# Evaluating Insulin Delivery Systems Using Dynamic Glucose Region Plots and Risk Space Analysis

## TL;DR

This study compares two automated insulin delivery systems using dynamic glucose plots and finds that one system better maintains glucose levels within a target range.

## Contribution

Introduces risk space analysis with dynamic glucose region plots as a novel method for evaluating insulin delivery systems.

## Key findings

- System A had a higher fraction of (RoC, glucose) values in the optimal risk space compared to system B.
- Using AID systems reduced the risk of glucose declining below the target range compared to non-AID systems.
- Dynamic glucose region plots and risk space analysis revealed clinically relevant differences between insulin delivery systems.

## Abstract

What are the main findings?
Our results reveal a clinically relevant higher fraction of (RoC, glucose) values in the optimal risk space with AID system A compared to system B.The risk of glucose declining from the target range to below target range was lower in persons using AID systems than in persons without an integrated CGM system.

Our results reveal a clinically relevant higher fraction of (RoC, glucose) values in the optimal risk space with AID system A compared to system B.

The risk of glucose declining from the target range to below target range was lower in persons using AID systems than in persons without an integrated CGM system.

What is the implication of the main finding?
Risk space analysis of dynamic glucose region plots is a novel strategy that contributes to the real-world evaluation of different systems for insulin delivery.

Risk space analysis of dynamic glucose region plots is a novel strategy that contributes to the real-world evaluation of different systems for insulin delivery.

Simultaneous values of glucose rate of change (RoC) and glucose can be presented in a dynamic glucose region plot, and risk spaces can be specified for (RoC, glucose) values expected to remain in the target range (glucose 3.9–10.0 mmol/L) or leave or return to the target range within the next 30 min. We downloaded continuous glucose monitoring (CGM) data for 60 days from persons with type 1 diabetes using two different systems for automated insulin delivery (AID), A (n = 65) or B (n = 85). The relative distribution of (RoC, glucose) values in risk spaces was compared. The fraction of all (RoC, glucose) values anticipated to remain in the target range in the next 30 min was higher with system A (62.5%) than with system B (56.8%) (difference 5.7, 95% CI (2.2–9.2%), p = 0.002). The fraction of (RoC, glucose) values in the target range with a risk of progressing to the above range (glucose > 10.0 mmol/L) was slightly lower in system A than in B (difference −1.1 (95% CI: −1.8–−0.5%, p < 0.001). Dynamic glucose region plots and the concept of risk spaces are novel strategies to obtain insight into glucose homeostasis and to demonstrate clinically relevant differences comparing two AID systems.