTs showed a target dose of about 0.6U/kg. However, in shorter TTT studies using CGM, patients were able to titrate basal insulin to ~0.25U/kg at a ~100 mg/dl fasting target. In these trials, there was an early meal. Nonetheless, with a 2-3 times higher dose, we would expect to see much more hypoglycemia.
- A trial of 20 patients previously titrated to target and then tracked with CGM demonstrated that switching to an earlier meal led to nocturnal hypoglycemia. The patients were titrated to target by conventional methods, and asked to eat dinner at 6pm. The subsequent CGM data showed that dinner had moved from 8pm to 6pm and that nocturnal hypoglycemia had significantly increased (from 0 events to 7 events).
- Another study showed that the starting dose of insulin can have a large effect on target dose, because of insulin resistance. Insulin resistance of patients was measured (in response to rapid acting insulin) and a higher starting dose developed for one group based on the data. Another group was titrated conventionally. Although the higher dose patients reached target more quickly, they ended up at double the insulin dose! It seems that the high dose caused insulin resistance.
- It seems that patients with a higher natural variability of fasting plasma glucose (FPG) are more likely to have a much higher target insulin dose. That’s presumably because the dose gets increased with positive fluctuations (per protocol) but not decreased with the negative ones. The ultimate dose seems to be correlated with fasting plasma glucose measured at the beginning of the protocol.
- In conclusion, it seems that many people are taking more insulin than they need – it might even be reduced by as much as 50%! We can address this by ‘starting low and going slow’ (to avoid induced insulin resistance), moving to an earlier meal, using CGM or measuring glucose during the night, and measuring the variability in FPG before starting, to avoid inadvertent dose escalation.
Q: For the Duo device, can you define the physical separation of the cannula?
Dr. Fran Kaufman (Medtronic Diabetes, Northridge, CA): It’s a minimum of 7-8 mm apart. We’re not totally finished with the device. We’ve done four studies with it and are looking to launch in Europe. We have really good stable sensor data with the Enlite.
Q: Are there any wearability challenges
Dr. Kaufman: It’s a Sure-T steel needle with the Enlite sensor. You cannot really rock it. There’s a little bit of a taping issue as well, but it’s not so different from what you’d usually expect.
Dr. Garry Steil (Boston Children’s Hospital, Boston, MA): In the pediatric population of seven years and under, parents have an incredible number of basal rates. What can we do for this population?
Dr. Jordan Pinsker (Tripler Army Medical Center, Honolulu, Hawaii): For pediatrics and adults, it makes no sense in a pump to have more than five to six basal rates. It takes a few hours for a change to equilibrate. I do agree that there are patients who have too many basal rates. One of the things you can offer to these patients is to talk about and show them the CGM devices. The commercially available devices in the US have different pros and cons. To some patients, having a separate receiver in another room is a tremendous advantage. To others, low glucose suspend is an advantage. It’s about individualizing.
Q: Lots of patients say they cannot live without CGM. Other patients aren’t on CGM. What are the barriers for patients and physicians? Is it economic? What can we do to take maximal advantage?
Dr. David Price (Dexcom, San Diego, CA): CGM has been an early stage technology. There have been significant reimbursement hurdles, but the landscape is changing daily. There were people one year ago who could not get CGM, and they may now get it. Early CGM did not perform well, and that was from any company – early Dexcom and early Medtronic. If you tried CGM and it didn’t work well, you would go back to the clinician. It discourages them from using it and recommending it subsequently. What we’re seeing with our current CGM is it’s more consistent and better, and a lot more clinicians are prescribing CGM a lot more. The uptake is pretty dramatic right now, both in terms of people prescribing a lot and the numbers of different clinicians that are prescribing.
Dr. Schwartz: Some early adopter patients were calibrating too frequently and when the glucose was not stable.
Dr. Kaufman: Having the CGM on all day would be ideal, but there is also some benefit in pediatrics to using it during illness, for changes in activity, etc. Once patients get used to it, there is an opportunity to use it more.
Q: What about physician data overload. How can we deal with all that data?
Dr. Schwartz: Automated data analysis is the key for the busy-practicing physician. You have to divide and conquer.
Dr. Garry Steil: The stock market is notoriously hard to predict. Is glucose any easier to predict?
Dr. Schwartz: It’s the same with glucose. If you don’t know the life event data, it really is very inaccurate. It’s amazing how accurate it is with experienced clinicians if they know that life information.
Dr. Steil: But if that data is not automatically acquired, it increases the burden on the patients.
Dr. Schwartz: Yes. For type 1 diabetes, I’ve heard an estimate that it takes up to three hours per day of management time.
Dr. Kaufman: One of my adolescent patients suggested we keep a camera on him all day.
Q: What about CGM in surgical patients?
Dr. Price: It can be used however clinicians want to use it. Dexcom is labeled for single patient use, and really for home use.
Dr. Schwartz: There are six or eight published papers of CGM use in the hospital setting.
Dr. Kaufman: In development are systems for the ICU. Medtronic has a system outside the US.
Workshop: Clinical Applications of Pump and Insulin Delivery Systems
Trends in Insulin Pump Therapy
Stuart Weinzimer, MD (Yale University, New Haven, CT)
Dr. Stuart Weinzimer provided a great overview of what he sees as the four major trends in pump therapy: 1) integration and communication (CGM and Internet/the Cloud); 2) new pumps (patch pumps, miniaturization, dual chamber pumps); 3) improving insulin delivery pharmacokinetics/dynamics; and 4) application of pumps to broader populations beyond type 1 diabetes and medications beyond insulin. He focused his talk on the latter two topics, first addressing insulin PK/PD using Halozyme’s hyaluronidase (very positive data from the first patient in the Yale closed-loop study that started recently) and InsuLine’s InsuPatch. Dr. Weinzimer next showed data suggesting benefits of pumps for type 2 diabetes (to date, only showed in cohort studies) and in patients with cystic-fibrosis-related diabetes. Dr. Weinzimer concluded with a brief review of pumping glucagon and pramlintide, which have both demonstrated glycemic benefits in several studies. The first audience response question was particularly notable – a striking 61% of attendees felt insulin pump “connectivity to other devices” (e.g., computers, smartphone, cloud) is the most important area for future development (the question purposefully excluded CGM integration).
- Dr. Weinzimer showed unpublished data from the first patient in Yale’s closed-loop study using Halozyme’s hyaluronidase (ClinicalTrials.gov Identifier: NCT01945099). Estimating from the chart, maximal blood glucose change was +100 mg/dl with hyaluronidase vs. +130 mg/dl with insulin only. In addition, the there was a much flatter and more normal return to baseline when hyaluronidase was add to closed-loop control. These are encouraging results and we very much hope they are replicated in the 20-patient trial.
- Insuline’s InsuPatch has been tested in one patient in another Yale closed-loop study – the benefit on glucose control was quite modest, though it’s pretty early on to draw any definitive conclusions. In a previous study of the warming device, insulin peaked in 77 minutes with the device vs. 111 minutes without it.
- While RCTs have not shown a benefit to using insulin pumps in type 2 diabetes, several cohort studies have shown a benefit. There was no advantage of pump therapy over MDI in RCTs from Raskin et al. (Diabetes Care 2003) and Herman et al., (Diabetes Care 2005). However, six cohort studies have shown an A1c benefit with use of pumps in type 2 diabetes: Kesavadev 2009 (-0.5% from a baseline of 8.1%); Edelman 2010 (-1.2% from 8.4%); Reznik 2010 (-1.8% from 9.3%); Courreges 2012 (-1.1% from 8.8%); Charras 2012 (-1.4% from 9.4%); and Leinung 2013 (-1.1% from 8.8%). The Charras study has an impressive 12-year follow-up, while the Reznik study has a five-year follow-up.
- Citing Drs. Ed Damiano and Ken Ward’s work, Dr. Weinzimer stated, “ There is a real benefit of adding glucagon.” However, he noted the challenges of developing a stabilized glucagon, which must be solved before this is is broadly applicable. In Q&A, Dr. Weinzimer asserted that glucagon was not absolutely necessary for closed-loop control.
- Dr. Weinzimer also discussed the use of pramlintide in addition to insulin, which can mitigate glycemic spikes after meals (Huffman et al., Endocr Pract 2009). He stated, “There might be some long term or larger clinical benefits to having basal infusion of pramlintide.”
- ARS #1. What is the most important area for development in insulin pump therapy besides CGM integration?
- Patch miniaturization: 7%
- Implantable: 12%
- Multi-chamber pump: 7%
- Connectivity to other devices (computers, smartphones, cloud): 61%
Avoiding Infusion Set Problems
Lutz Heinemann, PhD (Science & Co., San Diego, CA)
Dr. Heinemann believes that insulin infusion sets (IIS) are the Achilles heel of pump therapy. In this presentation he touched on a very deep mine of data and experience to point out that not enough general attention is given to the design and use of sets. He showed high level results of his survey of over a thousand German pump users, who reported many IIS problems per year. Some were minor acute problems (like tugging on the tubing), and some were more significant such as skin reaction, infection or inflammation. However, the number of set problems didn’t seem to influence overall glucose control. Dr. Heinemann put forward some valuable techniques to use IIS optimally. These included rotating sites, disinfecting the skin, optimizing the insertion and adherence, introducing an anchor for the tubing, and keeping wear time to 2-3 days. In a call to action, he suggested that set problems are underappreciated, that we need more data, and that there should be more widespread evaluation of sets pre- and post-market.
- There are many different insulin infusion sets (IIS), and they have many different designs. Design aspects include the catheter material, angle of insertion, manual or automated insertion. There is also de facto only one manufacturer (Unomedical, Denmark).
- Issues with infusion sets are the primary reason for people discontinuing pump therapy. Issues include acute problems (such as pullout, blockages, insulin leak along skin) or chronic issues (skin reaction, infection, inflammation).
- When a patient first changes their set, average glucose drops by 30 mg/dl, and gradually increases (by the same 30 mg/dl) over the next 5-6 days. This is considered to result from performance changes of the infusion set (although there might be other explanations).
- The PumpDiab survey (on the reality of pump therapy) showed that the vast majority of patients have more than one issue with infusion sets each year (a significant cohort had more than ten issues). (From the slide our rough estimate is that the median number of issues was roughly four per year.) The data was extensive, and there were some interesting takeaways. In Germany 39% of pump users used a steel needle and only 36% used an inserter. Most patients had used the same set for over five years, while the average duration on pump was about eight years.
- It’s important to note that there was no significant difference in glucose control between groups of patients with few or many set problems.
- Dr. Heinemann put forward some techniques to use IIS optimally. These included rotating sites, disinfecting the skin, optimizing the insertion and adherence, introducing an anchor for the tubing, and keeping wear time to 2-3 days. Just because some patients can wear sets for a long time, it doesn’t mean they are getting better control. Pumps can detect occlusion (for which there is not yet a good solution) but they should also detect other set problems.
- In a call for action, Dr. Heinemann suggested that there were many areas of focus for improving current sets, but he didn’t dwell on details. He noted that in some experiments, the insulin infusion pressure behaved highly irregularly, suggesting that there are some problem factors that need investigation. More information should be made obtainable about set performance and problems in different countries, users should be more involved in set development, there should be more evaluation pre -and post-market and clinical researchers should focus more attention on the topic.
Dr. Yogish Kudva (Mayo Clinic, Rochester, MN): For multi hormonal infusion, if you had to choose between pramlintide and glucagon, which would you choose first to become the next step?
Dr. Stuart Weinzimer (Yale University, New Haven, CT): Dr. Boris Kovatchev asked me that at the NIH meeting. [Pauses] I would choose glucagon for now, because of the obvious benefit for hypoglycemia reduction. But I’m not convinced that we can do that easily and safely until there are more stable glucagon formulations. I also don’t think either second hormone is needed for a completely functioning artificial pancreas. We can do it with insulin alone.
Dr. Kudva: If you would choose glucagon, for that reason, does the availability of low glucose suspend change the game of it?
Dr. Weinzimer: Glucagon will not save you from all hypoglycemia. Even in Drs. Russell and Damiano’s paper, they had a fair number of hypoglycemia episodes that glucagon could not pull patients out of – particularly related to exercise. We shouldn’t think that by slapping glucagon on someone we can prevent hypoglycemia.
Q: Counting carbs is a challenge for many patients. Does adding fat add another challenge?
Dr. Howard Wolpert (Joslin Diabetes Center, Boston, MA): Ultimately, what it’s going to come to is a more structured approach to nutrition recommendations. People are not eating a wide range of different foods – it’s usually a few. Most patients know from experience how to bolus for different meals and food types. It will be about customizing particular boluses for particular meals for particular individuals. Patients need to learn what they need to take for different foods. Everyone talks about carb counting – but if you speak to most patients, very few people do it. We have to face the reality and approach this from what’s realistic.
Q: You said that when you fed individuals the same meal – one high fat and one low fat – there was considerable variability between patients. Did you try to explain that through BMI, muscle mass, or the gut microbiome’s role in metabolism?
Dr. Wolpert: The only factor that came out was total daily dose. Those who needed more insulin were using more insulin at baseline. That’s only valuable for hypothesis generating. We’re actually selecting a group for another study with a wide array of insulin requirements. We need to be able to phenotype people and come up with dosing recommendations.
Barry Ginsberg, MD, PhD (Diabetes Technology Consultants, Wyckoff, NJ)
Dr. Barry Ginsberg provided a whirlwind overview of patch pumps, splitting his presentation into full-featured pumps (Insulet OmniPod, Roche Solo, Debiotech Jewel) and simplified patch pumps (Valeritas V-Go, Calibra Finesse, and CeQur PaQ). He devoted a few slides to each, highlighting the design features, costs, and pumping mechanisms. Overall, he concluded that patch pumps are likely to replace current insulin pumps due to lower cost, convenience, and discretion. The audience was optimistic on this question as well: nearly two-thirds thought sophisticated patch pumps would “definitely” (23%) or “probably” (32%) replace current insulin pumps (16% voted “unlikely,” 16% voted “no,” and 14% were “not sure”). Dr. Ginsberg was less enthusiastic about the data to support use of pumps for type 2 diabetes – he noted that these simplified pumps are still fairly primitive, expensive vs. syringes/pens, limited in their basal capabilities, and have shown limited clinical advantage. However, he believes they will get better over time. We think so as well.
- Insulet OmniPod: Dr. Ginsberg highlighted the much smaller size of the new pod relative to the previous generation. He also noted the 85-200 unit reservoir, described how the pod’s delivery mechanism, and mentioned the ~$4,000 per year usage cost.
- Roche Solo: The FDA-cleared Solo is “similar in size” to the first-gen OmniPod, though its advantage over the pod is the ability to bolus without the controller. Dr. Ginsberg highlighted the pump’s unique three-piece design, which consists of a cradle with the catheter, a reservoir, and the pump (lasts three months). The reservoir and pump can be removed for showering, etc.
- Debiotech Jewel pump: Out of the gate, Dr. Ginsberg made it clear that the Jewel is not approved in the EU or US, though the company is “seeking” approval. He listed several attractive design features of this pump: smaller size than the Insulet OmniPod and Roche Solo, the ability to bolus without the controller a “big” 450-unit reservoir, and an Android smartphone controller. Though the FDA is “very worried about Android phones,” Debiotech has taken a unique approach – the Jewel controller has two SIM cards, one for the pump and one for the phone. A patient can choose whether to use the phone card or the SIM card, but both cannot be used at the same time.
- Valeritas V-Go: This is the “first and only” simplified patch pump on the market. It is smaller than the OmniPod and Solo and bolusing requires a two-unit double button push (for safety). Dr. Ginsberg emphasized that is expensive (~$210 per month), but “interestingly” it is reimbursed – apparently, the company used Lantus as the comparator to demonstrate that the V-Go is a cheaper alternative (about $100 cheaper than Lantus, according to Dr. Ginsberg). One of the major flaws of the device – similar to CeQur’s PaQ and J&J’s Finesse – is there is no counter of how much insulin has been given.
- J&J/Calibra Finesse: The Finesse is “very small,” but “only designed as a pen replacement.” Dr. Ginsberg pointed out the same limitation as with the V-Go – no way to count how much insulin has been given. He noted that the Finesse’s buttons lock up when the device is clogged. In a 38-patient, six-week crossover study vs. MDI, those using the Finesse had equivalent glycemic control (the primary endpoint), though perhaps less glycemic variability (not statistically significant).
- CeQur PaQ. The CE Marked PaQ consists of two devices an infuser and messenger. The three-day reservoir three days, holds 330 units, much bigger than the Finesse and V-Go. The device also offers seven different basal options. Like the V-Go and Finesse, it does not have a dose counter. In a pilot clinical trial (presented at ADA), patients using the PaQ experienced a slight decease in blood glucose, a small decrease in CGM, and a small decrease in insulin dose.
- SFC Fluidics: Dr. Ginsberg is currently this company’s Consulting Medical Director. SFC Fluidics is developing a new pump mechanism – it utilizes a patented non-mechanical means for precisely moving very small amounts of fluid. Low voltage or current drives a pumping fluid across a selective membrane. This, in turn, causes an elastic diaphragm to expand and push a controlled amount of fluid. The advantage is a very compact pumping mechanism, highly accurate delivery, a 30-fold range of basal rates, and a “true pulseless basal.” Dr. Ginsberg believes it will “probably be a three-day device,” though he commented that it is several years away.
- ARS #2. Regulators are concerned about use of smartphones with insulin delivery devices. Which of the following do you believe?
- Smartphones should never be used in insulin delivery: 6%
- It would be ok to use tablets or other smart devices that have WiFi or Bluetooth, but no phone connections: 6%
- It is ok to use smartphone with dual SIMs so that either the phone. would be active or the controller, but not both at the same time: 40%
- Current phones are ok to control devices: 49%
Dr. Stuart Weinzimer (Yale University, New Haven, CT): On the order of CSII vs. CGM, I can only speak to the fact that in our clinic, 80% of our patients are on pumps. There aren’t many patients on CGM but not a pump. The granularity of information you get out of the sensor is almost lost if you don’t use it with a pump. It almost always starts with pumps first. Except when trying to show people not on a pump the advantage to using one.
Dr. Ken Ward: It seems to me that for patch pumps, for a certain build, you might not be able to see the patch pump on you. You could advance the dosing by feel, but having a dose counter on the pump itself might be hard. That’s when there’s an advantage to pumps like the one with the Android phone controller.
Dr. Barry Ginsberg: I agree. Mechanical pumps are not that inexpensive. Take Valeritas – it’s a one-day pump and costs $8 per day. That means a three-day pump is $24. That’s less expensive than OmniPod, but not way less expensive. You might wind up with electronic pumps with fewer features but more capabilities than current simplified pumps.
Dr. Ward: You showed different motive forces that push insulin in – the bladder, etc. Are some more accurate than others?
Dr. Ginsberg: That has not been looked at.
Dr. Saleh Adi (UCSF, San Francisco, CA): I want to add to the comment from Stu. I completely agree. In our practice, about 70-75% of patients are on a pump. You’re right – it’s the other 20% who are not on a pump. We can convince about half of them to go on a pump by putting them on CGM. Everything that is measured in studies is based on A1c level – mean blood glucose, doesn’t tell you the real variability. I want to caution and remind all of us that glycemic variability is going to be more important than A1c. Pump therapy reduces glycemic variability and that’s the one thing we should be looking at. Also quality of life. With pump therapy, quality of life is much better.
Dr. Hsin-Chieh Yeh (Johns Hopkins University, Baltimore, MD): I appreciate the comment. There is not a lot of good research documenting quality of life on pumps vs. MDI. But you consistently see better quality of life in the pump group and the CGM group. I appreciate your comments about choosing different measures beyond A1c. It’s just one of the many measures.
Dr. Adi: [Jokingly] In our practice, anyone with a fasting glucose over 125 mg/dl or a random glucose over 200 mg/dl qualifies for a pump or CGM.
Dr. Weinzimer: We can bring that up in healtheconomics talk.
Dr. Kudva: In the meta-analysis you did, it included studies over a long period of time. There have been lots of changes with the technologies that have been used. How do you factor that in? These meta-analyses will be looked at by payers. The second issue is that when you have single-center, small studies, there are different endpoints and different education techniques. In a large study, you involve multiple centers and have a lot more standardization. STAR-3 stands out vs. other studies.
Dr. Yeh: In older studies vs. newer studies, it’s not an issue in CGM – most of the studies were published after 1990. [Editor’s Note: Most everyone in the room seemed to think this comment was far off base.] In CSII, some of the studies were from years ago. We took a more conservative approach and included a smaller number of studies.
Dr. Kudva: even with real-time CGM, every generation improves. To my knowledge, real-time CGM has only been around since 2004; it’s a rapidly changing field. Conventional meta-analytic techniques have limitations.
Dr. Yeh: We excluded retrospective CGM. In the end, there were not a lot of studies. If we start to look at subgroup analysis by time and device, there are not enough studies. Any meta-analysis provides combined estimates based on published data. It’s one type of evidence.
Q: Dr. Heinemann showed problems with insulin infusion sets over time. Do patch pumps have the same problem?
Dr. Ginsberg: Of the patch pumps, only the OmniPod is on the market. There is some data suggesting as many problems with the catheter on the pod as other pumps, and maybe more.
Q: You didn’t think there was much of a price difference between traditional electronic pumps and patch pumps. Electronic pumps are $4,000 per year, while you said the Valeritas V-Go was $200 per month. That’s about 40% less expensive.
Dr. Ginsberg: What I was suggesting was that the electronic pumps can develop simpler models. When it’s 10 times cheaper, it’s hard to develop comparable stuff. When it’s 40% cheaper, it’s not that hard.
Q: From your experience, what are the criteria for type 2 patients and pumps?
Dr. Ginsberg: I don’t think there are yet. Things have been on the market for less than a year – only the Valeritas of mechanical pump. The other two are still going through clearance. People have not figured out what the appropriate criteria are. I’m a big believe that patients with type 1 diabetes or type 2 diabetes should be on intensive insulin therapy. That means a basal rate and boluses before each meal. If the easiest way to do that is on one of these pumps, they should be on one of these pumps.
Ms. Aimee Jose (Palo Alto Medical Foundation): On the question of what should come first, CGM or CSII, I really challenge everybody here. I’ve been initiating real-time CGM prior to pump therapy with many pump candidates. I’ve been doing it for a while, and I’ve gotten better at doing it more recently. You see the benefits of CGM before even thinking about pump therapy. From a behavioral standpoint, transitioning from MDI to CSII is crazy. It can take months for people to get comfortable with pump. If you’re on CGM, the transition to pump therapy can happen in one week. You cannot get by doing basal testing without CGM. I would challenge all of you to think about reversing the process and trusting your patients in learning how to interpret CGM and making the transition to the pump.
Dr. Yeh: There are no studies to support which should come first. There was a 16-patient study in Spain, which demonstrated that real-time CGM first before CSII can increase adherence to CGM. But there was no difference in A1c or quality of life. We need to encourage people to design better studies to see what comes first.
Workshop: Performance of Blood Glucose Monitors
The New Error Grid
David Klonoff, MD (Mills-Peninsula Health Services, San Mateo, CA)
Dr. David Klonoff presented the much-awaited new color-coded surveillance error grid, the Diabetes Technology Society’s (DTS) alternative to the currently used Clarke and Parkes error grids. He highlighted the underlying goal behind the development of this new error grid: to more accurately represent the clinical accuracy of blood glucose meters, “to create a modern era grid,” and “make a grid to show how diabetes is treated in practice.” Unlike the Clarke and Parkes error grids, the new surveillance error grid has five different color-coded risk ranks that depict the risk of an erroneous reading from 0 mg/dl - 600 mg/dl. The new grid is based on the average of 206 HCP risk rankings from a survey conducted by DTS. The grid has a blended tie-dyed look, a sharp contrast from the current stark boundaries of the Clarke and Parkes error grids. To develop the new tool, DTS surveyed HCPs on which of five actions they would take for patients facing particular blood glucose levels ranging from 0 mg/dl - 600 mg/dl. Survey takers were also asked about the risk associated with an incorrect blood glucose reading. Dr. Klonoff and Dr. Lias emphasized that this grid will be used for post-market surveillance of BGM quality – it was very clear that it will NOT be used for pre-market evaluation of devices.
- To develop the new error grid, HCPs ranked which of five treatment actions they would take with two different patients. The actions were: 1) emergency treatment for low blood glucose; 2) take oral treatment for low glucose; 3) no action needed; 4) take insulin; and 5) emergency treatment for high blood glucose. These designations served to create boundaries between treatments options.
- HCPs were given two of four patient profiles to access risk for: 1) a patient with type 1 diabetes using an insulin pump; 2) a patient with type 2 diabetes using insulin; 3) a patient with type 2 diabetes not using insulin; and 4) a type 1 patient using MDI and a CGM. None of the treatments were significantly different between these different types of patients.
- After HCPs assigned treatments to the glycemic range, they received a 5 X 5 grid and assigned risk values for treatment recommendations from the meter reading vs. treatment recommendations from the HCP. HCPs chose one of nine risk values: 1) none; 2) mild for developing hypoglycemia; 3) moderate for developing hypoglycemia; 4) great for developing hypoglycemia; 5) extreme for developing hypoglycemia; 6) mild for developing hyperglycemia; 7) moderate for developing hyperglycemia 8) great for developing hyperglycemia; and 9) extreme for developing hyperglycemia. Once the HCPs gave risks assignments, scores were assigned to each blood glucose level based on an eight-point scale (negative four to four), and the scores were averaged across HCPs. This gave a unique risk score for every point on the 360,000-point grid.
- Risk for hypoglycemia was given a score from -0.5 to -4, and risk for hyperglycemia was given a score from 0.5 to 4. “No risk” had an absolute value 0.0-0.5, “slight risk” ranged from >0.5-<1.5,” moderate risk” ranged from >1.5-<2.5, “great risk” ranged from >3.5-<3.5, and “extreme risk” ranged from >3.5-<4. These absolute scores were each assigned a color, which was mapped onto the grid. Dr. Klonoff noted that the differentiation between absolute values for risk was arbitrary, and the group is currently working to develop more fine-grained differentiation between “slight risk” and “moderate risk.”
- One error grid is effective for different types of clinicians, locations, practices, and patients. Dr. Klonoff noted that the results from the survey differentiated HCPs along the above four lines, and the research group found that there was no difference in the boundaries of treatment. The only slight discrepancy was location (in or out of the US); however, the group eventually decided that the differences were not significantly different enough (and the number of European diabetes experts was too small) to justify developing a second risk chart.
- Software can be used to find the percentage of meter readings between any two risk scores. Dr. Klonoff noted that this could be extremely helpful when determining the clinical significance of out-of-range tests; instead of having to visually determine where the test falls, computers will do this for the HCP.
FDA Postmarket Surveillence of Glucose Monitor systems
Courtney Lias, PhD (FDA, Silver Spring, Maryland)
Dr. Courtney Lias provided an overview of current FDA postmarket surveillance of blood glucose monitors, including where she believes DTS’ Surveillance Program could further improve this process. While we were pleased to hear Dr. Lias speak on the program that was originally proposed at the September 9 DTS meeting (read our report), we were disappointed that the majority of the information was the same. Looking at steps that the FDA has taken, Dr. Lias remarked that the FDA is working on internal solutions to improve how the organization deals with adverse event reporting for devices; according to Dr. Lias, most of the vast data that the FDA receives is low quality. However, the agency has improved its IT to get the most information out of what they are given. Additionally, the FDA is also working on long-term solutions to improve the standardization of reporting among companies. Turning to inspections, Dr. Lias acknowledged that postmarket surveillance is critical, as the accuracy of strips can change due to so many variables (e.g., the enzyme, a different strip coating procedure). Although the FDA currently does routine inspections and investigates complaints, Dr. Lias noted that the FDA is very supportive of a postmarket surveillance program. We hope to see this a postmarket surveillance program move along – the September 9 meeting wrapped up with a clear next action step: DTS would move forward with surveillance program development, starting with convening an expert steering committee to hammer out specifics.
- Dr. Lias highlighted that the new surveillance error grid is a great tool to talk with industry and make sure that everyone is on the same page. Earlier in the morning, we heard from Dr. Klonoff about the design of the new grid, which was developed to assess clinical risk of blood glucose meters. The grid was developed through a collaboration between scientists and FDA members. Hopefully, this signals that the Agency is receptive to adopting it.
- Dr. Lias discussed DTS surveillance program, highlighting many of the same points she made at the September 9 meeting. Specifically, she noted that although a surveillance program may discourage detrimental manufacturing issues, it will not identify the sources of an issue. She also raised many good questions, such as whether or not this program would be mandatory or voluntary. For more detailed discussion of the FDA’s views on a postmarket surveillance program, please see page eight of our September 9 report here.
Q: Do we have a best comparator method?
Dr. David Sacks (NIH, Bethesda, MD): We haven’t selected the best method yet. We are still evaluating that, and it’s not clear that there will be only one way.
Q: I work with prediabetes screening, and the healthcare system where I work is terrible. There are multiple counties without health departments. I have to screen for prediabetes in Latino and Native American populations. We just don’t know if the accuracy of these meters is good enough for screening. I don’t have access to A1c on regular basis. How can we do a real world test and determine if the accuracy of meters is reliable for prediabetes diagnosis? I think that there is an opportunity for the FDA to step forward and work together with the CDC on a solution. I challenge the group to think about that, and to also think about screening group capabilities.
Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA): You raised some very good points. Whichever method you use for screening has to be FDA approved for that purpose. Recently, the first A1c assay for diagnosis was approved, and the way it happened was that the FDA said it was approved but they have to conduct regular surveillance.
Q: All these grids do not take into considerations the frequency of measurements by users, as well as the re-use of the device. When you are doing all these grids and testing, even pre-market, you are testing them in lab conditions. In reality, the users don’t wash their hands, they use their first drop, and that dramatically impacts the results. All of us are familiar with the fact that there is often a difference between trials and real-world results.
Dr. Klonoff: You make some interesting points that I will try to address. It wasn’t our intention to resemble one or the other. In one regard, we resemble Parkes’ grid because we don’t have blocks that skip around, however, our intention was to create a modern era grid. We wanted to make a grid to show how diabetes is treated in practice and to show that the closer you are to analytical accuracy, the less the clinical risk. This graph also shows the diverse readings. You make a point about patients who test themselves with bananas on their fingers, but this grid is more intended to access clinical risk. If a meter is reading differently than the real measurement, then there is a discrepancy in treatment.
Q: About the error grid, you’ve done lots of work with the methodology. Have clinicians been consulted? And what about other types of diabetes such as gestational diabetes? Should the grid for them look different?
Dr. Klonoff: I recognize several clinicians in this room that participated in the process, but the panel decided that we wouldn’t want to publicize the names of the clinician respondents. As far as gestational diabetes, we couldn’t look at everything. We thought even our four scenarios was quite a load. We think that gestational diabetes patients might fit most closely with type 2 diabetes, so they could hopefully use that grid.
Q: Because the room for error becomes so much smaller with the new ISO standards, it becomes more important to have a good reference. When you use the hexokinase method and mixed samples, you are probably sure this method is correct good for assessing the trueness of devices.
Ms. Grete Monsen (NOKLUS, Bergen, Norway): We believe the hexokinase method is the best method. However, we call it the comparison method, not the reference. It is our job to document that it gives true values. We use serum values, not accurate solutions, and then we have controls from a reference method to verify the true values. So we use hexokinase in all of our operations.
Dr. Klonoff: I’d like to make a comment to the industry. It’s very impressive how the performance of these products has improved. Some may feel that it has not improved enough, but that’s how technology works. You guys have done a great job: the products are better, and the problems are much smaller. Everyone is helping and its good for the entire industry.
Dr. Howard Zisser (Sansum Diabetes Research Institute, Santa Barbara, CA): My first question has to do with the title of the error grid: surveillance. It sounds like it is on the back-end of things. What kind of output do you expect this grid to have? If you have a device that meets ISO standards, what will that look like in analytical numbers, and how will that differ from someone who does not meet ISO standards?
Dr. Klonoff: I am sure that those data would be in line with the dark green. In general, the better analytical accuracy, the less risk; however, it is not always linear. Going back to your other question of why we titled it “surveillance.” When we started, we asked what it would be useful for us to do. The FDA did not consider error grids useful for premarket, but, instead, it found it useful for postmarket. I think industry could use this internally to assess how their products are working, but my understanding is that if industry uses this grid and takes the results showing good clinical accuracy to the FDA, it would not add strength to the application.
Dr. Courtney Lias (FDA, Silver Spring, MD): What came up during discussions was that we don’t need new tools for premarket assessment of glucose data. Industry representatives on that group were opposed to an error grid that would be used for approval. We ended up with a grid for surveillance because everyone in that panel decided it would be useful. That doesn’t mean that it couldn’t be used for something else.
Legal Action and Enforcement
Marc Benson (J&J LifeScan, Milpitas, CA)
Mr. Marc Benson discussed the legal actions companies can take against counterfeiting of blood glucose test strips. He began by outlining the multifold dangers of counterfeit strips: 1) they can be inaccurate and lead to dangerously inappropriate insulin dosing; 2) this resulting improper treatment can burden an already overburdened healthcare system; 3) proceeds from counterfeiting often support other criminal behaviors; and 4) corporate anti-counterfeiting efforts drain valuable resources that could otherwise be spent on R&D. Mr. Benson noted with dismay that federal authorities are not necessarily helpful in fighting medical device counterfeiting due to financial, manpower, and jurisdiction limitations. He emphasized that succeeding in anti-counterfeiting efforts requires institutional commitment on the part of the manufacturing company, as even a relatively small investigation may require millions of dollars, drain company manpower, and take years to complete. Mr. Benson also pointed out that prosecuting and obtaining compensation from the counterfeiters is far from a certain outcome. The primary reason companies like LifeScan/J&J fight counterfeiters, Mr. Benson noted, is to deter future counterfeiting, protect their brands, and (most importantly) protect patients.
Workshop: Diabetes Technology Saves Money and Lives
The Current Economic Environment for New Diabetes Technology
Peter Ehrhardt (Simon-Kucher & Partners, Mountain View, CA)
Mr. Peter Ehrhardt discussed the current economic environment as well as payer perspectives and priorities on diabetes technology. For existing technologies such as BGM, he noted that payers are increasingly interested in cost-containment measures that restrict price or the volume of patient usage. Regarding newer products, Mr. Ehrhardt emphasized that market access restrictions are growing more stringent, making it more difficult to obtain premium reimbursement for new products. New technologies will need to be associated with proven improvements in outcomes or cost-effectiveness to appeal to payers. Overall, Mr. Ehrhardt characterized the current reimbursement environment as an area fraught with challenges and dilemmas for innovators. Payers want new technologies to be backed up by outcomes data, which are costly and difficult to obtain. Manufacturers who go the alternate route of appealing to payers through cost-benefit run the risk of cannibalizing sales of their other products. We find these observations rather worrying, as there are already enough roadblocks to innovation. Mr. Ehrhardt ended with a series of recommendations for innovators, including catering to payer preferences for big data and initiating cost-effectiveness trials early in the product development process.
- Payers are increasingly pushing cost-containment measures for existing diabetes technologies. Many measures focus on reducing the volume of reimbursement, including limits on the number of glucose testing strips reimbursed per patient in France, the UK, and Canada. In Germany, volume is controlled by largely excluding type 2 diabetes patients from BGM reimbursement. Other payer tactics focus on price, including Medicare’s competitive bidding system for BGM in the US.
- Increasingly, payers are looking for new technologies to demonstrate clinically meaningful outcomes or cost-effectiveness benefits to gain reimbursement. This is especially true for therapies that come at higher costs than current options. Mr. Ehrhardt cited the example of Insuline’s InsuPad as a technology that appealed to payers through cost-effectiveness data. The device is associated with clinically proven reductions in hypoglycemia and insulin usage, both of which have clear cost benefits as well as benefits for patients. Read our coverage of the Barmer reimbursement study from EASD (page 14).
- Mr. Ehrhardt characterized the current reimbursement environment as an “innovator’s dilemma.” Payers increasingly see currently available therapies as good enough, given the incremental nature of most advances, and are therefore more demanding of new technologies. They generally want proven efficacy and outcomes data for a candidate technology, which are difficult and costly to obtain. Some manufacturers have gone the alternate route of appealing to payers through cost-effectiveness data, but Mr. Ehrhardt pointed out that this could result in the cannibalization of the company’s existing products.
- Mr. Ehrhardt concluded with a set of recommendations for diabetes technology innovators. He noted that payers are particularly receptive to data-driven solutions, such as data collection and processing tools that could help them manage large patient groups more effectively. He discussed the idea of targeting the patient as the payer through technologies that improve patients’ lifestyle, but noted that the opportunity for reimbursement in this space is highly limited. He also recommended planning and initiating health-economic trials as soon as possible in the product design process, which we see as particularly valuable advice, assuming given that reimbursement decisions are as important as regulatory approval in terms of commercialization success. Although, of course, product design can change, seeing the magnitude of health economic impact from the start is important, even just to understand how product design may change to heighten impact.
Self-Monitoring of Blood Glucose: Impact of Accuracy on Clinical and Economic Outcomes
Oliver Schnell, MD (Forschergruppe Diabetes, Helmholtz Centrum Munich, Munich-Neuherberg, Germany)
Dr. Oliver Schnell discussed the effects that blood glucose monitoring accuracy (or inaccuracy) has on clinical and economic outcomes. He began by emphasizing that there is a clear association between glycemic variability and cardiovascular disease, with a number of potential mechanisms (including endothelial changes, atherosclerosis, and lipid levels) to explain the connection. He noted that hypoglycemia (one half of the glycemic variability equation) cannot be overlooked in this regard. Dr. Schnell proceeded to discuss the results of a health economic modeling study on SMBG meter accuracy conducted by his research group in Germany (Schnell et al., Journal of Diabetes Science and Technology 2013): a reduction in meter error from 20% to 5% led to a 10% reduction in severe hypoglycemia and a 0.39% reduction in A1c, which in turn drove a 0.5% reduction in myocardial infarctions. The cost savings conferred by these benefits was 24 Euros (~$33) per patient per year, which (applying the results to the 2.3 million insulin-treated patients in Germany) would yield annual savings of over 55 million Euros (~$75 million). Despite the fact that this was a modeling study, it’s positive to see evidence that investments in better meter accuracy can yield substantial cost savings (as well as improved outcomes for patients, of course). Dr. Schnell concluded that the benefit associated with improvements in BGM accuracy is substantial, although future modeling studies are needed to better understand the more nuanced effects and impacts of accuracy.
Q: Whenever you try and get a payer to cover something, you run into the payer attitude that a patient is only theirs for a couple years and then will be someone else’s problem. Does that change now that the Affordable Care Act is in effect?
Dr. Bruce Quinn (Foley Hoag LLP, Boston, MA): I think that payer mindset has been exaggerated. It’s true that they don’t believe in the longer-term projections. The bigger thing is that insurers need to compete in the market each and every year with their pricing. If a payer raises costs by 20%, now they will lose patients. Therefore, it isn’t feasible for them to raise prices 20% to save money five years in the future. With the ACA, you have guaranteed insurance and insurers are stuck with the patients. The other thing is that with policymakers, none of them read journals like JDST. You need to communicate things multiple times before they respond.
Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA): I have a comment on whether we should be issuing non-insulin-users lower quality, less accurate monitors. The ethics are questionable, but there is data looking at the question. In a study, we surveyed a number of clinicians on the risks of inaccurate readings, and found that both insulin users and non-insulin users were at similar risk for many of the outcomes of inaccurate readings.
Q: How much do you think device companies should be emphasizing what the cost benefit is for a device over a drug? I see cases where payers ask if they should cover a drug with a given A1c reduction, or cover a device. I think device makers have a good case for cost-effectiveness that isn’t being made.
Mr. Peter Ernhardt (Simon-Kucher & Partners, Mountain View, CA): I fully agree. In terms of doing that modeling and making that comparison, manufacturers have to catch up. Of course, on the drug front, it doesn’t look that rosy either, as getting reimbursement for medications has been getting harder. In Germany, we have limitations on reimbursement for insulin. Both medications and devices are on the cost-effectiveness radar screen. But I agree that the device industry hasn’t done a good job of demonstrating relative cost-effectiveness.
Excess Mortality in Type 1 Diabetes: Can Technology Make a Difference?
John Pickup, MD, PhD (King’s College London, London, UK)
Dr. John Pickup’s talk on the impact of technology on mortality in type 1 diabetes focused on severe hypoglycemia as a modifiable risk factor. As background, Dr. Pickup brought up data from the UK National Diabetes Audit – excess mortality was high in type 2 diabetes patients (60% increased risk), but even more elevated in type 1 diabetes patients (160% increased risk) vs. the background population. Young female type 1 diabetes patients, as well as patients in deprived areas, were at particularly high risk. Worryingly, only half of diabetes patient death certificates mentioned diabetes, indicating that diabetes mortality may be significantly underestimated. Turning to hypoglycemia, Dr. Pickup cited a recent study (McCoy et al., Diabetes Care 2012) demonstrating that severe hypoglycemia increased five-year mortality risk over three-fold, even after adjustment for multiple factors. Dr. Pickup expressed particular concern over nocturnal hypoglycemia, otherwise known as “Dead in Bed Syndrome” – this is responsible for approximately 6% of deaths in type 1 diabetes patients under 40 years old. A 2008 meta-analysis (Pickup and Sutton, Diabetic Medicine) demonstrated that switching from MDI to an insulin pump reduced severe hypoglycemia by approximately 75%. CGM-linked pumps with low glucose suspend (LGS) also reduce hyperglycemia (especially nocturnal hypoglycemia) by substantial margins (Choudhary et al., Diabetes Care 2011). In concluding, Dr. Pickup lamented that hypoglycemia and its effect on mortality are not appropriately factored into cost-effectiveness models, including the frequently-used CORE model, and as a result the cost-reducing effect of hypoglycemia prevention is often underestimated.
Cost-Effectiveness of Glucose Monitoring in People with Diabetes
Stephanie Fonda, PhD (Walter Reed National Military Medical Center, Bethesda, MD)
Dr. Stephanie Fonda gave a data-driven presentation on the cost-effectiveness of SMBG and CGM, beginning with how much the healthcare system is willing to spend to add one year to a patient’s life (a challenging question to merely consider!). Relatively recent research pegs the value of a quality-adjusted life year (QALY) in our healthcare system between $109,000 and $297,000, with a number of factors (such as patient finances and available medical resources) potentially altering the value. Dr. Fonda next surveyed data on the cost-effectiveness of SMBG, concluding that SMBG appears to have good value and that clinical efficacy and cost-effectiveness can be improved through structured paired testing (such as testing immediately before and after a meal to uncover trends) and appropriate data interpretation by patients and providers. Two studies on the cost-effectiveness of CGM in type 1 diabetes gave a range of $45,000 to $98,000 per QALY gained, which Dr. Fonda pointed out was a very broad range. Dr. Fonda next shared the results of her research group’s 12-month cost-effectiveness analysis of real-time CGM (using the Dexcom Seven) in 100 type 2 diabetes patients not on prandial insulin using the IMS CORE Model (this was based on Dr. Robert Vigersky’s intermittent CGM study from ADA 2011). RT-CGM use for 12 weeks led to a 0.6% placebo-adjusted A1c drop and a cost-effectiveness of $8,800 per QALY based off of data taken 12 months after the initiation of treatment, due primarily to a reduction in complications. For reference, that beat the cost-effectiveness seen with SMBG in type 2 diabetes patients, and appeared to be better than the cost-effectiveness of SMBG or CGM in type 1 diabetes patients. We certainly agree with Dr. Fonda that technologies such as CGM are cost-effective in the long run – a bit key is collecting and using the right data to prove the point, especially with the latest devices that are more accurate and reliable.
Q: One of the advantages of SMBG that is discussed less is its impact on healthcare utilization. If you offer SMBG technology with a self-management algorithm, you can reduce the time each patient requires with his or her provider. Is that sort of health economic analysis being brought into play with SMBG? Instead of looking at it in terms of outcomes, can we look at it in terms of health services utilization?
Mr. Michael Minshall (DJO Global, Fishers, IN): In short, the answer is no. We do our analysis on what payers want to see, and what they want is outcomes. You’re right, in that there are other possible advantages for technology.
Q: Measuring glucose doesn’t change A1c, which is only an intermediate endpoint. It’s only the intervention that changes the A1c. Nobody has put forward exactly what the actual intervention is. Is it that CGM or self-monitoring induce behavior change? We need to know about the intermediate step. Are there any good studies that look at the intermediate marker?
Dr. John Pickup (King’s College London, London, UK): You brought up A1c, but we know that CGM interventions reduce hypoglycemia. There, we aren’t talking about complications that only come twenty or thirty years down the line. We haven’t focused on hypoglycemia enough.
Comment: There are so many personal patient factors that need to be stratified to understand what’s going on with patients.
Dr. Minshall: I think payers would pay for information on what patients would respond and what patients wouldn’t, because at the end of the day some people are going to adhere and respond and some aren’t. That would be very interesting information to have.
--by Adam Brown, Hannah Martin, Manu Venkat, and John & Kelly Close