Long-Term Patterns in Automated Insulin Delivery and Carbohydrate Announcement: A 24-Month Follow-Up

Abstract

We evaluated long-term (24 months) consistency in carbohydrate counting and meal announcements among Control-IQ users using two parameters: auto-bolus percentage (i.e., automatic correction boluses divided by total boluses) and daily carbohydrate announcement (DCA). In this single-center retrospective cohort study (October 2021-October 2024), we analyzed trends in auto-bolus percentage-alongside DCA and metabolic control-and its associations with age, sex, DCA, and time in range (TIR) using mixed-effects linear models. Among 2751 person-quarters (57% women, mean age 37 years), the mean auto-bolus percentage was 61% and increased by 0.4% per quarter (P < 0.001). DCA declined from 132 to 100 g/day, while TIR slightly decreased from 61% to 58%. Auto-bolus percentage was inversely associated with age, TIR, and DCA, with the latter association strengthening over time. The modest change in TIR suggests sustained effectiveness of automated insulin delivery despite decreasing carbohydrate reporting-likely reflecting adaptive user behavior.

Keywords: Control-IQ algorithm; auto-bolus percentage; automated insulin delivery; daily carbohydrate announcement; type 1 diabetes.