2012 International Clinical Diabetes Technology Meeting and EXPO

April 20-21, 2012; Los Angeles, CA Full Commentary – Draft

Executive Highlights

Hello from downtown Los Angeles and the 2012 International Clinical Diabetes Technology Meeting. The two-day meeting drew around 300 attendees to its lively discussions and debates on CGM, blood glucose monitoring, insulin delivery, glycemic variability, and information technology. While little new data was shared, we certainly appreciated the nuance and opinion from a number of KOLs.

One major highlight came in the conference’s last session on the artificial pancreas. Drs. Howard Zisser (Sansum Diabetes Research Institute, Santa Barbara, CA) and Bruce Buckingham (Stanford University, Stanford, CA) gave valuable updates on the AP, including both teams’ upcoming trials. Importantly, the speakers have been very encouraged by the recent shift in oversight of the AP to the FDA’s Office of In- Vitro Diagnostics (OIVD) – we hope it translates into a meaningful acceleration of trials and technology.

Blood glucose monitoring was an important part of the discourse as well – we appreciated valuable talks from Dr. Barry Ginsberg (Diabetes Technology Consultants, Wyckoff, NJ) on blood glucose meter accuracy (iBGStar looks quite good, he says) and Dr. Bill Polonsky (Behavioral Diabetes Institute, San Diego, CA) on structured testing. We sincerely look forward to all the innovations coming to patients on the meter from in particular this year.

The session on insulin delivery featured engaging talks by Dr. Bruce Bode (Atlanta Diabetes Associates, Atlanta, GA) on pumps in type 2 diabetes, Dr. Jane Seley (New York Presbyterian/Weill Cornell, New York, NY) on what’s new in pens and pumps, and Al Mann (MannKind Corporation, Valencia, CA) on the benefits of Afrezza (including no mealtime dose titration for type 2s eating a range of carbs). Although the speakers were excited by the new tools in the pipeline, they also highlighted key areas where the needle needs to continue moving: reimbursement, device/data integration, and user- friendlier design.

We also picked up valuable learnings on the CGM front – the discussions included a 45-minute panel with six CGM users (an annual favorite at CDTM) and presentations on interpreting downloads, use of CGM in pregnancy, and when we might see implantable sensors. We particularly enjoyed Dr. Robert Vigersky’s (Walter Reed National Military Medical Center, Bethesda, MD) review of his work on intermittent CGM use in type 2 patients (see our detailed coverage of the Diabetes Care publication at http://bit.ly/zrqLK9). Notably, he has now identified three patterns of patient responses to real-time CGM in the study: “Poster Children,” “Immediate Responders,” and “Burnouts.”

Finally, glycemic variability (GV) came to the forefront during a debate between Drs. Irl Hirsch (University of Washington, Seattle, WA) and Eric Kilpatrick (Hull Royal Infirmary, Hull, UK). There was no shortage of Photoshopped Facebook photos or humorous teasing on either side, but in the end, the audience clap-o-meter seemed to indicate a victory for Dr. Hirsch. He included Helmsley Charitable Trust data to support his point, showing that the recent shift in diabetes care towards reducing GV (e.g., pumps, CGM, MDI) has brought a concurrent drop in rates of diabetic retinopathy. We hope his upcoming GV feasibility trial, FLAT-SUGAR, will shed more light on the issue as it begins recruiting in the coming months.



Detailed Discussion and Commentary

Session I: Self-Monitoring of Blood Glucose


Barry Ginsberg, MD, PhD (Diabetes Technology Consultants, Wyckoff, NJ)

To lead off the session on BGM, Dr. Ginsberg gave an excellent review of meter accuracy. He began by arguing that mean absolute error (also called MAD or MARE) is “a very good measure” of precision and accuracy, much better than correlation coefficient and linearity. Regarding clinical accuracy, Dr. Ginsberg was somewhat negative on error grids, noting they were developed nearly 20 years ago and have limitations in the hypoglycemic ranges. He next turned to what seemed like an obvious question – whether better blood glucose meter accuracy actually makes a difference. We appreciated Dr. Ginsberg’s very factual analysis: he concluded that better meter accuracy makes a significant difference for hypoglycemia and in calibrating continuous glucose monitors. He called accuracy “somewhat important” and something we should continually strive for, but we must do so without compromising other features (price, sample size, etc.). Dr. Ginsberg closed his presentation by explaining the encouraging accuracy in the LifeScan Verio, Sanofi iBGStar, and GE 100 meters. He asserted that few (if any) meters would meet a potential 10% ISO threshold.

  • The best measure of both precision and accuracy is mean absolute error – correlation coefficient and linearity don’t work. Dr. Ginsberg explained that it’s critical to know both precision (how each number correlates with the number before it) and accuracy of blood glucose meters. He called mean absolute error (also called MAD or MARE) a “very good measure” because it reflects both accuracy and precision through use of absolute values. In his view, correlation coefficient and linearity don’t work because they fail to get at the combined accuracy and precision question. For instance, if blood glucose meter measurements are consistently 100 mg/dl higher than the reference, the correlation would be perfect but accuracy would be low. The linearity measurement is also problematic, as it doesn’t reveal anything about variability. A reading could be 100 mg/dl high one time, one hundred mg/dl low the next time, and the average could look good.
    • Audience Response Question (ARQ) #1 – What is the best single measurement of blood glucose meter laboratory accuracy?
      • Correlation – 34%
      • Linearity – 12%
      • Mean absolute error – 27%
      • ISO 95% confidence limits – 27%
  • Clinical error grids (e.g., Clarke, Parkes) are problematic because they were developed almost 20 years ago. Both grids have particular problem in the low range. At the borders of the “A” region in the Clarke grid, a mere one mg/dl point difference in the 50-70 mg/dl range can move a point from the A to D regions. Additionally, Dr. Ginsberg’s noted both grids’ views that “if you’re below 50 mg/dl, we don’t care what the glucose is.”
    • ARQ #2: What is the single best measurement of blood glucose monitoring clinical accuracy?
      • Clarke Error Grid – 26%
      • Consensus Error Grid (Parkes) – 16%
      • Both – 33%
      • Neither – 26%
  • Blood glucose meter accuracy is not a significant issue when compared with the other large errors in calculating a mealtime insulin dose. Meter errors of 6-8% add a small additional amount of error (0-1%) compared to large mealtime insulin errors (10-30%) that patients typically make. Meter accuracy also has a small effect when compared to the large effect of food (a value of 225 mg/dl vs. 200 mg/dl is not a big difference), exercise, and titrating non- insulin medications (uses an average rather than a single point).

Mealtime Insulin Sources of Error

Likely Error

High Error

Carb Counting



Constants (I:C, ISF)



Insulin Absorption



Total Error



Total Error with a 6% meter error



Total Error with a 8% meter error



  • Blood glucose meter accuracy makes a “significant difference” for hypoglycemia. In an in silico study from Marc Breton and Boris Kovatchev appearing in the Journal of Diabetes Science and Technology (2010), sixteen thousand computer simulation trials were conducted using 100 simulated adults with type 1 diabetes to assess the impact of SMBG errors on detection of hypoglycemia. Meters with errors of 20% missed 10% of hypoglycemia episodes; 15% error missed 4% of episodes; 10% error missed 2% of episodes; and 5% did not miss any hypoglycemia episodes.
  • Dr. Ginsberg believes accuracy is “somewhat important” and “We should always seek more accuracy,” but other meter features are also important (e.g., sample size, pain, meter size, cost). To illustrate this point, Dr. Ginsberg compared a YSI analyzer to Sanofi’s iBGStar: a 25 microliter sample vs. 0.5 microliters, high pain vs. low pain, 65-second test time vs. six-second test time, $10 dollars per sample vs. $0.50 per sample. In this case, the better accuracy comes at a major cost.
  • The new LifeScan Verio, Sanofi iBGStar, and GE 100 are very accurate, although few (if any) current/new systems meet the 10% criteria.

20% Threshold

15% Threshold

10% Threshold

LifeScan Verio




Sanofi iBGStar




GE 100



  • A large number of factors limit accuracy: electrode noise, manufacturing variance, chemical variance, coding problems, patients failing to wash hands (“Four out of five times patients don’t wash their hands. The tiniest amount of sugar on the hands can raise blood glucose 100-300 mg/dl.”), hematocrit, urate, citrate, altitude, and temperature.
    • ARQ #3: Under competitive bidding, current strips will cost about 30 cents. For patients with type 1 diabetes, if strips were made more accurate to an ISO 95 of 10%, how much would they be worth?
      • 30 cents – 37%
      • 45 cents – 22%
      • 60 cents – 22%
      • A dollar – 20%

    • ARQ #4: Under competitive bidding, current strips will cost about 30 centsFor patients with type 2 diabetes, if strips were made more accurate to an ISO 95 of 10%, how much would they be worth
      • 30 cents – 69%
      • 45 cents – 19%
      • 60 cents – 6%
      • A dollar – 10%


William H. Polonsky, PhD (Behavioral Diabetes Institute, San Diego, CA)

Dr. Polonsky argued that self-monitoring blood glucose (SMBG) can be an important tool for non- insulin-treated type 2 diabetes patients, provided that they test in a structured manner and with healthcare provider support. Historical evidence is mixed as to the benefits of SMBG in this population, and a recent Cochrane meta-analysis of randomized trials was quite lukewarm: it showed a statistically significant A1c decrease of 0.3% at six months, no significant A1c change at one year, and no effect on patient satisfaction or well-being (Malanda et al., 2012). We believe caution in interpreting these results is critical. Indeed, Dr. Polonsky warned against combining studies of unstructured SMBG (in which there is typically little education or support) and structured SMBG (in which patients are taught how to collect actionable data and to make good use of that data). He said that the latter approach can drive behavior changes in both patients and healthcare providers, in turn fostering glycemic benefits (A1c declines of 0.5% were seen in ROSES, St. Carlos, and the per-protocol analysis of Dr. Polonsky’s own STeP trial). Compliance challenges for patients and time pressure on clinicians will doubtless remain formidable in the real world (and have been seen even in clinical trials). However, we agree that improvements must start with empowering patients to understand their test results and to then act efficaciously.

  • Dr. Polonsky examined the recent Cochrane meta-analysis that found SMBG to be ineffective in non-insulin-using patients (Malanda et al., 2012). The meta-analysis of 12 randomized controlled trials indicated that SMBG led to A1c declines of 0.3% at six months or earlier, but the benefits seemed to go away at 12 months (non-significant A1c decline of 0.1%). However, as Dr. Polonsky noted, 12-month data were available from only two studies, neither ofwhich showed a benefit even at six months. (Thus the issue is not that BGM’s effect “disappears,” as the Cochrane authors say, but that two of the negative studies lasted longer than any of the positive studies.) Overall, Dr. Polonsky warned against making major healthcare decisions based solely on limited data from randomized controlled trials. As one classic paper showed, no strong base of RCT evidence supports the use of parachutes in skydiving, either (Smith and Pell, BMJ 2003). Unfortunately in this tight economy, governments and private payors seem all too eager to cut costs around the margins. Session moderator Dr. David Kerr (Bournemouth Diabetes and Endocrine Centre, Bournemouth, UK) said, “In the UK, the Cochrane view is the only one people are listening to.”
  • Dr. Polonsky said that the Cochrane meta-analysis’ authors “came up with a carefully considered answer” but with one big caveat: “I think they asked the wrong question.” His fundamental criticism of pooling SMBG trials is that study designs can be quite different, making for an apples-and-oranges comparison. He defined “apple” studies as those in which patients are told essentially: “here’s a meter, good luck.” He reviewed ZODIAC, one of the two year-long studies included in the Cochrane Review (Kleefstra et al., Neth J Med 2010), in which patients were not instructed how to reach good control based on test results. He agreed with the authors that the study reflected daily practice in its current state, but he questioned whether it fairly assesses SMBG’s true potential.
  • In what Dr. Polonsky called “orange” studies, patients are taught to use structured SMBG and supported along the way by clinicians. One such study is ROSES, in which patients were asked to test before and after a single meal, three days a week, every two weeks. Dr. Polonsky said that ROSES had some limitations (including a small study size of n=62). However, he believes the significant results on A1c (six-month drop of 0.5%) and bodyweight (six-month drop of 4.0 kg [8.8 lbs]) are noteworthy nonetheless. Another “oranges” trial is St. Carlos, in which newly diagnosed patients (n=161) were instructed to take six-point SMBG profiles every three days, with frequent healthcare provider visits (every two weeks for the first three months and quarterly thereafter). Once again, the results showed an A1c decline of roughly 0.5% in the structured testing group. The St. Carlos data further indicated that those A1c benefits were driven by patients’ changing their lifestyles: by study end, the testing and standard care groups were significantly different in terms of exercise and consumption of vegetables, high-fiber cereals, and nuts.
  • In Dr. Polonsky’s STeP study, structured SMBG led to greater engagement from primary care providers, which contributed to significant A1c declines (Polonsky et al., Diabetes Care 2011). As a reminder, the Roche-funded study included 483 type 2 diabetes patients (not all of whom were SMBG-naïve at baseline – one reason that STeP was not included in the Cochrane meta-analysis, Dr. Polonsky noted). Patients were randomized to an active control group (ACG) that used SMBG without special instructions or to a structured testing group (STG) that used seven-point SMBG profiles for three consecutive days, once per quarter. Despite a large study effect (Dr. Polonsky said that the “usual care” intervention was “way too effective”), structured testing led to a relative A1c drop of 0.3% by intent-to-treat analysis and 0.5% when including only STG patients that were at least 80% compliant with the protocol. (We wished Dr. Polonsky had had time to elaborate on why so many patients were non-compliant even in a clinical trial setting, and what steps clinicians and researchers are taking to address this key issue.) Dr. Polonsky noted that although both groups had baseline mean A1c of nearly 9.0%, primary care physicians were much more active in changing the treatment regimen of patients in the STG arm. He concluded that a structured “snapshot” of glycemia over several days (which helikened to “a poor man’s CGM”) can be a valuable tool to prompt change in clinicians, and that this in turn can lead to better glucose control.
  • Dr. Polonsky outlined four key components of good SMBG interventions. The first is that someone must do something with the data, whether this means adjustment of medications or patient-initiated lifestyle change. The second is that testing be structured, so that patients can practice “discovery-based learning” and come to truly believe the educational principles that they have been told (e.g., exercise lowers blood sugar). Another important factor is that a healthcare provider helps patients to see the data as actionable rather than confusing or guilt-inducing. He noted that for many patients testing is a “de-motivational experience” (fingersticks can be frustrating, the results can be discouraging), but with a good outlook the process can actually become motivational. Finally, Dr. Polonsky said that healthcare providers must help patients see that the changes they make based on the SMBG results are “efficacious” – e.g., that they lower blood glucose over time. Dr. Polonsky forecasted that such patient-provider contact will increasingly occur online and through mobile applications.



Melissa Schaefer, MD (Centers for Disease Control and Prevention, Atlanta, GA)

Dr. Schaefer gave the CDC’s current perspective on disinfection of blood glucose meters in healthcare settings. She highlighted the recent preponderance of hepatitis B, especially prevalent in assisted living facilities of late. The virus can remain infectious in dried blood on surfaces for at least seven days, has a long incubation period (up to six months), and 50-70% of infections are asymptomatic. To prevent infection, the CDC recommends: (1) wearing gloves during blood glucose monitoring, changing gloves between patient contacts, and performing hand hygiene after changing gloves; (2) using single-use, auto-disabling lancing devices for each patient (“fingerstick devices should never be used for more than one person”); (3) never using insulin pens for more than one person (“there is absolutely backflow of blood and skin cells into the cartridge”); and (4) assigning a blood glucose meter to an individual (if possible), or if it must be shared, cleaning and disinfecting the device after every use. We appreciate the CDC’s commitment to this very important issue, though given overburdened nurses, provider time constraints, and cost considerations, we suspect few personnel are actively following all of these recommendations.


David Kerr, MD (Bournemouth Diabetes and Endocrine Centre, Bournemouth, UK), Barry Ginsberg, MD, PhD (Diabetes Technology Consultants, Wyckoff, NJ), William H. Polonsky, PhD (Behavioral Diabetes Institute, San Diego, CA), Melissa Schaefer, MD (Centers for Disease Control and Prevention, Atlanta, GA)

Dr. Robert Vigersky (Walter Reed National Military Medical Center, Bethesda, MD): I’d like to congratulate Dr. Polonsky on what is really a landmark study. What you’ve done is change the conversation and recognition on how important SMBG is. It really is a tour de force what you’ve done and it really will make a difference on patients with type 2 diabetes not on insulin. I’m wondering what physicians in primary care practices were given or told to do to modify their behaviors. In the ROSES study, there was an algorithm on how to interpret the patterns and how to adjust the doses for different medications.

Dr. Polonsky: Some of what we did is online and available. For PCPs, it was really quite surprising to see how little many of them understood about diabetes and blood glucose results in general. As a result of STeP training, they were taken with this idea that you could actually make sense of blood glucose meter data. We walked through a simple formula and a graphical display of the data. This required reviewing medications and some significant training. First, you start looking at hypoglycemia. Then you address fasting followed by postprandial. They liked that you could address these in order. We found that the limitations were dealing with the practices of formularies around the country. Most of the training we did was about medication titration. We touched on other things, which we should talk about over beer – but I believe most of what we did is available online.

Dr. David Price (Dexcom, San Diego, CA): Dr. Ginsberg, I think your comment about the importance of meter accuracy in detecting hypoglycemia was somewhat overstated. Glucose doesn’t just appear at 50, 70, or 80 mg/dl and stay stable. Even at a modest rate of change – say, 1 mg/dl per minute – a 100% accurate meter could measure 80 mg/dl, and within 20-30 minutes the result would be very different. We need sequential measurements for the detection of hypoglycemia, no matter how accurate the meter is.

Dr. Ginsberg: I agree.

Q: Can you comment on the accuracy of blood glucose meters for CGMs and the artificial pancreas?

Dr. Ginsberg: CGMs have significant error. For blood glucose meters, we’re talking about average errors around 6-8%. The average error for CGM is 14-20%. That error comes from a number of factors, not the least of which is that most CGMs are calibrated with a fingerstick monitor. The more accurate the fingerstick monitor and the flatter the glucose is at the time of calibration, the more accurate the CGM will be. The accuracy of a glucose monitor is important but not critical to the artificial pancreas. It depends on where you’re looking. For changes in the basal rate, you’re not dealing with large changes in blood glucose due to meals. Reasonable accuracy is fine. The place where you need true accuracy is around meals to determine if you’ve eaten. The blood glucose could be 90, 90, 90, then 95, 98, 102, 108. For something with 15% inaccuracy, the 95% confidence limits in that case become close to 50%. So for a blood glucose reading of 90 mg/dl, I may get something from 130 to 60. It’s hard to pick up that meal with all that variation. By the time blood glucose is rising, it’s also too late to give insulin. Accurate is somewhat important but less than critical. According to Drs. Frank Doyle and Roman Hovorka, we can do a lot a better with what we have.

Dr. Howard Zisser (Sansum Diabetes Research Institute, Santa Barbara, CA): I have a couple points on that. When we do these trials, usually the sensor is right on, biased high, or biased low – not jumping from 30 mg/dl high to 30 mg/dl low. These sensors tend to track very well. One thing we need to consider is the single-input / single-output model. In cars and airplanes, we are not relying on one sensor. Redundant CGM sensors would take a lot of the noise out. Overnight, this is very simple to do. But when is it most important? I think it’s most important late after a meal when people’s blood sugar is coming down. If they have a nice soft landing, everything’s fine. But if not, and the sensor is biased a certain way, people can have problems.

Q: Dr. Ginsberg, do meters’ accuracy and precision change with duration and frequency of use? If so, is there a critical period when we should change old meters?

Dr. Ginsberg: Not really. Almost all of the devices themselves almost all have internal calibration and self- checking. (I’m not talking about older meters like the SureStat, which is still used in hospitals.) Newer meters check their electronics, and if they are not working internally then they turn off. Each time you put in new strip, you in essence have a new system. One potential problem is deterioration of the contacts between the strips and the meter. This can lead to some inaccuracy even if they deteriorate just a little bit. The greater problem area is strips. Some people recommend using control solution to test the validity of strips. But doing a control solution on the meters themselves is a waste of time – and money, since the hospital meters that allow you to do that are roughly 20 times more expensive than regular meters.

Q: Have you seen any reports or suspicious leads from glucose meters in diabetes camps? Do you have any opinions on this? It’s something we’ve given thought to, but we don’t want it to turn into a hospital stay in the woods.

Dr. Schaefer: We would say standard cleaning and disinfectant precautions still apply there. Camps are an important educational opportunity for the kids – if we don’t do it right when teaching them there, they get older and then grandpa or a friend wants to borrow their meter. I think we’re missing an opportunity there.

Dr. David Horwitz (J&J Diabetes Institute, Milpitas, CA): Dr. Ginsberg, you never answered your own question about the economic value of increased testing. It seems that especially for people on insulin, decreasing rates of hypoglycemia would lead to less ambulance use and fewer auto accidents. Do you have any thoughts on the right price for test strips?

Dr. Ginsberg: I don’t. I wanted to see the recommendations of the audience, who I assume is mostly educators. I was surprised for type 1 diabetes. I would have expected people to pay more for greater accuracy, and I didn’t see that.

Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA): Could you comment on recent CDC recommendations for Hepatitis B vaccine for patients with diabetes?

Dr. Schaefer: The recommendations for Hepatitis B have changed to focus on an older group. I’m blanking on the specifics. The outbreak information was fed in to that.

Dr. Michael Murray (Mayo Clinic Hospital, Phoenix, AZ): Does CDC have any recommendations for how to clean glucose meters?

Dr. Schaefer: I don’t have specific steps. Each meter designed differently. Guidance to manufacturers (selecting EPA-registered cleaning agents) is on website. It would almost be irresponsible for us to make such guidance, since the process varies so much by product.

Dr. Ginsberg: The FDA now requires cleaning instructions to be written before it clears new meters.

Dr. Schaefer: There is also a distinction made between institutional vs. self-use meters.

Q: On those guidelines, is the expectation that HCPs can do it, or must they send it down to central supply?

Dr. Schaefer: The expectation is that the HCP can do that.

Dr. Larry Hirsch (Becton Dickinson, Franklin Lakes, NJ): I have two related questions for Dr. Schaefer. I assume the potential transmission of viral illness – HIV, CV, BV – is directly related to the inoculum. Do you have any thoughts about the transmissibility from lancet fingersticks, which involve a solid cannula going into a fairly richly vascularized part of body, vs. finer-gauge needles that we use for injection? Maybe with pen needles there is still a risk, but it’s less likely as compared to fingersticks.

Dr. Schaefer: The point we want to hammer home is that the risk exists for either. One case of contamination with syringes that comes to mind was at an Endoscopy clinic in Las Vegas. Syringes were being used to go into vials, which then got contaminated. I think the point is not the grade of risk; the risk exists for either. We are not seeing the lancets per se being used from patient to patient – though that may happen – but we do see contamination of the lancing devices.

Dr. Hirsch: In institutional settings, safely engineered devices are really the only option. Hospitals seem to be all over the place in terms of what they do for injections. We’ve been in facilities where they’ve gone from safety syringes to safety pen needles, only to go back to syringes. Do you have any recommendations for safety syringes or safety pen needles?

Dr. Schaefer: We have not released guidance on that. The auto disable syringes have not infiltrated here in the US.

Dr. Hirsch: The other major issue concerns patients newly started on insulin in the hospital. Do you train them with a safety-engineered device? We’re aware of reports where patients reported back to the hospital with DKA because they failed to take off the inner shield of the safety engineered pen needle. It sounds unbelievable but it does happen.

Q: I have two questions. In the Phoenix Veterans Affairs system, where we have over 20,000 patients with diabetes, what would be the procedure for checking for hemoglobin on point-of-care devices? And how do you check to see if one person is not doing quality control properly?

Dr. Schaefer: Our guidance is not just for blood glucose meters – it includes other POC tests. A theoretic risk applies there, too; it’s just a matter of time until outbreak linked. We recommend cleaning and disinfecting so that clinicians don’t have to check for hemoglobin on the meters. I’m not sure the best way to check for individual clinicians not following through; I think we need to make clear to everyone on the care team that this is a real problem, and that we all need to be part of the solution. I’m not sure I’d recommend specific tests within the clinic, unless to really emphasize: look, this problem exists here to. But I think the paper on this topic illustrates the issue nicely.

Q: We also need some guidance on using iPros with multiple patients. Do you clean it? We worry about getting chemical in the transmitter and ruining it.

Dr. Schaefer: I’ll take that back to CDC, thank you.

Dr. Kerr: Is there a need to revisit the guidance for manufacturers of meters, to make cleaning easier?

Dr. Schaefer: The CDC hosted a meeting that brought manufactures to the table to point this out as an issue that patients are facing. We discussed the challenges that some equipment can pose.

Dr. Ginsberg: I think it’s important for us not to throw the baby out with the bathwater. 90% of meters never see any possibility of use by different people. As we increase the burdens on manufacturers, we also increase cost, and then we complain that products are too expensive.

Dr. Schaefer: Yes, but we do need to think about what patients are doing with the meters. It should be clear on the labeling and in the product guide how it’s meant to be used, and this is an important point for clinicians to emphasize in education.

Q: Competition is supposed to reduce cost. In Honolulu, all the strips cost 75 cents a piece. Can you explain that, Dr. Ginsberg?

Dr. Ginsberg: With competitive bidding, it will go down to 30 cents or so in the next year.

Q: On the issue of POC testing and risk of infection, we do A1c checks for all patients. I’m sure they don’t clean the A1c machines between every test. Are there data on this?

Dr. Schaefer: Most of what we know is focused on BG meters. But as I said, the guidance is not just for meters. If it’s a similar device, there is likely similar risk.

Dr. Ginsberg: There is a little difference. With BG meters the clinician you change gloves, pick up the monitor, the device contaminates the new gloves, and then you touch the patient and contaminate them. With the A1c device, you collect the sample, go over to the device, put the sample in, and run the test. Then when you put new gloves on, you don’t touch the machine again before touching the patient. So the risk of contamination is significantly lower.

Dr. Schaefer: Yes, for a truly stationary device, the same risk is not there. But in some hospitals we have seen supposedly “stationary” devices placed at patients’ bedsides. I’m not saying that’s the case here – especially since A1c devices can be so heavy – but it’s important to define what is really stationary.

Q: Have you heard anything about contamination of test strips? What about dirty gloves that go into canisters of 50 strips?

Dr. Schaefer: For someone with bloody gloves, I would be concerned. I haven’t heard of that specifically but that doesn’t mean it couldn’t happen.

Dr. Kerr: Dr. Polonsky, what is the impact of the bolus calculator for non-pump-using patients who are switched to more intensive insulin therapy? In Europe, there are lots of patients on fixed insulin doses. What is the size of that benefit?

Dr. Polonsky: For insulin users, I don’t know. I’m not sure. I’m intrigued by that – I’m leaving it open.

Dr. Ginsberg: I think bolus calculators are very useful. They are not panacea, however. A lot of times the settings that are in them are not changed. But they are better than what Bruce Bode calls “wild ass guesses.” In some cases, they are part of the blood glucose meter themselves, meaning you just add the number of carbohydrates and it gives an approximation. I think they can be very useful. The last data I saw suggested that among pump users, about 25-30% of those who have the capability actually use them.


Session II: Continuous Glucose Monitoring (CGM)


Robert A. Vigersky, MD (Walter Reed National Military Medical Center, Bethesda, MD)

Dr. Vigersky reviewed his well-publicized study published in Diabetes Care last year (see our detailed coverage at http://bit.ly/zrqLK9) and focused on the positive results seen for A1c, weight, medication use, etc. Interestingly, recent analyses performed by his team have added nuance to the 50 patients in the CGM group. After looking at these patients’ glucose profiles over the first 12 weeks of the study (the CGM wearing phase), three typical response patterns have emerged – what he calls (1) “Poster Children,” (2) “Immediate Responders,” and (3) “Burnouts” (examples summarized below). Dr. Vigersky also proposed a conceptual model to summarize the beneficial impact of CGM in type 2 diabetes: perceived autonomy and self-regulation combined with real-time CGM and education leads to competence and self-efficacy. The result is a healthier lifestyle, which translates to a positive, self- reinforcing process. Dr. Vigersky closed his presentation with a few areas for further research – the optimal duration of CGM in type 2 diabetes, whether it is superior to structured testing, and study in larger/more diverse populations. Encouragingly, Dr. Vigersky mentioned that the approach used in his study appears to be cost effective based on “preliminary data.” We certainly hope to hear more about this in the future. We believe CGM as a behavior modification tool for type 2s could be very powerful, although we suspect payors may be hesitant given the cost. If Dr. Vigersky’s intermittent approach is indeed replicable, this could be one valuable way to get around the expense until more data emerges.

  • In a straw poll to start his presentation, the audience was overwhelmingly positive on use of CGM for type 2 diabetes. Dr. Vigersky described a 58 year-old man with 10 years of type 2 diabetes, an A1c of 8.3%, oral triple therapy (metformin, sitagliptin, glipizide), a BMI of 32 kg/m2, a blood pressure of 136/84 mmHg, mild retinopathy, a SMBG one to two times per day. Dr. Vigersky asked the audience if giving the man CGM might help him. Nearly every single hand in the room went up (no hands went up to the contrary).
  • Poster Child: Dr. Vigersky described a patient with an A1c of 11% that declined to 5.9% by the end of 12 weeks. Standard deviation of blood glucose also declined from 50 mg/dl to 26 mg/dl. An immediate effect of the CGM could be seen by the third day, and by the end of the year, the patient had lost 40 pounds! The main change the patient made was simply cutting out bread after noticing how quickly it skyrocketed blood sugars – to us, this speak strongly to the (1) benefits of CGM for behavior change in general and (2) how small behavioral changes can be truly meaningful in the long run.
  • Immediate Responder: This patient began the study with an A1c of 10.1% (~243 mg/dl), modified his behavior (no specifics provided), and dropped his average glucose down to 100 mg/dl within the first week of the study.
  • Burnout: This was perhaps the most interesting example given by Dr. Vigersky. This patient’s standard deviation of blood glucose rose from 27 mg/dl to 65 mg/dl by the end of 12 weeks. Despite getting worse over time with CGM, the patient started with a baseline A1c of 8%. Dr. Vigersky noted that about five out of the 50 patients exhibited this pattern.



Jeffrey Joseph, DO (Thomas Jefferson University, Philadelphia, PA)

In a clear and engaging overview of long-term implantable CGM, Dr. Joseph reviewed the (many!) physiological and technical challenges and mentioned several of the companies trying to overcome them. He explained that in order to minimize the trauma of insertion and the in-body motion of the sensor after it’s been inserted, smaller size is better. Small size is also helpful to reduce the chance of a scar-like fiber forming around the sensor, which would choke off access to oxygen and glucose (this is not necessarily a problem for optics-based sensors, such as the one he is working to develop with Animas). On electrochemical sensors, though, Dr. Joseph was particularly positive on the near-term potential of Glysens’ device (“this one is closer than you think”), which has demonstrated stability in animals for several hundred days (Gough et al., Sci Transl Med 2010). Other companies in the space include Dexcom, Sensors for Medicine and Science, MicroCHIPS, Biorasis, and others. Dr. Joseph concluded with a forecast that implantable sensors could reach patients in Europe in the next two-to- three years, and in the US in the next five-to-six years – a rather optimistic timeline in our view, but exciting to hear even as a best-case scenario.


Howard Zisser, MD (Sansum Diabetes Research Institute, Santa Barbara, California), Robert A. Vigersky, MD (Walter Reed National Military Medical Center, Bethesda, Maryland), and Jeffrey Joseph, DO (Thomas Jefferson University, Philadelphia, Pennsylvania)

Q: A question on the recently released guidelines for type 2 diabetes. This was a patient centered approach. Why was there no mention of CGM in type 2 diabetes?

Dr. Vigersky: Not everyone is as compulsive as you and I. You are referring to the ADA/EASD consensus panel that published their guidelines for management of type 2 diabetes yesterday. It was centered on use of various medications for management. I just scanned it quickly and my take away – they really made a major point of individualization of management. Rather than being prescriptive, it was very much patient centered. They did not mention monitoring at all – SMBG or CGM. And there was no mention of insulin pumps. It was very focused on medication use and individualization. That said, I think that being patient centered means we have to consider what our patients will do. Maybe there will be other questions about this. You could do the same thing with structured testing as I did with CGM. The question is will patients do it, how long will they do it for, and will they do enough to sustain it. Those are questions for RCTs.

Dr. Bruce Bode: I had thought that in your trial you did not teach them what to do with the readings. Is that correct?

Dr. Vigersky: Yes. We just handed it to them, much like the SMBG trials we all criticized earlier. We said that it would show their glucose levels, there would be alarms, and they should follow the advice of their PCP.

Dr. Bode: In the future designs would you recommend reinforcement at six months? Might that prevent some of the burnout?

Dr. Vigersky: I think some of the burnout would have happened no matter what. We don’t know how frequently you have to do it. But one way going forward would be to “dose” CGM for two weeks every three months, for example. Or maybe the protocol should be four consecutive weeks at the start, and then one week every quarter thereafter.

Dr. Bode: Based on [Dr.] Bill [Polonsky]’s earlier talk on structured testing, have you compared to structured SMBG?

Dr. Vigersky: We haven’t. We are about to begin trials in prediabetes; we have applied for an NIH grant to do this. This is another paradigm where CGM could be useful. As everyone in the room that has done research knows, it’s not easy or quick to get these studies off the ground. But you raise an issue that needs to be addressed.

Q: I’m concerned about the burnout patients. How did you determine it was burnout and not another life event?

Dr. Vigersky: We don’t know. We took a look at all 47 patients by spreading out all the graph on the table. We tried to group them by patterns. We just named them “Burnout.” We have no idea what went on. SO there’s still hope.

Dr. Irl Hirsch (University of Washington, Seattle, WA): If you are showing data to demonstrate a reduction in variability, I would suggest using coefficient of variation. When the mean goes down, the standard deviation goes down too.

Dr. Vigersky: When we analyzed the data, we looked at six variability measures. For the purposes of what I was trying to bring out, most people understand standard deviation. I don’t understand CONGA. That’s a very good point.

Dr. Hirsch: In the type 1 CGM studies. STAR-3, STAR-1, patients come in to clinical trials and it doesn’t match real life. That’s why in the Pacific Northwest, CGM is not covered for type 1 diabetes. I’m extremely supportive of what you’re doing. But we’re always going to be criticized for this – it’s a clinical trial. Would we be better off going into Kaiser and offering this to people? As opposed to the way we recruit now?

Dr. Vigersky: You’re between a rock and a hard place. If you did that, you’d be criticized for not having done a randomized controlled trial. There is practitioner here in California. He has given it to a bunch of his patients and it’s not randomized. He has found the same kinds of results. Hopefully that will get published. That will be evidence along the lines you’re suggesting. We need all these kinds of studies.

Dr. Karen Horowitz (Case Western Reserve University, Cleveland, OH): I see some patients at the VA. Patients say, “They tell me to test, and I’m always at 300 mg/dl. I get demoralized and I don’t know what to do, so I stop testing.” Have you looked back at the burnouts to see if the primary care providers actually made any interventions to move them along?

Dr. Vigersky: We just generated the data two days ago so haven’t yet analyzed this. But it’s an interesting suggestion, and we will certainly do that.

Dr. Saleh Adi (UCSF, San Francisco, CA): Could you use CGM in type 2 diabetes to exaggerate hyperglycemia? That way, you show patients they’re going to 300 mg/dl. Would that have a better impact on behavior change? … I’m just kidding. [Laughter] In the initiation of type 2 diabetes management, could putting patients on CGM influence what drugs you use? Would you start some on insulin immediately?

Dr. Vigersky: It really will influence what you do. If you can get an effect of 0.5% or more in the A1c range we’re talking about, 8-10%, you may in some of those patients, obviate the need for adding a medication. I would say the fewer the meds the better if we can effect behavior change that is sustained. As long as it can be sustained, it’s better off than adding another drug. The cost is not trivial with this, but if you dose it one week every few months, it’s pretty low cost to do this.

Q: You have talked about implantable devices. Are there endeavors in industry to develop non-invasive devices, analogous to pulse oximeters?

Dr. Joseph: There have been numerous attempts with near-infrared (IR), mid-IR, and different impedance measurements. Basically, as glucose changes there are osmotic changes, changes in the way light is polarized, and numerous other possible changes. When you use an external device, coupling of device with tissue is so variable that noise exceeds the small signal you have. In an experimental setting with anesthetized animals that are perfectly still, you can get a good reading. But in the real world, movement and other factors create situations where the noise exceeds signal. It really doesn’t work.

Dr. Vigersky: There is a noninvasive device that is worn around the waist.

Dr. Barry Ginsberg (Diabetes Technology Consultants, Wyckoff, NJ): It is from C8 MediSensors. It is not FDA-approved.

Dr. Vigersky: Nevertheless, there is such a device. Whether or not it is sufficiently sensitive, especially at the low end, remains to be seen.

Dr. Zisser: There is also a company called OrSense in Israel, which uses something like a blood pressure cuff to intermittently occlude blood flow. I’m not sure where they are in development.

Dr. Joseph: Again, there have been numerous attempts, and no one has gotten it accurate and precise enough in the real world to dose insulin.


Howard Wolpert, MD (Joslin Diabetes Center, Boston, MA)

Dr. Wolpert gave a clinically focused talk on interpreting CGM data downloads (what he calls, “A systematic approach for making sense out of spaghetti”). The first step is to identify glycemic trouble spots – on the Dexcom software, he uses the modal day chart plotting 24 bars of hourly data (mean, quartile range, median, and extreme values). Drawing lines at 70 and 180 mg/dl, Dr. Wolpert can easily identify times of the day where there are problems (e.g., frequent hypoglycemia or hyperglycemia, lots of variability). On the Medtronic reports, he is a particular fan of the quick view summary, which allows easy integration of insulin and CGM data. Dr. Wolpert transitioned to a discussion of postprandial glucose profiles, explaining that CGM can be helpful in differentiating between early and late postprandial hyperglycemia. He reviewed a few studies of dietary fat intake, which can dramatically increase insulin resistance and insulin requirements – in Q&A, this was a clear area of interest and sounds like an important avenue for future study. He concluded by discussing alarms – he recommends new CGM users should set alarms at 55-60 mg/dl and 250 mg/dl or higher; those with lots of hypoglycemia should set a low alarm at 80 mg/dl or higher; last, alarms should also be changed depending on patient preferences, responses, and sensor lag.



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

In a wide-ranging talk full of clinical insights, Dr. Jovanovic discussed the utility of continuous glucose monitoring in pregnancy. She opened with strong evidence that maternal normoglycemia (i.e., always below 120 mg/dl) is necessary to prevent large birth weight and other factors predictive of lifelong metabolic problems (or even death in infancy). She thus recommends that her patients eliminate refined carbohydrates, take prandial insulin (either Humalog or Novolog) whenever they eat, and use basal insulin round-the-clock (currently she gives patients thrice-daily NPH, but twice-daily Levemir seems to be a better option now that the latter is classified by the FDA as pregnancy category B). She said she used to think that her patients were perfectly adherent and their normoglycemia was perfect, but blinded CGM studies in her clinic revealed otherwise. Fortunately by a similar em>token, frequent use of real-time CGM allows patients and providers to optimize glucose control – though she said that reimbursement is currently difficult in California. She highlighted a few areas where CGM can be especially useful (steroid use, multiple [e.g., twin] pregnancies, and diagnosis of glycemic status early in pregnancy), and she looked forward to a large, JDRF-funded study that will hopefully convince payors that CGM in pregnant women with type 1 diabetes is well worth reimbursing.

  • Dr. Jovanovic shared her excitement that Novo Nordisk’s Levemir (insulin detemir) was recently reclassified from pregnancy category C (risk thought to outweigh benefits) to category B (human studies adequately demonstrate lack of fetal risk). She explained that NPH insulin, the current standard of care, must be given at strict eight-hour intervals. This thrice-daily schedule can create problems when women want or need eight hours of sleep at night – e.g., they might fall asleep without giving themselves NPH and then wake up the next day with hyperglycemia. By contrast, Levemir can be dosed twice daily, allowing patients a more flexible lifestyle. Dr. Jovanovic noted that she does not yet know the best way to dose Levemir in pregnancy, but she is quite excited to now have the opportunity. For more information on Levemir’s relabeling, see the April 10 Closer Look at bit.ly/Io80A0.
  • Presenting a few case studies, Dr. Jovanovic discussed circumstances in pregnancy when CGM can be particularly valuable. One such circumstance is steroid use, since steroids can have dramatic (and unpredictable) effects on the way patients respond to insulin. Also, the way to adjust the typical insulin dose for women with twins (or triplets, or more) is not always obvious, and CGM can help here as well. Finally, Dr. Jovanovic noted that short-term (three-day) use of CGM could be used instead of an oral glucose tolerance test as a way to assess glycemic control during pregnancy. She presented modal-day sensor data from four individual pregnant women – one with normoglycemia, one with gestational diabetes, one with type 2 diabetes, and one with type 1 diabetes – with higher average glycemia (and glycemic variability) in each disease state.
  • Dr. Jovanovic noted that JDRF is funding an international randomized study of real-time CGM vs. intermittent testing in pregnant women with type 1 diabetes toensure good fetal outcomes. The study, called CONCEPT, has a target enrollment of over 900 patients. The hope is that if the results are positive, the state of California (and other healthcare systems) might start reimbursing CGM in all women with type 1 diabetes during pregnancy – perhaps a long shot in some states currently, but quite a boon if it comes through. The investigators are scheduled to meet in September 2012; they hope to have enough data by 2013 to say whether CGM is safe and efficacious in pregnancy.


Howard Zisser, MD (Sansum Diabetes Research Institute, Santa Barbara, CA), Howard Wolpert, MD (Joslin Diabetes Center, Boston, MA), Lois Jovanovic, MD (Sansum Diabetes Research Institute, Santa Barbara, CA), Robert Slover, MD (Barbara Davis Center for Childhood Diabetes, Aurora, CO)

Q: I’m wondering about the timing for meal doses of insulin for pregnant patients. How have you used CGM? We used to think short-acting analogs were absorbed quickly, but now we know that’s not the case. And what about for children?

Dr. Jovanovic: We use aspart and lispro and our patients have no carb in the meals – they are all protein and fat. The postprandial response is from the gluconeogenic amino acids and the vegetables that eventually get into blood stream. We also allow them to drink milk and take multivitamins – the latter is because when we restrict citrus fruits, we worry about Vitamin C intake. They take insulin immediately with the start of a meal and measure blood sugar one hour later.

Dr. Slover: Our pediatric experience has been guided by artificial pancreas research. As you will hear, that research talks about the hybrid closed loop where you pre-bolus for meals. We tell people that unless they are hypoglycemic or under 100 mg/dl before eating a meal, they should take their insulin at least 15-20 minutes ahead of eating. It should be earlier if their glucose is higher.

Dr. Bruce Bode (Atlanta Diabetes Associates, Atlanta, GA): Lois, excellent talk. But I beg to differ – I’ve never ever been turned down for an insulin pump in pregnancy for either type 1 or type 2 diabetes. We’ve never tethered the insulin pump with NPH and I’ve never seen DKA in our patients. I’ve certainly seen hyperglycemia and we always check at 3 am. My worry about tethering with NPH is that it’s so unpredictable and it could confuse us. Once the basal rate is set for pregnancy, it doesn’t change much over time.

Dr. Jovanovic: We use NPH before bed as good luck in case the needle falls out.

Dr. Ginsberg: There is now insulin pump tubing which will generate a high pressure alarm if it falls out. (Editor’s Note: D. Medical’s Spring Universal Infusion Set).

Q: Based on the impact of dietary fat and the need for variation in insulin delivery, has there been thought to a magic formula based on calories?

Dr. Wolpert: We’re going to be doing some studies to address this. It’s a huge challenge due to the inter- individual variability. Extra insulin is also required for two reasons: 1) the fat delay in gastric emptying and (2) free fatty acids causing insulin resistance. It’s very challenging to model this. Initial data, in a small subset of patients, shows that the best predictor of increased insulin requirements from dietary fat is total daily dose. We’ll have to look at a larger subset. It’s going to have to be an iterative bolusing process. What if data show that the type of free fatty acids affects the degree of insulin resistance? Saturated fats are different from poly- or monounsaturated fats. There could also be a dose response. And does the fat to carb ratio matter? There are a host of issues here.

Q: I wasn’t clear on the slide about the enhanced epinephrine response. Was that plasma levels? Behavioral response?

Dr. Slover: That was basal epinephrine levels one year out, comparing the CGM group and the non-CGM group. It was not looking at adrenaline response to a specific event. People using CGM had higher basal levels.

Q: Can you speculate as to the mechanism?

Dr. Slover: I’m not sure.

Dr. Bode: If you prevent hypoglycemia, your epinephrine response stays normal. If you have hypoglycemia, your epinephrine response is depressed.

Dr. Irl Hirsch (University of Washington, Seattle, WA): I’m curious about the information on degludec in pregnancy, especially immunogenicity. It will be on the market reasonably soon.

Dr. Jovanovic: No. There aren’t enough studies to know anything related to pregnancy. Just don’t use it.

Dr. Hirsch: In animal modeling, do we have clues about IGF-1 binding? Eventually over time, we’ll decide if it’s safe or not.

Dr. Bode: Its IGF-1 binding is half of what Levemir’s is. It’s the lowest of any insulin ever developed. There have been some patients in the trials that got pregnancy and some case reports, but they were then immediately taken off degludec.

Dr. Hirsch: For the fatty acid moiety, I wouldn’t expect any difference vs. detemir. But there probably will be reports and we’ll start to do it over time. But I know that it took Novo Nordisk a long time for detemir.


Session III: Insulin Delivery Technology


Jane Jeffrie Seley, DNP, MPH, BC-ADM, CDE (New York Presbyterian/Weill Cornell, New York, NY)

Dr. Seley gave a comprehensive tour of pens and pumps, featuring her wish lists for both devices and a review of upcoming innovations. She was very upbeat about Diamesco’s pendiq (a downloadable pen), Valeritas’ V-Go, and Medtronic’s mySentry and Veo. She also addressed the “complicated” Bydureon injection process. Although administration includes 13 steps by her count, she does believe it can be mastered quickly and is worth it for the once-weekly dosing. We found the audience response questions particularly interesting – attendees were most in favor of pens with date/time/dose memory and pumps with CGM integration. Dr. Seley closed optimistically with a review of futuristic diabetes device prototypes from DiabetesMine.com’s Design Challenge.

  • Dr. Seley’s insulin pen wish list: smaller; more discreet; cool looking; reminders (one of her favorite possible features); dose memory; download capability; easier to depress the plunger when administering large doses (Dr. Seley described an elderly patient with an A1c of 16% that she immediately initiated on insulin; the patient had to use two thumbs to depress the plunger); dual hormone chambers; U200 and U500 insulin pens; integration with glucose meter and smart phone.
  • Dr. Seley was most positive on Diamesco’s pendiq and Novo Nordisk’s Flextouch pen; her feelings were more mixed on Novo Nordisk’s Echo pen and the Bydureon kit. She called Diamesco’s pendiq the “cool factor winner,” highlighting the device’s display and memory – for more information, see page 210 of our EASD 2011 report at https://closeconcerns.box.net/shared/0q4e4a4vrgkujjhs6y18. Dr. Seley also expressed enthusiasm for the Flextouch pen, which is a “really easy pen to use” and doses up to 80 units. Noting the Echo pen’s great ease of use, half unit dosing, and memory, she was not a big fan of the replaceable cartridges – Dr. Seley explained that patients may forget, may not do the refilling process well, or may drop the cartridge and break it. (She did concede that a reusable might be worth given other beneficial features.) Finally, Dr. Seley was frank about the Bydureon injection process, calling it “a complicated procedure for patients to do.” She counted out 13 steps, but noted that it can be learned and mastered in a short period of time. And the fact that “you only have to do it once a week” is “certainly worth it.”
  • ARQ #1: Insulin Pens are more cost effective than a vial and syringe when patients:
    • Are visually impaired – 25%
    • Are taking small doses of insulin – 55%
    • ​Are taking large doses of insulin – 11%
    • Are taking basal insulin only – 9% 
  • ARQ#2: Which one of the following insulin pen features do you think is most important?
    • Integration with glucose meter – 17%
    • Dual hormone chambers – 6%
    • Remembers date, time, and dose – 53%
    • Download capability to a PC or a mobile device – 23%
  • Dr. Seley’s insulin pump wish list: more integration with blood glucose meter and CGM; dual or multi-hormone chambers (could be amylin, glucagon, or GLP-1); pre-filled insulin cartridges (she later called the Asante Pearl “really cool”); more connectivity through Bluetooth,web, and mobile; new built-in algorithms and user friendly carbs lists to facilitate easier extended bolus calculations (a particularly important innovation); photo recognition of carbs on the plate using a built-in pump camera (show knows a doctoral student working on this); fixed dosing geared for type 2 diabetes (she called Valeritas’ V-Go “a really good idea” and “very, very easy”); more patch pumps with bolus buttons on pumps (“how many times does an Insulet user have to give a bolus but the remote is not handy?”); less button pressing to prime and deliver boluses (she gets this complaint from patients a lot); auto shut-off for hypoglycemia; flexible pumps that mold to the body; comfortable infusion sets for thin patients; durable/patch convertible pumps (e.g., Cellnovo); and smaller pump size (singling out the Tandem t:slim, Valeritas V-Go, and Solo Micropump).
  • ARQ #3: Which one of the following insulin pump features do you think is most important?
    • Integration with CGM system? – 61%
    • Dual hormone chambers? – 1%
    • Pre-filled insulin cartridges – 14%
    • More connectivity with Bluetooth, web, and mobile devices – 24%
  • Dr. Seley was very positive on the Medtronic mySentry and Veo (“Innovations we love”). Regarding the mySentry, she liked the device’s dual potential for remote monitoring (parents/caregivers) and for adult patients that don’t hear their alarms or have poor vision. She highlighted that insurance companies aren’t paying for this right now, although “that could change” and “Medtronic has brought the price down to a level people can afford” (Editor’s Note: the device is $3,000 with discounts available; see our coverage of the January approval at http://www.closeconcerns.com/knowledgebase/r/aca26119). Dr. Seley also discussed the Medtronic Veo, which she called “a very nice idea,” a “wonderful option,” and a “no brainer.”
  • Dr. Seley concluded with a review of prototypes from DiabetesMine.com’s Design Challenge. She highlighted the Charmer, which can be worn around the neck or on a key ring. The prototype’s display screen gives a quick indication of CGM status – sunny skies for good blood sugars and thunder/clouds for values that need attention. The Flex, a pump that molds to the body, was another prototype from the challenge. Also intriguing was a combined pump and sensor in a Livestrong-like bracelet (“wouldn’t that be wonderful?!”). The 2011 winner was the Pancreum, a prototype wearable artificial pancreas combining a patch pump, sensor, and glucagon in one unit. The device is controlled from a smart phone.
  • ARQ#4: Which automobile company is working with Medtronic to develop an in-car navigation system to communicate with the CGM? (Editor’s Note: see our coverage of the announcement at http://www.closeconcerns.com/knowledgebase/r/07f9d475).
    • Ford – 45%
    • General Motors – 25%
    • Honda – 14%
    • Mercedes-Benz – 15%

Questions and Answers

Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA): I’m speaking to ask David Kerr to come up and comment. He came up with the idea for CGM in a car and working with Ford.

Dr. Seley: I want to hear more about this.

Dr. David Kerr (Bournemouth Diabetes and Endocrine Centre, Bournemouth, UK): So do I. [laughter] This has been around for a few years. It is still in a very embryonic stage. There is lots of negotiation between Medtronic and Ford. There are other interesting variations of this as well. Ford is interested in heart rate and heart rhythms – using the car as a medical clinic and allowing real-time feedback for the doctor. One example is atrial fibrillation. A sensor measures oxygen levels in car. The most developed is a Japanese company that has an alcohol detector in the gear stick. They also can detect the alcohol level on the breath and it will shut off the car.

Dr. Seley: No one will buy that car.

Dr. Kerr: The idea is if you have a conviction, you have to install it. Watch this space.

Q: I would put reduced cost on the pen wish list. Is there a chance this could happen?

Dr. Seley: It’s a shame that some patients can’t afford it. There’s been a change recently. Pens used to be distributed by the full box. In the last six months in New York and California, patients get the exact number of pens that they need for that time. Maybe now that they are distributing them more wisely, then the pens will be more economical for insurance companies. Those strategies are probably smarter.


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

Dr. Bode gave a comprehensive presentation on pumps in type 2, including the therapy’s rationale, a literature review, the near-term product pipeline, estimated current market size (50,000-100,000 patients), and what we can expect on the clinical trial front. Although the two large RCTs on pumps in type 2 have not shown a benefit, Dr. Bode has faith that Medtronic’s OpT2mise trial will change that – the study has improved upon earlier trials by testing the benefit of pumps in MDI failures (A1cs of 8- 12%), and results will hopefully be published in 2013 or 2014. Dr. Bode made it clear that pumps have numerous benefits in type 2s and the introduction of simpler devices (e.g., Valeritas’ V-Go) should dramatically expand the market. On a separate but related note, we also learned from Dr. Bode that Tandem will be launched at ADA this year and the company has future plans to build a 500-unit (!) reservoir – we wonder if this would be accompanied by a simpler design for type 2s. Dr. Bode also mentioned that U200 pens will be coming out this fall and U300 for Lantus will arrive next year.

  • The biggest problem in treating type 2s with insulin is the reluctance of providers to start insulin. Much of this stems from lack of training – many providers never received the proper training to start people on insulin. Even Dr. Bode’s fellows in endocrinology did not know how to initiate insulin therapy. Dr. Bode noted that the launch of Lantus “helped a lot” and explained that insulin is cheaper than most orals.
  • There are numerous benefits of pump therapy for type 2s: improved glycemic control, reduced glucose toxicity, beta cell preservation, enhanced quality of life, improved patient satisfaction, and greater ease of management. Additionally, given that 89% of type 2s do not inject insulin outside the home (Roper Starch 2006), Dr. Bode believes that patients need more options to discreetly take their insulin.
  • Dr. Bode estimates that there are between 50,000 and 100,000 type 2 pumpers in the US. Medtronic data from 2008 estimated the type 2 pump market at 37,000 patients (<1% penetration) – according to Dr. Bode, people think it’s easily doubled by now. He also believes that the upcoming dedicated type 2 pumps will increase this number “dramatically.”
    • ARQ #1: How many type 2 patients are on an insulin pump in the US?
      • <25,000 – 59%
      • 25,000-50,000 – 30%
      • 50,000-100,000 – 8%
      • >100,000 – 3%
  • Pumps for type 2 diabetes must be cost-effective, simple, and most importantly, discreet. Regarding the latter, Dr. Bode reiterated that patients simply don’t inject insulin outside the home. Unfortunately, current pumps (Medtronic, Animas, Insulet, Roche) have been used in type 2 diabetes, but none have been specifically designed for type 2s.
  • Pump therapy indications are stringent, mostly restricting pumps to insulin deficient type 2s. Patients need an A1c >6.5 or 7%, glycemic excursions, hypoglycemia or hypoglycemia unawareness, and the need for a flexible insulin regiment. Medicare and Aetna have other criteria as well – fasting C-peptide <110% lower limit of normal or <200% lower limit of normal if CrCl <50 ml/min with concurrent FPG <225 mg/dl or beta cell autoantibody positive (+ICA or +GAD antibodies).
  • Two large RCTs comparing pumps to MDI in type 2s have shown no difference in A1c; however, both studies enrolled a suboptimal population and Medtronic’s ongoing OpT2mise trial will be a “definitive trial of pumps in type 2.” Dr. Bode believes the two major RCTs of pumps in type 2 (Raskin et al., Diabetes Care 2003 and Herman et al., Diabetes Care 2005) made an important mistake – they did not compare patients failing MDI to pump therapy. Fortunately, Medtronic’s OpT2mise study, examining pump therapy in type 2s with A1cs between 8-12%, will do just that. The 30-center, prospective RCT has a target enrollment of 400-600 patients (over 120 enrolled thus far) and an estimated primary completion date of December 2012 (clinicaltrials.gov identifier: NCT01182493). We look forward to seeing results hopefully published in 2013 or 2014.
  • Edelman et al., Diabetes Technology and Therapeutics 2010 demonstrated that the majority of type 2s can be controlled on only one basal rate. This 16-week, open-label, multicenter, pilot study initiated insulin pump therapy in 56 insulin pump-naive type 2s. Insulin doses were adjusted to achieve the best control with the simplest insulin regimen. By the end of the trial, 88% of patients were treated with two or fewer daily basal rates, and 70% were on just one daily basal rate at 16 weeks. Mean A1c dropped by 1.2% from a baseline of 8.4%. In patients with a baseline A1c >8.5%, the decline was an impressive 2.1%. Dr. Bode noted that this was also “highly cost effective,” although he conceded that weight did increase by an average of 1.9 kg (4.2 pounds).
  • Valeritas’ V-Go was launched in the last month, while the Calibra Finesse is “a good year away.” Dr. Bode highlighted the V-Go’s simplicity (non-mechanical, spring loaded) and high patient satisfaction. Beyond showing a brief picture of the Finesse patch/pen he only mentioned the 1+ year timeline before moving to his conclusion.
  • Dr. Bode presented two interesting case studies of problematic type 2 pumpers. Case#1 was a 57-year old female with non-alcoholic fatty liver disease and an A1c of 9.2%. After going on a pump, her A1c was still suboptimal at 7.9%. However, she was only averaging 0.9 boluses per day. After putting her on real-time CGM, her A1c declined to 6.3%, she’s had no severe hypoglycemia, and she “loves her CGM.” The key for her was being able to see her glucose excursions on the CGM screen. Case #2 was a 13-year old girl on a pump with continued to gain weight. She had an A1c of 10.4% at diagnosis, a weight of 165 lbs, a height of 63 inches, and a total daily dose of 65 units per day. Pump therapy was initiated and brought her A1c down to 6.9%; however, her weight had increased to 195 lbs. After Dr. Bode put her on liraglutide (titrated to 1.8 mg/day), her A1c declined to 5.3%, her weight declined to 159 lbs, and she required no basal rate on her pump.



Stanley Nasraway, MD, FACP, FCCP, FCCM (Tufts Medical Center, Boston, MA)

Dr. Nasraway discussed how insulin-dosing software can encourage frequent blood glucose measurement, improve protocol adherence, and help reduce hypoglycemia and glycemic variability. He noted that although insulin-induced hypoglycemia may not “cause” mortality, it could potentially be associated with long-term, minor cognitive effects – as well as acute risk of coma. Dr. Nasraway argued that insulin-dosing software is a valuable way to keep hypoglycemia low and encourage protocol adherence, and he gave a brief overview of the competitive landscape. He also discussed unpublished data from his own surgical ICU at Tufts Medical Center, where adoption of the GlucoStabilizer software has been associated with lower mean blood glucose, less hypoglycemia, and reduced glycemic variability. Looking ahead, he discussed the SHINE study of tight glycemic control in acute stroke patients, which is now enrolling patients toward a target of 1,400. The multicenter trial will compare a tight IV insulin regimen (targeting 80-130 mg/dl) to a more moderate protocol based on subcutaneous insulin injections (targeting below 180 mg/dl). Dr. Nasraway closed by urging that hospitals continue intensive insulin therapy on non-ICU hospital floors, which actually serve more patients for longer durations. This call was echoed by Dr. Irl Hirsch during the panel discussion and by Dr. Anthony Furnary during his presentation later in the day.

  • Dr. Nasraway briefly reviewed clinical evidence on the deleterious effects of hypoglycemia in the ICU. A 1997 brain imaging study of four vegetative patients illustrates the physical effects of prolonged low blood sugar (Fujioka et al., Stroke), and animal studies have supported the notion that cerebral damage depends on both intensity and duration of hypoglycemia. What is more, even moderate hypoglycemia (<70 mg/dl) has been associated with heightened ICU mortality (Egi et al., Mayo Clin Proc 2010).
  • Paper-based protocols have been associated with non-adherence leading to poor glucose control in several clinical studies. In a prospective study of 30 ICU patients in Belgium, nurses were given five-weeks or pre-study training and education on Dr. Greet Van den Berghe’s paper protocol (blood glucose target of 81-110 mg/dl). Despite the training, protocol adherence was only 71%, only 42% of the total 6,016 blood glucose values were within target, and 20% of patients experienced severe hypoglycemia of 40 mg/dl or less (Oeyen et al., Amer J Crit Care 2007). A retrospective study at Brigham & Women’s Hospital revealed protocol compliance of 47%, with “appropriate adjustment of infusion rates” occurring 68% of the time (Cyrus et al., Annals of Pharmacotherapy 2009).
  • Dr. Nasraway proposed that protocol non-adherence might have been one reason why tight glucose control was not successful in NICE-SUGAR, where the paper protocol was six pages long and involved 56 “action codes.” Such complexity often provokes nurses to devise “workarounds” that they can use to shortcut the protocol when they are forced to act quickly – not necessarily a bad thing as long as the patient is well cared for, Dr. Nasraway said, but certainly undesirable for a protocol that is supposed to be standardized. Indeed, according to an analysis of the first 100 hypoglycemic events in NICE-SUGAR, the biggest reason for low blood sugar was protocol violation.
  • Software-based algorithms may allow hospital staffs to maintain tight glucose control without high rates of hypoglycemia. Additionally, Dr. Nasraway noted that software allows clinicians to readily analyze results in real-time and retrospectively, with graphs that are easier to read than handwritten charts. He briefly reviewed several of the products currently used for ICU insulin therapy.
    • Glucommander, the original computer-based software for dosing insulin. Dr. Paul Davidson, inventor of the Glucommander, is the Medical Director of Glytec (formerly GlucoTec), which markets the G+ System for glucose management.
    • Hospira’s EndoTool, which runs a proprietary set of algorithms that Dr. Patrick Burgess designed to customize insulin regimens based on patient demographic and health data. Dr. Nasraway mentioned that Hospira had planned to combine the EndoTool with the company’s Sybmiq smart pumps, but these plans were hampered by an FDA recall of Symbiq pumps in March 2010. He noted that this obstacle is now removed – Hospira announced FDA clearance for a new version of Symbiq in March 2012.
    • MAS Informatics’ GlucoStabilizer, a network-based software adapted from the Glucommander algorithm. Dr. Nasraway and his colleagues use the GlucoStabilizer at Tufts Medical Center, but he noted that he has no vested interest in any software programs.
    • Pronia Medical Systems’ GlucoCare, which is designed to be compatible with multiple established paper protocols.
    • eProtocol Insulin, an NIH-funded system that is not commercially available.
    • GRIP (Glucose Regulation for Intensive care Patients), a free, unregulated,open-source software from the Netherlands. Dr. Nasraway noted that based on GRIP, the code needed for an insulin-dosing software appears relatively simple. He speculated that if FDA approval were not required for commercial viability, the market would be inundated with alternative glucose control systems. Costs will come down soon even despite the regulatory hurdles, he forecasted. He anticipates that companies will eventually have to start including software as an add-on to higher-margin products rather than trying to make money on software by itself.
  • Dr. Nasraway described his own experience implementing glucose control protocols at Tufts. In the late 1990s, he and his colleagues began using a paper-based protocol that used an “arbitrarily” defined target of below 180 mg/dl for intensive insulin administration. In 2001, after publication of the original Leuven studies, they lowered the target to 95-135 mg/dl. In 2008, in hopes of reducing severe hypoglycemia, the clinicians switched from paper protocol to the GlucoStabilizer.
  • During 2009-10, Tufts surgical ICU patients using the GlucoStabilizer had lower mean blood glucose (119.3 mg/dl) and lower rates of hypoglycemia below 40 mg/dl (0.12% of measurements) compared to patients from other wards who were “boarding” in the SICU but not using GlucoStabilizer. (Medical ICU patients on a paper protocol had mean blood glucose of 136 mg/dl and a severe hypoglycemia rate of 0.28%, while neurological ICU patients on sliding scale insulin had mean blood glucose of 145.9 mg/dl and a severe hypoglycemia rate of 0.44%. These data were based on over 9,000 measurements for each subgroup). Mean time to target range (95-135 mg/dl) was 6.2 hours with the GlucoStabilizer. Compared to other wards in 2010, surgical ICU patients had lower mean blood glucose, lower rates of hypoglycemia below 70 mg/dl, and lower rates of severe hypoglycemia below 40 mg/dl. He said that nurses hate the software because it is “fussier” and harder to work around than a paper protocol, but they tend to agree that software is worth using when they see data demonstrating its safety. Indeed, the GlucoStabilizer was recently adopted by all of Tufts’ ICUs.
  • To assess the effects of the GlucoStabilizer on glycemic variability, Dr. Nasraway and his colleagues retrospectively compared glucose control in their surgical ICU between 2008 (paper protocol) and 2009 (software). The resulting paper has been submitted for publication (Dr. Nasraway smiled that he and his colleagues are revising it based on a “scathing” response from one of the reviewers). Of the nearly 200 patients studied, the populations using the paper-based and software-based protocols were not statistically significantly different in mean age (59 vs. 60 years), mean APACHE-II score (16 vs. 15), or incidence of sepsis (35% vs. 39%), and all blood glucose measurements were taken at bedside with a Roche Accu-Chek meter.
  • Patients who were treated according to the software-based protocol achieved better glycemic control, despite being admitted with a higher mean blood glucose (156 vs. 181 mg/dl, p<0.01). Compared to the paper protocol, the GlucoStabilizer was associated with statistically significantly better mean blood glucose (135 vs. 117 mg/dl; p<0.01), time in target (52% vs. 68%; p<0.01), final blood glucose (145 vs. 99 mg/dl; p<0.01), percentage of measurements below 70 mg/dl (1.44% vs. 0.51%; p<0.05), and glycemic variability as measured by standard deviation (42 vs. 29 mg/dl). Dr. Nasraway showed graphs to illustrate that glycemic variability was improved for patients who were and who were not severely ill (APACHE-II scores above or below 20). He noted that the study was too small (n=189) to detect potential improvements in mortality, length of stay, or other clinical outcomes.
  • Turning to the present, Dr. Nasraway highlighted the recently begun Stroke Hyperglycemic Insulin Network Effort (SHINE) study of tight glycemic control in acute stroke patients. Headed by Karen C. Johnston, MD (University of Virginia, Charlottesville, VA), the large (n=1,400), multi-center (56) study is comparing an intensive IV insulin protocol (GlucoStabilizer with target 80-130 mg/dl) to a more moderate regimen using subcutaneous insulin (paper protocol with target below 180 mg/dl). The main endpoint is 90-day neurologic outcome, in which the researchers expect the tight protocol to confer a 7% absolute improvement in the Rankin Scale (a measure of disability/dependence following stroke). Dr. Nasraway, who is not participating in the trial, hopes that the use of a software-based protocol can help encourage adherence and maintain clear separation between the study’s two arms, avoiding problems from past multicenter trials.
  • A new software from Dr. J. Geoffrey Chase’s group in New Zealand may enable less- frequent glucose testing (every three hours, instead every one-to-two hours as with much current software); however, Dr. Nasraway is concerned about real-worldusability since the system requires nurses to input all of a patient’s carbohydrate consumption – quite difficult in the hectic ICU environment (Evans et al., JDST 2012). The researchers’ approach involves modeling insulin sensitivity with stochastic differential equations, rather than treating sensitivity as a constant. Estimated insulin sensitivity and carbohydrate are then used to predict glucose trends. As of the recent paper the system had been tested in a “virtual cohort” computer model and three ICU patients.
  • Dr. Nasraway argued that although clinics worldwide have made good progress in controlling glucose in ICUs, a much larger (and largely unmet) challenge is to maintain such good control once patients have moved from the ICU to other hospital wards. (He drew a pointed analogy to another critical intervention: “Why, if you survive the SICU, would I stop antibiotics once you get onto the hospital floor?”) He presented several years of data on blood glucose from every floor within his Tufts facility; the means ranged from roughly 160 mg/dl to nearly 220 mg/dl, and no floor showed any clear sign of progress from year to year. Given that 90% of hospital beds are on the floors rather than in the ICU, Dr. Nasraway said that ICU glucose control is just the tip of the iceberg and that technological advancements are “more desperately needed” elsewhere in the clinic.


Jane Jeffrie Seley, DNP, MPH, BC-ADM, CDE (New York Presbyterian/Weill Cornell, New York, NY), Bruce Bode, MD, FACE (Atlanta Diabetes Associates, Atlanta, GA), Stanley A. Nasraway, MD, FCCM (Tufts Medical Center, Boston, MA), and Robert A. Vigersky, MD (Walter Reed National Military Medical Center, Bethesda, MD)

Dr. Vigersky: Dr. Nasraway, you’ve achieved some very nice numbers. What can you say about mortality, length of stay, etc.?

Dr. Nasraway: That is the perfect question, which neither reviewer actually asked. We have a 10-bed SICU, and the outcomes are actually pretty good, so we need a large sample size to show a difference. Outcomes actually pretty good, so we need a large sample size. We widened the sample size, for each pool, and then we pulled out the sickest patients, since Dr. Van den Berghe says that the sick patients who are on insulin the longest show the biggest benefit from intensive control. We then looked at length of stay and mortality, and we saw only a statistically non-significant trend. We don’t have the statistical power with the current sample to be able to show significance. But it’s the right question to ask.

Dr. Irl Hirsch (University of Washington, Seattle, WA): Your system falls apart when transferred out of the ICU. I think that at this point, much of the good you did may be undone once they come out. I interpret the NICE-SUGAR data based on Brownlee’s finding that one hyperglycemic spike can impact inflammatory cytokines for a week. Once patients get to the floors, all the good you’ve done may get washed away for weeks.

Dr. Nasraway: I agree with most of what you said. But it’s not “my” problem. It’s your problem; it’s all of our problem. I think our group is lucky to have some data, at least. What do the rest of you have? [Laughter]

Dr. Hirsch: Don’t get me started on my data.

Dr. Nasraway: I have access to all the BG scores across the hospital. Now every quarter I gather these data and send it to the chief nursing officer and other key decision makers. What you are really doing with this sort of data is embarrassing everybody – that is what I try to do, in my mild way. When people see it quarter after quarter, they have to do something. We have now rolled out the Glucommander to all five ICUs.

Dr. Hirsch: I think the bottom line is NICE-SUGAR has set us back a decade in this area. A question for [Dr.] Bruce [Bode] – I was in that Raskin pump study. The problem is when patients are at Medicare age or graduate to Medicare age, they cannot get pumps. It’s hard to tell someone in their early 60s that you’re going to have to give it up. The real question is how will insulin therapy change when we start having access to degludec?

Dr. Bode: Degludec is an extremely flat, stable insulin for 48 hours. We are designing a trial to look at it against pumps starting after the launch. And we will want your input on that study, Irl. In the hospital, the key in these software guidance is you have got to take it away from the doctors and mandate it to the nurses. We are in the midst of launching subcutaneous Glucommander. The mean glucose on our floors is 120 mg/dl. And we do have outcome data on this. Clearly subcutaneous control means a heck of a lot. In our cardiovascular center, anybody using computerized software is in the top percentage of SCIP scores.

Dr. Nasraway: I think that the iceberg under the water is that we need more studies of floor patients as it relates to outcomes.

Dr. Bode: There are a couple of us trying to launch a study to give away software to centers around the world and let them study this in a prospective manner. We need this data. Hospitals are going to have to comply – there are all these ‘never events’ happening. You may not get paid. The hospitals don’t care what the ADA/AACE algorithm tells you – they care about their costs. For any infection that comes back into the hospital, they won’t get paid. All of this will be money-focused and driven by software. In our institution, we have not had a sternal infection for seven years. We are probably the top cardiac center in the Southeast and one of the top in the world. It’s all glucose control.

Dr. Seley: I think you touched on an important point: lots of drip protocols don’t have a transition protocol built in. I think that we need to make doing this standard.

I loved that you talked about nurse workarounds. I use this term a lot; if nurses can figure these out they will. If the protocol is on paper, I think nurse are much less likely to follow it, whereas people for some reason tend to think that computers are always right. They are more likely to follow its instructions blindly. Dr. Van den Berghe taught nurses how to adjust recommendation based on what they thought was right. Maybe one of the problems with most protocols today is that nurses aren’t empowered to make such changes in protocols. If they were, perhaps they wouldn’t be trying to find workarounds.

Q: I think we have a challenge in that we had these wonderful original results from Dr. Van den Berghe, and they kind of disappeared into thin air. We have seen some good studies since then. Everyone is doing something to some extent now, so maybe the background is more difficult. I completely share your opinion that it’s time to put effort into software protocols. The problem is that showing you can control blood glucose and then get hard outcomes takes a large number of patients. Maybe that is cause for some sort of multicenter study.

Dr. Nasraway: As someone active in medicine for 25 years, I’ve noticed like many of you that control – of glucose, of insulin, of other aspects of care – changes. In many cases, by the time you have started your trial, you are aiming for a target that has already been shifted due to new evidence. That happens over and over. Despite NICE-SUGAR, the world has voted with its feet – everyone is doing intensive insulin, just to varying degrees. I think that software has to come down in cost, simply because it’s not that hard to invent as I understand it. However, the way to pitch buying software now is not necessarily that it will change mortality, because that hasn’t been proven. But we do already have evidence showing that it is safer. It definitely attenuates hypoglycemia.

Q: Patients with type 2 diabetes on pumps need more than 25-unit boluses. In the new patch pumps, is there a plan to have larger boluses?

Dr. Bode: That’s a big issue in pump design. The max is 300 units. Tandem has made it known that they will go to 500 units. Even with that, they will be the smallest pump in the world. Most type 2s need a 300- unit cartridge, and potentially use of U200 or U500. That won’t happen. You’ll see U200 pens coming out this fall and U300 for Lantus next year. FDA will only allow concentrated insulin to come out in a dedicated device. You need to have the pump companies talking to each other to get concentrated insulin in pumps, and right now they’re not doing that.

Q: Will Lilly try to get U500 in pumps?

Dr. Bode: I doubt they will. They might do a concentrated Humalog in a dedicated device.

Dr. Karen Horowitz (Case Western Reserve University, Cleveland, OH): I have a comment about ICU patients. Coming from an academic medical center in Cleveland, my iceberg looks like your iceberg. Like you we do fairly well in the ICU, but control breaks down on the floors. Misinterpretation of ACCORD has had detrimental effects on the way people in internal medicine think about diabetes. Providers in the hospital say: “diabetes – not my job.” That gets passed down, and patients suffer. Maybe the goal should not be 7.o% anymore, but certainly we need to target below 8.0%. The papers we write are confusing to general internists because they don’t really know who these patients are.

Dr. Bode, I see a few patients with type 2 diabetes from the ACCORD trial that have pumps. Now these patients are over 70 years old, dementia, multi-system disease. How do we help them manage what they want as intensive management now that we know they don’t really need it? Maybe basal patch pump all we need.

Dr. Bode: I actually think that in people with decreased mentation, pumps work well. You can set the basal rate, and a spouse or family member can teach them how to bolus or just bolus for them. If the patient is going into a home, however, it becomes a huge program. Then the strategy becomes to use a long-acting basal insulin and just bolus as best you can. Today, patients with type 1 diabetes are living a long time. Many have been using pumps for years and then get to Medicare age and lose reimbursement. You will see more and more of these patients in nursing homes. Many of these patients do not do well on multiple daily injections: they go high, they go low, and they often die quickly.

Dr. Horowitz: Lots of providers in these homes don’t understand how to use pumps.

Dr. Bode: Yes, so the devices have to be simple – just as does software in the hospital. Someone must be able to learn how to use them in 15 minutes.

Q: The major problem with intensified MDI is hypoglycemia in type 2s. The major advantage of pumps for type 1 diabetes is the reduction in risk of aggravation of hypoglycemia. This happens while improving A1c. Do you think that pumps in type 2 diabetes may reduce the risk of aggravation of hypoglycemia?

Dr. Bode: Data in studies has not shown that. In the European OpT2mise study, hopefully we’ll see that. In that case, you have two groups, both being pushed to target. We don’t have data right now in type 2 diabetes. The data is overwhelming in type 1 diabetes. But I suspect type 2 diabetes would be the same.

Q: Regarding the second case you presented, that’s a very dramatic effect of liraglutide. In Gastroenterology recently, an analysis of FDA adverse effects was presented of all the GLP-1 mimetics and DPP-4s. There was a statistically significant increase in pancreatitis and pancreatic cancer rate was higher. How worried are you?

Dr. Bode: I’m not that worried. We have a huge amount of experience with Byetta. I reviewed all of those pancreatic episodes, and most were not related to exenatide at all. Victoza also has a very large database. Pancreatitis also happens on metformin, glyburide, and other medications. Is it greater? We cannot tell that right now. But I’m not worried. I hope it’s not the tip of the iceberg – if it is, we’re in trouble.


Alfred Mann, MS (MannKind Corporation, Valencia, CA)

Al Mann “couldn’t resist” the invite to CDTM after hearing the title of his presentation, stating emphatically, “MannKind’s inhalable insulin program is very much alive and well.” He discussed many of the advantages of Afrezza and gave a few interesting updates on his thinking. Most importantly, both the type 1 and type 2 trials are meeting enrollment targets to complete by year-end 2012, putting filing on track for the middle of 2013 and potential approval by year-end 2013. Mr. Mann also made an interesting case that Afrezza may require little need for a blood glucose meter or complex meal titration in type 2s. A meal study (n=6) from Dr. Zisser’s group at Sansum (being prepared for publication) showed that for meals with varying amounts of carbs in type 2 diabetes (0, 25, 50, and 100 grams), a single dose of Afrezza was sufficient to bring postprandial glucose into the ADA target range regardless of carb content. In Mr. Mann’s view, this suggests just a general idea of carbohydrate content and occasional fasting blood glucose measurements would be sufficient with Afrezza – we’ll of course be very interested to see larger studies to confirm these intriguing results. We suspect that PCPs would be especially intrigued by the potentially reduced need for dose titration. Mr. Mann briefly touched on pricing as well – the company hopes to provide Afrezza at no more than a 5-10% premium over pen delivery of rapid-acting analogs. When pressed on this during Q&A given the introduction of generic analog insulin in the US, he conceded that MannKind’s future partner will have a lot to say on the pricing front. We look forward to more updates in MannKind’s upcoming 1Q12 earnings call (date to be announced).

  • A meal challenge study demonstrated Afrezza’s encouraging ability to cover a wide variety of meals with a single dose. The study took six type 2s through four breakfast and four lunch meal challenges. Each meal challenge had a different number of carbs – 0 grams, 25 grams, 50 grams, or 100 grams – but the same dose of Afrezza. Blood glucose was measured every 50 minutes (50 minutes before to 250 minutes after the meal). For the breakfast challenge, the 100-gram meal led to the highest peak blood glucose change from baseline (+30 mg/dl), while the 0-gram meal resulted in the largest blood glucose decrease from baseline (-30 mg/dl). By 250 minutes, blood glucose for all four meals was within 20 mg/dl of baseline. A similar pattern was observed for lunch – the 100-gram led to the highest peak change from baseline (+20 mg/dl), while the 0-gram meal led to the largest blood glucose decrease from baseline (-30 mg/dl). By 250 minutes after the meal, all blood glucose values were within +10 mg/dl or -20 mg/dl of baseline. We see this as very encouraging data if it indeed holds up in clinical practice – both patients and HCPs will no doubt appreciate the reduced focus on carb counting and mealtime insulin titration.
  • Mr. Mann explained why Afrezza has not shown an A1c benefit in clinical trials. Using a schematic, Mr. Mann showed the pharmacodynamics of Afrezza and rapid-acting analogs. While Afrezza quickly brings postprandial glucose back to baseline, rapid-acting analogs have an extensive late postprandial persistence. This results in post-meal levels that are lower than baseline, which skews the mean blood glucose (and therefore A1c) down. In other words, Afrezza permits greater time in range, but the higher rate of hypoglycemia (and accompanying lower A1c) with rapid-acting analogs makes it hard to show superiority. Mr. Mann believes that by lowering fasting blood sugar to near normal, Afrezza will show superiority for A1c as well. He noted that  the current trials are going to enable the company to do that.
  • Mr. Mann believes that the future of type 2 diabetes lies with a simple, low cost basal patch pump combined with Afrezza. He stated that glargine is “not great,” and although degludec may be better, a basal patch pump would be optimal. Mr. Mann called the biggest challenge in pump therapy providing it at a cost comparable to insulin.
  • The major advantages of Afrezza include: (1) greatly improved seven-point glucose profile;(2) less risk of hypoglycemia; (3) lower fasting glucose; (4) comparable A1c control; (5) less weight gain; (6) simplicity/ease of use.

Questions and Answers

Dr. Robert Vigersky (Walter Reed National Military Medical Center, Bethesda, MD): Can you give more insight on how the same dose is able to achieve the same glycemic goal with a wide range of carbs?

Mr. Mann: In the meal ranging study in type 2, there was no change. Howard Zisser is preparing a paper - in type 2, there were flat results no matter what you did. For type 1, it wasn’t as forgiving. But if you’re talking about A1c, you don’t need to be at a precise rise for each meal. It’s really the average. There’s a lot more latitude than we thought. We don’t see any really significant need for close titration with this therapy.

Q: The whole audience is very grateful and thankful for all you’ve done on pumps and inhaled insulin. These would not exist without you. The FDA is giving you a hard time and two studies are so difficult to handle. I would encourage you to renegotiate some aspects of the studies. You are including patients with an A1c higher than 7.5%. These are not the best patients. Most of the other insulins didn’t have a lower limit – they had a high limit. Also, in the type 2 study, they wanted to include metformin and metformin plus sulfonylurea patients. They are quite different.

Mr. Mann: It’s a fascinating story about how the regulatory process has evolved. In the type 2 diabetes study, we will hopefully be able to differentiate Afrezza. About half the patients are on metformin only, and probably none on sulfonylureas, some are some are on DPP-4 inhibitors, and some on TZDs. There is enough sample in each category to differentiate there. We believe we’ll be able to show excellent results compared to any of those therapies. In the type 1 study, it is what it is. We needed to demonstrate that the new inhaler produces the same results as the old inhaler. The bioequivalence study did that. After a few hours you cannot find a trace of the insulin or the carrier. This whole thing doesn’t make a lot of sense, but that’s the regulatory world we live in today.

Q: Your study is fascinating. It’s the first demonstration of replacing first phase insulin secretion. What worries me is the pizza, steak dinner, and French fries. Those require three to four hours of insulin activity.

Mr. Mann: You’re absolutely right. Some foods take longer than roughly three hours. If these people are eating a large amount of high carb meals, they should take a dose before the meal and a dose after. There is still no need for complex titration. We typically see highs well within the ADA limits, even with a carb load up to 100 grams in type 2 diabetes. You just need a rough estimate – no need for complex titration.

Q: Will Afrezza be as effective in type 2s with COPD? Or contraindicated?

Mr. Mann: The FDA is concerned, as are we. We will contraindicate Afrezza in those with COPD or asthma. We are doing trial in these populations and you need a somewhat higher dose. In asthma patients, Afrezza seems to work quite effectively. For our own protection, we will probably label it against those people.

Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA): I’ve been an investigator on several of your trials. Our patients like your products and I hope you get approved soon. What do you say when the FDA says that A1c is not better with Afrezza? Do you talk about faster absorption, the lack of postprandial hypoglycemia? And since bioavailability is much lower than injectable insulin, how will this be a similar cost?

Mr. Mann: On the second question, our objective is to offer Afrezza for no more than a 5-10% premium over pen delivery of rapid-acting analogs. Could we do it for less? Yes we could. But we’ll launch from there and see where it goes. The processing cost is low. On the problem with FDA, I don’t really know. I have not been discussing these issues with them. We show them cartoon of late postprandial glucose that I showed. That really explains it. When Lilly got lispro approved, the FDA said there was no clinical benefit. They focused only on A1c without looking at other issues. The problem is that it’s hard to get clinicians to lower fasting glucose levels. Our objective in these trials is to get fasting under 110 mg/dl or under 120 mg/dl. If investigators do this, I would expect a fairly substantial number of patients will get to an A1c under 6%.

Dr. Barry Ginsberg (Diabetes Technology Consultants, Wyckoff, NJ): I’m wondering about the economics of Afrezza. By the time you come out, Biocon and Wockhardt will have generic lispro, aspart, and glargine in the US.

Mr. Mann: I won’t dictate the price. Because of the delay and the additional $500 million this is costing us, we’re going to have a partner on this. The partner will have a lot to say on the pricing.

Session IV: Diabetes Communication – Niche or Mainstream?


Majid Sarrafzadeh, PhD and Tannaz Moin, MD, MBA (UCLA, Los Angeles, CA)

Drs. Majid Sarrafzadeh and Tannaz Moin gave an update on the wireless health initiatives going on at UCLA. The university has a remote monitoring system called WANDA, which takes data from in-home sensors (e.g., blood glucose meter, weight scale, blood pressure monitor) and sends it to the cloud. The system also features advanced data analytics and patient/HCP feedback (e.g., “Patient 9936’s weight has fluctuated by more than three pounds in the past two days”). Social networking, pattern analysis, and patient questionnaires are also important parts of the system. Most interesting was the team’s ongoing Diabetes Telehealth Study. Patients in the intervention group are sent home with a wireless blood glucose meter (the brand/model was not specified) and receive feedback from their providers based on alerts (this has averaged about two 5-15 minute calls per week per patient). Preliminary results after three months have shown a benefit in the 51 patients currently enrolled: a reduction in A1c of 1.5% in the intervention group (baseline 8.8%) compared to a reduction of 0.96% in the control group (baseline 8.6%). These are intriguing results and we look forward to hearing more about the study as it completes. We’ll also be interested to see data on provider time – two calls per week per patient seems like a lot to handle, especially as the study scales up.

  • The UCLA Diabetes Telehealth study will evaluate the benefits of a wireless blood glucose meter, remote monitoring, and ongoing HCP feedback. The study will last until summer 2013 and has enrolled 51 patients thus far (target enrollment of 100 patients). The primary outcome is change in A1c at three months, and blood glucose, general health status, and medication compliance will be measured. Patients with an A1c >7.5% are randomized to one of two arms after being discharged from UCLA for an acute inpatient hospital stay. The control group receives monthly phone calls and diabetes education prior to hospital discharge. The intervention group receives the same diabetes education, but also gets a wireless blood glucose meter and ongoing monitoring/feedback based on their readings. Once home, patients in the intervention group receive phone calls based on certain alerts. Examples given include hypoglycemia <80 mg/dl, hyperglycemia >200 mg/dl, worsening health status, and medication noncompliance. On average, participants in the intervention group have received about two calls per week, lasting about 5-15 minutes on average. These calls may involve titration of medications.
  • Despite learning a number of valuable lessons, the team also has many unanswered questions. The program has taught them that this is not a one- size fits all approach. However, questions remain about how to select the most ideal patients and how to keep them engaged. Additionally, it’s not clear what the ideal timeframe is for enrollment in the trial and/or continued motivation. As more and more companies (e.g., WellDoc, Telcare) get in to the wireless health game, we’ll be interested to see how these and other issues shake out. Reimbursement of provider time will certainly be a critical part of the process, although we suspect this won’t come about until clinical efficacy is demonstrated in large RCTs. Even still, CGM has a deep and wide base of evidence, and reimbursement is still not where it should be. The space will certainly be fascinating to watch in the coming years…


David Kerr, MD (Bournemouth Diabetes and Endocrine Centre, Bournemouth, UK)

Dr. Kerr gave a broad presentation on all facets of social media; the major thrust was that patient use of it is increasing dramatically, although pharma companies and doctors have been reluctant to take part. We were interested to hear Dr. Kerr’s point about mobile apps for diabetes. One study of 137 apps found that there was a major gap between evidence-based recommendations and applications’ features. Most importantly, apps with personalized education were underrepresented (Chomutare et al., J Med Internet Res 2011). This concurs with what we sometimes hear about the absence of really useful apps for people with diabetes (though this is slowly changing). Dr. Kerr also reviewed recent QuantiaMD research on physicians and social media – the top concerns that held physicians back from interacting online included liability (73%), patient privacy (71%), reimbursement (41%), and lack of time (38%). Turning to social networking for diabetes, Dr. Kerr showed data that nearly one-third of advertisements on Facebook were for non-FDA approved products (Greene at al., J Gen Intern Med 2011) – this strikes us as a very un-policed and potentially dangerous area. Finally, he closed with a series of slides on the thorny issue of pharmaceutical companies and social media (“Pharma are terrified of social media), including examples of inaccurate breast cancer tweets from Novartis andRoche. Overall, Dr. Kerr believes cloud-based data upload holds tremendous promise, but an important part of driving down A1cs is making the data social.


Anne Peters, MD (University of Southern California, Los Angeles, CA)

Dr. Peters discussed the challenges of diabetes care in elite athletes (challenges that she said affect all patients who exercise, to varying degrees) and the ways that technology can help. She noted that modern devices have allowed people with diabetes to reach athletic levels previously difficult to achieve. (As a half-joking extreme example, she juxtaposed the famous image of the original ‘backpack’ insulin pump with a picture of a cyclist wearing an OmniPod.) She emphasized that she could give no “magic bullet” universal recommendations, since the elite athletes she’s treated (e.g., Olympic swimming gold medalist Gary Hall, Jr.) require more individualization than any other patients. However, she did emphasize that the first priority must always be safe diabetes control. For example, she has worked for three years with Indy racer Charlie Kimball foremost to keep his blood sugars stable throughout a race (he currently averages ~180 mg/dl in the car), and they are now working to lower the mean only gradually (given the extremely high risks of hypoglycemia). Athletes also tend to benefit from a scientific, data-driven mindset. On this note, she praised the now-analytical approach of skier Kris Freeman, whom she hasn’t met but whose blog chronicles his diabetes management (and whose Olympic performance in 2010 was jeopardized by not paying attention to glucose control). Dr. Peters also recommends contact with other patient-athletes (because they can compare each other’s personal experiences in detail; Dr. Peters mentioned Team Type 1 and Insulindependence as great avenues for people to learn from and work with each other).


Neal Kaufman, MD, PhD (DPS Health, Los Angeles, CA), Majid Sarrafzadeh, PhD (UCLA, Los Angeles, California), Tannaz Moin, MD, MBA (UCLA, Los Angeles, CA), Anne Peters, MD, CDE (University of Southern California, Los Angeles, CA), and David Kerr, MD (Bournemouth Diabetes and Endocrine Centre, Bournemouth, UK)

Dr. Moin: Dr. Kerr, what about confidentiality issues? How does someone own the information and then maybe provide advice to peers without it being identified?

Dr. Kerr: The key is that the information going to the cloud is anonymized. We would never get away with putting, “You need to take drug X” – it is the way you word it. It is up to us to create a content of words and phrases used.

Dr. Sarrafzadeh: Who owns the data? Who should?

Dr. Kerr: It should be the patient. At the moment, though, it’s often third parties. Rather than use established clouds, you can create a cloud that is owned by the organization that covers the institution or environment.

Dr. Kerr: In pump treated patients, what is the optimal time prior to exercise to make changes in the basal rate?

Dr. Peters: You could bet money on my answer – it needs to be individualized. Not every patient suspends their insulin; some need a little insulin. Some change just before exercise. Some people need to start changing two hours before.

Dr. Saleh Adi (UCSF, San Francisco, CA): Majid, who owns your data?

Dr. Sarrafzadeh: There must be a business incentive. Google owns a lot of our data, but that’s how we can search. Everybody should own some aspect of the data. If you don’t allow Facebook and Google to own some of the data, they won’t have any incentive to provide services.

Dr. Adi: Should medical centers own the data?

Dr. Sarrafzadeh: Part of it.

Dr. Adi: How did patients get the message at home?

Dr. Sarrafzadeh: We could communicate through the glucose meter and ask questions on it.

Dr. Moin: It’s the glucometer that prompts the patient to respond.

Dr. Adi: So there is no message from your server?

Dr. Sarrafzadeh: There is fax, email....

Dr. Adi: Can the method of alert transmission be chosen by the patient?

Dr. Sarrafzadeh: Currently not. For larger studies, I think it makes sense to have the patient choose.

Dr. Adi: Is there any thought to integrate data with EMRs or paper medical records?

Dr. Sarrafzadeh: We don’t do it. But there are companies that do it for a living.

Dr. Adi: How is it done at UCLA?

Dr. Moin: At UCLA, we’re not integrating it. They just purchased Epic. At the VA, providers can get access to patient data from the wireless blood glucose meter.

Dr. Kaufman: The information that these programs collect are microbehavioral. None of them are important in isolation, but the sum of them determines the outcome. Clinicians don’t want every detail. The more information there is, the more critical the data reduction question becomes. Information must be actionable and ready to be used. You don’t need to know every single blood glucose every minute.

Dr. Adi: So you are able to actually open the two systems so that they are communicating? Your software is communicating with the VA medical record software?

Dr. Moin: There are five vendors in the greater Los Angeles VA. The data is integrated in CPRS. 

Dr. Robert Vigersky: For Dr. Peters – this relates actually to the patient CGM panel yesterday. I was struck by something they said. They were waiting for you to tell them what to do with their retrospective data, yet you showed examples of athletes who are clearly able to integrate this data. We talk about the overwhelming amount of data… how do we get them to understand the data so they don’t have to come to us to make changes?

Dr. Peters: In reality every one of them makes changes every day based on data, as does anyone with type 1 diabetes. They do receive instructions on downloading, and in other things they don’t want to do. Of all the patients that spoke yesterday, Toby is the most integrated, best controlled, yet she insists on coming in every month. Each time she brings reams of data and proposes changes that I inevitably agree with; I tell her to just tweak her basal rates, for instance, but she wants to run it by me. I think type 1 diabetes management is about the relationship of the patient and the care team. I think this disease is so hard to have, you can’t remove the love component. These are long-term relationships. I’m lucky to have them. I have followed Toby for over 25 years, since I was a fellow. You can’t overlook the personal touch; technology will never replace this. Though they are encouraged, trust me, to make decisions on their own.

Dr. Kerr: Yes, a machine won’t replace me. But I could spend my time with greater value if technology could remove some time burden. A problem is that most downloads are uninterpretable to most patients. We need to do work on improving imagery.

Dr. Kaufman: Our approach is to try to use technology to go to scale with public health interventions. The basic premise for me is that almost all the things we’ve talked about are tools. They allow patients to get more support, more knowledge. All the use needs to be in the context of both a therapeutic relationship and an intervention. Technology can help with motivation and understanding for many, but most people at some point need some extra help as well. To me, the future of clinical care is integrating this all together.

Q: You mentioned many athletes competing as individuals. Do you have any experience with those on teams? Especially for disconnecting from a pump and CGM, how did you get athletes to modify apparel? Or could you get them to wear CGM, but not look at it, so there is retrospective data?

Dr. Peters: Almost all the athletes I’ve had are individual. I haven’t done as well with team participants at getting the data when they’re competing. Generally I seem to use less technology and have people do more fingersticks and logging – I’ve been more conventional with some of that. I’ve done better with frequent glucose monitoring than with sensing when people are away from the receiver. But most of my experience has been individual athletes. I have one that’s a pole-vaulter – now that’s a very easy sport to play with type 1 diabetes. It just depends on the sport.

Q: I had one mother who would run down the sidelines with a Dexcom to get the information… is there a better way?

Dr. Peters: I’ve been trying to do this for Charlie Kimball – they have telemetry in the car, so it’d be great to get his CGM information coming to the pit. Now, he radios his blood sugars in. I want to have it so he doesn’t need to think about diabetes. That’s exactly the problem for Charlie that we could get fixed.

Dr. Kaufman: There are also ways to monitor physical activity and keep track of activities remotely as a person is running around on the field.

Dr. Peters: The person on the field can stop and check their blood glucose. Racecar drivers cannot check.

Q: Dr. Peters, do you work with mountain climbers? I have a type 1 patient that recently climbed Mt. Kilimanjaro. He wears Dexcom but calibrates with only one meter. I did a lit review to see which meter is best at high altitude.

Dr. Peters: I have had the same problem. I can’t be sure which is best, given the small sample size and the rarity that my patients climb mountains. I usually just tell them not to use very much insulin and to stay well hydrated.

Dr. Barry Ginsberg (Diabetes Technology Consultants, Wyckoff, NJ): The least altitude-sensitive would be a GDH-PQQ meter. Roche has a GDH-PQQ meter, and Abbott’s FreeStyle line uses this enzyme as well.

Dr. Peters: I used a FreeStyle meter. I didn’t want to say, as I didn’t think I was expert enough to comment.

Comment: My patient used FreeStyle; that is what I could glean from my research as well.

Dr. Kerr: There was actually a paper on this. Someone took a number of meters up Kilimanjaro; I can’t remember who.

Dr. David Horwitz (Johnson and Johnson Diabetes Institute, San Francisco, CA): On the amount of time you spent on the phone with patients? It sounds reasonable. But is that economically sustainable as number of patients grows?

Dr. Moin: The question of costs is an interesting one. A provider who is in clinic from 8-5 won’t have the time. It takes less time than I thought, but it’s still time intensive. We have 51 patients enrolled, and when we get to 100, that will be several hours a week. And sometimes you call a patient, they’re not there, so you have that back and forth. Someone definitely needs to think about that. These programs will be successful when there are dedicated providers who have time set aside to run them.

Dr. Horwitz: Are you collecting health economics data?

Dr. Moin: Not with this pilot.

Dr. Kaufman: There are a whole range of communication channels – one on one in person, a phone call, text, twitter, email, and in-person group sessions. There should be no one door. Patients should have options. The health system will push patients away from one-to-one encounters because they are most expensive. If you could imagine it, once a protocol is defined, you could automate large portions of it. Individuals can get and give as much as they can, and then get more. We’ve used asynchronous secure messaging. We never use telephone or face-to-face unless the patient has additional requirements. The scaling cost of the systems is the human cost – how do you keep that at a minimum.

Dr. Peters: In the county of LA, we run intensive programs for diabetes patients, and we allot one-third of nurses’ time to be on the phone. So that time is already incorporated. All of my patients have access to my email, cell phone, however they want to contact me. Coming to this meeting, I wanted to calculate how much all this costs me. I figured that my involvement in electronic medicine is about $43,800 per year. It’s two-to-three hours every single night. It’s an unbelievable personal cost, but one I gladly give – it’s just hard to expect others to do.

Dr. Kaufman: And reimbursement will cover that. Things will change as it becomes more obvious that outcomes are better.

Q: Dr. Kerr, were you to develop a free-flow information exchange, how do you think we can measure or assess the quality of discussion and information being provided? How will we ensure relevance?

Dr. Kerr: Automated content analysis is at an embryonic phase, but you can search out key phrases. From the pharma point of view, this will be critical – to monitor social media for adverse event reports, for example. As to relevance, people who don’t use Twitter have this bizarre view that people on Twitter talk trivia all the time, but that is not actually the case.

Dr. Kaufman: I saw a talk about an evaluation of informal online health discussions and whether misinformation was being given – a frequently voiced concern. In almost every case of inappropriate discussion or bad advice, the environment was self-regulated. Obviously this won’t always happen, though, but there are fortunately ways to moderate content. Most healthcare settings in the US and Canada are unwilling to have chat sessions on their brand without moderation, so they will set up a chat at a fixed time and then shut down the forum when the time is up. That is one way to mitigate the risk.

Dr. Sarrafzadeh: eBay, for example, manages online forums quite well. Whether it can be extended to doctors remains to be seen.

Dr. Robert Vigersky (Walter Reed National Military Medical Center, Bethesda, MD): How can outcomes be linked to social media interactions? Can you envision knowing whether social interactions at an individual level make a difference?

Dr. Kerr: You have to choose what outcomes you are interested in. People vs. clinician outcomes differ. One idea is using Facebook to make young people more aware of some of the technologies out there. That way, we can increase penetration of pumps in the UK. Or increase demands so that CGM can be reimbursed – it’s currently not. Patients will still be seen, but probably less frequently face-to-face. The outcomes are still there to be collected, but you have to devise a mechanism. With peer support, education programs, and advertising some of the things out there, quality of care is likely to improve.

Dr. Vigersky: Is there a way to link lab data such as A1c with the number of social interactions a patient may have engaged in?

Dr. Kerr: We have a proposal that we’re hawking around. That’s been one of our unsuccessful grant applications.

Dr. Kaufman: This is one important question – how do you evaluate the impact of technology. There are challenges – if I want to do randomized controlled trial, it takes about six months to a year to plan, a year or year and a half to enroll, and then six months to write and publish it. That’s three years. At that point, we’re two to three generations beyond the technology, so what you’re reading as modern is several generations past. How do you change the paradigm? I don’t have a good answer. But RCTs are not the only method. We can also research and identify the core principles that matter – does social support improve individual well-being and outcomes? That way, whether it is delivered through one mechanism or anotspan>her is less important.

Dr. Kaufman: Are there any parting comments?

Dr. Kerr: Social media is an interesting, rapidly changing area and patients are all using it. It’s a mistake to avoid it and it’s not going to go away. It’s actually quite a lot of fun.

Dr. Peters: In terms of using social media, these are methods for getting patients to connect. We’re trying to do that in east LA with the underserved population. At baseline, they don’t do much of this. They don’t have smart phones, but they do have phones that can text. We need to take technology and adapt it for various circumstances. We have to be clever and always keep in mind this concept of burnout. You like getting a text for a month, and then after that it’s a pain in ass. There’s a lot we can and should do to help patients improve.

Dr. Moin: All of us cannot live without our cell phones and these other wonderful technologies. We need to think outside the box to benefit our patients. These issues with burnout and confidentiality are challenges we need to face.

Dr. Sarrafzadeh: I once read this book on how Silicon Valley was built as an accidental empire. That’s where we are with information and health. We are building this accidental empire. We must tackle problems one at time.


Session V: Is This the Modern Management of Diabetes?


Bruce Buckingham, MD (Stanford University, Stanford, CA)

Dr. Buckingham gave an excellent update on the broad swath of rapidly advancing closed-loop research. We were most fascinated by his comments on FDA, infusion sets, and upcoming trials. First, he is very encouraged by the recent shift of the AP to FDA’s Office of In-Vitro Diagnostics (OIVD). Since February, he has a seen a noticeable acceleration in responsiveness – this is excellent to hear from a patient perspective. Dr. Buckingham also discussed some intriguing infusion set studies. Interestingly, around 40% of infusion sites last out to seven days, although the biggest predictor was the individual. We hope research continues on this front; it seems like an under-investigated area, although it is absolutely critical for the AP and patients’ day-to-day variations in glycemic control. Finally, Dr. Buckingham gave an update on the nocturnal hypoglycemia minimizer his team is studying. FDA recently approved an outpatient study of the system in 45 patients over 42 nights (1,600 nights total!) – the team will complete 200 outpatient nights in 20 subjects in the next 30 days. Regarding the JDRF control-to-range study, the hope is to have all subjects’ data in by the beginning of ADA, though the investigators are “all sleepless over this.” Finally, Medtronic’s treat-to-range system, using a PID algorithm with insulin feedback gain, has been tested in nine subjects, with the tenth to occur next week. We always appreciate Dr. Buckingham’s enthusiasm and passion and can’t wait to hear more at ADA in June.

  • Following the recent change to OIVD, Dr. Buckingham has been encouraged by the FDA’s responsiveness (“It seems like a new day”). On February 13, Dr. Buckingham and other closed-loop investigators received an FDA communication that oversight of the AP would be moved to the Office of In Vitro Diagnostics (OIVD), the same unit that reviews CGMs and meters. The goal of the move was to streamline things, and that’s exactly what Dr. Buckingham has seen thus far. For the team’s most recent studies, FDA has reviewed, processed, and responded to requests within 30 days, on par with Dr. Buckingham IRB experiences. Overall, he believes this move has made the study process “much, much quicker.” This is very encouraging to hear and echoed Dr. Zisser’s thoughts during Q&A.
  • Dr. Buckingham’s team will be testing a predictive hypoglycemia minimizer in 45 patients over 42 nights (totaling 1,600 nights of in-home use). The nocturnal system will suspend insulin delivery based on predicted hypoglycemia. The team recently made a minor revision to the algorithm after the first few patients. Encouragingly, this change was approved by FDA in just ten days. The team will complete 200 outpatient nights in 20 subjects in the next 30 days. These are some of the first in-home studies in the US, and Dr. Buckingham is notably “excited that it’s going on.”
  • Infusion set issues are very common in pumpers and will be an important part of AP algorithms. Dr. Buckingham explained that 66% of patients have occlusions or unexplained hyperglycemia. With this high prevalence, the closed loop is going to need to detect when an infusion site is failing. Additionally, since sensors typically last a week and infusion sets need to be changed out every two to three days, it may be challenging to have a single port artificial pancreas. To demonstrate further areas for improvement, Dr. Buckingham showed pictures of patients with chronic scarring and abscesses due to infusion sets. Many patients repeatedly used the same areas for their infusion sets.
  • A study of Humalog and Novolog revealed that infusion site occlusion rates were similar between the two insulins – about 40% of sites lasted one week. The small trial included 20 subjects (mean A1c 7.6%) wearing Silhouette infusion sets. Participants were blinded to which insulin they were using (In fact, when asked to guess whether Novolog or Humalog had been used, exactly 50% were able to guess correctly). Patients were asked to leave the infusion set in for one week or until the site failed. To look for occlusions, Dr. Buckingham and his team then stained the infusion site catheter tips. Staining was seen in every single catheter. There was no difference between the insulins, and about two out of five sets lasted for one week.
  • A similar study compared the occlusion rates between Sure-T (steel) and Quick-set (Teflon) infusion sets; interestingly, the best predictor of length of wear was the individual. Twenty participants were randomly assigned to wear each infusion site twice.

Patients were instructed to wear the set for one week unless there was set failure (defined as a blood glucose >300 mg/dl and a failure to decrease 50 mg/dl in one hour). A greater percentage  of Quick-sets failed on insertion (about 15%), although the two sets had similar failure rates thereafter. Overall, about 35-40% of sets lasted seven days. Interestingly, the biggest predictor of infusion set survival was the individual: some people could not go more than 2-3 days, while some could go seven days. Encouragingly, those lasting seven days had no deterioration in glycemic control. This led Dr. Buckingham to conclude, “Length of wear is a biological factor.” The investigators also measured erythema and induration (similar between both sets), which did not necessarily result in infusion set failure or a deterioration in glycemic control. Based on this data, Dr. Buckingham and his team have developed an algorithm to detect infusion set failure. We believe infusion sets are a major area for further innovation and are glad to see Dr. Buckingham conducting this critical work, especially as AP trials move to more ambitious settings.



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

After reviewing the history of artificial pancreas development and thanking JDRF for its important advocacy role, Dr. Zisser discussed various approaches to improving insulin speed: intradermal needles (BD), hyaluronidase (Halozyme), intraperitoneal (Roche; Medtronic), heating the infusion site (InsuLine), pramlintide (Amylin), and inhaled insulin (MannKind) – the last of which will be studied as a “priming dose” in a partially closed-loop study this summer, pending FDA approval. He said that time lag is relatively less important for glucose sensors, but that CGM accuracy and reliability remain concerns (one step forward is the Dexcom Gen 4 sensor, which the Sansum team hopes to start using this summer.) Beyond faster insulin and better sensors, Dr. Zisser also expressed his interest in remote monitoring (to alert loved ones of hypoglycemic events) and glucagon as a failsafe measure (perhaps to be stored in a patch-pump-like device on the patient’s body, though we think that few patients would want to wear this). Finally he announced that the day before, researchers from the University of Virginia had conducted the first-in-US outpatient study of partially closed-loop control (n=1, ~42 hours). Excitingly, the Sansum team plans to go one better in May when they conduct an outpatient study with five patients at the same time.

  • The Sansum researchers recently finished incorporating the Dexcom Gen 4 sensor into their closed-loop platform; pending FDA approval they plan to start using their updated system in trials this summer. Dr. Zisser especially looks forward to the new sensor because the receiver can pick up sensor signals from up to 50 feet away – a big improvement from the Dexcom Seven Plus. He noted that Dexcom is developing a modified version of the Gen 4 with algorithms optimized for closed-loop research, though until that is available his team will use the “off-the-shelf” consumer version.
  • Dr. Zisser believes that remote monitoring is critical for research on (and eventual use of) closed-loop devices. He envisions a system whereby risky sensor readings cause an alert to be sent to some designated friend or family member; this alert could include the glucose level, its rate of change, and the GPS location of the patient. Dr. Zisser indicated that such a system will be subject to many regulatory hurdles; he said that the FDA took “a long time” to approve a study in which a text message was automatically sent from the closed-loop device to his own phone.
  • On April 20, the University of Virginia conducted an outpatient study of partially closed-loop insulin control – the first in the US, to Dr. Zisser’s knowledge. Similar to recent studies conducted in France and Italy, the 42-hour experiment used the UVa team’s Android-based handheld device and involved a single patient engaging in daily activities (e.g., walking around a city, eating in restaurants). The Sansum team plans to conduct a similar study with five patients simultaneously in May, an important evolution of the one-patient-at-a-time experiments thus far. Dr. Zisser explained that a main goal of the research is to lay groundwork for pivotal outpatient studies of overnight closed-loop systems. He says that researchers will also continue to conduct inpatient trials, as these will be necessary for the early tests of fully automated artificial pancreas products.


Howard Zisser, MD (Sansum Diabetes Research Institute, Santa Barbara, CA), Bruce Buckingham, MD (Stanford University, Stanford, CA), David Klonoff, MD, FACP (Mills- Peninsula Health Services, San Mateo, CA), and Selah Adi, MD (University of California, San Francisco, San Francisco, CA)

Questions and Answers

Dr. Klonoff: What do you see as the biggest problem with the predictive alarm system you’re working with?

Dr. Buckingham: It’s very hard to make predictions, particularly when they’re about the future – a Niels Bohr sort of thing. One thing is turning the pump off too frequently. We can shorten and prevent lows, but the price is raising average glucose overnight a bit. I don’t think it’s going to be a big problem. But we’re thinking about that.

Dr. Jeffrey Joseph (Thomas Jefferson University, Philadelphia, PA): Talking about insulin pump catheter failures – by day three to day five, failure is common. Is this mainly due to insulin precipitation, or to a biological issue within the wound?

Dr. Buckingham: I don’t think we really know; there is not much data on the topic. I think you see cytokine release and inflammatory response. Clearly, as I showed, there is not a big difference between steel or Teflon cannulas. Clearly there is interindividual variability, suggesting a strong biological component. Whether the issue is irritation due to crystallized insulin, to the phenol used in the preservative, or for some other reason – I don’t know.

Dr. Klonoff: Dr. Zisser, can you discuss the closed-loop program at UCSB?

Dr. Zisser: We’ve got five to six projects going simultaneously. We got started in 2006 with our first grant from JDRF. Back then, we went directly to the IRB and started the trials. We built the Artificial Pancreas System (APS) to communicate with pumps and sensors and got approval right away. We did about five studies with just IRB approval and that’s when JDRF and FDA started talking to each other. FDA told us to stop trials until we got IDE approval. It was lots of paperwork and it has been difficult in the last couple years. When we started doing this, FDA didn’t know what to look at and they didn’t know how to guide us. Since February and the switch to OIVD, things have been moving much faster. One project we’re working on is control to range to eliminate highs and lows. That has been a very difficult study to get going because we’re working internationally at seven or eight sites. We had to wait for FDA approval to get that started. We are also figuring out how to make the systems more efficient. For me, that’s a more stable sensor. The sensor lag not too bad – if you have a good sensor, control sessions are very easy. When a sensor is off, it’s usually biased in one direction or another. When YSI is low and the sensor is high, that tends to be the hardest thing to control.

Dr. Robert Vigersky (Walter Reed National Military Medical Center, Bethesda, MD): Dr. Buckingham, you alluded to a two-year study in newly diagnosed type 1 diabetes patients to preserve beta-cell function. What can you tell us? Do you have C-peptide data?

Dr. Buckingham: Yes, but I’m not allowed to see it. At the end of October we will review C-peptide data and see where we are. The data safety monitoring board said that the control group could have whatever therapy they wanted, including pumps and sensors. Meanwhile, there were a few patients in the treatment group with requirements for psychotropic drugs or suicidal ideation, so this may affect the results. But when you do a randomized study, you have no idea what you will get.

Dr. Vigersky: The patient you showed had a very low insulin dose – 0.2 or 0.3 U/kg. One could infer from this that the patient is making insulin. Is this typical?

Dr. Buckingham: A number of other patients have entered a honeymoon period in this way, but it is extremely rare for this to last 2.5 years as it did for her.

Q: In the DirecNet trial, the curves are so flat it’s envious. Do you have any closed loop data on how many basal rates are appropriate for the average patient?

Dr. Buckingham: These are the wrong subjects to test that. They have endogenous insulin production. You’re much better testing that in someone who is a true C-peptide negative. What this study has taught us is a lot about meals. When you put a lot of fat in a meal, it’s way different than a carbohydrate only meal.

Q: Do you have data on insulin-depleted patients and basal rates?

Dr. Buckingham: No, I don’t have any.

Dr. Saleh: It sounds like with remote monitoring you are taking the technology further, beyond just closing the loop.

Dr. Zisser: I like to say it’s like the space program. We are headed to the Moon but will get Kevlar and Tang along the way. One thing we talk about with Bruce is the potential for meal detection. Another is the idea of remote monitoring – had we not been downloading data in real time wouldn’t have been able to do this.


Erik Dutson, MD (University of California, Los Angeles, Los Angeles, CA)

Dr. Dutson discussed bariatric surgery in a broad overview that included mechanistic speculation, clinical outcomes data, and squirm-inducing videos. He estimated that only ~5% of surgery’s mechanism is actually well understood; many previous theories have been complicated by new data. For example, researchers once thought that intestinal remodeling was necessary to affect metabolism, but this does not explain the efficacy of vertical sleeve gastrectomy (which involves cutting out a part of the stomach but preserving the intestines). He noted that despite the benefits of bariatric surgery on diabetes, the UCLA team will operate only on patients that have already lowered their A1c to 7.5%. He explained that too-high long-term glucose is associated with delays in post-surgical healing that can cause serious complications in death.

Questions and Answers

Q: We have a patient that received bariatric surgery and now has severe hypoglycemia. What is your experience in treating this?

Dr. Dutson: Thankfully we have seen this only a couple of times. Sometimes cutting out refined carbohydrates can improve the situation, but this may not work. The problem is that after bariatric surgery, insulin peaks very quickly after meals; this can be very severe for some patients. In some rare cases we have had to reverse the surgery.

Debate: Glucose Variability Significantly Contributes to Diabetes Complications


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

Dr. Irl Hirsch took the pro side of the glycemic variability debate, beginning with a historical review of the topic. He emphasized that in the DCCT, the treatment group could only explain 6.6% of the variation in retinopathy risk. Of this, A1c explained 96% of the variation. A separate analysis combining A1c and duration of diabetes only explains 11% of the variation in retinopathy risk – Dr. Hirsch suggested this leaves ample room for glycemic variability to play a role in development of retinopathy. He next turned to some of the clinical data supporting the detrimental long-term effects of glycemic variability and hypoglycemia, cautioning that DCCT seven-point profiles are biased for this type of analysis. Dr. Hirsch concluded by using unpublished Helmsley data to link the recent shift towards pumps and MDI (i.e., reducing glycemic variability) with reduced rates of retinopathy. We appreciated his clever arguments and hope to see supportive results from his upcoming feasibility study, FLAT-SUGAR.

  • Dr. Hirsch summarized the history of the debate around glycemic variability. The basis for glycemic variability comes from a 1995 analysis in Diabetes comparing the risk for sustained diabetic retinopathy in the DCCT conventional and intensively treated groups. Conventionally treated subjects at a given A1c had a higher risk of complications then those with the same A1c in the intensive group. Many had hypotheses that the difference was due to glycemic variability. In 2006, a study from Monnier et al. in JAMA found a strong link between urinary markers of oxidative stress and glycemic variability (r=0.86, p<0.001). Drs. Hirsch and Brownlee then wrote an editorial in JAMA that same year arguing that glycemic variability may be an A1c- independent risk factor for complications. In 2008, the 1995 analysis was retracted due to a “statistical artifact” (Lachin Diabetes 2008).
  • Dr. Hirsch emphasized that very little of the risk of retinopathy progression in the DCCT is actually explained by A1c – in his view, this leaves ample room for glycemic variability to play a role. If 100% is the total risk of progression to retinopathy, the amount of variation in retinopathy risk explained by DCCT treatment group was only 6.6%. Of that 6.6, A1c explains 96% of the variation. In a separate analysis, glycemic exposure (A1c plus duration of diabetes) explained only 11% of the variation in retinopathy risk in the DCCT. Dr. Hirsch thus believes there are other factors that may explain the remaining 89% of risk – glycemic variability, environment, and possibly genetics.
  • There is clinical data to suggest glycemic variability is detrimental. Dr. Hirsch’s “favorite study” appeared in Circulation in 2004 (Esposito et al.) – 87 patients with type 2 diabetes were treated with either repaglinide or glyburide and followed over 12 months. Although both arms had an identical drop in A1c (0.9%), carotid IMT regression (the primary endpoint),was observed in 52% of patients receiving repaglinide and in 18% of those receiving glyburide. The reduction in carotid IMT was associated with changes in postprandial but not fasting hyperglycemia. This suggests a beneficial effect of targeting glycemic variability. Dr. Hirsch also cited two ICU studies (one in adults and one in pediatrics) and one transplant study that associated higher glycemic variability with increased mortality.
  • Studies of the DCCT data have not linked glycemic variability to complications, but Dr. Hirsch believes the seven-point profiles are unreliable for this type of analysis. He emphasized that that these are retrospective observations that have a potential human behavior bias. Back in the 80s and 90s, lancets “were barbaric” – Dr. Hirsch knows of DCCT participants that would do a seven-point profile at one time by filling up all seven tubes with a single sample! We’d certainly agree with his view that “human behavior can have a huge impact on how we look at these studies.” Hypoglycemia, a key part of glycemic variability, has also been linked with long- term complications. First, Dr. Hirsch noted that hypoglycemia is more complicated than neurocognitive changes and cardiac arrhythmias, both of which can be fatal. Just one hypoglycemia episode can result in proinflammatory cytokines, markers of lipid peroxidation, and reactive oxygen species. Moreover, data from the ADVANCE study linked hypoglycemia with microvascular and macrovascular complications, death, cardiovascular disease, and non- cardiovascular death.
  • Dr. Hirsch made a case that the recent philosophical shift in diabetes care towards reducing glycemic variability (e.g., MDI and pumps) has coincided with reduced rates of retinopathy. Using unpublished data from the T1D Exchange, Dr. Hirsch examined a subgroup of 1,440 patients diagnosed around 1980. The group had an A1c of 7.8% and a rate of proliferative retinopathy right between the DCCT intensive and standard care groups. He noted that this outcome aligns with the shift in diabetes care that occurred over that time period, from a standard of one to two injections of NPH and regular per day in the 80s and 90s to increasing use of pumps and MDI in the early 2000s (especially with the introduction of glargine in 2003). To further bolster his case, Dr. Hirsch discussed a study (Diabetes Care 2011) in 1,604 adolescents stratified over four time periods since the 1990s. Over the 20-year span, the risk of retinopathy declined from over 50% to just 10%. In parallel, the use of MDI and pumps increased from 16% to 90%. This was accompanied with just a minimal change in A1c, suggesting the reduced glycemic variability may be responsible for the improvement in retinopathy. Finally, he mentioned a recent study appearing in Diabetic Medicine (Kim et al., 2012) that found a marker of glycemic variability (1,5 AG) was highly correlated with diabetic retinopathy in 567 type 2 patients.
  • ARQ #1: Dr. Kilpatrick will tell you glycemic variability doesn’t matter, however if you had diabetes, what pattern would you want your CGM to look like?
    • 1. (A wavy CGM trace) – 6%2.
    • 2. (A very wavy CGM trace) – 4%3.
    • 3. (A trough-shaped CGM trace) – 0%4.
    • 4. (A flat CGM trace) – 90%
  • ARQ #2: You are about to hear an anti-glycemic variability talk. What is your best way to prepare for it?
    • Read a book? (British Sign Language for Dummies)
    • Watch a James bond film festival?
    • Ski jumping lessons from Eddie the Eagle?
    • Single malt scotch charged to Dr. Klonoff’s account? (Majority vote)
    • Dinner with Elizabeth Hurley?


Eric S. Kilpatrick, MD (Hull Royal Infirmary, Hull, United Kingdom)

Using mainly retrospective analysis of large clinical trials, Dr. Kilpatrick argued that glycemic variability has not been convincingly shown to correlate with microvascular or macrovascular complications in either type 1 or type 2 diabetes. He concluded that if asked to rule on whether glycemic variability is “guilty” or “not guilty” of causing vascular disease, he would choose a third option that is used in Scots Law: “not proven.”

  • To begin, Dr. Kilpatrick noted several circumstances in which he thinks variability in blood glucose does seem to be a relevant risk factor. Some evidence suggests that variability in A1c (i.e., month-to-month variability) might be an independent risk factor for microvascular complications in type 1 diabetes (Kilpatrick et al., Diabetes Care 2008; Waden et al., Diabetes 2009), and two upcoming papers support that such a relationship exists in both type 1 diabetes and type 2 diabetes. Dr. Kilpatrick also accepted that intra-day glycemic variability is a risk factor for hypoglycemic episodes (Monnier et al., DTT 2011) and that it is important to avoid in intensive care settings.
  • Dr. Kilpatrick questioned the premise that glycemic variability raises the levels of reactive oxygen species (ROS), a frequently proposed mechanism for hyperglycemia-induced cellular damage (Brownlee et al., Diabetes 2005). Monnier and colleagues have famously found that mean amplitude of glycemic variability (MAGE) seems to correlate with urinary isoprostanes, a measure of ROS (Monnier et al., JAMA 2006). But subsequent analyses by Hans deVries’ team, who used a more specific measurement of urinary isoprostanes, found no relationship between urinary isoprostanes and MAGE in either type 1 (Wentholt et al., Diabetologia 2008) or type 2 diabetes (Siegelaar et al., JDST 2011).
  • To assess the effects of glycemic variability on microvascular complications in type 1 diabetes, Dr. Kilpatrick focused on the seven-point SMBG data from DCCT/EDIC. (He acknowledged Dr. Hirsch’s point that some of the data may have been faked, but he said that actually very patients had all seven of a day’s BG values the same.) Although standard deviation in the DCCT data has been seen to correlate with the risk of retinopathy, two separate analyses showed that this relationship disappears after controlling for mean blood glucose (Kilpatrick et al., Diabetes Care 2006; Service et al., Diabetologia 2011). Two other measures of variability, MAGE and glycemic risk assessment diabetes equation (GRADE), were also not found to correlate with microvascular complications risk in DCCT (Hill et al., Diabetic Med 2007). Further, no relationship emerged within the four years of publicly available data from DCCT’s follow-on study, EDIC (Kilpatrick et al., Diabetes Care 2009). Dr. Kilpatrick concluded his look at DCCT/EDIC by disputing Dr. Hirsch’s claim that at any given A1c level, intensively treated patients had lower risk of complications than patients in the conventional treatment arm. DCCT lead statistician Dr. John Lachine has written that this phenomenon was actually an artifact of the modeling technique that he and his team originally used (Lachin et al., Diabetes 2008).
  • Dr. Kilpatrick reasoned that within the intensive treatment arm of UKPDS, patients randomized to insulin likely had higher glycemic variability than those randomized to sulfonylureas; he thus used UKPDS treatment assignment as a proxy measure to see whether GV causes microvascular complications in type 2 diabetes. Each arm of the intensive group had highly similar results with regard to A1c change and diabetes-related endpoints (and myocardial infarction, he also noted). Dr. Kilpatrick thus concluded that GV seems not to affect complications risk in type 2 diabetes.
  • Dr. Kilpatrick also rejected the hypothesis that hypoglycemia increases microvascular risk. In ADVANCE, patients that experienced one or more severe hypoglycemic episodes seemed to be at increased risk, as Dr. Hirsch noted. But the ADVANCE authors have proposed that this effect may be more easily explained by other factors about the patients with severe hypoglycemia risk – e.g., older age and longer diabetes duration. Dr. Kilpatrick and his colleagues also searched the DCCT data for a relationship between severe hypoglycemia and retinopathy and nephropathy, but they did not find one (Kilpatrick et al., Diabet Med 2012).
  • Glycemic variability has not been shown to affect macrovascular outcomes, in Dr. Kilpatrick’s interpretation. Dr. Kilpatrick said that many plausible mechanisms have been proposed for how postprandial hyperglycemia could increase cardiovascular risk: abnormal vasodilation, hypercoagulability, protein kinase C activation, etc. (Metabolism 2006). However, he questioned whether postprandial hyperglycemia is really special (“How does the body know whether glucose is pre- or postprandial) or whether it might just be a proxy for mean blood glucose. To explore the issue he turned to HEART2D, in which high-CV-risk type 2 diabetes patients were randomized to either thrice-daily prandial insulin or once-or-twice daily basal insulin. Patients in the prandial arm had slightly less intra-day GV, but after a mean of 2.7 years follow-up the two groups were identical in mortality and extremely similar in number of CV events (indeed, the study was stopped due to “statistical futility”). Perhaps even more compellingly, subsequent analysis of both groups together showed no relationship between GV and whether patients experienced an event (p=0.49-0.57) (Siegelaar et al., Diabetes Care 2011).


Bruce Bode (Atlanta Diabetes Associates, Atlanta, GA), Irl B. Hirsch, MD (University of Washington, Seattle, WA), Eric S. Kilpatrick, MD (Hull Royal Infirmary, Hull, United Kingdom)

Debate moderator Dr. Bode gave each speaker a few minutes for rebuttal, and then called a vote by “clap-o-meter” before opening the discussion to the audience. Although both debaters got loud (and well- deserved) applause and Dr. Bode smiled that both had won, Dr. Hirsch’s applause sounded louder to us.

  • Dr. Hirsch: You were the first to prove, and others followed, that A1c variability seems to have an impact on retinopathy and other complications. I would argue that A1c as a measurement is easier and more standardized than 7-point profiles, which are fraught with problems. The consistency of risk with A1c variability gives a message, I think. (I don’t want to minimize the issue of hyperglycemia per se. You’ve showed that A1c variability has the greatest impact in those with the highest mean blood glucose – the A1c variability at lower A1c values did not have as big an impact.)You did not talk at all about Dr. John Lachin’s analysis indicating that A1c is responsible for only 11% of the variability in outcomes. Maybe this is not completely appropriate, but I am stillstruggling for the other 89%, and somewhere in there I think is glycemic variability. How much is not clear.

Throw out the UKPDS data from your argument, because in that study they did not use soluble insulin, which in the US we call regular insulin. They used Ultralente – no prandial insulin until the end of the study. So trying to bang on variability with UKPDS data is ludicrous.

HEART2D was not just underpowered. Though statistically significant differences were shown in post meal glucose control, the difference was only 16 or 17 mg/dl between the prandial and basal arms. This difference is probably too trivial to show a difference in outcomes status post-MI.

I think the biggest message from the ADVANCE data, even though we both showed it in our discussions, is that we can look at all this retrospective data but we really need a prospective study. The NIH appreciates that we don’t know why some people get into complications and others don’t, whether their A1c is 11.0% or 9.0%. I think that we have made it much too simplistic with this one number of A1c, which is crude and has many limitations. Hopefully as a clinical chemist you would agree with that.

  • Dr. Kilpatrick: I don’t have much to add. You mentioned the 89% of complications risks that is not explained by glycemia in the DCCT and EDIC analysis. However, I don’t think you were able to show that glycemic variability is any more than 0% of that. You also mentioned about carotid intima thickness and the intensive care data. The majority of patients with diabetes are worried about MIs, strokes, and kidney disease. Talking about CIMT is not enough. On the point about type 1 patients having less retinopathy than patients at the same A1c treated a long time ago – we’re using more ACE inhibitors. There may be other confounds unrelated to glycemia as to why that risk of retinopathy or nephropathy is lower. I put on my slides that glycemic variability has no effect on complications risk. It’s not proven. I agree with you that we need a prospective study. We need to show once and for all if it matters or not.

Questions and Answers

Q: It seems like we’re only looking at endpoints and one or two surrogate markers – A1c or glycemic variability – to draw conclusions. Should we be looking at other variables like genetic differences?

Dr. Hirsch: I think that who your parents are has a huge impact. The important part is that 89% that we cannot explain. How much of that is in the bucket of genetics? And then which are the best candidate genes? I know the VEGF gene is being looked at. It really does need to be better worked out.

Dr. Kilpatrick: DCCT complications risk seems to run in families. But it doesn’t explain 89%, nor nearly 89%. There must be other things as well.

Dr. Jeffrey Joseph (Thomas Jefferson University, Philadelphia, PA): Are there any data that relate GV and abnormal lipid profiles?

Dr. Kilpatrick: Excellent question. I am not aware of any.

Dr. Hirsch: Nor am I.

Dr. Bode: Some data indicates that that higher postprandial glucose is linked with higher lipids. Postprandial glucose would go along with GV.

Dr. Larry Hirsch (Becton Dickinson, Franklin Lakes, NJ): Reducing A1c lower the risk of microvascular complications by 50-75%, depending on the cohort. But that only explained 11% of the between group difference. What would you expect the other 89% of factors to do to ultimate clinical course of retinopathy? If you have eight times the variation not explained, and you could lower that, you would have to drive progression of retinopathy to zero. I cannot imagine that significantly reducing another factor would have the type of effect that mean A1c reduction has.

Dr. Hirsch: I don’t know how to answer that. When we first started using A1c in the early 80s, it was very rare to see someone with type 1 diabetes in single digits. We didn’t have the tools. We developed A1c around the same time as we had human insulin and home blood glucose monitoring. I think what has happened, taking off my variability hat, is a combination f things. Things went from average A1cs of 11- 12% to 9%. That’s had a huge impact. As important, if not more important, has been mealtime insulin. Those are incremental benefits as far as complications.

Q: I am fascinated by the story of hypoglycemia leading to more events in the year after. Can you speculate on the mechanism? Some speculate that the rate of recovery and potential rebound hyperglycemia is harmful, as well as the hypoglycemia itself.

Dr. Kilpatrick: I’d be keen to separate out macro- and microvascular complications. Even in ADVANCE the authors were not certain that a relationship existed with microvascular. With macrovascular you can see a number of mechanisms: catecholamines, inflammatory cytokines, etc.

Dr. Hirsch: What was shown in ADVANCE, even though the investigators rode the fence on it, is what we’ve all been seeing for some time. The DCCT population was younger and more robust than that of ADVANCE. This would expect why there was more of a problem for the patients that got hypoglycemia in ADVANCE.

Dr. Robert Vigersky (Walter Reed National Military Medical Center, Bethesda, Maryland): You both agreed that we need good studies to either prove or not prove this. The problem is that to power such a study, you would need so many patients that it would be financially impossible to fund this. You would have to reduce the rate of retinopathy to a pretty low level, if not zero.

Dr. Kilpatrick: My own view is that it may be done more by fortune than by design. New therapies like GLP-1s and SGLT-2s all reduce postprandial hyperglycemia, and therefore glycemic variability. If there are studies of those agents versus others, we may be able to show a complications risk difference. But then, is it the agent itself or glycemic variability?

Dr. Hirsch: We’ve been thinking about this in a couple ways. One, if you’re looking at complications, retinopathy is the wrong one. The better one is albuminuria. We are getting ready to start a feasibility trial called FLAT-SUGAR. The two treatment groups will be type 2 diabetes patients randomized to basal-bolus insulin or basal insulin with exenatide. We hope to be recruiting in the next couple months. Bruce has his patients ready to go. To do the study right you must have few things in place. You must show a clinically significant difference in variability. We think we can do that with the GLP-1 exenatide. You must have identical A1cs between the two treatment groups. If we show a difference in variability, then we hope to do a definitive trial after that. If we look at retinopathy, I agree with you. But the nephrologist has told me we can do this with albuminuria.

Patient Panel of CGM Users


Moderators: Irina Nayberg, RN, CDE, BSN (Mills-Peninsula Health Services, San Mateo, CA) and Donna Miller, MSN, FNP-C, CDE (University of Southern California, Los Angeles, CA)

Q: Could everyone please introduce himself or herself?

Eric: I’m 28 years old and was diagnosed at 22. I’ve hard Dr. Bruce Bode and Dr. Anne Peters as my doctors. I’ve used a Dexcom Seven Plus for 2.5 years. I had a seizure about three years ago from an overnight low. I put on the sensor after that. It’s one of the most important pieces of technology in my arsenal now. I wear the OmniPod and use Apidra SoloStar pens, which are not available right now. So I just use the pod.

Jim: I’ve had type 1 diabetes for 33 years. I’m a dinosaur and I still use injections. I also use the Dexcom Seven Plus and have had very good success.

Ana: I’m type 1 and have been wearing a Medtronic insulin pump for four years. I’m using the sensor as well with that.

Neil: I was diagnosed with type 2 about 16 years ago. I’ve progressed to a 1.5 or 1. I’ve worn the Dexcom Seven Plus for about three years. I also began using it in response to a super low overnight incident. I use injectable insulin.

Toby: I’ve been type 1 for 46 years. I use a pump and sensor.

Q: Can you tell the story as to why you went on the sensor? Was it your healthcare provider? Did you read about it yourself? Has it met your needs?

Toby: My care advisor played an instrumental role. I had many years of just not doing that well and wanting to do better. I personally get a lot of security from having the sensor. It’s taken me out of the Stone Age and my control is better.

Neil: I was overseas in Italy and had a typical pasta, bread, and alcohol dinner. I should not have woken up with a blood glucose of 37 mg/dl. I had never had a hypoglycemia episode like that. It scared me. My provider had been suggesting a sensor. I didn’t understand the need for it, but it was clearly something I needed to do. I truly depend on it now. The printouts over a period of time are so clear. You can see the pattern and they make a big difference if I’m eating regularly. Ironically, my sensor ran out of juice Thanksgiving weekend. I was pretty well bereft waiting until Monday to call Dexcom and get a new unit. The positives completely outweigh the negatives.

Ana: I wouldn’t say there was one event that caused me to use the sensor. It was my healthcare provider bugging me for year to use a sensor and pump. I was hanging out in the 200s and 300s and I think that was our main concern. It’s definitely brought those levels down. Knowing that I can go to sleep and it will wake me up at night with an alarm is really big. Before I’d wake up three to four times in the middle of the night. We’re meant to sleep at night, and now I can sleep.

Jim: I went on the sensor because my blood glucose control was erratic to say the best about it. Dr. Peters was instrumental in getting me on the sensor. I’m being kind with that word – she suggested that. I tried one system that did not work out; it was unreliable. I finally got to a point for international travel and I didn’t travel with the sensor. I came back and she suggested I switch over to the Dexcom. It’s been a revelation and I kind of feel naked without it. Every once in a while I will go without it for a day and I go back to testing more regularly. You become reliant on it. One of those little side benefits is that my fingers are in much better shape right now. The things are not all calloused. The sensor definitely helps with exercise too. I’m one of the lucky ones in that I do wake up when I’m low. But I cannot count on that; in the future, I will probably need the sensor even more.

Eric: The reason I was initially drawn to Dexcom was an event I had. I won’t go a day without it. It brings peace of mind during the middle of the night and wakes me or my wife up. You also see an improvement in A1c. Over the past couple years I came on to the Dexcom. I try to set a range as tight as I can, forcing myself to have better numbers. I’m part of the team type 1 triathlon team. The CGM is helpful for having the confidence when I’m out there on a four-hour bike ride. I would feel bad having the team stop for me so I could check my blood glucose. I now have my monitor mounted on my back. For me, it’s the trending ability – as I’m starting to go down, I need to do something. If I’m going for a mile swim, it’s tough to check my sugars in the water. Knowing where I’m headed with the trend is the biggest thing for me. It has met my needs and then some. Aside from the A1c improvement, analyzing the data with my doctor and looking at a specific event or race has been helpful. Was there a relation between what I did? Looking at those things to make changes accordingly is important, and it becomes second nature. The first time it’s a bit of a pain, but the benefit outweigh the trouble. I certainly won’t go a day without it.

Q: How has your control changed?

Toby: I never could have had this control.

Neil: The data have made things easier to manage. Again, I think it’s a function of seeing trending and being able to react to it. I also tell myself that next time, I am sure it will be better – there’s an element of eternal optimism.

Ana: Seeing trends and where I am headed rather than reacting after the fact. With fingersticks you wind up playing the rollercoaster game, whereas with CGM I am able to go up and down much more gradually.

Jim: The first thing I thought of when you asked that question was [Dr.] Anne [Peters] looking at me and saying, “Not well enough.” Seeing the printouts and charts has really helped a lot. To see the charts and to be able to pinpoint problems, besides just eating – to see when to inject and what to inject. We recently added NPH insulin, which I had never used – I had used Humalog and Lantus. Adding NPH has made a difference.

Eric: The improvement is there, and it’s undeniable. Aside from the actual data, there’s the knowledge of what I eat and how it affects me. It’s about being able to see what happens if I eat a certain meal, if I take insulin at different times. This helps me see when to give that injection and how much to dose.

Q: What was the longest period of time you wore a sensor?

Toby: Usually it’s working really well when I have to take it out. That’s on the fifth day. I never like to be without it.

Neil: The meter tells you to change it. Can I wear it longer? I’m guessing it could. But I know once the countdown runs out and I don’t really let it go much beyond that, particularly if I’m away from home.

Ana: I do trick the sensor for another couple days. This is for Medtronic and I wear it about five days. I’m sure with more experience, I could get it to wear it longer. If it tells me to change it, I change it.

Jim: I just follow it and change it after seven days. It gets kind of itchy at the site, so I’m happy to change it and put it on the other side of my body. I tend to keep my injection sites away from the sensor on one side of the body or the other. It’s nice to give that side a rest too.

Eric: I try to not ever have a break. As you know, Dexcom Seven Plus is “approved” for seven days. I have “heard” of people wearing it for more. I’ve had one go for almost two weeks more. I’ll try to change it once the numbers start to go wrong. As long as the numbers are good, I throw some tape on there.

Q: Do you have site issues? Adhesive issues?

Eric: I swim three to four days per week. With OmniPod and Dexcom, I’m amazed why the companies can’t use the same adhesive. Dexcom for the most part is not a problem. But my pod will fall off occasionally. If it’s a long training week, sweating and in the pool, I will put tape over the Dexcom. But for the most part it’s not necessary every time. For the OmniPod, I occasionally will have to tape it down if I spend lots of time in the pool. Site issues are only occasional with Dexcom, and that’s if it’s on for too long. I don’t know if chlorine helps but it doesn’t seem to get too bad.

Jim: The site gets kind of itchy. I also have to be careful with placement. If it’s on my side too close to my ribcage, I then can’t sleep on my side. With the previous system, one of the final straws – besides its unreliability – was that I was standing in the middle of winter in a post office parking lot, and I felt the sensor just fall off. The Dexcom adhesive has been pretty great. In the more recent sensors I’ve had, the adhesive wipes off easier. It kind of looks like a battle with a squid, but it comes off fairly easily.

Ana: When I was first introduced to the sensor, I was given swabs that made it stick better. For both the sensor and pump the adhesive is great. I actually have a hard time taking it off. For the site, this is completely a personal issue: I need to call Medtronic and see if they make smaller cannulas. I am somewhat limited to where I can put the pump site; I feel it can get into my muscles. I now use a 6 mm cannula and am not sure if a smaller version exists. I don’t have any issues when I lie on it; it’s just the cannula depth.

Neil: I wear the sensor on my upper thigh or abdomen. Generally, I don’t have any problems. I need to use alcohol to get the adhesive off. It is better if the site is clear of hair. Sometimes I get bruising, but far less than with regular injections. Sometimes the sensor adhesive needs to be supplemented with tape, but generally this is not a problem.

Toby: When I first started I had site allergies; I have changed the tape since then and this is no longer an issue. With regard to the site, I tried using a 6 mm cannula, but I need a larger cannula. The downside is that the 9 mm then pokes into my muscles. It’s hard to get it comfortable, but it’s the most comfortable infusion site I’ve had yet.

Q: I’m wondering about alarms. You’ve made positive comments. We want to hear some issues. Do you have any alarm nuisance? How do you manage those?

Toby: Randomly, I’ll get a low blood sugar alarm and I’m not low. But that’s a random event. I would appreciate a louder alarm, but other than that, no.

Neil: On the Dexcom, you cannot adjust the low alarm – it’s like a foghorn. If you are trying to react to a low, you cannot tell it to shut up if you’ve eaten. You’ve gotten up, you’ve had some juice or a snack, and you want to go back to bed. But it takes a while to react. Generally, it just ends up in our drawer. Once I’ve corrected it, I don’t expect to have another problem during the night.

Ana: For my sensor, I would prefer more options than low, medium, and high. I can ignore my phone alarm for hours in the morning, let alone this under the covers. It’s not loud enough at night. I teach at a community college. I thought ‘oh cool, I can put it on vibrate.’ But I’ve learned that if you ignore it, it blows up really loud. This happened in front of my class once. I’d rather have nighttime taken care of with the louder alarm. During the day, I’d rather be safe than embarrassed.

Jim: Dexcom has a great system. My one major complaint is that once you’ve fixed the low, it keeps alarming. You need to have a setting or button, “I just took care of it, now leave me alone.” I have to resist the urge to throw it against the wall. I just go and put it on the other side of the house. I put it in a timeout. [laughter] That’s the one flaw in that system for me. I do the same thing for the settings in terms of a high alarm. I also teach and I’m a poet. I give readings sometimes. That’s the last thing I want in the middle of a reading. Adjustability of alarms – on, off, vibrate – would be good. There needs to be something to say “thank you.”

Eric: I would echo all that’s been said. We’re used to phones that have so many options. Having flexibility to change it would be good. At nighttime, the vibration sometimes doesn’t wake me up. My wife usually does, but it’s not fair to her. Having the flexibility to change the settings – now it’s either a foghorn or a vibration that you may not feel. That would be my only thing.

Q: What are your obstacles to wearing a sensor? What prevents you from wearing it all the time?

Eric: In combining with the pump, one obstacle is placement. There aren’t many places I can insert the Dexcom.

Jim: I wear it all the time, but I will go for a day without it every once in a while, just to be away from a machine.

Ana: I recently started wearing the sensors, so maybe I will have more to say on this in six months. I already have few places with pump. With another item, it might get difficult.

Neil: I don’t not wear it. The receiver is big – that’s the only deterrent for me. It’s a nano-world, and the receiver is the size of a pager. Looking for that, in addition to my wallet, keys, and phone, is a hassle. Then, having big enough pockets to hold everything is a problem. It is bigger perhaps than it needs to be, given the functionality of all these other small devices available today.

Jim: I would like to second that; that is a great point. I carry it in my front pocket, and it is big. Neil: Yes, then there are the times you forget and jump in the pool…

Toby: I share others’ concerns about territory on the body, and concerns about building scar tissue such that I have even fewer possible places to put it in the future.

Q: What do you want to see in the next generation?

Toby: Better accuracy

Neil: No calibrating against a test strip and a smaller design.

Ana: I completely agree. Something more in line with true blood glucose readings. Sometimes I read it and it’s one point off, and I’m amazed. And other times it’s as big as 60 mg/dl. I see the trends, but my sensor is saying 140 mg/dl and I check and I’m 90 mg/dl. I could use a little more accuracy.

Jim: That 20% error can grow on either end. For the future, I would love to see islet cell transplants. I was told at the time I was diagnosed that it was ten years away. In the interim, these technologies are helpful. I would like a smaller device with no calibration.

Eric: It would be great to have a combined phone, pump, and Dexcom. It’s so many devices. I know it’s on the way, but that’s the main thing for me. It’s literally four pieces of equipment to carry around at all times.

Q: How has your significant other reacted to your having this?

Toby: My first husband would not have liked it, even though he was a scientific-type person. My current husband is very understanding; he says he doesn’t even see it. I don’t know if I could be as generous.

Neil: My hubby – the whole diabetes thing scares the bejeezus out of him, and this is just one more aspect of it. I don’t think he understands why it’s beeping or going off; I’ve tried to reassure him it is a security device. Other than it being a visible reminder that I have diabetes, it hasn’t changed our relationship much.

Ana: My husband could care less; I am quite fortunate. Donna was bothering me for years and years before I went on the sensor. At the time I was single and dating. It is hard going out at night, pulling out the receiver and looking at it freaks out a lot of people. It really does. That is hard. After I first met my husband, I was discussing it with him, and he said “Do it.” It is now wrapped around his leg half the time in bed; but it doesn’t bother him.

Jim: I am also fortunate. My wife is part of treatment, “Don’t you have to change your sensor today?” The only real drawback I mentioned is that she wakes up with the alarms.

Eric: My wife works for JDRF in LA; she often comes home with facts that are new to me. I am lucky in that regard.

Q: How long did it take to get used to downloading? What was your learning curve? How many of you download at home?

[Only one panelist downloads at home]

Jim: I work on a Mac. It was not very compatible with a Mac. I don’t want to go spend $100-200 on a cable.

Q: When would you download?

Eric: I just got access to the IBGStar. I love it. I’m a data person and it’s really easy to use. It has great accessibility.

Neil: Dr. Peters will take the email PDF and interpret it. To have her be able to see it is quite astonishing. I will say one thing about functionality. On the Dexcom system, if you’ve forgotten what you’ve entered, you cannot call back up an insulin or blood glucose entry. The only place you see it is on the printouts. It doesn’t tell you that on the receiver. I always have to go back and check my meter. I forget between the bathroom and bedroom.

Q: Is it unrealistic to ask you all to download your pumps and CGM?

Toby: We all ask Dr. Peters to interpret it for us. It might be asking a lot for me to interpret it.

Q: That’s my expectation. I always think I can teach my patients to make their own changes. I guess that’s kind of an individual thing…

Jim: This is a constant guessing game, a juggling act. It’s hard. To expect that you’re going to sit down and read this graph that has seven lines in multi-colors? It’s not going to happen. I’m a person that has a smart phone that can do amazing things. I just want it to be a really good phone. At the same time, it’s a real balancing act. We really are our own mangers. There’s only so much you can do day to day.

Ana: There are preset reports. But I want the seven lines and this. I want an option to pick things. I want three basic items dumped into there and an area where I can dump food in. I want more flexibility. I have no desire to sit and look at this rainbow and trends.

Neil: Dr. Peters has thousands of patients, so she has enormous experience interpreting the data. For me to interpret it, I’m not really trained that way.

-- by Adam Brown, Joseph Shivers, and Kelly Close