The Artificial Pancreas Scares Me

Type 1 and LADA
1

After 40 years of testing urine, and not really knowing my blood sugar levels, it was wonderful to have my first glucometer in the mid 1980’s. My control was really bad those first 40 years. After using a meter it was obvious that my BG had been much too high during those years. I was seeing numbers in the 200’s very frequently, and numbers in the low 300’s, too. I felt comfortable at those levels, but I felt uncomfortable below 150, and hypo below 120. It took a lot of adjustments with my diet and insulin dosages to feel good while having numbers in the low 100’s. Control was better, but still not so good, until synthetic insulin was available in the 1990’s. By the end of the 1990’s I was able to use carb counting, and a fast acting insulin. Then my numbers were so much better, and my A1c’s were in the high 5’s. I do not know how I survived for 50 years on animal insulin, and type 1, without having any serious diabetes complications.

Improvements in the 21’st century are happening so fast, perhaps too fast. There are several artificial pancreas devices being developed, and tested on human subjects. They depend on a CGM for measuring the blood glucose levels. I am using a Dexcom G4, and it is very accurate about 80% of the time, but there are times it is as much as 30 or 40 points different from my meter. If I was using an artificial pancreas device, and I had a false high, I could be given insulin that I did not need. That would give me a hypo, maybe one that would put me in the hospital! There are devices being developed that would give us insulin when we are too high, and glucagon when we are too low. We would not be making the decisions. The device would be making the decisions for us. I feel very uncomfortable with that. The almighty CGM is not accurate enough to have the device do the decision making. I know that the Dexcom G6 is being developed, and it will be more accurate, but I cannot believe that the accuracy will be good enough that we would never have occasional false highs and lows with that CGM. For that reason I will not use an artificial pancreas, even if one was given to me as a gift.

The “Guardian® sensor 3 continuous glucose monitoring sensor Introducing the most accurate sensor from Medtronic, now with up to 7 day wear and easy insertion. It is the FIRST and ONLY continuous glucose monitoring sensor FDA approved and trusted to control insulin dosing.” That one is being sold now, but I will not use it. I am chicken! I am eligible for a new pump in May, but I will be getting the Medtronic 630g. That one can be used without the CGM supplement. I will be wearing my Dexcom G4 CGM, and I will be making all of my own decisions!! I have made my own decisions for 71 years, and I have done very well. I want it to stay that way for the rest of my life. :slight_smile:

17 Likes
2

I agree with everything you said. That was a great post. I never dose off CGM. It’s incredibly sporadic.

@Richard157, I’ve been managing it for 45 years. Us old-timers gotta stick together Richard! :grinning:

5 Likes
3

I’ve “only” had Type 1 for 25 years, but I am looking forward to the AP systems. In the beginning, I am sure they will still allow some level of manual control until the systems before perfected (maybe being able to switch automatic control on or off, or being able to bolus insulin/glucagon manually in addition to the automatic adjustment). I imagine users will still need to test regularly to confirm the CGM readings. More accurate CGM sensors and faster-acting insulins and stable liquid glucagon on the horizon will make these systems better than what’s possible today.

I have never been someone who’s been able to hit a 5% A1c or who can flatline for days on end. So many things affect my blood sugar that I will be happy to offload some of that work onto an automatic system, especially overnight (I love my CGM, but I certainly don’t sleep as well as I used to). I get accurate CGM results most of the time, so am not too concerned about accuracy. I’m sure in the beginning there will still be highs and lows, but I already get highs and lows daily, so that won’t be much of a change. My pump has been out of warranty for over a year and is (literally!) being held together by tape, becaus I’m “saving” my next pump coverage for a semi-automatic system.

6 Likes
4

@Richard157 As usual, Richard, your wisdom of years of dealing with diabetes through several technological eons is sage. Thank you for posting this.

I too have been dealing with T1 for many years, before glucometers were available, and have grown to depend on my Dexcom. But you are so correct, it is only a tool that needs a human brain to interpret.

I look forward to your giving us your next dose of wisdom, and your next for years to come.

4 Likes
5

I cannot wait to get the 670G. I went to the 630G tow months ago, I am having excellent luck with the Medtronic sensor and I can not wait to get the new item. I have been at type 1 for about 43 years now and I am all about the tech.

6 Likes
6

Great post! +1 prior to glucometers here also.I am also about the tech now and l also am getting a fortnight from most Enlites and am really excited by medtronics tech at the moment. I think a sensor that reads venous glucose like the proposed optical BG sensors and faster acting insulins with reduced half life ( to mimic venous insulin ) would be ideal for completely closed loops. However using partial closed loops where the system does almost everything but asks the ‘user’ is also an exciting prospect,especially if commercially developed.

1 Like
7

I say bring it on! The less I have to weigh all the many variables what feels like 300 times a day, the less I have to do math calculations before I get to sit down and enjoy a meal or go for a bike ride, the less I have to be constantly, moment by moment all day and half the night aware that I have a chronic condition that needs constant monitoring, the better.

I, too, was T1 in the days before the first glucometer – the mid-70s big beige box that took three or four minutes to do a test. But then as now, with all our tech advances, we don’t really “make our own decisions.” Our meter or our CGM says our blood sugar is X, our pump says our dose should be X, and for the most part we believe it. In the “old” days we didn’t look at the Tes-Tape or the bubbling-up acetest tablet in the test tube and say, “It could be a false high. I’m not trusting that. I’ll make my own decisions.”

Early pacemakers sent one set, steady pulse to your heart. Modern pacemakers will monitor things like your breathing rate and blood temperature and automatically adjust your heart rate – no human intervention, no decision making, no doubting. Would one resist having such a miraculous life-saver because it might give a false reading, because it might send the wrong signal?

I’m going to assume – and call me naive – that all new technologies will be rigorously tested, and the chance of error reduced to the absolute possible minimum. (They already are, which is one reason it takes so long for these things to come to market.) Eradicating error altogether is unlikely. Your car’s on-board computer, too, is liable to error, and might slam on the brakes when it’s unwarranted, but that doesn’t prevent most people from driving.

Having started out a diabetic in the first-void-wait-30-minutes-test-again-scrub-the-test-tube days, when you visited your family doctor once every six months and he’d change your one daily dose of insulin by a unit or two and we’ll see how that goes, through to the early glucometers, to my first pump in 1980, to today’s Omnipod and CGM with the upside of tighter control and the downside of constant awareness, constant observation and intervention, I truly look forward to some kind of system that allows me to go just five minutes without thinking, “I have diabetes. I have to weigh that in the equation.” I truly look forward to some kind of system that Just Does It.

5 Likes
8

I do agree that I would never dose our son off of what the CGM says. Although it’s nice to have “an idea” of what his current BG is, anytime we’ve “corrected” with the assumption that that number was right…we ended up taking him low, or too high…it’s a gauge, but it’s certainly not accurate for any type of correction. Finger sticks are the only way and, until a pancreas can work out a means to get this right, I, like you, feel like it’s a dangerous technology (especially for someone my sons age where even small doses can mean the difference between life and death.)

2 Likes
9

@Richard157, I understand your lack of confidence in the current technology to make automatic dosing decisions using current CGM technology. I have now lived with the experimental open source hybrid artificial pancreas system, Loop, since November 14, 2016. I use it every day. I use the Dexcom G4 CGM.

This system sets temp basal rates above and below the pump’s programmed basal rate to control blood glucose. The user sets a max basal rate function and that sets an important barrier to keep the pump from over infusing. If the controlling app loses contact with the pump or CGM, the system fails to the settings programmed into the pump.

I have not had any serious lows below 54 mg/dL without being aware of it and treating as needed. I fingerstick twice per day and any time that a trending high CGM needs a correction beyond the temp basal rate that calls for a manual correction, I do a fingerstick first and will often correct with Afrezza, if needed. Sometimes I find that the CGM is reporting a false high and I calibrate the CGM if it’s > 20%.

What I’m trying to say is that if Loop determines a large insulin correction is needed, it is still limited by my max basal rate that I choose. If the CGM shows a trending high, whether actual or false, it is not capable of delivering that large dose. Correcting anticipated highs using a temp basal does not allow the pump to stronge-arm an aggressive trending high. My BG simply trends higher and alarms let me know a manual intervention may be needed.

My system alarms me and I decide, after a fingersick, if a correction is needed and how much it will be. Temp basaling has its limits of control and works best when all other factors are well tuned, including insulin to carb ratio, insulin sensitivity factor, and duration of insulin action. When the BG is finely balanced, the system gives nudges with a little bit more insulin and a little bit less insulin.

After six months of using the Loop, I’m here to say I do not want to go back to my previous pump treatment with the help of the temp basal algorithm. My control is better; my time in range is usually > 90%, my average BG is below 100 mg/dL, and my standard deviation is often in the low 20’s. And my time spent below a verified 54 mg/dL is less than 1%. I do have some false lows due to sensor inaccuracy but the system makes me aware with its alarms.

Here’s my trace from yesterday. I’m attending a conference, eating conference food, and dealing with all the social distractions. I did go high in the afternoon but I attribute that to my carb count being too low, a manual mistake. This shows you that the system is not set up to deliver unlimited amounts of basal to bring that high down by itself. I intervened with a fingerstick and an IM correction.

Screen Shot 2017-04-23 at 8.11.35 AM.png868×222 16.9 KB

These early systems are not for everyone, just as pumps are not for everyone. Later commercial systems will not be for everyone, either. But the safety built into the Loop gives me strong confidence to use it without fear. In fact I think my overall safety would go down if I had to stop using it. And its ability to consistently take me safely through the night is amazing.

8 Likes
10

Terry, thanks for this reply. It gives me more respect for the loop system. I have heard that the G5 CGM is more accurate. Will you be using that one in the near future?

2 Likes
11

I’ll likely switch to the G5 next January as I attempt to get on Medicare’s system for reimbursement. I’m glad you raised these issues. We’ll all need to work our way through them as the pace of D-tech quickens.

2 Likes
12

Are you still eating low-carb? Or is this even eating higher-carb food?

13

My son has had type 1 diabetes for 10 years and used Dexcom for seven of them. We’ve seen remarkable improvement in the accuracy of the CGM and we do indeed dose off of it and I have yet to find that backfire. We are cautious about doing that within the first day or two of a new sensor, but once we get comfortable, Dexcom is relied upon heavily.

I remember when the Artificial Pancreas idea started to gain traction and dexcom was not as accurate as it is now. I had the same reservations you’ve expressed. I no longer do.

My mindset is that I’m calibrating the CGM off of a meter whose accuracy is considered acceptable even if it’s 20% higher or lower than my son’s actual bg. To me dosing off of a current day CGM is not any different than dosing off a meter. If the result seems out of whack for either, we will challenge it, confirm as necessary and use our judgement.

Like Terry, my son has also been Looping for a couple of months. We are just starting to get into a groove after learning a new approach to managing blood sugar, but again, we have yet to have an issue bc of dexcom’s accuracy. We remain cautious that first day or two and ensure there are two calibrations a day and continue to use judgement. From that perspective it’s been smooth sailing with positive results.

And Caleb is not low carb. :slight_smile:

5 Likes
14

Great discussion here. Good strong arguments on each side of the question. Here’s my own personal 2¢.

I’m in Richard’s camp, at least until there is a practical CGM technology that tests blood directly and not at one remove (e.g., interstitial fluid), and possibly not even then. I have worked with computers my entire adult life and I know too well the fallibility of programming, even when done by the most expert and conscientious people. Airliners have crashed because of software bugs or design oversights. This is the real world and it happens.

What it really boils down to is the question, how much risk is acceptable risk?—an intensely personal choice that each of us must make for ourselves.

Here’s an analogy. I have lousy eyesight; I’ve been severely myopic since childhood. So why have I never had the laser surgery that could fix it? Very simple: that surgery has a great success rate and a very small rate of failure or complications. But though small, those rates are finite and real. And should I turn out to be one of the very few unlucky ones . . . well, I have only one pair of eyes. I can’t get another set.

Similarly for an artificial pancreas, even if it makes correct decisions 99.99% of the time, I’m not willing to be the one who gets a massive dose of insulin when I shouldn’t. Not if I can help it.

Again, this is a deeply personal judgement and everyone has to decide it for themselves.

Great discussion! :sunglasses:

1 Like
15

I can understand this concern. I’m a fairly skeptical person by nature. I think my level of comfort has come from experience. Have you used a CGM? I’m sincerely interested and I hope that doesn’t come off like I’m challenging your opinion, bc I truly respect it.

With a background in financial auditing and what I now consider years of testing this equipment, I’m comfortable with its accuracy within an acceptable margin of error, of which we remain aware, and have adequate internal controls in place to mitigate against the risks that exist. To me, there are inherent risks with diabetes and using Dexcom, without a doubt, is a control procedure which greatly protects us against some of those risks.

I find this all fascinating because, as I’ve said, I’m skeptical by nature and I’m generally risk averse. The reason we started Looping is because Caleb is traveling to Europe this summer without me. Knowing that Loop will be helping him, particularly while he sleeps, brings me great comfort bc I’m confident it will help him stay safe.

But, I also got Lasik surgery, so… :wink:

1 Like
16

@Richard157 thank you very much for bringing this important issue up!

Could not agree more. I’ll add a few more thoughts based on my personal experience. I’ve had T1D for 41+ years now, and I’ve lived through urine testing, animal insulins, etc. In Nov 2015, from a post on TuD no less, I became aware of a group of people who had hacked MM pumps and who have been working on building a closed loop system known as OpenAPS. My background is in engineering, and when I found that my MM523 pump was hackable, I just could not resist giving this a try. I was initially very skeptical for two reasons: first, just as @Richard157 and others are pointing here, I feared that CGM inaccuracy could lead to potentially dangerous errors. Second, by that time I was already doing pretty well “sugar surfing” manually, so I did not think the closed-system would be able to give me any tangible improvements or advantages. Turned out I was completely wrong on both counts. After spending entire Christmas holiday digging through how the system really works and how to put it together, I finally closed my DIY loop on January 1, 2016. I have run DiY closed-loop systems (OpenAPS first, then Loop) ever since. My results have been similar to what @Terry4 is reporting, and I have no intentions of turning back and exposing myself to increased risks associated with open-loop pumping or any other manual insulin delivery. .

I have two points to add regarding perceived risks related to CGM inaccuracy: (1) compared to many other variations, inaccuracies, and unknowns, including meter inaccuracies, variations in metabolism, insulin absorption, food absorption, and (above all, perhaps) many kinds of human errors, etc, a well calibrated CGM objectively presents relatively small risks even for manual dosing. (2) unlike humans, closed-loop systems do not ever deliver large amounts of insulin based on a single meter or CGM reading. Instead, they execute small corrections every 5 minutes based on a number of past readings. As it turns out, G4 or G5 CGM are plenty accurate for running such closed-loop systems.

Lets not kid ourselves, diabetes is risky. Injecting insulin externally in any form is risky and potentially dangerous, no matter how skilled or experienced you are. Yet, we expose ourselves to these risks many times daily, and with more or less success, because we have no other choice. Do closed-loop systems eliminate these risks? No, and that’s important to highlight and understand - we are not talking about cure here, not even close. But, most importantly, do closed-loop systems increase these risks? No! The opposite is true: closed-loop systems reduce the risks associated with taking insulin externally, and are likely to result in better bg control with less effort.

6 Likes
17

Your points are good ones. I’m impelled to comment on this one, though.[quote=“Dragan1, post:16, topic:61035”]
closed-loop systems do not ever deliver large amounts of insulin based on a single meter or CGM reading
[/quote]

True today. I’ll use the aviation analogy again. The early digital controls only interpreted pilots’ commands and could be ignored easily. As time went by and confidence grew, they were made more and more powerful and given control of additional functions. Nowadays, in the case of one manufacturer, they can actually override (read: ignore) pilot inputs and make decisions on their own. At least two and probably more of those accidents I alluded to occurred for precisely that reason: the pilot said, in effect, “do this” and the computer said, “No, I know better than you.” Hundreds of fatalities resulted.

The whole history of technology teaches that new things start out small and grow.

We’re all really saying the same thing: there is a continuum of acceptable risk and we each have to decide where on the continuum we’re willing to go. I have a somewhat cynical view of technology; perhaps it’s the old cliché about familiarity breeding contempt, I dunno. Everyone is free to make their own choices.

18

I have used injections for 61.5 years, and a pump for almost 10 years. I have a lot of scar tissue. My upper ab is so scarred that I cannot inject or use an infusion set there. There are spots on my lower ab and legs that are somewhat scarred, so I experience slower absorption there. When I change my infusion set, I watch my Dexcom G4 and determine the absorption levels of the insulin. Then, after a day of carefully recorded BG numbers, I alter the basal rates on my pump accordingly. This is necessary every time I change sets.
There is a recommendation in the Dexcom manual that suggests that we not place a sensor in an area that has scar tissue. I do not know exactly where my scarring is located, and exactly how much scarring there is at any given location. Maybe that is why I see my CGM and glucometer differ by 20 or 30 points at times. I had a 20+ difference this morning, and a 30+ difference this afternoon. The rest of the day has been much better, with differences that were less than 10 points.
It is these big differences that would make me scared to use a closed loop system. People who do NOT have scarring or any other problems that cause unpredictable differences between the CGM and glucometer would be much more likely to trust a closed loop system.

3 Likes
19

I have the same concerns as Richard. I’ve had T1 since 1961 so I appreciate all the new developments since then, I don’t know how I ended up with so few side effects & able to live the life I wanted when I had no idea what my BG was & using a fixed insulin dose. But I don’t thing CGM technology or Pumps, specifically infusion sets are ready yet. My BG control is so dependent on infusion set insertion. BG can be low several hours then high for several hours, when I change sets there is often a little bit of dried blood at the site (invisible under the set). So I assume that was the reason. How would an AP program cope with all the variations. I spent all my working life writing computer software, the program is only as good as the design, did the programmer allow for every possible permutation. I think the more hopeful developments are islet transplantation or smart insulins that work as needed.

2 Likes
20

Is this an answer ?