If you look through the various social media groups for DIY closed loops, there are a lot of different views as to what’s the best strategy when it comes to using basal rates in DIY closed loops.
The first and foremost of these is that the basal rates on the pump should be the fall back for when (and not really if) something goes wrong and you need to return to manual pumping.
But beyond this, there are various views on what’s the correct approach to take.
In the AndroidAPS users group, there’s a strong preference for using a circadian profile (with similar variations to ISF and CR), based on various research (examples here and here) that suggests glucose homeostasis is driven by a circadian rhythm, with some research suggesting that pancreatic islets align very strongly to this, from which I infer that background insulin release is not uniform throughout the day.
Separately, members of the Loop and Learn group have promoted a single basal rate throughout 24 hours, which is based on 0.5 x TDD or 0.4 x TDD. This comes from the idea that the pancreas produces a constant background insulin level and that the liver produces glucose as required to counter the insulin levels in the body, which can be managed by the closed loop and potentially having multiple difference ISFs and/or CRs.
People operating with both models seem to be getting reasonably good results and there is a robust discussion as to which is the best approach to take.
Given this variation in outlook, I figure it would make sense to try both out myself and see what happens.
Over the next month or so, the plan is to run a two week stint on the flat basal/variable ISF model and a similar two week stint on a circadian basal/variable ISF model, followed by reversion to the almost circadian, single ISF model that I’ve historically always used. With this dataset, it will hopefully be possible to identify which approach works best for me and provide some indication to others as to what the best model is.
The parameters for this test will be to run both of the aforementioned basal models with historic modus operandi for meal management. This means announcing meals and partially bolusing for meals. The insulin in use will be a Lyumjev/Fiasp mix, at a 66.6% Lyumjev, 33.3% Fiasp. For now, I’ll be putting fully closed loop modifications on hold while I go through the process. There is no intent to change my eating style.
This will be running using AndroidAPS as the DIYAPS system, with Autosens operating fully.
In the meantime, you should always have a stored profile on the pump that you can go back to as one that you know works if, for whatever reason, closed loop isn’t possible.
Let’s see which approach generates the best outcomes for me!