- advertisement -

Is a "closed-loop" CGM and Pump a Cure?

Discussion in 'Stickies' started by Sarah Maddie's Mom, Jan 6, 2010.

?

Is a "closed-loop" system a cure?

Poll closed May 6, 2010.
  1. Yes

    4 vote(s)
    1.7%
  2. No

    224 vote(s)
    96.1%
  3. I don't know

    5 vote(s)
    2.1%
  1. Darryl

    Darryl Approved members

    Joined:
    May 8, 2008
    Messages:
    4,313
    Christopher,

    The AP project, as far as has been published thus far, uses commercially available CGM sensors and insulin pumps, and proposes to link them with software.

    While those of us who use CGM's and pumps have found that they can be amazingly effective with human intervention, we know equally well that they would be both ineffctive and unsafe without human intervention. One example I have mentioned before is when a child sleeps lying on their sensor, the corrective action is to roll the child over, something a pancreas can not do, particularly an artifical one. There are many other practical realities that can't be addressed by a closed loop system - at least insofar as the system being any more "automatic" that it is today - such as having different bolusing rules when resting vs. exercising vs. during illness, or during days when there are persistent lows even at a low basal rate, and the now-admitted reality that the AP will not be able to bolus for meals, that will also have to be done manually as we do today.

    I mentioned FDA testing, because once this project tries to move beyond the realm of a 10th grade engineering exhibition, they will need to pass FDA phase I, II, and III trials. At that point, with large scale testing they will encouter kids who roll over onto their sensor and end up in DKA, or gross instabilities when sensors read the inverse of the actual blood sugar trend as they occaisionally will do, and the FDA is going to send them back to the lab with the instruction: "design a reliable sensor and come see us again when you have done so."
     
  2. Christopher

    Christopher Approved members

    Joined:
    Nov 20, 2007
    Messages:
    6,771
    Thanks Darryl, I understand your view better now.
     
    Last edited: Jan 18, 2010
  3. Toni

    Toni Banned

    Joined:
    Sep 14, 2009
    Messages:
    2,882
    Hmmmm.... If Darryl feels AP is not feasible, even with the reliable results his DD gets from cgms, it may very well not be so. I am wondering, though, if AP could be feasible as a hypo/hyper manager in the very low and very high ranges? Certainly DN would not be one of the first to use the AP unless the sensors become far more accurate for her. I would also like the computer to analyze daily blood sugars and make a computerized recommendation for basal and bolus changes. If AP could make basal suggestions every evening at 9pm, 12am and alert us through the night. Changes that are not automatic but that we could try. AP has kept subjects in range overnight in the absence of food so far using the computer to make changes in individual's basals. That has been successful.
     
  4. Toni

    Toni Banned

    Joined:
    Sep 14, 2009
    Messages:
    2,882
    Add to Darryl's suggested list of questions for JDRF:

    Why did J&J, immediately upon purchase of Animas, cease R&D on Animas' implantable sensors? I had thought Animas was pretty far along on this. Is it because a sensor, implanted once, or once a year would not garner the same revenues as sensors that must be changed every seven days?
     
  5. Flutterby

    Flutterby Approved members

    Joined:
    Nov 11, 2006
    Messages:
    14,623
    Kaylee has had ONE instance where she rolled over and the cgms dropped low immediately.. its NOT an evernight occurance.. she's been using a cgms for over 2 years.. somehow, I think the engineers that make the cgms and sensors know about this, they are well aware of whatever problems the cgms has.. they certaintly have more than a 10th grade education.. Not sure why the bashing of the cgm is needed. We all know its not perfect. I think the scientists know what they are doing. Far more than anyone on this board.
     
  6. Darryl

    Darryl Approved members

    Joined:
    May 8, 2008
    Messages:
    4,313
    This is feasible, and is not much of an extension from what pumps already do. This could be stand-alone software used right now by people who use CGM's and pumps, or it could be integrated into a CGM or pump.

    Yes, in 1-day time periods, in a hospital setting. In our experience, the CGM and pump could indeed be put on auto-pilot 3 nights out of 4 without incident. But not indefinitely.
     
  7. Darryl

    Darryl Approved members

    Joined:
    May 8, 2008
    Messages:
    4,313
    Not bashing CGM's, in fact I think they are the best thing in the world for someone with T1. The rolling over on the sensor happens more often than you might have been aware of, because if the child rolls over again on their own, the problem resolves. Any given little dip in the prior night's CGM graph could be sleeping on the sensor. Under closed loop control, the insulin would have been cut off. I have great respect for the scientists who are enthusiastic about this project, but I also have a masters degree in engineering and 15 years designing closed loop electronic systems, and 3 years using a CGM in real life, and I can't see any way for them to be successful until a better sensor comes along!
     
  8. Flutterby

    Flutterby Approved members

    Joined:
    Nov 11, 2006
    Messages:
    14,623
    Kaylee's typically a straight line at night.. some gradual movement, but typically straight line.. we do have our problems at night, but I don't see any evidence of the sensor problems with her rolling over. The one time it did happen, it was obvious.. she went from mid 100s to 40 in the 5minutes.. it was obvious what happened.. any other dips are proven with a bg check.. Maybe it has something to do with the fact that she sleeps on her belly almost 100% of the time and the sensor is on the back of her arm or her tush.. so she's not usually ontop of it.. we did try a sensor on her belly last week.. it was HORRIBLE.. kept getting week signals and large jumps.. so the sleeping on the belly definitely effected it. so I guess I can't say it happened once, because it did happen when she had the sensor on her belly for 2 days last week. We had to pull it early because it was horrible.. its also an expired sensor, by 6months, so its possible it could be that too.. although all of our sensors are currently expired, and we don't have the issue when its placed elsewhere.
     
  9. buggle

    buggle Approved members

    Joined:
    Mar 24, 2008
    Messages:
    4,267
    We have issues with false-lows at night from Brendan lying on his sensor all the time. Whenever we get a low alarm, we always make sure he's not lying on it and then watch for a few minutes in case he might have rolled over after the alarm and before we could check him.
     
  10. hawkeyegirl

    hawkeyegirl Approved members

    Joined:
    Nov 15, 2007
    Messages:
    13,157
    We do get the occasional false low at night. It's very obvious when that is the issue, because the dip is so very fast.
     
  11. Darryl

    Darryl Approved members

    Joined:
    May 8, 2008
    Messages:
    4,313
    I'm sure that not everyone experiences the lying-on-the-sensor issue to the same degree, it probably depends on sensor type, site placement, angle of insertion, the child's age and weight, what kind of bed they sleep on, etc. Hmmm... I wonder if they are including those things in the AP algorithms ;)?

    This brings up another issue... as we all know, new sensors need anywhere from a few hours to 12 hours be become "stable", then they are pretty good after that. Some sensors never become very accurate, but we use them anyway because they are "good enough" to maintain control for a few days and we'd prefer to not insert a new sensor when we can deal with the innacuracy using our best judgement.

    So with the AP... exactly when do you switch the sensor to "auto pilot"? Certainly not in the first few hours. And what do you do with the sensors that are just "OK"... or the sensors that radomly "cut off" or are innacurate every day or two? do you disable the auto pilot and run under manual control like we do now? More than likely, the AP studies done so far (the 1-day tests done in the hospital) were with sensors that were already stabilized and working well.

    So many questions, and so much money being spent on software, when the sensor is the real issue.
     
  12. Toni

    Toni Banned

    Joined:
    Sep 14, 2009
    Messages:
    2,882
    We have had false lows, but only twice, I think. Dex seems to immediately go over to ??? as soon as we hit the pillow.:( If we are still awake, we can roll her over and usually Dex will pick up a reading. Putting Dex in bed with her does not help, she drinks an awful lot of water, sensor still only worn on tush so have not trouble-shooted that theory. You all know how I hate to praise Minimed's Minilink, but no, we did not have this problem at all with the Minilink or lose the sensor overnight.
     
  13. Flutterby

    Flutterby Approved members

    Joined:
    Nov 11, 2006
    Messages:
    14,623
    Please, correct me if I'm wrong, or if I'm thinking about someone else. But doesn't Leah depend on her cgms to do a lot of the correcting? she eats something is she gets a low alarm, boosts her basal as she goes higher (or gives a correction)? She pretty much stays where she's suppose to (or use to before the whole growing up thing;).)
     
  14. Darryl

    Darryl Approved members

    Joined:
    May 8, 2008
    Messages:
    4,313
    Yes, we still use the CGM data to determine boluses and basal adjustments inbetween BG checks - that is, except when the sensor has not yet had it's 2nd calibration, when the sensor is on day 7, or when the data looks suspicious. In these cases we will do a BG check. Most days 3 BG checks are sufficient, on other days (mostly days 1 and 7) we do more BG checks. In this regard we could use a closed loop system on some days, on those days when the sensor behaves perfectly, but we would not know that without looking at the data with our own eyes.

    It's been harder since September with her TDD basal anywhere between 10u and 60u depending on the day and high's coming on faster than in prior years, but she is still maintaining an 30-day average BG around 102 with no symptomatic hypo's. There are more days now now where BG gets into the mid to high 200's, and occaisionally a low 300 for a little while until she increase the insulin enough bring it back down.
     
  15. Flutterby

    Flutterby Approved members

    Joined:
    Nov 11, 2006
    Messages:
    14,623
    So, if you trust the information that the cgms is giving you, and you believe its accurate (to be able to correct off of, it must be accurate).. why don't you believe the AP would work..

    I'm totally not trying to be a pain the A here.. I'm really curious.. Kaylee's sensor is extremely accurate but we do not do any corrections off of it.. we still check between 8-10 times a day.. (it was 15+ times before cgms..).. I also assume her A1c has to be fairly close to her sensor average and what the sensor is showing you, otherwise you wouldn't be trusting it like you do.
     
  16. Darryl

    Darryl Approved members

    Joined:
    May 8, 2008
    Messages:
    4,313
    The issue isn't accuracy, it's reliability. Sensor readings are often 90% or more accurate when the sensor is working reliably, and yes, over a 3-month period, the average sensor reading predicts A1C rather nicely. However, on those days when not working reliably, the accuracy can be way off, and worse, the sensor's trend can be the opposite of the real BG trend.

    When a sensor is unreliable, it is not fit for inclusion in a closed loop system. Sensors become unlreliable in a number of ways:

    - Until around the 12th hour of operation, the cal. ratio drifts, so a rising BG could appear to be a falling BG or vice-versa.
    - On the last useable day of the sensor (day 7 in our case), the same thing happens
    - Pressure on the sensor, such as while sleeping
    - If the sensor becomes partially dislodged, or the site goes bad
    - Some sensors are randomly unreliable throughout the time they are used.
    - A single bad calibration can cause a CGM to report wildly innaccurate readings.

    In a closed loop system, accuracy needs to be around 90%. However, if even for a brief time the sensor becomes unreliable - such as reading the opposite of the BG trend as happens when there is pressure on the sensor or drift in the sensitivity when the sensor is either new or old, the insulin delivery will be the opposite of what is needed, which leads to unstable operation.

    There is no way that software algorithms are going to address this issue well enough to trust a child's life to. It requires human eyes looking at the data, and correlating questionable data with knowlege of the age of the sensor, the track record of the sensor, whether the child just ate, or was exercising, or might have bumped the sensor or is sleeping on the sensor, etc.

    The best analogy I can think of is cruise control in a car. We can trust it most of the time. But what if the speedometer was only reliable on some days? What if on some days is showed that we are slowing down, stopping, when we are really speeding up? In that case, the cruise control would make things less safe than driving manually. With closed loop control of CGM's and pumps these situations will arise often.
     
  17. Flutterby

    Flutterby Approved members

    Joined:
    Nov 11, 2006
    Messages:
    14,623

    I see your point.. I don't know if I could ever trust a machine to do the job I'm doing.. case in point.. our cgms is extremely accurate (with the occassion opposite trending, which is really frustrating, and last day readings are sometimes off.. first day is typically very good.. ) but we still test 8-10 times a day, and we don't dose off of the cgms alone.. although, now with the guardian and its predictive alarms.. I can see where it would come in handy.. and be able to use those to give some glucose to help avoid the low..we avoid two lows today, the third we didn't avoid but would have been horrible if not for that predictor alarm.. we are actually still dealing with it.. she had a predictor alarm go off.. she still had 1.3u of insulin active from dinner.. she's had two juiceboxes.. and still alarming low.. I have NO idea where these lows are coming from..


    sorry to get off track.. but I see your point.. I think it would take an extremely long time for it to get FDA approved, and it should.. its something that should be gone over with a fine toothed comb.
     
  18. Darryl

    Darryl Approved members

    Joined:
    May 8, 2008
    Messages:
    4,313
    Yes, that is exactly what I am saying... in control loop design, "opposite trending" as you put it, or as we refer to it in engineering as "180 degree phase reversal", is a recipie for disaster in a closed loop system, as is any change in the sensor's sensitivity (in this case, glucose sensitvity).

    When CGM sensors are as reliable, repeatable, and accurate as car speedometers, then we can legitimately pursue a closed loop insulin control system, which at its best will provide a similar kind of semi-automatic control with constant human oversight.
     
  19. frizzyrazzy

    frizzyrazzy Approved members

    Joined:
    Dec 23, 2006
    Messages:
    14,141
    it begs the question: If this is the best they have to offer us now or in the foreseeable future, how can ANYONE call it a cure?
     
  20. Toni

    Toni Banned

    Joined:
    Sep 14, 2009
    Messages:
    2,882
    Not only is this not a cure, it is NOT an Artificial Pancreas either. More like a smarter pump. Which I am all for....... but a cure, no.
     

Share This Page

- advertisement -

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice