#Fiasp and #OpenAPS – is this what we’ve been waiting for? A week in the life….

I’ll cut to the chase. This has been one of the more interesting weeks in my life with diabetes. There aren’t many occasions when you can say that you’ve put two things together and seen immense results, but I think this is one of them. It gives me more hope for future treatments, and I don’t think this is the end game, but it’s a great step in the right direction. Is it what we’ve been waiting for? Not quite, it’s not a cure after all. That’s still ten years away… But in terms of injectable insulins? More please!

I put my hand up and admit that when I saw the data on Fiasp from Novo at EASD last year, I wasn’t all that impressed, in fact, I said:

“the BCLispro data seems way better than that of Faster Aspart”

I then stated I was really looking forward to the BC Lispro and kind of ignored the Faster Aspart. Roll forward four months and we learned that Lilly had ended its association with Adocia, so there was never going to be “BC Lispro”…

Then the release date of Faster Aspart came about, and as I’ve previously stated, I went all out to find it and get going on it. And now, after a week, I feel that I am getting to know it. There’s still plenty of learning going on, but I’m more familiar with how it and I interact, and how it works with my OpenAPS set-up.

So what have I learned in this week? A few things.

1. It gets to work more quickly. Substantially so.
2. The effective tail feels less, although technically under clamp testing it isn’t any shorter.
3. You really can bolus just before eating with little detriment.
4. I haven’t needed to adjust my basals on the pump.
5. It works really well in a hybrid closed loop system
6. Changing insulin means you do have to think about what you’re doing and monitor what’s going on.
7. I’m not 100% sure how you pronounce it. ‘Fee-asp’ or ‘Fye-asp’?

I’ve talked about it being more rapidly available and feeling like it has a shorter tail at great length in my previous posts on the topic, so rather than dig too much into that, I’ll discuss some of the other things I’ve tried this week to see how it works.

Whilst the Krispy Kreme test proved to be really effective, the “Salted Caramel and Walnut” cake test was less effective. I hugely underestimated the carb content of it and spiked, however the correction dose with Fiasp took effect more quickly, which is a result. So while Fiasp is fast, it’s not a magic bullet if you don’t get your carb count right.

I also enjoyed the “accidentally delayed lunch” effect, where I bolused, someone came to my desk and interrupted me for twenty minutes, and as I was starting to eat, my glucose levels were heading down faster than I would have liked. A learning point.

As I mentioned in my first post about Fiasp, I experienced the “double bolus blow” as well, where I bolused roughly two hours after eating for eating something else and dropped in an unexpected way. I’ve not been able to repeat that experience, as I don’t really fancy another pair of Krispy Kremes, but it’s one to have at the back of my mind, just in case it happens again. I’m aware that another user in their first couple of days saw some unexpected behaviour with insulin stacking, so it seems to be one to watch.

In terms of basal rates and ratios, I’ve not really seen much in the way of change. My basal slots have remained the same and so have the associated ISF and Carb ratios. I ran without the loop at first and although I thought they may have been wrong, the fasted period I undertook suggested otherwise. Now though, I have OpenAPS Autotune and Autosens looking out for variations, so it becomes a moot point.

But overall, it’s been very positive. Comparing the aggregate data from February (which was entirely on NovoRapid) with the last week on Fiasp, we see that the statistics look good.

First up, February. Not really especially low carb, and the low limit is 3.8 and high limit is 9.0. There’s also a duff sensor in there, but the average low is too low on that one.

Then the last week on Fiasp:

What surprises me most is the reduction I saw in hypo amounts over the week. I mentioned that I thought it had less of an effective tail, and I’ve a feeling that this shows through in the stats, but at this point, I also have to suggest that it being the first week and me being more aware may also have played a part.

How does this look when we look at the AGP graphs though? Well February looks like this:

What we learn is that post lunch I’m not great at managing prolonged glucose release after lunch and that effect lasts for a while, but most of the time, OpenAPS does a good job of keeping me in line. For the last week I’ve seen this:

Okay, it’s only a week of data, but what really stands out for me is that the 25th-75th percentile band is generally slightly tighter and that overnight, the 10-90th and 25th-75th are almost aligned. It seems that “faster in” is making the hybrid closed loop adjustments work more effectively, and reducing the swings that occur. You can also see that one day’s cake test affects the bands far more on a single week AGP….

Overall then, off a very small set of data, it looks as though Fiasp is having a beneficial effect on glycaemic variation.

Testing the Hybrid Closed Loop

Now as I mentioned, Fiasp really does work well with a hybrid closed loop. Back in October, I suggested that faster insulins might enable fully closed loops, so it would have been remiss of me not to test this and see how it worked with OpenAPS. For a while I’ve been testing an alpha version of the OpenAPS algo, and I thought that testing it with Fiasp was too good an opportunity to miss.

So I tried the “Meal Announcement” test.

This works by letting the system know I’ll be eating soon with an estimate of carbs/carb equivalents (CE) (i.e. adding a value to take account of protein), then letting it get on with working. The test I gave it was a fairly tough one. An announced 85g of carb/CE, sometime in the next couple of hours. Then leave it to get on with managing my glucose levels. This is what it did:

The area in the blue circle is the loop doing its thing. The actual macros for the meal were:

• CHO: 63.6g
• Fat: 29.8g
• Protein: 51.5g

The excursion from the meal was from 6.8 mmol/l at the start of the announcement to peak at ~10.5 mmol/l about two hours after eating, dropping back to where we started four hours later.

Whilst those are not numbers that I’d want every day, they are an exciting start to the idea of a truly closed loop in the Open Source community, and I feel they demonstrate that faster insulins are the key to this.

The Chicken Katsu Curry test – really pushing it

But if we are to truly understand how these insulins work, we needed to see the effects with a “difficult meal”, using it more normally. I elected to use Chicken Katsu curry as my test subject because I like it and because it’s really difficult to manage being high carb and high fat. And I eat it with brown rice, just to add to that. This is the offending meal:

With the associated Wasabi peas, that’s a lot of carbs in one hit. 160g to be precise. Previous attempts to tame this beast with NovoRapid have looked like this:

Where I pre-bolused too much, too early (16u), so didn’t spike too much but was hypo by the time I ate and this:

Where I struggled to balance the insulin timing with the food (and this is more typical, sad to say) resulting in a peak up at around 14 mmol/l, which is just too high. You’ll note the velocity of the initial climb thanks to the sauce, then the later, slower rate of change thanks to the combination of ingredients.

This is a real challenge, so given the same algorithm, we got this:

 

This was handled with a bolus of 12u beforehand and the loop left to get on with it. Beforehand means “I initiated the bolus about 5 minutes before I was due to eat but it took five minutes for the old Medtronic pump to deliver it”. I’d had eating soon set up for 90 minutes previously so I had some background insulin already on board. I bolused when I did because I really didn’t fancy having 12u of this stuff on board with a gap before eating. I know roughly what my absorption time is, and frankly, the full force of 12u hitting me with no food was not an idea I wanted to consider.

It worked remarkably well, and the insulin delivery that resulted kept the Katsu well under control, staying below 9 mmol/l. I take that as a win. You could interpret the delivery, if you were using a non-clever pump as a Dual-wave/Combi bolus with a 70/30 split of the insulin, but the key for me was that it was fire and forget, so I didn’t need to consider that.

You also see the spike at 19.00. That’s beer that is. The 15g was an announced beer, but for the two pints to be cut off on their inexorable rise so quickly was impressive. Especially as I’d bolused for the second after drinking half of it.

Does it affect hypos or anything like that?

After a week, I’ve a couple of observations in this respect. There is a slight change to my hypo symptoms. it seems that with the Fiasp, I’m noticing them earlier and it feels a little like I’m drunk. As I mentioned previously, the use of a loop to tighten my variation has resulted in feeling hypo symptoms earlier. Well now it’s earlier still. I think it’s too early to say whether this is a permanent effect though.

That’s really the only non-numbers related item I’ve noticed. Other than being slightly freer. What do I mean?

Having the knowledge that I don’t need to pre-bolus in the same way for food is freeing. You do it just before you eat and don’t suffer the same level of consequences as when using the current fast acting insulins. That’s pretty liberating, and something that every type one can benefit from.

So what do you take away from your experience?

Firstly, what looks like a small difference on paper makes a bigger difference in the wild. I’m surprised at how much of an improvement this really is. I wasn’t expecting it, and the study data didn’t suggest to me, at least that it would make much of a difference. But it seems to.

I also think pumpers will benefit from this more than those using MDI. Why? The change in the insulin tail. I’m not saying that you can’t use it for MDI, but the ability to set dual wave boluses on  a pump is enhanced by the way this stuff works. I can imagine that on MDI it’s going to require more jabs for some of those trickier meals see the Katsu curry).

Alternatively, it may be worth modifying MDI for those who want to to allow Faster analogue and something like Actrapid to be taken together for those Pizza moments. Maybe that’s something that Novo might consider. A kind of Fast/Medium insulin mix to mimic Dual Wave bolusing for MDI.

I am pleased that I pursued getting hold of this. It’s better than I had hoped and I think it does make a difference with the closed loop and the proximity of the action to the decision. As a result I look forward to whatever happens next. Whether that’s HDV-Humalog from Lilly/Diasome, or Adocia teaming up with Sanofi or Novo to make Bio-Chaperoned faster insulin, either would be another step on the path.

But for now, I’m happy that, for the first time in 20 odd years, we have a noticeably faster insulin, and not one that’s only a bit different. I just hope that the next steps on this path don’t take another 20 years.

Maybe we need to make those Pharmaceutical patents shorter to encourage faster innovation….