Diabetes is a complex chronic condition that requires constant management and monitoring. It currently affects an estimated 537 million adults worldwide, and this number is projected to increase to 740 million over the next 15 years.[1] People with diabetes face significant challenges in managing their glucose levels. Even simple everyday activities, such as driving or exercising, require careful consideration of how they will affect blood glucose levels. With limited healthcare professional guidance, individuals must make up to 180 therapy decisions daily on their own.
Over the last 25 years, advancements in diabetes care technology have played an important role in improving the lives of people with diabetes. These innovations have provided better and more accurate methods of measuring blood glucose, thereby easing the burden of managing therapy.[2],[3] Today, continuous glucose monitoring (CGM), a technology that has revolutionised modern diabetes management, is considered the gold standard for glucose monitoring in insulin-treated diabetes. With the advent of artificial intelligence (AI) in diabetes management solutions, a new era of predictive care has begun. This era promises more individualised management and a shift from reactive to proactive approaches.
Finding clarity amidst complexity
Achieving good blood glucose control is essential to avoiding diabetes-related complications. However, glucose dynamics are incredibly complex and volatile, influenced by numerous endogenous and exogenous factors. This makes them historically very difficult to predict.[4]
Most existing CGM solutions provide trend arrows, indicating whether a user’s glucose level is likely to increase, decrease, or remain stable in the near future. However, these trend arrows are based solely on simple mathematical calculations that account for recent glucose measurements, without considering factors such as insulin doses or food intake. As a result, they offer only limited guidance for therapy adjustments. Additionally, constant monitoring of glucose data can lead to data fatigue, further increasing the burden of managing diabetes.
This is where AI comes into play. While CGM is powerful on its own, combining it with AI holds the promise of revolutionising personalised diabetes management. AI can enhance the recognition of glucose patterns and effectively forecast glucose dynamics, translating vast amounts of data into useful, actionable insights for people with diabetes and their care providers.[5]
This is the technology underpinning our Accu-Chek SmartGuide CGM solution, which predicts glucose levels across various timeframes to help adults with type 1 and type 2 diabetes on flexible insulin therapy proactively manage their condition. With this solution, we aim to reduce the burden of diabetes, improve glycaemic control, and enhance overall quality of life.
From reactive to proactive diabetes management
The impact of glucose prediction on the well-being of a person living with diabetes cannot be understated. In the words of diabetes advocate Hanna Boëthius at a recent event at the European Association for the Study of Diabetes annual meeting, “seeing into the future as a person with diabetes is like a superpower”. Clinical trial data demonstrates the relief this technology could provide for people with diabetes, with our Accu-Chek SmartGuide CGM solution’s two-hour blood glucose prediction reducing diabetes distress and fear around hypoglycaemia.[6]
One key area where predictive technology makes a highly significant positive impact is in addressing nocturnal hypoglycaemia – episodes of very low blood glucose that occur during nighttime. Nocturnal hypoglycaemia negatively impacts the mental, physical, and social well-being of people with diabetes. The advent of CGM has revealed that nocturnal hypoglycaemia is a more common problem than previously thought.[7] A single severe incident of nocturnal hypoglycaemia can have dire consequences, as it can be fatal. The ‘dead-in-bed’ syndrome is responsible for approximately 5-6% of deaths in people with type 1 diabetes under the age of 40.8 It is no wonder that many people with diabetes fear nocturnal hypoglycaemia – it impacts well-being, causes anxiety and tension, and makes getting a peaceful night’s sleep difficult.[8]
To tackle this issue, an AI-enabled CGM solution offers the ability to predict the likelihood of hypoglycaemia throughout the night, with preliminary data showing a potential significant reduction in the number of nighttime hypoglycaemic episodes.[9] Now, people with diabetes can maximise their chances of a restful sleep by knowing when preventative actions, like eating a snack, are needed to stay in range.
The intelligence behind predictive CGM
Without the integration of AI, current CGM solutions typically use a very small amount of glucose data – usually the past 15 minutes – to identify trends or predict the immediate future.[10] With the incorporation of AI, predictions can be provided with much higher accuracy across a range of timeframes.[11]
The Accu-Chek SmartGuide CGM solution incorporates three distinct predictive features to meet the different needs of people with diabetes. Glucose Predict continuously forecasts glucose excursion over the next two hours. Low Glucose Predict monitors the user’s risk of entering hypoglycaemia within the next 30 minutes and notifies them, allowing them to act accordingly. Night Low Predict informs about the likelihood of nocturnal hypoglycaemia, promoting improved sleep and overall quality of life.10 Each model employs a sophisticated machine learning approach to analyse both current and up to 28 days of historical individual data.11
With this glycaemic foresight, we hope to equip people with diabetes to take preventative action to stay within therapeutic glucose range. By doing so, they can avoid unforeseen glucose excursions, tackle their fear of hypoglycaemia, reduce the amount of troubleshooting needed, and ultimately alleviate some of the overall burden of living with the condition.
The future of AI innovation in diabetes management
This is just the beginning of the AI-driven health technology era, and the potential impact on diabetes care is already substantial. Looking ahead, AI could play a pivotal role in the evolution of diabetes management, improving the lives of millions of people worldwide. As we move into the future, we will continue to empower people living with diabetes and develop innovations that can reduce the overall burden of their condition.
Article by Dr Pau Herrero, Lead Research Engineer at Roche Diagnostics
References
[1] International Diabetes Federation. IDF Diabetes Atlas 10th Edition. 2021.
[2] Kulzer B, Heinemann L. JDST. 2024;18:1000–3.
[3] Hirsch IB. Introduction: History of Glucose Monitoring. In: Role of Continuous Glucose Monitoring in Diabetes Treatment. Arlington (VA): American Diabetes Association; 2018 Aug. Available at: https://www.ncbi.nlm.nih.gov/books/NBK538968/ (Accessed: October 2024).
[4] Butt H, et al. Diagnostics (Basel). 2023;13:340.
[5] Medanki S, et al. World J Exp Med. 2024;14:87916.
[6] Ehrmann D, et al. Diabetes. 2023;72:1794-PUB.
[7] Kulzer B, et al. J Diabetes Sci Technol. 2024;18:1052–60.
[8] Cappon G, et al. IEEE Trans Biomed Eng. 2023;70:3105–15.
[9] Special issue on CGM; Journal of Diabetes Science & Technology, October 2024. Available at: https://journals.sagepub.com/toc/DST/0/0 (Accessed: October 2024).
[10] Elbarbary N, et al. Diab Vasc Dis Res. 2021;18:14791641211062155.
[11] Herrero P, et al. J Diabetes Sci Technol. 2024;18:1014–26.