Abbott Freestyle Libre results vs BG meter - Review part 2

Last day of 2nd sensor. Complete with crazy-flat overnight graph.

In this update to my initial review of the Abbott Freestyle Libre I wanted to share my experience of what it was like to use and also to provide the promised comparison with results from my regular fingerstick meter (the Contour Next Link USB). If you have been living in a cave and have no idea what I'm on about you can get an introduction to the Libre here.

It is probably worth pointing out at the outset that the Libre WILL NOT give you identical readings to your BG meter all the time. It just won't. They are not even measuring the same things. If that's what you want the Libre to do, you can pretty much stop looking, because it's not going to happen - not with the Libre and not with *any* technology that measures glucose in interstitial fluid and then converts that into an estimate of plasma glucose values (so all current CGM options). More on that later. But first - what is the Abbott Freestyle Libre like to use?

Do the sensors come unstuck?
I found the Libre to be extremely comfortable and unobtrusive to wear and both sensors stayed firmly stuck for the full 14 days. I was almost never aware of the sensor being there, with the exception of one or two times when I leant on it or absent mindedly scratched near it (having forgotten it was there). I did deliberately try to remember it when towelling off after showering to make sure I didn't accidentally dislodge it, but apart from that I barely gave it a second thought. There was very little in the way of itchiness for me and at the end of 14 days the sensor removed to reveal, well... nothing very much. Insertion of the second sensor was as painless as the first. I have seen some comment online where people have found that sensors came loose after just a few days, but that certainly didn't happen to me. With infusion sets I have found that the adhesive seems to need 12-24 hours to get up to full strength. I suspect it is the same for the Libre sensors. If it gets loosened early on you may need to over-tape it with Opsite Flexifix, Tegaderm or something similar. There's not a lot of 'edge' on the sensors and I can see the possibility of the sensor getting 'levered off' if knocked in just slightly the wrong direction.

I wonder what my levels are now?
I think this picture sums up my experience of living with the Libre for 28 days. Since I started attempting to *actively* manage my blood glucose levels four or five years ago I spend quite a lot of time wondering what is going on between my BG checks. I have never really known how often this happened until now - but the Libre released me from any monitoring constraints. So there you go. An average (average!) of 31 checks a day. Something like once every 30 minutes during waking hours. I had expected that my frequency would drop off after the initial excitement, but I remained fairly consistent through the full life of the sensors - checking more frequently when levels were changing more rapidly and leaving hours between checks at times in the day where things are generally more stable. The inconvenience and discomfort of 'traditional' punctured-finger-and-strip-fiddling monitoring means I would never check this often via a BG meter, but for 28 days whenever I wondered what my levels were, I found out. It is easy to underestimate how much of a transformation this is. Any time. Every time. All the time. I could know what my levels were *and* what they had been over the last 8 hours. There was no longer any reason not to look. No 'avoiding' the numbers (whatever they might be!) in between my 'normal' testing routine. Rather than feeling swamped and judged by data I felt completely liberated.

Basal test every day
In my previous post I mentioned how amazing it was to suddenly be able to see what had happened overnight every morning. Overnight basal tests are - as everyone who has undertaken them will know - a real chore. Waking every couple of hours to check blood glucose via fingerstick is a very effective management technique - but it's no fun. But look at what happened when I had 28 consecutive basal tests - the result is at the very top of this post. I couldn't help but tinker with my overnight basal profile gradually moving it from a bit wobbly, to ridiculously level (at least for a short while!). Seeing the patterns every day meant it was easy to spot the general trends and ignore the one-offs. And I made alterations to the 'shape' of my overnight pattern that I would *not* have made without those data.

New technique
After I had been living with the Libre for around 7 days I noticed that I had added a whole new technique to my management armoury. The preventative TBR (temporary basal rate). I've regularly used them since I started with Artoo to cover activity and other things, but this was slightly different. Because of the constant availability of data and the trend arrow that accompanies each check, I found myself setting short sharp TBRs - perhaps 50% for 30 minutes or an hour - to head off an impending dip in glucose levels. I was slightly surprised to realise that the 'level' trend arrow on the Libre doesn't actually mean level as such. It just means not rising/falling fast or very fast. In the first week on sensors I ignored a lowish level because of the 'flat' arrow only to dip below 4.0mmol/L an hour or so later. By using preventative TBRs for flat or more accurately reading-flat-but-slightly-falling readings that were just 'a bit near the edge' I managed to avoid several low level hypos and without bouncing up into double figures. Similarly, micro-boluses of small fractions of units when levels were high-ish (but not yet out of range) allowed me to be bolder in preventing BG drift. WIN!

Data analysis
I really like the simple, clear overview screens on the Libre. They might be a bit simplistic for some, but by dividing the day into four chunks and averaging 7, 14, 30 and 90 days of results in those sections I found it very easy to spot patterns and filter out the ebb and flow of 'diabetes randomness'. I particularly liked the 'Daily Patterns' graph which only appears after 5 days of results are stored and offers an average of daily results along with a 90th and 10th centile shaded area (it's a simplified version of the Ambulatory Glucose Profile graph - see image). It becomes very easy to see which periods in the day are providing the most challenges. You get a similar view (but with many more options, and helpful traffic lights) when you connect the Libre to your PC (or Mac - hooorah!). The Libre software niftily allows you to create smart PDF reports recording all sorts of averages, graphs, low glucose events and mealtime patterns which can really help to understand what has been going on. Additionally any mealtime notes and/or carbs and other details that you may have chosen to record on the libre are transferred for review too. If you have a particular penchant for number crunching you can also download up to 90 days results and details as a 'tab separated' txt file ideal for import into your spreadsheet package of choice. Look for the 'File' menu at the top of the screen, and choose 'Export Data'.

The end result of my 28 days? I had some of my BEST results all year. Lower post-meal spikes, fewer lows, fewer highs. Not bad Abbott! Thank you very much Freestyle Libre.

Ready to insert second Libre sensor

Libre results vs BG meter
Well... this is all very well you say. But if the results the Libre is providing are complete garbage, then it doesn't matter a hoot how nifty the downloaded PDF reports might be.

So how reliable were the results?

It's a good question. And not necessarily an easy one.

As I suggested earlier the Libre is reading glucose levels in interstitial fluid (via factory calibrated sensors) and then interpreting those through an algorithm to present values intended to reflect plasma glucose levels. That's a whole lot of steps in between what the Libre reads and the results you receive. And much as I realise a lot of work will have gone into ensuring that those results will be mostly OK for most people most of the time - clearly it is likely or at least possible that some variation will occur. Not least because the glucose values in interstitial fluid will 'lag' behind blood glucose values (typically 10-15 minutes, though the Libre aims for 5). Things are complicated further by the +/-20% variability which is allowed between BG fingerstick strips (though again these days meters tend to perform more like +/-5% or 10% most of the time).

What matters more to me is not whether results are identical all the time - it is more whether the results I get are useful. Whether any variation is modest enough that the readings, trends and analysis help me manage my diabetes better.

In order to understand what I was getting from the Libre I tested both sensors against my standard BG fingerstick testing routine. This is usually between 6 and 10 fingerstick tests a day, including premeal and post-meal tests as well as those around driving, exercise and activity. Whenever I took a BG fingerstick test on my 'normal' meter (the Contour Next Link USB that works with my pump) I cross-checked with the Libre. At the end of each sensor I then downloaded all the data and compared it on a spreadsheet to see what (if any) differences there had been. I do not for a minute pretend that this is a scrupulously scientific test - or that the results here might apply to anyone else. I took these comparisons purely for my own interest, and share them here in case others find them useful.

Here's a summary table from the first sensor - there were 94 pairs of results (BG meter vs Libre) over the 14 days:

Sensor 1

	BG meter	Libre	Avg +/- % (against BG meter reading)	Avg +/- mmol/L (against BG meter reading)
Average	7.8mmol/L	7.4mmol/L	10.1%	0.7
SD	2.8	3.0	8.9%	0.6
Distribution of readings
Number of readings where Libre higher			18	19%
Number of readings where Libre lower			70	75%
Number of readings equal			6	6%
Number of readings within 0.5mmol/L			43	46%

I don't know about you - but that looks pretty darned good to me. Yes, in amongst that there were some readings which were substantially different. A handful of times around 40% different from the BG meter - to put that into context one was a 4.4mmol/L that read as a 6.1mmol/L - one needs watching the other doesn't. There were also results where the Libre read hypo, but my BG meter confirmed I was in the 5's.

But the vast majority of the data feed is well within usable limits for me. The SD line gives a useful estimate of the range of most of the results. So from an average of 0.7mmol/L difference the most of the results were somewhere between 0.1 and 1.3mmol/L out.

You will notice that in most cases (but not always) the Libre was reading lower than my BG meter - with the result that overall the average of all results for the Libre comes out at 0.4mmol/L lower than my Contour Next Link, but with a slightly wider range (SD of 3.0 vs 2.8). Amazingly though, almost half the time the Libre was reading within 0.5mmol/L of my BG meter - and don't forget that at least some of the differences in readings could be down to the 5-10 minute 'lag' between BGs and interstitial glucose - especially if BGs were moving fairly rapidly at the time.

What I think this means for me in practical terms is that if I was running off results from the Libre more or less full time, with only a few cross-checked results where BG was moving rapidly or if things 'didn't feel right' I would most likely run a little higher most of the time. This would probably do wonders for my avoidance of hypos, but I suppose it may also have a small knock-on effect on my HbA1c.

But.

Isn't there always a 'but'?!

Things get a little more interesting when I switched to Sensor 2. As soon as that sensor started reading the graph 'jumped' down the screen. If the first sensor had tended to read a little lower, this sensor was taking that to new levels. Now Abbott do suggest that the Libre sensors may not read quite so accurately on day 1, but that days 2-14 should be relatively steady. Even on the second day though readings on the Libre were consistently 2mmol/L or more lower. And since there is no calibration option there was absolutely nothing I could do about it.

Here are the results for the first 5 days of that sensor:

Sensor 2 - days 1-5

	BG meter	Libre	Avg +/- % (against BG meter reading)	Avg +/- mmol/L (against BG meter reading)
Average	7.1mmol/L	5.4mmol/L	26.2%	1.7
SD	2.3	2.3	13.7%	0.8
Distribution of readings
Number of readings where Libre higher			1	2%
Number of readings where Libre lower			43	98%
Number of readings equal			0	0%
Number of readings 1.5mmol/L or more out			26	59%

You can see the difference in performance for yourself - over twice as far out. Suddenly almost 60% of the time the readings were 1.5mmol/l or more different and almost always lower. I had a few overnight traces that looked as it I'd been on the verge of a hypoglycaemic coma all night - when in reality I'd just been bobbling along in the low 5's (I know because I panicked and checked). On several occasions the Libre just reported 'LO' - with (allegedly) a level too low for it to record.

After several days of 'waiting to see' I contacted the lovely folks at the Abbott helpline who ran through a few checks to make sure I had the sensor in the right place and generally could tell my Libre from my elbow. I had been given a handful of Freestyle Optium Neo strips with my pack of goodies and the helpline person suggested I tried cross-checking against the Libre's inbuilt BG meter rather than some other technology (interestingly these strips also read a smidge lower than my ususal meter). As it turns out the reading on that occasion was pretty close (typical!) so I left it at that and wished I'd called them earlier.

The plot thickens
A couple of days later I had a call from Fiona who I'd met at the Libre pre-launch meeting called to say that a software glitch had been discovered (and fixed) in some of the very early sensors which was causing some interruptions in data and other concerns. I related the issues I'd been having with the second sensor and while they didn't seem to exactly fit with the software glitch described, Fiona offered to replace the second sensor.

In the meantime, and certainly into the second week, the second sensor seemed to be settling down considerably and behaving much more like the first. Here are the results averaged from days 6 to 14:

Sensor 2 - days 6 to 14

	BG meter	Libre	Avg +/- % (against BG meter reading)	Avg +/- mmol/L (against BG meter reading)
Average	6.9mmol/L	6.5mmol/L	11.7%	0.8
SD	2.2	2.1	10.2%	0.6
Distribution of readings
Number of readings where Libre higher			15	27%
Number of readings where Libre lower			38	69%
Number of readings equal			2	4%
Number of readings within 0.5mmol/L			21	38%

Not exactly back to the performance of the first sensor, but pretty darned close. Of course it did take those 5 days to settle. And if I'd been stumping up hard-earned cash for that sensor I suspect I might have been quite miffed that data from 36% of the life of the sensor was fairly useless to me.

I would have expected that a relatively consistent error would have still yielded useful 'trend' data - but in reality I found it very difficult to detach myself from the uneasy feeling that 'red' (hypo) results gave. However much I knew, or suspected, that my *actual* BG level was 1.5-2.5 points higher it was impossible for me to use the information in the same way when the margin of error was that great.

So in the end I am left with one spare sensor yet to use (which I plan to put into action around Christmas time) and just a very slight sense of unease as to what future sensor(s) might bring. When it worked at the 10% MARD that Abbott promise in the marketing literature I found it an incredibly powerful tool. If it were available on prescription and performed consistently like that I would be banging the table and asking to swap my (fairly generous) fingerstick allowance for Libre sensors - topping up with as many strips as I needed on top of that out of my own pocket.

But.

If other sensors only perform to the accuracy of the first 5 days of Sensor 2 then it's a whole different ball game. I wonder if Abbott may come to regret their decision to go for factory calibration. From a user's perspective I have already seen several posts from people who would find it very useful to have a 'manual override' setting where you could offset Libre readings which are consistently out to bring them more in line with our own fingersticks.

All in all I love the Libre and I am really hoping that future sensors live up to overall positive experience I have had so far.

Now that the online shop has opened I'd be interested to hear your experiences if you have used the Libre yourself.

Final verdict: 4/5.

Update: Freestyle Libre now licensed for use in under-18s