Automated Insulin Delivery: Six Universal Observations and Understandings

By Dr. Laurel Messer

Six universal facts about automated insulin delivery systems, and the things you should keep in mind about this revolutionary technology

Automated insulin delivery (AID) systems are moving towards the forefront of diabetes management. AID systems combine continuous glucose monitors (CGM) with smart algorithms to automatically adjust insulin delivery.

As more systems move through the pipeline and eventually into the commercial market, important patterns are emerging in user expectations and user experience. As a diabetes nurse, certified diabetes educator and research investigator, I, along with my team at the Barbara Davis Center, have worked with nearly every AID system in the pipeline, and other systems that will never make it to market. Here are six insights we have gleaned, which seem to be universal (thus far) to all AID systems:

1. You can always beat an AID system with compulsive diabetes management

Many people with diabetes compulsively attend to diabetes care in order to achieve ultra-tight glucose ranges – and are the first to ask about automated systems. What ends up happening is that these “super-users” are invariably frustrated that the system is not yielding the same results that they were able to achieve with their own calculations and management.

An important point is that many automated systems are excellent at reducing mental burden for taking care of diabetes, excellent at reducing hypoglycemia, and adequate at improving glucose levels. Humans can beat automated systems if they attend to diabetes care near-constantly. The individuals who will likely be satisfied with AID are those who are comfortable with an A1C in the 7s or above, but they want to reduce the mental load of adjusting settings and micromanaging high glucose levels.

The most important question to ask is, “Why do I want to start using an automated system?” If it is to achieve near-perfect glucose levels, the system will likely disappoint. If it is to reduce the burden of “thinking like a pancreas” all the time, it may be a good option. AID will excel at the marathon of diabetes care but may disappoint in the hour-to-hour sprint.

2. Systems work best when you let them work

Using both research and commercial systems, we have seen all the ways to “trick” AID systems – entering phantom carbohydrates, changing set points, performing manual corrections, overriding recommended doses. More often than not, these behaviors lead to glucose instability – reactionary highs and lows from the system destabilizing. All systems will perform best if they are used according to user instructions. This is difficult for the individual who would prefer to micro-adjust settings or desire control over all insulin delivery. Most systems work best when users learn to trust them.

3. Give the system a chance – 2-4 weeks before deciding long term potential

It may benefit us to think about AID like a new significant relationship – it can take some time to “settle.” I mean this both on an interaction level (learning how to respond to alerts, when to intervene, when to let it ride) and on an algorithm level (allowing the system to adjust internal algorithm parameters based on usage). In addition, programmable user settings may need some adjustment in the first few weeks of use, so working with diabetes educators can be helpful for initial set-up and early follow-up.

4. Bolusing is still king

If I could go back in time, I would caution device manufacturers against any whisper of not needing to bolus with AID systems. Bolusing is the singular most important action a person with diabetes can do to optimize insulin delivery on current and near-future automated systems.

This will be true until insulin action time gets exponentially faster or artificial intelligence gets better at predicting human behavior, neither of which is on the immediate horizon. In order for people with diabetes to see the best performance on any system (automated or manual), they need to bolus before carbohydrates are consumed. Specific to AID, the timing of the bolus (prior to carb intake) is especially important, as the system will automatically increase insulin delivery after an initial rise of glucose levels, so a late bolus (e.g., after the meal) could lead to insulin stacking and hypoglycemia.

5. Rethinking low treatments

Low glucose levels (hypoglycemia) still happen when using automated systems. What is different with AID is that the system has been trying to prevent the low by reducing/suspending insulin, possibly hours before the low occurs. This means that an individual may need to consume significantly fewer carbs to bring glucose levels back into range – perhaps 5-10 grams of carb at first, reassessing 15-20 minutes later. This can be difficult when wanting to eat everything in sight; however, it can reduce the chance of rebounding into the 200s after over-treating.

6. Infusion sets are still infusion sets

While AID algorithms are revolutionary, the infusion set is not. It is the same plastic or steel cannula that occludes, kinks, or inflames. This hardware limits automated systems and can very quickly lead to hyperglycemia or diabetic ketoacidosis (DKA). It is important for people using AID to recognize signs of infusion set failure – persistent hyperglycemia, boluses that do not bring glucose levels down, ketones, vomiting, etc. Knowing how to treat ketones (via syringe injection of insulin and set change) can prevent a hospital admission or worse.

I love that the diabetes community learns from its members and experiences. Check out our Barbara Davis Center PANTHER (Practical Advanced THERapies for diabetes) website for our team’s latest insights on automated insulin delivery, and tools for people with diabetes, clinicians, and engineers.

Are you considering AID? Feel free to share this article with your healthcare team. For more information about AID systems that are currently available or in the pipeline, click here.

About Laurel

Dr. Laurel H. Messer is a nurse scientist and certified diabetes educator at the Barbara Davis Center for Diabetes, University of Colorado School of Medicine, Aurora, CO. She has spent the last 15 years studying how to best utilize new diabetes technologies, and remembers fondly teaching families to wrap up their corded CGM system in a plastic shower bag for bathing. Ok, not that fondly, but look how far we have come! Dr. Messer works with the Barbara Davis Center PANTHER team (Practical Advanced Therapies for diabetes), conducting clinical research trials on promising technologies to make life better for children, adolescents, and adults living with type 1 diabetes. Get in touch at Laurel.Messer@cuanschutz.edu