What type of insulin is best for AAPS insulin pumps?

I have been using Humalog with my Omnipod using the AAPS algorithm. I was wondering from your experience from using these automated algorithms. What insulin performs the best. I just dont think Humalog is the quickest when it comes to fast acting insulin.

I don’t think there’s a such thing as “best for AAPS”, given that there’s a setting to specify exactly what type of insulin you’re using so the insulin action algorithm matches your reality. In the config builder, under the insulin setting, do you have “rapid-scting 0ref” selected?

It’s more about timing things correctly. Do you pay attention to your IOB graph vs your BG? If you’re BG is spiking before the insulin action graph spikes, then you’re just not pre-bolusing soon enough, or not setting an eating soon temp target appropriately. It doesn’t necessarily mean your Humalog isn’t a good match for AAPS, but rather that you’ve not yet learned to match your behavior to the insulin.

Now there may absolutely be a better insulin for YOU, if not specifically for AAPS. Humalog, Novolog/NovoRapid, and their generic equivalents are considered rapid-acting insulins. (Rapid compared to the Humalin R that was previously used.). Medical and insurance people will tell you Humalog and Novolog are identical and interchangeable, but the actual patients know this often isn’t true. Personally, I DESPISE Humalog. It’s like injecting water. Novolog works so much better for me, but there are others who prefer Humalog. Now we also have newer insulins which are considered to be Ultra Rapid-Acting, because they work even faster than Humalog and Novolog. This category includes Lyumjev, Apidra, and FIASP. The base insulin is very similar to Humalog or Novolog, with extra added ingredients to make them work faster. So you if you have a preference between Humalog or Novolog, you might want to stay within the same manufacturing family.

There isn’t actually a massive difference between the rapid-acting insulins and the ultra rapid-acting ones. The newest ones peak in activity about 15 minutes sooner, though that varies greatly between individuals. Many report no difference at all, or that they work noticably faster in the initial weeks/months, but quickly lose all effectiveness shortly thereafter. Maybe that 15 minutes is really important to you, though. The only way to know if you like them is to try them for yourself. Unfortunately, depending on your insurance, access may be limited. I would like to try Lyumjev, since it seems to be the most favorably received, but it isn’t on an accessible tier on my insurance formulary. It requires a prior authorization and to get that you have to demonstrate failed treatment with the preferred insulins first. Since I’ve got good numbers and no undesirable side effects, I don’t qualify.


Tha k you for the detailed response. I do have that setting correct in config. The issue is what insulin the insurance will cover. I will look into Lyumjev. I think that one is covered.

1 Like

It really depends a lot on YOU and what your body likes. Everyone is a little different and you’ll have to experiment. :grimacing:

My experience is similar. I pay oop for novolog since humalog is on insurance but not novolog. But use a discount card for novolog.

Humalog works faster than novo log for me. My insurance dropped humalog for a while and forced me into novolog. And then after a few years they switched back to humalog.
In reality they are both very similar and I’m happy to use either.

I haven’t tried either of the fastest ones yet, but they need to solve the clogging and all that before I would bother pumping w it.

1 Like

The net effects of all U100 insulins are the same over their durations and there’s never been any proven difference in outcome from using one vs another. There’s a slight difference in duration before clumping/occlusion but it also varies between pump and infusion sets models. Potency doesn’t vary enough over 3 days that the algorithm can’t compensate for it.

The biggest variable is still the users, and those haven’t changed enough in the past 30 years that available insulin has made a significant difference.

So I’d look at cost effectiveness. Since cost parity still doesn’t exist, I’d use the least expensive variant that the pumps were tested with in documented clinical trials.

I used Novolin R and Humulin R in the past and saw no significant difference. My choice was dictated by availablilty and cost. Changing to Novolog was just more expensive, and when I lost insurance coverage, changing back was unit for unit.

Today I use Novolog rather than Humalog with an AIIS because Lilly bought the license to produce insulin in the 1920s, and as soon as it could, “forgot” its agreement to keep the product affordable. Then it intentionally made it impossible for any other company to license and produce the expired patent variations. So other companies had to reverse engineer them from limited information and submit them for approvals as new products. A company with a long history of putting profits ahead of pubic health is not a company I can support when I have another choice.

I need to use extended boluses, so there’s no advantage to me using a faster onset insulin. I’m not aiming for “normal” BG hour by hour. I’m content to keep in range,avoid hypo and have an A1C low enough that my endo was initially concerned.

I’m not an early adopter when it comes to my life, so I’m not using the AndroidAPS software until its commercially adopted and implemented but if it were, I’d have no qualms about using Novolin Regular or any of the faster acting insulins provided the algorithm can be tweaked for their onsets and durations.

FWIW, I’ve been pumping Fiasp for about four years and haven’t experienced any of those issues. I was a little concerned about it with switching to a Tandem pump because I’d Heard Things, but It’s been three weeks and no problems. I asked about it during the decision process because the informational materials specify that you should only use Novolog or Humalog, but the representatives assured me it wasn’t really a problem, just an FDA thing. My Tandem trainer has been using Fiasp for years in her T:Slim pump, so that was reassuring as well.

Re Fiasp “fade,” that’s obviously different sort of issue, being dependent on individual metabolism rather than the pump mechanism. Enough people report it that I have to acknowledge it’s a thing, but for my part I still find it faster than the two leading analogs. Still not as fast as I’d like sometimes, but I was at a point with the others where I was waiting 45 minutes or more before my curve would start to bend. They seemed much more susceptible to things like stress level. Fiasp seems much more stable and reliable in that respect for me.

Everyone has a different world view and imo has the right to do anything to themselves that doesn’t predictably injure another person. Each of us has biases formed by experience, knowledge, and our tolerance for risk.

My tolerance for risk is moderate (and gets lower as I get closer to EOL). If I believe it’s necessary, I’ll go “way out on a limb” but cautiously, one tested step at a time. I don’t trust things that can’t be measured and I don’t use measurements that don’t have high correlations to a measurable outcome.

I don’t trust anecdotes. My skepticism about things that “someone said” is why I don’t rely on them, whether they support what I want to do or oppose them. When it comes to my health. I only consider anecdotes after I run out of options, when nothing tested by “the system” has worked for me.

I have gotten disturbingly good results using tested off the shelf technology approved for use in managing diabetes. That are disturbing because I’ve been told that they are unusually good while I’m relatively lazy and not smarter than average. So I’ve wondered why so few people who have been using the same technology don’t get better results than they do. I’ve wondered why those who get results as good a mine aren’t satisfied to wait for the next stage of technology.

I believe the former lack a combination of knowledge and motivation, and the latter are a combination of risk takers and those seeking instant gratification. I think of the former as settlers and the latter as pioneers. I believe I’m in the middle, a successful developer, and that we need all three kinds of people.

As a developer, I look for studies where the primary outcome studied was closely aligned with what I’m interested in. The algorithms developed by the AAPS and Loop have both been evaluated for safety and effectiveness in studies, but the effect of using different insulins with them hasn’t.

I have no doubt that every fast and rapid insulin on the market is safe with AAPS because they are safe when used with MDI, and used by people who pay far less attention to what is happening than any sensor driven system can. The AAPS and Loop system were developed by insulin-dependent PWD who wanted better results than they could manage manually. If those systems couldn’t do that, then they would not have been adopted and continued to be used by thousands of pioneers. Stupid pioneers who don’t quickly learn from experience what doesn’t work don’t survive,

I’m not a pioneer. I let them take the risks and follow in the trails of the ones who have been the most successful.

I have no objective basis for believing that outcomes are improved by further decreasing BG variability of a PWD who has an A1C of 6.5 or lower, SD under 20% , and an average BG level lower than 140mg/dL. And all those could be accomplished manually using human synthetic insulin, which is biologically closer than any of the analogs to what our bodies are evolved to use.

I don’t care about changes in numbers unless they are sensitive indicators of safety or outcome. My A1C only served to help confirm that I had diabetes and warn me if I was getting out of control. It never told me why or what I should do.

Having a low BG detection alert and high rate of change alert is useful for maintaining my safety, but if an automated system’s algorithm can reduce highs and lows to keep them in a safe range, I need hard evidence that any change I make to to adjust it will be beneficial- to me. I did a series of very careful experiments to tune my system to my body and my lifestyle to my health. All these adjustments were based on known science using 3rd grade math.

I’ve never found a confirmed study that seriously studied PWD who manage their BG and determined that using any one of them in an automated system will make a significant difference in average glucose level, SD, in % high and low - or outcomes.

So, as an developer, an engineer, I look at what we know, what works, and its cost effectiveness.

Pioneers are welcome to experiment to the limits of their resources. I just hope that they find a way to share their collective experience through a systematic study so it becomes reliable evidence, not anecdotes.

I think those points are well taken, but I would argue for an additional measure of success that is necessarily subjective: quality of life. A few years ago someone here linked to a study indicating that the outcomes for T1s using basal-bolus MDI were not all that markedly better than for those using the old R/NPH regimen. Having been on the latter for twenty years before finally getting switched to carb-counting with Lantus-Novolog, the difference in quality-of-life was enormous. I remember sitting there literally weeping at my desk the day I was able to keep working past my previously hard-wired lunch time and nothing happened!!! I still have some bitter feelings toward the non-specialist PCPs I had (HMO wouldn’t do a referral to an actual endo) who kept me on the old regimen for years, despite my pleas, because “Your numbers are fine let’s not change anything.” I.e., “We’re deeply uncomfortable with this whole insulin-T1 thing and let’s not rock OUR boat just cuz you want to try something you think you might like better.”

Switching to Fiasp was not nearly so epochal an evolution for me, but it does make things easier in terms of meal-blousing in time to limit post-prandial peaks. I haven’t kept exhaustive data on it but I could see the effect in my CGM tracks compared to how I was responding on Novolog, and those don’t seem to have altered significantly since going on it, so I prefer to stick with it. As a quality-of-life matter, it just seems to make things a bit easier.

Funny, I’d expect as one has less to lose, risk tolerance should go up!

But then as I further consider this tiny snippet (taken out of context from your comment) it is clear that the risk would remain great and the expected benefits reduced since there’s less time to realize the benefits.

Anyway sorry, didn’t mean to derail the conversation.

Done and done.


I actually linked the wrong study up there. I had meant to link this one from the New England Journal of Medicine.


Those were observational studies, limited, uncontrolled and unblinded as they would have to be, with unsurprising findings. Something that automatically does what a person following a procedure should do, will perform more reliably minute by minute 24 hours a day than a person could.

What they studied was safety and effectiveness of the algorithms not comparative effectiveness of different insulins, the core topic of this thread.

That nobody has done a long or intermediate longitudinal study comparing insulins to show whether there is any long-term difference on outcome is the basis of my skepticism of all of the new insulin variants. Onset speed- acceleration - is only one metric of a complex system. A dragster isn’t a better choice for an endurance race than a Honda Accord.

The reason I switched from MDI to using a Tandem pump was to reduce the effect of my inconsistent human management on my nocturnal glycemia.

Imo We humans, not the insulins, not available technology have been the reason that there has been virtually no improvement in outcome for 30 years. There are many times in history and fiction where human ability to use technology effectively has been more influential on outcome than technology.

"It’s not the plane it’s the pilot." Captain Pete “Maverick” Mitchell (fic)

A more recent meta study shows that there has been no obvious improvement in outcomes between 1988 and 2020- a period where multiple analog insulins were introduced with differences. All were FDA-approved as U100 products having the same glycemic effect over their durations, NOT any claim of superior outcome on complications or mortality.

2022 JAMA Trends and Disparities in Glycemic Control and Severe HyperglycemiaAmong US Adults With Diabetes Using Insulin, 1988-2020.pdf (933.9 KB)