BSPED2025 Poster Presentations Diabetes 5 (10 abstracts)
Actual automode use offers superior discrimination of HbA1c outcomes compared to relative automode use in children with type 1 diabetes using AID systems
Proteek Sen 1 , John Pemberton 1 , Zainaba Mohamed 1 , Melanie Kershaw 1 , Leslie Drummond 1 , Renuka Dias 1,2 , Suma Uday 1,3 , Vrinda Saraff 1 , Ruchi Nadar 1 , Chamila Balagamage 1 , Louise Collins 1 & Ruth Krone 1
1Birmingham Children’s Hospital, Birmingham, United Kingdom. 2University of Birmingham, Department of Applied Health Research, Birmingham, United Kingdom. 3Department of Metabolism and Systems Science, University of Birmingham, Birmingham, United Kingdom
Background and Aims: Automated Insulin Delivery (AID) systems have transformed glycaemic management in children and young people (CYP) with type 1 diabetes (T1D). Metrics such as percent sensor wear time (Sensor%) and percentage automated mode during sensor use (PAM%), may misrepresent actual automode use. This study compared Sensor%, PAM%, and actual automode percentage over a defined period (AAM%) to assess which best predicts HbA1c 6 and 12 months after AID initiation.
Materials and Methods: This retrospective cohort study included 170 CYP with T1D initiated on AID therapy between 2021 and 2024 at a UK tertiary centre. HbA1c values at 6 and 12 months were paired with corresponding 90-day CGM/pump use data. Sensor% and PAM% were taken directly from clinical reporting tools. AAM% was derived by multiplying Sensor% by PAM% and dividing by 100, i.e., (Sensor% × PAM%) / 100. Repeated-measures mixed-effects models assessed the relationship between each usage metric and HbA1c. To explore clinical utility, the proportion of individuals achieving HbA1c thresholds (≤48, ≤52, ≤58 mmol/mol) was calculated across use bins (<80%, 80–89%, 90–94%, ≥95%).
Results: Across the 340 observations, mean HbA1c was 55.3 mmol/mol (SD = 9.1), and mean age was 11.3 years (SD = 3.3). All three metrics significantly predicted HbA1c (P < 0.001). All three metrics explained ~80% of HbA1c variance, however, AAM% demonstrated the lowest Bayesian Information Criterion (BIC), indicating superior model parsimony (see Table). AAM% showed the clearest gradient across all HbA1c cut offs, for example, the proportion achieving HbA1c ≤58 mmol/mol rose from 38% at <80% use to 83% at ≥95% use, a greater differential than seen with Sensor% (44% to 75%) or PAM% (41% to 79%), further supporting AAM%’s stronger discriminatory power.
| Predictor | Beta Coefficient | 95% CI Lower | 95% CI Upper | Marginal R² | Conditional R² | BIC |
| Sensor% | -0.3 | -0.41 | -0.18 | 0.792 | 0.665 | 2362 |
| PAM% | -0.34 | -0.42 | -0.25 | 0.812 | 0.696 | 2334 |
| AAM% | -0.28 | -0.35 | -0.21 | 0.796 | 0.679 | 2330 |
Conclusion: AAM% integration into clinical reporting tools may enhance patient stratification and inform AID optimisation strategies.
Volume 111
Article tools
My recent searches
My recently viewed abstracts
Authors
Endocrine Abstracts
ISSN 1470-3947 (print) | ISSN 1479-6848 (online)
© Bioscientifica 2025 |
Privacy policy |
Cookie settings
BiosciAbstracts
Bioscientifica Abstracts is the gateway to a series of products that provide a permanent, citable record of abstracts for biomedical and life science conferences.
Find out more