Written and maintained by CASRAI Editorial Board
Last updated
Continuous glucose monitoring (CGM) is a wearable biosensing method that automatically measures glucose concentrations in dermal interstitial fluid at regular intervals (typically every 1 to 5 minutes) over multi-day wear periods. Unlike a conventional capillary blood glucometer that captures isolated point-in-time measurements, CGM systems generate continuous time-series datasets that quantify glycemic excursions, rate of change, and standardized endpoints such as Time in Range (TIR) for clinical care and metabolic clinical trials. This page describes device standards, informatics architecture, and clinical research practices; it is not medical advice.
What Is Continuous Glucose Monitoring and How Is It Used in Clinical Research and Healthcare?
Most clinical CGM sensors consist of a percutaneous electrochemical enzyme electrode coated with glucose oxidase (`GOx`)—or an implantable fluorescence-based polymer sensor—inserted into the subcutaneous adipose tissue. Interstitial glucose diffuses across a semipermeable membrane and undergoes enzymatic oxidation, generating an electrical current (measured in nanoamperes) proportional to local glucose concentration. Because glucose must diffuse from capillary blood across the capillary endothelium into the interstitial space, interstitial fluid glucose exhibits a physiological transport and sensor-filtering lag time (commonly 5 to 10 minutes relative to venous or capillary blood glucose), which calibration and signal-processing algorithms compensate for.
In clinical practice and decentralized clinical trials (DCTs), CGM serves three distinct operational functions:
- Real-Time and Intermittent-Scanned Glycemic Surveillance: Real-time CGM (rtCGM) systems transmit continuous streams and predictive hypo/hyperglycemia alerts, whereas blinded (professional) CGM modes record interstitial glucose without displaying values to the participant—a critical study design control in baseline run-in phases or blinded randomized controlled trials to avoid behavioral Hawthorne effects.
- High-Resolution Digital Endpoints Beyond HbA1c: Hemoglobin A1c (HbA1c) reflects average glycemia over 2 to 3 months but cannot detect acute glycemic variability, postprandial spikes, or asymptomatic nocturnal hypoglycemia, and is confounded by hemoglobinopathies, anemia, chronic kidney disease, and pregnancy. CGM directly measures the frequency, duration, and nadir of hypoglycemic and hyperglycemic events.
- Automated Insulin Delivery (AID) Control Loops: When cleared as an integrated CGM (iCGM), sensor output feeds directly into closed-loop control algorithms governing subcutaneous insulin infusion pumps.
Standardized Ambulatory Glucose Profile (AGP) Metrics in Clinical Trials
To harmonize clinical trial outcome reporting and routine endocrinology interpretation, the 2019 International Consensus on Time in Range (Battelino et al., endorsed by the American Diabetes Association, EASD, and ATTD) standardized the Ambulatory Glucose Profile (AGP) based on at least 14 consecutive days of CGM wear with at least 70% active data capture:
| AGP Core Metric | Glycemic Threshold | Standard Adult Target (Type 1 & Type 2 Diabetes) | Clinical Trial & Informatics Role |
|---|---|---|---|
| Time in Range (TIR) | 70–180 mg/dL (3.9–10.0 mmol/L) | > 70% of readings (> 16 hr 48 min/day) | Primary or key secondary efficacy endpoint; each 5% TIR increase correlates with clinically meaningful vascular risk reduction |
| Time Below Range (TBR) Level 1 | 54–69 mg/dL (3.0–3.8 mmol/L) | < 4% of readings (< 1 hr/day combined below 70 mg/dL) | Hypoglycemia alert threshold for safety monitoring and dose-titration protocols |
| Time Below Range (TBR) Level 2 | < 54 mg/dL (< 3.0 mmol/L) | < 1% of readings (< 15 min/day) | Clinically significant biochemical hypoglycemia; regulatory safety endpoint in insulin and incretin trials |
| Time Above Range (TAR) Level 1 | 181–250 mg/dL (10.1–13.9 mmol/L) | < 25% of readings (combined > 180 mg/dL) | Quantifies postprandial and basal hyperglycemic exposure |
| Time Above Range (TAR) Level 2 | > 250 mg/dL (> 13.9 mmol/L) | < 5% of readings (< 1 hr 12 min/day) | Severe hyperglycemia exposure and ketosis risk indicator |
| Glycemic Variability (%CV) | Coefficient of Variation (`SD / Mean × 100`) | ≤ 36% | Separates stable from labile glycemic profiles independent of mean glucose |
| Glycemia Management Indicator (GMI) | Regression formula from mean CGM glucose | Individualized (typically < 7.0%) | Estimates laboratory HbA1c equivalent over short 14-day trial windows |
What Standards or Regulatory Requirements Govern Continuous Glucose Monitoring?
In the United States, the U.S. Food and Drug Administration (FDA) regulates continuous glucose monitoring systems across distinct risk tiers depending on whether the device is adjunctive, non-adjunctive, or interoperable. In 2018, the FDA established the de novo classification codified at 21 CFR § 862.1355 for the Integrated Continuous Glucose Monitoring System (iCGM) as a Class II medical device with special controls (product code `QBJ`).
While summary metrics such as Mean Absolute Relative Difference (MARD) are widely reported in literature (see CGM Accuracy and MARD), 21 CFR § 862.1355(b)(1)–(2) mandates far stricter distributional accuracy and trend performance verified against an FDA-accepted laboratory reference method (such as YSI 2300 STAT Plus) across the entire measuring range. For every point-accuracy criterion below, the statistical lower one-sided 95% confidence bound of paired sensor-to-reference measurements must exceed the specified threshold:
| Reference Glucose Range | Tight Accuracy Band | Required Lower 95% Confidence Bound (Tight Band) | Wide Safety Band | Required Lower 95% Confidence Bound (Wide Band) |
|---|---|---|---|---|
| < 70 mg/dL (Hypoglycemia) | Within ±15 mg/dL | > 85% of measurements | Within ±40 mg/dL | > 98% of measurements |
| 70–180 mg/dL (Euglycemia) | Within ±15% | > 70% of measurements | Within ±40% | > 99% of measurements |
| > 180 mg/dL (Hyperglycemia) | Within ±15% | > 80% of measurements | Within ±40% | > 99% of measurements |
| Entire Measuring Range | Within ±20% | > 87% of measurements | No clinically significant bias | Verified across sensor wear life and insertion sites |
Beyond point accuracy, 21 CFR § 862.1355 requires special controls covering rate-of-change trend accuracy (limiting extreme discordant trend arrows), clinical validation in pediatric and adult cohorts, biocompatibility, interference testing (for example, demonstrating resilience against therapeutic levels of acetaminophen, ascorbic acid, or hydroxyurea), and cryptographic cybersecurity controls for Bluetooth Low Energy (BLE) data transmission.
How Does Continuous Glucose Monitoring Integrate With EHR and Health Informatics Workflows?
Integrating high-frequency CGM data streams (up to 288 observations per patient per day at 5-minute intervals) into hospital EHRs and clinical trial databases requires avoiding transactional database bloat while preserving raw source data provenance:
- HL7 FHIR and LOINC Representation: Under the IEEE 11073-10425 standard and HL7 FHIR R4 `Observation` profiles (driven by the Cross-Paradigm / Argonaut and Gravity/diabetes informatics initiatives), individual interstitial fluid glucose readings are coded separately from capillary whole blood glucose using specific LOINC terms (such as LOINC `99504-3` for interstitial fluid glucose by CGM and structured LOINC panels for AGP summary metrics including TIR, TBR, TAR, and GMI).
- Summary Report vs. Time-Series Ingestion in EHRs: Health systems typically configure middleware or vendor cloud APIs to push two payloads into the EHR: a discrete flowsheet summary containing 14-day AGP statistics (`Observation` resources) and an embedded PDF Ambulatory Glucose Profile report (`DocumentReference` / `DiagnosticReport`) viewable inside the clinician’s chart review tab without flooding the laboratory results table with thousands of 5-minute points.
- Clinical Trial eSource and DHT Governance: When CGM devices are deployed as Wearable Devices and Digital Health Technologies (DHTs) in Clinical Trials under the FDA’s December 2023 DHT guidance, sponsors ingest raw 1-to-5-minute time-series streams (in UTC timestamps with device serial numbers and firmware versions) into CDISC SDTM domain datasets, applying pre-specified statistical analysis plan (SAP) rules for sensor warm-up exclusions, compression artifacts, and missing-data imputation.
Common Methodological Misunderstandings in CGM Research
- Treating Interstitial Glucose as Identical to Instantaneous Capillary Blood Glucose: During rapid glycemic changes (post-meal absorption or exercise), interstitial fluid glucose lags blood glucose by 5 to 15 minutes; clinical trial protocols must specify whether confirmatory blood glucose meter testing is required for symptomatic hypoglycemia adjudication.
- Ignoring Sensor Compression Artifacts (“Compression Lows”): Mechanical pressure on the subcutaneous sensor site during sleep reduces local tissue perfusion and can produce false transient hypoglycemic readings that resolve immediately upon posture change.
- Comparing Unharmonized Sensor Models Across Study Arms: Different sensor generations and calibration algorithms exhibit distinct error distributions in the < 70 mg/dL range; multi-center trials must use a single standardized iCGM model and firmware family across all randomized arms.
Frequently Asked Questions
What is continuous glucose monitoring and how is it used in clinical research and healthcare?
Continuous glucose monitoring (CGM) uses a subcutaneous biosensor to measure interstitial fluid glucose every 1 to 5 minutes. Clinicians and researchers use CGM to track 24-hour glycemic patterns, detect nocturnal hypoglycemia, drive automated insulin delivery (AID) pumps, and compute standardized Ambulatory Glucose Profile (AGP) endpoints such as Time in Range (70–180 mg/dL).
What standards or regulatory requirements govern continuous glucose monitoring?
In the United States, the FDA regulates integrated continuous glucose monitoring systems (iCGMs) as Class II medical devices with special controls under 21 CFR § 862.1355. The regulation mandates strict one-sided 95% lower confidence bounds for accuracy (>85% within ±15 mg/dL below 70 mg/dL, >70% within ±15% from 70–180 mg/dL, >80% within ±15% above 180 mg/dL, and >87% within ±20% overall) alongside interference and cybersecurity controls.
How does continuous glucose monitoring integrate with EHR and health informatics workflows?
CGM data is transmitted via Bluetooth Low Energy (BLE) to mobile apps and cloud portals, which exchange discrete 14-day AGP summary metrics (TIR, TBR, TAR, GMI, %CV) and visual AGP PDF reports with EHRs using HL7 FHIR R4 `Observation` and `DocumentReference` resources and LOINC codes, while streaming raw timestamped time-series data into clinical trial data lakes.
Sources
- U.S. Food and Drug Administration (FDA), 21 CFR § 862.1355 — Integrated Continuous Glucose Monitoring System (Class II Special Controls), Electronic Code of Federal Regulations, ecfr.gov.
- Battelino T, Danne T, Bergenstal RM, et al., Clinical Targets for Continuous Glucose Monitoring Data Interpretation: Recommendations From the International Consensus on Time in Range, Diabetes Care (2019; 42(8): 1593–1603), PubMed Central PMC6973648.
- U.S. Food and Drug Administration (FDA), Digital Health Technologies for Remote Data Acquisition in Clinical Investigations: Guidance for Industry, Investigators, and Other Stakeholders (December 2023), fda.gov.
- Regenstrief Institute, LOINC (Logical Observation Identifiers Names and Codes) — Interstitial Fluid Glucose and Ambulatory Glucose Profile Terms, loinc.org.
