• Guest - w'd love to know what you think about the forum! Take the 2025 Survey »

Advice regarding discrepancies between Libre 2 sensor readings and blood glucose monitor readings

Introverted_And_Proud

Well-Known Member
Messages
52
Type of diabetes
Type 1
Treatment type
Insulin
Dislikes
Warm weather, insects
Hello there. First time poster on this forum, so I apologise in advance if I've put this post in the wrong place.

I'm seeking some advice from fellow diabetics who are Type 1 and use the Libre 2 sensors, as I've recently encountered an issue that I'm unsure about.

For context, I've been Type 1 for over six years now. Up until last year, I used a blood glucose monitor to keep track of my glucose levels. Due to issues I was having with my background insulin at the time (I was prescribed Levemir until the end of February 2024), the consultants at my diabetes clinic suggested I try the Libre 2 so I could get a better picture of what my blood sugars were doing. I was a little sceptical at first, but overtime, I came to really like the sensor and how convenient and helpful it's been in managing my diabetes, especially this year when I was switched to Tresiba (still having a lot of teething problems with it at the moment). Over the past year, I was heavily relying on the sensor to keep track of my levels and adjusting my basal and bolus insulins where appropriate, only doing a finger-prick test if I felt hypo-ish or the Libre app alerted me I was going low. But as I wasn't having any serious problems with my injections and the finger-prick readings were usually within 1mmol of the sensor readings (I was already aware that the sensor readings and the blood glucose readings weren't going to be exactly the same and would vary), I was under the impression that there wasn't any issues. However, that impression changed at the beginning of last month.

I encountered a sensor that, with only one day left of its life, started giving me inaccurate readings, claiming my glucose levels were lower than they actually were, the discrepancy being between 3-4mmol. I frequently conduct control solution tests with my blood glucose monitor, so I knew the high readings they were reporting were accurate. I changed my sensor over to a new one as soon as I could, but I'll admit the experience did shake me up a little. So, since then, I became a bit more proactive in conducting finger-prick tests and comparing their results to the sensors', just so I could get a better idea of how long my sensors were actually lasting from the moment of activation.

Over the past month, I've noticed that quite a few of my sensors have started out fairly close to the finger-prick readings, staying under 1mmol difference. But upon reaching four to three days left of their lifespan, that discrepancy increases between 1.5-2mmol, with the sensor primarily reporting my levels to be higher than they actually are. Naturally, this has left me a little anxious. Tresiba has been causing my levels to fluctuate a lot ever since I started using it, so I've relied on the Daily Patterns graph to try to tweak my dosage per the advice of a DSN. But because of the recent discrepancies, I'm worried about how accurate my readings have actually been and I'm unsure of how to proceed. It doesn't help that the consultants at my diabetes clinic are also very reliant on the actual figures (they never bother to look at the reports or graphs, just the numbers) and I've gotten into a couple of debates with them regarding my chosen Tresiba dose. They're under the belief that my levels are too high and I'm not injecting enough, refusing to believe me when I've told them that the Tresiba primarily causes my levels to go low a couple of hours after specific meals and things because 'it isn't in the data' (actual words from the mouth of one consultant I saw a few months ago).

So, here's my question: when it comes down to the discrepancy differences between the Libre 2 sensor readings and the blood glucose readings, what range is considered 'accurate' when it comes to the sensor readings? Until now, I thought that anything above 1mmol is too much. Though, I have seen others on this forum say that 1.5mmol is considered 'normal' for them. Is it that the discrepancy is dependent on the individual, or that there is a set number where anything above it is inaccurate? I just want to be sure that I'm not wasting sensors by constantly changing them if the discrepancy falls between, say, 1.5-2mmol, for instance. Any advice would be greatly appreciated.
 
Hello and welcome @Introverted_And_Proud (love the name I am also introverted)

I've previously been on the FL1 and FL2, I am currently using the Guardian 4 and honestly would never go back to the FL2. Whilst I am sure it works great for some it just wasn't what worked for me. From what I've heard a CGM should be within 15% either way of a meter value so depending on what your blood sugar is at the time it could vary greatly. Obviously, you've also got to consider the fact that the values can differ greatly if your blood sugar is rapidly changing as a CGM can be up to 15 minutes behind an actual blood sugar reading. The higher your sugar generally the less accurate the readings are and you can just as easily also get compression lows depending on where the sensor is and what you are doing. Personally, I wouldn't be overly bothered by a 1-2mmol difference unless I was below 5 or above 10 but my target range is 3.9-10 and that's on hybrid closed loop. For me the FL2 was just never accurate (we're talking 4+ mmol difference between that and a finger prick) and I much prefer a CGM where I can calibrate it if it is slightly out but I know that isn't an option for everyone.

I never personally did this but have heard of people inserting their sensors 24 hours before they need to to let the CGM "settle in" before using it as the first 24 hours can be inaccurate, the same also goes though when the sensor is coming to the end of its life it can read more inaccurately. From reading your post I would also consider your options for a different consultant or even a diabetic nurse who is more versed in CGM's and can look at the data with you as it can be very valuable and helped me greatly that my consultant and nurse both look through mine during every visit.

Although calibrations are technically not required on the Guardian 4 I still do a blood sugar at the beginning and the end of the day just to make sure things are where they should be, the only downside for you with a FL2 is if they aren't you can't calibrate your sensor.

Edited to add info.
 
Introverted_And_Proud said:
So, here's my question: when it comes down to the discrepancy differences between the Libre 2 sensor readings and the blood glucose readings, what range is considered 'accurate' when it comes to the sensor readings?
Click to expand...
Neither your home finger prick meter or your libre should be considered absolutely accurate, but it seems to be generally accepted that the finger prick readings are more reliable. The Libre is calibrated for the 'average' person and some people find it accurate and others don't and normally the only way to be sure how accurate it is for you is to compare your predicted HbA1c from the libre with the actual HbA1c from a lab.

As far as I know a meter/sensor is considered accurate (or accurate enough to sell) if 95% of it's readings are

Within ± 0.83 mmol/L of laboratory results at concentrations of under 5.6 mmol/L
Within 15% of laboratory results at concentrations over 5.6 mmol/L

So it is perfectly possible for your true reading to be 10 and your meter say 8.5 and your libre say 11.5 (or vice-versa) and both would be considered 'accurate'. Also means if your true reading is 4, your meter may say 4.8 and the libre say 3.2.

But all this is theory and not very helpful or practical and I know a few people who assume the finger-prick reading is accurate and don't like the differences and use an alternative app (like Xdrip) which allows them to calibrate the figures from libre. (Of course this assumes that the difference is always consistent)

Whether the difference is acceptable depends on what you use the libre for. If you are lucky enough for your BS to spend 95% in a range between 4 and 5, then a difference of 1 or 2 may be unacceptable. But if like me you want to know when your BS is going high then I am not all that concerned if it says 12 when it is actually 10. In both cases I would do something to bring it down.
 
Not Type 1 I'm afraid though have been doing some learning in this area as I'm also interested in the accuracy of CGMs.

'Accurate' is a relative term in this context. There are basic standards for the accuracy of finger-prick meters and CGMs but the range of values that is considered acceptably 'accurate' is quite wide.

On the finger-prick meter side the specific device model you use to compare with the CGM graphs matters. This is a study from 2022 where 18 meters were compared with two different types of lab testing machine for accuracy - Link The most accurate meter tested was the 'Contour Next One', which was shown to give readings that were within 10% of both lab testing machine measurements 99.5% of the time. That meter had a relative bias of +3.3% or -1.1%, depending on which lab machine you compare it to. That means that if you were to do hundreds of readings they would average out at a level of accuracy that falls within the range that separates two lab-grade machines. Pretty good. The least accurate meter in that study gave values that were within 10% of objective accuracy only 71% or 51% of the time, depending on which lab machine you compare it to, and the most biased, the 'Freestyle Freedom Lite' had a relative bias of -6.6% to -11.1%. If you're using a meter that is either inaccurate a lot of the time or heavily biased then it's not ideal for gauging the accuracy of a CGM. Is someone were comparing CGM readings to the 'Freedom Lite' for example the CGM would always appear to be giving high readings even if the CGM was actually more accurate. The best meters available at the moment, for which study data is available and which are available in the UK are the 'Contour Next One' and the 'Contour Next'. The higher-end Accu-Chek meters have also done very well in studies in the past but the specific models that are currently available in the UK have not appeared in any independent studies that I am aware of.

Abbott will replace a Libre 2 as faulty if it gives very inaccurate readings compared to a good finger prick meter. I've read that Abbott will replace if finger-prick and CGM readings differ by more than 20% at least three times when BG levels are stable - flat. If it's less than 20% that is considered good enough as far as Abbott are concerned, apparently.

In my own case I'm currently wearing a Libre 2 that is off (high) every day by around 1 mmol/L when the CGM is giving readings at around the 5 mmol/L mark, and off by around 2 mmol/L (high) when the CGM is giving readings of around 10 mmol/L, compared to a Contour Next. This is right on the margin of inaccuracy that might get me a free replacement, but I can't be bothered. I do a finger-prick reading a few times per day when levels are steady and that enables me to 'eyeball' the graph and see what's really happening. I don't inject insulin so objective accuracy in real time isn't critically important to me. To compensate for these levels of inaccuracy there are 3rd party apps that enable a person to calibrate the data from a Libre 2 sensor. The xDrip+ app (which is used with a data analysis tool called Nightscout) is the most sophisticated. This app 'intercepts' the signal from the Libre 2 and provides minute by minute data points on the graph instead of the 15 minute averages that appear on the official Libre 2 app. There is a reason the Libre 2 system uses averages on the graph though - minute-by-minute readings tend to be 'noisy' - a reading at any given moment can be quite inaccurate and it's the trend that is most informative. The official Libre 2 app and sensor together detect 'outlier' data points and delete them from the graph and so give a fairly consistent indicator of past trends. Not accurate exactly, but consistent.

Using xDrip+ and Nightscout, calibrating xDrip+ properly to account for sensor inaccuracy on any given day (using a good meter) and using that system to make decisions that relate to your health requires significant learning to be sure you're doing it right. The official Libre 2 app is an approved medical device while xDrip+ is a DIY software project. Sophisticated, but not approved for use as a medical tool. Someone who is very experienced with this system might be able to provide information on how to learn how to use it safely. Depending on how accurate the Libre 2 typically is for you it might be good enough for your purposes to take finger-prick readings a few times per day and note them in the Libre 2 app. You'll be able to put the Libre readings into context that day, and these notes will then appear on the blood glucose report than can be generated from the LibreView website. When I wake up every morning for example I prick my finger, scan the sensor, and add a note to the log entry for that scan reading with something like 'Waking. Libre 5.0. C Next 4.1' or whatever. This enables me to look back at the blood glucose report and see indicators of how inaccurate the sensor was on any given day. That's good enough for my purposes and it's as good as you can get with the official Libre 2 app.

Again I'm not a Type 1 and do not have great experience with these devices so someone else might have better information and advice.
 
Thank you for responding @Pipp and @Nicola M . I appreciate the help.

@Nicola M I do remember reading somewhere about the 15% range, but maths in general isn't my strong suit. So whenever I did try to make sure the sensor reading was 15% off from my BG, I just kept making myself confused. Lol. I have tried to make sure that I conduct my finger-prick tests whenever the LibreLink app's trending arrow is horizontal. But in some cases shortly after, it looks like I get what I suspect are 'phantom spikes' or 'phantom drops', where my levels on the sensor randomly start rising or falling for a few minutes and then start going the other way to 'catch up' to the BG reading (though, it never really does. The discrepancy is still quite a bit out). I have tried to finger-prick test again after fifteen minutes to account for the lag, but again, it's rare the two results get within 1mmol of each other.

Compression lows aren't usually a problem for me, because I am a still sleeper and never sleep on the arm that my sensor is on. But as I've said, the Tresiba I've been prescribed has been causing my levels to fluctuate too much, either sending my blood sugars too low or too high out of my target range (5-10mmol), and I have to add any correction doses from my Novorapid to my mealtime bolus to avoid insulin stacking. So, having accurate-ish readings is crucial to avoid injecting one unit too much or too little. I wouldn't be too bothered for the discrepancy to be between 1-2mmol if my readings were, say, 7.8 for BG and 9,2 for ISF, for instance. But when BG is 6.5 and ISF is 8.5, for example, with the latter not catching up to the former, it admittedly isn't very reassuring for me. Up until now, I was under the impression that my fluctuating levels were because of how my body has been reacting to my insulins, but I have been wondering if a lot of the lows and highs I have been getting could have been caused because I unknowingly under or overestimated my doses strictly following the sensor readings. It's frustratingly tricky, especially as my consultants are constantly on me for having high levels, yet the advice they give me to correct doesn't help, not helped further by them not listening to me.

I am also aware of the 48 hour calibration issue. I have been considering trying to insert my sensors earlier before activation, as I have noticed a good portion of my sensors over the past month have been reporting higher readings within the first two days over 1.5-2mmol. Some actually stayed below 1mmol, but it generally seems my body doesn't react well to the sensors for the first two days. As for seeking out help from different consultants, I unfortunately don't have control over who I speak to. I'm automatically sent appointments from the hospital every 3-4 months and see a different consultant by default every time from the diabetes team, so it's usually a case of them picking up the slack from the last consultant and then giving me conflicting advice. I only got to see a DSN once back in April of this year about my predicament because a consultant decided to refer me to them 'as a precaution'; I agree the DSN was much more helpful and actually did check the reports to dig deeper into the problem.

I think the whole experience has shaken up my faith in the sensors in general, to be honest. I don't want to stop using them because they have been very convenient and helpful in terms of not bruising my fingers too much or allowing me to check my levels when I can't find anywhere to finger-prick test. But considering my recent issues with my insulin and trying to figure out the correct doses, I also want to make sure that it's definitely the insulin doses causing the problem and not me trusting inaccurate readings. And also not wasting sensors purely out of paranoia. I think I've changed more sensors over in the last month or two than I did since starting them. Lol. XD
 
Last edited:
Thank you for the reply. Really appreciated.

Funnily enough, at my last hospital appointment last month, it was revealed my HbA1c was about *7.4mmol (I think; poor memory), while the LibreLink app estimated it at *8.0mmol at the time. As the two were fairly close, it does seem like the sensors I've been using up until then were fairly accurate, so that's put my mind at ease a bit.

I'm aware that the monitor can also have its inaccuracies, but like you've said, I was told by the DSNs and consultants to always refer to my monitor when checking my levels. They suggested I switch to the sensors so I could get a better picture of how my body is handling the insulin I'm injecting, so that is primarily what I use the app for. I don't mind the 1-2mmol discrepancy as much if my blood sugars on the monitor are, say, 7.4 and 9.0 on the sensor, because that is still close enough for me to add a correction dose alongside my mealtime dose (I have to do both at the same time to avoid insulin stacking). But because of the recent problems I've been having with my insulins and fluctuating levels, I have grown concerned that a lot of the highs and lows have actually been caused by me injecting per the sensor readings rather than my monitor and not realising where the discrepancy is at the time. The consultants are also heavily reliant on the figures shown on the app and have given me their advice purely based on that, a lot of which has not been helpful (especially as they also don't listen to me when I try to explain things I've noticed that the libre hasn't picked up). So, it's currently really frustrating for me in trying to figure out what the actual problem is. I want to make sure that the data they're looking at is as close as it can be to where my levels actually are, so I can't help wondering if part of the reason their advice hasn't been very effective is purely down to the sensor giving inaccurate readings are specific points of its lifespan.

Edit:

*7.4%
*8.0%

Used the wrong unit of measurement and didn't notice. Whoops. XD
 
Last edited:
Understood. It's perhaps worth having a look at the GMI (Glucose Management Indicator) figure on the blood glucose report from the LibreView site also, for comparing with HbA1c. If you have no unusual blood conditions that can make the HbA1c blood test inaccurate for you, and if the Libre sensors have been very close to accurate, then the GMI figure on a blood glucose report, with the date range for the report set to the 2 to 3 months leading up to the blood test, should be very close to your HbA1c result. The A1C number displayed in the app uses an older formula for estimating HbA1c. GMI does the same thing but uses a newer formula created using data from CGMs. If the sensors were perfectly accurate and if your red blood cells are being replaced at the rate of the average person then HbA1c and GMI should match up nicely.
 
Thank you for your responses. I wasn't actually aware that this was a thing, as I primarily use the app and haven't logged in to the LibreView website before now. I'll be sure to give that a look and see what the GMI says. Hopefully, it'll give me a better idea as to how well the sensors have actually been working.

Regarding your other post, the monitor device I use is Accu Chek Aviva Expert. I do frequent control solution tests and when doing finger-prick tests in the past, I've found the sensor results have been fairly close to it, hence why I haven't really questioned it until now. Despite the varying discrepancies I've had over the past month, my sensor results have lined up with my BG results quite closely 90% of the time (either under 1mmol or up to 1.5mmol), so it does seem like my monitor is doing pretty well. The DSN who prescribed it to me at the time said a lot of positive things about it, but in terms of how well it's done in studies and whatnot for accuracy, I can't say. It is five years old now, though, so I will probably have to change it out for a different model soon. I have also tried to look into the calibration apps, but as I have an IPhone, I can't really get any of the apps mentioned by others on the forum. It seems they are primarily for Android phones.

Edit: Took a look at the GMI up to the day I originally had my blood test done and it says it was around 7.5% (59mmol). That is only 0.1% off my HbA1c result, so it looks like up until that point, despite the blip I had with my sensor the week prior, they have been recording my levels fairly accurately. I'll have to see if that stands come my next appointment, but at least I know the results prior to August were okay. XD
 
Last edited:
Introverted_And_Proud said:
, it was revealed my HbA1c was about 7.4mmol
Click to expand...
Are you sure about that unit of measure?

In the UK, the HbA1c tests are usually reported either as a percentage or as a value in mmol/mol (not to be confused with mmol/L which your normal blood sugar is measured in). It is possible to estimate your daily average based on your HbA1c (which you may have done), but I don't think it is normal for the labs results to do this.

8 mmol/mol isn't a valid HbA1c result, but 8% could be and would be the same as 64 mmol/mol which would mean that and estimate of your average blood sugar for the previous 90 days would be around 10 mmol/l.

There are a lot of confusing things related to diabetes and one of the most confusing seems to be the various ways of measuring blood sugar and the whole HbA1c vs GMI vs eA1c.
 
Introverted_And_Proud said:
can't really get any of the apps mentioned by others on the forum. It seems they are primarily for Android phones.
Click to expand...
Have no experience of them, but there are alternatives apps for the iphone including Shuggah and xDripIOS (which despite the name isn't related to xDrip on Android)
 
Ah. Apologies. I think I got a bit muddled with the units and didn't notice. XD I was meant to say 7.4% and 8.0%. So, my HbA1c was around 7.4% and my Libre app estimated it at 8.0%. So, in terms of mmol measurement, my app estimated it was 64mmol, but the HbA1c was around 57mmol. But both are still pretty close.
 
Sorry for the late reply @Introverted_And_Proud and welcome to the forum.
I just wondered if you'd done a basal rate test recently? If you do say 3 over a couple of weeks or so, and record your finger prick tests every 2 hours, you could show these to the hospital team and also have a better idea yourself if your basal insulin needs altering.
 
Hi there. Thanks for the reply.

I haven’t done a basal rate test, no. I can’t find an appropriate time in my daily schedule where I could perform it safely. However, I have spoken to a DSN about it. They checked the Daily Patterns graph Libre report and found that my blood sugars start to lower rapidly at around 3 in the morning instead of remaining flat. I eat a very small snack before bed that doesn’t require me to inject any Novorapid, so the DSN determined that I was injecting too much Tresiba and needed to reduce my dose until that trend flattened. At the time, I was on 7U. Now I’m constantly switching between 6U and 5U. Same thing is still happening, but with 5U, it’s a steeper drop.

However, throughout the rest of the day, my sugar levels rise too high within the first two hours of some of my meals, and then it starts lowering very quickly during the latter two hours. Whenever I inject 6U of Tresiba, sometimes my levels don’t rise very high, so when they drop after two hours, they plummet too low, sometimes into hypo territory. I tried adjusting my Novorapid to carb ratio to see if that helped and it didn’t. But when injecting 5U, the drop isn’t as bad, but it means the rise after my meals is worse. It’s like my levels are constantly on a rollercoaster. When I was on Levemir, I had a similar problem, just less extreme on the low side. So, I have a feeling the basal insulin is the culprit.
 
How long before eating do you usually inject your novorapid?
Many of us have found we need a (much) longer prebolus than the generally advised 15 minutes.

There are also quicker acting insulins available, Fiasp and Lyumjev.
 
Antje77 said:
How long before eating do you usually inject your novorapid?
Many of us have found we need a (much) longer prebolus than the generally advised 15 minutes.

There are also quicker acting insulins available, Fiasp and Lyumjev.
Click to expand...
For the majority of my meals, it's right before I eat. The only exception is my lunch during the weekdays, which I delay eating by around 20 minutes after injection. I have attempted to delay my eating time for some of my meals, but the results haven't been great. In some cases, because of the food I'm eating, it doesn't get into my system fast enough to counteract the novorapid, so when my levels do start lowering, they drop into the red. Other times, I've managed to curb the rise, but because of my blood sugars plummeting after two hours, it worsens the drop.

The last consultant I spoke to about these issues recommended I might benefit from switching to Fiasp. I'm currently doing some research into it first before making up my mind, though. I never did that with Tresiba and then found out after I started using it that the consultants didn't give me a thorough enough explanation regarding how it works, among other things. It came back to bite me pretty hard, as you can see. XD
 
sgm14 said:
including Shuggah and xDripIOS (which despite the name isn't related to xDrip on Android)
Click to expand...
XDripIOS started as an old version of xDrip+ many years ago but the developers took a "branch if code" and they have never been referred.
 
You're not alone, I often have quite large offsets between my sensors and fingerprick readings, I calibrate using the Android version of XDrip+, I'm not sure what the iPhone version can do but if you do try it out please do let us know. These offsets tend to change over time and it's quite interesting seeing how the calibration line which XDrip+ displays moves over the life of the sensor. I try to do two calibration blood tests per day, ideally reasonably far apart (but still within usual levels - not massively high or low) and with reasonably constant BG (as seen from the 1min data, which can show vastly different behaviour to the "smoothed" 5min data.)

Sometimes the discrepancy is so large (libre reading too low) that XDrip+ refuses to calibrate it (as minimum value from the sensor would be above the hypo warning level) at which point it's of no use any more so I ask for a replacement and send back the faulty sensor.
 
In Response said:
XDripIOS started as an old version of xDrip+ many years ago but the developers took a "branch if code" and they have never been referred.
Click to expand...

I don't know if I am being pedantic, but that's not how I would categorise it. I think it is find to say Shuggah started as an old copy of xDripIOS as they copied and used all source code from xDripIOS.

But XDripIOS was started as a completely new project when one of the people who used xDrip (and contributed to the project) switched to use an Iphone and wanted something similar. But as it had to be written in a completely different language, it could not contain any of the existing code - which undoubted made it a huge task.

I think the name has been changed several times, but I still think it is confusing.

It is obvious that LibreOffice Calc for linux is inspired by Excel, but there is no way they would get away with calling it ExcelForLinux.
 
Thanks for the reply.

I gave the App Store a quick look and it appears the Shuggah app is the only one available for IPhone at the moment. I'll probably hold off from trying it for now, though. I'm going to see if activating my sensors one or two days after application might help in the accuracy department. Over the course of last month and the previous, my sensors were mostly fine for the first ten days, but the problem with the discrepancy worsening seems to be occurring when it reaches eleven to twelve days. The only exception was a sensor I applied four to five days ago, which I've since swapped out because the discrepancies were all over the place, even when my BG was in normal range. Sensor kept reading my ISF levels to be over 1mmol to 2.5mmol higher compared to my BG levels a good 90% of the time and it was giving me so much anxiety.

My new sensor was applied yesterday on my other arm and I activated it today, so fingers crossed.
 
You must log in or register to reply here.
Menu
Log in
Register
Cookies are required to use this site. You must accept them to continue using the site. Learn More.…