Wednesday, May 19, 2010

El-Khatib's Group Announces Phase-I Results

If you are following Artificial Pancreas (AP) research, this announcement has it all: good experimental design, good results, a bright future, and some interesting side discussions.

First, the study:  These guys are testing an artificial pancreas which uses both Lispro insulin and glucagon.  Other AP projects are only dosing insulin.  Obviously, this gives the AP more control, it can dose insulin for highs and glucagon for lows.  The study was 11 patients none of whom generated their own insulin (so honeymooners excluded).  The study ran 27 hours and included 3 meals.

Second, the results: the 11 patients fell into two groups as far a results go, as described in the abstract:
In six subjects, the closed-loop system achieved a mean BG concentration of 140 mg/dl, ... There were no instances of treatment-requiring hypoglycemia. Five other subjects exhibited hypoglycemia that required treatment; 
So basically in the initial trial, it worked completely for six patients, but only partially for the other five.  So then they "tuned" their software, and got these results:
prevented hypoglycemia in both groups while achieving an aggregate mean BG concentration of 164 mg/dl.
So after tuning there were no more problems with lows, but the average BG was 164.
Now, remember that the current ADA guidelines for BG control is 154.  So the 140 is just within range, but the 164 is just out of range.  (Assuming you have a really accurate BG meter, of course. :-)  So for a phase-I trial this is a great result.

But the actual news is even a little better than averaging 164, because the researchers were able to discover why half the people did so much better than the other half in the first set of results.  The people who did well absorbed lispro insulin in about an hour, while the group that had low BGs absorbed it in about two hours.   One of the things that these researchers found, was that insulin is absorbed at different rates at different times, even by the same person.  So it is not a question that some people absorb it quickly or slowly, it's also that even in the same person, sometimes it will be absorbed quickly and sometimes slowly.  To quote Dr. El-Khatib (per. comm.):
We saw variations in the rate of insulin absorption by as much as 40–50% in some subjects between their first and second visits, as well as in an additional screening test that we ran on each subject.
(Think about this the next time you are wondering why the same dose of insulin has a different effect the next time you use it.)

I saw only three issues in this trial which made me nervous about it's applicability, and two are going to be addressed in research these guys hope to start very soon.  The first issue is activity.  For this test, the patients were lying in bed or sitting in a chair, so that was not realistic.  Second, they were using blood measurements of BG levels, not standard CGM measurements.  Third is the regular use of glucagon.  I discuss these issues below.

Third, the future:  For the one day the experiment was run, for 6 out of 11 people, the algorithm worked.  If these phase-I results carry forward into phase-II, phase-III and general availability, then this algorithm is likely good enough for a commercial AP.  Obviously, I would expect improvements so that phase-II and III would have even better results than these.  But these results by themselves are very good.

This research group is already planning to run follow on trials in Boston.  This first trial was run with very limited activity. But a follow on clinical trial will last longer (two days instead of one), use CGM technology, and more accurately represent real activity.  Patients will be able to move around (although they will be pushing around a pole with equipment on it), and they will spend some time on treadmills to model exercise. After that, there is a third trial, a follow on to the two day test.  In that third trial, people will use the bihormonal AP for a total of 5 days while "out and about" (doing what they would normally do).

Another difference in the future trials will be a partial pre-meal bolus.  This will be based on weight, and given with meals.  The researchers expect that this will keep them below the ADA's BG guidelines, especially patients who absorb insulin more slowly.  So if you know the JDRF "stages" of AP development, this would be a stage 6 AP because it includes glucagon, but a stage 4 AP in that you need to tell it when you are having a meal.  In either case, a big improvement over nothing, which is the kind of AP we (in the US) have now, or stage 1, which is available in parts of Europe.  You can read more about JDRF's stages here:

The use of standard CGM technology, starting in the next trial, is another important step forward, in my opinion.  CGM technology has some real limitations, and one of the big open questions in AP development is this: is current CGM technology good enough to feed a successful AP?  The completed trial doesn't directly address this, because they were using BG measurements from the blood stream, not interstitial spaces where CGMs work.

This work is based on giving Lispro for highs and glucagon for lows.  We have a lot of experience giving people insulin when their BGs are high, but much more limited experience giving people glucagon for lows.  Sure, we all have needles in red boxes, and some have even experimented with "microdoses" of glucagon.  But still, this AP will use glucagon much more than has been done in the past, and that is something that must be tested.

For the software nerds out there: basically, the software to control the pumps has one basic algorithm.  It's this basic algorithm that they are testing, and they are not going to change that. However this algorithm has a couple of parameters which the researchers can change.  Two of these involve how quickly insulin is absorbed, and these are the two that they changed in order to "tune" the algorithm for people who absorbed insulin more slowly.

Finally, a little fantasy of mine:  For one of the big diabetes meetings in 2011 or 2012 (ADA, or EASD, or "Friends For Life") there will be enough successful AP research projects, so the can have a "bake off" right there at the convention.  Each team brings a couple of APs, and some convention goers use them for the duration of the convention.  Then in the last session, they become a panel that discusses the current state of APs from a patient point of view.  Someone can chart their BGs on a couple of slides to get the discussion rolling.   I'd love to get a clinical trial number for that......

Research web site:
Clinical Trial Record:
Full paper:
Extra data:
    How to Summarize AP Clinical Trial Status

    I'm also thinking about how I can summarize an AP clinical trial is just a few lines, that are understandable.  For this trial, I'm thinking about something like this:

    A phase-I trial of 11 people over a 1+ day time period which includes meals but not activity.
    Results: The algorithm had 6 people averaging BG of 140 and no lows.  After tuning, all 11 averaged BG of 164, and no lows.
    Notes: This AP could dose both insulin and glucagon.  Commercial CGMs were not used (blood BGs were).

    If anyone has any thoughts on how to summarize AP clinical trials, please do share them with me. 
      I would like to thank Dr. Steven Russell (a principal investigator in this research) for answering my questions, and also Dr. Firas El-Khatib (a Co-investigator in this research) for providing additional information.  All mistakes are my own.

      Joshua Levy
      All the views expressed here are those of Joshua Levy, and nothing here is official JDRF or JDCA news, views, policies or opinions.