Background: People with type 1 diabetes or type 2 diabetes requiring insulin are struggling to achieve optimal glucose control. This is problematic, since hyperglycemia, hypoglycemia and high glycemic variability are all associated with adverse clinical outcomes and poor quality of life. Multiple factors affect glucose levels in people with diabetes, including carbohydrate intake, stress, physical activity and health conditions. Interventions based on artificial intelligence (AI) that consider these factors can benefit daily decision-making of insulin therapy and improve glucose control. The main aim of the MELISSA project is to develop and validate an effective and affordable AI-based mobile health application to improve the management of diabetes by providing individualized basal and bolus insulin dose suggestions.

Methods: This study aims to enroll participants with type 1 diabetes (n=381) and type 2 diabetes (n=90), from the Netherlands, Denmark, Germany, and Greece. They should be aged ≥16 years and on multiple daily insulin injections. Participants will be randomized to the MELISSA study-arm or to continuation of regular treatment, with a follow-up of 24 weeks. The MELISSA application consists of two AI-driven features: 1. the adaptive basal bolus advisor; 2. the goFOOD^TM application to estimate carbohydrate content based on photos. Blinded continuous glucose monitoring (CGM) will be used to monitor glucose profiles. An activity tracker will be used to measure physical activity and sleep. 

Results: The primary outcome is time spent in glucose target range (3.9-10.0 mmol/L) based on blinded CGM. Secondary outcomes include hypoglycemia, CGM-glucometrics and insulin doses, as well as patient-reported outcomes, such as quality of life, treatment satisfaction, diabetes distress and other health-related parameters.

Conclusion: The MELISSA application has the potential to improve glycemic control in people with diabetes by providing individualized AI-driven recommendations on insulin dose adjustments, thus enhancing people’s self-management skills.