I think I am finally nailing down how the sensors work. Most of this information has been posted here, and some of it I got other places and personal experience. I've never seen it compiled in one place, so I think I'll give it a shot. Felix over on IP gave me the idea for this. Please note that this is not intended for brand-new users.

Sensor glucose (SG) is calculated by multiplying the ISIG by the calibration factor. The calibration factor is calculated by dividing the blood glucose you enter to calibrate with by the ISIG. The software actually uses the average of the last 4 calibration factors it calculates, I believe, to determine the SG it reports.

Thus, understanding the ISIG is fairly important in making the CGM work right. As Felix explained, "I" is the international symbol for electrical current. The 'SIG' either stands for 'signal' or is an acronym for 'signal [for] interstitial glucose'. Either way, I think it is important – not to beginning users, but those that have used it for a few weeks and are starting to get a feel for it, or even experienced users who have had problems.

There are a few things that affect the ISIG values. When the reagents (enzymes?) in the sensor react with the glucose in the interstitial fluid, a current is created due to voltage being sent by the transmitter. That current signal is measured in nanoamps. You can see this on the Carelink data table report.

Thus, the voltage that the transmitter sends to the sensor must be consistent, and strong. So, it is important to make sure the transmitter battery is charged enough. I haven't heard of a transmitter battery lasting more than about 18 months, but I'm taking a couple of (hopefully) proactive steps to increase my transmitter battery's life (the transmitter battery cannot be replaced without getting a new transmitter – about $600).

First, I am changing the battery on the transmitter charger every three months. I remember panicking when I got the flashing red light on my transmitter when my charger battery died - after six months. I won't go that long again. I think I probably wore the transmitter battery down a little by charging it with a AAA battery in the charger that was wearing out.

Next, I am letting my transmitter charge overnight at least once a month. I am fairly certain that a Medtronic electrical engineer would tell me I am wasting my time – that when those lights quit flashing, the transmitter battery is charged as much as it will ever be. Still, I figure it can't hurt, and I just have an intuitive feeling about this.

The next important factor for the ISIG is the amount of interstitial fluid in the sensor cannula. This has been discussed at length here. There are a number of successful users who insert their sensors the night before attaching the transmitter the next morning to get them completely wet. I don't have that problem, and I'm glad, because every few Sundays, I get a morning when I have neither infusion set nor sensor poked in me. Gosh, I love those showers.

I think this is the problem Dave (Spike) had with his sensors (where's he been, anyway?). Given his history of cannula kinking, I think his sensor cannulas just did not get enough interstitial fluid. Thus, his numerous SG low alarms at night - which were, of course, bogus. This theory is supported by the fact that his first sensor worked great for him – he just got lucky and found a site that didn't kink the cannula.

Finally, the more sugar in the fluid, the more the reagents get used up. That creates a higher current (and thus the ISIG), and that causes the sensors to get used up more quickly. This has also been discussed here. So, the longer the sensors are used, the greater the likelihood of them not giving a good ISIG (and thus a rotten SG).

Also, when the BG's remain high for a while, the same is true. If your BG's remain over 200 for a while, your sensors won't last as long. I've always changed my sensor every week, but I started leaving them in for 2 weeks. I've somehow been under extremely good glycemic control lately, so I've not had a problem with this. I'm fairly sure I can get 3-4 weeks out of a sensor, but I can't stand the scars after just 2 weeks.

So, anyway, those ISIG values are closely tied in with the calibration factors. I've found that if your calibration factors are pretty stable, you'll have really good luck with the sensors. If, on the other hand, you have one calibration factor at 5 and another at 12, you're almost doomed to not get good SG readings. I've looked at this fairly closely, and determined that my ideal calibration factor is somewhere between 6-6.5. Of course, everyone is going to be different.

It shouldn't matter if the BG is 60 or 200. If you calibrate when the calibration factor is close to the last three you did, you should do well. I sure wish MM would document the calibration factor and its importance. MM put a chapter in the back of their 'Sensor Features User Guide' that discusses sensor performance and calibration stability that I just now re-read. I still don't completely understand it. It mainly discusses the accuracy while hypoglycemic, and the importance of NOT calibrating too much. I still don't get that.

If you're still reading this, you're either extremely interested in this subject, or you're REALLY bored. Either way, I appreciate your taking the time to make it this far. If I've said anything wrong, please post it here immediately. I've spoken fairly authoritatively here, and I'd hate for someone to rely on something I've said that's just plain wrong. Final note: I am neither a health care professional nor anywhere close to this business: I'm just a successful user who thinks he has finally started to figure this out. I love this forum.


Tom

Tom,

I have one thing to add - from the perspective of your post its not that important. Since you are summarizing a lot for sensor information, it should also be noted that when the transmitter sends an isig value to the pump, it's not sending the value that it just read. What the sensor does it over a five minute period, it determines an isig reading once every minute. About every five minutes, the average of these values is sent to the pump. It is this average that is used to determine the SG reading.

Your post is a great - a comprehensive summary of what the sensor is doing.

Great post, Tom.

Like others have said, excellent post. What about the second sentence in your fourth bullet which starts with "When the reagents (enzymes?) in the sensor react with the glucose..."

Could you point me to the source please? Before I got my graduate degree and started work in my current field, I spent the first 10 years working in a medical research laboratory. Yet again, you've piqued my curiosity. What I can't figure out I'm sure my husband (chemical engineer) or father (electrical engineer) can assist me.

Thanks again Tom. It is nice to see everything compiled into a single post.

Hi Susan,

It came from a frequent poster on IP - Felix. Maybe you've never seen that board, but I won't post about it here. He is EXTREMELY knowledgeable. I think he knows more about DM than most endos. His general science knowledge is quite keen.

I may be breaking somebody's rules here, but here's what he posted:

"The sensor measures glucose concentration by applying a voltage
across a little cavity filled with fluid (blood) that has an
enzyme dissolved in it (also supplied by the sensor). The enzyme
converts glucose and produces some free ions in the process, which
make the fluid more conductive to electricity. The voltage applied
therefore results in a higher current flowing if the reaction is
more intense, and lower if it is less intense. This current is the
"Isig" ..."

I knew that somehow there were reagants in the cannula that created that current, but he describes it a lot better than I could ever hope to. What he wrote makes absolutely perfect sense to me, and I'd be shocked if he is not correct.

The only thing I would disagree with him about (it is extremely dangerous to disagree with Felix) is that the fluid really is not blood. It IS bloody, but I'm surprised he did not refer to it correctly as interstitial fluid. It really does make a huge difference.

I will send you the entire email.
Tom

Here's the trick with the calibration factors. The pump tries to use the last four calibrations to predict how the calibration factor will change over time. As a simplified example, say that your first calibration is 100mg/dl at an ISIG of 12.0 for a calibration factor of 8.33 (100/12). 4 hours later, you calibrate 100mg/dl at an ISIG of 11.50 for a calibration factor of 8.70 (100/11.5). The pump then assumes that the calibration factor is increasing at a rate of 0.0925 per hour. As you do additional calibrations, it refines the estimate of change.

Once it has 4 or more calibrations, it has a pretty good idea of where the CF should be at any time and will resist changing this too much. That's why it will sometimes seem to ignore a calibration that's too far off of where it's currently at. The calibration adjustment is being biased toward it's internal estimate. On the other hand, I believe that it will affect the "curve" so it effectively shows up later.

This also shows why you don't want to calibrate too often, since the process works from the last 4 calibrations only. If you do 4 calibrations in 4 hours, the algorithm is working with samples from only 4 hours instead of 24-48 hours.

Is this at all clear? I'm not sure how clear I've made it.

BTW, I don't know this for a fact but it matches what I've observed and makes sense to me as an engineer.

Ted,

This absolutely makes a ton of sense. I think you are right on the money, and I sure do appreciate the insight. I knew it had to be more complicated than just a straight average, but I didn't know what their algorithm was.

If you're not right, you SHOULD be, because this is just way too logical.

Tom

Thanks for your e-mails Tom. I answered back. If you feel there is anything of value in there, please feel free to go ahead and post or let me know and I can.

Ted, you did a wonderful job, and yes, you are correct based on some mathematical analyses that I've run on my own values.

Tom,

I'm going to refer to my earlier post where I asked for the resource regarding the statement "When the reagents (enzymes?) in the sensor react with the glucose...".

I have followed up on this and the reaction is definitely not biochemical (reagents) but electrical. My sources tell me that the internal sensor probe is probably copper. I inquired as to whether or not it could be a possible copper/polymer combination because of the flexibilty of the probe. Although that was not ruled out, I was told that copper itself could indeed be that flexible. The same sources also confirmed our suspicion that the transmitter provides the nanoamps for the Isig readings.

Two things of note here:
1. The limiting factor is the battery life of the transmitter not the "reagents" of the sensor. This is corraborated by the fact that the same sensor has been used effectively for at least 21 days per other users posts.

2. Copper is used in glucose meter test strips for an electrical reaction rather than a biochemical one. One caveat--I'm not familiar with all test strip types, but I believe this to be true as a biochemical reaction would need to be measured by a colorimetric method (light wavelength) similar to the earlier versions of glucose meters. Can you say "blot and flip the strip?"

Susan,

Thanks for the interesting post. I had forgotten that there was a thread similar to this that was started by Cara and last updated 6/7/2008 - "Maybe a weird CGMS question... ". Aaron posted a fascinating link that unfortunately I cannot understand. Dave (Spike) also posted a link to the patent office.

I got to poking around through the patents and I did find something I think I can understand.


It may not be a biochemical reaction, but it is certainly an electrochemical reaction. The abstract for the patent (5798031) I found is:

"The present invention concerns an electrochemical sensor made up of an insulating base having an electrode layer on its surface and a lid of deformable material which comprises a concave area in the central portion thereof, so that when it is mated with the base, the lid and base form a capillary space containing the electrode layer. When the electrode layer is in operative contact with a reaction layer comprising an enzyme which will cause the production of mobile electrons when contacted with a suitable analyte, the concentration of analyte, e.g. glucose in blood, can be measured by measuring the current created by the flow of mobile electrons when contacted with a suitable analyte, the concentration of analyte, e.g. glucose in blood, can be measured by measuring the current created by the flow of mobile electrons. "


What I highlighted in boldface is almost exactly what Felix described in my previous post.

With all due respect, I think that the sensor's life is limited by the amount of enzyme remaining in the sensor. If not, the sensors would last a lot longer than the 21 days you cite.

I'm not certain - I can't understand much of the patent past the abstract, but it does go on to say that the "electron acceptor is a ferricyanide" - not copper.

One of the interesting things about this patent is that it was assigned to Bayer. However, it is referenced by a Medtronics patent for the physical design of the sensor. I looked at that patent, and it just had drawings that were not in the patent. So, I'm not entirely certain that this patent is what the current MM sensor does, but I'd be shocked if it's not.

Once again, many thanks for reviving this. I actually have better things to do than research this, but this absolutely fascinates me. I'm going to go back and try to read the reference Aaron cited. I think I'll just get a bigger headache. It's been 35 years since I took a biochemistry class.

Tom

Thanks Tom for the research. We didn't get around to actually performing an analysis of the probe itself, but the ferricyanide deal makes perfect sense.

I'll share this with others while I'm still here at the symposium. It's made for great cocktail discussions.

I am almost sure that the enzyme being used is glucose oxidase (GOx). When it is under voltage and in contact with glucose and oxygen, it somehow becomes hydrogen peroxide, and being unstable resolves itself to H2O2 --> 2H +O2 + 2e. I don't know how to add sub- and super-scripts, or I'd be more precise. Thus the electrons are separated, and the current created is measurable. This is discussed as reactions 1,2 and 3 in the article Aaron cited (I don't know how he found that - but it is fascinating - and gives me a raging headache - and I understand maybe 5% of it).

One of the things that I don't understand from all this reading is how the oxygen comes into play. I'm fairly certain that we don't just have free oxygen molecules floating around, so I have to think that the oxygen comes from hemoglobin. However, I've never seen any reference to anything about anemic patients. I don't get it.

This all reminds me of when I was in college and thinking how ignorant I was, and that feeling is extremely reinforced now. I sure wish I hadn't thrown out all my chemistry texts.

I suspect you and I are the only ones interested in this level of detail, but it is really interesting. I wish I was at your symposium, but I'm sure I would feel stupider than a bullfrog. I hope you had fun (and learned a lot!).

Tom

I have had ISIG numbers between 22 anc 40 here lately, kinda concerning. The sensor is still relativelu close to what my sugars are running, so I'm not sure what this means.

Any help is appreciated.

Thanks

Jessica

ISIG numbers between 22 and 40 are normal when the sensor is working properly. These are nothing to be concerned about. I've usually found when the isig number drops below about 5, this is a sign of a failing sensor. Read carefully through the thread regarding how to calculate the calibration factor. The ISIG number by itself is meaningless unless you understand what you are looking at.

First of all, thanks for the post, I realy enjoyed reading it and found it very intresting.
What i want to know is, why dont the people from medtronic tell us all this stuff. I have had my sensor for two weeks now, and i have had some inconsistencies which i dont like. like having the pump tell me that my sugar is 10.2 but when i check the levels are sitting at 14.9. (and i am taking into acount the 15 min delay that the sensor has)
I want to now, what is the best way to determin what my calibration factor should be? cause sometimes my ISIG is sitting in the high 50's and then sometimes it is in the low 10's

I remember the feelings of outrage and betrayal when I was new to the sensor and it would say "everything is fine" and I would discover that actually everything is not at all fine and I need glucose right now (or insulin if high). So I would try to correct the erroneous sensor by calibrating at that time, which I now believe is counterproductive.

My best suggestion is only calibrate if your blood sugar is normal. I find that if I calibrate when I am high, then the sensor can't reliably detect lows, and if I calibrate when I am low, then the sensor can't reliably detect highs. To be concrete, for me, normal is 80-120 mg/dL. If I calibrate when I am 170 mg/dL (high), then I may find later that the sensor says I am at 92 (ok), when actually I am at 62 (too low). And if I were to calibrate then, with my blood glucose at 62 to try to "correct" the erroneous 92, then I may find later that the sensor says I am at 118 when actually I am at 148. By calibrating only when my blood sugar is normal, the qualitative advice from the sensor (too low, too high, or OK) is generally trustworthy even if the numbers aren't spot-on.

I do calibrate when I'm higher than the sensor says. I find that if I don't, the sensor will continue to read low. In the middle of the night recently my alarm actually woke me with a high reading. The sensor said I was 141 but my meter was 227, I think. I did calibrate, did the correction and went back to sleep. By morning the sensor & meter were both very close. I know the sensor has a hard time catching highs soon enough when I sleep so my High alert is set for 128 overnight.

Most times it would be impossible to wait for my BG to get back to normal. Sure, doing a correction in the middle of the night and waiting until morning to calibrate wouldn't be bad but if I'm 20 minutes away from eating dinner, I don't want to wait hours to correct a high, wait for my BG to come down and make sure I don't have much IOB left. I want to calibrate, correct, wait half an hour or so and then eat. I have long days and spent over 2 1/2 hours a day commuting to & from work. I don't have the time or patience to wait for perfect readings. I am very happy overall with the sensors and performance, so it works for me.

Edited to add: Sometimes if my BG and sensor are off by a lot, I'll just restart the sensor. I got a calibration error the other day because my sensor was over 150 and trending up, but my BG was in the 70s/80s (tested with two meters to make sure). After the error, I restarted it, calibrated and it was fine. The sensor was brand new, just started that morning, and was still being a little finicky.

Hi, Liz. Thanks for offering the other side of the story. Here's my take on it. If the sensor is calibrated too low and we recalibrate it with a higher value, the sensor calibration moves in the direction of the new number, but not all the way to meet that new number. So sometimes it improves the accuracy of the calibration, but sometimes it can overshoot too far and end up erroneous in the other direction. I've had both kinds of outcomes, including the one that you relate where it ends up about perfect. But I think my approach has the property that it will move the sensor calibration in the direction of more accuracy (possibly not far enough) without overshooting, and this may be helpful especially for beginners.


If I'm significantly high before dinner, I'd wait to calibrate. In part, this is because when I've been high for a while I tend to develop some insulin resistance so it can take an unusually large insulin correction to get into a good range. So I'd skip calibration, take an oversized correction bolus, and wait 45 minutes or an hour to see what is happening. In the good case, my sensor glucose will start dropping quickly, so I will count the IOB (or what my 723 calls "active insulin") toward reducing my meal bolus, and eat. If my sensor glucose doesn't start dropping, or worse, if it continues to rise, then I'll take another oversized correction, intentionally "stacking" to build up enough active insulin to force the blood glucose down. I don't worry about the stacking of corrections, because when my blood glucose __starts__ to fall, I will catch it by eating the meal promptly so that I don't go too low. As you point out, this means that I won't be calibrating for a few hours. And that means I'd better be prepared to rely on watching the isig if I miss the 12-hour calibration deadline and Medtronic goes draconian on me and turns off the sensor glucose display.

Do you find, as I do, that with experience we learn when to trust the sensor glucose value, and when to mentally adjust it ("it says I'm 145 up 5, so probably I'm actually around 175 or 185"), and when to disregard it entirely? ("That sensor reading can't be right. Better take a finger stick and try to get this sensor calibrated better.")


I came to a similar practice (night high 125) for a slightly different reason. Early on with the sensor I set the sleeping hi value at 165. But whenever it triggered, I saw that the graph had started at a normal value and was just climbing steadily until it hit the alert threshold. My blood glucose was just going to keep rising until it hit whatever Hi Limit I set. If it's going to wake me up anyway, I'd rather solve the problem before it gets out of hand, because then I can take a smaller correction, which I find to be more predictable. And I don't spend those extra hours driving my A1c up.


Great tip. Starting over can be much faster than waiting for several calibrations to finally get the sensor reading right.

This whole discussion just shows again that there are lots of different ways to get a good result, and that it's not the same for everybody. And to the beginners, I'd like to say that you may initially have disappointments and frustrations because sometimes the sensor is __just wrong__, but with with time it should get better ("Well, I guess I can work with this...."), and experienced sensor users generally agree that it is quite a useful tool.

Liz and bkh (sorry I don't know your name) have covered a lot of ground, and don't follow the same procedures.

To make it simple for people who are new to the sensors, there are a few guidellines that a very experienced MM tech gave to me:

Try to calibrate when your BG is STABLE. For most people, this does not have to be NORMAL. You can tell if it is stable by watching the SG (the sensor glucose number) for 2 or 3 cycles (it changes every 5 minutes). If it stays within 5 or the most 10 points total during those 15 minutes, then your BG is relatively stable. It should be fine to calibrate then, even if it is high or low.

What throws it off is if you calibrate when your BG is changing rapidly. If it is climbing or falling quickly, since it takes 15 minutes to calibrate, by the time it is finished 'digesting' the number, the number it has will be very wrong.

Try not to calibrate if you have taken a bolus within the past 2 hours.

Don't wait for the system to ask for a meter BG, calibrate when you know your BG has been stable for awhile. This can be between meals or at meals or at bed or during the night, it doesn't matter.

Hope this helps,
Linda

... but if you are having trouble with accuracy, you could try calibrating only when your BG is normal, to see if this helps you. It made a huge difference in my case. Linda points out that most people may not need to do this --- just keep it in reserve as something to try if needed.

Thanks alot for the tips. i am sure i will master the sensor soon. what fustrated me is the fact that it feels, for the price we pay for the sensor, it should have been more accurate. but i guess this is only my third sensor that i have used, so you could still call me a unexperienced. lol

I also noticed that the copper wire wasnt straight when i took it out last night, it almost look like a (S). so i guess that also affected the accuracy.
I will keep you all posted on how it is going.

I've found that when the sensor is really bent out of shape, such as an S shape like you described, it doesn't work well. I guess maybe it hits something during insertion and bends it. I now use my legs for most of my sensors and find that they never bend there. In my arms they do sometimes get bent. A slight curve doesn't affect it, but when it bends at sharp angles they're usually sensors that never worked that great.

I think the sensor wire escapes out the side of the needle during insertion and gets bent into different odd
shapes and sometimes this will damage one of the three very tiny electrodes.

The sensors that have the slight downward curve probably stayed in the groove during insertion and have a
tendency to curve because the gel softens up on the exposed side of the plastic tube.

I think the soft sensor has is a design flaw and the new smaller 90° sensor will be more consistent during insertion...JMHO

Okey, i took out the last sensor yesterday and the sight i used, worked much beter this time. the wire wasnt bent and the results were much better. i also found that for the calibration to work the best for me, i had to get my BG at 6.5-6.8 and the ISIG at roughly 40. i hope the sensor that i started today, wil work even better.. Thanks for all the help. i will keep everyone posted on what works for me, and what doesnt work.

i do have one other qustion, how do i control the ISIG readings?

You cannot control the ISIG Readings. They are the raw data the sensor sends to the system. They will vary with different sites, and even within the same site over a periiod of time. They usually start out higher and gradually decrease until they are too low for a calibration to be effective.

It is not important for the ISIG to be the same number all the time. You said yours was 40 - when my BG is normal, mine is never more than 20! That is why calibration is necessary. An ISIG number means NOTHING without the correlation to YOUR BG at that moment in time.

Linda B.

I'm not sure that bending is always a bad thing. My longest lasting sensor ever (34 days) was shaped like a fish hook when it came out. My theory was that the bent shape helped it stay firmly in place. Of course, breaking part of the sensor during insertion would be bad, as John says, but virtually all of my sensors are bent and I have very good luck with accuracy and longevity.

Marty