Harnessing wearables and mobile phones to improve glycemic outcomes with automated insulin delivery

There is increasing interest in incorporating wearable and mobile phone data, including heart rate and accelerometer data, into automated insulin delivery (AID) systems to improve glucose outcomes and reduce burden on people living with diabetes. A challenge with any type of insulin delivery system is that once insulin is delivered, it cannot be removed from the body, and so improved ability to adjust and reduce insulin concentration in advance of or in response to physical activity is one area of opportunity for improving outcomes. The study by Peter Jacobs and colleagues1Jacobs PG Resalat N Hilts W et al.Integrating metabolic expenditure information from wearable fitness sensors into an AI-augmented automated insulin delivery system: a randomised clinical trial.Lancet Digit Health. 2023; (published online Aug 3.)https://doi.org/10.1016/S2589-7500(23)00112-7Google Scholar suggests that it is possible to integrate activity data (ie, heart rate and accelerometer) from smart-watches into AID systems. Jacobs and colleagues examined two different insulin-only algorithms integrated with exercise data: an exercise-aware adaptive proportional derivative (called exAPD) and an exercise-aware model predictive control (called exMPC). The exAPD system requires user confirmation when physical activity crossed a threshold, whereas the exMPC operates without such confirmation, responding also to unscheduled or varying daily activity, like housework or easy cycling or walking. These systems were studied in-clinic with structured exercise as well as in a subsequent free-living situation where participants were able to exercise freely. There was no significant difference between the systems in-clinic for time below range or time in range, although the exMPC (no user confirmation of exercise required) system had significant lower mean glucose in the 2 h post-exercise period, with comparable time below range. The study was small (n=25) and among a cohort with a starting A1c average of 6·4%, and could be considered to be a prototype for future studies, as authors indicate they plan successive studies of longer duration and will address other study design limitations including not being powered to determine factors contributing to glucose changes. They instead point to Riddell and colleagues work2Riddell MC Li Z Gal RL et al.Examining the acute glycemic effects of different types of structured exercise sessions in type 1 diabetes in a real-world setting: the type 1 diabetes and exercise initiative (T1DEXI).Diabetes Care. 2023; 46: 704-713Crossref PubMed Scopus (3) Google Scholar indicating baseline glucose, rate of change of glucose before activity, and bolus-only IOB at start of exercise, among others, as contributors to changes in glucose during exercise. Notably, both Riddell and colleagues and Jacobs and colleagues, along with most AID researchers, have not traditionally calculated net insulin on board (net IOB) when evaluating exercise-related glucose metrics. Net IOB accounts for all insulin activity changes (increase or decrease) relative to hourly baseline insulin requirements.3Lewis D Automated insulin delivery in real life (AID-IRL): real-world user perspectives on commercial AID.J Diabetes Sci Technol. 2022; 16: 500-503Crossref PubMed Scopus (5) Google Scholar Particularly in Jacobs and colleagues' work, where two different automated insulin delivery systems are being assessed, it was striking to review the study materials and observe that participants were given guidance to take anywhere from 15 grams to 35 grams of carbohydrates for glucose concentration to be in range of 100–145 mg/dL before exercise. This is based on Moser and colleagues' guidelines,4Moser O Riddell MC Eckstein ML et al.Glucose management for exercise using continuous glucose monitoring (CGM) and intermittently scanned CGM (isCGM) systems in type 1 diabetes: position statement of the European Association for the Study of Diabetes (EASD) and of the International Society for Pediatric and Adolescent Diabetes (ISPAD) endorsed by JDRF and supported by the American Diabetes Association (ADA).Diabetologia. 2020; 63: 2501-2520PubMed Google Scholar and the carbohydrate intake is not adapted based on IOB. The amount of carbohydrates intake before exercise following this study guideline was also not analysed as a correlate for during or post-activity glucose outcomes, but one could imagine and extrapolate that someone on these AID systems with glucose concentration trending slightly down would have already experienced reduction in insulin delivery by the AID even before starting to exercise, and thus the carbohydrates intake of 15–35 grams could have a strong effect on the during and post-exercise glucose concentrations. This might play a role in influencing the 2 h outcomes for mean glucose. It is promising to see systems that are evaluating outcomes based on different AID algorithms and different UX designs that require different levels of user input to the system, particularly related to exercise. However, future studies related to AID should more thoroughly evaluate the net IOB as an input to any user behaviours (such as carbohydrate intake before exercise), as this would aid a user in balancing the selection of actions that could be taken relative to exercise, without causing more work for themselves in the future by triggering rebound increases in glucose concentrations despite the exercise. Advising and informing users when carbohydrates—and precisely how much—might be warranted, based on recent insulin delivery (eg, net IOB), CGM data, and impending physical activity, remains an underexplored opportunity for users of commercial AID systems. Those designing and studying AID systems should consider taking advantage of some of the UX patterns including carbohydrate recommendations pioneered by open source AID systems5Burnside MJ Lewis DM Crocket HR et al.Open-source automated insulin delivery in type 1 diabetes.N Engl J Med. 2022; 387: 869-881Crossref PubMed Scopus (24) Google Scholar and study the myriad of user behaviours that influence the outcomes of AID systems related to exercise. I declare no competing interests. Integrating metabolic expenditure information from wearable fitness sensors into an AI-augmented automated insulin delivery system: a randomised clinical trialAIDs can integrate exercise data from smartwatches to inform insulin dosing and limit hypoglycaemia while improving glucose outcomes. Future AID systems that integrate exercise metrics from wearable fitness sensors may help people living with type 1 diabetes exercise safely by limiting hypoglycaemia. Full-Text PDF Open Access