Thanks for the info @jredel, that’s interesting.
What were you using before the 670G?
I suppose you could do it the other way around. I don’t generally have overnight excursions to worry about other than that damn DP, which it wasn’t handling. But a lot of people do like it precisely for that feature of keeping you out of overnight lows.
1 Like
Yup. The alternative work-around is to give yourself a “shadow bolus”–i.e., tell it you’re having carbs when you aren’t and get it to give you a bolus that way. I’ve read that that can also negatively affect how the algorithm interprets things, though there seem to be two schools of thought on that.
1 Like
Well it wasn’t always in auto mode, which effectively made it a 630 for the first two weeks. 7 years ago I had an animas and then switched to pens for five years ago because of the nuisance with all the adhesive. But I went to 670g specifically for auto mode because I thought it would bring such a high quality of life. I don’t think waking up every night is really worth it.
3 Likes
Yeah I really didn’t want to trick the system because it feels like that would lead to a feedback loop of needing to do it every time I go high.
1 Like
@drbbennett, I was told today at the diabetes clinic that the algorithm in the 670G uses a moving window of 6 days, and throws away older data. Does this ring right to you?
Yes it does. It fits with the other thing I was told, which is that it takes six days to see results from changes in your settings. Which is part of what makes it so hard to tweak. Change one parameter, wait 6 days, see if it made any difference–and try to separate out that signal from all the usual noise in the T1 system, the vagaries of stress, diet, exercise and god knows what else. Eventually that was what drove me out of auto. I’d made changes that seemed to be slowly improving my DP, and then things started to slide back the other way again, and I couldn’t tell if it was something I’d done, or the algorithm or god knows what, should I tweak again and wait another six days or… At which point it occurred to me that I had been controlling it just fine in manual and what exactly was I getting out of auto in return for all this aggravation? And I decided the answer was, well, not very much. It took all my usual controls out of my hands and in return, it wasn’t doing as good a job as I’d been doing on my own.
4 Likes
Regarding “shadow bolusing” as you call it – what about doing it with a “fake” BG instead of “fake” carbs? When my sensor reading is high and doesn’t seem to be going down fast enough, I am tempted to just enter my sensor glucose as “BG” instead of actually testing my BG. Do you think this would “mess with the algorithm” too or is it OK to do this? They make such a big deal about BG and SG being “different things” but having come from using the Dexcom I am having a hard time adjusting my thinking to this. Curious if anyone else does this?
1 Like
I know I’m about a year late, but good instinct…
I won’t fake carb… I think it’s probably risky and might cause auto to deliver insulin based on a pattern that’s not real. In fact, I’m not even in auto any more, so I’m no longer doing any of it. 
About putting in fake BGs though…this I would do from time to time when I needed auto to accept a number. I would not just feed it my SG though… I’d “cross-calculate”, if you will. In a case where my SG is significantly delayed, I would calculate 35% of my SG and then do a finger stick. If that value was more than 35% outside of my SG, I would then calculate 20% of my BG. I’d look for whatever number that fell inside the overlap that was closest to my BG… the justification was that even if I weren’t using my exact BG, even by FDA guidelines, there’s a 20% window on our meters. So if there’s a number within 20% of the meter and still in the overlap, it’s not perfect, but it’s “close enough”. I wouldn’t calibrate though… calibrating with a “close enough” isn’t ideal.
And then my noisy, demanding little pump would be quiet again.
I know you were having this discussion a long time ago, but I’m very curious what the “two schools of thought” were on the shadow boluses?? If you can remember that far back??
Have you since? If not, any plans to? I keep telling my group I’m planning on another try, but what a drag…
In auto over night, you avoid hypoglycemia by not getting as much insulin as you normally do in manual. The reduction caused a rise that looks like DP. In manual over night, you get all of your insulin. “DP” solved.
Just that some people thought they were a bad idea because they distorted the data going to the algorithm, while others claimed that they didn’t make any difference. My feeling was that if I was going to use the auto mode, then I was going to use the damn auto mode and keep my data as clean as I could make it. GIGO, in other words. Because if I don’t want to have it do its thing what’s the point of it? I can always just run it in manual where I can give myself whatever bolus I want and it’s not a shadow-anything.
Nope. Actually, it’s worse: it has so many irritating habits even in manual (11 clicks to Suspend, including the one that says “Do you want to suspend?” after it took me ten clicks to get there) that I actually went back to my old pager-style Paradigm 730 or whatever the number is, and went back to my Dexcom CGM–basically reverted to what I was doing before switching to the 670G.
So right now it’s a doorstop, but the most recent development is that I started pumping FIasp as of a couple of weeks ago, and it’s really working great for me. Enough so that I’ve been wondering whether it might be fun to try it in the 670 on auto just to see. The faster onset might effectively make the algorithm more aggressive about spikes and corrections in spite of itself. But I dunno. Switching this stuff around ends up being non-trivial and I’m pretty ok with where I am.
3 Likes
I’m actually resentful at the idea of having to fake out my insulin pump for insulin. I’m with you 100% on that. Everything about this disease seems to run by the numbers… except auto, which seems to do best on luck, magic, faith, and deception.
2 Likes
That’s hilarious. 
What kind of difference do you see?? There are a couple of people in my group who have tried Fiasp in auto, and they didn’t like it. That’s not exactly enough to go on because there could be a million and one reasons WHY, but…
IF you ever do go back into auto, and we’re both still alive and using FUD, let me know. I’ll venture in with you. I just can’t find any motivation to do it otherwise…
Is auto mode worth it?
When I am in auto mode my BS run 30-50 points higher. Im surprised Medtronics does not make any statements regarding this mode.
2 Likes
I came across a series of articles from someone who tried to make the 670G work with auto mode. Despite the help from Medtronic, she eventually gave up after a rollercoaster of “It’s wonderful” / “Now I can’t stand it.” The articles have numerous tips on things to try when there’s trouble, and they give a good picture of the emotional impact that the troubles can have.
The emotional aspect turns out to be more important than I had recognized: there can be the simultaneous occurrence of technical indicators that show the system is working ok or at least not too bad, together with a user experience that is so frustrating and out of control that the system is perceived as completely intolerable. I’ve heard this here from drbbennett and nickyghaleb, but didn’t really grasp the full weight of the message until encountering this series of articles (I came across the link on Tu at https://forum.tudiabetes.org/t/i-am-so-done/68913.)
For hybrid closed-loop systems to become widely useful, it seems that we need improvements in sensor technology so that it is broadly trustworthy and accurate, not just in favorable or typical circumstances. Moreover, this must be combined with user interfaces that are much more advanced in managing how they effect the emotional component of the user experience. If the user ever gets a strong urge to rip the pump off and smash it on the floor, the designer of the user interface has failed badly, yet we see numerous reports of this kind of experience.
2 Likes