Advanced Technologies & Treatments for Diabetes (ATTD) – 5th Annual Conference

February 8-11, 2012; Barcelona, Spain Day #3 Full Commentary

Executive Highlights

Greetings once again from Barcelona, where the sun has set on the penultimate day of this year’s ATTD. As we rambled around Barrio Gotico this evening, searching for churros and chocolate, we took a few moments to say a silent gracias for the day. Not only from an academic perspective to be learning so much about diabetes technology, not only from a patient perspective seeing all the great new tools and clinical strategies being presented, and not even from the perspective of excited tourists trying to take in all the beautiful architecture (is that another Gaudi?). No, today we were grateful to be Americans, because it’s now clear any beyond doubt that our country has – if not a full-blown registry – one of the richest information sources on type 1 diabetes that the world has ever seen.

Of course I’m talking about today’s presentation of data from the Helmsley Trust’s T1D Exchange, a cohort of 25,000 people with type 1 diabetes in the US. As Dr. Richard Bergenstal (International Diabetes Center, Minneapolis, MN) put it in his introductory talk, this “real life” data will help answer which treatments and practices actually improved patient outcomes, and it could also challenge or confirm many of the assumptions held about diabetes care. Of course like any observational dataset, the Exchange cannot directly address important questions of causality (e.g., is Therapy X beneficial, or is it just more likely to be used by the type of patients who would somehow have achieved better outcomes anyway?) – but some very, very strong associations do jump out just from looking at the data. For example, in what mined the registry’s wealth of data on SMBG and CGM use. In what Dr. Satish Garg (University of Colorado, Denver, CO) dubbed “the most important slide of the presentation,” patients who tested their blood glucose ≥10 versus 0-2 times per day were seen to have nearly 2.0% better mean A1c . This relationship was maintained when results were broken down by age group and insulin delivery method, provoking Dr. Garg to ask, “How many more randomized trials does one need?” Pumps and CGM were associated with notable benefits as well: if this isn’t evidence that diabetes technology’s health benefits are worth its short-term costs, we aren’t sure what is. We are grateful for the brainpower put into the T1D Exchange thus far, and we look forward to seeing how the data will benefit future patients, physicians, and researchers, particularly in shaping policy and payments.

The patient experience was also emphasized in Dr. Tadej Battelino’s (University of Ljubljana, Ljubljana, Slovenia) presentation of initial results from the outpatient DREAM 3 trial, which demonstrated the benefits of overnight closed-loop control (MD-Logic Artificial Pancreas) in young patients at diabetes camps. Fulfilling Dr. Moshe Philip’s (Schneider Children's Medical Center, Petah Tikva, Israel) promise from earlier in the week, the DREAM Consortium researchers screened a dramatic video documenting the first phase of the study in Israel, with a focus on the children who bravely entrusted their sleep to science. As a reminder, the video was in part a response to the iAP researchers’ movie of their own outpatient closed-loop efforts – one piece of a semi-serious debate about which team was first to outpatient trials. We are certainly glad to see the international competition heat up, as it reflects the field’s ever-increasing cachet. On the other hand, we would ec/em>ho the call of Dr. Eyal Dassau (UCSB, Santa Barbara, CA) for greater standardization of study protocols and analysis – across academic institutions as well as companies. As closed-loop control moves into patient’s homes, researchers, funders, and regulators will all need to know which approach offers the fastest way forward: things may be moving fast, but that doesn’t mean there’s any time to waste.

Table of Contents 

Detailed Discussion and Commentary

Oral Presentations: Clinical Benefits of CGM


Pratik Choudhary, MD, MBBS, MRCP (King’s College London, London, UK)

Dr. Choudhary presented a one-year observational study that gave CGM to patients with hypoglycemia unawareness. Encouragingly, the study led to a significant reduction in the number of severe hypoglycemia events. While there was no change in A1c or measures of hypoglycemia awareness, the reduction in severe hypoglycemia is a compelling finding and one we would think is especially attractive to payers. (Editor’s note – we are often crestfallen in our heads when people seem to think results aren’t notable because A1c doesn’t decline. In this case, fewer severe hypo events reflects less hypoglycemia and less hypoglycemia could actually prompt higher A1cs – but also a “higher quality” A1c, in our view.)

  • Twenty patients with hypoglycemia unawareness were given CGM in a retrospective clinical audit over one year. Patients had a mean age of 35 years, a mean A1c of 7.9%, a mean GOLD/Clarke score of 5.6 (>4 predicts hypoglycemia unawareness), and a severe hypoglycemia rate of 3.8 per year. One patient was on the Animas Vibe (Dexcom CGM), four were on the Abbott Navigator, and 15 used the Medtronic Veo. (Wow. What would life be like as a patient if you livedin Europe? The number of choices is incredible and the quality of technology is significantly higher than in the US.)
  • The addition of CGM led to a statistically significant reduction in severe hypoglycemia. At baseline, the rate of severe hypoglycemia was 3.8 per year. This declined to0.9 episodes per year at the study’s one-year mark (p=0.006). The study had a total of five severe hypoglycemia episodes, which occurred in three patients on CGM; four of these were due to overcorrection with rising glucose, one occurred at night (patient did not hear the alarms). Notably, none occurred in patients on the Veo pumps with low glucose suspend.
  • The study had no effect on A1c. Baseline A1c was 7.9%, which dropped to 7.7% at one year but was not statistically significant. There is no reason in our view to expect A1c to improve but certainly fewer severe hypos represents much better health in our view.
  • The addition of CGM did not lead to a restoration of hypoglycemia awareness. By the end of the study, there was no significant change observed in the Clarke or GOLD score. Dr. Choudhary called this fact “surprising” and emphasized that to restore hypoglycemia awareness patients must avoid all hypoglycemia, i.e., even short duration hypoglycemia is enough to suppress the counterregulatory response to hypoglycemia. Given that prolonged hypoglycemia is common in patients on CGM, the finding seems to fit with what is known about hypoglycemia unawareness – at any rate, “officially” restoring hypoglycemic unawareness is quite challenging, but we assume much lower glucose variability is a major benefit to patients and that it also may well reflect fewer ER visits.


Ignacio Conget, MD (Hospital Clínic i Universitari, Barcelona, Spain)

Dr. Conget reviewed the results of the SWITCH study, which we first reported on at ISPAD 2011 (see our report at The study used a crossover design to test the independent effect of CGM in patients on pumps. After a one-month run-in, subjects were randomized to either a sensor on/sensor off sequence, or a sensor off/sensor on sequence. The on and off periods were six months each, separated by a four-month washout. During the sensor on period, patients dropped their A1c by 0.43% on average from an 8.4% baseline. When the sensor was subsequently removed after the washout (this was only the case in the group with the on/off Sequence), A1c reverted back to baseline. Dr. Conget gave additional per-protocol data that was not presented at ISPAD – the A1c drop increased to 0.52%. We think the SWITCH study is a striking endorsement of CGM and are happy to see evidence continuing to mount for the benefits of CGM.

  • The SWITCH (Sensing With Insulin pump Therapy to Control Hemoglobin A1c) study examined the benefit of adding CGM to existing pump users. The 17-month, eight center, crossover design study had a unique protocol: a one month run-in period, followed by either six months with the sensor on or six months with the sensor off. A four-month washout followed, and then each group crossed over to the other therapy (in other words, participants’ sequence was either sensor On followed by sensor off, or sensor off followed by sensor on, where the off and on periods are six months and separated by a four month washout). During the sensor off period, subjects still wore a blinded guardian CGM. All patients in the study used the Bayer Contour Ascensia meter.
  • 153 participants with mean A1c of 8.4% and mean age of 28 years were randomized to off/on (n=41 adults, 35 children) or on/off (n=40 adults, 37 children) study arms.  Mean time since diagnosis was 15 years, mean baseline BMI 23.5 kg/m2, and mean time since start of pump therapy was five years. There were no significant differences in baseline characteristics between the two groups. The primary outcome was the difference in A1c levels between the sensor on and sensor off arms for the intent to treat population.
  • For the intention to treat population, the mean difference in A1c was -0.43% (p < 0.0001) in favor of the sensor on arm. The difference was -0.46% in children [n=72] and - 0.41% in adults [n=81]. Stopping sensor use (in the on/off sequence) led to A1c levels reverting to baseline.
  • For the per-protocol analysis (n=90), the mean difference in A1c was -0.52% (p < 0.0001) in favor of the sensor on arm. The difference was -0.67% in children [n=34] and - 0.45% in adults [n=56] – we thought it was notable that the difference in children was especially big.
  • Rate of sensor use was high during the study. Overall, 71.9% of the 153 study participants used the sensor 70-100% of the time. This represents a striking change from the JDRF trial some years back – mainly, we would assume, because today’s technology is far better than it used to be and although there are many improvements still needed in CGM, the progress made in accuracy and reliability has been particularly profound. Mean sensor use was 79% overall – specifically, 73% in children and 85% in adults. Again, especially the adult percentage reinforces to us that sensors are becoming perceived easier to use and are perceived as increasingly reliable.
  • Overall numbers “in zone” looked very positive for the “Sensor On” arm.
  • Notably, secondary endpoints for median glucose, standard deviation, time spent in range, and AUC were all statistically significant and in favor of those wearing the sensor. Specifically, the “Sensor On” group spent 16% more time in zone (71-180 mg/dl) than did patients in the “sensor off” group. Area under the curve was ~30-40% less for the “sensor on” group for both hypoglycemic and hyperglycemic ranges. Perhaps surprisingly, rates of severe hypoglycemia were not different between the two groups although given the higher A1cs, perhaps this group in aggregate doesn’t experience as much severe hypoglycemia as would patients who are closer to an A1c of 7% or less.

Sensor On

Sensor Off

P value

Median daily glucose, mg/dl




24-hour SD, mg/dl




Median time (minutes) in 71-180 mg/dl




AUC <70 mg/dl




AUC >180 mg/dl






Kirsten Nørgaard MD (Hvidovre Hospital, Hvidovre, Denmark)

The INTERPRET study is a large, international observational study of sensor-augmented pump therapy at 25 clinics in 15 countries. High A1c at baseline, more frequent sensor use, and being an adult predicted target A1c after 12 months. The average sensor usage during 12 months was fairly low to start, at 37%, and use declined over time to about 25%. Fear of hypoglycemia declined throughout the study.

  • The INTERPRET study is a large, international observational study of sensor- augmented pump therapy at 25 clinics in 15 countries. Pumpers were enrolled when they started using a sensor, with sensor usage recommendation of at least 10% of the time during one year. The number of pumpers considered in the analysis was n=186.
  • In a regression, only three factors were predictive in A1c reduction as a result of CGM – these were high A1c at baseline, more frequent sensor use, and being an adult. There were many factors included in the regression.
  • There were only two predictive factors associated with high sensor use - low A1 levels at baseline, and high sensor use at month 3. There was no effect of age group, gender, etc.
  • The average sensor usage during 12 months was 37%. Use declined over time to about 25%. Despite this decline, the ratio of patients with acceptable metabolic control remained largely constant. Patients who had high sensor usage at 12 months were predicted by high usage at three months.
  • Fear of hypoglycemia declined throughout the study. A survey scored fear of hypoglycemia, amount of worrying, and types of behaviors used to avoid hypoglycemia. On a positive note, all of these measures moved in a positive direction throughout the 12-month period.


Richard Bergenstal, MD (International Diabetes Center, Minneapolis, MN)

This short presentation showed results for the continuation phase of the STAR 3 study. In the original study, patients we switched from MDI to pump plus CGM (SAP). After 12 months, an A1c reduction of 0.6% was sustained. In the continuation phase, MDI patients switched to SAP and experienced a 0.4% drop in A1c after six months. In both phases of the trial, children 7-18 used the sensor less and were less able to sustain the A1c benefits over time. The groups that benefit the most from SAP are those with high baseline A1c, those older than 36 at diagnosis, and those older than 17 when they started SAP.

  • The STAR 3 study enrolled 485 patients with type 1 diabetes in a study of sensor- augmented pump therapy (SAP) versus MDI. The trial commenced in 2010 and lasted one year. At the end of the study, patients in the MDI arm had the chance to switch to SAP. In the continuation phase, 245 patients crossed over to SAP from MDI. The results presented today are after six months continuation.
  • In the original trial, the headline result was a 0.6% reduction in A1c for the SAP compared to the MDI group. Adults in the SAP arm achieved a 7.3% A1c at 12 months – closeto target. Children 7-18 achieved a similar 0.6% difference between the two arms, but with higher A1cs that drifted up a little over the course of the study. As might be expected, the study showed that sensor usage is very important. The threshold for improvement is about 40% usage, and better results were obtained as usage improved to over 80%. Results in this sort of trial should become much more significant over time as reliability and usability of sensor become.
  • In the continuation phase, the crossover group lowered A1c from 8.0% to 7.6% - very similar to the prior result. For the group already on SAP, A1c remained unchanged.
  • For children aged 7-18 years we saw a similar picture as th/strong>e main study. A1c dropped after switching to SAP therapy, but then started to rise back up from months 15-18. An analysis of sensor usage shows that usage was above 40% of the time at the switch, but then usage fell to 30%, which was not enough to sustain the benefit. We would expect usage to improve significantly with future generations of sensors.
  • The groups that benefit the most from SAP are those with high baseline A1c, those older than 36 at diagnosis, and those older than 17 when they started SAP.
  • Fun fact. Those in the lowest A1c quartiles bolus more but take smaller boluses – we assume this reflects a higher number of corrections.
  • Also note that glycemic variability is lower with SAP than with MDI for all levels of A1c. Severe hypoglycemia was not necessarily increased by improving glucose control.



Birthe Olsen, MD (Herlev Hospital, Herlev, Denmark)

This follow-up presentation examined quality of life and treatment satisfaction in the SWITCH study. Children wearing CGM in the study had no significant change in quality of life as measured by the Pediatric Quality of Life Inventory (23 items divided into physical functioning, emotional function, social functioning, and school functioning). However, adult treatment satisfaction as measured by Diabetes Treatment Satisfaction Questionnaire (DTSQs) significantly improved among those wearing the sensor (p=0.012). The study also measured days in the hospital and missed school/work. While hospital days and missed work were unchanged, children wearing the CGM missed significantly fewer school days over the course of the year (13 vs. 42, p=0.004) – we see this as quite a striking result. Dr. Olsen concluded that adding CGM to existing pump therapy increases treatment satisfaction in adults and does not impair it in children. This is a fairly significant finding in our view considering the ~0.5% A1c benefit demonstrated in this study.

Questions and Answers

Dr. Garg: How much education was involved in the study? Was it only at baseline?

Dr. Olsen: The education for sensor use occurred during the sensor on period. After the washout period, the patients about to use the sensor had CGM education. There was no extra education during the washout period.

Dr. Garg: What was the time spent per patient?

Dr. Olsen: I cannot say, but when I saw the patient it was half an hour.

Dr. Larry Hirsch (Becton Dickinson, Franklin Lakes, NJ): For the quality of life questionnaire for children, the scores differed by up to 6 points. But you said all were not clinically significant. What is the total range for that score? What is the maximum score difference?

Dr. Olsen: If you are feeling very well, the score is around 100. If you are feeling very bad, the score is around 20. It depends. When you see the negative figures, it is parents that felt their children had better quality of life without the sensor on. But it was not clinically significant.

Dr. Hirsch: You said the adult scores were significant, but these only around one or 1.5.

Dr. Olsen: It was a completely different scoring system that was used. It goes to 36.


Eda Cengiz, MD (Yale University School of Medicine, New Haven, CT)

Previous evidence has suggested that periods of hyperinsulinemia are associated with CGM underestimation of plasma glucose values (Monsod et al., Diabetes Care 2002; Yale’s closed-loop studies). Dr. Cengiz presented a study examining whether high doses of insulin affect the accuracy of CGM sensors. Fifteen adolescents underwent hyperinsulinemic euglycemic clamps and CGM values were compared to plasma glucose values. Ultimately, the CGM and plasma glucose values did not differ during times of maximum insulin action, suggesting the discrepancy is not attributable to the presence of high insulin levels.

ATTD Yearbook 2011


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

Dr. Hirsch highlighted two studies on self-monitoringem> of blood glucose (SMBG), illustrating the promise of smarter testing strategies and the still-relevant limitations in meter accuracy. He first reviewed Roche’s yearlong STeP trial, which suggested that a structured protocol can significantly improve the glycemic benefits of SMBG in poorly controlled, insulin-naïve patients with type 2 diabetes – without requiring any increase in the average number of fingersticks per month. While quite enthusiastic about STeP’s implications, Dr. Hirsch noted that the results could be difficult to translate into the real world, where primary care providers often do not enough time counseling their patients with type 2 diabetes. Dr. Hirsch also discussed an analysis of seven marketed blood glucose meters, which suggested that several do not meet current ISO accuracy standards and that hardly any would meet the newly proposed, stricter guidelines. He noted that meter accuracy has wide-ranging implications in diabetes care – from home insulin dosing to ICU glycemic management to CGM calibration – and he wondered how below-threshold meters would fare once the new guidelines are officially adopted (we think this will happen later in 2012).

  • Dr. Hirsch began by reviewing STeP randomized controlled trial, a Roche-funded randomized controlled trial of structured SMBG in poorly controlled, insulin-naïve patients with type 2 diabetes (Polonsky et al., Diabetes Care 2011). The yearlong studyincluded 483 adults with type 2 diabetes (mean age 55.8 years, mean A1c 8.9%, mean BMI 35.1 kg/m2). All patients received free Accu-Chek meters and test strips and were instructed in their use. Patients in the active control group (ACG) were simply told to test, while those in the structured testing group (STG) were instructed to perform a seven-point SMBG profile on each of three days prior to each quarterly visit with their primary care providers.
  • In the intent-to-treat analysis, the STG patients experienced a 0.3% A1c decline relative to the ACG (-1.2% vs. -0.9%; p=0.04). The per-protocol analysis, which Dr. Hirsch said he found more interesting, showed a 0.5% between-group A1c difference (-1.3% vs. -0.8%; p<0.003), with no significant difference in A1c decline between ACG patients and nonadherent STG patients. Dr. Hirsch noted that a higher percentage of STG patients started on insulin during the study (a result he said was “perhaps unsurprising,” presumably because doctors who have a clearer picture of their patients’ glycemic profiles are more likely to prescribe therapeutic changes). He also pointed out that the structured protocol did not require additional testing – in fact, in the ITT analysis, the STG patients performed statistically significantly fewer tests – a positive sign that SMBG efficiency can be greatly improved.
  • Dr. Hirsch said he agreed with STeP’s implication that structured testing (and perhaps other variations on SMBG) would likely be more effective than the once-to- twice-daily unstructured testing recommended today, though he expressed caution about translating the results directly to clinical practice. Outside the clinical trial environment, healthcare providers generally have less time to spend with patients. Thus Dr. Hirsch would be interested to see how structured testing fares in a more real-world environment.
  • A group of researchers from Taiwan evaluated seven blood glucose meters, concluding that most would fall short of the proposed updates to ISO accuracy standards – with some not even meeting today’s standards (Kuo et al., Diab Technol Ther 2011). The seven meters tested were: Bionime’s Bionime Rightest GM550, Roche’s Accu- Chek Performa, J&J LifeScan’s OneTouch Ultra2, Abbott’s MediSense Optium Xceed, Terumo’s Medisafe, Taidac’s Fora TD4227, and Bayer’s Ascensia Contour. These devices were used by 107 adults, with results compared to hexokinase reference values.
  • Strikingly, only three of the seven meters met the ISO standard for accuracy below 75 mg/dl (95% of values within 15 mg/dl of reference), and only four of the seven met the ISO standard for accuracy above 75 mg/dl (95% of values within 20% of reference). The three meters meeting the <75 mg/dl threshold were Bionime’s GM550, Roche’s Accu-Chek Performa, and J&J LifeScan’s OneTouch Ultra2; the four meeting the >=75 mg/dl threshold were the GM550, the Performa, the Optium, and the Contour. As for the proposed new ISO accuracy standard (95% of readings within 15%/15 mg/dl for reference values greater/less than 100 mg/dl), only one meter met the bar – the Bionime GM550. As a sizeable grain of salt, we note that three of the paper’s six authors were employed by Bionime. We continue to question decisions regarding which lot is used, how strips are stored, etc.
  • In light of the generally underwhelming accuracy results, Dr. Hirsch raised several questions about the implications for SMBG in different contexts. Meter (in)accuracy is important not only for traditional home-use insulin dosing, but also in the hospital (especially the ICU, where blood glucose meters are often used despite lacking an indication in these patients) and with regard to CGM calibration (a particular concern for Dr. Hirsch himself, and for other clinicians who widely prescribe CGM). More broadly, Dr. Hirsch wondered how today’s meters will fare under the upcoming ISO standards – i.e., whether and how regulatory agencies will retroactively interpret the revised rules.



Tadej Battelino, MD, PhD (University Children’s Hospital and University of Ljubljana, Ljubljana, Slovenia)

Dr. Battelino covered four papers in his review of CGM in 2011: (1) Dr. John’s Pickup’s meta-analysis; (2) an MDI vs. sensor-augmented pump study from Hermanides et al. (Diabetic Medicine); (3) a study in the ICU from Brunner et al. (Critical Care Medicine); and (4) the buzzed about type 2 study from Dr. Vigersky presented at ADA and recently published in Diabetes Care (see our Closer Look at All studies were positive and pointed to a glycemic benefit of using CGM. Dr. Battelino emphasized that CGM is both a “very clinically meaningful device” and “likely to be cost effective,” although the right patients must be selected and they must actually wear the sensor. As noted, we think wearability stats will improve enormously with the next generation to be approved.

  • Pickup et al., BMJ 2011 – “It’s kind of a turning point when a field gets a meta- analysis.” Dr. Battelino discussed John Pickup’s meta-analysis of CGM. He noted the large number of patients included in the study (892) and the fact that individual data was incorporated. There was a nice 0.3% A1c decline in favor of CGM – while some argued this was not clinically relevant, stratifying by sensor usage and baseline A1c revealed up to a 0.9% decline in A1c for seven days of sensor wear per week and a starting A1c of 10%. For hypoglycemia, both a fixed effects and a random effects model revealed a significant decrease in hypo AUC for CGM wearers (“A significant achievement.”). Dr. Battelino concluded that baseline A1c is the strongest predictor of glycemic improvement, every additional day of use of CGM decreases A1c by 0.15%, and age has a small effect. He urged that we need to use CGM in the right patient group – it is both a “very clinically meaningful device” and “likely to be cost effective.”
  • Hermanides et al., Diabetic Medicine 2011. Dr. Battelino characterized this sensor- augmented pump (SAP) vs. MDI study as very similar to STAR-3. The six-month parallel arm study included adults with type 1 diabetes with an A1c over 8.2%. He believes the “outcome is extremely, extremely significant” – a 1.2% A1c benefit of SAP therapy vs. MDI. Time spent in hyperglycemia dropped by 17% and there was no significant change in hypoglycemia. There were also significant improvements in quality of life. We were very interested to hear an account of the healthcare provider time spent with each group: 690 minutes with SAP vs. 240 minutes with MDI patients. We think this speaks to the need for better reimbursement for clinicians actively using pumps and CGM.
  • Brunner et al., Critical Care Medicine 2011. This study revealed safe and successful use of CGM in 174 patients in the ICU. An insulin titration error grid analysis revealed 99% of points in the acceptable treatment zone, 0.5% in the unacceptable treatment zone, 0.4% in the next zone, and 0% in the life threatening zone. Additionally, 92.9% of the CGM values met the ISO standard. Dr. Battelino highlighted that using real-time CGM was actually more accurate than using apoint-of-care device.
  • Vigersky et al., Journal of Diabetes Science and Technology 2011. This buzzed about study first displayed at ADA (see page 133 of our ADA Full Report at examined episodic real-time CGM use in type 2 diabetes over 12 weeks (four cycles of two weeks on CGM/one week off). A significant 1% decline in A1c was observed for the CGM group compared to 0.5% in the control group. Although there was no change in hypoglycemia, those using the sensor more often had better outcomes, and the improvement was sustained over one year. See our Closer Look for coverage of the Diabetes Care version of the article at



John Pickup, MD, PhD (Kings College London School of Medicine, London, UK)

For this presentation, Dr. John Pickup focused on two pump topics. Dr. Pickup stated that there is increasing evidence that giving a bolus 15-20 mins before a meal significantly reduces post-prandial hyperglycemia (but probably only when blood glucose is >90 mg/dl). He also noted that managing CSII in the hospital is gaining increasing attention. It seems that further optimization is possible via specific protocols and the advice of inpatient pump experts.

  • Dr. Pickup selected the following themes for the yearbook: (1) Optimizing control on pumps, (2) pumps in special circumstances, (3) pumps for people with type 2 diabetes, (4) the best type of rapid acting insulin for pumps, (5) pumps plus CGM.
  • Regarding the timing of the meal bolus, Dr. Pickup stated that there is increasing evidence that giving a bolus 15-20 mins before a meal significantly reduces post- prandial hyperglycemia. This seems very logical to us. He cited work from Dr. J. Hans de Vries et al who gave ten type 1 pumpers a breakfast bolus either 30, 15 or zero minutes before eating and followed them with CGM for 4 hours afterwards. With a 15-minute bolus, a 40 mg/dlmmol/l) reduction in peak blood glucose level was observed with no hypoglycemia.  Curiously, with a 30-minute advance bolus there seems to be only little effect. Another paper by Dr. Peter Chase et al discussed giving the bolus -20, 0, or +20 minutes from start of eating, and demonstrated a very dramatic glucose lowering effect of bolusing 20 minutes before. It’s probably best to bolus before only when blood glucose is greater than 90 mg/dl (>5 mmol/l), it was stressed..
  • Managing CSII in the hospital is gaining increasing attention. Dr. Pickup mentioned a USA based survey of 65 pump users admitted to hospital on 125 occasions for an average of 4.7 days. Most would have preferred to remain on their pumps. But only 66% of the patients continued on pumps in the hospital, 14% discontinued, and 19% had intermittent use of pumps. Nonetheless, there was similar mean blood glucose in all three groups. Dr. Pickup suggested three possible explanations: (1) Pumps can’t overcome all the challenges of glycemic control in the hospital; (2) there are relatively few staff experienced on inpatient pump management; (3) the hospital stays were relatively short and there wasn’t enough time to adjust pumps appropriately (basals, waveforms etc.). A specific protocol is needed for hospital pump management, and advice from a team experienced in pumps is required when patients are admitted.



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

Dr. Dassau noted that the number of manuscripts in this year’s ATTD Yearbook grew to 14, up from nine each of the previous two years. He sees this as a sign of how much the field is moving forward (he especially highlighted advances in overnight closed-loop control). To close, he listed several near-term goals for artificial pancreas research, including: standardized study designs and outcome measures, trials that involve tougher and more realistic challenges to glucose control, larger studies, outpatient studies, and more-wearable closed-loop systems.

  • Dr. Dassau noted that this year’s chapter on Closing the Loop is longer compared to the first two editions of the Yearbook – a good sign in his view. He explained that themanuscripts each year fall into three general categories: clinical studies (single- or dual- hormone), “under-the-hood” advances in closed-loop algorithms or control strategies, and review or editorial papers. This year Dr. Dassau and his co-editors selected seven clinical studies (compared to seven in 2010 and four in 2009), five under-the-hood papers (up from only two in 2010 and three in 2009), and two reviews (compared to zero in 2010 and two in 2009). He interpreted the higher number of manuscripts (14, as compared to 9 in both 2010 and 2009) as a reflection of the field’s progress, and he forecasted that the number of manuscripts would be even higher next year.
  • Dr. Dassau highlighted two studies on overnight control – a subfield he found particularly meaningful in 2011. Both studies compared an MPC algorithm to open-loop control. In one paper (n=20), researchers from the international iAP consortium (UVA, Padova/Pavia, Montpellier) reported a statistically significant reduction in hypoglycemia of almost fivefold compared to open-loop control (Kovatchev et al., JDST 2010). Further, in both the iAP manuscript and a 25-patient paper from the University of Cambridge (Hovorka et al., BMJ 2011), closed-loop control led to statistically significantly greater time spent in the near-normal range of 70-140 mg/dl. Mean blood glucose in both studies was roughly 120 mg/dl. Dr. Dassau emphasized the benefits for patients of waking up with a consistent, normal blood glucose value: such days offer “a new start at avoiding the roller-coaster effects of type 1 diabetes.”
  • Dr. Dassau mentioned several next steps for closed-loop research. He called on his colleagues to unify study design and outcome measures, allowing for more consistent interpretation across trials. Also needed are studies that present tougher, more realistic challenges to artificial pancreas systems; larger multi-center studies (Dr. Dassau noted that several such studies are slated to report data next year), outpatient studies (which have already begun), and a move toward more-wearable closed-loop systems (for example, toward either a small mobile device or the pump itself).



Jan Bolinder, MD, PhD (Karolinska University, Stockholm, Sweden)

  • Dr. Bolinder discussed Novo Nordisk’s degludec in his yearly review of insulin and insulin therapy. Three proof of concept studies were published in 2011, although Dr. Bolinder focused on the type 1 study from Birkeland et al. appearing in Diabetes Care. The 118-patient study compared aspart plus degludec to aspart plus glargine. There was no significant difference in A1c between the two groups, although overall confirmed hypoglycemia was 28% lower (not significant) in the degludec group. Encouragingly, there was a statistically significant 60% lower rate of confirmed nocturnal hypoglycemia with degludec. Dr. Bolinder believes degludec may allow for more flexible basal insulin regimens and a reduced risk of nocturnal hypoglycemia. He cautioned that data from long term RCTs are “much awaited” to clarify the insulin’s safety and efficacy. (As a reminder, degludec has a PDUFA date of July 29, 2012 – for more information, see our extensive Novo Nordisk 4Q11 report at Dr. Bolinder briefly mentioned Biodel’s Linjeta, Halozyme’s PH20, and insulin and cancer (“the burning question from previous years”) at the beginning of his presentation, but he did not discuss any studies or provide any opinions on these topics.



Lutz Heinemann, PhD (Profil Institute for Metabolic Research, Neuss, Germany)

In 2011, there were no papers on ARIA (alternate routes for insulin administration) for nasal, dermal or transdermal administration. Oral insulin has also been relatively silent. Biocon’s phase 2 trial of their oral insulin candidate, IN-105, was not successful and there were no publications by Oramed. Buccal insulin has lost the momentum it once had. Dr. Heinemann cited a paper by Patton et al., which was a survey on injection rotation practices in children. 22% self-reported using only one site, and 36% only two sites. This is important, as site rotation is one the best methods of reducing the occurrence of lipodystrophy. (Editor’s note – this reminds us of a piece in the February 9, 2012 NEJM on insulin- induced lipodystrophy. See

  • 2011 was not a good year for new ways of giving insulin, Dr. Heinemann began. There were no papers on ARIA (alternate routes for insulin administration) for nasal, dermal or transdermal administration. At ATTD, KK Pirkalani et al. have a poster on insulin administration in the auditory channel – but the poster was not up when Dr. Heinemann checked.
  • Oral insulin has also been relatively silent. Biocon’s phase 2 trial of their oral insulin candidate, IN-105, was not successful. Novo Nordisk also announced that they were stopping development of NN-1952 – one of their candidates, and there were no publications by Oramed. However, there is a lot of science. Dr. Heinemann commented that: “In rats, oral insulin is working quite nicely.” Selected papers show that for oral insulin, there is a nice increase in insulin kinetics with dose. But so far, differences in post-prandial glycemia are not so pronounced with dose.
  • Buccal insulin seems to be a promising area that has lost momentum according to Dr. Heinemann. The Canadian company Generex used to be quite prominent at scientific conferences, purchasing large booths, and working with a number of renowned diabetologists. Dr. Heinemann understands that their product is still on market in Ecuador and India. A Phase 3 study was started, but data never presented. There has been a change in the management team, and little information is available on what happened.
  • Regarding insulin pens, Dr. Heinemann takes the view that there is a feature war using scientific publications as ammunition. Many studies are head-head comparisons sponsored by a manufacturer, designed solely to showcase some known point of differentiation. However, he cited a paper by Patton et al., which was a simple survey on injection rotation practices given to 201 children with type 1 diabetes. 22% reported using only one site, and 36% only two sites. Fear of pain was the most common barrier to using multiple sites. This is important, as site rotation is one the best methods of reducing the occurrence of lipodystrophy.



Neal Kaufman (DPS Health, Los Angeles, CA)

Dr. Kaufman described three papers that exemplify different aspects of health information technology research, using each as a case study for general lessons in health IT. He said that research on electronic- health-records-integrated clinical decision support was positive, but that clinical decision support must be only one of several goals for next-gen EHR. Discussing a successful intervention targeted at diabetes- associated depression, he said that the next steps are to identify traits predictive of participant success (and to consider targeting the intervention around patients with these qualities). Finally, he said that a review of Veterans Administration telehealth efforts show the possibility of innovation even within a bureaucracy – but only if the innovators have enough political will, incentives are favorable, and resources are available.

  • O’Connor and colleagues reported encouraging results from a clinic/strong>-randomized study of electronic health records (EHR)-based clinical decision support (Ann Fam Med 2011). Their study included 11 clinics, 41 physicians, and 2,556 patients. The researchers’ system recommended a medication regimen (including warnings about medication contraindications), laboratory tests, and more-frequent visits for patients not at goal. After one year, clinicians using the decision support software were generally satisfied, and patients had improved levels of mean A1c and systolic blood pressure (though not diastolic blood pressure or triglycerides). Dr. Kaufman said that the results were encouraging, though he noted that EHR must offer more than just clinical decision support – they must also improve the efficiency of the diabetes care team, encourage patients to increase participation in their medical care (e.g., involving them in appointment scheduling), and helping patients to self-manage their conditions outside the clinical visit.
  • Many patients with diabetes also face depression – a difficult double-diagnosis that Van Bastelaar and colleagues attempted to address with a 12-week online intervention (Diabetes Care 2011). In the program, 255 patients with type 1 or diabetes participated in eight lessons, watched videos of other depressed patients describing how they had benefited from the same course, and received feedback on course homework from coaches. Encouragingly, the intervention reduced depressive symptoms as well as diabetes-specific emotional distress. Although no glycemic benefits were observed at study end, Dr. Kaufman said this was to be expected given the study’s short length. Over time, he said, anything that improves depression will improve glycemic control as well. Dr. Kaufman added that the next steps for this (or similar interventions) will be to identify the traits of patients who benefit most, assess the advantages of personalizing the intervention to these patients, and then weigh the costs and benefits.
  • Finally, Dr. Kaufman discussed an analysis of 19 papers on various Veterans Health Administration telemedicine efforts, published between 2000 and 2009. Dr. Kaufman drew encouragement from the possibility of developing telemedicine interventions even within such a vast bureaucracy as the Veterans Administration. However, he emphasized that success in such environments requires political will, skilled and dedicated people, aligned incentives, and – as with any intervention – the availability of resources.



Jay S. Skyler, MD MACP (University of Miami Miller School of Medicine, Miami, FL)

As always, Dr. Skyler gave an authoritative overview of progress towards preventing type 1 diabetes. He selected six papers relating to immune interventions. Overall, the headlines were not all positive, but within the data there were many provocative findings. Bright spots included a paper suggesting hydrolyzed casein infant formula has some impact in preventing antibodies, and evidence that atorvastatin, teplizumab (anti-CD3) and abatacept delay (but not stop) C-peptide decline. Dr. Skyler concluded his presentation (somewhat ironically) with the words of Salvador Dali: “Have no fear of perfection – you’ll never reach it.” Although we may not be reaching perfection it’s not through want of trying, and the hope is that the learnings will translate one day into an effective combination therapy.  Dr. Skyler presented six selected papers from July 2010 – June 2011 (in only seven minutes). The presentation included:

  • The DPT-1 Follow Up Study showed no change after more than ten years. In 2005 the DPT-1 reported that oral insulin alone had no effect on delaying the appearance of type 1 diabetes. With longer term follow up, the time to diabetes curves decline in the same way. However, in a subgroup of individuals with insulin auto-antibodies >80 nU/ml, the original DPT-1 showed separation of the curves (meaning a delay). The separation is constant at more than a median nine years of follow up. The use of oral insulin was discontinued after the original study. At this point, we can still see the original delay but we don’t know if this would have improved or stayed the same if treatment had been continued.
  • A ten year follow up study shows that the survival rate for developing 1 autoantibodies is improved with the use of casein hydrolysate baby formula in infants at risk for diabetes. We are impatiently waiting the results of the TRIGR trial as it reports in the next few years.
  • The DIATOR study of atorvastatin (Lipitor, Pfizer) showed a slower decline in C-peptide over time in the intervention group. It turned out to be statistically non-significant but suggests that we should investigate further to see if the anti- inflammatory effect of statins could have a role in preventing diabetes.
  • The Protégé trial of teplizumab anti-CD3 (MacroGenics) was reported by Sherry et al. in the Lancet. In the paper we see a statistically significant change in C- peptide from baseline compared to placebo for the full dose 14-day regimen. Unfortunately, the trial used a different primary composite outcome measure (low A1c together with low insulin use), which did not show any differences to placebo, making it technically a negative trial. But it has yielded many provocative observations.
  • A recent trial of abatacept by Orban et al. showed a clear separation of C- peptide curves, although there was still a decline in C-peptide over time in the intervention group, just like in the placebo group. This implies some delaying effect, but a single intervention (as opposed to combination therapy) may not be enough to ultimately prevent diabetes. This paper concerns the effect of blocking T-cell co-stimulation for 24 months. At ADA this year we will see the results for 36 months. Will the effect be sustained or will the curves start to come together?
  • A study of GAD-65 vaccination yielded a negative result, in contrast to a small trial a few years ago. The trial compared patients with two or three courses of GAD vaccine versus placebo, and results were identical. In the prior work, it had appeared that GAD-65 vaccination worked in a sub-group, but this was only ~25 participants, showing the dangers of scaling up too quickly. In the previous week (so it will be in next year’s Yearbook) the results of a European Phase 3 trial reported. The trial compared two and four doses of GAD vaccine; again, stimulated C-peptide declined similarly in all groups.
  • Dr. Skyler concluded his presentation with the words of Salvador Dali: “Have no fear of perfection – you’ll never reach it”. Although strongly positive results have not been obtained, there is progress. He reiterated his hope that effective combination therapies would emerge in the next few years.



Lois Jovanovic, MD (Sansum Research Institute, Santa Barbara, CA)

Dr. Jovanovic was quite excited to see hundreds of articles related to pregnancy and diabetes in 2011 (“Diabetes and pregnancy has exploded.”). The yearbook chapter includes ten articles, including three that Dr. Jovanovic focused on in this presentation. The first was a closed-loop study in pregnant type 1s --while the study achieved a mean blood glucose of 104 mg/dl and showed the safety of closed loop control in pregnant women, Dr. Jovanovic emphasized a number of the study’s limitations that should be addressed in future trials. She also discussed studies on predicting post-pregnancy complications and the importance breastfeeding infants.

  • Murphy et al., Diabetes Care 2011. This study examined closed-loop insulin delivery in ten type 1 pregnant women. Women were studied during both early (12-16 weeks) and late (28-32 weeks) pregnancy. Overnight glucose control was very good, with 84% time in range in early pregnancy and 100% in late pregnancy. Dr. Jovanovic called the potential of an AP for pregnancy “wonderful,” although she spent much her talk noting many of the study’s limitations. First, there was a long period of hyperglycemia after breakfast that should have been incorporated in the algorithm. Second, the mean blood glucose of 104 mg/dl and target range of 63-140 mg/dl in the study were too high in her view. Dr. Jovanovic believes that these glucose targets put the fetus at risk considering a non-diabetic blood glucose average is 84 mg/dl. She also questioned the time periods used in the study, which were more akin to studying women twice in the middle of pregnancy. Future studies should look at women even earlier and even later in pregnancy. Finally, the high carb meals used in the study were not typical of the lower carb diets more commonly used by type 1 pregnant women.
  • Yogev et al., Journal of Maternal Fetal and Neonatal Medicine 2011. This study looked at hundreds of pre-pregnancy factors related to post-pregnancy complications in 32 women with type 1 diabetes. The only factor that predicted complications was pre-pregnancy BMI. Dr. Jovanovic emphasized that overweight and obese women with type 1 diabetes must have weight counseling prior to pregnancy.



Shlomit Shalitin, MD (Schneider Children’s Medical Center, Petah Tikva, Israel)

Dr. Shalitin spent most her talk discussing a very intriguing study of real-time CGM in adolescents with hypoglycemia unawareness – after just one month of use, patients had improved epinephrine responses during hypoglycemia, suggesting a valuable benefit of CGM for those with hypoglycemia unawareness. We thought this was an impressive finding considering the short duration of the study. Dr. Shalitin also discussed a crossover study testing strategies to mitigate nocturnal hypoglycemia following exercise (Taplin et al., Journal of Pediatrics 2010). Both strategies (reducing overnight basal insulin or using a low dose of terbutaline at bedtime) were effective at preventing nocturnal hypoglycemia, although the number of readings >250 mg/dl was higher in both cases compared to the control group.

  • Hewitt et al., Diabetes Care 2011. This study examined the benefits of real-time CGM for improving hypoglycemia unawareness. Eleven adolescents used real-time CGM for four weeks, with counterregulatory hormone responses to hypoglycemia measured before and after the intervention. The epinephrine response during hypoglycemia after the intervention was greater inthe CGM group than in the standard therapy group, suggesting real time CGM is a “useful clinical tool to improve hypoglycemia unawareness in adolescents in type 1 diabetes.” Dr. Shalitin emphasized that this was a small study of just one month, but it is important for its implications.



A. Liberman, PhD (Schneider Children’s Medical Center of Israel, Petah Tikva, Israel)

Patient outcomes are strongly influenced by the ability of the patient to translate physicians’ recommendations into everyday actions. A study of missed boluses in young people determined that A1c is correlated with the amount of missed boluses. Fear of hypoglycemia in type 1 pumpers is strongly associated with poor glycemic control.

  • Patient outcomes are strongly influenced by the ability of the patient to translate physicians’ recommendations into everyday actions. Evidence shows that one of the biggest problems in getting to target is poor adherence to primary daily diabetes management tasks.
  • A study of missed boluses in young people determined that A1c is correlated with the amount of missed boluses. 34% of respondents revealed that they had missed a breakfast bolus. The most common reasons for missing a bolus were ‘forgetting,’ social embarrassment, or being distracted. It’s theorized that some patients experience diabetes management as an exhausting burden and repress or deny the need to give boluses.
  • Fear of hypoglycemia in type 1 pumpers is strongly associated with poor glycemic control. There is a vicious circle from fear to poor control to dissatisfaction to fear. Some patients deal with a lower quality of life in order to get better glycemic control. However, avoidance of severe hypoglycemia is the best way of decreasing fear of hypoglycemia.



Satish Garg, MD (University of Colorado Denver, Aurora, CO)

Explicitly instructed to speed through his talk so that the session would finish in time, Dr. Garg gave rapid-fire themes related to basal insulins (e.g., Novo Nordisk’s degludec, Lilly’s pegylated lispro, and multiple new variants of glargine), ultra-rapid-acting insulins (he mentioned Halozyme’s PH20-based therapies), incretin therapies (notable papers included late-stage research on Bydureon and early-stage investigations of incretins for type 1 diabetes), SGLT-2 inhibitors (he said that dapagliflozin’s potential long-term risks require further assessment), and diabetes prevention (low-dose pioglitazone could be promising, but Dr. Garg believes more evidence is needed).

  • Basal insulin: Dr. Garg called attention to three recent studies and one editorial on Novo Nordisk’s insulin degludec. He seemed particularly interested in the degludec+aspart coformulation degludecPlus, especially for type 2 diabetes, given its dual effects on fasting and postprandial glucose. Of course, other companies and researchers are also looking for ways to build a better basal insulin, he said. Uehata and colleagues have improved the PK/PD of insulin glargine with the addition of SBE7-β-CyD, other groups have worked on U-200 or U-500 formulations of glargine, and Lilly recently began phase 3 research on its pegylated version of insulin lispro, LY2605541.
  • Diabetes prevention: Dr. Garg noted that low-dose pioglitazone might be a good pharmaceutical option to prevent or delay diabetes onset; however, he suggested that confirmation of this effect would require a larger study than the 600-patient investigation published by DeFronzo and colleagues (NEJM 2011). On a more basic level, Dr. Garg wondered aloud whether the glycemic criteria for diabetes diagnosis should be revised – a potential discussion at ATTD 2013, he said.

Symposium: Closing the Loop


Moshe Philip, MD (Tel Aviv University, Petah Tikva, Israel) and Eran Atlas, MSc (Schneider Children’s Medical Center, Petah Tikva, Israel)

Dr. Phillip and Dr. Atlas discussed the DREAM consortium and summarized the basics of the MD-Logic Artificial Pancreas System. The project is a four-step approach to bringing overnight closed-loop control to the home environment: feasibility study (DREAM 1), an inpatiem>ent overnight study (DREAM 2), an overnight study at a diabetes camp (DREAM 3), and an overnight home study (DREAM 4). The control algorithm is based on fuzzy logic, which seeks to emulate the line of reasoning of diabetes caregivers by using if-then statements. The system includes remote monitoring of multiple patients through the Internet, features individualized and learning capabilities, and controls to a range of 5-8 mmol/l (90-144 mg/dl). The MD-Logic AP can communicate with the Dexcom CGM, Medtronic pumps/CGM, a Bayer meter, and Sanofi’s iBGStar. The DREAM studies used the Medtronic Veo insulin pump, Enlite CGM sensor, and Bayer’s meter. Dr. Phillip closed with a presentation on the DREAM 1 results, a feasibility study of the system in 12 patients. The system demonstrated good performance and patients kept in the target range for most of the night. Wow!


Thomas Danne, MD (Kinderkrankenhaus auf der Bult, Hannover, Germany)

Dr. Danne summarized the results of the previously conducted DREAM 2 study, an inpatient overnight trial of the MD-Logic AP. The system performed well in 12 patients – time in range (62-140 mg/dl) was 29% during open loop and 76% during closed loop (p=0.03); mean blood glucose was 155 mg/dl in open loop vs. 131 mg/dl in closed loop (p=0.02); and three events of hypoglycemia (<63mg/dl) occurred during open loop and none occurred during closed-loop control. This inpatient study paved the way for the recent outpatient DREAM 3 study (see below).


Tadej Battelino, MD, PhD (University Children’s Hospital and University of Ljubljana, Ljubljana, Slovenia)

Dr. Battelino presented brand new results from DREAM 3, which is studying the MD-Logic AP in overnight “transitional” studies at three diabetes camps in Europe. This was a very inspiring talk. He first discussed the pilot study that took place last October – the overnight system significantly increased time in range and decreased hypoglycemia. Dr. Battelino followed with a preliminary analysis of the DREAM 3 main studies – although the data was not statistically significant, it certainly trended in the right direction of reduced hypo- and hyperglycemia. We expect as more data is collected, the benefits of the overnight closed-loop system will be more convincingly demonstrated. We are very glad to see successful outpatient studies – this meeting has given us a real tangible sense of just how far research has come and how quickly it is moving. Hopefully, regulators can keep up! Dr. Battelino ended the presentation with a dramatic video filmed on the night of the DREAM-3 pilot closed-loop study. To view it, visit (As an aside, Dr. Moshe Phillip has been characterizing the DREAM 3 pilot study as the first outpatient trial of the closed-loop, though there is some debate as to whether it was truly as “outpatient” as Dr. Claudio Cobelli’s more ambitious study a few weeks later. Dr. Cobelli’s study featured a hotel room stay, meals out in restaurants, a mobile phone based portable system, and even a bicycle ride around the city! The fact that we can have such a debate is pretty amazing – just a couple years ago, many thought the possibility of an outpatient AP study was an event for the distant future. For coverage on the other outpatient study, see page 15 of our ATTD Day #1 coverage at

  • DREAM 3 is designed as a transitional overnight closed loop study taking place at diabetes camps at three centers in Europe. Participants were randomized to two arms (overnight closed-loop control with MD-Logic and open loop) in a crossover design study. Prior to the study, basal and bolus settings were adjusted according to ISPAD regulations. Alarms were set at 75 and 350 mg/dl for the control group and low glucose suspend was turned off.
  • The DREAM 3 pilot study took place on October 9, 2011 in 18 patients (mean age: 13.5 years, mean A1c: 7.7%, mean diabetes duration: six years, mean duration on pump: four years). Fifteen patients were entered into the final analysis due to two sensor failures and one patient that wanted to go home. A customized program was built for patients prior to the study after collection of iPro CGM, insulin pump, glucose meter, and meal diary data. Patients at the camp were randomized to overnight closed-loop control with the MD-Logic AP or open-loop therapy. They crossed over to the other arm of the study on the following night. One major downside of the study was the use of large laptops at bedside to run the closed-loop algorithm; the researchers are currently developing a smaller, more portable system to hopefully run on a cell phone. Participants were permitted to eat freely and engage in unrestricted physical activity at camp. All patients were simultaneously monitored from a central command and control center – we think this is one of the coolest and most impressive parts of the system.
  • Time in range increased significantly (p=0.039) alongside a reduction in hypoglycemia. Events less than 70 mg/dl numbered 16 during open loop compared to 5 during closed loop. The number of events under 63 mg/dl was 2 during closed loop and 8 during open loop. The duration of time less than 63 mg/dl was 25 minutes during open loop, compared to 2 minutes during closed loop. There was no significant difference for time above 180 mg/dl (47 minutes vs. 21 minutes [p=0.337]).
  • A few changes were made from the pilot DREAM 3 study before the main studies commenced. The major change was a protocol amendment that meant investigators would stop following the blood glucose of the control group continuously. When they had done so in the pilot study, it was difficult to avoid intervening. The investigators also improved the automatedcommunication protocol, fine tuned the system alarms, and changed the post-hypoglycemia controller settings.
  • The DREAM 3 main studies occurred in 38 patients at two other diabetes camps from November 26-28, 2011 and January 28-30, 2012. Patients had a mean age of 14 years, a mean A1c of 8.1%, a mean duration of diabetes of 6.9 years, had been on the pump for 4.9 years, and a mean total daily dose of 0.8 u/kg. Thirty-six patients were included in the final analysis because of two sensor failures. The study design was identical to the pilot study. Dr. Battelino explained that he “was personally scared to ask parents to put a child on a computer.” However, he was surprised to see families were very excited and noted, “we had a huge line” and “not a single family was excluded.” The study’s primary end points were hypoglycemic events under 63 mg/dl, time less than 60 mg/dl (an FDA requirement), and mean overnight glucose level.
  • In an interim analysis of 18 patients, closed-loop therapy demonstrated benefits in the primary endpoints, though none were statistically significant. The total number of events less than 63 mg/dl was four in the control group compared to one in the closed-loop group (not significant). Time less than 60 mg/dl was 18.5 minutes in open loop compared to 1.2 minutes during closed-loop (p=0.07). Average blood glucose level was 134 mg/dl in the control group vs. 124 mg/dl during closed-loop (p=0.17). We imagine that open loop success wouldn’t necessarily be as great in a “regular” setting but we’d like to find out more about the help that children received – we can imagine they were on “best behavior” more than they otherwise would have been had they not been in the trial. “Control groups” can be challenging to create.
  • A number of secondary endpoints showed a benefit to closed-loop therapy, including increased time in range, a decreased number of blood glucose readings over 140 mg/dl, fewer events under 70 mg/dl (six during open loop vs. three during closed loop; not significant), time less than 63 mg/dl (27 minutes during open loop compared to 1.8 minutes during closed loop; p=0.04). There were no severe side effects, no episodes of severe hypoglycemia, and no episodes of DKA.

Questions and Answers

Dr. Satish Garg: Great work. I hope that this makes it to the next step. Before the next step, what is the justification for the sample size at each stage?

Dr. Battelino: Zero. When we applied for the European grant, we did a power calculation. The size for endpoints like A1c and events for hypoglycemia was 185 patients and a six-month continuous experiment. We cannot do that.

Q: Did you not have an opportunity to sample blood during the experiment? Are all your outcomes based on CGM?

Dr. Battelino: No, that’s not correct. We were required by ethics committees to add blood glucose measurements every two hours.

Q: It seems like you had a sensor failure rate of ~5%? How do you feel about that?

Dr. Battelino: There are always different ways to see trouble. One way is to say, “Well, we have sensor failures. We cannot do it.” The other way is we need to detect it and switch to manual operation to maintain safety. We’ve use sensors for a long time – we know some sensors fail or drift. The task is to detect this.

Q: So SMBG was done during them middle of the night?

Dr. Battelino: Every two hours.

Q: How long can this be extended? The two-hour blood glucose checks are obviously not realistic.

Dr. Battelino: We several issues to overcome. One thing we might do is use a redundant sensor.

Q: You started the system at 4pm. From the time that program was operating, you included a meal that they bolused for and the algorithm also took care of the period around the meal?

Dr. Battelino: We showed data from 23:00 to 7:00. We also have data from 7 to 7.

Dr. Ragnar Hanas (Uddevalla Hospital, Uddevalla, Sweden): How did you treat the alarms?

Dr. Battelino: Alarms in the closed loop were linked to the closed loop. It was the physician’s decision whether to intervene or not.

Dr. Hanas: How often did you manually intervene?

Dr. Battelino: Out of my study, maybe four in 80 patients in two nights.



Bruce Buckingham, MD (Stanford University School of Medicine, Palo Alto, CA)

This was a very interesting presentation, providing valuable evidence on the real-world longevity of infusion sets. Dr. Buckingham showed that 66% of patients studied have recently experienced occlusions or unexplained hyperglycemia causing infusion set failure, so he set out to understand root causes – and emphasized that a future closed loop system needs to detect impending failure. During these studies, patients wore infusion sets to failure (because of hyperglycemia or skin problems). Survival of the sets was about 100% at 3 days, and 40% at 7 days, implying that a majority of patients could get to five days in principle. Steel and Teflon sets had the same all-cause failure rate at seven days, although Teflon sets had a 15% early failure rate, presumably because of kinking. There was no difference between the two leading rapid acting insulins. Finally, some clever modeling work was able to predict the failure of a set from glucose and insulin data, even if glucose was in the normal range.

  • There are many problems with wearing infusion sets and sensors including ‘too much stuff’ on the body, skin issues, and accuracy issues. Dr. Buckingham showed many photos of skin problems, including lack of real estate in children, erythema (skin redness), and induration (lumps, bumps, hard, flaky skin). But curiously, many patients don’t change their sets until they fail, presumably because of the inconvenience and pain of changing. This reinforces to us the importance of the development of better sets.
  • Failed infusion sets show a range of features, such as zinc crystals (from precipitated insulin), even macrophages coming up the catheters. Deposition of zinc crystals was similar between Novolog (insulin aspart, Novo Nordisk) and Humalog (insulin lispro, Lilly).
  • Dr. Buckingham presented results of his studies on infusion set failure. He asked 20 patients to wear multiple sets until they failed (because either of sustained hyperglycemia or erythema). Survival of the sets was about 100% at 3 days, and 40% at 7 days. There was no difference between Novolog and Humalog. Induration and erythema generally doesn’t occur before about four days and then the longer you wear the set, the more you get.
  • Steel and Teflon sets performed similarly with one exception – Teflon sets had an early failure rate. 20 patients wore both a steel and Teflon twice for a week (80 set-weeks). The Teflon sets experienced a 15% early failure rate, which wasn’t seen with steel (presumably because of kinking). After the first day, failure rates were parallel. Steel has slightly higher tendency to pull out than Teflon. By seven days, Teflon and steel were more or less similar for all-cause failure.
  • Modeling work has been able to predict impending infusion set failures well. A mathematical model takes into account glucose and insulin trends and can predict a set failure several hours before it happens, even when blood glucose appears normal (at least it could in the examples that Dr. Buckingham showed).
  • The failure rate of infusion sets implies that most people could get to five days with current infusion sets. By seven days, two thirds of sets have failed. It seems that the maximum days of wear is dependent on the person – not everyone can wear sets for an equally long time, although all can get to three days. Dr. Buckingham suggested that this effect is probably related to immune response.

Questions and Answers

Q: Dr. Francine Kaufman (Medtronic, Northridge, CA): Have you considered coating or impregnating the catheter, for example with anti-inflammatories…

A: Dr. Buckingham: Certainly! But first we need to find out the root cause – macrophages? Fibrin? Plasmin? We need to understand the biology before we can fix it.


Bruce Bode, MD, FACE (Atlanta Diabetes Associates, Atlanta, GA)

Dr. Bode discussed the management of both in-hospital intravenous and outpatient subcutaneous insulin delivery using the Glucommander computer-based solution. As a reminder, the Glucommander system automates in-hospital management, providing various safety guardrails and conveniences for tracking multiple patients. Previous results have shown that the Glucommander can bring patients under 120 mg/dl within three hours, with only 2% experiencing blood glucose levels under 50 mg/dl – leading to reductions in readmission rates, rates of surgical site infections, and the length of ICU stay. Moving forward, Dr. Bode suggested the Glucommander could also facilitate the transition to subcutaneous insulin within the hospital following recovery, providing case studies for how the system could aid in calculating dose requirements. Notably, a study is ongoing examining this application of the Glucommander with outpatient values measured using Sanofi’s BGStar and sent to a server for monitoring by a nurse – with this complete integration, Dr. Bode suggested computerized dosing of subcutaneous insulin would be “the way of the future.”

  • Given the complexity of inpatient insulin protocols, an automatic and effective method for achieving goal with minimal risk of hypoglycemia was desired. While nurses have traditionally administered protocols with increasing complexity of calculations, the Glucommander system automates management, providing various safety guardrails and conveniences for tracking multiple patients; the 2.0 system also allows for integration with the hospital’s mainframe computer and data monitoring from handheld devices.
  • Results from a study of over 2,000 cardiovascular patients indicated the Glucommander brought patients under 120 mg/dl within three hours, with only 2% experiencing blood glucose levels under 50 mg/dl and 0% under 40 mg/dl; evidence from thePiedmont Hospital in Atlanta also indicates the system reduced readmission rates, rates of surgical site infections, and the length of ICU stay, producing $994,000 in savings per year. Dr. Bode suggested that as a result of these successes the system would be used in a large NIH- sponsored trial of bypass surgery patients – we hope to hear more about the design of this trial to gauge how it might clear up controversy surrounding in-hospital glycemic targets. Although the 2% sounds very low, we believe nurses in hospitals become very worried about very low blood glucose although the same is not necessarily as true for high blood glucose.
  • Computer-based systems can also benefit the transition to subcutaneous insulin in the hospital. Protocol suggest that the transition to subcutaneous insulin is best managed in the hospital following recovery; if the patient has known diabetes or an A1c of over 6.0%, they should be initiated on basal/bolus therapy. Dr. Bode indicated that the Glucommander could facilitate this process, determining the patient’s 24-hour insulin requirement through the use of a simple multiplier. The regimen is then calculated with one-half of the requirement administered as basal and one-half administered as total bolus therapy. A study is ongoing examining this application of the Glucommander, with outpatient values measured using Sanofi’s BGStar and sent to a server for monitoring by a nurse – in this manner, Dr. Bode suggested computerized dosing of subcutaneous insulin would be “the way of the future.”


Symposium: Innovative Tools for the Treatment of Diabetes


David Simmons, MD (Chief Medical Officer, Bayer Diabetes Care)

Dr. Simmons discussed Bayer’s continuous glucose monitor in development – a product that the company first mentioned publicly at last year’s ATTD (see coverage at, and for which we last saw data at the 2011 Clinical DTM in April (see coverage of Bayer’s poster at A big focus of this latest talk was BGM integration, which Dr. Simmons announced will involve Bayer’s highly accurate Contour Next strips (see DTM 2011 coverage at rather than the Contour strips as previously planned. Dr. Simmons also highlighted positive preliminary accuracy data from a recent weeklong study, described the strength of the latest adhesive patch, attested to the fidelity of signal transmission (a key issue that we have heard mentioned many times at this meeting, especially by closed-loop researchers), and emphasized the importance of comfort (yes!). We are glad to see that work is moving forward, though we had hoped that Dr. Simmons might discuss the near-term timeline.

  • Dr. Simmons presented preliminary data on the first eight patients (n=16 sensors) from a recent weeklong study of CGM accuracy. Clinic visits took place on day one (which included a meal challenge) and day seven (which included an acetaminophen challenge); during the week patients wore the CGM in standard home use. Over both clinic visits, mean average percentage difference of 16.4% (for reference values at or above 75 mg/dl) and mean average difference of 11.3 mg/dl (for reference values below 75 mg/dl). Broadly speaking these numbers do not seem indicative of an advance beyond currently available CGM. Of course, comparison across studies is imperfect at best due to differences in study design, patient population, statistical analysis, etc. – especially for early-stage studies that don’t necessarily follow a standardized format for data reporting, and especially when the sensor and algorithms are still in development as in this case.
  • The Bayer CGM features a straightforward startup process, with calibration that can be performed after only one hour. The patient connects the receiver, transmitter, andadhesive patch, then attaches the entire assembled product onto his or her body and inserts the sensor with a needle-free insertion mechanism (as we understand it, the 33-gauge signal wire itself acts like the insertion needle). One-hour calibration has been validated in an early clinical study (n=16 patients, 32 sensors), which indicated that one-hour and three-hour calibration are not statistically significantly different (mean average percent difference 17.4% for one hour vs. 18.8% for three hours). Thereafter, twice-daily calibration is required throughout the sensor’s seven-day wear time.
  • Early data suggest that once the adhesive patch is attached, it tends to stay attached.  Dr. Simmons said that Bayer’s latest patch uses a new, improved adhesive (as demonstrated in two standard adhesive performance bench tests) on a non-woven material selected for strength, flexibility, durability, and breathability. In previous pilot studies, adhesive patch failure rates were approximately 10% after seven days. But in the most recent seven-day pilot study (for which preliminary accuracy data were reported above), no failures occurred in the first 14 patients (who each wore two patches at once). The last two patients of the study are expected to complete the study early next week.
  • Another emphasis of Bayer’s R&D program has been signal telemetry – an area where current CGM often falls short, as we’ve heard in several artificial pancreas talks at the meeting. The transmitter sends data to the receiver every minute, at a range of three meters (roughly 10 feet). In case of signal interruption, the transmitter also features up to 48 hours of “walk-away” data storage and recovery (allowing for interruptions during showers, airline takeoffs, receivers that get lost in the car for a day, etc.). Once the transmitter and receiver synch up again, the old data can be processed by the system (as long as a valid blood glucose reading is available for calibration). In a seven-day home-use study (120 days of data total), the latest version of the antenna performed with high fidelity. Specifically, 96.3% of data were transmitted on the first try, 99.1% were transmitted within five minutes, and 100% were transmitted within 56 minutes. In the seven-day pilot study of sensor accuracy, interstitial glucose readings were available for roughly 97.6% of wear time. (Dr. Simmons indicated that the remaining 2.4% of readings were lost due to issues with expired calibration or the algorithm rather than signal transmission.) As for data transfer from the CGM to computers, the receiver features a USB connection for direct downloading, similar to the one used in the Contour USB blood glucose meter.
  • We were glad to hear Dr. Simmons refer to comfort as “another critically relevant area,” considering that comfort affects who uses CGM, how often they use it, and how much the technology improves their overall quality of life. On this note, he briefly discussed data from a comfort questionnaire given to 31 patients. Subjective factors like comfort can be difficult to quantify precisely (especially without a clear comparator), though this early patient feedback seems positive.


Strongly disagree


Neither agree/ diagree


Strongly agree

I experienced little or no discomfort from the sensor












I experienced little or no itching from the adhesive
















It was easy to calibrate the CGM system   1   10 20
I knew when to calibrate the CGM system       11 20
The CGM system was easy to use       9 22
The CGM system was easy to wear   2 6 13 10



Howard Zisser, MD (Sansum Diabetes Institute, Santa Barbara, CA)

Dr. Zisser gave a broad overview of what’s holding us back and what we can look forward to in artificial pancreas development. The talk summarized a paper written by Dr. Zisser in the Journal of Diabetes Science and Technology in September 2011. Dr. Zisser considers time lag (both sensor and insulin action) the primary hurdle in the development of an artificial pancreas. To get around these problems, he reviewed some of the approaches in the works: intradermal needles (BD), hyaluronidase (Halozyme), intraperitoneal (Roche; Medtronic), heating the infusion site (InsuLine), pramlintide (Amylin), and inhaled insulin (MannKind). On the latter, Dr. Zisser noted that the FDA recently approved a closed-loop study using Afrezza for pre-meal priming boluses. The study should be completed by the end of the year with data expected in 2013. On the topic of better sensors, Dr. Zisser focused on the Dexcom G4 in this Bayer symposium – as we learned in Dr. Tom Peyser’s talk, the special version of the Dexcom G4 for the artificial pancreas will use a different receiver with new algorithms tailored to the needs of researchers. Dr. Zisser closed by discussing the solid progress on remote monitoring and outpatient studies, including the recent outpatient DREAM study summarized earlier in the day and a UCSB poster on remote monitoring.



David Simmons, MD (CMO, Bayer Diabetes Care); Howard Zisser, MD (Sansum Diabetes Research Institute, Santa Barbara, CA); Andrea Scaramuzza, MD (University of Milan, Italy); David Harlan, MD (University of Massachusetts Medical School, Worcester, MA)

Questions and Answers

Dr. Zisser: David, who funds the project?

Dr. Harlan: The MyCareTeam project is privately funded pretty much. I’m currently paying from my own budget and we have a grant application out.

Q: I’m a pediatric endocrinologist. With small children, it’s hard to do the 15-minute pre- meal bolus. Parents don’t know if the children are going to eat. I’ve been following Peter Chase’s work by telling them to do at least 50% before the meal. Then, if the food is completely eaten, the rest of the bolus is given.

Dr. Scaramuzza: We’ve tried square boluses. From what we’ve seen, it’s always better to inject the bolus prior to the meal. Otherwise, they have a spike after eating.

Q: But what about only 50%?

Dr. Scaramuzza: I don’t know how to answer you. I think we have to do some investigation.

Dr. Zisser: I think doing 70% would still give you risk of hypoglycemia.

Plenary Lecture (Sponsored by Novo Nordisk)


Bruce Bode, MD, FACE (Atlanta Diabetes Associates, Atlanta, GA)

In the Novo Nordisk-sponsored plenary lecture, Dr. Bode reviewed clinical evidence for the use of the three rapid-acting insulin analogs on the market (Eli Lilly’s insulin lispro, Novo Nordisk’s insulin aspart, and Sanofi’s insulin glulisine) in insulin pumps. Following a review of the physiochemical composition of the analogs, Dr. Bode questioned whether variations translated to meaningful clinical differences. His discussion centered primarily on rates of unexplained hyperglycemia and infusion set occlusions between insulin aspart and glulisine – while an initial trial suggested some trend towards increased line occlusions and unexplained hyperglycemia with aspart, a later randomized trial reversed these findings, showing higher monthly rates with glulisine versus aspart.

  • The rapid-acting analogs have differences in their physiochemical composition.  While similar in principle (small changes in amino acid structure to promote rapid formation of active insulin monomers), Dr. Bode reviewed the subtle differences in makeup employed by the manufacturers, such as the inclusion of a detergent in insulin glulisine to promote stability versus zinc in insulin lispro and aspart. Given the tendency to precipitate, these differences have led to slight variations in the FDA’s temperature and storage time recommendations for the analogs when used in insulin pumps.
  • Dr. Bode questioned whether this physiochemical variation translated to meaningful clinical differences. On aspart versus lispro, Dr. Bode showed an initial 16-week study of pump use in type 1 patients that indicated insulin aspart performed better at controlling postprandial glucose levels following dinner with less overall hypoglycemia, though there were no differences in A1c; he also noted some limited case studies suggesting some evidence of precipitation and occlusion with lispro. On aspart versus glulisine, he presented a later 16-week trial suggesting no differences in primary endpoints but some trend towards increased line occlusions and unexplained hyperglycemia with aspart, which FDA allowed Sanofi to include in their labeling. However, a study of five-day simulated pump delivery (at a low infusion rate [0.1 U/hr] to promote occlusions) with all three insulins instead indicated the lowest risk of occlusions with aspart, though Dr. Bode noted that only two-day use was approved at the time.
  • Dr. Bode concluded with a final randomized trial in which patients used each of the three analogs in succession. While no differences occurred in the overall number of patients with >1 unexplained hyperglycemia and/or confirmed infusion set occlusion, monthly rates of unexplained hyperglycemia and occlusions were higher with glulisine versus aspart, in contrastwith previous trends. Glulisine also showed an increase in hypoglycemia (though Dr. Bode noted glulisine at the time was a new insulin no one had used before) but no apparent differences in A1c.

Corporate Symposium: Type 1 Diabetes Technology and Therapeutics in the US – A Report from the T1D Exchange (Sponsored by T1D Exchange)


Dana Ball (Program Director, Helmsley Type 1 Diabetes Program, Boston, MA)

Mr. Ball opened the session with an introduction to the Helmsley Charitable Trust’s type 1 diabetes program and the T1D Exchange. As Mr. Ball described, the type 1 diabetes program is dedicated to improving the speed of type 1 diabetes research and development through the creation of new tools and collaborations. The grant portfolio has thus far included $36 million to research, $19 million to technology (Mr. Ball indicated a large donation to this sector was due in 2012), $12 million to outreach, and $33 million to systems – about 95% of this systems donation has gone toward the T1D Exchange. As a reminder, the T1D Exchange is a cohort of 25,000 type 1 diabetes patients in the US. The Exchange includes a heavily detailed clinic registry, to be integrated with information from a Biobank; it also includes a patient website “Glu,” which serves as a community forum and blog for patients involved. In the Q&A, Mr. Ball indicated that the Exchange continues to expand as well, with aims in the future to leverage its clinics as a trial network to facilitate research. To say the least, we sensed great excitement for the potential applications of this data during the session – we are incredibly grateful for the work put into the T1D Exchange thus far and look forward to seeing how the data will benefit patients, physicians, and researchers in the future, particularly in shaping policy and reimbursement.


Richard M. Bergenstal, MD (International Diabetes Center, Minneapolis, MN)

Dr. Bergenstal described characteristics and demographics of the participants in the T1D Exchange Clinic Registry. To date, the Registry includes 22,300 type 1 diabetes patients from 67 clinical centers across the United States, many smaller and not associated with an academic institution to better reflect -- as many asserted in the Q&A – “real life.” Data collected from patient questionnaires and medicalrecords indicated a broad representation of ages, ages of diagnoses, and disease duration in the dataset, though perhaps reflecting the nature of the disease Caucasian patients (83%) were in the vast majority. To be explored in later sessions, Dr. Bergenstal suggested the data collected would help answer whether various treatments and practices (particularly insulin delivery method and frequency of blood glucose monitoring) actually improved patient outcomes in real life settings. He indicated that the data could challenge or confirm many of the assumptions held about diabetes care as well, noting in particular low rates of weekly downloading (at most 2% in patients over 13) and CGM use (at most 3% in patients under 25 though 14% in patients over 26 – we believe actual use nationally is even lower); percentages of patients experiencing severe hypoglycemia (11-14% in patients over 26) or DKA (4-10% across all ages) in the past year were also surprisingly high.

  • The T1D Exchange Clinic Registry includes type 1 diabetes patients from 67 clinical centers across the United States. Many were smaller and not associated with an academic institution, reflecting – as many asserted in the Q&A – “real life.” Recruitment has increased steadily since initiation, with about 22,300 participants to date and a goal of 25,000 participantsby March 2012. Dr. Bergenstal indicated the program would be employing targeted methods to increase recruitment in underrepresented areas in the future.
  • The Registry includes responses from a detailed participant questionnaire as well as information from medical records. The questionnaire was designed to focus on personal use data unavailable in medical records, such as demographic information, initial therapy, history of severe hypoglycemic events, insulin practice, glucose monitoring habits, and lifestyle data; importantly, questionnaires were completed anonymously rather than administered by the clinician, in hopes of reducing biases in reporting.
  • Speaking to the strength of the dataset, demographics indicated a broad representation of ages, ages of diagnoses, and disease duration. Patient ages ranged from less than 6 to over 50 years, with the majority from 6-17 years; diabetes duration ranged from under 1 to over 50 years, with the majority from 1-9 years. Both genders were represented equally, though perhaps reflecting the nature of the disease Caucasian patients were in the vast majority (83% vs. 5% black, 8% Hispanic, and 1% Asian). While a large proportion (34%) of patients fell in the >$100,000 bracket of annual household income, a broad distribution of incomes was represented (11% <$25,000, 19% $25,000-$50,000, 11% $50,000-$75,000, 18%$75,000-$100,000).
  • To be explored in later sessions, Dr. Bergenstal suggested the data collected would help answer whether various treatments and practices actually improved patient outcomes in real life settings. Data collected included insulin delivery method (pump and injection/pen use were divided mostly 50-50 in individuals of 6-25 years – a potential overrepresentation of pumpers – with greater injection/pen use [66%] in patients <6 years of age and greater pump use [~60%] in patients >25 years of age) and self-reported frequency of SMBG testing (the majority testing 3-4 [31%] or 5-6 [34%] times per day), addressing some long-debated issues in diabetes care.
  • Dr. Bergenstal also indicated that the data could challenge or confirm many of the assumptions held about diabetes care. As he presumed many expected, self-reported rates of downloading were low, though possibly even lower than anticipated (at most 2% of patients downloading at home at least once a week in patients over 13, with slightly higher rates of 4-5% in younger patients) – while lack of reimbursement is likely a factor, he suggested such low rates compelled further research. Similarly, despite the large number of presentations on the technology at scientific meetings, he indicated that CGM use in the dataset was relatively low (at most 3% in patients under 25, though 14% in patients over 26) and asked what barriers remained to wider adoption. Rates of severe hypoglycemia (4-7% of patients under 25 having one or more severe event of seizure/loss of consciousness in the past year; 11% in patients 26-50 and 14% in patients over 50) and DKA (4-10% across all ages) were high. Some said this suggested that they may not be what many have considered complications of the past, though we certainly have in now way seen severe hypoglycemia and DKA as complications of the past – we consider them both grim realities of life with diabetes today for many patients.



Satish Garg, MD (University of Colorado, Denver, CO)

Dr. Garg presented a wealth of rich data on the use of self-monitoring of blood glucose (SMBG) and continuous glucose monitoring (CGM) in the Registry. In the cohort of SMBG users, demographics of frequent testers were as to be expected, with patients on insulin pumps and of higher incomes testing more frequently. In what Dr. Garg dubbed “the most important slide of the presentation,” results indicated a near 2.0% improvement in A1c in patients testing ≥10 versus 0-2 times per day – this relationship was maintained when results were broken down by age group and into patients on pumps versus MDI as well, provoking Dr. Garg to ask “How many more randomized trials does one need?” A similar trend was seen amongst CGM users, with about a 0.5% benefit in A1c versus non-CGM users across all age groups, with mean A1c values declining with increasing frequency of use. About 65% of patients reported discontinuing previous CGM use, with most frequently cited reasons including not wanting to wear the device anymore, too many alarms, accuracy concerns, and cost - we believe this is similar to SMBG and pump use in earlier days, when development of technology was in earlier days and there were still many aspects of the technology that could be improved. From a patient perspective, it is frustrating to see that regulatory delays are keeping so many patients from better technology. Overall, these results are a positive for CGM and monitoring companies, with a clear demonstrated benefit for their products – results we hope will be presented to payors and policymakers – though the data certainly make the current obstacles to wider adoption and use more apparent as well. It is clear that companies must be nimble in improving their products and we believe this is probably quite challenging in the current regulatory environment.

  • The SMBG cohort included 16,287 participants with data on self-reported SMBG not using a CGM. Patients tested roughly 5-6 times per day on average, with frequency of testing generally greater in younger patients (6.5-7.0% in patients ages 1-12), which Dr. Garg said likely reflected more intensive parental care. Patients of both genders tested with similar frequency.
  • As could be expected, patients on insulin pumps tested more frequently than those using syringes/pens. Of patients testing ≥7 times per day, 70% were pump users; of patients testing 0-2 times per day, 60% were using syringes/pens. Similarly, and unsurprisingly, patients of higher income were more likely to test more frequently, with ~60% of patients testing ≥7 times per day in the ≥$75,000 annual income bracket.
  • In what Dr. Garg dubbed “the most important slide of the presentation,” results indicated a near 2.0% improvement in A1c in patients testing ≥10 (7.6%) versus 0-2 (9.3%) times per day. Dr. Garg felt that these results suggested testing even more frequently than standard recommendations produced increasing benefit, as mean A1c declined steadily with increased frequency of testing (8.5% with 3-4 times per day, 8.1% with 5-6, and 7.9% with 7-9); this relationship was maintained when results were broken down by age group and into patients on pumps versus MDI. He made particular note that amongst patients over 25 the ADA target A1c was only approached as frequency neared 7-9 times per day, asking “How many more randomized trials does one need?”
  • Of CGM users (n=1,123; 6% of the cohort), 43% were on Dexcom versus 57% on Medtronic – brand data was not broken down by insulin delivery method, though previous reports have indicated penetration of Dexcom amongst MDI users. CGM use was lower amongst individuals under 25 (at most 3% vs. 14% in those over 25), which Dr. Garg suggested likelyreflected challenges in reimbursement in younger individuals. We believe that younger patients may also be less willing to wear more technology, though it is difficult to say.
  • As might be expected, the vast majority of CGM users were on insulin pumps. This was true across age groups, with the percentage of pump users hovering above 80% (near 90% in patients ages 6-17) amongst CGM users. Interestingly, despite having the sensor data available and calibration required only twice daily, many CGM patients continued testing with regular frequency (~30% testing 3-4 times per day and ~30% testing 5-6 times, with only ~5% testing 0-2 times). However, regardless of this maintained vigilance, a strong percentage of patients reported never downloading CGM data (around 40-50% amongst patients over 18) – Dr. Garg worried that without downloading, many patients were not seeing the full benefit of the technology, thus increasing the chance of discontinuation. We believe “hassle factor” with downloading, particularly perceived hassle factor, continues to be high. CGM use was more prevalent amongst non-Hispanic white patients, patients with income >$50,000, and patients with private insurance.
  • About 65% of patients reported discontinuation of previous CGM use. Rates of discontinuation were highest amongst individuals of 13-17 (82%) and 18-25 (79%) years of age. Of the many reasons for discontinuation, those cited most frequently (patients could report more than one reason) included not wanting to wear the device anymore (33%), too many alarms (22%), accuracy concerns (19%), and cost (19%) – speaking to the need for improvements in accuracy, form factor, and ease of use – with only 15% suggesting the technology was not helpful. Dr. Garg felt increased patient education could also reduce discontinuation rates.
  • CGM users showed about a 0.5% benefit in A1c versus non-CGM users across all age groups. Similar to SMBG results, mean A1c values decreased with increasing frequency of use (7.9% with less than 3.5 days per week of use down to 7.4% with greater than 6 days). Interestingly, in patients under 26 years, both pump use and CGM use produced increasing benefit in A1c (8.7% with injection only, to 8.2% with pump only and 7.7% with pump and CGM), while in patients over 26 those on CGM showed similar A1c values regardless of insulin delivery method (7.4% with injection and CGM vs. 7.3% with pump and CGM) – Dr. Garg suggested this warranted further evaluation to determine which therapies garner maximum returns in various patient groups.


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

Dr. Hirsch presented some stunning data from the largest group of insulin pumpers ever studied. He remarked that understanding what the data is telling us is a unique opportunity. 55% of the T1D Exchange sample used an insulin pump – a group of over 10,000 people. In this sample, a large majority use Medtronic pumps, followed by Animas and Insulet. Insulin aspart (Novolog, Novo Nordisk) is the preferred insulin, particularly in children for reasons unknown. The headline news is that use of insulin pumps is associated with a lower A1c than MDI for all ages. (We will see later that pumps improve incidence of DKA and severe hypoglycemia too). Younger people bolus more and use the bolus calculator more; although this did not confer an A1c advantage, we wonder about glucose variability broadly. Patients aged 26-64 leave in the infusion set for longer than recommended. Pump use is also greater for higher socio-economic status. There is a hugely exciting opportunity to further explore this dataset for more clinical insight, and to assist in reimbursement and setting public policy.

  • The T1D Exchange sample consists of 18,428 patients of which 55% (10,065) used an insulin pump. Of the remaining patients injecting insulin, 46% used insulin pens, 37% used vial and syringe and 17% used both pen and syringe.
  • Some of the key findings for pump usage were:
    • Pump use rises across age groups, ranging from 43% in the <6 years cohort to 60-63% in the 26-64 age group. 56% of the over 65 segment wear pumps (who are covered by Medicare).
    • The majority of pumpers use a Medtronic pump, but there was a decreasing trend with age (see table for market share by age group):


<12 years old

13-17 years old

>18 years old













  • Pumps are not perfect: 39% of Animas users returned their pump to the manufacturer at some stage for some problem, compared with 25% of Medtronic users and 48% of Insulet users (this is not really apples to apples comparison, since Insulet patients could have sent back either a pod or a handheld).
  • Novolog (insulin aspart, Novo Nordisk) is preferred overall, with a stronger preference in children (see table for market share by age group):

<12 years old

13-17 years old

>18 years old













  • Younger people bolus more per day (6.2-6.5 up to 25 years), but the number of boluses declines to 4.8 at over 65 years old. Dr. Hirsch noted that the reasons were unknown - perhaps older individuals don’t snack as much as youngsters!!!
  • Overall, 71% of pump users reported using a bolus calculator as a method to decide how much insulin to take. There is no difference in usage between manufacturers. However, bolus wizard usage declines smoothly with age. Interestingly people who use the wizard more don’t have appreciably different rates of severe hypoglycemia or better A1cs.
  • The median length of infusion set wear is three days. The number of days’ wear varies with age – from 2.8 days for youngest, to 3.3 days for the 26-64 age group. But there is a wide range – from 1-7 days in teens and adults.
  • Pump use appears to be strongly associated with a higher socio-economic status.  Double the percentage of whites use pumps compared to blacks, and pump use is associated strongly with higher income and access to private insurance.
  • The data show categorically that A1c is lower for pumpers than patients using MDI for any age group. The A1c difference is ~0.6% for the 18-25 age group, for example. The gapnarrows with age, as A1cs get lower. We also get essentially the same result if we adjust for socio- economic status (since pump use is highly correlated).
  • The percentage of people who pumped for at least three months and then stopped was 14% in the 18-25 age group (notorious for body image concerns). For those 31 and older, it was in the range 7%-9%. The most common reasons for stopping are problems with the pump, and ‘didn’t want to wear.’ Other reasons included cost, pump not helpful, and problems with DKA (which we imagine reflects some user error).
  • This is the largest survey of insulin pumpers ever recorded. It’s important to remember that it’s all self-reported. It’s observational and may not be representative of all pump users in USA.


Steven M. Willi, MD (The Children's Hospital of Philadelphia, Philadelphia, PA)

Dr. Willi’s objectives for this presentation were to assess the incidence of DKA and severe hypoglycemia (SH) and to identify clinical and demographic factors that are associated with acute complications and to examine how they relate to insulin delivery method. About 7% of respondents have had one or more episodes of DKA with the majority (5%) having had only one episode in the last year. ‘Tricky’ teens and young adults had the most DKA. Patients with lower A1c had much lower rates of DKA. Both pump and CGM use were associated with a reduced incidence of DKA. Overall, it appears that a certain maturity and using the latest devices to achieve good control greatly lowers the risk of DKA. At the other end of the glucose scale, severe hypoglycemia (SH) – defined as a seizure or coma – was scarily common. 7.5% of patients had at least one episode in the last 12 months, 4% had two or more, and 0.75% had 9 or more. There is a ‘frequent flyer’ effect here. In contrast to DKA, children and young adults had less severe hypoglycemia, which could be because of more physiological resilience. Hypoglycemia rises with age to levels higher than the DCCT. So no progressem> there. Surprisingly, there is no clear relationship at all between A1c and the incidence of severe hypoglycemia, except possibly for younger people (where higher A1c corresponds with more severe hypoglycemia). Again, both the pump and CGM (to a small extent) are associated with fewer episodes. Although conventional wisdom has been that tighter control leads to more severe hypoglycemia, clearly educated pump and CGM use can get you better health. Finally, it was very clear that both DKA and SH disproportionately affect minorities and those of lower socio-economic status.

  • DKA is defined as an episode of hyperglycemia requiring overnight hospitalization at some time in the last 12 months. Severe hypoglycemia (SH) is defined as an episode of low blood glucose that resulted in seizure or coma in the past 12 months. The dataset consisted of 18,109 people with >1 year duration of type 1 diabetes. Dr. Willi excluded people who hadn’t been pumping for more than a year in some of the analysis, since there would be confusion over their method of insulin delivery at the time of any DKA or SH episodes.
  • About 7% of respondents had one or more episodes of DKA in the last 12 months. 5% had only one event, and 2% more than one event.
  • The key learnings for DKA incidence were:
    • Teens and young adults had the highest incidence of DKA – nearly double that of other groups.
    • Blacks and Hispanics had pronounced higher incidence of DKA – more than double that of whites. There was a slight female predominance.
    • There was markedly more DKA in lower income groups – those earning over $75K had only a quarter of the DKA of those earning <$35K.
  • For all age groups, patients that had zero DKA events had a significantly lower A1c compared to those who had one or more events. In the 26-49 group those without DKA had A1c nearly 1.5% lower! When patients are stratified by A1c, there is a strong, consistent trend to more DKA at higher A1cs. Nearly one in four patients with A1c >10% had had a DKA event in the last year.
  • Pump users had a lower average incidence of DKA than MDI patients, except in the over-50 age group, when rates were similar for the two methods.
  • CGM users had a significantly lower average incidence of DKA – patients aged 13-49 had about five times more DKA if they didn’t use CGM, compared to those who did. If we combined both pump and CGM use, then the reduction in DKA is better than just pumping alone. But patients need to use CGM for four or more days/week to effectively prevent DKA.
  • In the last 12 months, over 7.5% of patients had experienced at least one episode of SH. 3.5% had one SH event, but over 4% had 2 or more episode. Indeed, 0.75% had experienced>9 events. There seems to be some kind of ‘frequent flyer’ effect.
  • The key learnings for SH incidence were:
  • Adults over 26 had an increasing incidence of SH, markedly higher than in the DCCT, although most children and young adults were well below that threshold.
  • Blacks had double the rates of SH compared to whites and Hispanics, and there was no bias by gender.
  • There was a strong association with lower socio-economic status (SES) - similar to DKA - double the rate from low income to high-income patients.
  • Somewhat surprisingly, there was no relationship at all between A1c and the incidence of severe hypoglycemia. If patients are stratified by A1c, there is potential evidence of a trend to more SH with increasing A1c for children under 12, but no relationship above 12 years.
  • Insulin pump use is associated with about a 25% lower incidence of SH, with the exception of the over 50 group, where rates are equivalent between pumps and MDI.
  • The relationship between CGM and SH is more complex. CGM use is generally associated with a marginally lower incidence of SH, except in the ‘tricky’ 13-25 age group, where CGM use appears to make things worse. (However, there could possibly be some selection bias here, e.g., CGM is prescribed to teenagers because a history of hypoglycemia was identified). There is also no clear difference between occasional and continuous CGM use. We might postulate that CGM allows patients to ‘sail closer to the wind’ – ending up with a similar risk of hypoglycemia but with lower average glucose.



Moderators: William Tamborlane, MD (Yale University School of Medicine, New Haven, CT) and Kellee M. Miller, MPH (Jaeb Center for Health Research, Tampa, FL)

Panelists: Richard M. Bergenstal, MD (International Diabetes Center, Minneapolis, MN), Dana Ball (Program Director, Helmsley Type 1 Diabetes Program, Boston, MA), Irl B. Hirsch, MD (University of Washington, Seattle, WA), Satish Garg, MD (University of Colorado, Denver, CO), and Steven M. Willi, MD (Children’s Hospital of Philadelphia, Philadelphia, PA)

Questions and Answers

Dr. Tamborlane: A few disclaimers. I get a little nervous when people make big claims about this being the “biggest” and “best” dataset available – obviously there are other registries, each with advantages and disadvantages. It’s also not a population-based study, which makes analyses a bit tricky but does not preclude us from making analyses within this group.

Dr. Bruce Bode (Atlanta Diabetes Associates, Atlanta, GA): One thing I was curious about was severe hypoglycemia – in all our studies on CGM, it is associated with a reduction in severe hypoglycemia. I know you asked the question if you are on a CGM. I think the question should have asked if you had severe hypoglycemia when on CGM – what could have happened was that they got severe hypoglycemia and their doctor put them on CGM in response.

Dr. Willi: I have to agree with you, particularly for patients in that particular age group. There is certainly that relationship that hypoglycemia is identified first. Many payors also often develop criteria for CGM, and they include frequent hypoglycemia as part of the justification.

Dr. Bode: One comment also on patient-reported collection. It would be great if we could get true data from pump downloads and meter downloads. Clearly looking at data from CareLink, it suggests 4-4.5 days until change out, which is longer than ours. When you ask a patient they tend to bias towards better appearances. That gives you some bias on what’s truly there and what’s not.

Dr. Willi: Those are all very good points. We tried to improve the fidelity of the questionnaire by making it anonymous, completed ideally online and then directly loading to the Exchange, so we hope that improves the fidelity of the data. I can also say our patient report data tends to be more consistent than the actual medical report data.

Dr. Tamborlane: One where the data is most suspicious is adolescents when they report the number of boluses, but we’re familiar with that bias. There are some surprises in the data, but also some confirmations, justifying what you already believe to be true.

Dr. Garg: Can I just add one more point – some of the patients specifically asked me, saying that they were not going to say if they changed it longer than three days. They thought if their insurance company came to know that they changed really every four to five days, they were going restrict the number of catheters they gave. Despite what I said, they were convinced it could land in the hands of insurance companies. So those are some of the fantasies that affect reporting.

Q: I was really impressed with the data. I was curious though – I’ve seen data for greater declines in A1c with pumps than you found. Why the discrepancy?

Dr. Willi: If you’re asking why our results are different from other centers reporting data, perhaps that’s a reflection of the heterogeneity of our group versus those centers. Many of the centers included are not the type that would pull the data together and would report it in aggregate. A relatively high percentage are small centers, and some don’t exist at academic institutions. That gets to the point that Dr. Bergenstal made that this might be a more of a real world study – a little bit larger slice than we usually see.

Dr. Tamborlane: If you’re talking about meta-analyses of trial data showing greater amounts, in this presentation, we’re looking at one variable. But to really hone it down you need to do a multivariate analysis and control for many other factors such as socioeconomic status, what determines who gets on a pump, etc. So you have to look at it different than randomized controlled trial data.

Dr. Hirsch: To state the same thing a little differently. The meta-analyses were dealing with randomized controlled trials, and this is an observational study. There’s a clear difference in socioeconomic status, self-selection, and maybe even physician selection in who goes on a pump. It needs to be taken into account. This is real life.

Dr. Aaron Kowalski (JDRF, New York, NY): These will be valuable data for reimbursers and clinical care standards, as well as the FDA and other regulators. What is the plan for dissemination and access for other investigators?

Ms. Miller: We have a list of 30 manuscripts we’re working on now. There will also be in the future public datasets available.

Dr. Tamborlane: We had a patient who wanted to know what to expect for the future of his care – how likely and how long it might take for someone in his health until a kidney transplant might be needed. This data is going to be tremendously beneficial for those types of questions as well.

Dr. Stuart Weinzimer (Yale University, New Haven, CT): I think you got to my question before I could, about a linear regression for socioeconomic status and rates of ethnicity because they’re probably covariates. But I guess it hasn’t been don’t yet.

Ms. Miller: For the difference in pump use and A1c we adjusted for that, but that wasn’t done yet for the DKA results.

Dr. Weinzimer: For what Dr. Hirsch said in the use of bolus calculators that there were no differences in A1c, but have you broken that down even into large age chunks?

Dr. Hirsch: Not yet. We certainly want to. Every time we did something in processing this data, additional questions kept coming to mind – the data is so rich. But I think the bigger issue is that the question for the bolus calculator was too broad to answer the question in a way you or I would like to. For instance, I have patients who just use it for correcting highs and not for meals.

Dr. Larry Hirsch (BD, Franklin Lakes, NJ): I may have missed this, but for patients who are injecting have you attempted to split the data into those using syringes versus those using pens?

Ms. Miller: All we’ve broken out is the number using pens versus syringes, and I was about 40-50%. We’ll add it to the list of calculations to do.

Dr. Bode: On SMBG frequency and CGM use, if you wore the CGM more often did SMBG frequency decrease?

Dr. Garg: From the data I showed it doesn’t look like it. It looks like people who did continue CGM use did continue to monitor SMBG 4-5 times a day.

Ms. Miller: We have not broken it down yet by days using CGM.

Dr. Hirsch: We had a huge problem in the state of Washington where the state attempted to eliminate SMBG for children under 19. Fortunately at the last minute we convinced them it made no sense and as a matter of fact they agreed to pay for CGM for children with hypoglycemia unawareness. I actually think one of the things we can do with this data is make a recommendation for our own set of guidelines. It’s very powerful, and I’m tired of having to justify strips because it’s taking away from my ability to justify for CGM.

Dr. Bode: Being an ex-president of ADA, Dr. Bergenstal do you want to comment?

Dr. Bergenstal: I think anyone making policy decisions needs real data. So this sort of data is what we need. I think this puts the pressure on organizations to take this data and say we have a message.

Mr. Ball: The Helmsley trust is forming a project around standards of care with the JDRF – we consider it a first-generation attempt at changing standard of care; this data will give us the second generation.

Dr. Tamborlane: Could you outline how the exchange is looking at leveraging the 67 centers to partner with industry?

Mr. Ball: There are 150,000 patients at the 67 centers. The Exchange office in Boston is expanding, and we are establishing a strategic alliance to act as concierge for our clinical network, the patients, and the investigators. At this point it’s data collection; in the next years we’re going to convert to a clinical trial network. We built it as a system we can drive dollars through. We’ve spent a lot of time developing relationships with all the companies and think that by creating the system we can break down barriers; we want to increase the efficiency of discovery.

Corporate Symposium: Insights into the Future of Glucose Management (Sponsored by Medtronic)


Henk Veeze, MD, PhD (Diabeter, Rotterdam, Netherlands)

Dr. Veeze gave an overview of the current and future status of diabetes technology by mostly focusing on the use of eHealth at the Diabeter clinic (~1,200 patients, 65% on pumps, and 25% on CGM). The team processes an astounding 40,000 emails per year at his clinic and has developed a custom multidisciplinary patient care database. Dr. Veeze and a number of parents actually write programming code for the database, and it uses a CareLink batch export program to increase efficiency. The system has improved outcomes since the center formed a few years ago (no specifics provided). Many during Q&A were quite impressed by the center’s commitment to eHealth and patient contact – we’re certainly glad to see this as well, though we hope the needle starts to move on reimbursement to facilitate the move toward great patient-provider contact.

  • Dr. Veeze discussed the Diabeter model, emphasizing the size of the center is important for cost-efficient operation. He said a minimum of 400-600 patients is required for break even/reasonable facility costs. Diabeter has a physician to patient ratio of 1 to 300-400 and an educator ratio of 1 to 150-200. The clinic has a very low hospital admission rate of 3-5%. During Q&A, he mentioned that the center’s reimbursement is not tied to appointments but an overall negotiation with the insurance company. This helps prevent the eHealth focus from being a major money loser.
  • In-clinic visits to Diabeter are more strategy focused, while eHealth consultations are more tactical. In other words, appointments in the clinic focus on things like “Where doyou want to go with your diabetes? What’s your next step?” The more tactical eHealth focused consultations would address glucose trends, user interaction, education test results, pump settings, and lifestyle events.

Questions and Answers

Q: How many patients have access to this?

Dr. Veeze: It’s touching about 15% of all patients. That number will go up to perhaps 30%.

Q: Are the patients that come to you more motivated?

Dr. Veeze: That might be the case. Half the patients travel more than 15 km. These are very motivated patients. The Netherlands is a small country. Traveling for one hour is the same as sitting in a waiting room for one hour.

Dr. Ragnar Hanas (Uddevalla Hospital, Uddevalla, Sweden): Very impressive. What is the reimbursement policy for insurance companies?

Dr. Veeze: Different countries have different challenges. Germany and Holland count visits. If you shift to eHealth in these cases, you won’t get paid and you must do it on your own expense. We have a case price for one year per patient. We are a non-profit. We do annual negotiations with the insurance companies. There are no ridiculous salaries and money does not drift away. It’s a long-term contract and there is no admin on each patient every year.

Q: Inspiring. I’m wondering about the other members of the team. You talked about the physician and the educator. But what are your numbers for dietitians and social workers?

Dr. Veeze: We have psychologists, dietitians, and nurses are our biggest group. We have a few admin people for weighing and blood pressure and managerial person.

Q: What are the specific numbers?

Dr. Veeze: The dietitian is roughly one-third of appointments. Pediatrics is twice as expensive as adult medicine.


Timothy Jones, MD (Princess Margaret Hospital for Children, Perth, Australia)

Dr. Jones discussed the low glucose suspend (LGS) feature of the Medtronic Veo as the latest evolution of glucose monitoring – a way to address both hypoglycemia incidence and hypoglycemia-related fear. Although CGM by itself can help address hypoglycemia, Dr. Jones said that incidence of severe hypoglycemia is fairly high even in CGM clinical trials (8.5% of nights in the JDRF CGM trial included two consecutive SMBG readings of 3.3 mmol [50 mg/dl] or lower). Also, CGM alarms may not be enough to rouse patients in deep sleep (Ly et al., Diab Technol Thera 2012). The Veo’s LGS functionality has shown potential to reduce the incidence of severe hypoglycemia in short-term studies (Danne et al., Diab Technol Thera 2011; Choudhary et al., Diabetes Care 2011). To evaluate the system on a larger scale (n=100), Dr. Jones and his colleagues have begun a six-month, head-to-head comparison of the Veo vs. standard pump therapy. The study will not complete for several months, but preliminary results (n=74) suggest that LGS events successfully address nocturnal hypoglycemia without severe rebound hyperglycemia or diabetic ketoacidosis. Also importantly, patient satisfaction is high (85% of participants have chosen to continue in a six-month follow-on study) – we hope things go this well in the LGS pivotal clinical trial that is getting underway now in the US!

  • A small recent study by Dr. Jones’ team suggests that hypoglycemia does not statistically significantly affect the amount of noise required to rouse patients from sleep (acoustic arousal threshold), but that CGM alarms may not be loud enough for many patients regardless (Ly et al., Diab Technol Ther 2012). The researchers studied seven adolescents with type 1 diabetes (mean age 14, mean diabetes duration 2.5 years, mean A1c 8.1%) on two separate nights, once under euglycemic clamp and once under hypoglycemic clamp (5.5 and 2.8 mmol/l [50 and 100 mg/dl], respectively). During deep (slow-wave) sleep, the mean acoustic arousal thresholds were not statistically significantly different between the euglycemic and hypoglycemic conditions (79 vs. 71 dB; p=0.353). By comparison, most CGM alarms are closer to 50 db, Dr. Jones said. This difference can help explain why patients so often fail to wake up to alarms, and illustrates a reason why backup measures like LGS can be valuable (as a side note, we are curious to know if this study will spur the development of alarms that can be set louder at night).
  • Dr. Jones presented interim data from his team’s six-month, head-to-head randomized trial comparing the Veo to standard pump therapy. The study includes both children and adults with type 1 diabetes and A1c of 8.5% or less (target n=100); thus far data are available on 74 patients (baseline A1c 7.2%-7.4%). These patients experienced a total of 2,832 low glucose suspend events, 47% of which were short (<10 minutes) and 13% of which lasted the full two hours. No diabetic ketoacidosis (DKA) or severe hyperglycemia has been observed in the study so far.
  • The preliminary results on two-hour nocturnal LGS events look positive overall. Of all 365 two-hour LGS events, 282 occurred at night (accounting for roughly 36% of all 775 nocturnal LGS events). Of two-hour nocturnal suspensions, roughly half occurred because the patient didn’t respond to the alarms, and the other half reflected the patient’s decision to let the event continue (143 and 139, respectively). Using a graph of aggregated data, Dr. Jones discussed the general pattern of two-hour overnight LGS events that occurred before 3 am (n=108). In the typical two-hour nocturnal LGS event, the patient returned to euglycemia by the time insulin delivery resumed (mean two-hour sensor glucose ~5.2 mmol/l [~94 mg/dl]), with the rise in blood glucose continuing for another two hours or so afterward (mean four-hour sensor glucose~8.7 mmol/l [~157 mg/dl] – although data beyond four hours weren’t shown, Dr. Jones said that glucose tends to plateau at this point). We are curious about how these results would look with inclusion of LGS events that occurred after 3 AM; presumably sensor glucose would rise faster and higher in these cases due to the dawn phenomenon. We also note that the typical patient in the study seemed to use an LGS threshold of roughly 3.3 mmol/l (~60 mg/dl), as opposed to the 70 mg/dl threshold that will be used in the US version of the Veo – presumably a higher threshold would translate to higher post-suspension glucose values.
  • So far, 85% of the patients completing the six-month study have elected to continue for another six months. (In this follow-on period, the patients on standard pump therapy continue as before, and the Veo patients are re-randomized either to stay on the Veo or to switch to standard pump therapy.) Dr. Jones interpreted the high rate of re-initiation as a sign of high patient satisfaction. He contrasted this number with historically lower rates of follow-up in trials of CGM alone, arguing that sensor-augmented pumping is more acceptable to patients because it serves a clearer function. Still, the average percentage of time that patients spent wearing the sensors was only 63% (higher in adults, lower in children); Dr. Jones said that frustration with sensor alarms was the biggest factor reducing wear time.

Questions and Answers

Dr. Ragnar Hanas (Uddevalla Hospital, Uddevalla, Sweden): Most people like to sleep at night. Did some of your patients ask for version with the alarms disabled at night, so that the pump would go straight to suspension?

Dr. Jones: Some patients do try to play with the buttons to switch the alarms off at night. But obviously that’s not possible.

Q: Wouldn’t it be better to suspend insulin for less than two hours, to prevent reactive hyperglycemia?

Dr. Jones: I think two hours has been selected as a safe period. After that, mean blood glucose goes back to just over 5 mmol/l (90 mg/dl) at two hours. There was no significant reactive hyperglycemia.

Q: Not right away, but two hours after that there was hyperglycemia.

Dr. Jones: It’s difficult. You’d probably want to customize it for different individuals to account for different pharmacokinetics of insulin. Maybe a future version would let people customize the suspend duration.


Louis Monnier, MD (University Institute of Clinical Research, Montpellier, France)

Dr. Monnier again gave us the benefit of his extensive understanding of glycemic variability. The key points of his presentation today were (1) the contribution of post-prandial glucose to A1c is about 70% for patients with low A1c and only 30% for high A1c, and in absolute terms it’s about 1%, (2) glycemic variability leads to oxidative stress, which it’s postulated then leads to micro- and macrovascular complications, (3) there is experimental evidence which appears to contradict the notion that glycemic variability impacts complications, such as recast data from the DCCT, the Kilpatrick retinopathy paper, and the HEART 2D trial in which the arms had similar A1c, different glycemic variability and the same outcomes, (4) the use of insulin appears to inhibit oxidative stress, explaining the clinical data, and implying that glycemic variability is less important for those treated with insulin, (5) glycemic variability is an important factor in severe hypoglycemia, and post-prandial excursions can disproportionately impact overall A1c. To conclude, Dr. Monnier stated that it would be nonsense to completely exclude post-prandial glucose and glycemic variability as a risk factor for complications.

Corporate Symposium: Alarm Fatigue: A Fresh Look With an Eye Toward the Artificial Pancreas. An Interactive Workshop (Sponsored by Animas)


Howard Zisser, MD (Sansum Diabetes Research Institute, Santa Barbara, CA)

Dr. Zisser led an interactive session (laden with quizzes and prizes – thank you to Dr. Zisser for the copy of Morrissey in Conversation!) on alarm fatigue. As Dr. Zisser described, alarm fatigue refers to the feeling of exasperation and anxiety brought about by frequent alarms. Given patients begin to ignore alarms and quality life declines in the face of fatigue, there is the need to design more intelligent alarms in the future, particularly as devices become increasingly complex. A well-designed alarm should signal to a patient accurately and when further action is needed; the alarm should also reliably catch the patient’s attention and be convenient for the patient. Future directions put forth by Dr. Zisser included remote monitoring (when patients are unable to respond to an alarm, signals can be sent to another party for intervention – he showed the example of a text message sent to family or staff when a patient nears dangerous levels of hypoglycemia), combining multiple algorithms for improved predictive accuracy, dynamic alarms (different signals for different events and at different times of the day), and watchdog alarms (default to a safe setting when there is no patient response). Notably, he suggested considering such factors would become increasingly important in the development of the artificial pancreas, particularly given FDA’s increasing focus on human factors in draft guidance documents.

Questions and Answers

Q: You had a slide with words like “clear” and “recognizable” for alarms. How do you translate those into actual measurable features?

Dr. Zisser: There are people who specialize in human factors and how they respond to technology. There was an article I read about a man who works on iPhone alarms, and he worked relentlessly to perfect that sound that chimes when you unlock your phone. There a people who work in these fields of certain things we take for granted. But you have to have someone specialized in creating that and then test that with users.

Comment: I want you to think about something from a patient perspective – we all have insecurities and to have something on our hip that plays off our insecurities, often we’ll turn it off due to negative reinforcement.

Dr. Zisser: Oftentimes you can’t turn it off. When we were doing early trials with Dexcom, one of our patients went on a trip, and due to inaccuracy the alarm kept going off and there was no “off” button. That’s designed appropriately; you don’t want it to be turned off. But when it’s inaccurate it’s annoying. She didn’t know what to do; those are difficult problems. But you don’t want to be alerting unless you need to do something, and that’s the importance of precision and better accuracy.

-- by Adam Brown, Eric Chang, John Close, Joseph Shivers, and Kelly Close