Publications

The Role of AI in Improving Insulin Dosing for People with Diabetes

Diabetes is a growing global health challenge, affecting an increasing number of people each year. Many people living with diabetes rely on insulin therapy to manage blood sugar levels*. However, insulin therapy is complex – many factors like meals, activity, stress, and hormones affect blood sugar, leading to risks like low (hypo) or high (hyper) blood sugar. People often have to calculate insulin doses manually, which is stressful and prone to errors.

The publication explores how artificial intelligence (AI) – especially reinforcement learning – can improve insulin dosing and diabetes management by making it more personalised and automated.

The researchers of the MELISSA project reviewed recent developments in AI-powered insulin delivery systems, such as closed-loop systems, and assessed their benefits, challenges, and future potential.

Key findings include:

• AI can improve blood sugar control and reduce the risk of dangerous highs and lows by analysing individual patterns, meals and activity.
• Reinforcement learning helps systems learn from each person’s unique patterns – making personalisation possible.
• Personalisation is central in diabetes care – AI systems tailor insulin recommendations to each person, improving blood sugar control and reducing complications.
• Clinical outcomes may improve when AI-driven systems are used, with more stable blood sugar levels and fewer episodes of hypoglycaemia due to better matching of insulin doses to individual needs.

In conclusion, AI shows great promise for safer, more effective diabetes care. That said, some important challenges, like ensuring data privacy and security, making AI decisions transparent, and keeping the technology accessible, remain. More real-world testing, clear regulations, and efforts to ensure fair access are needed to bring these tools into everyday use.

Read the full publication here.

*The term blood sugar level refers to the concentration of glucose in a person’s blood.

The effect of bolus advisors on glycaemic parameters in adults living with diabetes on intensive insulin therapy

The study looks at how using bolus advisors - devices or software that help people with diabetes calculate the correct dose of insulin at mealtimes - affects blood sugar control in adults with diabetes on intensive insulin therapy.

The researchers reviewed, summarised and analysed data from 18 studies, involving 1,645 adults living with diabetes, most of whom had type 1 diabetes and used multiple daily insulin injections.

Key findings include:

1. HbA1c Reduction: The use of bolus advisors slightly reduced HbA1c levels (a marker of long-term blood sugar control) by an average of 0.11%.
2. Improvement in Low Blood Glucose Index: There were minor improvements in preventing the risk for low blood sugar episodes.
3. Treatment Satisfaction: People using bolus advisors were generally more satisfied with their treatment.
4. No major changes in preventing hypoglycaemia: The number of hypoglycaemic events (dangerously low blood sugar levels) didn’t reduce much.
5. Other glucose outcomes: No changes in other glucose outcomes such as glucose fluctuations were found.

In conclusion, bolus advisors can help improve overall glucose control and make people more satisfied with their treatment, but they don’t reduce hypoglycaemic events. The authors recommend more research to personalise these advisors using artificial intelligence to make them more effective.

Read the full publication here: https://dom-pubs.pericles-prod.literatumonline.com/doi/10.1111/dom.15521