Posted by on Saturday, 30 December 2017

CGM, diabetes time travel, and lessons learned from go-karting

Image by 'aurorasognatrice' used under cc.
2017 is hurtling to a close, and despite my giddy optimism about having 'loads of time' over the festive break to do all sorts of things that I don't generally get around to, the days have passed in a blur of hopelessly guestimated carbs, fun, friends, family and alcoholic excess. Consequently I am once again hastily cobbling together a round-up of the year type post - you lucky lot.

Almost exactly 12 months ago I was placing my order for a transmitter and first box of sensors for my MM640G insulin pump. I'd trialled the system over the summer of 2015, and I have always opted for Medtronic pumps with the vague notion of possibly, some day, self-funding CGM - but this was the first time our family finances had permitted it. I'd been using Freestyle Libre sensors intermittently for a few years (which you can use sporadically without the additional £500 for a transmitter), so I was interested to see how occasional SmartGuard coverage worked out for me.

In the end 2017 turns out to have been by far my most sensor-filled year. I was invited to trial Medtronic's Guardian Connect in April, and was unexpectedly and very generously gifted some short-dated sensors by someone who was switching systems and could no longer use them.

I always try to get the maximum use out of every self-funded sensor I insert, and I am fortunate to be able to restart almost all of them for at least another 6 days while retaining good performance. This almost halves the cost - or more accurately for me, doubles the sensor coverage. I had hoped to spread 10 'stretched' sensors (approx £500-worth) across the year to give me somewhat less than 50% coverage, but in the end, have been able to use them continually for quite a number of months which has been a very interesting contrast to my usual pancreas impersonation guesswork. It's interesting to reflect on the changes I've noticed myself making to day-to-day management decisions, and how it has felt as an experience. Here are a few basics:
  • On the whole it has felt far easier to live with diabetes this year.
  • I've not done any complex analysis of BG results, but my basic monthly spreadsheet analysis (nerd alert) shows significantly better results.
  • My A1c has fallen by 7mmol/mol (0.5%) and my hypoglycaemia has dropped significantly
  • Sustained reduction in hypoglycaemia has really improved the reliability and timing of my warning signs.
  • For the most part I seem to be operating with around 80-85% of results in range (4-9mmol/L) - even over Christmas. This is ridiculous.
  • On average I've only been getting 1-2% of results below 3.9mmol/L, nocturnal hypoglycaemia has been almost completely eradicated and I am having days and sometimes weeks at a time with all but no readings below 4.0.
  • These are not results I can achieve without continuous data, not matter how hard I try.
  • Even with the benefit of Freestyle Libre these are results I find it impossible to achieve. The alarms of full CGM provide me with significant added benefit, especially for catching lows.
  • Occasionally life with CGM has been rage-inducingly frustrating, and alarms have sometimes driven me to distraction.
  • I've had a few duff sensors and made some very poor choices based on inaccurate sensor data.
  • Additionally, SmartGuard is a bit of a liability with an inaccurate or under-performing sensor, sometimes sending me into double figures with a cancelled basal (and sometimes half of a slowly-delivered bolus!) when I would have been fine if left well alone.
  • SmartGuard is amazing for me, but very often I can't resist overruling it, ending it early and/or adding some carbs. Sometime this works better, sometimes I suspect I create more problems for myself than is strictly necessary 
  • Additional data is a significant help to me day-to-day. I've come to rely on it and feel quite lost without it, but there are times when some of the subtle details of life with CGM have created their own challenges - which leads me to...
CGM and diabetes time travel
I've always been one for a thinly stretched analogy. I can't help myself. There are two coming up... You have been warned.

With absolutely no apologies for the shameless 'Christmas Dr Who' reference, I've also been thinking quite a lot this year about the diabetes time travel that you get involved in when tinkering with continuous data. I've considered this before in terms of the repeated half hours you can spend with diabetes waiting for various management decisions you have made to start working - periods of waiting which can seem interminable. With more time spent in CGM-land this year I've noticed additional time travelling shenanigans with the lag between 'sensor glucose' read via interstitial fluid and actual 'what's happening now' blood glucose information.

This was brought more keenly into focus with my brief dalliance with the faster-acting Fiasp. An insulin surrounded by feverish hype of very rapid action - which sadly for me rather failed to live up to expectations. But the promise of faster acting doses has occasionally made me feel the sluggishness of Novo-not-very-Rapid all the more keenly.

Mostly I find 'sensor lag' is barely noticeable, but with a whiff of irony, it is when my BG is on the low side and I'd really appreciate accurate information that this 'time travel' is most clearly noticed. Looking at a Libre or Enlite trace at those times, shows you what was happening something like 10 or 15 minutes ago. And any 'rapid' carbs you take to bring up those low levels, or turn-around a dip towards hypoglycaemia won't change your blood glucose for 10 or 15 minutes, and may not show on the eagerly-watched line for a further 10 or 15 minutes after that. More than enough time to double-treat, only to watch your levels climb into double figures a little while later.

In just the same way, when you have a trace to watch, there is a powerful urge to see a high-and-rising BG trace turn around. But correction doses for me are unlikely to show any noticeable effect before 60 minutes have passed (unless I add increased activity into the mix) - plus the mandatory 10-15 minutes, of course. And at each of these moments the graphed time-travel of results, and interminably upward direction of travel gives few clues as to when it is 'just about' to level, or begin to dive downwards. Threepio proudly suggests 'insulin on board' from which you might think I could make an educated guess (and sometimes I do luckily seem to drag some precious clues) but all too often I can find my IOB dwindling away to nothing, coping only with mis-guessed carbs, rather than the elevated BG I had hoped to squash. Other times I can bear it no longer and wade in with an additional dose, only to see my BG trace drop off a cliff and plunge downwards - awash with both insulin and various types of carbs. Which reminds me...

CGM and lessons learned from go-karting
I have only been go-karting three times I think. By which I mean the crash-helmeted-boiler-suited-whiff-of-2-stroke-engines style go-karting rather than the sliding-down-a-hill-in-a-fruit-box-with-pram-wheels-bolted-on style. I don't think it's an exaggeration to say that I'm absolutely terrible at it. I was reminded of my go-karting prowess when thinking about some of my... erm... more questionable diabetes decisions in response to a more frequent CGM data-feed. I suspect I am not the only one who has fallen foul of the double-dose and/or double-treat temptations - and for all the benefits of all that extra information, it is unmistakeably one of the risks.

The very first time I sat in a go kart, in a dimly lit, oily, industrial shed on the outskirts of our city, my driving style was essentially binary. The accelerator was either fully down or entirely untouched. The steering wheel locked at either edge or dead centre. Brakes were applied with sledgehammer-like gracelessness. As a result I kangarooed around the indoor circuit, making full use of the amply-supplied tyre walls and doughnuting my beleagured kart in furious circles. I proceeded at lightning speed from one collision to another and made very little effective progress around the twists and turns of the circuit.

This is pretty much the way I drive my diabetes when things are not going well. Frustrated by apparent lack of action of more reasonable measures I heftily over-correct with hugely inflated insulin doses or swigs of Lucozade. Lurching and stumbling from high to hypo and back again. Crash! Clonk! Screech! Everything becomes overblown and chaotic.

The second time I went karting was every bit as 'successful' as the first, but I had something of a lightbulb moment on the third occasion - which was at an outdoor circuit. For whatever reason, this time I adopted a more subtle approach. The accelerator was rapidly pressed full down at the start, then backed off in time for the first corner... rather than stamping the brakes I found myself making little feathered dabs. Steering was altogether more considered. Confidence grew. Speeds into tricky corners increased, a late firm braking followed by a hard turn of the wheel and full throttle at the mid-point of the turn allowed me to power-slide out of the corner. I was still making strong adjustments in some circumstances, but more often my choices were much more subtle, more measured.

This is what I see when my diabetes management is more successful. It is when I am able to make smaller adjustments that I do better. Multiple big overlapping doses and rapid-carb 'rescues' can leave me in a flat spin and going nowhere. Smaller tweaks, spaced further apart are often significantly more successful. I have found this to be a really interesting and important thing to think about when Threepio is merrily warbling away. Those alerts can seem equally frantic, but I need to pause and consider my position on the 'circuit' of my day. Can I just coast through this corner without stamping on the brakes (adding insulin) knowing that the turn of the wheel or dab on accelerator I've already made is enough, or is this the devious hairpin and do I need full-lock and firm braking before powering-on with additional carbs to make the turn.

In general terms, when I notice that I am oversteering and stamping on the brakes and accelerator of my diabetes managment, I am trying to remind myself to make a couple of slower laps and build up to speed again more gradually with more gentle adjustments.

Widening the access to continuous data?
I was delighted to be invited by Abbott to attend DxAmsterdam in July, and then in September the eagerly awaited news was released that Libre was to be placed on the NHS tariff and could theoretically be made available on prescription, subject to local CCG decision-making (and perhaos an emerging postcode lottery). Abbott's real-world data, shared in Amsterdam, backs up my own experience of access to continuous data - that the more information I have on which to base my guesswork, the better things tend to go for me.

Diabetes still has the capacity to be hugely annoying of course, but for me - more data certainly leads to better results. Continuous data is not without its challenges, and it will be vital for people living with diabetes to be given appropriate support and help in order to make best use of the information and avoid the pitfalls. Both in terms of their support from their clinic and also perhaps, those who have made the mistakes before them and can share their experiences. It will be really interesting to see what effect wider access to these technologies delivers as access to Freestyle Libre and CGM increases in the years to come.

It remains to be seen quite how much life my transmitter has left in it. They are warrantied for just 12 months. I will continue to use Enlite sensors for as long as I can, especially since my CCG seem rather reluctant to take any decisions on the finding of any kind of sensors for anyone - however great the need. I would imagine I would be very far down any list they eventually decided to draw up. After that I'm not sure if I will revert to Freestyle Libre, spring for another Guardian transmitter or take a continuous-data break.

Wishing you all a tip-top 2018. And thanks as always for reading.

Posted by on Saturday, 2 December 2017

Keya Smart Meter review - Ketone and BG results in a single strip

Every once in a while, some device manufacturer or other drops us a line to see if we'd like to take a look at their fancy new gadget and see what we think. When it comes to blood glucose (BG) meters, it is very rare that something genuinely new comes to the table, often it's just a cosmetic way of getting the fingerpricker, strips and reader in one amorphous gloop of plastic - a fascination which escapes me. A couple of years ago, at a diabetes conference I had somehow managed to get into (not at all easy if you lack a medical qualification) I was struck by a display in the exhibition hall advertising a new BG meter, that seemed to break this mould, and be offering something new and really rather interesting. The ability to get blood glucose AND blood ketone results from a single strip. A strip that would be priced more or less in line with the general BG meter market.
 
For those mercifully unaware of the significance of that, blood ketones are generally regarded to be a Very Bad Thing if you live with type one diabetes. If left to build up in the blood due to insufficient insulin and raised BG they can quickly turn a normal day into one filled with paramedics, wailing sirens and everyone's second favourite urgent diabetes complication - Diabetic Ketoacidosis (DKA) - which, joking aside, is potentially lethal.

Ketones not keytones
Keytones are the annoying bleeps your phone makes as you type, unless you have the good sense to switch them off. Ketones take annoying to a whole new level and might easily send you scampering to A&E.

Ketone monitoring is generally advised for people with type 1 during periods of illness, or if BG levels become elevated (say above 13mmol/L / 230mg/dl) and remain there for any length of time. For some people on insulin pumps, ketones can be particularly problematic, because a blocked cannula or bad infusion site can leave little, if any active insulin after a matter of hours. I am lucky in that I do not seem to create ketones as readily as some, but I know that for many, especially parents of children with T1, ketones are a real source of anxiety and worry.

There are two ways to check for ketones, one is through urine strips, the other blood ketone monitors. Urine ketones offer a significantly delayed picture because it takes ages for them to filter through, and blood ketone strips until now have been fiercely expensive (approx £2.50 each strip) which means that those who secure them on prescription usually only have a limited supply, that they must use carefully when they need them most. But here was the promise of a meter that would give an instant blood ketone check, effectively free of charge, alongside every single BG result. This was a genuinely exciting prospect and I asked the Keya Smart folks to keep me informed.

It took a while for them to send me a meter to look at, but here is my n=1 experience of the Keya Smart. I can't say how it would work for anyone else, but this is what happened when I tried it out for a month.

Left to right: AccuCheck Expert, Contour NextLink and Keya
Pouch, pricker, strips and basics
One of the first things I noticed was that the case for the Keya Smart was a bit on the large side. The meter itself has a good sized touch-screen which is bright, responsive and easy to read. There is a USB cable/adapter for charging and data transfer, though data transfer can also apparently be done via Bluetooth - a feature which I did not investgate. The fingerpricker is perfectly serviceable and fairly pain-free with adjustable depth, but is not going to snatch my affections away from my trusty Multiclix. The strips, unusually in my experience, come in a flat container which nestles neatly in the pouch. This made strips extremely easy to remove when the container was full as they lined up like soldiers, but as the tub emptied the strips were able to slosh about a bit more and became a little fiddlier to remove. One nice design feature of the pouch was that the meter say inside four little 'corners' which neatly held it, but also allowed you to easily remove and replace it whenever you needed and also access the touch-screen completely unhindered. None of that faffing about with elasticated transparent bands.

Set-up, battery life and data display
You are walked through a basic set-up when you first switch the meter on (time, date, language and so on). You can then set glucose targets and activate other functions as you need them from the 'set-up' screen - for example whether you would always like to see blood ketone results, or only be alerted if they are elevated.

The battery seemed to last well - I would guess I'd get a good couple of weeks' use before I needed to recharge. The large screen must take some juice, and rather strangely there did not seem to be a way to turn it off when finished, you had to rely of the screen timing out and switching itself off after a few seconds. Pressing-and-holding the 'On' button didn't seem to do anything. This might be a bit annoying if you were running low on battery when out and about and needed to squeeze a few BG checks out of it until you could recharge.

The meter stores results for both BG and ketone results which can be viewed in a logbook table, or averaged over 7, 14, 30 and 90 days. I particularly liked the 'trends' screen which allows you to view results for either BG or ketones in a pie-chart style over 7, 14, 30 or 90 days.

You can flag results as being before or after meals, fasting, during sickness, for exercise or with insulin. These flags can then be applied to your 'trends' view to filter your results and see how your numbers stack up for those different times of day/activities. Importantly for me, you can add or edit those flags whenever you want to, there is no time limit. You can go back and add a meal tag several days later if you forget on the day. It frustrates me beyond belief when diabetes gadgets set arbitrary time periods for these kinds logging options. They are my data - I want to be able to access and update when it suits me.

Ketone values are highlighted either each time you check BG or only if elevated (amber) or high (red). Ketone values of up to 0.6mmol/L are considered fine, between 0.6 and 1.5mmol/L rates as 'Elevated' and any ketone reading above 1.5mmol/L is flagged as 'High'.

Build quality?
It all seems pretty slick and solid to handle (with the possible exception of the finger pricker which feels a little flimsy), but I did have an unfortunate time with my first Keya Smart meter which stopped turning on after about 4 days. Neither the strip port, nor the on button, nor recharging would bring it back to life, and there was no way to extract any data from it that I could find. It was fine one evening and just would not turn on the next morning. It was an ex-meter.

Possibly the most annoying error screen, ever.
Slurpiness
One aspect I always enjoyed reading in Tim and Alison's meter reviews on the venerable and much-missed Shootup related to strip slurpiness. How keen, or otherwise, a BG strip is to avail itself of a proffered droplet of blood. Sample size for the Keya Smart meter is a piffling 0.5μL, but unfortunately the strips themselves are rather bashful in welcoming your freshly squeezed fingers. It takes my Contour NextLink less than a third of a second to slurp up it's required sample, and the strips hungrily home in on blood from half a room away. By contrast the Keya Smart meter took a full 2 seconds, and seemed to need you to place your finger in a very precise alignment before it would deign to begin its dainty sipping. Not only that, but there was no opportunity to 'have another go' if a sample was fractionally short, as there is with some meters. Not quite enough blood on a strip and you had to abandon the check, and start again. Even more annoyingly, given it was completely outside of your control - the meter would reject some tests because the strip was filling too slowly - to which I would frequently shout, "I KNOOOOOW!!".

What was very unfortunate was the frequency of these strip errors. I don't know if I was just unlucky, or if it is something to do with the engineering of the strips to allow the dual results, but I was having significantly more checks rejected by the Keya Smart than by any other meter I have owned. Particularly irritating when you had a slow fill error, followed immediately by a underfilled strip error, before you finally managed to get a result.


Elevated BG, but ketones 'all clear'.
Results
Again I must stress that these are just my n=1 observations, but while I was using the Keya Smart I decided to check against my current NextLink USB meter. I often do this when evaluating a new piece of BG technology, because it helps me to know whether the new one generally reads higher or lower than I'm used to which can help inform my BG management decisions. The NextLink USB was said to be one of the most accurate on the market in a recent review, and I have always found it to be very reliable when double-checking a value, rarely differing by more than a few decimal points if I am not sure of a BG check and want to make sure it's not a rogue result. The official results in the Keya Smart handbook look similarly impressive with 94% of results within 0.8 mmol/L of a lab reference.

Sadly this was nothing like my experience.

Alongside BG comparisons, I also acquired a handful of Optium Blood Ketone test strips for my Freestyle Libre to cross-check any occasions where the Keya Smart meter registered elevated ketones.


Disappearing ketones
On multiple occasions I would recheck a value from the Keya Smart, only to be given a completely different number from the previous strip. Both blood ketone and BG values were subject to significant variation between two strips, checked moments apart. If either (or both) of those rechecks involved strip rejection(s) you can imagine the florid and colourful language which freely flowed.

Here you can see an initial check which alerted for high ketones despite in-range BG, by contrast the Libre shows only a trace of ketones. A recheck with another strip and while the BG value is very similar, the ketones have now dropped from 'A&E here we come' to 'Nothing to worry about here, sonny'. 

It did seem odd for me to get a high ketone alert at 4.0mmol/L BG, so this was easy to spot - but if it had been the earlier photo where I'd run in double figures all night I might easily have taken the first value as accurate. I have to say that in all the elevated/high ketone results I cross-checked, all the subsequent Keya strips and the Libre blood ketone checks only registered a trace.

And so, within a few days of using the Keya Smart, I had seen enough rogue values of either BG or ketones which had come out very differently a second time around that I knew that any value that was even slightly unexpected needed to be rechecked.


Keya Smart reads almost double.
But it's not always quite as easy as that in the day to day business of pretending to be your own pancreas. Sometimes you almost expect to get a high value. And sometimes the difference between Keya Smart and other glucose sensing technologies I was using at the time were actually quite alarming. If I had corrected for 12, when I was actually 6.8, then my correction dose, based on the data I'd received from the Keya Smart (had I not rechecked), would have aimed to send my blood glucose to 0.8mmol/L.  

That is more than a little worrying.

Keya Smart vs Contour NextLink - BG readings

NextLinkKeya SmartAvg +/- %

(against NextLink)
Avg +/- mmol/L

(against NextLink)
Average6.9mmol/L7.8mmol/L19.6%1.2
SD2.32.417.2%1.0
Distribution of readings
Number of readings where Keya Smart higher3682%
Number of readings where Keya Smart lower818%
Number of readings equal00%
Number of readings within 0.5mmol/L1739%

The table shows the results from 43 pairs of readings. Not a huge dataset, but the differences are quite marked. The Keya Smart almost always reads higher, sometimes significantly so. And only 4 times out of 10 does it read within what I think of as being the benchmark for 'pretty much the same' - less than +/- 0.5 mmol/L.


Conclusion
I really wanted to like this meter. There's a lot about how it has been put together that I really like, and the promise of 'ketone checks every time' I see as having real value for many people. Sadly though, the variability in the results I got means that it really isn't the meter for me. I will keep it and use up the strips I have for blood ketones as and when I need them, but I'll make sure I check three times on each occasion so that I can be sure of the result I'm getting.


I wish Keya Smart every success for the future and hope they can iron out these wrinkles for their next version. At the moment though, this feels like it's a product that's not quite ready for the real world.


Final verdict: 1/5


Disclaimer. I was offered a Keya Smart meter and sent it free of charge. I was not asked to write about the Keya Smart meter and I've not been paid to write this post or publicise the product in any way.

Posted by on Wednesday, 13 September 2017

Libre available on NHS - big news and bigger caveats

Last night the rumour mill reached fever pitch as Non Disclosure Agreements were stretched to their very limits. And this morning the Twitterweb was a-buzz with the news that Abbott's Freestyle Libre flash glucose monitor is to become available on the NHS from 1st November*. JDRF issued this nifty press release and everyone's second-favourite cat-loving pyjama-wearing T1 ex-schoolteacher and all-round good egg, Adrian Long, was even glimpsed on Sky News in the early morning undertaking some top Libre punditry and sharing his love of 'Libs'.

* subject to local healthcare economic approval, CCG friendliness, moon in jupiter, blah blah blah.

Of course, no sooner had the long-awaited announcement been made than people began to get a bit sniffy about it, or unbelievably optimistic - depending on their frame of mind. Either Libres were about to be handed out to everyone immediately, whether they wanted them or not; or it was going to be a postcode lottery / the end of CGM funding / a complete disaster.

The official announcement from Abbott covers the whole of the UK, including Scotland, Wales and Northern Ireland. Freestyle Libre will be on the 'drug tariff'. Which means that it will be able to be prescribed, and reimbursed by the NHS.

BUT (and depending on your point of view this might be a small niggle, or a deal breaker), this is subject to local health economy approval.

It's an important step, but it might not be the end
I think I probably come down more on the side of 'wildly optimistic' about the announcement. But the 'local health economy' / local clinical commissioning group (CCG) approval thing is a bit of a worry. Even the JDRF announcement is rather cautious, stressing how important it is that the technology actually does end up reaching people. There may still be some work to do in your area to encourage the bean counters to play fair.

Balance of costs
The cost to the NHS of one Libre sensor is going to be £35. For people using intensive insulin therapy who might be using 8 or more finger stick test strips a day, the costs more or less balance out. Assuming an average-ish strip cost to the NHS of £14.50/pot it costs the NHS about £2.32 a day for 8 strips, versus £2.50/day for Libre sensors lasting 14 days where you can be checking 15, 20, 30 times a day or more. I am reminded of the real-world data that Abbott shared recently. In general, across all their users, the more people used Libre, the better their results. Fewer hypos, fewer highs, more time in range and a lower predicted HbA1c.

Even if you are not prepared to take on the heaving behemoth that is your local CCG and try to turn them around to the idea, I can certainly imagine myself having an interesting conversation with my GP (who, of course, runs their own business) about exchanging my strips for sensors for all the added benefits that gives me. It may be that as part of that negotiation I suggest paying for my own strips for DVLA and other occasional requirements. The cost analysis undertaken by NICE for T1 demonstrates that 8-10 strips a day can be cost effective (more BG information is associated with better BG outcomes and reduced complication risk). There may be niggling details and rules about 'local formulary', but it's certainly a conversation I'd be interested in having with my GP if the local CCG drag their heels (as they have been known to do in my area).

Getting your GP and/or hospital clinic on-side and banging the table for you is a good plan too. Speak to them and get them to apply pressure to the CCG to ensure readers and sensors are listed in the local formulary (which is located at Hogwarts just down the corridor from potions and defense against the dark arts).

But what about 'proper' CGM?
Some people have worried that all these funds getting diverted to Libre will spell the end of CGM funding. Personally I don't see that happening. CGM is currently only weakly recommended in national guidance for people who have significant problems with recurrent hypoglycaemia and have lost all or almost all of their hypo warning signs. The submissions to the NHS for approval were very clear that while Libre can really help some people reduce their exposure to hypoglycaemia with extra information, they are not a substitute for CGM alarms/sensor augmented pump for those with no awareness.

Diabetes UK have put together a position statement on Flash Glucose monitoring which I was pleased to be involved in, and which I think clarifies many of the issues about the Freestyle Libre. What it is good for and who can benefit from it. Thankfully it involves people with Type 1 and Type 2 diabetes - and actually I think should be applied to anyone with any of the many types of diabetes who are intensively using insulin. The recommendations on page 4 are very interesting.

Of course some people have been able to carefully negotiate the fiery hoops to secure full or partial NHS funding for CGM in their own particular case. I'm not sure I see the availability of Libre as affecting the clinical reasons which led to their funding being granted - unless they wanted to swap of course! Freestyle Libre is not a CGM, and does not issue alarms. If those alerts are important to you, then CGM is the better option. But for others the lack of 'alarm fatigue' is a positive benefit of Libre.

Onward and upward
I am really encouraged by this announcement. It's been a long time coming and a lot of work has been done behind the scenes to get to this point. Huge thanks to Lesley and Melissa at INPUT, the team at JDRF and Diabetes UK, and not forgetting Dr Partha Kar in getting us this far.

I am absolutely convinced that Freestyle Libre has a huge potential to help thousands of people who are quietly struggling with their diabetes management. Not in extreme enough need with frequent A&E visits to attract CGM funding, but just keeping going not knowing what they don't know about their BG fluctuations. I really hope the technology can be made available so that those quiet strugglers can go from doing OK to doing really well. Can reduce their long-term complication risk and improve their quality of life.

Posted by on Tuesday, 8 August 2017

Fiasp review, fun with 50:50, and the mystery of the missing insulin

I have been using NovoRapid for many of my 'pretending to be my pancreas' years. I had a brief dalliance with Humalog not long after we started writing this blog, but switched back to NovoRapid when I started with Artoo as my DSN had worries about accounts of Humalog crystallising in pump tubing.

One of the challenges with NovoRapid, as many users are keenly aware is that it's not very... well... rapid. NovoSluggish perhaps? NovoOhForGoodnessSakeGetAMoveOn!

When your blood glucose is stubbornly high and you dose a correction only to find it even higher an hour later it is very tempting to rage bolus it into submission with multiple additional units, only to find that you crash into low blood glucose hours later once they have all started acting together. I once wrote a post about speedboats and oil tankers outlining my frustrations around slow insulin action where everything else seems to act very fast indeed.

For those that don't know Fiasp (Faster Insulin Aspart) is the latest insulin from NovoNordisk. It is similar to NovoRapid, but has some additional ingredients that have improved the speed of onset. Official trial data shows a relatively modest improvement, but there has been much excitement in the DOC, and the early experiences shared seemed to suggest some people saw significant differences with faster action and a shorter duration.

I was keen to try Fiasp to see if I could do away with my reasonably lengthy pre-bolus at breakfast and lunch (taking insulin at least 30 minutes before beginning eating) and also to see if I could see improved/faster action of correction doses. This blog is my n=1 trial of 3 vials, I cannot say how Fiasp would work for you or anyone else, but this was what happened when I tried it.

But before we get started...
(jump to the Fiasp bit here if you're in a hurry, but you'll be missing out on some *sparkling* blogrambling)

The mystery of the missing insulin #1
When I spoke to my pump clinic about getting Fiasp they could not prescribe it because it was not yet on the hospital formulary. I had to get it from my GP who could use the PIP code to prescribe it directly. I planned to try it for 3-6 months to give me long enough to experiment with it and give it a decent go. I generally get 3 vials at a time when I order insulin. each u100 10ml insulin vial contains 1,000 units, which at my general Total Daily Dose of 30-35u should last me around a month each, more or less.

Except that they haven't.

Looking back, my insulin use has been pretty much as expected, but I used up those vials in 57 days not 90. And I am always careful to draw out every drop out of each vial, conscious of the huge privilege it is to live in a country where insulin is provided free for me by the NHS.

I had never realised before how much insulin prescribed to me goes unused. If I lived in the US with their absolutely horrendous price issues it would make a huge difference. Every set change for my insulin pump requires the tubing to be filled, and at the end of the site's life, that full tubing is discarded. I'm careful to only fill reservoirs with just enough insulin for their 3-day life, and have run them to all but empty more than once, but even then, there is a substantial measure of insulin left at the neck which you can't get to.

What this has showed me though, is that if planning a lengthy trip away, I would need at least a third more insulin than I might have guessed.

Oh and one more thing before we get going...

The mythical 50:50 split and other 'rules'
Immediately before switching - NovoRapid doing very well
I knew that changing the insulin I was using would most likely involve resetting a lot of things - ratios, factors and so on. So I decided (looking back this was probably not one of my brightest moments) that it would provide me with a useful opportunity to finally experiment with that mythical 50:50 split that some Healthcare Professionals seem quite keen on where exactly half your insulin is used for basal and half for meals. Along with the 500/100 'rules' that often get mentioned to me in clinic. I have always thought these 'rules' to be useful starting points - academically interesting, but no real substitute for systematic (and repeated) self-testing of basal insulin with fasting basal tests for example. Here was a chance to see how it worked for me.

So when I switched to Fiasp, I took an average of my Total Daily Dose (TDD) over the previous 30 days and split it exactly 50:50, then set a flat basal profile to spread that much insulin over the 24 hours. Apparently in most people with a functioning pancreas, the body uses half the insulin for food and half for background. Quite how they have worked this out is beyond me. And I've always thought, "Well surely, doesn't that depend on what you are eating??". But nevertheless the 50:50 thing still floats around and some HCPs raise eyebrows when your split is more 40% basal to 60% bolus as mine is.

Part of the reason why I half-thought this might be a useful experiment (apart from my own curiosity and thinly veiled desire to prove that it was nonsense and wouldn't work for me), was that I had read accounts by a few people who had tried Fiasp already that found it had a shorter action. By boosting my basal split to half of my TDD, I reasoned I might soften that out a little. Take less insulin with each meal, but still have some being fed-in continually in the background that I could dial down with a Temporary Basal Rate during exercise/activity.

Additionally I was wearing a CGM sensor during this period, and could watch what was going on, plus I had Smartguard to catch me overnight, just in case.

The first morning of my changing-absolutely-everything Fiasp trial showed a dramatic drop overnight - caught by Smartguard and low prevented, but enough to confirm my suspicions that a significant hike in overnight basal insulin would cause me problems going forward. Undeterred, and wanting to give the experiment more of a go I adjusted the pattern to shift some of the basal insulin into the daytime and keep the pattern at 50% of my TDD.

500 and 100 rules
The other half of my Great Big Reset experiment was to use the 500 and 100 'rules'. These are a suggestion of calculating your insulin:carbohydrate and correction factors using your TDD as a starting point:

1u of insulin covers: (500 / TDD) grams of carbohydrate
1u of insulin lowers BG by: (100 / TDD) mmol/L

The correction factor always works out very similar to the one I generally find works OK for me, but the meal ratio is always a bit of a surprise. More than once in clinic when the subject of hypoglycaemia has come up a calculator has been tapped and mentions made of what my ratio 'should be' according to the 500 rule - I often use 1:10 and 1:11, the 500 rule suggests 1:15. I've always been of the opinion that if my meal ratio were 50% out, I might have noticed, but this Brave New World was an opportunity to have a go and see what happened.

'500 Rule' boluses really struggling
After a couple of days I took stock. I had been experiencing a lot of high glucose alarms and had needed to dose several extra corrections to bring my levels back into range. Hilariously when I looked at the splits between basal and bolus I noticed that the extra corrections I had needed pushed me back almost exactly to 40:60 rather than 50:50. My diabetes can be extremely stubborn sometimes.

Additionally, I soon realised that the 500 'rule' was massively messing with my attempt to aim for 50:50. Even though I was using my TDD as a starting point, I simply do not eat enough carbohydrate most days for the 500 rule to generate half my TDD. I usually eat around 130-150g of carbohydrate per day. Don't get me wrong... I'm no sandal-wearing low carb evangelist. Sometimes I can eat 120g of carbs in a single meal - but on the whole, I find around 150g is all I need, and helps keep my BG a little more stable. The 500 rule seemed to assume I would be eating 250g of carbs a day. Which I can do, but carbier days are often the less predictable ones in my experience.

So I gradually began to tweak my basal profile and opted for more of a mid-point for meal ratios, and the experiment continued. I lasted around a week before I threw the towel in. I only hit the mythic 50:50 split on one or two days (about as many as I do using my own system to be honest). Most of the days with the 500-rule-ratios involved significant corrections due to rising levels, however quickly Fiasp may have been working. And more often than not these pushed my basal:bolus split back to where it normally sits.

Finally! The Fiasp part of the Fiasp Post
If you've waded through these ramblings so far (congratulations, some sort of perseverance medal is clearly in order) you will understand why I am choosing to pretty much ignore my first week's experience with Fiasp.

Looking back at that first week though, I will just briefly mention in passing that we were away on holiday and so there were a good few treats to test Fiasp's rapid action. I was also experimenting with not pre-bolusing for breakfast or lunch. Early results were promising once I had tweaked my ratios a little. Doses for other things, like white bread, ice cream, cake, beer did seem to be starting to act more rapidly, and where I'd misjudged things and was dosing for corrections they seemed to be starting to act within 25-30 minutes rather than my expected 60 minute wait with NovoRapid before I see much BG reduction.

I think it's fair to say that Fiasp had its work cut out because in the months before trying it, partly powered by the occasional CGM sensors I've been running this year, NovoRapid had been unusually cooperative. Many weeks with more than 80% of sensor readings in range and with almost no minutes below 4.0mmol/L.

Faster acting
After my slight false-start and once I had my ratios and basal back to more like where I would expect them to be I began to find my feet with Fiasp. During this period, here's what I found:
  • I did still need to pre-bolus, but only about half as much. Perhaps 15-20 minutes at breakfast and 10-15 minutes at lunch. Much more than that and I risked dipping low before the carbs kicked-in.
  • Corrections were acting faster, just as I hoped they would. This meant that my errors were resolved more quickly
  • Meals where I would not normally need to pre-bolus and where I'd expect reasonable results from an 'all up front' approach I actually needed to delay the meal insulin. Setting all or part of it as a square wave/dual wave/combo
  • Smartguard occasionally mangled these square and dual waves, cutting basal insulin and stopping the remaining bolus following a small dip in BG and just as the carbs began to hit, resulting in the dose only being delivered later on when I noticed what was happening. This was intensely frustrating.
  • The insulin action did seem to be shorter than NovoRapid for me, or at least the way that Fiasp makes more of the dose available sooner meant that the tail was less pronounced and I reduced my duration of insulin action to better reflect 'insulin on board'
  • Breakfast was my biggest challenge. Lower carb weekday ones (15-20g carbs) were relatively OK, but bigger weekend ones (45-50g carbs) were a nightmare. At some points in the year I can find I have to add an extra mini-bolus to account for my liver dumping glucose when I crawl out from under the duvet (even though my basal pattern always kicks-up at this time), but even that tried and tested strategy didn't keep me out of the teens after breakfast at the weekends. In the end I used a surprisingly strong bolus ratio that scaled the doses upwards where I was eating more.
Fiasp performing pretty well at 3-4 weeks in.
Finding the Fiasp sweet spot
There was definitely a point, when I'd been using Fiasp for about 3-4 weeks where I began to see distinct potential. There were still some horrendous numbers to be found, but there were some great successes too. For example, a Tapas meal out one Sunday with delicious breads, patatas bravas, beers and all sorts of incalcucables that was bolused late, in a series of guesses and to correct my earlier underestimates of carbiness where I could actively see Fiasp's faster action helping me out.

It was also at this time that my results around breakfast greatly improved, which helped a lot in improving my time-in-range.

What it made me realise, I suppose, was that after something like 15 years of using NovoRapid I had memorised a lot of 'exceptions to the rules'. Little tricks and strategies that I use, almost without thinking, to work around NR's particular activity profile and my individual BG response to different foods. When switching to Fiasp, I was needing to re-invent a lot of these, and discover a whole lot of new ones. If the switch was to become permanent, it would take time to build up this knowledge.
Things improving around breakfast time with Fiasp

Increasing resistance and the mystery of the missing insulin #2
Unfortunately my successes were fairly short-lived. I can see the Standard Deviation (how spread apart my BG results were) taking a leap upwards after about 10 days of beginning to feel I was making progress. During this phase of my Fiasp experiment my basal and bolus requirements seemed to be heading inexorably upwards once again (they had kicked upward after a couple of weeks, but I'd seen that happening to others and didn't stress about it too much). At the same time I was finding my earlier shorter pre-boluses less and less effective, and had more or less reverted to exactly the timings I would use with NovoRapid. Additionally, I no longer needed to dual or square wave those well known 'all up front' meals as I had in the first few weeks.

Even more perplexingly, rather than acting more rapidly, sometimes my corrections of high BG values seemed to have no effect at all. I would be watching a sensor trace waiting for a high or rising BG to be corrected and nothing would happen. I began to throw in 2u and 3u speculative 'turnaround' corrections to try to halt a rising BG only to see it continue to rise, and where I was expecting to have to mop-up the excess insulin with carbs later, the dose seemed largely to disappear entirely.

As an example, in the image you can see my BG rising after an early evening meal. The blue dots along the bottom represent corrections. The first, before 7pm was in response to an 'alert before high' which indicated I would be rising to 11mmol/L within 30 minutes. I gave a small correction (0.7u) which aimed to take the edge off the rising BG. Over an hour later, not only had the remainder of the meal bolus not reduced my BG, but the additional correction was not doing much either. By 8.20pm or so I was getting a little frustrated and bolused 3u planning to watch and wait -  mopping up with some tasty carbs once my BG had begun to drop. In the early days of Fiasp I would have expected even a modest correction to begin to lower BG within 30-45 minutes (unless immediately after eating), but over the next hour my BG continued to rise. The two corrections already on board almost doubling my meal dose. A further small correction at around 9.20pm did finally provide some BG lowering effect and I went to bed mid-range after a small snack. For anyone wondering about the condition of the infusion set - it returned to much more expected behaviour overnight and the following morning. But it was odd events such as this that rather cast a shadow on my Fiasp experiments. I began to opt for 3u and 2u over-corrections fairly often.

I was also increasingly aware of a stinging sensation at the infusion site. Not always, and sometimes stronger than others. But many infusion sites were noticeably tender to the touch.

Losing faith with Fiasp. Averages and SD rising.
Calling it a day
It was about this point where I decided that Fiasp was not going to work for me. I was nearing the end of the third vial of Fiasp and needed to put my repeat prescripton request in to restock. I decided to return to NovoRapid.

I am sure I could have made it work given enough time, but I was losing trust with it and finding it not altogether reliable or predictable. This was relatively manageable when I was wearing a CGM sensor to keep track of where doses were not behaving as expected, but I generally only use sensors occasionally and I really need an insulin that I can trust while I'm not able to watch it like a hawk.

Ultimately, I had wanted to try Fiasp to reduce or remove the need for pre-boluses, and to improve the speed of action of corrections. I had seen some evidence of these early on, but not for several weeks. And those positive attributes had apparently been replaced by a less than reliable action.

I am quite disappointed if I am honest. I continue to see lots of accounts of people getting on really well with Fiasp, enjoying lightning speed and seeing significantly improved post-meal numbers. I have also seen other accounts very similar to my experience though. So it seems that Fiasp may be an insulin that just does not work well for some people.

But for me - despite all its faults, NovoRapid has brought an immediate relief and return to significantly better results. Well... for the time being at least!


Posted by on Monday, 31 July 2017

Does Abbott Freestyle Libre improve Hba1c? - belated thoughts from Dx Amsterdam

More than 20 bloggers from across Europe gather for Dx Amsterdam

"I love deadlines" (as Douglas Adams used to say) "I love the whooshing sound they make as they fly by."

If I had set myself a reasonable time limit to write some reflections on my time at Dx Amsterdam in June I suspect it would have been rather sooner than this. But even though this is rather late, there were some bits and pieces I picked up over that weekend that have stayed me since, and while I now have several other blog posts jostling for position in my head, I've decided that this post is better late than never.

I was really lucky to be picked out of a hat and selected as part of a 7-person UK contingent at Dx Amsterdam, which gathered 21 bloggers from the UK, Ireland, Netherlands, France, Germany, Spain, Belgium, Poland, Greece, Sweden and Turkey in the wonderful city of Amsterdam for a weekend of friendship, conversation, shared experiences, learning, support, wobbly bike rides, and the dreaming of dreams.


There were lots of opportunities for us to chat and exchange experiences together over the weekend, and just like the Dx event I was lucky enough to attend in Stockholm there were useful and inspiring presentations - from practical things like taking better photographs, to the astonishing and humbling story of Claire Lomas and how she adapted to life paralysed from the chest down.

Mercifully, Abbott are careful at these events not to bombard us with announcements and product news - it's really more of a chance for us to learn and grow together as bloggers and to make connections across international boundaries. However once you've got some real actual Abbott peeps in a room, there are inevitably questions you want to ask and things you want to find out. For the UK folks, there was a brief breakfast meeting on Sunday which included updates on the tantalising prospect of getting the Freestyle Libre available on prescription on the NHS, which seems to be creeping ever closer (including a new campaign by Diabetes UK as the discussions and tortuous process seem to be nearing a final decision). It is interesting that the Libre is already fully or partially reimbursed in 9 countries across Europe.

But the thing that my mind keeps returning to following the weekend was the presentation of analysis Abbott have done on their anonymised real-world data which Tim gave on Saturday afternoon.

As Libre users may (or may not!) be aware, when they connect their reader to the computer to access the PDF reports and swanky graphs a de-identified anonymised copy of their data is also uploaded to Abbott. I remember some people being a bit huffy about that early on, but if I'm honest it has really never bothered me, as long as the data is completely anonymous (which it is) I don't mind Abbott having a set of my random BG craziness to see how their gizmo is working out in the real world.

And it is an absolutely massive dataset, almost 400 million data points from 55,000 users over 20 months.

What I found really interesting was some of the observations Abbott were able to make by analysing and filtering the data.

Average scans per day
The average number of scans per day per user is much higher than they had originally imagined. On average Libre users check their glucose 16 times every day. 16 times! A number of checks that would be unsustainable, or at the very least very uncomfortable and tiresome with traditional fingersticks. Additionally, while there are a few people that hardly scan at all, there is a real cluster at that 16x a day level, and many, many users who check between 20 and 30 times a day or more.


More scans associates with better outcomes
Abbott were also to stratify their results and confirm that people who scan more often are more likely to see reduced glucose variation when compared to those who scan less frequently. There are also observable improvements in the three biggies: reduced hypoglycaemia, reduced hyperglycaemia and increased time in range.

What about HbA1c?
Well, of course, the truth is that because of the way these data were collected no one can really know for certain (though there are plenty of anecdotal accounts from people who report their HbA1c reducing alongside Libre use). But I still find the data compelling. After all HbA1c is really only a proxy for glucose management, average blood glucose and time in range. The very things the Abbott real-world data is actually collecting.

At the moment HbA1c tends to be the focus of many academic/research studies at least in part because it is relatively easily collected, standardised, and has a long history going back to DCCT analysis that offers the promise of reduced complication risk. It also makes it easy to compare multiple studies by using the same HbA1c outcome. But I know I'm not the only person with diabetes to know that A1c is often a pretty poor indicator of what is actually going on day-to-day, and that the number you get does not always reflect the average BG or variation you are seeing.

I am aware that some will see these 'real world' data gathered by Abbott as a poor relation when compared to a 'proper' randomised controlled trial. There is no proper structure, no control group, no baseline data, no monitoring and observation of what is going on, no assessment of the relative education and support (or otherwise) of participants. But actually I think personally that is exactly what gives these data their real power. These are the numbers of ordinary individuals at all different stages of their diabetes journey, living messy, complicated real lives with type 1 diabetes. The only common factor is that these people wanted more information to support their diabetes management - and the more information they had, the better they were able to manage their blood glucose.

In a sense, of course, this is a self-selecting group. People had to be interested enough, and committed enough to fund the use of the sensors for as much of the year as they could manage. And people who couldn't get on with Libre are likely to have dropped out - but drop-outs and careful selection happen in clinical trials too. So while I don't for a minute imagine that chucking Libre sensors around like NHS-funded Smarties will instantly solve all the T1 diabetes woes in the UK, I do find these real-world data very encouraging and empowering.

Living with type 1 diabetes is complicated. And for many people, additional information about what their BG is doing 24 hours a day can go a long way to help them make better management decisions.

Disclaimer: Abbott Diabetes sponsored my attendance of Dx Amsterdam including flights, accommodation and the programme they organised. They also treated us to lovely meet-and-greet nibbles on Friday night and a slap-up meal on Saturday evening at Amsterdam's Van Pufflen restaurant. I was not paid to attend and I have not been asked to write this or any other post about the weekend or the Freestyle Libre.