Glucagon me… n=1 experiments with microbolusing glucagon part 2: quantifying use

Glucagon me… n=1 experiments with microbolusing glucagon part 2: quantifying use
Glucagon me… n=1 experiments with microbolusing glucagon part 2: quantifying use

In part one of this microseries, I looked at lifespan, dose sizes and expiry of the GlucaGen sets that we get in the UK, and cost around £20 per set to the NHS.

Whilst this is all important information, it was included as part of a larger experiment to see what microdosing glucagon looked like, and whether there was an appropriate model that could be used with open source APS systems in a dual hormone set up using two pumps.

This article sets out what I’ve learned from very limited use of two GlucaGen kits, and what it might mean in adapting an open source codebase.

The Hypothesis

Based on what we know about pancreatic activity, the effectiveness of exogenous glucagon will be in proportion to insulin on board, with both hormones acting as switches to the liver.

Insulin, however, works with skeletal muscle, so when provided exogenously, has limited effect on the liver.

The same skeletal muscle has no glucagon receptors, so glucagon only really affects the liver and the brain. When delivered subcutaneously to various body parts, it mostly affects the liver.

The key data being captured was:

  • Time to effect – how long before the glucagon dose countered a drop
  • Size of effect given insulin on board – what positive effect was observed and did it vary with IOB
  • Unexpected side effects
  • Effects with and without alcohol

In addition to these points, any other observations were captured in an attempt to understand the effects.

The approach

The intention here was to figure out how glucagon might be dosed to minimise hypos in the context of the oref1 model.

Within that there is a “Carb required” function that offers up suggestions for carbs to counter future hypos. The idea was to use that as a trigger and see whether it was a valid indicator for dosing glucagon.

Data was captured in Nightscout, using the notes function to identify the glucagon doses.


Time to effect

Firstly, time to effect. Given various literature about intramuscular administration of glucagon, the expectation was that it would take somewhere between 10 and 15 mins for subcutaneous application of glucagon to take effect. This proved to be a reasonable estimate, and indeed, it was possible to observe changes in glucose attributable to glucagon administration in that time frame.

As the example in the image below shows, the effect could be very quick.

If we zoom in on the noted glucagon dose, we can see the quick response to it. The red line shows the point of the 5iu dose.

This, however, is an example in the first period of testing, where alcohol was not included. During this phase, the application of glucagon proved to be very effective, however, during the second phase, where there was alcohol involved, there was zero effect from the glucagon after drinking (which is expected given the single process focus of the liver).

Size of effect and IOB

After capturing a few data points (bearing in mind that this is a very small data set), my hypothesis was that the following curve could be applied to determine the anticipated rise given the IOB levels, after 5iu of glucagon given to me.


Below basal rate and between 100% and 150% of basal rate, the curve worked, however, once the percentage of IOB increased above 150% relative to basal rate, the relationship between the difference became more between expected drop and the level above that that the glucagon maintained. 

Essentially, where IOB was greater than 150% of basal rate, multiple doses of 5iu of glucagon would be required to stop a hard drop, and there’s a fair bit more work to do to figure out the relationship.

Unexpected side effects

Perhaps less surprising than it should have been, using glucagon to handle low levels or the risk of low levels, results in a “hungry liver”. As people with type 1 diabetes, we get used to the liver not trying to convert glucose to glycogen because it doesn’t get any signalling, however, I noticed that my postprandial highs were lower, for a good 12-18 hours after a glucagon dose, as the liver looked to replenish its missing glycogen. Overall, in the alcohol free period, the overall time in range was good, with very low time below range. Is it possible that using glucagon as an additional hormone reduces the sleeping liver effect?

During the three days of testing without alcohol, the results were good, given the food eaten (this includes snacks in the office, and relatively high carb eating). I was able to allow the AID algorithm to overdose for mealtimes and then catch as necessary with glucagon. Noticeably, there  wasn’t any reduction in average total daily dose.

Effects of alcohol…

With alcohol though, the effects were rather different. Given that the Boost algorithm hadn’t been modified to take into account the glucagon dosing, aside from glucagon having no effect when there was alcohol on board (AOB), there was a secondary side effect that it caused lows overnight that I wouldn’t normally see. This appeared when the Boost functionality is disabled and was as a result of standard oref1 function. I essentially became more sensitive overnight than normal.

This came as a bit of a surprise. While I expected that there would be no effects from dosing glucagon with alcohol, I wasn’t expecting some kind of delayed effect. This occurred on multiple occasions so I’m reasonably sure that it was linked to the glucagon administration. 

What can you take away from this?

I think that it’s fair to say that using glucagon to stave off low glucose events is a reasonable approach to using an AID. It does come with some caveats though, namely, it doesn’t work with alcohol, and where IOB is significantly greater than basal rate, it may not have much of an effect, as the liver is shutdown by whatever of the IOB is getting to it, in spite of that being very small levels. 

Whilst we’ve established that reconstituted glucagon does seem to survive for 3 days once made up, and in theory could provide a low cost option when using a system, it’s not the answer to everything.

The issues observed with alcohol and IOB represent things that need to be taken into account when incorporating glucagon into a system and as a part of a multi-hormone system, and self monitoring for these effects would be a key part of introducing its use into any form of open source AID system.

This, however, is an n=1 observation using very small amounts of glucagon and providing very small amounts of data. I’d love to see what the likes of Betabionics have found in their dual hormone approach.

At this stage, I’d suggest that while it presents promising options, and I can see how it staves off low glucose levels effectively in many circumstances, I’m not sure it’s the answer to everyone’s hopes in the world of type 1 tech. As an adult, there are many cases that would deem it ineffective. I think the major concern is the complete unreliability after alcohol, which is when you’d be most likely to need that safety net.

I’d be intrigued to learn more about the Inreda system, and how it manages in the context of alcohol. I assume that within the algorithm, there’s a “no response” function that accepts that if there’s no response to glucagon, it should stop trying and do something else. I think there would be the need for that within an open source version of a system that incorporated glucagon. There’s also a need to modify the system’s view of sensitivity post glucagon dosing, which would also need measurement and incorporation.

All of these things present options and I think it’s still possible to improve management of type 1 with the addition of glucagon in an AID.

It’s simply that diabetes is far more complex than two hormones and an algorithm, so while we might make it better, it won’t be perfect.


  1. The reference cited in the post was tested on mice rather than humans, according to the author. The behavior of humans would probably differ from that of animals. A long-term use of PPIs unnecessarily for their indication can cause major side effects. DPP4 is not indicated for type-1 and is used off-label. Gaba is a shelf supplement that is not regulated as a pharmaceutical. The quality of supplements has been questioned, with different brands offering different levels of quality. Some people with T1D may benefit from GLP-1 and metmorphone by improving their insulin resistance, especially those who are obese.

      • It is not my intention to argue whether the treatment is beneficial or not. An abstract for a conference and a nice model of 56 individuals are two very small drops in the ocean. My research led me to the conclusion that there is nothing related to the 2020 ATTD presentation that I could find after doing a few searches. Every day there is a new hype or trend promoting a new promise for people with type 1 diabetes. It is important to keep things in proportion.

        • I agree. The team behind this has set up a business in Israel, and I’ve seen results from a couple of people that are quite remarkable.

          I’m not sure of the number of people they’ve taken this approach with, but it certainly is having an effect.

Leave a Reply

Your email address will not be published.