Temporal Discontinuity in Continuous Glucose Monitoring: How Midnight Segmentation Generates Systematic Measurement Artifact

View at publisher Summarise with VIRAT
Authors: Mark E. Paull
Cited by: 0
References: 24

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Continuous glucose monitoring (CGM) generates dense physiological time-series data sampled every 5 min across 24-h periods. Standard analytical approaches impose calendar-based temporal boundaries—treating midnight as a natural segmentation point—despite glucose homeostasis operating through continuous circadian oscillators that recognize no such delimiter. This structural misalignment introduces a systematic measurement artifact: biologically continuous overnight patterns are bisected at 00:00, artificially inflating glycemic variability estimates and obscuring individual circadian phase relationships. We analyzed approximately 60 days of CGM data, comparing three binning strategies: (1) 24 linear hour-bins (conventional), (2) 36 linear 40-min bins (resolution control), and (3) 36 angular 10° bins (circular topology). Shannon entropy with variance-weighted probabilities quantified information content. Bootstrap resampling (1000 iterations) and null topology permutation (random within-day time-permutation, 1000 iterations) distinguished genuine temporal structure from mathematical artifact. Circular representation demonstrated 12.1% higher information entropy compared to linear binning at matched resolution (3.56 vs. 3.18 bits, P < 0.001, bootstrap percentile method), with nonoverlapping confidence intervals (95% CI: 3.41–3.71 vs. 3.06–3.30). Increasing from 24 to 36 bins in linear space produced zero entropy change (3.18→3.18), isolating topological continuity as the information-preserving factor. Midnight boundary created 2.8-fold reduction in continuity correlation ( r = 0.31 vs. r = 0.87, P < 0.001) for biologically adjacent timepoints. Analysis showed nonrandom angular variance structure ( P < 0.001), with 1.97-fold variance differential between evening (21:20–00:00) and midday (11:20–14:00) zones. Midnight segmentation introduces quantifiable information loss through temporal discontinuity. Circular time representation—mapping 24-h cycles onto angular coordinates using established directional statistics—eliminates this artifact while preserving temporal information. Current glycemic variability metrics (coefficient of variation, time-in-range) calculated within midnight-bounded periods inherit discontinuity artifacts, potentially misclassifying normal circadian oscillations as pathological variability. Adoption of circular frameworks would align CGM analytics with chronobiological principles and enable individual circadian phenotyping without data manipulation. This represents methodological infrastructure requiring prospective validation for clinical utility.

Fields of Study & Subjects

Diabetes Management and Research Circadian rhythm and melatonin Mental Health Research Topics

Publication details

Journal / source
Diabetes Technology & Therapeutics
Volume
28
Issue
10
Pages
1104–1109
Publication date
23 March 2026
Language
EN