670G Auto Mode Insulin Sensitivity

Wouldn’t this whole system have been a lot better without the 150? It just seems like they packed a bunch of unneeded garbage into the algorithm.

@Jason99 a PID controller needs reasonable limits. Once a controlled value leaves a reference area PID doesn’t work well anymore. Also I guess it’s a safety concern as they’re trying to avoid lows and there is the risk of a good deal of user error involved. The next version will have auto corrections that are not user initiated but they need FDA approval for that first. IMO a predictive ML model would be much better or something similar to what OpenAPS does but they’re notoriously hard to get through FDA approval as they require a fair amount of technicality on the patient side to tweak tuning parameters. I wish they’d let us configure the targets I’d want to lower them by at least 20 mg/dL

2 Likes

I don’t mind the 120 target, it should be user selectable to a certain point. The correction seems to take forever, as @DrBB said, waking every morning with high BG is not appealing. This thing sounds like it needed more time to develop instead of being rushed out to the public half baked.

Yeah dawn phenomenon is tricky in automode unfortunately. I’m personally giving it some more time as it has worked fairly well for me but I will likely go back to manual mode as I can achieve same TIR with better targets. I used to have a 80 md/dL target so to me this whole automode correction stuff takes way too long. And I’m sure Medtronic won’t just give us a software update once their new approach is FDA approved. Kind of regret not going with the Tandem which is a lot more forward thinking in that regard (free updates, yay!)

Do you suspect that Tandem will run into the same issues with their algorithm as the 670?

I think every HCL system will have issues somewhere. The good news is that Tandem has had time to observe feedback to 670G and has a chance to do it better. I assume it will probably be closer to the oref0 specification of OpenAPS. However that is just speculation based on the fact that Tandem has a unique opportunity to stand out in the market by adopting and integrating with open source technology. The other thing that Tandem has going for it is that they can (and will) just ship it as an update to X2 users for free.

I love my X2. The integration with Dexcom and the ability to update the pump software was what sold me. I don’t think Medtronic is a bad company by any means. The 670 is just to complicated for me, and requires way too much babysitting. Thanks for all the info

1 Like

Yeah I wish I had gotten an X2 but I was in a time crunch as my old out of warranty accu check pump had failed and Medtronic were the first to contact me and get the process going. I didn’t have much time to do research or anything and my endo seemed to prefer the 670G. In retrospect I should’ve definitely taken the X2 as I was already on the Dexcom G6. At the end of the day I did make my peace with the 670G though and I’m making it work for me. I don’t have nearly as much issues as other people report and I don’t know if that’s because I’m a software engineer and mathematician and just intimately understand it better or if I’m just lucky :man_shrugging:t2:

Your success with the 670g is most likely a combination of all those things. Some users are having great success while others want to smash the pump with a hammer. I feel that there should be a happy medium somewhere in between. The 670 was the first iteration of this technology, it will evolve.

That was what interested me, too, but…

As was I, and Tandem took forever to get back to me after I contacted them, whereas MedT was up in my grille, offering to come to my house, meet me near my office etc.

Very much so, but as @mbensch suggests, the whole system is designed around fear of lows more than strict control (as also indicated in the quote I pulled about patients who won’t be well suited to the system), which is a pervasive attitude in the industry for mostly understandable reasons. I kinda think most commercial HCL systems are going to have the same bias in one form another for a whole host of reasons, not just FDA approval but also that they are aligned with medical culture generally on this point, and that culture has never been very comfortable with this whole thing of handing over a potentially dangerous drug (well, hormone) to patients who then have to determine their own doses and administer same multiple times a day. My experience with the 670 reflected this, and it was unexpected. But basically the philosophy is “take the uncertainty of dosing decisions as much as possible out of your hands and do it for you” (Yay!), which in practice translated as “Lock up the parameters that you really need to customize it to your own metabolism and abilities because you might hurt yourself.”

3 Likes

@DrBB to add to your las post: More recent research also shows that TIR and avoiding severe lows contributes much more to long term health and lowering the risk of complications than aggressively looking at the A1c like we are used to. I was always more comfortable risking lows for chasing that 5.x A1c but just being more steady in TIR feels much better than the rollercoaster ride from before.

2 Likes

@DrBB
As was I, and Tandem took forever to get back to me after I contacted them, whereas MedT was up in my grille, offering to come to my house, meet me near my office etc.

This is exactly what happened to @Dave44 a few weeks ago. I hope this is not a trend for Tandem

That is really the single most critical tweak they could make that would have kept me with it. There are users who struggle with too many lows as well as those of us for whom it just wasn’t aggressive enough. I don’t think it would have taken much of an adjustment in my case—maybe dropping the target to 110 might have been sufficient, and there are others who would struggle a lot less if they could raise it a bit. Because the target is, in a sense, just notional, like AIT and the other parameters that don’t carry over from the ones you used in your manual pump. It isn’t a perfectly objective and “real” number, it’s just a parameter that constrains the system to behave in a certain way, and when that meets the actual biology of the individual patient, it may need to be tweaked to get the actual result that the notional “120” is aimed at. Being able to adjust this one within a certain range to accommodate the needs of different patients with different metabolisms would be a huge step forward.

Also, @Jason99 Re DP and Tandem’s eventual HCL: a huge factor is Dexcom G6 vs Guardian G3. I found the Guardian about equivalent to my Dex G5 (I wore both simultaneously for a week or so), except with respect to overnight/pre-dawn values, and neither in my experience comes anywhere close to the G6 for reliable accuracy 24/7 (excepting initial 24 hrs which is something that can easily be ameliorated by “pre-soaking”). My problem was simply that the Guardian didn’t pick up my rising BGs toward dawn. What it didn’t “see,” the basal rate couldn’t account for. So I’d get up with my sensor reading, say, 120, but my fingerstick would show 180. It really was the sensor’s fault, much more than the HCL algorithm. I tried various things, including getting up at 3:30 a.m. and doing a calibration, but it was just ridiculous to have to resort to this kind of thing when the whole point of the system was to make it less onerous to operate, not give you all kinds of other crap to do to try to keep it on track.

  • If the Guardian had been anything like as accurate as the G6, I would probably still be using automode.
  • I think the decision to develop their own in-house CGM system was a mistake from the start, because they’re never going to be able to keep up against companies for whom CGM is their core business. I think interoperability is the future, and trying to freeze everyone into the MedT ecosystem is eventually going to prove short-sighted.
1 Like

Agree about TIR, yes. My TIR in auto was ok, but not as good as I was doing without it. I didn’t mean to imply that tolerating more lows would be a positive, but that the heightened concern about them drives a lot of the design decisions that made auto a difficult fit for me and I think for others who are/were already achieving good results.

@DrBB
I think the decision to develop their own in-house CGM system was a mistake from the start, because they’re never going to be able to keep up against companies for whom CGM is their core business. I think interoperability is the future, and trying to freeze everyone into the MedT ecosystem is eventually going to prove short-sighted.

Let’s be brutally honest, Medtronic have never made a sensor worth bragging about. The G3 is miles ahead of the Enlites, you were still experiencing a 60 point difference. If the pump is completely at the mercy of the sensor readings, the damn thing better be accurate.

I have the exact opposite experience. I was on the G6 for roughly a year and for me the Guardian is on par when it comes to accuracy and I don’t have any less sensor issues than I already had with the G6. I personally find them both to be mediocre when it comes to longevity and compared to my fingersticks I’m rarely off by much. The only thing that can happen (and it was the same for me on the G6) was false lows or lower reading due to external factors (compression lows, sensor lifted out of skin partially due to adhesive failure, sweating, poor hydration, etc.).

100% agree here. I think that for any technology those “walled gardens” are more of an achilles heel in the long run than something that helps business. Interoperability is the key in our modern technological society.

Exactly. And my experience with the G6 vs G5 is similar: my confidence in a HCL system based on the G6 would be significantly higher than one based on the G5, good as it is. I’ve pretty much stopped doing fingersticks, maybe one or two a day, since upgrading to the 6, whereas I never quite trusted the G5 to that extent and almost always used a fingerstick for bolusing.

I never quite trusted the G6 but it was my first ever CGM (after that weird wannabe Libre v1 that I only used for a few weeks and got so annoyed with it). I do have to say that I do calibrate and fingerstick more on the Guardian but it’s also because I’m much more involved with taking care of myself again. Hard for me to say if I had done it on the G6 as well if I hadn’t been in one of those pesky denial/ignore phases at the time.

1 Like

Yes, my understanding is that sensor systems can also be very sensitive to differing metabolisms, and I haven’t had any of the problems of longevity or compression effects that others have reported with the G6. So, lucky me! :wink:

Which is also why I’m intrigued by the fact that MedT has struck a deal with Tidepool around developing a G6-compatible pump. Dunno what will come of it, but just the fact that they’re looking at it is encouraging.

1 Like

I do a fingerstick at least once a day, to make sure the G6 is giving me accurate info. I’ve noticed that the G6 starts to go wonky around 190. My G6 at least will be off by 20-30 points once it reaches 190. Anything below that number it will be 2-6 points difference from my fingerstick. My G5 did not exhibit this behavior.