What are Your Thoughts on Closing the Loop?

Hello Tudiabetes Folks! My thougths are sceptical in nature, like way to many human variables aka “Diabetic Curve Balls” that any computer can possibly predict. I posted my take on http://www.three2treat.com, but I really want to know if you think it is a good solution?


Agreed. Too many factors. I have to think through every choice I make when taking insulin. I cant even explain it to a doctor, so how am I going to let a machine figure it out?

I’ll pass on the whole deal. I would still like a CGMS type deal without needles though.

Too bad the glucowatch was a failure.

Have to agree on variable like stress etc.
Sometimes my BG can drop 200 points from one unit, sometimes not. Sometimes food hits one way, sometimes it hits a totally different way.
Not ready to let anything mechanical make decisions for me.

Personally I think you’re wrong.

First of all the loop doesn’t have to be fully closed to be useful. An open-loop system where meals are announced to the system so initial bolus could be started would be fine, maybe even preferable since it could track BG closer to normal.

Second this is a miniaturized high-tech engineering problem, and this is an area that has seen real and dramatic success in recent history.

Third the components already exist. Pump - check. CGM - check. Micro portable computer (e.g. iPhone) - check. The power of each of these components is sufficient to handle the full task. There are two components missing. First is a robust algorithm to tie them together safely - but there are approaches that are being worked now that have already shown to be successful. Second is portable glucagon (also pumped) which appears to also allow tracking BG closer to normal - but some are working on systems without this component so maybe this component isn’t really required.

Beyond this, what is needed is regulatory approval. And the JDRF has recently commited to helping with this.

All in all, I think this is the best hope we have of getting a system in place that would allow T1’s a bit of relief from the 24/7 concern of where their BG is at. It will be interesting to see what A1c these systems will be able to achieve.

I think the concept of “closing the loop” oversimplifies things. How does a non-D body know when to release insulin? Certainly not based on what your BG (plus or minus 20%) was fifteen or twenty minutes ago. When does the insulin take effect? Not with a 15-30 minute peak, and a lag that could last four hours. A “closed-loop” system has to refine these two pieces before calculating how much insulin to deliver.

A closed loop, based on today’s measurement and delivery techniques, wouldn’t be a “set-it-and-forget-it” system, in my mind. It would put patients and their devices in a constant “reactive” mode. If, however, the first generation of closed-loop patients still bolused before meals, and still had a modest basal – proactively – and let the machine do all the fine-tuning, that would be a step in the right direction.

I am not a big fan of spending a lot of resources on closing the loop. To me it is useful to perceive changes and think about them, rather than letting the machines run things.

Hey there!
Thanks for the thoughts! Glucagon would be a great component to the closed loop system. It would be very cool!
As a Type 1 for 30 years, with 2 type 1 kids there is nothing I’d like more then to forget about the big D and let a device do the decision making for me!


Here is my complete Blog Post;

When I have idle time at work I use it to review any and all information related to Diabetes. Yesterday I stumbled across a site called www.diabeteshealth.com and was scrolling through the numerous summaries regarding new research, nutrition etc…

I reviewed the topic related to closing the loop, since a cure is too distant at the moment, perhaps the focus should be on designing a complex computer feedback system whereby a Continuous Glucose Sensor communicates with a microscopic mega computer that in turn uses fancy algorithms to communicate with the pump telling it what to do. Sounds pretty cool, except for a couple of things…

My issue and the researchers mainly in Europe feel the challenge is how the computer program will account for things like, spontaneous activity, stress, and the human bodies biochemistry, aka “Diabetic Curve Balls” that get thrown our way on a daily diabetic basis. I have to say I can’t see the system being too successful, just my humble opinion.

I know, I am a natural sceptic.

I think they need to focus on insulin analogues that respond to micro changes in blood glucose, like a non-diabetics where the minute the body senses a micro increase the beta cells send out the first responders. Now I am not a biochemist, nor really that smart, I get by, but this just seems logical Don’t know if its possible and I am sure they are working on it.

When I reflect on such things, I can’t help to marvel at how truly amazing the human body is designed, how complex our inner workings really are.

What do you all think about this?



That’ll be the day when I let a computer run my body totally. No, I want input ability at every phase. When I don’t have a brain …

How much time do you spend every day managing your diabetes - including this thinking? Include in this all the time you spend checking your CGM and thinking about what you ate and when you ate it to cause your BG to react the way it has, and exercising when you don’t really want to, and how hard you exercised and how much that is affecting your glycogen restocking and all the other attempts you make to flatten your BG curve. You seem to be pretty obsessive with your CGM, so I would guess it’s a lot of time. Wouldn’t you like to have that time back?

To me a working semi-closed loop artificial pancreas that could achieve a good A1c would be HUGE.

I agree with Jag1–this is probably the closest thing that T1 diabetics will see to a cure for a long time, and all the problems that people are identifying are basically math/engineering problems. Obviously the system has to allow some sort of manual override, and I think it will probably have lots of kinks early on, but 10 or 20 years from now I really see this as being the gold standard of care for T1 diabetics and look forward to having one. I am fine letting a computer run my endocrine system (which is not my whole body). It doesn’t have the last word, and I expect it will generally do a better job than analog insulin and whatever is left of my pancreas will do.

I don’t think I would ever trust a machine to be able to beat me consistently. I am always leading food w/ boluses by a bit to cut off post meal spiking. Sometimes as much as an hour. I don’t see anyone affiliated with the medical community programming a machine like that but, for DP a small CB left to it’s own works way better than CB + bolusing for carbs. I’d also be concerned about, at least the way my CGM works, that the interstitial fluid seems to run out of gas after long runs. I’ve had a couple of times where it seemed to read atrificially low for hours after multiple days in a row of long workouts. Obviously I couls stop but I’d rather be more engaged with managing diabetes than “spam in a can”, to continue my “Right Stuff” allusion.

A healthy pancreas naturally controls its secretion of insulin in response to stress, exercise, hormonal activity, etc. Those curve balls hit healthy people too. We already have devices that can secrete insulin (although mimicking the profile of insulin diffused through the portal vein is a challenge); the key is just to find an algorithm that can replicate the signaling functions that cause a healthy pancreas to secrete appropriate amounts of insulin. That’s obviously a big challenge, but fundamentally it’s a technical challenge and is well within our conceptual understanding.

I have sort of written off a cure. If it happens, great but I am not anticipating getting rid of this junk any time soon. I think there’s more metabolic wild cards, jokers and full houses that are floating around that are not going to be programmable and may not be regular?

Add me to the skeptical list. As much as I love having a Dexcom and the alerts that I get from it, it is not 100% accurate. I would not want a machine deciding to give me a correction bolus because it thinks I am 200 when I am really 60 nor would I want it cutting off my insulin because it thinks I am 60 when I am 200.

So imagine that you had a CGM that took 6000 readings per day (roughly every 15 seconds) and was actually accurate within, say, +/- 2 percent. How do your feelings change?

If you think this is unrealistic, remember that the first blood glucose monitoring systems were the size of a shoebox, took 5 minutes to use, and were wildly inaccurate. What’s to say that the technology couldn’t develop to the point of having a nearly-flawless and ultra-rapid CGM in the next decade or 2?

I think it would have to measure a bunch of stuff but a +/- 2% setup would be great!!

I would love to have that kind of CGMS but I would not want the pump side of it or the machine deciding what I need to do. I tried a pump for 8 months and besides the pump itself breaking twice, I had a lot of site problems. My body does not like having infusion sets in me and I easily form scar tissue, even with properly rotating sites. Pumping is not for me. I love my CGMS though.

The closed loop may not be the direction I would choose but it means a lot of brainpower and resources are being focused on improving our lives. We all know first hand how difficult it is to manage all the factors that affect bg and we suffer the consequences directly. I hope these efforts will cause a generation of researchers to better understand our balancing act and therefore they can design and implement better programming and technology. For example, it might inspire pump companies to finally implement the Super Bolus. I can just imagine researchers observing meal spikes and wondering how to tweak the algorithms and hitting upon the idea of borrowing from basal to knock down the spike and only later realizing this is not a new idea!

Sure, which is where an algorithm to model the action of the insulin analog you’re using comes in. Totally accurate? Probably not for a long time, but I’d rather have something to do some automatic steering than nothing. Your argument is akin to saying that cruise control is pointless because it can’t drive the car without human input. Of course it can’t, but it can do something to help keep BGs in range, and then individuals can do the fine-tuning on their own or manually override the system if need be. Besides, even if research into a closed-loop system isn’t going to turn out, it’s not money completely wasted, because the research is likely to result in better pumps and better CGMs, even if it doesn’t figure out a way to close the loop between the two.

The only way I see this working is if someone develops a 99.99% accurate CGM that is implanted in the body in such a way that it is measuring whole blood BG’s every ten seconds or so – not interstitial fluid and not peripheral capillary blood. Think of a pacemaker-quality thingy implanted in the body with a probe in a major vein.

My doctor told me last year that someone on earth is working on this; it gives me hope.

If we could wear something like a super-comfortable, water-proof, shock-proof wrist-watch that was giving us up to the minute readings that were lab quality – so we could ride herd on the pump – and a pump could also receive these readings via telemetry and handle the insulin with little to no human intervention except in cases of malfunction (bad infusion sites, kink in the system), we’d be in business.

Our liver can respond by dumping out more glucose as needed – what it cannot do is shut off the flow of insulin from a pump. We need a FAST and ACCURATE CGM device to do that before any of us could truly rely on a closed loop system.