Research

An Adaptive Model-Based Approach to Personalized Basal Insulin Initiation in Type 2 Diabetes

In DTU Compute PHD-2019, 2019

Read the article

Abstract

Type 2 Diabetes is a growing global problem. Despite advancement of medication in recent years, the eﬃcacy demonstrated in clinical trials fails to materialize in real world data and the majority of patients does not reach treatment targets. The primary reason for this discrepancy is low adherence to treatment, caused by 1) a lack of perceived need for the medication, 2) the complexity of the treatments, and 3) fear of hypoglycemia. To address the need for support to improve glycemic outcomes, we developed and tested the feasibility of a novel concept for basal insulin initiation. The concept leverages data from diabetes management devices during insulin initiation to model dose response and estimate the dose a patient needs to reach the glycemic target. This estimated dose can be used as guidance for the remaining treatment period in combination with adaptive dose guidance algorithms. The estimated dose is expected to improve patients perceived need for insulin, reduce fear of hypoglycemia, simplify the treatment, and eventually improve adherence and outcomes. We developed a dose response model and dose estimation method by analyzing glucose and insulin data from previous clinical trials on Insulin Degludec treatment and tested the feasibility of the dose estimation in a clinical study. Learnings from the clinical feasibility study lead us to propose two approaches to adaptive model-based dose guidance, as well as an automatic glycemic target setting method. We compared the performance with standard of care and other model-based approaches using in silico simulations of low adherence scenarios. The studies suggest that dose estimation is feasible, and the adaptive dose guidance algorithms show potential for improved glycemic outcomes, even in case of low adherence to medication.