Assessing Glucose Variability Metrics: Nerve Conduction Velocity in Children and Adolescents with Type 1 Diabetes
Journal
medRxiv
Type
working paper
Date Issued
2025-03-30
Author(s)
Marc Robin Gruener
;
Erin West
;
Ute Muhitira
;
Katrin Heldt
;
Sarah Sophie Oberhauser
;
Dagmar l'Allemand
;
Mia Jovanova
;
;
Philip Julian Broser
Abstract
Background Diabetic peripheral neuropathy is a common microvascular complication of type 1 diabetes (T1D) that can manifest early in pediatric patients. While chronic hyperglycemia is an established risk factor, the role of glucose variability in the development of neuropathy remains controversial, particularly in children and adolescents.
Methods In this single-center retrospective cross-sectional study, we evaluated associations between height-adjusted nerve conduction velocity (aNCV) and multiple glucose variability metrics in children and adolescents with type 1 diabetes. We analyzed continuous glucose monitoring data from the 90 days preceding standardized neurophysiological assessment to calculate established glucose variability metrics (SD, CONGA, ADRR, MAG, GVP) alongside our novel Glucose Fluctuation Moment Index (GFMI). The GFMI uniquely integrates both the velocity of glucose changes and their distance from the glycemic target. Using linear regression models adjusted for diabetes duration, we assessed the relationship between each glucose variability metric and aNCV, employing leave-one-out cross-validation to evaluate predictive accuracy despite limited sample size.
Results Among 42 eligible participants (mean age 12.2 ± 3.2 years, diabetes duration 4.1 ± 3.2 years), mean aNCV was negative in the peroneal nerve (-2.9 ± 2.9 m/s). Linear regression models adjusting for diabetes duration revealed no statistically significant associations between any glucose variability metrics (SD, CONGA1, CONGA24, ADRR, MAG, GVP, or GFMI) and peroneal nerve aNCV (all p>0.05). In predictive modeling using leave-one-out cross-validation, the Glycemic Variability Percentage (GVP) demonstrated the best predictive performance (RMSE 2.99 m/s with all observations, 2.46 m/s after removing three identified outliers), but offered minimal improvements (4.28 % and 6.08 %) over the baseline- (simply predicting the mean aNCV) or the glucose SD-based model. Outlier analysis revealed notable clinical factors potentially affecting nerve function, including physical activity.
Conclusions While early neurophysiological changes were observed in our pediatric T1D cohort, neither established glucose variability metrics nor our novel GFMI demonstrated superior predictive accuracy for the peroneal aNCV compared to the glucose SD, the clinically established GV measure, nor achieved clinically relevant accuracy. Further research with larger cohorts and longitudinal designs is needed to account for relevant confounding variables like C-peptide and physical activity to better understand these complex relationships.
Methods In this single-center retrospective cross-sectional study, we evaluated associations between height-adjusted nerve conduction velocity (aNCV) and multiple glucose variability metrics in children and adolescents with type 1 diabetes. We analyzed continuous glucose monitoring data from the 90 days preceding standardized neurophysiological assessment to calculate established glucose variability metrics (SD, CONGA, ADRR, MAG, GVP) alongside our novel Glucose Fluctuation Moment Index (GFMI). The GFMI uniquely integrates both the velocity of glucose changes and their distance from the glycemic target. Using linear regression models adjusted for diabetes duration, we assessed the relationship between each glucose variability metric and aNCV, employing leave-one-out cross-validation to evaluate predictive accuracy despite limited sample size.
Results Among 42 eligible participants (mean age 12.2 ± 3.2 years, diabetes duration 4.1 ± 3.2 years), mean aNCV was negative in the peroneal nerve (-2.9 ± 2.9 m/s). Linear regression models adjusting for diabetes duration revealed no statistically significant associations between any glucose variability metrics (SD, CONGA1, CONGA24, ADRR, MAG, GVP, or GFMI) and peroneal nerve aNCV (all p>0.05). In predictive modeling using leave-one-out cross-validation, the Glycemic Variability Percentage (GVP) demonstrated the best predictive performance (RMSE 2.99 m/s with all observations, 2.46 m/s after removing three identified outliers), but offered minimal improvements (4.28 % and 6.08 %) over the baseline- (simply predicting the mean aNCV) or the glucose SD-based model. Outlier analysis revealed notable clinical factors potentially affecting nerve function, including physical activity.
Conclusions While early neurophysiological changes were observed in our pediatric T1D cohort, neither established glucose variability metrics nor our novel GFMI demonstrated superior predictive accuracy for the peroneal aNCV compared to the glucose SD, the clinically established GV measure, nor achieved clinically relevant accuracy. Further research with larger cohorts and longitudinal designs is needed to account for relevant confounding variables like C-peptide and physical activity to better understand these complex relationships.
Publisher
Cold Spring Harbor Laboratory