Artificial pancreas trial

[azure]

Thanks @tim2000s - it explains very well

The system I read about a while ago only used a TBR, which I didn't like the sound of. I'd want a faster solution if I was high.

 

[tim2000s]

Thanks @tim2000s - it explains very well

The system I read about a while ago only used a TBR, which I didn't like the sound of. I'd want a faster solution if I was high.

The limit isn't really whether you use a TBR or a Microbolus. It's that the insulin is too slow, and the CGMs update every five minutes. What you're trying to balance is safety versus effectiveness.

For example, my dexcom sensor lost signal overnight (one of the issues with these systems, whoever makes them). It re-established connection at about 4.40am, it detected my glucose level at 13.8 and took action by delivering a temp basal of around 3.95U/hr, then 3.18U/hr which persisted for 15 mins, giving about 1u. It then provided slightly higher TBR for the next hour or so. The key aim is to drop you safely without you going Hypo, knowing what the insulin action curve looks like and what your insulin sensitivity is. The result was that I was back down in single figures in just over an hour.

The difference between 1U at 4.40 and 1U over fifteen minutes after 4.40 is actually quite limited in effects, in terms of rate of drop, but until you've run with one of these systems, you don't realise how little that 15 mins makes a difference. That's what I mean by needing faster insulins. What you also have to take into account is the max that will be delivered in one hit.

In the open source systems, Using the TBR model, we set a new TBR every five minutes, on the basis that in the worst case, the TBR will last for 30 mins and only deliver as much insulin as we can handle with fast carbs. The upper limit is controlled by the user, but shouldn't be set higher than the previous statement.

Likewise, in the microbolus model, there will be a maximum bolus that can be applied safely, and a max IOB that the system won't exceed in order to allow recovery in the event of something going amiss. Realistically, a microbolus model is going to operate in a way that is not too dissimilar to that of the TBR approach, in that you are more likely to have multiple smaller microboluses over a time period rather than one big one when higher, simply because it's a safer way to deliver the insulin and to monitor the outcome of the reaction.

Interestingly, the open source systems tell you what IOB looks like, including basals and boluses. From what I've heard about the Medtronic 670G it doesn't tell you this information when it lowers basal, so you don't have a full view on what your real IOB is.

 

[iHs]

Sadly, a fast acting insulin with short duration action hasn't materialised as yet. Biodel in the US was on brink of developing Viaject but it failed along the way. Novolog/Novorapid has got its problems due to its slow onset with longer duration but weirdly is a very common insulin used in pumps. I hope that Lilly will launch a faster version of Humalog

 

[azure]

Fantastic explanation and example there @tim2000s That demonstrates how it works perfectly.

I suppose for me my underlying concerns are safety/trust and the fact that in order to be safe, any system might not be able to aim for as strict control as I do. I also worry about the fallibility of some of the components eg the CGM as detailed above.

I suppose I'll only know if it would,be an option I'd consider if I had a go with it. Like many things, a trial is the best proof of usefulness to the individual.

I was initially very keen on the closed loop as it was mentioned to me when I first got my pump years ago, but now, having used a pump all this time, I'm more sceptical about how well any system could do.

 

[tim2000s]

I was initially very keen on the closed loop as it was mentioned to me when I first got my pump years ago, but now, having used a pump all this time, I'm more sceptical about how well any system could do.

That's an interesting statement. And one that goes to the heart of the closed loop and open source question. The commercial closed loops have to limit the amount of variability on the parameters that the user can affect so that they are considered safe for anyone to use. Expert users (who really would be expert patients) will be frustrated by them. There are ways around this that you can use to adjust the factors you do have at your disposal, but I think you'd struggle to get an Hba1C below 6.5% with what's currently available without fudging them.

The open source systems are different because you have that level of interaction and can manage them differently, specifying behaviour more finely. As a result, I get the same results from using OpenAPS as Sugar Surfing with CGM, but importantly, I have to put so much less effort in to get there (high is greater than 10, low is below 3.9, as per the trial comparison and those are affected by food more than anything else).

 

[diabetic0312]

I am using a Medtronic 640g pump

 

[tim2000s]

I am using a Medtronic 640g pump

Interesting - have they told you who built the artificial pancreas system they are using?

 

[azure]

Very interesting Yes, that confirms it for me really. My HbA1C has always been in the 5s apart from one in the 6s soon after diagnosis, and I wouldn't want that to be any higher.

The idea of putting in less effort is very appealing, but for now I'll be going on with my own micro-managing, I guess. I am interested to hear about the open source systems though, as a possible future option, so thank you

 

[tim2000s]

Also, @diabetic0312, would you be able to give us a screenshot of the software on the phone? I'd be really interested to see it!

 

[slip]

Watching with interest!

 

[diabetic0312]

Yes I will post some photos later when I have a chance!

 

[tim2000s]

Yes I will post some photos later when I have a chance!

Cool. Thanks. Another question. They gave you the phone, I presume? How locked down is it? Can you use it as a normal phone or is it just the app? If it works as a normal phone, do you mind if I PM you?

 

[diabetic0312]

It is simply one app, you can't even go off the app

 

[tim2000s]

Oh well makes it harder for participants to hack!

 

[AndyS]

I dunno, if it is just software then there is bound to be a way to reverse engineer it unless they are locking it down with some form of trusted certification in the pump and app perhaps.

@diabetic0312 How did you get on the trial? Are you treated out of Addenbrookes? I know they have done plenty of work in the hospital and had heard there was supposed to be something starting up (also a side study on the super fast novorapid) but nothing since.
Seems I don't qualify for anything they are running there as am now too far down the line for DX and don't have (or qualify) for a pump.

/A

 

[tim2000s]

I dunno, if it is just software then there is bound to be a way to reverse engineer it unless they are locking it down with some form of trusted certification in the pump and app perhaps.

@diabetic0312 How did you get on the trial? Are you treated out of Addenbrookes? I know they have done plenty of work in the hospital and had heard there was supposed to be something starting up (also a side study on the super fast novorapid) but nothing since.
Seems I don't qualify for anything they are running there as am now too far down the line for DX and don't have (or qualify) for a pump.

/A

I was having a chat off line outside the forum with someone. The interesting part to this is the communication protocol the pump uses. It's a modified form of Zigbee, which means that normal Bluetooth wouldn't work, so with they've modified the pump or the phone. So even if you could reverse engineer the app, there's still something else that needs to be resolved.

 

[AndyS]

I was having a chat off line outside the forum with someone. The interesting part to this is the communication protocol the pump uses. It's a modified form of Zigbee, which means that normal Bluetooth wouldn't work, so with they've modified the pump or the phone. So even if you could reverse engineer the app, there's still something else that needs to be resolved.

Ah ok, that's pretty interesting in itself, I guess the obvious would be to go messing about with an additional box to add a ZigBee transceiver unit but you are rapidly heading down the line of OpenAPS anyways and from your adventures in that space I would almost say if you need to go to that effort then go your route since there is far more customisation for the expert patient

/A

 

[diabetic0312]

So the red dotted line is my blood glucose, the blue dashed line is my normal basal without the artificial pancreas and the solid blue line is my basal I'm getting now I'm using the artificial pancreas. I know it looks confusing so ask away

 

[diabetic0312]

 

[donnellysdogs]

By the looks of the blue dotted line were you on the standard instruction to have 4 or 5 blocks of rates a day? As it looks like the Artificial Pancreas is doing differently??