Hello from Melbourne, where the fourth day of IDF featured torrential downpours, an aboriginal concert under the Southern sky, new data, fierce debates, and, of course, lots of insight into how to best care for the hundreds of millions of people with diabetes worldwide - 382 million, to be exact, according to the IDF’s sixth Diabetes Atlas.
We attended an outstanding early morning session on diabetes technology on day #4 (which historically constitutes a very small slice of IDF sessions), where Drs. Roman Hovorka (University of Cambridge, UK) and Moshe Phillip (Schneider Children’s Medical Center, Petah Tikvah, Israel) shared the latest on the closed-loop front, including brand new at-home data, upcoming study timelines, and key challenges to overcome.
A jam-packed oral presentation session later in the morning featured a sub-analysis from the DUAL I examining postprandial glucose control with Novo Nordisk’s IDegLira (not surprisingly, the liraglutide half of the combination was responsible for most of the postprandial effect – it is great to see new products driving forward real change), as well as promising data from a small study of Victoza (Novo Nordisk’s liraglutide) in nine type 1 diabetes patients. As background, both IDegLira and the use of liraglutide in type 1 diabetes featured prominently during Novo Nordisk’s recent Capital Markets Day update (read our quick take on the presentation).
In the afternoon, Dr. Robert Ratner (ADA, Alexandria, VA) debated against Ricardo Cohen (Oswaldo Cruz Hospital, Sao Paulo, Brazil) on the need for controlled trials on bariatric surgery. Dr. Ratner made a particularly well laid-out argument in favor of more RCTs, highlighting the gaps that exist in our understanding of bariatric surgery (e.g., its mechanism of action and which procedure works best in a given person) and pointing to the limitations of prior research. Dr. Ratner concluded, “We should operate in order to learn, to discover new things, and perhaps to treat.” Dr. Ratner’s persuasively argued case led debate moderator (and longtime bariatric surgery hotshot formerly of Cornell, now at King’s College) Dr. Francesco Rubino (King’s College, London, UK) to remark, “I think Dr. Ratner is winning.”
Our appendix has a deeper dive on many of our top five highlights and other valuable presentations from the day, including a presentation and lively Q&A session with Dr. Ralph DeFronzo (University of Texas Health Science Center, San Antonio, TX) on triple therapy and a presentation by Dr. Bo Ahrén (Lund University, Lund, Sweden) on a study comparing Lyxumia (lixisenatide) administration before breakfast or before the main meal of the day.
- Executive Highlights
- Symposium: Technical Developments — Type 1 Diabetes
- Debate: Should Controlled Trials of Bariatric Surgery be Mandatory?
- Oral Presentations: Treatment of Type 2 Diabetes
- Initial Triple Combination Therapy Is More Effective and Safer Than Stepwise Add On Conventional Therapy In Newly Diagnosed T2DM
- IDegLira, A Novel Combination of Insulin Degludec and Liraglutide, is Efficacious and Safe in Subjects with T2D: A Randomised Phase 3 Trial
- Postprandial Glycemic Control With a Combination of Insulin Degludec and Liraglutide vs Each Component Individually in T2DM
- Oral Presentations: Islets and Incretin Biology
Top Five Highlights
1) Dr. Roman Hovorka presented brand new, quite compelling, data from the Cambridge team’s 21-day in-home, overnight, closed-loop study. The crossover study randomized adolescents (n=16) to use of sensor augmented pump therapy for 21 days, a 2-3 week washout, and then overnight closed-loop for 21 days (or the reverse order). The trial used an Abbott Navigator CGM and a companion receiver, a Dana R insulin pump, and an ultra portable laptop running a control algorithm. The overnight data (11 pm-7 am) looked excellent – closed-loop control significantly improved time-in-target (70-144 mg/dl) from 46% to 68% (p<0.001). Interestingly, time spent in hypoglycemia (<70 mg/dl) was not significantly different (1.4% during closed loop vs. 0.9% during open loop; p=0.13), a finding that probably mostly reflects the very low overall rate of hypoglycemia during open-loop therapy. However, the number of nights with glucose <63 mg/dl was nearly halved, dropping from 17 nights during open loop to 10 nights during closed loop (p=0.01). In discussing future challenges in the closed-loop domain, Dr. Hovorka mentioned that CGM accuracy and reliability “is not a major issue” as many have suggested (they would have been right with early versions of CGM); rather, he emphasized that optimal “patient selection is quite important at this stage,” and will ultimately be very important for obtaining reimbursement. We certainly agree, especially because this can so profoundly influence trial results which will then itself have such big regulatory and reimbursement implications. See the appendix for the complete detailed write-up and subsequent panel discussion.
2) Dr. Moshe Phillip provided an enthusiastic overview of the literature supporting use of insulin pumps, explaining his view that “the worst patient is the best patient to put on a pump.” (This was fairly fascinating to hear and opens up an interesting debate, as we imagine some clinicians would argue that it’s the most diligent patients who are best suited for pumps. We note that “worst controlled” might be a better way to characterize patients) Dr. Phillip believes we the field is moving into an era of a “pump-centric approach to type 1 diabetes management” – more details are in the full write-up in the appendix. Certainly the more physiologic the better, though long acting insulins are also becoming more physiologic. Dr. Phillip also reviewed the DREAM consortium’s inspiring overnight closed-loop work, which has now gathered an impressive ~1,000 nights of closed-loop control in patients’ homes. The DREAM 4 at-home pilot study #2 (four nights) finished in May 2013 (n=44), and a similar study is ongoing in Europe (n=30). An interim analysis of that study was already published in Pediatric Diabetes in August 2013. Notably, a six-week at-home study is (n=22) just finished yesterday (!), and Dr. Phillip mentioned that the hope is to conduct a pivotal study in the second half of 2014. He did not give details on what the pivotal study device would look like or how long/large the study might be, though we will be interested to hear what the regulatory requirements would look like in Israel and Europe. We salute IDF for organizing this data to be shown here and to Drs. Hovorka and Phillip for their ongoing work on the closed loop.
3) Dr. Stephen Gough (University of Oxford, Oxford, UK) presented an analysis of the postprandial glucose reductions seen in the DUAL I study on Novo Nordisk’s IDegLira, a fixed-ratio combination of the ultra-long acting Tresiba (insulin degludec) and GLP-1 agonist Victoza (liraglutide). As background, the overall study’s primary outcome showed a remarkable A1c reduction of 1.9% from a baseline of 8.3% (!) after 26 weeks of treatment with IDegLira (see page 4 of our ADA 2013 Incretins Report for more on the initial data presentation). In this analysis, IDegLira significantly reduced postprandial glucose excursions over Tresiba monotherapy as assessed by nine-point SMBG in the entire study population (n=1,663) and CGM readings from 260 patients. IDegLira’s postprandial efficacy was comparable to that of Victoza monotherapy. We didn’t find these results surprising, given that GLP-1 agonists like Victoza have a stronger postprandial effect while basal insulins like Tresiba have a stronger effect on fasting plasma glucose but it does reinforce the magic of the combination (we don’t use that word lightly). Data such as this will undoubtedly be critical in helping convince payers that IDegLira is worth a premium over the combination’s constituent components.
4) Dr. Yoshiyuki Hamamoto (Kitano Hospital, Osaka, Japan) presented striking early results on the use of liraglutide over one or two years in nine type 1 diabetes patients with residual insulin secretion (C-peptide > 0.3 ng/ml) and positive for GAD or IA-2 antibodies. Group A received insulin therapy and 0.9 mg of liraglutide for the full two years (n=5), while Group B (n=4) used insulin only for one year and then added on 0.9 mg of liraglutide for the second year. A1c declined by 0.6% (baseline: 7.9%) with two years of liraglutide use in Group A, while Group B saw a 0.6% gain in A1c (baseline: 7.6%) with insulin only (year one), followed by a striking 1.1% A1c decline after adding liraglutide over the course of the second year. Liraglutide “significantly decreased” hypoglycemia, though the results were not quantified. Average C-peptide response to arginine stimulation showed gradual deceases with the passage of time; however, the individual C-peptide responses to the addition of liraglutide varied amongst patients, suggesting there were also other individual factors at play – Dr. Hamamoto did not speculate on why this was the case, though we wonder if differences in autoimmunity could have played a role. We thought this data (though from a very small study) confirmed many other encouraging trials testing liraglutide in type 1 diabetes – as a reminder, Novo Nordisk initiated the 52-week ADJUNCT ONE study in November 2013 (n=1,400), the first of two phase 3a trials investigating the efficacy and safety of liraglutide in people with type 1 diabetes.
5) In a debate on whether more randomized control trials (RCT) of bariatric surgery should be conducted, Dr. Robert Ratner effectively argued that that more RCTs are needed to elucidate the “great many” unanswered questions that remain (e.g., which procedure is best for a given population, at what stage of a disease will a procedure be better for a person, what is bariatric surgery’s mechanism of action?). Indeed, Dr. Ratner concluded that given the current gaps in the evidence surrounding bariatric surgery “we should operate in order to learn, to discover new things, and perhaps to treat” – at which point he reemphasized the “perhaps.” In particular, Dr. Ratner pressed that sufficient data does not exist to justify the indication for bariatric surgery in adults with a BMI under 35 kg/m2. He also stressed that the data do not convince him that bariatric surgery prevents type 2 diabetes, “certainly not to the level of the DPP.” We agree and note that recently we’ve heard less about bariatric surgery causing a “cure” and more about it prompting “remission.” Overall, Dr. Ratner feels that the benefits, cost-effectiveness, and risks of bariatric surgery need to be studied further in well designed, RCTs with optimal medical therapy and lifestyle interventions as the comparator. The room overwhelmingly agreed with Dr. Ratner, as the panel discussion dissolved into a discussion of trial design instead of pro- or con-RCT; even Dr. Ricardo Cohen (Oswaldo Cruz Hospital, Sao Paolo, Brazil), who was ostensibly arguing against RCTs, highlighting the need for these trial in order to determine optimal treatments. Moderator Dr. Francesco Rubino (University College London, London, UK) put it best: “I think Dr. Ratner’s winning.” In our view, the need for more RCTs in bariatric surgery is quite clear, and the real question is how to make these trials more feasible. RCTs in bariatric surgery are notoriously difficult given that it is very hard to find patients who are truly well informed about both approaches and are willing to let fate determine whether they receive one treatment or the other.
Symposium: Technical Developments — Type 1 Diabetes
Roman Hovorka, PhD (University of Cambridge, Cambridge, UK)
Dr. Roman Hovorka presented brand new, very exciting data from the Cambridge team’s 21-day in-home, overnight, closed-loop study. The crossover study randomized adolescents (n=16) to use of sensor augmented pump therapy for 21 days, a 2-3 week washout, and then overnight closed-loop for 21 days, or the reverse order. The trial used an Abbott Navigator CGM and a companion receiver, a Dana R insulin pump, and an ultra portable laptop running a control algorithm. It was great to see this very real world study, with patients connecting to closed-loop at their convenience before bedtime (9:34 pm on average). (Dr. Hovorka did not discuss whether there was remote monitoring throughout the study; it did not seem like there was.) The overnight data (11 pm-7 am) looked excellent – closed-loop control significantly improved time-in-target (70-144 mg/dl) from 46% to 68% (p<0.001) and from 69% to a notable 85% for the wider 70-180 mg/dl range (p<0.001). Interestingly, time spent in hypoglycemia (<70 mg/dl) was not significantly different (1.4% during closed loop vs. 0.9% during open loop; p=0.13), a finding Dr. Hovorka attributed to a very low overall rate of hypoglycemia during open-loop therapy. However, the number of nights with glucose <63 mg/dl was nearly halved, dropping from 17 nights during open loop to 10 nights during closed loop (p=0.01). Notably, the tighter control with closed-loop, which ended at 7:37 am on average, also improved open-loop control up until ~12pm the following day. There were no severe hypoglycemia or adverse events during the study. In discussing future challenges, Dr. Hovorka mentioned that CGM accuracy and reliability “is not a major issue” as many have suggested; rather, he emphasized that optimal “patient selection is quite important at this stage,” especially for obtaining reimbursement.
- This study evaluated the safety, efficacy, and utility of overnight closed loop vs. sensor-augmented pump therapy in the home setting. The Abbott Florence prototype system was used, which included a Navigator transmitter, a companion receiver (one minute communication), a control algorithm running on a small ultra portable computer (connected by USB cable to the companion receiver), and a Dana R insulin pump. Every 12 minutes, a control decision was sent to the pump via Bluetooth. In the evening, patients connected their CGM receiver to the small computer and pressed a button to initiate closed loop. The system then downloaded information and closed-loop control began.
- The crossover study randomized participants (n=16) to use of sensor augmented pump therapy for 21 days, a 2-3 week washout, and then overnight closed-loop for 21 days (or the reverse order). Participants had a mean A1c of 8%, a mean age of 15 years, a mean duration of diabetes of seven years, a mean duration on the pump of three years, and a mean total daily dose of 55 units per day. Prior to the study, patients underwent a two-week optimization and training phase (minimum of eight days on CGM). Closed-loop training of about an hour was provided at the beginning of the closed-loop phase.
- Overnight closed-loop control (11 pm - 7 am) significantly improved time-in-target, reduced time spent in hyperglycemia, and reduced the number of nights with a glucose <63 mg/dl. Time spent in hypoglycemia (<70 mg/dl) was not significantly different, a finding Dr. Hovorka attributed to a very low overall rate of hypoglycemia during open-loop therapy.
Overnight Closed Loop (n=16)
% Time in Target
% Time <70 mg/dl
- Mean glucose over 24 hours was also lower with overnight closed-loop insulin delivery: 153 mg/dl vs. 162 mg/dl during open-loop (p=0.006). To us, this speaks to the power of better overnight control to improve overall glycemic control, even if daytime delivery is via open loop therapy. Indeed, Dr. Hovorka showed a graph of morning glucose values, which were clearly lower and less variable following closed loop nights. This effect wore off by noon.
- Closed loop insulin delivery achieved better glucose control with less overall insulin use. Interestingly, nighttime closed-loop delivery required about one more unit of insulin than open loop therapy (8.1 units vs. 7.2 units). However, total daily dose was ultimately three units lower with closed-loop delivery (49.9 units vs. 53.2 units). Dr. Hovorka attributed the higher nighttime dose to the patient population, who had little hypoglycemia and seemed to generally run higher; consequently, closed-loop delivery overnight required more insulin to bring their glucose values down. This implies that the improved overnight control reduced insulin needs the following day.
- Dr. Hovorka showed an “important graph” of controller effort, which “shows why closed loop is needed” and “what is limiting current use of open loop.” The graph depicted the percentage of closed-loop insulin given relative to the pre-programmed basal (i.e., a 200% controller effort means the controller gave twice as much insulin as the pre-programmed basal). Dr. Hovorka highlighted that the controller effort varied dramatically night after night within the same subject, suggesting that basal insulin needs change noticeably over time: “If you program something on the pump, it’s unlikely to be the optimal setting on the pump for the future. The past does not predict the future.” That said, he did mention that for some subjects closed-loop controller effort was 100%, suggesting their basal rates were appropriately set.
- An ongoing seven-day study is testing 24-hour closed-loop control in adults. Dr. Hovorka flashed a picture of a healthcare provider with diabetes wearing the system at work – she had the portable computer clipped to her belt. Participants are under free living conditions at work and at home over the seven-day study. The system is not fully reactive; patients are required to bolus for meals and do carb counting. Since bolus patterns are thus similar to open loop control, the system seems to bring the most benefits overnight. Dr. Hovorka did not mention any remote monitoring aspects or on-site surveillance.
- Coming up, the team plans to do three-month multicenter overnight studies in both children/adolescents and adults, as well as a seven-day in-home study testing night AND day closed-loop in children/adolescents. In 2014 home studies, the Cambridge team plans to use the same basic system but to replace the portable computer with an Android smartphone.
- Dr. Hovorka also mentioned closed-loop control in type 2 diabetes and critically ill patients. A recently published critically ill study (Leelarathna et al., Crit Care 2013) used a computer running an MPC algorithm, insulin and dextrose pumps, and a Navigator glucose sensor. The picture made the outcomes abundantly clear – an immense improvement in variability and time in range. A closed-loop study in patients with type 2 diabetes was also recently published (Kumareswaran et al., Diabetes Care 2013). Said Dr. Hovorka, “This technology could also be beneficial for this group of patients.”
Questions and Answers
Q: Did you have a chance to compare total insulin delivered in first half of night to the second half. Was there a dawn phenomenon?
A: We didn’t see anything specific. There was no indication of the presence of a dawn phenomenon.
Dr. Ragnar Hanas (Uddevalla Hospital, Uddevalla, Sweden): Various groups are using existing devices. When you look ahead into the future, do you envision an open source algorithm so people can combine various units? Or will it be one package that goes through the regulatory process?
A: All existing evidence in medical devices suggests it should be one package together. There is an interest in creating standards that would allow combination. If this is going to happen, major changes need to be made in terms of regulation. My feeling is it would have to be selected devices. There needs to be collaboration between algorithm developers and pump/sensor developers.
Dr. Hanas: Would you prefer a standard interface that all systems used?
A: If you want to get a new sensor on the market, you need to do a new pivotal study. I think from my view, I would like to get the devices to as many patients as possible. Manufacturers need to devise how to do that. It could be the main manufacturers or smaller companies.
Q: The variability of insulin dosing overnight explains why we’re struggling with standard therapy…
A: We tried to explain it, but couldn’t find anything. It’s difficult to say. I have my theories, but they were not confirmed by the real data.
Dr. Moshe Phillip (Schneider Children’s Medical Center, Petah Tikvah, Israel): I wonder about the total amount of insulin on closed loop vs. open loop. Was it different than in the control?
A: We gave on average one unit more on closed-loop nights than open-loop nights.
Dr. Phillip: In our DREAM studies, we didn’t see a difference in our calculations. There was a distribution difference. But I suppose one unit of difference is not clinically significant.
A: That may reflect the difference between your patients and our patients. Our patients spend about six minutes a night in hypoglycemia in the control arm vs. your patients spending 60 minutes a night. Your patients are generally running lower. Our patients generally run higher. The reason to get them down was we needed to give more insulin.
Q: How much of the overnight insulin variability is due to exercise and variations in physical activity?
A: We didn’t collect that data. I agree.
Future Pumps Automatically Regulated
Moshe Phillip, MD (Schneider Children’s Medical Center, Petah Tikvah, Israel)
Dr. Moshe Phillip provided an enthusiastic overview of the literature supporting use of insulin pumps, explaining his view that “the worst patient is the best patient to put on a pump.” (This was fairly fascinating to hear and opens up an interesting debate, as we imagine some clinicians would argue that it’s the most diligent patients who are best suited for pumps.) Dr. Phillip explained that we are moving into an era of a “pump-centric approach to type 1 diabetes management,” which started with advances like sensor-augmented pumps and will continue with more automated insulin delivery (see below). Dr. Phillip then reviewed the DREAM consortium’s inspiring overnight closed-loop work, which has now gathered an impressive ~1,000 nights of closed-loop control in patients’ homes. The DREAM 4 at-home pilot study #2 (four nights) finished in May 2013 (n=44), and a similar study is ongoing in Europe (n=30). An interim analysis of that study was already published in Pediatric Diabetes in August 2013. Notably, a six-week at-home study is (n=22) was just finished yesterday (!), and Dr. Phillip mentioned that the hope is to conduct a pivotal study in the second half of 2014. He did not give details on what the pivotal study device would look like or how long/large the study might be, though we will be interested to hear what the regulatory requirements would look like in Israel and Europe.
- “We are going to an era of a pump-centric approach.” Dr. Phillip’s slide highlighted that today we have sensor-augmented pumps with flexible basal rates, bolus calculators, the ability to download and view data, and threshold suspend. In the future, we are moving to predictive threshold suspend, data sharing and cloud storage, remote monitoring, automated carb counting, automated patient setting adjustment, and closed-loop augmented pumps.
- Dr. Phillip briefly mentioned the MD Logic Pump Advisor, a system that helps with optimizing pump settings. The DREAM team is testing it right now to provide suggested advice to physicians. Said Dr. Phillip, “I believe that one day we will trust it enough to give it to patients.” That would be tremendous in our view, since optimizing pump settings is so challenging, even for the most advanced HCP.
Questions and Answers
Q: In Britain, very few patients have an A1c <7%. I wonder whether pumps will make much of a difference in all of these people.
A: In the UK, NICE has recommended to use pumps only in patients who have failed any other attempt. Maybe [not using pumps] is one of the reasons for these results.
Dr. Roman Hovorka (University of Cambridge, UK): Wonderful data; I love the work. You have the largest experience of closed loop overnight. You talked about who the best patients for pump are; who do you think are the best patients for closed loop?
A: We indeed have a lot of experience. You believe you know what patients are doing at night, until you have the data. They don’t sleep at night. There is not a perfect patient for closed loop. Patients can hook up to the system when they are low or high. They can connect at 7 pm or 11 pm or 2 am if they just came back from a party. It’s extremely interesting, especially if you have a remote monitor. You would not believe it. [Laughter] There is no perfect patient.
Dr. Hovorka: What I meant to ask was which patients will benefit the most from closed loop?
A: It’s too early to comment, even though we have 1,000 nights. My impression is that the worst patients will gain the most out of it. The perfect patient measuring 10-12 times per day will benefit much less than others.
Q: I can’t understand how closed loop is any superior compared to sensor-augmented pump therapy with predictive suspend. Both of them stop infusion.
A: We are debating this issue constantly. The easy answer for you, is, “What about the mean blood glucose level?” If you suspend delivery only, and don’t deliver insulin when the patient improves, you must have a system like Medtronic to wait for two hours. You get a better mean blood glucose level with closed loop compared to just prediction. Suddenly you have jumps and excursions with suspension, which you don’t have with closed loop insulin delivery. Definitely as physician, not as an engineer, I can see many benefits to this approach.
Q: I’m from the United Arab Emirates. What do you do when dealing with non-compliant patients? You said the best candidate for a pump are the worst patients. That should be taken with a lot of caution. We’ve had many cases of unsuccessful users. We have 400 patients on pumps, but many of them are not entering enough information into the pump. How do you deal with this?
A: As I disclosed earlier, the A1c at our entire clinic, for 1,700 patients, is 7.8%. It’s not 7%, and it’s not 7.5%. Patients have their own opinions on how they want to live their life. I think the pump is the queen in the battle against type 1 diabetes. It should be available for every type 1 patient. For the worst compliant patient, we give them a combination of a long acting injection per day, and a pump just for boluses at the beginning. We train them slowly but surely. We make sure the parent gives the long-acting, and the patient gives the boluses. We train them slowly. With an automatic pump, we do not need collaboration so often.
Q: What do you think about daytime using fuzzy logic? Is it able to handle exercise and meals?
A: [Pointing to slide] This is an example of a patient on closed loop for 36 hours. She is exercising. She gives boluses only for meals. The system corrects for her mistakes. If she gives too little, it increases insulin delivery. Her mean blood glucose over 36 hours was 139 mg/dl, which translates to an A1c of 6.5%.
Q: And there was no restriction in carb intake?
Q: How does the system detect hypoglycemia?
A: It’s not complicated. It’s based on the CGM measurement, but also on the consideration of when and how much insulin was injected recently, and the trend and speed of reduction. We sometimes suspend delivery even when the blood glucose is 170 mg/dl, because the system is anticipating a crash based on the amount of insulin, timing, and speed of reduction.
Debate: Should Controlled Trials of Bariatric Surgery be Mandatory?
Robert Ratner, MD (ADA, Alexandria, VA)
Dr. Robert Ratner framed the debate by remarking that the question posed “goes to the heart of science.” He continued to explain that researchers conduct RCTs in order to avoid bias, determine who does best with a given treatment, and elucidate when a certain treatment should be used. Given the limitations of previously conducted trials (including RCTs) or bariatric surgery, and the “great many” unanswered questions that remain (e.g., which procedure is best for a given population, at what stage of a disease will a procedure be better for a person, what is bariatric surgery’s mechanism of action?), Dr. Ratner concluded that “we should operate in order to learn, to discover new things, and perhaps to treat” – at which point he reemphasized the “perhaps.” In particular, Dr. Ratner argued that sufficient data do not exist to justify the indication for bariatric surgery in adults with a BMI under 35 kg/m2. He also pressed that the data do not convince him that bariatric surgery prevents type 2 diabetes, “certainly not to the level of the DPP.” Overall, Dr. Ratner feels that the benefits, cost-effectiveness, and risks of bariatric surgery need to be studied further in well-designed RCTs with optimal medical therapy and lifestyle interventions as the comparator.
- Dr. Ratner expressed his appreciation for the Swedish Obese Subjects (SOS) study; however, he pointed out that it did have limitations. First, Dr. Ratner reminded attendees that the SOS study was not an RCT; the control group was a cohort of people who refused surgical treatment suggesting they had different motivations and priorities than those in the experimental arm. As a result, the study might be subject to allocation bias. Second, Dr. Ratner remarked that while there is no question that bariatric surgery conferred a “profound” initial weight loss, the small proportion of participants available for long-term follow up makes it difficult to draw conclusions about surgery’s long-term results. As evidence, he drew attention to participant follow up being less than 50% at year 10 and about 10-15% at year 20. It is possible that characteristics that led to a person remaining in the trial were tied to how well they responded to bariatric surgery in the long term.
- Though RCTs of bariatric surgery against medical therapy have been conducted, they have weaknesses in Dr. Ratner’s view. For example:
- The RCT that Dr. Francesco Rubino’s (King’s College, London, UK) group conducted was small (n=60; Mingrone et al., NEJM 2012). Additionally, Dr. Ratner noted that thanks to the trial’s randomized nature one could discern what the characteristics of success versus failure were for each treatment. His concern was that people enrolled in the trial must have already failed medical therapy. Those randomized to the medical arm continued to receive the same treatment they had requisitely failed. Dr. Ratner expressed his desire for such a trial to either enroll people before they fail medical treatment or give people an alternative medical therapy to what they had previously been on.
- Dr. Phil Schauer’s (Cleveland Clinic) STAMPEDE trial, while larger (n=150), was still relatively short in Dr. Ratner’s view. Thus, while Dr. Ratner thinks STAMPEDE demonstrated impressive reductions in BMI, A1c, medication use, etc. in the short-term, he thinks the question remains what bariatric surgery’s impact is in the long-term.
- Looking at the evidence that does exist on bariatric surgery’s long-term effects raised a few questions for Dr. Ratner on the durability of surgery’s short-term efficacy. For example, Dr. Ratner highlighted that while ~40-60% of people with type 2 diabetes who undergo bariatric surgery might go into remission in the first few years following the procedure, the remission rate stabilizes at about year two or three. Over the next ten years, about two-thirds of people whose type 2 diabetes had been in remission relapse. Notably, Dr. Ratner does not think that a person’s weight trajectory strongly explains whether one will go into remission or if one will remain in remission. This is because the weight trajectories for these various sub-populations are quite similar.
- Similarly, Dr. Ratner has concerns about bariatric surgery’s safety. Dr. Ratner highlighted that bariatric surgery’s short-term safety appears to be “quite positive,” with a 30-day mortality rate of about 0.3% in The Longitudinal Assessment of Bariatric Surgery (LABS). However, LABS was not a RCT, as the surgeon and patient chose what procedure a person would receive. Thus, Dr. Ratner thinks it is hard to conclude what a procedure’s safety is from this data. Dr. Ratner was more uneasy about bariatric surgery’s long-term safety. In particular, he expressed distress about evidence existing that 71% of surgery patients experience dumping syndrome (stomach emptying) several years later (Pories et al., Ann Surg 1995). This is concerning because over 99% of these patients who have dumping syndrome have episodes of severe hypoglycemia. Dr. Ratner noted that hypoglycemia can also be an adverse effect of bariatric surgery independent of dumping syndrome, and that it typically does not present until two or three years after bypass. Hypoglycemia, of course, can be life threatening and in the context of post-bariatric surgery can require a partial or full pancreatectomy to treat. Furthermore, Dr. Ratner pointed to some vitamin and mineral deficiency rates of ~40% as evidence that bariatric surgery patients must receive tight follow-up.
- Dr. Ratner is especially worried about unskilled bariatric surgeons performing risky procedures. Dr. Ratner referred to a study in which a panel of bariatric surgeons judged (on a scale of 1-5, with five being optimum) the skills of 20 other bariatric surgeons whose performance of a procedure was videotaped. The study found that these surgeons – who were not beginners, having conducted ~10,000 procedures over the last 10 years – had a very broad distribution of capability (2.6 to 4.8). The lowest quartile by skill was associated with three-fold higher complications rates when compared to the top quartile (14.5% vs. 4.2%; p<0.001) and five-fold higher mortality risk (0.26% vs. 0.05%; p=0.01). Additionally, those in the bottom quartile had double the reoperations, and triple the readmissions and ER visits. Dr. Ratner remarked, “If I were to have bariatric surgery, I would want either Dr. [Ricardo] Cohen [Oswaldo Cruz Hospital, Sao Paulo, Brazil] or Dr. Rubino to do it, but they may not be available.” He went on to question whether bariatric surgery has a positive benefit/risk profile outside or clinical trials (where expert surgeons perform the procedures), since presumably the level of skill of surgeons outside of trials is lower than in trials.
- Dr. Ratner noted that bariatric surgery does not appear to prevent later hospitalizations or save money. In the 30 days following bariatric surgery, unsurprisingly, patients have more hospitalizations than their counterparts who do not undergo bariatric surgery. A study (using data from the SOS study) published in JAMA suggested, however, that these two groups of patients have essentially the same hospitalization rate in the following 20 years (Neovius et al., JAMA 2012). In line with this, Dr. Ratner argued that current evidence points to bariatric surgery not being a cost-effective option. The operative period costs about $28,000 per person, and surgical patients continue to have higher healthcare expenditures in the 30-days post-operation. An analysis has found that these patients continue to have the same medical costs as their non-surgical counterparts over the subsequent six years (Weiner et al., JAMA Surg 2013 – see our coverage of the article here). Nonetheless, in our view it may be naïve to consider only the cost-savings potential of bariatric surgery instead of the cost-effectiveness. We hope that future discussions about the value of bariatric surgery consider benefits to patients’ health, well being, and quality of life rather than on cost savings alone.
Dr. Francesco Rubino (University College London, London, UK): Is there a way that you two can find common ground? You both agree that randomized controlled trials (RCTs) are needed to answer many questions. The problem is using the evidence for clinical practice. When should surgery be considered the best option for patients?
Dr. Robert Ratner (ADA, Alexandria, VA): Dr. Cohen, your last slides make my point. Studies comparing surgery and medication to medication alone are studies that need RCTs. We are absolutely working together, and we are on the same wavelength. The available data supports the ADA standards of care. We do view surgical intervention as treatment, but for those with a BMI of 35 kg/m2 with complications. There is no question that the data that has been generated supports the use of bariatric surgery. However, there are so many unanswered questions and potential pitfalls that we need to be cautious. The data from the Michigan survey (Birkmeyer et al., Annals of Surgery 2013) should give us pause that despite all of the procedures that had been done, half of them were judged to be unskilled work, and complications with those procedures went up substantially. RCTs that get rid of allocation bias and mandate tight follow up so that we can know long-term complications and benefits.
Dr. Rubino: I think it’s important to point out that Dr. Ratner stated that surgery for patients with a BMI above 35 kg/m2 with complications might be an option. Right now 1-2% of those that meet those criteria receive surgery in the US, and only 0.3% that meet the criteria get surgery in the UK – this is even lower in other countries. If we don’t offer surgical treatment to candidates, we’re doing something wrong and we need to fix that.
Dr. Ratner: I wouldn’t go that far. I think a great number of individuals with a BMI above 35 kg/m2 with diabetes are doing perfectly well. We have seen A1c levels falling progressively in the past years; right now, the mean A1c is 7.3%. We are reducing the risk of complications quite effectively. I don’t think we should, nor do I think we can afford, to do surgery on 90%, or even 50%, of those individuals. It also goes beyond financials; why would we perform surgery if there is no need? We shouldn’t be performing surgery just because some patients meet a threshold.
Comment: If there is one thing that we’ve learned in the past 20 years, it’s that things we are confident to be 100% true are refuted when we test them with objective trials – and those are RCTs. We didn’t think we needed an RCT for PCI vs. CABG because we thought it was unethical to withhold CABG treatment, but that’s not true. Dr. Cohen, at the beginning of your talk you cited RCTs when you said that diet and lifestyle were ineffective. When people were randomized in ACCORD, people said that it was unethical not to give people diet and lifestyle intervention. History shows that what we claimed should be standard of care should not be. Knowledge from RCTs has provided huge insight into what we should and should not do. We don’t need a trial to show that jumping out of an airplane with a parachute is safer than jumping without a parachute because the outcome for jumping without a parachute is always the same. When you’re looking at reducing outcomes by 20-30%, you need trials because you can’t always be certain about how to treat people. We need evidence, and the only way to get that is through RCTs. I have to agree with Dr. Ratner on this one.
Dr. Ricardo Cohen (Oswaldo Cruz Hospital, Sao Paolo, Brazil): Comparing surgery alone vs. medical treatment alone is unfair. We don’t need an RCT like this. We want to optimize medical treatment. We do need an RCT comparing gastric bypass in uncontrolled diabetes PLUS best medical treatment vs. best medical treatment alone. But we don’t need one comparing surgery vs. no surgery.
Comment: If this is a RCT design discussion, that’s fine.
Dr. Rubino: I think Dr. Ratner is winning!
Comment: I think it’s very problematic to look at patients with diabetes and patients without diabetes together. In the SOS, women without diabetes were included. Only 9% had diabetes at the beginning of the study. Patients without diabetes didn’t show any benefits; all the benefits were caused by the small number of patients with diabetes. Most of the women who were obese had a small CV risk and the event rate was very, very low at ten years – this is different than patients with diabetes. We need RCTs.
Q: There are an increasing number of teenagers who are obese and have type 2 diabetes. However, treatments are limited; metformin is the only approved treatment. Do you think this changes the scenario for RCTs?
Dr. Ratner: There is an enormous difficulty with obesity in adolescence and children, and you’re right: We don’t have a solution. I believe there is currently an RCT looking at this group?
Dr. Cohen: Yes, there is.
Dr. Ratner: There is a need to do surgery in this population, but we need to do in a manner where we learn about the effectiveness of different interventions and how we can move forward with treatment.
Comment: What we feel as HCPs doesn’t really matter. One thing I’m taking away from these presentations is that there’s so little data about what these patients feel or what they think should be outcomes of surgery. We don’t get a good sense of quality of life (QoL) from the follow-up data, and we don’t have a consistent way of measuring this. We need to clarify how QoL changes with surgery. We need to listen to the patients instead of sitting here talking as HCPs.
Q: There is a problem if we perform an RCT comparing basic medical care plus surgery vs. basic medical care. Basic medical care is too good in diabetes, and we don’t get enough events. If we perform an RCT with patients with diabetes and diabetic kidney disease in order to enrich the trial population, does that mean that the RCT only shows information about patients with diabetes and diabetic kidney disease? Are we able to extrapolate back to patients with just diabetes?
Dr. Ratner: Studies have to be designed to answer very specific questions. The results then either prove or disprove the null hypothesis, and then we ask questions. We can increase the delivery of care by finding new indications through research. However, the questions we need to ask need to be focused and answerable. I’m not sure we need to answer the question of whether bariatric surgery alters the CV event rate in 35 years because it’s not a feasible question. We need to ask whether surgical interventions in patients with prediabetes prevent diabetes compared to other interventions. The SOS is observational; it’s looking at the natural history of disease. Why don’t we have a study looking at early bariatric surgery vs. known effective therapies? Why not see if bariatric surgery prevents diabetes even better than these therapies?
Comment: As a statistician, I love RCTs. However, we need to think about feasibility. We should do a complete trial of the primary prevention of diabetes in high-risk population. That would be doable. But if we require a RCT with CV endpoints – that’s not doable. We have to think about what’s doable and what is not because we need evidence to move forward.
Dr. Rubino: I want to make the point that sometimes as diabetologists and surgeons, we tend to be far apart on what we think should be part of an RCT. A pill and surgery are not the same things, and we should not compare them. Surgery is almost never a first line treatment. If a patient comes into my office looking for treatment, what’s my first course of action? Surgery or medication? I would never say surgery, and I’m a surgeon. You need to do RCTs looking at options that are comparable to surgery – not first line options. However, RCTs are expensive and pose a lot of ethical concerns.
Dr. Ratner: I completely agree with you. In order to do an RCT we must feel as though both arms in the trial both have potential benefits, and there can’t be any ethical concerns in either arm. We have enough medical questions to ask; we have no effective therapies for obese adolescents. We need to start looking at this problem and testing to see which therapy is the most effective. Maybe we should do an RCT comparing surgery to high dose GLP-1 agonists and look at what gives a bigger benefit for these adolescents. There have been issues with previous RCTs in bariatric surgeries because the medication therapies have not been on the same level of intensity as bariatric surgery. In the SOS, there was no control intervention; we were looking at the impact of natural history. We need RCTs. I view bariatric surgery as having huge potential. What we’ve learned is phenomenal: The increase in GLP-1 levels post-operatively and the changes in the microbiome are critically important pieces, and we need to take next steps with these. If we can figure out what the bariatric surgery changes about the microbiome, maybe we won’t need this surgery. We need to figure out the mechanisms behind the changes we see with surgery.
Comment: I’m from the SOS study. When we did the diabetes prevention study, we did look at the control group and active treatment vs. no active treatment; there was no difference. The effect was so great; if you’re operating on 13 patients and 10 of them don’t have type 2 diabetes for 15 years, that’s incredible. No lifestyle modification provides results like that. If you perform an RCT, you have to make sure you see the whole picture. If you examine bariatric surgery, what will the endpoint be? We shouldn’t forget that we need to consider all of the serious events – we can’t just look at one of them. If we do a short-term study looking at just one endpoint, we may be missing something else.
Comment: Just like in all other areas of science, the most important thing you can do is identify the question you want to explore. I disagree with the lost comment, that it’s hard to choose which outcome you want to explore. We do that all the time. Patients with diabetes have 30 different outcomes, and there have been trials assessing only a small number of them. The question has to consider the feasibility of ensuring that participants will accept their randomization, as well as the feasibility of follow-up. It may be difficult to do an RCT with participants who have a BMI of 38 kg/m2 that are already committed to surgery, but it’s different to do an RCT with participants who have a BMI of 30 kg/m2 with diabetes and other complications; these participants may have a high risk of health outcomes, and the RCT may look at whether or not we can mitigate them. The hardest thing is agreeing on methodology that has a lot of merit. If you do that and get the question right, who cares how much it costs? The ACCORD trial was half a billion dollars. What matters is that we are pushing forward the questions.
Dr. Rubino: As the moderator of this session, I have the honor of asking the last question. You have a patient who has had type 2 diabetes for five years and has a BMI of 34 kg/m2. You’ve tried every possible oral medication, and now you’re starting this patient on insulin. This patient is not well controlled, but you know that he’s been compliant. Do you need an RCT to advise this patient to get surgery?
Dr. Ratner: The ADA guidelines have a BMI threshold set at 35 kg/m2, so clearly you’re challenging that. Guidelines are exactly that. They are not prescriptions. The data are convincing in people with BMI above 35 kg/m2. If my patient has a BMI of 34 kg/m2 and if they have complications, I would consider it.
Dr. Cohen: It’s a no brainer. A BMI of 34, 33, 32 kg/m2. I have several of those patients in RCTs, and they are doing well.
Dr. Rubino: This is a very exciting session to moderate. In July of 1999, an RCT of surgery vs. medication for diabetes was turned down because it was unthinkable – it was considered not unethical. The use of RCTs is extremely exciting and will lead to great advances.
Oral Presentations: Treatment of Type 2 Diabetes
Ralph DeFronzo, MD (University of Texas Health Science Center, San Antonio, TX)
Dr. Ralph DeFronzo presented his triple therapy study, which was first presented at ADA 2013 and re-presented at EASD 2013. See the complete coverage from ADA 2013 on pages 386-388 here. Seeing the presentation again reminded us just how striking these data are and how well they jive with the results of the subsequent presentation on IDegLira’s DUAL I study – better glucose control with less hypoglycemia and weight gain. This study will continue for one more year, at which time Dr. DeFronzo hopes to extend the original grant (from the ADA) and continue to follow patients.
- The open label, randomized controlled trial assigned 155 drug-naïve, newly diagnosed type 2 patients to either triple therapy (metformin plus pioglitazone plus exenatide) or conventional therapy (metformin followed by the sequential addition of a SFU and then basal insulin). The drug doses were titrated to avoid hypoglycemia and to achieve an A1c <6.5%.
- At 24 months, those in the triple therapy arm experienced a lower average A1c (5.9%) compared to those in the conventional therapy arm (6.7%; p<0.001) from a baseline A1c of 8.6%. Compared to the conventional group, the triple therapy group had a greater proportion of participants who achieved an A1c <6% (27% vs. 60%; p<0.001) or an A1c <7% (72% vs. 92%; p<0.001). The triple therapy approach also provided greater improvements in body weight (-1.2 kg [-2.6 lbs]) compared to the conventional therapy approach (+4.1 kg [+9.0 lbs]). Notably, the triple therapy group had a lower rate of hypoglycemia compared to the conventional group (15% vs. 46%, respectively), despite having a lower average A1c.
Questions and Answers
Q: I was confused by the study design. You compared two different treatments and two different timings. Is the early treatment better or the type of treatment better?
A: There are a thousand things you can do in designing study. We could have given metformin, the sulfonylurea, and insulin together. It shouldn’t be confusing at all. The ADA algorithm says that you give metformin, after you fail, you add the sulfonylurea, and after you fail – and you will fail – you add insulin. It should be called the treat to fail algorithm. I disagree with that. At the state of IGT, you have lost more than 80% of your beta cell function. Triple therapy is a more rational approach to attack all these problems.
Q: But this study doesn't tell you which aspects improved things…
A: If you would like to know that, you can design your own study. There are newer medications coming out every day. You could do triple therapy with other medications. These data are pertinent to when we did the study.
Stephen Gough, MD (Oxford Center for Diabetes, Oxford, UK)
Dr. Stephen Gough presented the results from the DUAL I trial of Novo Nordisk’s IDegLira (insulin degludec plus liraglutide), first presented at ADA 2013 and re-presented at EASD 2013 – see our coverage on pages 102-104 here. We’ve heard a lot of enthusiasm for this combination – on Day #1 of IDF, Drs. Bernard Zinman, Chantal Mathieu, and Luigi Meneghini all characterized IDegLira as the preferred method by which to initiate (!) insulin. This reinforced our view that so much therapy will be combination in form ten years from now (perhaps sooner, but definitely by that point) and that monotherapy may be mostly used in titration but not so much for long stretches for patients.
- The phase 3 DUAL I trial compared a fixed dose combination of insulin degludec and liraglutide (IDegLira) with insulin degludec (Tresiba) and liraglutide (Victoza) alone. The trial studied 1,663 participants for a 26-week period titrated to target in the case of the IDegLira and insulin degludec arms, and titrated in the usual ramp-up fashion for liraglutide. After 26 weeks, the IDegLira group was taking 38U/day of insulin, compared to 53U/day in the degludec group. A1c reduction from a baseline of ~8.3% was 1.9% for IDegLira, compared to a reduction of 1.4% with degludec and 1.3% with liraglutide. The IDegLira group achieved and maintained a very low A1c of 6.4%. Even better was the result that 81% of the IDegLira group reached the target of a 7.0% A1c (compared to 65% of the degludec group).
Questions and Answers
Q: What was the limiting factor of not increasing the insulin dose further in the degludec arm?
A: In the insulin degludec arm, there was no upper limit to the dose of degludec. Some patients were up at 60 or 70 units.
Q: But still the A1c was higher. Was this because the treating physicians could not increase the dose?
A: What it’s showing is that if you titrate insulin on your own, you can get to a certain level. But degludec on its own did not lead as low of an A1c. That suggests degludec on its own is not addressing the postprandial glucose. The liraglutide component in IDegLira is addressing the postprandial glucose values.
Dr. Ralph DeFronzo (University of Texas Health Science Center, San Antonio, TX): Where do you envision this combination therapy used – early in the natural history of the disease? When you run out of oral agents, would you then start with a GLP-1 first and then add the basal insulin? Or would you go straight with the combination?
A: Each time I’m asked this question I seem to provide a different answer. Like you, I believe in treating aggressively early with multiple therapies. I would be keen to use this therapy as early as possible, certainly after metformin. But it’s a difficult question.
Dr. DeFronzo: The data are very impressive.
Q: Is there a lesson to be learned from the fact that the frequency of GI side effects, was lower with IDegLira vs. liraglutide? Does that mean we should change our way of giving liraglutide?
A: It’s an interesting question. It means two things. The final dose of liraglutide in the combination was lower – 1.4 mg vs. 1.8. for liraglutide on its own. But that doesn't explain the early changes. I think the lower rate of GI side effects is due to the slow titration. Not everyone does experience nausea. This was very interesting. It showed you can reduce the nausea.
Stephen Gough, MD (University of Oxford, Oxford, UK)
Dr. Stephen Gough presented results from a secondary analysis of the DUAL I trial on Novo Nordisk’s IDegLira, a fixed-ratio combination of its ultra-long acting basal insulin Tresiba (insulin degludec) and GLP-1 agonist Victoza (liraglutide). As background, the data were first presented as a poster at this year’s EASD, and our take on the primary results of the full DUAL I trial can be found on page 4 of our ADA 2013 Incretins Report. Postprandial glucose levels for the entire 1,663 patient study population were assessed through nine-point SMBG, and the postprandial glucose excursion was calculated as the difference between blood glucose readings before and 90 minutes after each meal. Additionally, a subgroup of 260 patients underwent CGM at the initiation and the end of the study. The SMBG data demonstrated that IDegLira provided significantly better glycemic control than insulin degludec alone (effect size of around ~10-20 mg/dl), and comparable postprandial control to liraglutide. CGM data were consistent with the SMBG findings: the average postprandial area under the curve (AUC) was significantly lower with IDegLira than with insulin degludec for all meals except lunch, for which the reduction did not reach statistical significance. Mean AUCs for IDegLira and liraglutide monotherapy were similar. This study, along with the results of DUAL II, demonstrate that liraglutide plays a significant role in the IDegLira’s overall efficacy, which we imagine will help convince payers that the combination is worth a premium over Tresiba monotherapy.
Questions and Answers
Q: The most dramatic effect seemed to occur in postprandial readings after dinner — is there an explanation for that?
A: I’m afraid I don’t have one. The timing of the injection was always in the morning, and the biggest meal for most people was in the evening. Perhaps it was because the evening meal was usually the biggest meal of the day for most patients. Other than that, I don’t have any explanations.
Q: Insulin degludec acts mostly on fasting glucose, while liraglutide acts mostly on postprandial glucose. Given that the combination is in a fixed ratio, how do you handle patients who need more assistance with either the fasting or postprandial component?
A: I think increasingly, as is evidenced by recent guidelines, we are looking to individualize treatment. What we are seeing with this combination is that it works effectively in the vast majority of patients. In the future, you might see different ratios, or (as we tend to do anyways now) we may use each individual component in different combinations. However, the fact remains that most patients do seem to do well on this combination.
Comment: What struck me was that even though there was a decrease in postprandial glucose, the decrease is quite modest with combination. It looks like the major reduction in A1c is not from decreased postprandial glucose but rather from a resetting of the basal glucose level.
Q: The device limits patients to a maximum dose of insulin and liraglutide. I’m sure that a certain proportion reached that limit. How was glucose controlled in that subgroup?
A: Only around 40% of patients reached the top dose, and of those, 70% achieved their glycemic target on that top dose.
Oral Presentations: Islets and Incretin Biology
Effects of Lixisenatide Administration Prior to Either the Main Meal of the Day or Breakfast in T2DM Patients
Bo Ahrén, MD, PhD (Lund University, Lund, Sweden)
Dr. Bo Ahrén presented the results of a study comparing the safety and efficacy of Lyxumia (lixisenatide) administration either before breakfast or before the main meal of the day. As background, the phase 3 GetGoal-M study tested administration in the morning and evening, but was not designed as a comparative study. Dr. Ahrén’s 24-week trial enrolled 451 type 2 diabetes patients (mean A1c = 8.0%, mean BMI = 33 kg/m2, mean age = 57) on background metformin therapy, categorizing them by their main meal of the day before randomizing them to pre-breakfast or pre-main meal Lyxumia administration. After 24 weeks, there was no significant difference in the mean A1c reduction between the two groups (~0.7%; p=0.2664), and no significant differences in the groups’ mean reductions in body weight or FPG. The seven-point SMBG profiles demonstrated key differences between the groups, stemming from Lyxumia’s short half-life relative to “long-acting” GLP-1 agonists. Patients who took the drug before breakfast saw significant glucose lowering efficacy during breakfast and (to a lesser extent) lunch, but the drug effect seemed to have largely worn off by dinner. Those who dosed before dinner saw a near-total elimination of the postprandial glucose spike for that meal, but no major difference during breakfast or lunch. These findings underscore the fact that surrogates such as A1c do a poor job of telling the whole story on glycemic control, although the clinical relevance of the differences in SMBG profiles is likely limited. In terms of safety, there was no significant difference in GI tolerability or hypoglycemia between the groups, and rates of hypoglycemia were low. Overall, the data indicate that patients can administer Lyxumia when it is most convenient for them, without undue fear of diminished efficacy or safety.
-- by Adam Brown, Hannah Deming, Hannah Martin, Manu Venkat, and Kelly Close