Predicting failure date of medtronic cgms sof sensors

for the longest time I couldn't understand how some sensors last 5 days and some will last 11 days and some will last 9 days.

I want to know when they are going to fail to be prepared,for example if I have a big presentation next Tuesday should I change the sensor on Sunday or will this one go past Tuesday? every extra day I can squeeze out of a sensor is money I don't have to spend (especially when insurance won't pay for them).

so being the geek I am,I began logging calibration data to look for patterns. when the pump asked for a calibration I wrote down the date & time,BG value, sensor isig, then divided the isig by the BG to get what I'm going to call the sensor health ratio.after a year of recording data I started graphing the data and the pattern becomes apparent

the sensor generally fails when the linear (average or sometimes called best fit) line crosses 1.0 on the y axis

if anyone wants to check if they get the same results with their pump I can give you an excel file.

wow that is really cool! :)

Sorry, but I don't follow. What are the X and Y axis in your graphs? Is this three different sample sensors, or a summary of your year's worth of data?

yes each graph is a different sensor

the horizontal axis is time in days, the vertical axis is the number i got after dividing the isig (read off the pump immediately after entering the bg value) by the bg everytime the pump asked for calibration(every 12 hours)

9:06pm BG 7.7 ,pump isig 19.46 ratio = 19.46/7.7 = 2.53
9:13am BG 10.7 ,pump isig 24.36 ratio = 24.36/10.7 = 2.28
9:20pm BG 13.7 ,pump isig 29.74 ratio = 29.74/13.7 = 2.17

the main thing that the graphs shows is the ratio of isig to bg continuously drops from the time you insert the sensor until the time the sensor dies. so after 2days based on the slope of the best fit line you can extend the line to see where it crosses 1.0 on the y axis and the x value is the number of days after insertion that sensor will last

on the first graph the best fit line would have crossed y=1.0 at 14 days and that sensor made it to day 13.9 which is approximately 14 days

on the second graph the best fit line crosses y=1.0 at 6.1 days and that sensor didn't make it to day 7

on the third graph the best fit line crosses y=1.0 at 6.9 days and that sensor didn't make it to day 7

I can also be geeky, but what surprises me are your results versus my extended sensor experience. I put in a new sensor every Thursday, following protocols. I restart that one as a "new sensor" on Sunday and again on Wednesday, changing again on Thursday. I do not go longer than 7 days on a sensor.

Works like a charm and the only sensor failures I have ever seen are at insert, and an occasional CAL ERROR, usually because I ate something stupid. Sometimes BGs are off from senor readings, but if I ride it out it will readjust--it sometimes takes a while for the sensor to adjust to the actuality of the blood glucose.

Could you define what makes you think a sensor failed?

I've never felt I got good data after 6 days w/ either the Sof Sensors nor the Enlites, although I am too impatient with the whole take off and recharge the battery thing so maybe I need to play with it more. Lately, they've been running out the battery around 5.5 days, which is also annoying. I'm thinking that it may be time for a new xmitter but I'm suspecting that Medtronic will be squidgy about sending me another one and will try to leverage me to the new 530 pump instead of the 523 I have, so I'm being lazy and not calling them. Although it's running season so I sort of have to be careful not to have a no sensor long run....

There are two factors related to sensor failure, one is the signal strength which is related to how long the wound canal stays wet (when the calibration ratio drops below a certain level then the pump rejects the sensor), and the second is how long the anti rejection chemical fools the body before it attacks the sensor and spoils the ionization process causing sensor failure. You can just look at the sensor data sheet and watch the sensor Isig degrade over time but I doubt you can actually determine with any accuracy when a sensor will reach it's last day. My suggestion would be to pick a time frame that gives reliable service and stick with that schedule.....JMHO

here's what happens when I combine the graphs for the past 13 sensors leaving only the best fit lines,can you see a pattern developing?

10 of the 13 have very similar slopes (sensor rate of decay),if it was random the slopes would all be different.

I think the variance in that group is due to a margin of error

For us Dexcom users, what's isig?

oh I hate having to recharge,I get it stuck down nicely with a couple iv3000s (transparent adhesive bandaid type things) then a few days later I have to rip it all off and recharge. why can't it recharge wirelessly through magnetic induction (see: duracell powermats) by holding a charger up to it for 5 minutes without having to take it off.

xerox, you have shown a correlation between your created number (best fit #) and the point at which it crosses the y line. But you still don't know what factor(s) cause inconsistency of the failure my opinion of course. What causes the # to cross that line? Mainly, I think it is because it is Medtronic technology rather than Dexcom! My Dex sensor has never lasted less than 14 days (I "restart" at day 7 for the second week). I change at day 14 even though I could get more days. Regardless, I am totally impressed with your due diligence data collection and innovative research.

Dave, I have looked at isig explanations on other sites many times and I still don't get it! It is an additional data set that MM cgm folks have to deal with that us Dex folks don't need to worry about. We calibrate and that is it. But, perhaps someone will give a simpler explanation than I have seen or more likely, you can understand the full explanation. So glad I don't have to deal with it.

I had to look that up:

The sensor measures electrical Current (in nano-amps) created by changes in your interstitial fluid and it takes an average of these readings using a smoothing algorithm and transmits this information to your pump every five minutes these average values are called Isig.

so I'd imagine your cgms system does the same thing they just might call it something different.

yeah I'm still working on the why part of the variation,I've been recording lot numbers and serial numbers to see if there's a correlation in that.

I totally get what you're saying to just get into a schedule and keep it,when I had insurance that covered them I would change sensors every sunday. now that it's coming out of my pocket I want to stretch them as far as they will go.

most people with medtronic cgms never look at the isig,but it's available for people that do want to see some technical information buried 3 screens deep at the very bottom of the screen. same as some people have never looked under the hood of their car, it works or it doesn't,that's all they want to know.

Translation: isig is the average current flowing through the sensor during the last measurement interval.

Makes sense. The G4 does work the same way, but does not expose this datum to the user. We have no way to get to it directly.

Your post inspired me to try something similar with the G4. I'm recording CGM reading, calibration value (from meter), and date/time for each calibration.

I plan to plot the error (measured BG - CGM value) per time since calibration, vs. date. Theory being that the rate of increase in error vs. time increases as the sensor ages, and (hopefully) exhibits a knee curve at the point of failure. I'll post my results (in a few months, that's how long it'll take to use three sensors).

I'm very intrigued by what Xerox is try to do, however -- find a consistent pattern to sensor data that can predict failure -- not just show after the fact that, yeah, when you replaced it, measured data confirms it wasn't working any more.

The range I get out of G4 sensors is pretty variable -- between 14 and 18-20 days, skewed to the lower end of that. I'd love to have some way to predict that my sensor, still giving solid, accurate readings, is going to go kaput in another 2-3 days.

Inspired, I'm going to try and figger somethin' like this out fer us, dagnabbit!

Your assumption is wrong…many of us long time users have watched our Isig.
and calibration factor…the soft sensor Manuel even said what the high and low
calibration ratios where and when the sensor would fail based on the ratios.

I have been using my Dexcom for about 2 years now. (I received one of the first G4s when they started shipping in the US) They have a fixed 7 day time. I did have to replace the sensor once when I got it caught taking off some clothes, but Dexcom replaced it without charge.

I also used the sof sensors for several years. MM actually define a parameter the Calibration Factor where CF = BG(mg/dl)/isig. There have been numerous posts on different forums over the years explaining how to use CF to assess sensor performance.

I also live somewhere where BG is measured in mmol/L and converting this into mg/dl and then doing the arithmetic was too much of a hassle. Like you I worked out that you could use an alternative, which is what you define as Sensor Health Ration, where SHR = isig/BG(mmol/L). In fact the two are related by the equation SHR = 18/CF.

I used to routinely work out SHR to see how my sensor was performing.

HOWEVER.... I used the opportunity to upgrade my pump to switch from MM to an Animas Vibe. My motivation at the time was financial. I have to pay for my own sensors, and given an average life of ~18 days for the Dexcom G4 sensors compared to around 8 days for the MM sensors, the G4 sensors work out at less than half the annual cost. The Dexcom sensors are also much much more accurate and reliable, so you don't need any of the faffing around and babying that you needed to do to get a half-way accurate reading from the MM sensors.


I'll be in the market for a new cgms system soon,my transmitter is almost 4 years old (I was told their expected lifespan is only 2 years) and it's getting harder and harder to recharge.

I don't suppose pump companies operate like cellphone companies where they give you a good deal to stay with them when your contract is up or alternatively give you a good deal to switch to a different company?

how do the dexcom sensors costs compare to the sof sensors ($47.50CAN or about $43.31Usd each)?