How do you handle off sensor readings in "auto mode"?

Diabetes Technology Commercial Closed Loop Systems
24

I’m an outlier here- an engineer with 40 years experience with T1D. I’m skeptical about complex technology reliability but more skeptical about anything that is simple technologically being a larger cause of application failures than human factors.

This tech is less complex than a cellphone, simpler than a programmable thermostat and a heat pump system. Instead of keeping the indoor temperature in a comfortable range by cycling between chill/idle/heat, it tries to keep glucose within a comfortable range by cycling a insulin pump on and off. It’s smaller with a glitzier UI, but it’s not new technology, just repackaging.

A CGM sensor is a wire with a coating that uses the same kind of electrochemical reaction as a blood glucose meter. The “transmitter” is a one or two chip integrated AD converter and BT transmitter. The probability of either being defective as shipped is much less than 1%. Both are simple enough to be tested 100% before shipping, and are. They can be damaged during transit by extreme, prolonged heat or cold

In comparison, the probability of a skin site everr being bruised, bleeding, or scarred in the interstitial layer or an inserted wire being moved within the interstial layer after insertion is much greater. Insulin flow is a larger molecule than glucose. Except for being slower, insulin flow in the interstitial layer to capillaries for basal delivery is nearly identical to glucose flow from capillaries to sensor wires. Glucose has the easier job.

I’ve done MDI +3x/day on average for 44 years. I know that I have injection sites that work poorly and I automatically avoided them or used longer or shorter needles - until I started using a CGM and pump and temporarly forgot what I learned. Once I remembered, I was able to start systematically mapping out my good sites, using this chart. https://usercontent.one/wp/www.bubblan.org/wp-content/uploads/2021/08/ijectionsites.png?media=1670715774

I had a much higher failure rate with plastic cannulas than with steel even though I utilized secondary strain relief. The steel cannulas eliminated one variable. If I had a “bad infusion site” I could move the cannula to the last known good site. Invariably the relocated system worked normally. Nearly every time I’ve used a known good infusion site for CGM sensor placement, it has been problem-free.

My experience suggests that the plastic cannula is more likely to cause or encounter underskin trauma than steel, and the thinner sensor wire has an easier time penetrating to an area clear of fluid resistance than either cannula.

I’ve used every kind of tool available for managing BG, delaying the adoption of each until I saw solid evidence that it not only “worked”, but was cost effective AND improved outcomes. AFAIC if a tech doesn’t increase my lifespan and make my QOL better, it’s worthless. Only one new tech has proven to be measureably effective in improving outcomes since 1980. It’s not “better” insulin, not pumps. It’s CGMs.

An insulin pump has an order of magnitude more mechancal complexity than a CGM and two orders of magnitude more electronic complexity. So I spent time more testing a Tandem pump manually, with Basal IQ and Control IQ software, than I did CGM sensors. The one I have is as reliable as my cellphone. All it lacks, imo, is a simple way to do a factory reset. The closest equivalent is to apply a software update. The next closest is like resetting a browser - disable the IQ software, create a new profile, power it off and on, and re-enable the IQ software.

My conclusion is that if I want excellent BG control, I test every site, infusion and CGM, for proper operation, and I look at to the readings before mealtime, neither a malfunctioning CGM nor the pump can prevent me from maintaining excellent control.

Knowing how to manually manage my BG based on glucose readings, but not doing it, will.

3 Likes
26

It should become more accurate as time goes on.

27

@Sam_I_am1

I don’t think that you received a specific answer to YOUR question.

Yes, even though I am lucky enough NOT to need to do this, a number of people insert a new sensor into their arm, abdomen, or other favorite location and let it sit there for 12-24 hours without a transmitter attached. I believe many people us the term “soaking” or even “marinating” during which time the sensor is getting settled it, biochemically, with your body chemistry. Then, when the in-use sensor stops at the end of their (hopefully) 10 day run, they move the transmitter to the new, hopefully acclimated, sensor and then start the sensor with it without a code as is their wont. If this is timed properly, there is no more than the normal 2 hours of warm up during which you have no CGM data. Of course, at the end of 10 days, that sensir will have actually been attached to you for 10 1/2 to 11 days … but if it results in more reliable data during the first 24 hours of data collection, most would consider that a big win.

Best of luck,

John

4 Likes
28

So, the G6 generally works fabulous for me. I’m soft and drink like a fish, and that seems to be what Dexcom loves best. I’ve also experimented with enough sensor sites to know where I get the best data. My arms, chest (arms and chest locations are all right around the front side of the armpit), and thighs give way better readings than my abdomen.

In my experience, if the sensor is reading off, it can generally be fixed, unless it’s at the end of it’s life. First step is making sure the sensor isn’t squished. Second step is always to drink some water. Seriously, the sensors read interstitial FLUID. There’s not enough fluid circulating for the sensor to read it you’re dehydrated. If it’s not either of those, I find the G6 responds really well to calibration, provided you only attempt them at the appropriate time: when your BG is trending steady/flat, when you’re in range, and when you’re not expecting an imminent change (right after eating, bolusing, existing, etc…), because those are the only times you should ever expect your fingerstick and CGM to read similarly. And if it really is a botched sensor, Dexcom has a fairly liberal replacement policy. They’ll replace it for tons of reasons with no questions asked if you fill out the web form, including big fingerstick discrepancies and 3 hours total of no data in a 24 hour period. Basically, if the sensor isn’t providing dependable data for my insulin dosing, I don’t wear it.

I also LOVE a restarted sensor, except the first hour or so of data, which reads really high. I simply calibrate the sensor in small increments (generally no more than 40 mg/dl at a time) until it reads true. I probably SHOULD turn Control-IQ off during that time, but I never remember to and it’s never been a problem. This is probably because I run my X2 w/ Control-IQ in “sleep mode” all the time. This targets a much lower range than the default programming, but sacrifices automatic correction boluses. It can only treat highs, real or misreported, with small bits of extra insulin given as a basal adjustment. It’s like trying to treat a high with an extended bolus. It can do it, but will take hours to do so. Sleep mode works it’s magic by helping to prevent the high in the first place, but isn’t very powerful once you’re already up there.

As an example: I try really hard to be at or below 100 mg/dl when I restart a sensor, because it makes it much easier to calibrate after the warmup. If my BG is 100, then the first restarted data points will read 180-220ish usually, let’s say 200 for this example. Control-IQ will decide to increase my basal to fix this. Instead of 1 u/h, it would ramp up to 1.3ish u/h. I immediately calibrate it down to 160, but that takes up to 10 minutes for the pump and CGM to talk back and forth. Since the pump delivers 1/10 of your basal rate every 6 minutes, I got 1 basal dose with an extra boost in those first 10 minutes until the calibration took hold. That would equal 0.13 units instead of 0.1. I’m just not worried about 0.03 extra units. You can enter a new calibration every 15 minutes, so at the 15 minute mark I’m going to calibrate it down from 160 to 120, but that 120 doesn’t fully register until the 25 minute mark. The pump has now delivered 3 basal doses of 0.13 units instead of 0.1, for a total of 0.09 extra units of insulin. 0.09 units of extra insulin just isn’t going to rock my world any. Remember, that’s spread out over a good 3+ hours of action time, too, and Control-IQ will be making further adjustments from better data in that timeframe. At this point I’m going to fingerstick (I never do for the first calibrations because I KNOW it’s high and 40 pts is the most I dare change it, there’s just no point to waste a strip) and enter a true calibration, and the system is back in alignment with me.

If the discrepancy were bigger, then the inappropriate correction would also be bigger, but you can see how a few bad readings won’t have much affect. It’s only a problem if you let the system run for an extended period of time with bad data.

1 Like
29

Both Omnipod 5 and Tandem dose based on predictive algorithms, not the current data point and trend arrow associated with it. Part of the algorithm includes statistical analysis to see if the data is trustworthy. It’s looking out for nonsense data. I don’t know what the O5 looks like in those circumstances, but the T:slim will display “- - -” when it doesn’t have data dependable enough to dose from, and revert to your programmed settings during that time.

4 Likes
30

My wife and I laugh over “great minds think alike” because actually it’s more true of mediocre ones. The smartest people I know don’t agree about very much that isn’t etablished, proven factual, but are always looking for a loophole or a new angle. Their easily accepted ideas are often proven wrong and their greatest ideas sound like fiction and fantasy until they prove that they are right.

2 Likes
32

I looked up the complete phrase. It negates any value of the first observation, making it useless for discriminating great minds from fools.

That fits well with my interpretation if it being an unintentional joke, but I’m language-impaired. I interpret everything literally, speak and write the same way. It gets harder for me every day to avoid misunderstanding or being misunderstood.

As I hear the phrase normally spoken, as a self-congradulatory exclamation, it’s not ironic or sacastic, just silly and untrue. I suppose that it could be used a sneer or with intonation sarcastically at different degrees, but not when used as a simple written quote.

Considering the implosion of langauge here, I wouldn’t be surprised if the current American translation of the phrase was, “Like, duh…” or an emoji.

2 Likes
34

I couldn’t find the original context of the expression. Without knowing what a person making the statement believes, the circumstances, or intonation, any declaration could be ironic.