2nd International Conference on Advanced Technologies and Treatments for Diabetes (ATTD)

February 25-28, 2009; Athens, Greece — Day #4 Draft

We wrapped up the final day of the 2009 ATTD in Athens with a full schedule of great talks. Most prominent in our view (although only given ten minutes!) was Dr. Lori Laffel’s presentation of the JDRF results from the under 7% cohort. Things look positive for CGM, and we hope that these results will join those from the main cohort to make a strong statement for the use of CGM. We also saw an interesting study about factory calibration of the Navigator, a new prototype non-invasive glucose sensor from Solianis Monitoring, and a new and improved version of the MPC algorithm that may make it better at coping with meals and other day-to-day fluctuations.

It’s truly an exciting time in the diabetes device arena, and this meeting has done a fantastic job of bringing together an incredible faculty – thank you to them all! – and some wonderful learning. This upcoming week we’ll be bringing you news from the Avalere Diabetes Forum 2009, starting on the 3rd.

Highlights

  • Lori Laffel, MD, MPH (JDRF Study Group) presented a welcome result from the JDRF Study Group, showing that well controlled type 1s (A1c <7%) on CGM can maintain that control while greatly reducing time spent in the hypoglycemia zone, effectively improving their safety and reducing glycemic variability. In this 26 week study, there was a dramatic decrease in time spent below 70 mg/dL per day in the CGM group (37 minutes/day less) although this wasn’t statistically significant because of technical problems with outliers – certainly, however, it’s a clinically relevant finding. There was also a significant reduction in time spent <60 mg/dL. A1c did not change in the CGM group, but worsened in the control group, leading to a difference of 0.3% after 26 weeks. The percentage improving A1c by 0.3% or more was much higher in the CGM group versus the control group, and conversely the percentage worsening A1c by 0.3% or more was much lower. The risk of severe hypoglycemia wasn’t different, despite the lower A1c in the CGM group – we would not necessarily have expected this given the low number of events. At ATTD, many presenters have commented that patients in control are good candidates for CGM, since they tend to display more concordant behavior. This trial reinforces that belief and at the same time should reassure reimbursement bodies that they can still see further health and safety improvements with CGM even in well-controlled patients. We always say that it’s not just the A1c, it’s the quality of the A1c that matters …
  • Irl B. Hirsch, MD (University of Washington, Seattle, WA) gave a talk on CGM, saying that we already have seen enough to understand CGM and make the case for using it. We know that it can be very effective, and we know that success has a behavioral component. In that sense, it’s no different from SMBG or insulin therapy. Just because the behavioral component makes it harder for conventional trials to deliver ‘clean’ answers, doesn’t mean we are paralyzed waiting for proof, and can’t move forward with the clinical work. The message is that academia and diabetes organizations are applying artificial and pedantic hurdles to the acceptance of CGM, by insisting on multiple randomized controlled trials, and incorrectly interpreting trials that had “negative results” instead of learning from them. Dr. Hirsch asserts that we don’t need any more trials like STAR1 or JDRF – but in fact there are a dozen or more underway. His approach was a welcome counterpoint to Dr. Bolinder’s sober talk yesterday, which made the case that we have formally proven very little about CGM to date. But Dr. Bolinder was talking realpolitik, which is that the trials are necessary for reimbursement and policy makers, despite physicians’ enthusiasm for CGM. Dr. Hirsch’s pragmatic perspective is welcome in the debate about CGM.
  • The work of Eyal Dassau, PhD (University of California Santa Barbara, CA) demonstrates the lightning progress of closed loop algorithms, and was one of the most exciting closed-loop presentations at ATTD this year. The new L-MPC (learning- type model predictive control) approach upgrades the MPC algorithms that are already showing strong performance for the artificial pancreas. Dr. Dassau has a process engineering background and has created an algorithm that learns about typical daily glucose patterns. A supervising algorithm, called the ILC, learns about an individual’s lifestyle, and then controls the settings for the MPC to achieve superior performance than just the MPC alone. This approach effectively learns about mealtimes and minimizes user intervention – so no need for meal announcements. The in silico results show great control, and excellent robustness to meal variations or subject variability – but as with all simulations, we have to be cautious about extrapolating to real life. The next step is to test L-MPC in the clinic. Dr. Dassau mentioned that in this setting we can also fine-tune L-MPC more closely to each subject, which was not done in the simulation.
  • Peter Chase, MD (Barbara Davis Center, Colorado, United States), and Bruce Buckingham, MD (Stanford University, Palo Alto, CA) discussed the use of an algorithm to shut off insulin delivery and avoid hypoglycemia, leading to outpatient studies. The daytime studies showed a 57% reduction in hypoglycemia, while the night time work showed an 84% hypoglycemia reduction. The algorithm uses a more sophisticated ‘voting system’ in which five prediction algorithms contribute to the insulin shutoff decision-making process. This is a very impressive result, and takes ‘Low Glucose Suspend’ a step further into the prevention realm. Note that at no time does this algorithm dose insulin – it’s all about improving patient safety. Which parent wouldn’t want their child to wear a pump that reduced nocturnal hypoglycemia by 84%? More broadly, we hope that while 2008 was the year of algorithm development, 2009 will be the year of inpatient trial results, and 2010 the year of outpatient trial results.
  • Chiara Dalla Man, PhD (University of Padova, Padova, Italy) discussed the role of algorithms in reproducing physiological insulin secretion. Like the other speakers in this session, she emphasized that artificial glucose control is not likely to replicate endogenous glucose production, but instead should focus on achieving functional control. She briefly mentioned an in silico study completed by her group that compared the proportional-integral- derivative (PID) and MPC artificial pancreas control algorithms, and showed that MPC was able to better control both hypoglycemic and hyperglycemic excursions compared to the PID algorithm. While it seems like most research groups currently are using the MPC algorithm, Medtronic continues to work using the PID algorithm. They contend that, while MPC is more accurate, PID is easier to implement and is accurate enough for functional control. We don’t have a firm position on which approach is better, although there may be some sense in using an imperfect but adequate algorithm if it helps to get the first artificial pancreas to market.
  • Arleen Pinkos, BS, MT, ASCP (US Food and Drug Administration, Washington DC) is a strong champion of the artificial pancreas, and is working to see it approved as quickly and efficiently as possible. However, it’s not going to be easy, and in this presentation she gave us many ‘clues’ as to where the problems lie. Specifically we learned not to over-reach on the intended use, to avoid complexity in hardware and software, to focus above all on safety and risk mitigation. It seems that we will get to the full artificial pancreas by a series of careful steps. Ms. Pinkos did an excellent job of representing the position of the FDA while advocating for patients: “I love CGM, it’s a great device. If I were a diabetic I would use CGM. I lost my mom to diabetes. If she were still alive I would have her use one.”
  • Jay Skyler, MD (University of Miami, Miami, FL) discussed possible peptide approaches for the treatment of obesity. He went through studies of a number of candidates, including GLP-1, PYY, oxyntomodulin, pramlintide, and leptin. In contrast to most discussions of this topic we’ve seen before, Dr. Skyler spent a lot of time discussing combination therapy. In particular, he discussed the combinations of amylin with leptin and amylin with PYY, and referenced triple therapy with amylin, leptin, and PYY. In addition, he briefly discussed a new second-generation amylin mimetic called davalintide, which has been optimized for obesity and shown to produce up to 15% weight loss by four weeks. The first generation amylin mimetic, pramlintide (Amylin’s Symlin) was designed as a diabetes drug and typically involves complex dosing. This newer generation is expected to have greater weight loss efficacy and simpler dosing. Phase 2 results are expected in 4Q09 – we’re looking very forward.
  • Henry Buchwald, MD, PhD (University of Minnesota, Minneapolis, MN) discussed the results of a soon-to-be-published meta-analysis researching the effects of bariatric surgery on weight loss and diabetes resolution. The work included studies from 1990 to 2006, comprising a total of 621 studies and 135,246 patients. Overall, the average excess weight loss seen was 56% (46% with banding, 55.5% with gastroplasty, ~60% with gastric bypass, and 63.6% with biliopancreatic diversion or duodenal switch). 76.8% of diabetes cases were resolved, 83.9% were either resolved or improved, 1.1% were worsened, and 14.9% remained unchanged. Patients undergoing surgery saw an average A1c reduction of 1.3%. Dr. Buchwald also discussed some experimental procedures to treat diabetes, like the EnteroMedics VBLOC technology and bypass surgeries that produce resolution of diabetes without weight loss. Notably, he also mentioned the potential role of implantable pumps in delivering incretin infusions.
  • Udo Hoss, PhD (Abbott Diabetes Care, Alameda, CA) discussed the results of a compelling study suggesting that the Navigator may be accurate with a factory calibration, eliminating the need for repeated fingersticks for this purpose. A 12- person pilot study done by Abbott showed that the accuracy of factory-calibrated sensors as measured on the Clarke Error Grid was no different from the accuracy when the sensors were calibrated with fingersticks. Dr. Hoss expressed enthusiasm for the results, but cautioned that more data from diabetic subjects are needed before firm conclusions can be made. This could be a big win in terms of patient convenience with the Navigator, were it to be proven to the FDA’s satisfaction.
  • Andreas Caduff, PhD (Solianis Monitoring, Zurich, Switzerland) discussed a new approach to spectroscopic non-invasive glucose monitoring being pioneered by Solianis Monitoring AG. Spectroscopic systems are extremely sensitive and have been plagued with interference from a variety of sources, but the use of addition sensors to detect interfering factors may allow a detection device to compensate for them and produce accurate readings. Initial results using the sensor combination in a controlled setting have been promising, with 86% of readings in the A or B region of the Clarke error grid. Like all diabetes technology, the more relevant question is performance in the real-world environment. Dr. Caduff will be presenting additional data at ADA 2009, and hopes to be ready to start clinical validation of the system in 2010.
  • R. Nimri, MD (Schneider Children’s Medical Center of Israel, Tel Aviv, Israel) presented work with the MD-Logic closed loop algorithm we learned about earlier in this conference. As far as we can tell, MD-Logic is a kind of fuzzy logic system. We understand that the system learns, but not automatically – it has to be ‘taught’ manually. This insulin/glucagon closed loop work in pigs reminds us of that of Dr. Ed Damiano, (who did it with sub-cutaneous infusion), who has now moved on to human trials. The learnings from insulin/glucagon are (i) that the system can be very effective at preventing hypoglycemia and (ii) that the system can be tuned to more aggressive at reducing hyperglycemia because of the ‘safety net’ of glucagon. It’s interesting to think about the effect of having insulin and glucagon administered closely together as in the more aggressive case above. Also, it might be that frequent use of glucagon could exhaust the liver, leading to ‘failure of the emergency brakes”.
  • Andries Smit, MD (University Medical Center, Groningen, Netherlands/DiagnOptics) discussed the use of skin autofluorescence as a non- invasive test for type 2 diabetes. Skin autofluorescence (AF) can be measured very simply and easily, and turns out to be able to detect type 2 diabetes. It’s a sort of cumulative damage marker. In studies, skin AF was reported to be a better diagnostic tool for the disease than FPG or even A1c (we note neither fasting blood glucose nor A1c is a standout diabetes diagnostic tool). So it’s a great, quick way of screening for diabetes without a blood test, they said.
  • Roger Lehmann, MD (University Hospital Zurich, Zurich, Switzerland) discussed the efficacy of various types of transplantation in restoring insulin secretion, and examined some of the problems facing islet cell transplantation. He presented data showing that about half of whole pancreas transplant patients have been shown in some studies to achieve normal glucose tolerance, while the auto-transplantation of islets results in an insulin- independence rate of 50-80%, and allo-transplantation has an insulin independence rate of about 26%. Major problems with transplant that he identified included the limited mass of permanently engrafted islets - which can experience 60-70% death within two weeks - the lack of a reliable marker for rejection, and the large number of islets required—300,000 for auto-transplantation and 600,000 for allo-transplantation. He cautioned that the target for transplantation should not be insulin independence but a reduction in A1c, because independence may not always be possible.
  • The work of Paolo Pozzilli, MD (University Campus Bio-Medico, Rome, Italy) establishes that we should put people with type 1 diabetes on the pump immediately after diagnosis to preserve beta cell function for the longest time. The age of diagnosis is associated with different levels of C-peptide - levels increase with age. Knowing your C-peptide is not an academic exercise, but might help you in future. Medtronic is sponsoring a pump/CGM trial in newly diagnosed kids – this should be a good opportunity to collect further data on C- peptide and relate this to glycemic control.
  • Tihamer Orban, MD (Joslin Diabetes Center, Boston, MA) reported the 104-week phase I results of a novel drug involving the reintroduction of the self-antigen to insulin with the goal of establishing immune tolerance. The trial had 12 patients randomized to control and treatment groups, and the treatment was given as one injection of insulin B-chain peptide. Patients were newly diagnosed type 1 diabetics between 18-35 years, with any positive autoantibodies. The average A1c was ~8.5%, although C-peptide levels were lower in treatment group to start. The drug caused no serious adverse effects. There was no difference in C-peptide values or A1c between treatment and control groups. Insulin antibody titers were much higher in the treatment group, as expected. These antibodies had no effect on the metabolic effect of insulin treatment. Vaccinated patients had a significant T-cell response, peaking at six months. The vaccine was shown to produce an insulin B-chain-specific immune response. We’ll be interested in seeing further data—companies have begun to show an interest in this area, and insulin-based vaccines may soon emerge as a viable therapeutic target.
  • Tadej Battelino, MD (University Medical Center, Ljubljana, Slovenia) gave a somewhat high level overview of closing the loop, covering the different aspects of system design and some of the inherent problems in the technology we have today. He touched on sensor design, pump design, potential controllers, and new insulins, among others.
  • Roman Havorka PhD (University of Cambridge, Cambridge, UK) discussed the possibility of replicating physiologic insulin secretion by artificial means. He thinks that the various components of physiologic insulin secretion in healthy individuals make it nearly impossible to replicate. However, he thinks that replication is not necessary to achieve good glycemic control, and that the focus should be on functional replacement rather than replication.
  • Yong Zhao, MD, PhD (University of Illinois, Chicago, IL) and his team have identified cord-blood-derived stem cells that may modulate the response of T lymphocytes to diabetes auto-antigens when co-cultured with them in the laboratory. Treatment with these modulated cells (mCD4CD62L Tregs) normalized the blood sugar of diabetic mice, and increased insulin production by increasing beta cell mass in these mice. This technology would pose no rejection risk, and would be very cost-effective (if it proves effective in humans), but this is still very early-stage research.
  • George Eisenbarth, MD, PhD (Barbara Davis Center for Childhood Diabetes, Aurora, CO) discussed his work in attempting to identify the genetic foundations of the immune response in type 1 diabetes. He thinks that insulin is the primary autoantigen responsible for diabetes, at least in the NOD mouse. The B-chain 9-23 peptide is the specific region of insulin that provokes an immune response, and T-cells targeting this region share a specific fragment of their alpha chain recognition protein. He believes that once the structural determinants of insulin antigen recognition are better understood in humans, there will be different ways to interfere with the process and potentially create a vaccine.
  • C. De Block, MD, PhD (University of Antwerp, Belgium) described mounting evidence showing that tight control of glycemia in the ICU is critical. CGM will make insulin titration easier and safer and save nurse time. He predicted that closed loop control would reduce nurse workload and hypoglycemia.
  • It’s great to see more process control engineers entering the artificial pancreas fray with new ideas for algorithms. H. Cormerais, PhD (University of Rennes, France) is introducing Passivity Based Control, which has been used with success elsewhere. This paper was intriguing, but more work needs to be done before we can properly compare the performance of this approach with PID and MPC approaches.
  • Addressing a hot topic in the world of islet replacement, Paola Maffi, MD (Instituto Scientifico San Raffaele, Milan, Italy) discussed the challenges of the islet transplant process and some potential solutions. She talked specifically about the challenges inherent in procurement of a pancreas, islet culture, and islet engraftment, and discussed alternative implant sites and pretreatment with immunosuppressants as possible solutions to enable longer-term islet survival.
  • Ralf Schiel, MD (Seeheilbad Heringsdorf, Rostock, Germany) discussed a new telemedicine intervention to help young people with diabetes improve their insulin therapy by better monitoring physical activity and eating habits. The assessment used a wireless sensor integrated in a mobile phone that monitored the kind, intensity, and duration of physical activity, and a camera that could be used to take pictures of food for later analysis by a dietician to improve carbohydrate estimates. The exercise measurement technology can identify ten exercises with near-perfect accuracy. The technology was used in two patients at a summer camp over four and nine days. There were large differences between the measured assessments of the duration of physical activity and patients’ subjective reporting. Dr. Schiel thinks that technology could be used to enable better calculation of physical activity and eating.
  • Joachim Ficker, MD (Nurnburg, Germany) talked about the connection between diabetes and obstructive sleep apnea (OSA), both of which he titled epidemics. Obesity is a risk factor for both diabetes and OSA, OSA is a risk factor for diabetes, and both OSA and diabetes interact as risk factors for cardiovascular disease. The IDF recommends OSA screening for type 2 patients.
  • Vincenzo Bacci, MD (Universiti de Roma, Rome, Italy) discussed gastric balloon technology, which in our understanding is used with some frequency in Europe and Asia but is much less common in the US. The balloons are a small plastic reservoir that can be filled with saline solution or air, expanding to take up space in the stomach and promoting weight loss. The balloons can only be used for a period of six months, and some studies have shown that they produce a BMI decrease of about 3.2 kg/m2 in that time, and an average A1c reduction of 0.5% (from 7.5%, in one study). Weight is usually regained after balloon removal, although Dr. Bacci thinks that a subset of patients may exhibit weight loss maintenance after removal.

Corporate Symposia Highlights

SANOFI-AVENTIS SPONSORED SESSION

  • Jay Skyler, MD (Diabetes Research Institute, University of Miami, Miami, FL) set out a compelling explanation of what happened in ACCORD, ADVANCE, and VADT and the implications for clinical guidelines. It seems that we have learned that (i) if patients in trials are relatively healthy, and we start early, then we get macrovascular protection from intensive management, (ii) if patients have established cardiovascular disease then it may be too late to help them via tight glycemic control (and we can even harm them if therapy is too aggressive). The corollary is that if we want a trial to prove that early intensive control in relatively healthy people has cardiovascular benefits, we have to wait a long time. The net effect is to prevent the adoption of yet lower A1c targets (such as 6.5%), but to allow that higher goals might be appropriate for those with hypoglycemia, cardiovascular disease or who are older. For everyone else, getting as low as possible without hypoglycemia makes sense for the microvascular benefits alone.

NOVO NORDISK SPONSORED SESSION

  • Giorgio Sesti, MD (Unviersita di Catanzaro, Catanzaro, Italy) gave an overview of the liraglutide clinical program and recent studies describing the effects of the drug. He showed a slide from Blonde et al., 2008 showing that nausea with liraglutide is transient, while nausea with exenatide is more lasting. Nauck et al., 2009 showed that liraglutide can reduce visceral adiposity, and Jendle et al., showed that weight reduction with liraglutide is due to a loss of fat mass. Courreges et al., 2008 showed that treatment with liraglutide is associated with improved biomarkers of CV risk. Further studies have shown that liraglutide produces antibodies in about 8.6% of patients, while exenatide plus metformin produces antibodies in 49%, and Dr. Sesti suggested that this was due to liraglutide’s greater homology with native GLP-1.
Table of Contents 

Detailed Discussion and Commentary

AIDPIT Symposium: Reproducing Insulin Secretion

PHYSIOLOGY OF INSULIN SECRETION: WHAT IS NEEDED FOR RESTORATION?

Roman Havorka PhD (University of Cambridge, Cambridge, UK)

Dr. Havorka discussed the possibility of artificially replicating physiologic insulin secretion. He thinks that the various components of physiologic insulin secretion in healthy individuals make it nearly impossible to replicate. However, he thinks that replication is not necessary to achieve good glycemic control, and that the focus should be on functional replacement rather than replication.

  • There are many important components to insulin regulation in healthy people. In healthy adults, glucose ranges between about 59 and 168 mg/dL. Fasting insulin is secreted by the pancreas in high frequency, pulsing into the portal vein every four minutes. At least 70% of insulin secretion is pulsatile, involving the synchronization of islet cells. After a meal, there are a variety of insulin secretions—there is a pre-absorptive insulin response (cephalic secretion) that lasts about ten minutes, and is required for a normal postprandial glucose tolerance. The release of incretin hormones amplifies insulin secretion relative to the size of the ingested meal. Once the cells start to secrete insulin, the rate of secretion is approximately linear with glucose concentration. In type 1 diabetes, hyperglycemia does not delay gastric emptying as it does in people without diabetes, and glucose intake does not suppress hepatic glucose production.
  • Dr. Havorka thinks it would be very difficult to restore physiologic insulin secretion by artificial means. With subcutaneous delivery, it takes insulin 40 minutes to reach a peak blood concentrations, 30 minutes longer to get to peak action, and then 70 minutes more to get the peak rate of glucose disposal. He said that the restoration of physiological secretion by the subcutaneous route is not possible, but this may not be necessary, and functional replacement may be enough. Taking into account what we know about bodily processes, the variability between different people, and the limitations of current insulin devices, he believes we can formulate an approach to insulin delivery that works well enough to treat diabetes. He thinks that insulin delivery could be improved to have faster absorption, additional meal detection, and perhaps intra-peritoneal delivery to more closely mimic physiologic secretion. Priorities for improving functional replacement remain to be clarified.

HOW WELL DO TRANSPLANTED ISLET CELLS RESTORE INSULIN SECRETION?

Roger Lehmann, MD (University Hospital Zurich, Zurich, Switzerland)

Dr. Lehmann discussed the efficacy of various types of transplantation in restoring insulin secretion, and examined some of the problems facing islet cell transplantation. He presented data showing that about half of whole pancreas transplant patients have been shown in some studies to achieve normal glucose tolerance, while the auto-transplantation of islets results in an insulin-independence rate of 50- 80%, and allo-transplantation has an insulin independence rate of about 26%. Major problems with transplant that he identified included the limited mass of permanently engrafted islets, which can experience 60-70% death within two weeks, the lack of a reliable marker for rejection, and the large number of islets required—300,000 for auto-transplantation and 600,000 for allo-transplantation. He cautioned that the target for transplantation should not be insulin independence but a reduction in A1c, because independence may not always be possible.

  • Dr. Lehmann began with a discussion of full pancreatic transplants. People need 50% of beta cell mass for normal glucose homeostasis, while insulin-independence requires about 20% of beta cell mass. With whole pancreas transplantation, the first phase insulin response is about normal, and second phase secretion is larger because of insulin resistance. About half of whole pancreas transplant patients have been shown in some studies to achieve normal glucose tolerance. Pancreas transplanted patients have increased amplitude of insulin pulses compared to normal.
  • He moved on to discuss the auto- and allo-transplantation of islet cells. The auto- transplantation of islets results in an insulin-independence rate of 50-80%, depending on the number of islets transplanted. Glucose disposal and insulin secretion is about halved compared to normal patients, even though A1cs may not be very different. Allo-transplantation was reported to have an insulin independence rate of about 26% up until 2000. The Edmonton protocol called for multiple transplantations, and still resulted in only 10% insulin-independence after five years, which was ultimately quite disappointing, as we understand it. After 12 months, glucose clearance was much lower, and there was only about 10% of normal insulin secretion regained.
  • The biggest problems with transplants are hypoxia and engrafted islet mass. The insulin dose required after transplantation is related to islet mass. On average, Dr. Lehmann performs about 2.2 transplants per patient to attempt to get adequate mass. However, islet mass is not predictive for A1c as long as patients have some C-peptide present. In animal studies, more success is seen in transplant if the pancreas for transplantation is taken from a live donor before brain death. Hubert et al., 2008 showed that response to arginine in brain dead donors could predict islets with a lower chance of survival, and 56% of these islets could be avoided. In addition, the size of the islets may make a difference; the portal vein does not have enough oxygen pressure to oxygenate the islets, and so big islets become necrotic in the center. Because of this, small islets survive hypoxia much better than big ones. Mac Gregor, 2006 saw that transplanting rats with small islets produces much better outcome, and that small islets have double the insulin secretion of large islets.
  • Critical obstacles to transplantation are: the limited mass of permanently engrafted islets, which can experience 60-70% death within two weeks, the lack of a reliable marker for rejection, and the large number of islets required—300,000 for auto-transplantation and 600,000 for allo- transplantation. It may help to choose smaller islets, and may be beneficial to do islet-kidney or islet after kidney transplants to reduce the need for immunosuppression. The goal should not be insulin independence but improvements in A1c, as independence may not be possible.

HOW CAN ALGORITHMS CONNECTING GLUCOSE SENSORS TO INSULIN DELIVERY DEVICES MIMIC INSULIN SECRETION?

Chiara Dalla Man, PhD (University of Padova, Padova, Italy)

Dr. Dalla Man discussed the role of algorithms in reproducing physiological insulin secretion. Like the other speakers in this session, she emphasized that artificial glucose control is not likely to replicate endogenous glucose production, but instead should focus on achieving functional control. She briefly mentioned an in silico study completed by her group that compared the proportional-integral- derivative (PID) and MPC artificial pancreas control algorithms, and showed that MPC was able to better control both hypoglycemic and hyperglycemic excursions compared to the PID algorithm. While it seems like most research groups currently are using the MPC algorithm, Medtronic continues to work using the PID algorithm. They contend that, while MPC is more accurate, PID is easier to implement and is accurate enough for functional control. We don’t have a firm position on which approach is better, although there may be some sense in using an imperfect but adequate algorithm if it helps to get the first artificial pancreas to market.

  • There are many features of glucose secretion. There is a short-term adaptation to glucose levels and a long-term adaptation to insulin resistance. In addition, there is a big difference in concentration of insulin in the portal and peripheral veins, and this gradient allows suppression of endogenous glucose production. The ‘optimal’ region for insulin infusion would be in the portal vein, but this is not possible. Intra-peritoneal delivery helps to mimic portal insulin delivery, but such delivery comes with complications and increased insulin antibodies.
  • An artificial pancreas controller may be able to provide improvements in glycemic control. She thinks that algorithms should not be focused on mimicking physiological secretion, but should provide functional control in reducing A1c, glycemic variability, and avoiding adverse events. The algorithms need to be robust against perturbation by meals and exercise. Possible closed-loop strategies are PID and MPC. Dr. Dalla Man thinks that PID is not suitable to control a system with a large delay (presumably referring to the lag time with CGM and subcutaneous insulin), while the MPC approach can handle such delays.
  • An in silico study by Dr. Dalla Man compared PID with MPC. Qualitatively, MPC resulted in fewer excursions into the hyper- and hypoglycemic range compared with PID. An artificial pancreas system will probably never mimic physiologic secretion, but should employ a ‘control to range’ strategy.

HOW TO IMPROVE ISLET TRANSPLANTATION IN HUMANS?

Paola Maffi, MD (Instituto Scientifico San Raffaele, Milan, Italy)

Addressing a hot topic in the world of islet replacement, Dr. Maffi discussed the challenges of the islet transplant process and some potential solutions. She talked specifically about the challenges inherent in procurement of a pancreas, islet culture, and islet engraftment, and discussed alternative implant sites and pretreatment with immunosuppressants as possible solutions to enable longer-term islet graft survival.

  • The first important consideration in islet transplantation is the procurement of the pancreas. Ideal donors are 35-55 years of age, with a BMI 25-30, death from cerebrovascular cause, didn’t abuse alcohol, non hypertensive or hypotensive, and have no history of pulmonary or cardiac arrest. The procurement team has to be very experienced, because islets can be damaged during transportation. Islet isolation can be difficult, and inconsistencies in the equipment and enzymes used can produce wide variations in islet viability after isolation.
  • Islet culture is the second important factor. It is necessary to control the islet preparation after isolation, to allow the patient to arrive at the hospital and undergo pretreatment, but culture does reduce the islet recovery between 20% and 30%. Although the culture medium does reduce islet recovery overall, it has been shown to be important for preservation purposes for islets over 24 and 48 hours.
  • Islet engraftment also poses a particularly significant challenge. Many sites of implantation have been tried, as the optimal site needs easy access, a rich oxygen supply, good metabolic function, and must be able to support long-term function. The liver seems to be the best site thus far, although it does not have a rich oxygen supply, immunoprivilege, or good support of long term function. Improved islet survival may entail blocking inflammation, changing the microenvironment, and improving support for the islet. Bone marrow has been an experimental site of implantation in animals, and has shown good islet survival rates and local insulin production. Glucose control in animals is better when islets are placed in bone marrow rather than in liver.
  • A modified version of the Edmonton protocol using pretreatment with the immunosuppressant tacrolimus has been shown to improve results compared to the original. The Edmonton-Miami group, using the CAMPATH protocol has shown improved islet survival compared to the original protocol.
  • For success with islet transplants, it is important to optimize the enzyme blend, optimize early engraftment, optimize immunosuppressive therapy, and tolerance induction.

CONTINUOUS GLUCOSE MONITORING IN THE TISSUE: DO WE REALLY NEED TO CALIBRATE IN VIVO?

Udo Hoss, PhD (Abbott Diabetes Care, Alameda, CA)

Dr. Hoss discussed the results of a compelling study suggesting that the Navigator may be accurate with a factory calibration, eliminating the need for repeated fingersticks for this purpose. A 12-person pilot study done by Abbott showed that the accuracy of factory-calibrated sensors as measured on the Clarke Error Grid was no different from the accuracy when the sensors were calibrated with fingersticks. Dr. Hoss expressed enthusiasm for the results, but cautioned that more data from diabetic subjects are needed before firm conclusions can be made. This would be a big win in terms of patient convenience with the Navigator, were it to be proven to the FDA’s satisfaction.

  • All current subcutaneous sensors require calibration with a fingerstick, and recalibration at frequent intervals. Is it possible to eliminate calibrations?
  • The requirements for factory calibration to be possible include reproducible sensor manufacturing, stable sensor sensitivity on shelf, a stable signal over use period, constant background signal, constant inter- and intra-subject sensitivity, and a known and unchanging relationship between in vitro and in vivo sensitivity.
  • The Navigator sensor may be suitable for factory calibration. The Navigator sensor has very low background interference and very stable chemistry, and has a known average in vitro sensitivity. Notably, an Abbott study showed a stable sensor life up to 30 months at up to 30 degrees Celsius. Biofouling may be a problem for some sensors, but further Abbott studies have shown that the sensor does not experience significant drift over a five-day period. There is a known average in vitro/vivo sensitivity factor for the Navigator and no change in background signal/interferences because the device runs at a very low electric potential. Constant inter- and intra-subject sensitivity is a bigger problem yet to be resolved.
  • An Abbott study with 12 subjects over five days using two sensors in parallel per subject demonstrated that the Navigator may be as accurate with factory as with fingerstick calibration. The sensors showed a coefficient of variance of 4.6% in vitro and 9.4% in vivo. This change was characterized as ‘not unexpected’, but was significantly different using an ANOVA statistical model. Although there was an inter-subject variation seen, 99.8% of sensor values were in the A+B range of the Clarke Error Grid with factory calibration, with 88.1% were in the A zone. Only 0.2% of subjects were in the D+E zones with factory calibration, and the MARD was 10.4% (compared to YSI, we assume). Using a standard finger-stick protocol, these numberswere about the same: 86% in A zone and a MARD of 10.9%. A 20-hour graph showed that factory calibrated sensor readings can be biased low, high, or can be on-target, but that they are generally no more inaccurate than sensors calibrated with fingerstick readings. Dr. Hoss expressed enthusiasm for the results, but cautioned that more data from diabetic subjects are needed before firm conclusions can be made.

Sanofi-Aventis Sponsored Session

WHAT'S CHANGED? THE RECENTLY INTRODUCED EASD/ADA GUIDELINES FOR THE MANAGEMENT OF T2DM

Jay Skyler, MD (Diabetes Research Institute, University of Miami, Miami, FL)

There has been a lot of talk about ACCORD, ADVANCE and VADT in the context of DCCT and the UKPDS. Dr. Skyler set out a compelling explanation of what happened and the implications for clinical guidelines.

It seems that we have learned that (i) if patients in trials are relatively healthy, and if we intervene early, then we get macrovascular protection from intensive management, (ii) if patients have established cardiovascular disease then it may be too late to help them via tight glycemic control (and we can even harm them if therapy is too aggressive). The corollary is that if we want a trial to prove that early intensive control in relatively healthy people has cardiovascular benefits, we have to wait a long time.

The net effect is to prevent the adoption of yet lower A1c targets (such as 6.5%), but to allow higher goals as appropriate for those with hypoglycemia, cardiovascular disease, or who are older. For everyone else, getting as low as possible without hypoglycemia makes sense for the microvascular benefits alone.

  • The 2009 revisions to ADA guidelines push for an aggressive lowering of certain cardiovascular risk factors. They also recommend the daily use of aspirin (or clopidogrel) for any person with diabetes over 40 (or with cardiovascular risk).
  • The ADA now targets an A1c of 7%, higher than the IDF’s and AACE’s 6.5% target. The 2006-8 ADA target used to be ‘as close to normal as possible without hypoglycemia’. This has changed in the light of recent clinical trials.
  • To summarize the key trials – DCCT, UKPDS, ACCORD, ADVANCE, and VADT – we see evidence that microvascular complications are reduced with intensive control. Cardiovascular disease is reduced in the long term, but not in short term. We saw short term increased mortality for intensive control in ACCORD and long term reduced mortality in UKPDS.
  • We have to wait for impact on mortality and cardiovascular disease. DCCT patients were young and healthy (mean age of 27), so it was only with the EDIC follow up that we saw twice as many cardiovascular events in the conventional group. There was a 57% reduction of any kind of event (MI, stroke etc). In the UKPDS there was a 15% risk reduction in MI or sudden death, and a 13% reduction in all-cause mortality, but this only became statistically significant many years after the beginning of the trial.
  • All three modern trials (ACCORD, ADVANCE, and VADT) had a trend towards risk reduction on their primary endpoints, although it was not significant. But the hazard ratio for mortality was 22% in ACCORD in the intensive group. Somehow, intensive therapy was killing people, so that arm of the trial was stopped early. ADVANCE had a trend towards reduction in mortality and VADT was essentially neutral.
  • Patients in ACCORD tended to be less healthy and they were pushed to achieve an A1c of <6% “no matter what”. They had to change their regimen every month if they were over 5.9%. This is probably not realistic in clinical practice. As a result they took much more insulin and combination drug therapy, and experienced more weight gain and hypoglycemia. “Did we push too hard?” asked Dr. Skyler. “My bias is that we did.”
  • In both ACCORD and ADVANCE, intensively managed patients who had not had previous history of cardiovascular disease got statistically significant protection on the primary outcome. In ACCORD it was the group who’d had a previous event who weighted the primary outcome so it wasn’t significant. In ADVANCE we also saw protection if patients were not previously treated with statins. In VADT we see suggestions of the same thing – those with shorter duration of diabetes had hazard ratios that benefitted them, only those with longer duration (presumably more cardiovascular complications) had more risk. The bottom line – early treatment may be beneficial for cardiovascular disease. But intensive management later might just be too late. All modifiable cardiovascular risk factors should be treated.
  • So the ADA recommends keeping A1c <7% (to protect against microvascular complications), although higher targets are acceptable in patients with established cardiovascular disease, those with hypoglycemia, and older patients. All modifiable cardiovascular risk factors should also be treated.
  • We still don’t know if treating glycemic variability leads to better outcomes. Now we have CGM, we should do a trial to figure this out, since we can’t infer glycemic variability from A1c.

Questions and Answers

Q: What is the role of CGM here?

A: What patients learn with CGM is impact of food on glucose, so it can really help achieve safer control. In the future CGM will have factory calibration. It will be used for couple weeks at a time. “It will be easier for the type 2 patient and will sweep the management of the disease just as blood glucose replaced obsolete urine tests.”

Q: Isn’t it silly to use long acting insulin first in the algorithm? We should use mealtime insulin.

A: I studiously avoided discussing the algorithm. It’s not a position statement, it reflects the consensus of the five or six people who created it. I personally don’t think it’s the right algorithm at all. It is driven by cost considerations. We need to re-examine the disease on a pathophysiological level and address that in the algorithm.

Novo Nordisk Sponsored Session

LIRAGLUTIDE: THE FIRST HUMAN GLP-1 ANALOGUE FOR ONCE DAILY ADMINISTRATION

Giorgio Sesti, MD (Unviersita di Catanzaro, Catanzaro, Italy)

This lecture was an overview of the liraglutide clinical program and recent studies describing the effects of the drug. Dr. Sesti briefly discussed the Liraglutide Effect and Action in Diabetes (LEAD) program, emphasizing its completeness. He showed a slide from Blonde et al., 2008 suggesting that nausea with liraglutide is transient, while nausea with exenatide is more lasting. Nauck et al., 2009 showed that liraglutide can reduce visceral adiposity, and Jendle et al., showed that weight reduction with liraglutide is due to a loss of fat mass. Courreges et al., 2008 showed that treatment with liraglutide is associated with improved biomarkers of CV risk. Further studies have shown that liraglutide produces antibodies in about 8.6% of patients, while exenatide plus metformin produces antibodies in 49%, and Dr. Sesti suggested that this was due to liraglutide’s greater homology with native GLP-1.

  • Current therapy for type 2 diabetes leaves several unmet needs, including the problems of declining control, weight gain, increased risk of hypoglycemia, and worsening of CV risk. GLP-1 improves glucose control through a wide range of mechanisms, and can improve glucose control, increase c-peptide levels, and reduce glucagon in a glucose- dependent way. They reported a T1/2 of 13 hours and tmax of 10-13 hours.

Questions and Answers

Q: Can you speculate about the difference in nausea?

A: This is a good point—I think that this is related to the fact that this is once/day while exenatide is twice/day. There are no other studies showing the mechanism behind this. Unfortunately, this is only data recorded by patients, and there is no data comparing the GI effects of the two drugs (like gastric emptying).

Novel Strategies for the Treatment of Obesity and Diabetes

PEPTIDE HORMONES FOR OBESITY

Jay Skyler, MD (University of Miami, Miami, FL)

Dr. Skyler discussed possible peptide approaches for the treatment of obesity. He went through studies of a number of candidates, including GLP-1, PYY, oxyntomodulin, pramlintide, and leptin. In contrast to most discussions of this topic we’ve seen before, Dr. Skyler spent a lot of time discussing combination therapy. In particular, he discussed the combinations of amylin with leptin and amylin with PYY, and referenced triple therapy with amylin, leptin, and PYY. In addition, he briefly discussed a new second- generation amylin mimetic called davalintide, which has been optimized for obesity and shown to produce up to 15% weight loss by four weeks. The first generation amylin mimetic, pramlintide (Amylin’s Symlin) was designed as a diabetes drug and typically involves complex dosing. This newer generation is expected to have greater weight loss efficacy and simpler dosing. Phase 2 results are expected in 4Q09– we’re looking very forward.

  • Dr. Skyler gave an overview of peptide hormones important in obesity. On his list were GLP-1 analogs, PYY, oxyntomodulin, amylin analogs, and leptin.
  • GLP-1 reduces food intake and promotes weight loss by a number of mechanisms. Dr. Skyler used liraglutide data to demonstrate the effects of the GLP-1 class on body weight.
  • PYY also reduces food intake and reduces body weight. There is a potential for combination therapy with GLP-1 and PYY. Talsania et al., 2005 showed that an exendin-4 plus PYY infusion produced greater weight loss than either alone in a rodent model. This was studied in humans by Neary et al., 2005 with similar results.
  • Oxyntomodulin has shown reductions in food intake in human subjects, and is reported to be associated with ~2 kg weight loss over a four-week period.
  • As shown by Gormally et al., 1982 and Smith et al., 2007, the amylin analog pramlintide reduces a ‘binge eating score’ and decreases total caloric intake, and produces weight loss of about 8 kg (17.6 pounds, or 6.8% body weight) over one year. About 43% of patients taking pramlintide for a year lose > 10% of their body weight. Davalintide has been shown to produce up to 15% weight loss by four weeks, compared to 10% with pramlintide, and decreases caloric intake by 34% - presumably by increasing satiety. Full- scale trials with davalintide are ongoing. Amylin and PYY have been shown to produce 18% weight loss in DIO rats in just two weeks. Amylin + leptin + PYY produces almost 25% weight loss (as presented at ADA this year), and animals treated with this combination have no detectable fat mass left after treatment.
  • Leptin replacement causes weight loss in leptin deficient animals, but does not seem to have an effect in non-deficient obese people or animals. Rosenbaum et al., 2005 did suggest that leptin replacement after weight loss helps patients maintain weight loss and reverses changes in the brain associated with food desire after weight loss. Can there be combination therapy with leptin? Amylin and leptin combined have been shown to produce dramatic and sustained weight loss that is not explained by hypophagia.
  • Dr. Skyler thinks that peptide therapies for obesity carry a relatively low likelihood of CNS side effects or off-target toxicities (as compared to, for example, the CB-1 antagonist class). He thinks that combination therapy can leverage multiple satiety and adiposity signals that are key in the regulation of food intake.

Questions and Answers

Q: Were there any issues with compliance in these trials?

A: In theory it could be a problem, but people who are interested in losing weight are happy to take injections to lose this degree of weight. I’m not familiar with the side effect details of these trials.

GASTRIC BALLOON: CLINICAL EXPERIENCE AND METABOLIC EFFECTS

Vincenzo Bacci, MD (Universiti de Roma, Rome, Italy)

Dr. Bacci discussed gastric balloon technology, which to our understanding is used with some frequency in Europe and Asia but is much less common in the US. The balloons are a small plastic reservoir that can be filled with saline solution or air, expanding to take up space in the stomach and promoting weight loss. The balloons can only be used for a period of six months, and some studies have shown that they produce a BMI decrease of about 3.2 kg/m2 (7 pounds) in that time, and an average A1c reduction of 0.5% (from 7.5%, in one study). Weight is usually regained after balloon removal, although Dr. Bacci thinks that a subset of patients may show exhibit weight loss maintenance after removal.

  • Gastric balloons have been used since the 1980s to control weight. Early models were quite poor. The balloons can be used for six months. They can be used as an intermediate weight loss step before more invasive surgery or for control of co-morbidities of obesity. 98% of patients have nausea for one to two days, and 75% have regurgitation for two to three days. About 1.8% had early removal voluntarily, and 0.9% for pain. Persistent nausea has been seen in ~9% of patients.
  • Balloons are moderately effective in reducing weight. Imaz et. al., 2008 showed a -3.2 kg/m2 BMI decrease and 6.7 kg (14.7 lbs) weight loss vs. placebo. This is much less than weight loss from bariatric surgery, but enough for effects on comorbidities to be seen. Weight is often regained after balloon removal. Genco et al., 2008 showed greater weight loss with balloon versus diet after six months, but greater weight regain with the balloon at 24 months. Balloon treatment can be effective in resolving diabetes, sleep apnea, and hypertension, and improves both A1c and insulin resistance, as expected. They saw an A1c decrease from 7.5% to 7.0% at six months.

Questions and Answers

Q: What do you think are the long-term benefits of putting a balloon in for six months?

A: There is a subset of patients that will show persistent weight loss—maybe 20-30%. There is an idea to do sequential two or three balloon insertions. I’m not aware of long-term weight loss data.

METABOLIC/BARIATRIC SURGERY IN THE TREATMENT OF TYPE 2 DIABETES

Henry Buchwald, MD, PhD (University of Minnesota, Minneapolis, MN)

Dr. Buchwald discussed the results of a soon-to-be-published meta-analysis researching the effects of bariatric surgery on weight loss and diabetes resolution. The work included studies from 1990 to 2006, comprising a total of 621 studies and 135,246 patients. Overall, the average excess weight loss seen was 56% (46% with banding, 55.5% with gastroplasty, ~60% with gastric bypass, and 63.6% with biliopancreatic diversion or duodenal switch). 76.8% of diabetes cases were resolved, 83.9% were either resolved or improved, 1.1% were worsened, and 14.9% remained unchanged. Patients undergoing surgery saw an average A1c reduction of 1.3%. Dr. Buchwald also discussed some experimental procedures to treat diabetes, like the EnteroMedics VBLOC technology and bypass surgeries that produce resolution of diabetes without weight loss. Notably, he also mentioned the potential role of implantable pumps in delivering incretin infusions.

  • Currently there are 300,000 bariatric procedures performed annually worldwide. He discussed gastric banding, biliopancreatic bypass, and the duodenal switch. He highlighted the surprising efficacy of bariatric surgery in resolving hyperglycemia.
  • A meta-analysis soon to be published in the American Journal of Medicine examined the effect of bariatric surgery on patients with type 2 diabetes. The study included studies from 1990 to 2006, comprising a total of 621 studies and 135,246 patients. 44.4% of the patients were from Europe, and ~43% from North America. The mean age was 40.2 years, the mean BMI was 47.9 kg/m2, and 22.3% of the subjects had type 2 diabetes. Overall, the average excess weight loss seen was 56% (46% with banding, 55.5% with gastroplasty, ~60% with gastric bypass, and 63.6% with biliopancreatic diversion or duodenal switch). 76.8% of diabetes cases were resolved, 83.9% were resolved or improved, 1.1% worsened, and 14.9% were unchanged. Patients undergoing surgery saw an average of a 1.3% A1c decrease. There was a more pronounced effect on glycemic control and the resolution of diabetes with the more intensive of the surgeries, including a 95.1% resolution of diabetes with the duodenal switch or biliopancreatic diversion.
  • He discussed several experimental procedures in diabetes treatment, include a duodenal bypass while preserving the whole stomach, which doesn’t produce much weight loss, but does ameliorate diabetes. Limited data show significant reductions in blood sugar with this procedure—biliopancreatic diversion may reduce fasting glucose levels from 222 to 117 mg/dL over eight months even in patients who are not morbidly obese, but it does not cause much weight loss. Similarly, a sleeve placed in the duodenum of diabetic rats has been shown to ameliorate diabetes, and the disease returns when holes are placed in the sleeve. Another experimental approach is neuromodulation—Dr. Buchwald used the EnteroMedics device as an example. He also mentioned the potential role of implantable pumps in delivering incretin infusions.

Questions and Answers

Q: How low would you go with the age for bariatric procedures?

A: Well, there are now pediatric bariatric surgeons. I’ve operated on adolescents for a long time—I’ve even operated on a boy who was nine years old. The team approach is important, and you have to have nutritionists, psychologists and others involved. Yes, I think this can be extended to adolescents.

Ongoing Studies – Sponsored by NiliMEDIX

IMPLICATIONS OF CGM STUDIES – WHERE ARE WE NOW?

Irl B. Hirsch, MD (University of Washington, Seattle, WA)

This was a particularly significant and memorable talk by Dr. Hirsch, notable for its straightforward and well-argued message. In brief, Dr. Hirsch is saying that we already have seen enough to understand CGM and make the case for using it. We know that it can be very effective, and we know that success has a behavioral component. In that sense, it’s no different from SMBG or insulin therapy. Just because the behavioral component makes it harder for conventional trials to deliver ‘clean’ answers, doesn’t mean we are paralyzed waiting for proof, and can’t move forward with the clinical work.

The message is that academia and diabetes organizations are applying artificial and pedantic hurdles to the acceptance of CGM, by insisting on multiple randomized controlled trials, and incorrectly interpreting trials that had “negative results” instead of learning from them. Dr. Hirsch asserts that we don’t need any more trials like STAR1 or the JDRF CGM trial– though there are a dozen or more underway. He pointed out that in addition to improved accuracy, lower costs and an economic incentive for clinicians to support and train on pumps and CGM are needed in the US.

His approach was a welcome counterpoint to Dr. Bolinder’s sobering talk earlier during the conference, which made the case that we have formally proven very little about CGM to date. It is important to understand that Dr. Bolinder was talking realpolitik, which is that the trials are necessary for reimbursement and policy makers, despite physicians’ enthusiasm for CGM.

Dr. Hirsch’s pragmatic perspective is welcome in the debate about CGM. Notably, Dr. Hirsch said that the form and subject of this debate is ‘déjà vu all over again’ – we had it in the late 1970’s with SMBG when it was considered inaccurate, costly and only suitable for some patients.

  • Dr. Hirsh posed the question, “Do we judge CGM by different standards than other tools for the treatment of diabetes?” In the past, we’ve never had a randomized controlled trial for insulin, urine glucose testing, or SMBG in type 1. Yet we are getting dozens for CGM.
  • CGM accuracy is getting better and it will continue to improve. In 2005, the original Medtronic sensor accuracy was 60-68% of readings within 20% of the actual value. In 2009, DexCom and Navigator work showed 67-86% within 20%.
  • In addition to better accuracy, we need lower costs and an economic incentive for the clinician for support and training on pumps and CGM (in the USA).
  • The JDRF trial showed us that with patients who are willing to change their behavior, glycemic control will definitely improve with CGM. Unfortunately, some people simplistically and erroneously quote the result as, “CGM doesn’t work for kids”. Successful diabetes management has always required a large amount of effort from patients and their families. So one day, because we will avoid a lot of human factors, the artificial pancreas will be very helpful.
  • In the STAR1 CGM study, there was no difference in A1c between the sensor and control groups, so the trial is remembered as a ‘negative result’ - but the takeaway message was identical to the JDRF – those who wore the sensor 60-80% of the time had a 0.7-0.9% reduction in A1c. The trial taught us that it wasn’t that the tool didn’t work, it was just that the patients weren’t interested in using it – i.e., a better form factor and easier ways to use were needed.
  • For some reason, Dr. Hirsch said, academia expects CGM to have randomized controlled trials that look like trials for hypertension drugs. CGM trials are much more complex since we have to factor in the behavioral issues.
  • So, in Dr. Hirsch’s words…..
    • We don’t need major diabetes organizations dismissing the technology because of the lack of a randomized controlled trial.
    • We don’t need major diabetes organizations dismissing CGM after we’ve already done a definitive randomized controlled trial.
    • We don’t need another STAR1 or JDRF-like trial. We’ve done that.
    • We don’t need physicians stating that there is no place for CGM in their practices because of lack of good study data (including the JDRF trial).
  • We need more research addressing behavioral issues – such as teenagers sabotaging their control; and more work on dealing with mental health issues which affect behavior in type 1 and type 2 diabetes.
  • Today, we are getting much closer to true closed loop systems, so hopefully the light is at the end of the tunnel.

Questions and Answers

Q: (Dr. Moshe Phillip) Why focus on the JDRF CGM trial so much?

A: JDRF is the first six-month trial – that’s like the gold standard for a drug. So that’s why I’m focusing on it. We will get 12-month data at ADA in June.

Q: When we do trials, I don’t like having to select only people who will comply – that’s artificial. For example, adolescents just don’t comply.

A: These are complex issues. STAR1 took all comers, but if you look at the science it was a negative trial, even though the people who wore it did very well. But that’s not what it is going to be remembered for. To get patient and provider buy-in we have to pick a study population where you are more apt to get success. That’s the reality of what you have to do to get reimbursement. And reimbursement is what’s driving the entire topic right now.

Q: (Dr. Satish Garg): Have you had a crack at creating a behavioral tool that works?

A: We have talked about this – we really need this. We need someone from the psychosocial community to come up with a tool, I’ve been using A1c over time. This is crude, but if someone is over 9% they are not doing the basics, so I am not sure they are the right person for CGM.

A: Final comment – didn’t we have the exact same discussions for SMBG about cost and applicability in the 1980’s?

A: (Jay Skyler) – Yes! But it was the late 1970’s!

PUMP SHUT OFF TO PREVENT NOCTURNAL HYPOGLYCEMIA

Peter Chase, MD (Barbara Davis Center, Colorado, United States) and Bruce Buckingham, MD (Stanford University, Palo Alto, CA)

This was a joint presentation between Dr. Chase and Dr. Buckingham who are collaborating on some important clinical aspects of developing a practical artificial pancreas. Together, they discussed the use of an algorithm to shut off insulin delivery and avoid hypoglycemia.

The daytime studies showed a 57% reduction in hypoglycemia, while the nighttime work showed an 84% hypoglycemia reduction. The algorithm uses a more sophisticated ‘voting system’ in which five prediction algorithms contribute to the insulin shutoff decision-making process. Notably, there was no rebound hyperglycemia or ketonemia from a pump suspension of 90 minutes, and no severe hypoglycemia.

This is a very impressive result, and takes ‘Low Glucose Suspend’ a step further into the prevention realm. Note that at no time does this algorithm dose insulin – it’s all about improving patient safety. It’s hard to imagine a rationale that would not advocate reducing nocturnal hypoglycemia by 84%.

More broadly, we hope that while 2008 was the year of algorithm development, 2009 will be the year of inpatient trial results, and 2010 the year of outpatient trial results.

  • Dr. Chase outlined a series of closed loop hypoglycemia prevention studies. First were daytime and nighttime studies in which an algorithm shuts off the pump to prevent hypoglycemia (this is different from Medtronic low glucose suspend, which doesn’t avoid hypoglycemia, just tries to mitigate the effects). The next step anticipated is software and hardware safety data studies, culminating in home studies of 48 adults, and finally, studies with children. Truly a smart pump – sounds amazing from a patient perspective.
  • In this daytime work with 15 patients, the team consistently induced hypoglycemia then evaluated the safety of algorithms that predicted incipient hypoglycemia and turned off the pump for 90 minutes. There were two hospital visits – the first to (carefully!) induce hypoglycemia (<60 mg/dl), the second to assess whether a hypoglycemia prediction alarm could prevent patients going below 60 mg/dl. The patients were in effect acting as their own controls. The algorithm was called a ‘statistical prediction alarm’.
  • In the first visit 85% of the patients went below <60 mg/dl, while the predictive algorithm was able to reduce this to 28% of the patients in the second visit, a 57% reduction in the incidence of hypoglycemia. The team also studied the standard FreeStyle Navigator low glucose alarm, which was characterized as less successful at prediction, as it’s a straight linear alarm. Notably, there was no rebound hyperglycemia or ketonemia from a pump suspension of 90 minutes, and no severe hypoglycemia.
  • Dr. Chase showed an interesting picture of the ‘Florence’ system, consisting of an Abbott Navigator sensor/transmitter, the “Aviator” insulin pump, a Companion handheld combined controller, and a control algorithm device (CAD). No more details were available at this stage, but it looks like Abbott’s artificial pancreas research system. It’s compact and we assumed that it would be used for the outpatient trials, although this wasn’t clear.
  • Dr Buckingham took over to describe the nighttime studies. In order to induce hypoglycemia at night, basal rates were increased 5% to 25% every 90 minutes based on current glucose and glucose rate of change. The idea was to create consistent lows in order to find out if they could be prevented. The goal was to prevent hypoglycemia 80% of the time. Hypoglycemia was obtained in >80% of study nights. 90-minute adjustments were chosen since it takes 60 minutes to see a meaningful glucose change after changing the basal levels and then 30 minutes more to reach a stable value.
  • The prediction system used five-way voting. Five prediction algorithms (statistical linear prediction, Kalman filter, hybrid infinite impulse, numerical logical algorithm, linear projection) ran independently. As blood glucose decreases, more of these alarms go off. The overall system was triggered when any two out of five algorithms were predicting impending hypoglycemia. This seemed to correspond to a prediction horizon of 50 minutes.
  • In trials with 14 patients, 84% of all hypoglycemia events were avoided, or in 75% of the subjects. This seems a quite significant result. Some nights there was more than one basal shut off. There was a 30 minute mandatory basal restoration after the 90 minute shut off, but it appears that in some cases, it would have been better to have a two hour shut off if blood glucose were still low or declining.

ISSUES WITH CLOSING THE LOOP: CRITICAL APPRAISAL

Tadej Battelino, MD (University Medical Center, Ljubljana, Slovenia)

This was a somewhat high level overview of closing the loop, covering the different aspects of system design and some of the inherent problems in the technology we have today.

  • The artificial pancreas consists of three components:
    • The Sensor, whose considerations include:
      • Location (sub-cutaneous, intravenous, non-invasive). Currently subcutaneous placement seems the only practical choice. Intravenous placement (placed in the right atrium of the heart) works well for up to a year, though it’s highly invasive.
      • Measurement method - glucose oxidase (DexCom, Medtronic, Abbott), microdialysis (GlucoDay), new infrared fluorescence (under development). Glucose oxidase was reported to be more practical than microdialysis for ambulatory use.
    • The Controller algorithm of which there are three main types:
      • PID (Proportional-integral-derivative) control algorithms as used by Medtronic. PID looks at the current glucose level, it’s rate of change and how much has been dosed historically and then determines the next dose – so it’s continually ‘homing’ in on the target. It’s an older system, usually seen as less successful than MPC. But there is trial data that appears successful.
      • MPC (Model Predictive Control) algorithms create a model of the body and use it to predict the effect of insulin on glucose level. It repeatedly solves this model to recommend the right course of action at any point in time.
      • MD-Logic a new approach from Tel-Aviv for which details are sparse. It’s a fuzzy logic learning approach tailored to the individual, so appears less practical.
    • Insulin delivery routes:
      • Subcutaneous (classic pump)
      • Intraperitoneal - done only in France, can be an implantable pump or a port, it’s a great idea, but there are practical problems with insulin stability.
      • Intravenous - this is largely a historical concern. The Biostator, which keeps getting mentioned at this conference was a practical closed loop artificial pancreas that worked with IV insulin and IV glucose.

Sensor delay is a problem, and at least 15 minutes is physiological. Algorithms can account for it, but is the ever-present delay still lessens the responsiveness of the system. That said, Sensors are continually improving and are getting more accurate, longer lived, with the possibility of implantation.

  • Delays are also a problem with insulin delivery. Insulin can take 30 minutes to have a first effect and two hours to reach peak – making it harder for algorithms to adjust to rapid glucose changes.
  • Faster insulins are in the pipeline, but are not being tested quickly enough because of commercial considerations. Dr. Battelino knows of a company that has an ‘ultra-short acting’ insulin that acts within 10 minutes and has shorter duration. This could be the fastest way of reducing the lag time in the whole loop. “Please help to lobby for a clinical trial with ultra short acting analogs”. This reminds us of the enthusiasm for Biodel’s VIAject we saw at AIDPIT last year.
  • There are dual infusion systems under consideration including insulin/Symlin or insulin/glucagon delivery systems. Dr. Battelino was very skeptical about these two approaches. He felt that pumping glucagon and insulin close to the same time was ‘difficult to imagine’ since it was not physiological.

REALISTIC EXPECTATIONS OF A CLOSED LOOP; THE FDA PERSPECTIVE

Arleen Pinkos, BS, MT, ASCP (US Food and Drug Administration, Washington DC)

Arleen Pinkos is a strong champion of the artificial pancreas, and is working to see it approved as quickly and efficiently as possible. However, it’s not going to be easy, and in this presentation she gave us many ‘clues’ as to where the problems could lie.

The presentation reiterated that a great deal of careful due diligence and communication/consultation is required to get to approval of the closed loop. Specifically, she said, we learned not to over-reach on the intended use, to avoid complexity in hardware and software, to focus above all on safety and risk mitigation. It seems that we will get to the full artificial pancreas by a series of careful steps.

Pinkos did an excellent job of representing the position of the FDA, while advocating for patients: “I love CGM, it’s a great device. If I were a diabetic I would use CGM. I lost my mom to diabetes. If she were still alive I would have her use one.”

  • The FDA is seeking to formally establish that “Probable benefits… outweigh any probable risks…. in intended and probable use”. So we should be thoughtful about ‘intended use’, since this sets the bar for safety and efficacy.
  • Getting the right trial protocol is 90% of the job, so it’s important to talk to the FDA as early as possible. When the FDA evaluates clinical trial design, it’s important that the study design matches indications (often this is not done). This includes representative populations, conditions of use, what is or what is not to be achieved. There should be a data collection scheme during all times of use, tight definition of endpoints that are clinically relevant, and it needs to be unbiased.
  • To approve the artificial pancreas, we need to show that benefits outweigh risks. But it’s hard to be more specific. Like the many models of cars, there are different features and various performance standards. But minimum safety requirements are critical. Like brakes on a car, we need built in safety, and we need to know what we can do if they fail.
  • “If you claim the world, you have to prove the world”. So tighten expectations for product. Focus on “First, do no harm”. But you do need to provide ‘significant’ clinical benefits. Because of the ‘study effect’ – borderline improvements in a study translate to no improvement in the real world.
  • There are four elements to proving safety and efficacy - overall design, human factors, adequate risk mitigation, drug safety. We discuss the first three here:
    • Overall design
      • Select reliable components – previously FDA approved is preferred
      • Software validation has to be done in the context of the entire system

      • Integration of multiple components should consider all feasible unintended consequences.
    • Human Factors
      • There has never been a system where human factors haven’t been the most important issue. The potential for operator error still exists and remains the largest source of error.
      • The design has to account for variable skills and abilities, teenagers, geriatric users.
      • There is low compliance - one third discontinue use.
      • People don’t always understand the consequences of their actions.
      • We can assume that people will ignore alarms and not follow directions (e.g. not calibrating at the correct time).
    • Safety mechanisms have to be built into any artificial pancreas design for risk mitigation. “I like the idea of remote monitoring”. It’s important to lower complexity, since more complexity for either patients or physicians leads to more problems. It’s to be expected that patients won’t read the manual, or physicians won’t follow guidelines on patient selection or the setting of target ranges.

FEASABILITY OF INTRAVENOUS CLOSED LOOP WITH IV MD-LOGIC SYSTEM IN THE DIABETIC SWINE

R. Nimri, MD (Schneider Children’s Medical Center of Israel, Tel Aviv, Israel)

We first learned something about the MD-Logic closed loop algorithm earlier in this conference. As far as we can tell, MD-Logic is a kind of fuzzy logic system. We understand that the system learns, but not automatically – it has to be ‘taught’ manually.

This insulin/glucagon closed loop work in pigs reminds us of that of Dr. Ed Damiano, (who did it with sub-cutaneous infusion), who has now moved on to human trials. The learnings from dual insulin/glucagon infusion are (i) that the system can be very effective at preventing hypoglycemia and that the system can be tuned to more aggressive at reducing hyperglycemia because of the ‘safety net’ of glucagon.

Dr. Nimri likened insulin and glucagon to hot and cold water. She showed a picture of a happy pig taking a shower in which hot and cold water were being mixed to get just the right temperature. Of course, this is not exactly physiological.

It’s interesting to think about the effect of having insulin and glucagon administered closely together as in the more aggressive case above. Also, it might be that frequent use of glucagon could exhaust the liver, leading to ‘failure of the emergency brakes”.

  • This system uses the MD-Logic closed loop algorithm, and a combination of intravenous insulin and glucagon. Some work has shown that glucagon can be stable at body temperature for seven days as required in this application. IV closed loop work was first performed in 1963, and 1974 - the introduction of the Biostator proved it in a clinical setting.
  • Pigs are a good model for humans. They have similar glucose homeostasis, similar PK/PD, similar insulin and glucagon structure, the GI tract is similar, and pigs are omnivores, so meals can be similar.
  • In open loop PK/PD experiments, it was established that insulin took 20-30 minutes to take effect and glucagon worked almost immediately.
  • In fasting closed loop sessions, blood glucose came down smoothly into the target range, and the time spent in the hypoglycemia range (50-70 mg/dl) was <1%. Glucagon was only used by the algorithm in two out of twelve sessions – almost like an emergency brake. By lowering the target range, the controller was more aggressive with both insulin and glucagon and dramatically shortened the time to reduce blood glucose into the target range.
  • In closed loop mealtime sessions, the post-prandial peak was reduced from 245 to 180 mg/dL, with time in the hypoglycemia range of only 4%. The algorithm was even more aggressive, using frequent boluses of glucagon. The control was very tight – an average blood glucose of 126 mg/dl, spending 85% of the time in the target range. We noted that there was an oscillatory appearance to the blood glucose data as glucagon was dosed to maintain the target.
  • In summary, glucagon can be used for hypoglycemia avoidance, but also as a counter-regulatory mechanism for tighter control.

Oral Presentations I

RESULTS OF THE JDRF CGM TRIAL IN PATIENTS WITH BASELINE A1C LEVELS < 7.0%

Lori Laffel, MD, MPH (JDRF Study Group)

A welcome result from the JDRF Study Group, showing that well controlled type 1s (A1c <7%) on CGM can maintain that control while greatly reducing time spent in the hypoglycemia zone, effectively improving their safety and reducing glycemic variability.

In this 26-week study, there was a dramatic decrease in time spent below 70 mg/dL per day in the CGM group (37 minutes/day less) although this wasn’t statistically significant because of technical problems with outliers – certainly, however, it’s a clinically relevant finding. There was also a significant reduction in time spent <60 mg/dL.

A1c did not change in the CGM group, but worsened in the control group, leading to a difference of 0.3% after 26 weeks. The percentage improving A1c by 0.3% or more was much higher in the CGM group versus the control group, and conversely the percentage worsening A1c by 0.3% or more was much lower. The risk of severe hypoglycemia wasn’t different, despite the lower A1c in the CGM group, – we would not necessarily have expected this given the low number of events..

At ATTD, many presenters have commented that patients in control are good candidates for CGM, since they tend to display more concordant behavior. This trial reinforces that belief and at the same time should reassure reimbursement bodies that they can still see further health and safety improvements with CGM even in well-controlled patients. We always say that it’s not just the A1c, it’s the quality of the A1c that matters …

  • The previous JDRF study included 322 subjects, and concluded that control was improved in adults without increasing hypoglycemia. In adolescents and children, usage was tightly linked to improvements in control.
  • The primary study excluded individuals with A1c <7% because lowering A1c was the primary goal. But there was a secondary cohort aimed at assessing the efficacy and safety in patients with A1c <7% at baseline.
  • For the secondary cohort, the primary outcome was the time spent in minutes/day with CGM glucose <70 mg/dL over a 26-week period. Secondary outcomes included the time spent <70 mg/dL at 13 weeks, time spent <60mg/dl and < 50mg/dl, change in A1c, percentage of people maintaining A1c in the target zone (<7%), and percentage of subjects increasing or decreasing their A1c by 0.3%.
  • 129 participants who used CGM for six out of seven days/week during a blinded run in phase were selected. They were randomized to the CGM group or control. The control group used blinded CGM at baseline, at 13 weeks, and at 26 weeks for comparison. As with the primary trial, there was a lot of education and follow up, which was >95% completed. 80-90% of the subjects used pumps.
  • The time spent in hypoglycemia range decreased from 91 to 54 minutes/day in the CGM group, compared to 96 to 91 minutes/day in the control group. That’s 37 minutes/day of hypoglycemia avoided with CGM use. Dr. Laffel said that because of some technical problems with outliers, this difference wasn’t statistically significant (ed. note – this surprised us), although it was characterized as “clearly clinically significant”. Combining the data at 13 and 26 weeks gave a significant change, as did truncation of the outliers.
  • The CGM group reduced time spent <60 mg/dl from 40 minutes/day to 18 minutes/day compared to 40 to 35 minutes/day in the control group. This result was statistically significant. Time spent <50 mg/dl decreased from 7 minutes/day to 4 minutes/day in the CGM group compared to a decline from 9 minutes/day to 8 minutes/day in the control. Of course, Dr. Laffel said, there weren’t really enough events here to be significant.
  • The CGM group maintained their average A1c (6.4%) for the entire 26 weeks, but the control group’s A1c rose, despite being matched in attention/follow up. The difference in the groups was 0.34% at 26 weeks. 88% of the CGM group maintained target A1c compared to only 63% of the control group.
  • In terms of changes in A1c 31% of the CGM group were able to reduce their A1c by more than 0.3%, compared to only 5% of the control group. On the other hand, A1c increased by 0.3% or more in 28% of the CGM group, compared to 52% of the control group.
  • Severe hypoglycemia was equivalent across the groups, despite the lower A1c of the CGM group. We were initially a little surprised to see that there was no difference in severe hypos although this makes sense in light of the time at <50 mg/dl per day – very little.

Questions and Answers

Q: How did the results vary with different age groups?

A: A subgroup analysis showed that results for those <25 years were similar to those participants >25 years old.

Q: What were the usage rates and how did they vary with age?

A: 80% of the population used the sensor for at least six days/week. In the first 13 weeks it was 79% and the last 13 weeks it was 60%. The percentage that used the sensor for six or more days/week was 79% for over 25 years, 53% for 15-24 years, and 65% for 8-14 years. These figures are all higher than in the previous trial.

Q: Why wasn’t it significant?

A: A 37-minute/day difference is clearly a huge drop. We could make it significant if you removed statistical outliers that distorted the result. The study was in fact powered to show a difference of 27 minutes.

Q: What’s next?

A: We are moving to a crossover phase for the control group after the six months point, which we will follow. But both groups will be put back into routine care, rather than the intensive follow up.

SLEEP APNEA AND DIABETES –TWO INTERACTING EPIDEMICS?

Joachim Ficker, MD (Nurnburg, Germany)

Both diabetes and obstructive sleep apnea (OSA) are epidemics. Obesity is a risk factor for both diabetes and OSA, OSA is a risk factor for diabetes, and both OSA and diabetes interact as risk factors for cardiovascular disease. The IDF recommends OSA screening for type 2 patients.

  • Obstructive sleep apnea (OSA) is an epidemic, affecting 5-15% of the population. Overnight screening devices can diagnose OSA easily. CPAP positive airway pressure) is the standard treatment for OSA.
  • Obesity is a risk factor both diabetes and OSA. The question is – could diabetes and OSA be related?
  • Epidemiological data suggests there is a linkage between diabetes and OSA. Perhaps OSA is a risk factor for diabetes. 70,000 nurses were grouped into non-snoring, occasional snoring, and snoring. The snoring group had a 2x risk of developing diabetes within 10 years. CPAP in patients with OSA but not diabetes improves insulin sensitivity. This effect is more pronounced in the less obese population.
  • OSA and diabetes interact as risk factors for cardiovascular disease.
  • The IDF consensus statement says that people with OSA should be screened for diabetes and cardiovascular risk. People with type 2 diabetes should be screened for OSA, particularly if they have classic symptoms - snoring, tired after a night of sleep, high blood pressure.

EVALUATING RATE OF CHANGE AS AN INDEX OF GLYCAEMIC VARIABILITY USING CONTINUOUS GLUCOSE MONITORING DATA

C. Whitelaw, MD (Kings College Hospital, London, UK)

The work from Dr. Whitelaw’s group looked at measures for glycemic variability, concluding that they were all useful, but didn’t yield much insight as to which were the best to use and why.

  • There are multiple indices for glycemic variability. There has been controversy over which index is the most valid.
  • The most widely accepted are
    • Standard deviation (SD)
    • MAGE (mean amplitude of glycemic excursion) – the drawback is that MAGE weights the largest fluctuations higher.
    • CONGA (Continuous Overall Net Glycemic Action)
    • Average rate of change (ARC) – excessively influenced by small and insignificant fluctuations
  • Using CGM data, physicians made a clinical assessment of glycemic variability on a scale of from 0-10 (from perfectly stable to highly labile). This score was known as CLS (clinical lability score).
  • All the indices above show a strong correlation with CLS except ARC. Also, ARC doesn’t show significant differences between healthy and type 1 subjects. If we smooth ARC with 50 minute smoothing, ARC has good correlation with CLS.

LEARNING-TYPE MODEL PREDICTIVE CONTROL (L-MPC) IN ARTIFICIAL PANCREATIC BETA CELL

Eyal Dassau, PhD (University of California Santa Barbara, CA)

This work demonstrates the lightning progress of closed loop algorithms, and was one of the most exciting closed loop presentations at ATTD this year. The new L-MPC (learning-type model predictive control) approach upgrades the MPC algorithms that are already showing strong performance for the artificial pancreas.

Dr. Dassau has a process engineering background and has created an algorithm that learns about typical daily glucose patterns. A supervising algorithm, called the ILC, learns about an individual’s lifestyle, and then controls the settings for the MPC to achieve superior performance than just the MPC alone. This approach effectively learns about mealtimes and minimizes user intervention annulling the need for meal announcements. The in silico results show great control, and excellent robustness to meal variations or subject variability. As with all simulations, we have to be cautious about extrapolating to real life.

The next step is to test L-MPC in the clinic. Dr. Dassau mentioned that in this setting we can also fine- tune L-MPC more closely to each subject, which was not done in the simulation.

  • The MPC algorithm uses a model of the body to repeatedly predict the next optimal move to achieve the desired glucose ‘set-point’. Closed loop experiments show good performance with MPC algorithms, but we often see deviations or ‘offsets’ from the glucose set- point - post-prandial hyperglycemia being an example.
  • The concept behind L-MPC is that offsets in previous days can be used to update the set-point in the current day. In other words, we can learn, for example, that there is usually some high glucose after lunch and adjust the MPC algorithm to shoot for a lower glucose target at those times, giving a compensating effect.
  • The architecture of the L-MPC algorithm is a two-tier approach, incorporating a classic MPC algorithm (termed the ‘local controller’). Iterative learning control (ILC) sits on top of the MPC and is used to update the MPC glucose set-point. The ILC receives the glucose target from the physician, monitors the daily performance of the MPC and varies the MPC set point accordingly to get the best in target performance. The ILC takes less than 10 days to converge, and is continually updated with previous day’s performance. The sampling frequency is five minutes.
  • Initial in silico testing of L-MPC with standard meal sizes gave a remarkably flat glucose profile. But the MPC set-point was changed a lot during the day – sometimes to be very low or even negative! Note that the MPC has built in safety constraints that shouldn’t allow hypoglycemia, but the team imposed some further constraints on the ILC, to reduce performance but make it safer.
  • Adding these constraints and simulating variations in (unannounced) meal sizes, the L-MPC algorithm had superb performance, spending 96% of the time in normoglycemia! Of course, this is a simulation and not real life. But we gather that the system is tolerant of missed meals or unusual boluses, since the MPC will always do its job of regulating glucose to be in range.

PASSIVITY BASED CONTROL OF BLOOD GLUCOSE: AN ALTERNATIVE CLOSED-LOOP STRATEGY FOR THE TREATMENT OF TYPE 1 DIABETES

H. Cormerais, PhD (University of Rennes, France)

It’s great to see more process control engineers entering the artificial pancreas fray with new ideas for algorithms. Dr. Cormerais is introducing Passivity Based Control, which has been used with success elsewhere. This presentation was intriguing, but more work needs to be done before we can properly compare the performance of this approach with PID and MPC approaches.

  • Currently, the two main closed loop algorithm approaches are PID and MPC.
  • The presenter proposed using Passivity Based Control, an increasingly common approach in non-linear process control. The presenter’s specific version is called Error Dynamics Shaping (EDS).
  • A simulation of this algorithm with SC insulin and IV blood glucose measurement results in blood glucose controlled between 75 and 159 mg/dL with 100g meals.
  • The team is building a subcutaneous sensor model. They also plan to add in extra time delays to account for insulin absorption.

 

NONINVASIVE SKIN AUTOFLUORESCENCE, AN INDEPENDENT PREDICTOR OF COMPLICATIONS IN DIABETES, ITS RELATION TO GLYCEMIC VARIABILITY

Andries Smit, MD (University Medical Center, Groningen, Netherlands/DiagnOptics)

Skin autofluorescence (AF) can be measured very simply and easily, and turns out to be able to detect type 2 diabetes. It’s a sort of cumulative damage marker, Dr. Smit said.

In studies, skin AF was reported to be a better diagnostic tool for type 2 diabetes than FPG or even A1c (we note neither fasting blood glucose nor A1c is a standout diabetes diagnostic tool). So it’s a great, quick way of screening for diabetes without a blood test, they said.

  • Skin autofluorescence (AF) in the 400-600 nm range can provide a measure of advanced glycation endproducts (AGEs). AGEs are a carrier of ‘metabolic memory’ – they denote cumulative oxidative stress and cell injury.
  • Skin AF shows a good correlation with diabetes complications in type 2. This technique had additional predictive value on top of the UKPDS risk score in type 2 diabetes. Skin AF can also predict mortality risk.
  • Skin AF is not affected by glucose levels, and so doesn’t change before and after the OGTT. But there is a very small correlation with A1c over three years.
  • Skin AF is superior in the detection of diabetes to FPG and A1c (using OGTT as the gold standard for diagnosis). Skin AF yielded a false negative rate of 18%, which is reduced to 9% if serum creatinine levels are taken into account.

NON-INVASIVE GLUCOSE MONITORING IN PATIENTS WITH TYPE 1 DIABETES: REPEATABILITY IN THE SAME SUBJECTS

Andreas Caduff, PhD (Solianis Monitoring, Zurich, Switzerland)

Dr. Caduff discussed a new approach to spectroscopic non-invasive glucose monitoring being pioneered by Solianis Monitoring AG. Spectroscopic systems are extremely sensitive and have been plagued with interference from a variety of sources, but the use of addition sensors to detect interfering factors may allow a detection device to compensate for them and produce accurate readings. Initial results using the sensor combination in a controlled setting have been promising, with 86% of readings in the A or B region of the Clarke Error Grid. Like all diabetes technology the more relevant question is performance in the real-world environment. Dr. Caduff will be presenting additional data at ADA 2009, and hopes to be ready to start clinical validation of the system in 2010.

  • The use of non-invasive devices in uncontrolled conditions causes interference. The use of the ‘multi-sensor concept’ may allow for the measurement of interfering conditions and compensation for them. The use of sensors with different mechanisms can monitor various factors that interfere with the main sensor independently and enable the system to compensate. Dr. Caduff‘s team has put several sensors on a single device.
  • He described a study utilizing the combined sensor in ten patients. The study went for two days, and the sensors were attached to subjects’ upper arms. Each day consisted of a calibration period followed by a glucose challenge. On the first day, two glucose excursions were simulated, and on the second day, there was a single glucose spike. The sensor package included components to measure. Overall, the MARD was 27.3%, the error grid A+B was 86%, and R2 was0.68. While these initial results are promising, the system needs to be tested with a larger subjectpopulation and in more realistic and uncontrolled settings.
  • The results of an additional outpatient study will be presented at ADA 2009 (we’ll be watching out for those), and a next generation model will be ready for studies in 2Q09.

Questions and Answers

Q: When will the system be available to clinical trials?

A: We think we’re getting to the level where we can get it together with a better algorithm and clinical validation. 2010 is when we would like to start with clinical validation.

Oral Presentations II

TELEMEDICAL ASSESSMENT OF PHYSICAL ACTIVITY AND EATING HABITS TO IMPROVE INSULIN THERAPY IN CHILDREN AND ADOLESCENTS WITH TYPE-1 DIABETES MELLITUS

Ralf Schiel, MD (Seeheilbad Heringsdorf, Rostock, Germany)

Dr. Schiel discussed a new telemedicine intervention to help young people with diabetes improve their insulin therapy by better monitoring physical activity and eating habits. The assessment used a wireless sensor integrated in a mobile phone that monitored the kind, intensity, and duration of physical activity, and a camera that could be used to take pictures of food for later analysis by a dietician to improve carbohydrate estimates. The exercise measurement technology can identify ten exercises with near-perfect accuracy. The technology was used in two patients at a summer camp over four and nine days. There were large differences between the measured assessments of the duration of physical activity and patients’ subjective reporting. Dr. Schiel thinks that technology could be used to enable better calculation of physical activity and eating.

Questions and Answers

Q: How do you download the information from the mobile phone?

A: At present, data is transferred to a central server and then to my office. I think this is only temporary, however, and we will soon have a direct connection.

AUTOANTIGEN-SPECIFIC REGULATORY T CELLS INDUCED IN PATIENTS WITH TYPE 1 DIABETES MELLITUS

Tihamer Orban, MD (Joslin Diabetes Center, Boston, MA)

Dr. Orban reported the 104-week phase 1 results of a novel drug involving the reintroduction of the self- antigen to insulin with the goal of establishing immune tolerance. The trial had 12 patients randomized to control and treatment groups, and the treatment was given as one injection of insulin B-chain peptide. Patients were newly diagnosed with type 1 diabetes, and were between 18-35 years, without any positive autoantibodies. The average A1c was ~8.5%, although C-peptide levels were lower in treatment group to start. The drug caused no serious adverse effects. There was no difference in C- peptide values or A1c between treatment and control groups. Insulin antibody titers were much higher in the treatment group, as expected. These antibodies had no effect on the metabolic effect of insulin treatment. Vaccinated patients had a significant T-cell response, peaking at six months. The vaccine was shown to produce an insulin B-chain-specific immune response. We’ll be interested in seeing further data—companies have begun to show an interest in this area, and insulin-based vaccines may soon emerge as a viable therapeutic target.

HUMAN CORD BLOOD STEM CELL-MODULATED REGULATORY T LYMPHOCYTES REVERSE THE AUTOIMMUNE-CAUSED TYPE 1 DIABETES

Yong Zhao, MD, PhD (University of Illinois, Chicago, IL)

Dr. Zhao and his team have identified cord-blood-derived stem cells that may modulate the response of T lymphocytes to diabetes auto-antigens when co-cultured with them in the laboratory. Treatment with these modulated cells (mCD4CD62L Tregs) normalized the blood sugar of diabetic mice, and increased insulin production by increasing beta cell mass in these mice. The cells were shown by histochemistry to increase replication of pancreatic beta-cells, and to preserve islet architecture. Treatment with these cells likewise reduced insulitis in these diabetic mice, possibly by modulating Th1, Th2, and Th3 cytokine production. Treatment may induce the apoptosis of infiltrated leukocytes in pancreatic islets, protecting them from immune attack. This technology would pose no rejection risk, and would be very cost- effective (if it proves effective in humans).

Questions and Answers

Q: Did the mice have any residual beta-cells?

A: We tested this in new-onset and recent-onset with good success. However, in late-stage patients when beta cells are almost gone, this therapy may not work.

Frontiers in Diabetes Technologies

USE OF CONTINUOUS GLUCOSE MONITORING IN CRITICAL ILLNESS/ICU

C. De Block, MD, PhD (University of Antwerp, Belgium)

Mounting evidence shows that tight control of glycemia in the ICU is critical. CGM will make insulin titration easier and safer and save nurse time. He predicted that closed loop control would reduce nurse workload and hypoglycemia.

  • Hyperglycemia occurs in 50-80% of critically ill patients and is associated with increased mortality and morbidity. In trials, intensive insulin therapy reduced mortality and morbidity. Achieving strict glycemic control requires extensive nursing efforts. So CGM should help by reducing workload. Dr De Block prefers a target zone of 80-110 mg/dl.
  • Inherent clinical alterations produce frequent changes in hourly insulin needs, so we need a display of glycemia for optimum insulin titration. However, we need high accuracy for the ICU – even Clarke Zone B is not accurate enough.
  • The best way to evaluate glycemic control is using time spent in blood glucose zones. CGM is the best tool for obtaining this data. The greatest limitation of CGM is the operating time of two to seven days. Dr. De Block cited more than 10 studies of CGM in the ICU showing that CGM performs well, and even detects more hypoglycemia and hyperglycemia than previously thought.
  • It’s not clear to Dr. De Block whether hypoglycemia is an independent contributor to mortality in the ICU. It’s correlated with tight control - NICE SUGAR is seeing more hypoglycemia in the tight control group, so we look forward to the results (due shortly).

NEXT GENERATION GENOME SEQUENCING: IMPLICATIONS FOR TYPE 1 DIABETES

George Eisenbarth, MD, PhD (Barbara Davis Center for Childhood Diabetes, Aurora, CO)

Dr. Eisenbarth discussed his work in attempting to identify genetic risk determinants for type 1 diabetes. He thinks that insulin is the primary autoantigen responsible for diabetes, at least in the NOD mouse. The B-chain 9-23 peptide is the specific region of insulin that provokes an immune response, and T-cells targeting this region share a specific fragment of their alpha chain recognition protein. He believes that once the structural determinants of insulin antigen recognition are better understood in humans, there will be different ways to interfere with the process and potentially create a vaccine.

  • Dr. Eisenbarth and his group have attempted to identify genetic risk determinants for type 1 diabetes. The appearance of next-generation sequencers have enabled more complicated genetic analyses—he thinks that individual sequencing will cost only $1000 dollars in two to three years. He believes that most genetically susceptible individuals will develop type 1 diabetes given enough time. Differences can be seen in the appearance and composition of the pancreas, caused by targeting of T cells to destroy islets.
  • He thinks that insulin is the primary autoantigen responsible for diabetes, at least in the NOD mouse. The B-chain 9-23 peptide is the specific region of insulin that provokes an immune response, and there is a specific T-cell receptor sequence that targets this peptide. An NOD mouse with one mutation in this region of the insulin peptide is protected from all diabetes. When T-cells develop, they randomly ‘pick’ elements of antigen recognition chains to assemble chain recognition proteins that enable them to recognize a specific antigen or combination of antigens. In the NOD mouse, most T-cells that recognize the insulin B-chain peptide express a particular portion of the alpha chain recognition protein. The sequence encoding this portion of protein can be introduced into a mouse to generate insulin antibodies and diabetes.
  • This information may eventually lead to a specific way to prevent auto-immune attack on beta cells. Cadaveric studies may help to uncover the details of autoimmunity in humans, and Dr. Eisenbarth’s group would like to perform studies on the bodies of people who have died with insulitis to better clarify what T-cells and other immune processes are present.

Questions and Answers

Q: What are your thoughts on the role of the MHC?

A: The MHC is an antigen binding complex that varies between individuals. In man, we haven’t had enough receptors studied yet to really know what the structural determinants of the interaction are. There might be different peptides in different individuals.

Q: Can there be something to block the immune response?

A: Once you get down to the structural determinants, there are different ways to interfere with this and potentially create a vaccine. When we introduce B9-23 genetically, we can prevent disease.

CSII TO PROTECT RESIDUAL BETA CELL FUNCTION IN TYPE 1 DIABETES

Paolo Pozzilli MB BS MD (University Campus Bio-Medico, Rome, Italy)

This work posits that we should put people with type 1 diabetes on the pump immediately after diagnosis, to preserve beta cell function for the longest time. The age of diagnosis is associated with different levels of C-peptide levels increase with age. Knowing your C-peptide is not an academic exercise, but might help you in future. Medtronic is sponsoring a pump/CGM trial in newly diagnosed kids – this should be a good opportunity to collect further data on C-peptide and relate this to glycemic control.

  • There is lots of variation in C-peptide at the time of type 1 diagnosis. Patients with nearly undetectable C-peptide need more insulin, are prone to higher A1cs, and tend to get complications later on. C-peptide is an independent protective factor for microvascular (but not macrovascular) disease.
  • The age of diagnosis is associated with different levels of C-peptide, which increases with age. Knowing your C-peptide is not an academic exercise, but might help you in future.
  • In a study in 1999-2000, newly diagnosed patients were immediately randomized to insulin pumps or MDI plus nicotinamide (which may keep C-peptide levels higher). The goal was to see if beta cell function could be preserved with intensive insulin therapy. Pump patients started in hospital. They never took shots first. Pumpers received more insulin but there was no increase in BMI between the two groups. The A1c was around 6.5% in both groups.
  • Mean C-peptide levels after 2 years rose 38% in the CSII group and +27% in the MDI group. Preservation of even some beta cell function improves control and impacts long term outcomes, so this rise was very helpful.
  • In a follow-up cross-sectional study, pumpers were divided into four groups based on time after diagnosis they started the pump, to assess the effect on C-peptide levels. The work showed that there was a window of about 2 years to start pumping after diagnosis, beyond which, meaningful C-peptide levels were not preserved.
  • Medtronic is sponsoring a pump/CGM trial in newly diagnosed kids – this should be a good opportunity to collect further data on C-peptide and relate this to glycemic control.