Continuous Glucose Monitoring

Table of Contents

1. How CGM Physically Works
2. The Device Landscape (10-Day, 14-Day, Over-the-Counter and Implanted Sensors)
3. Time-in-Range as the New Gold Standard
4. Why Hemoglobin A1c Hides What Matters
5. Individual Variation (Zeevi, Hall Glucotypes)
6. CGM in Non-Diabetic Adults
7. Accuracy, Lag, and Interference
8. Practical CGM Self-Experiment Protocol
9. Key Research Papers
10. Connections
11. Featured Videos

How CGM Physically Works

Modern CGM sensors use the same fundamental electrochemistry as fingerstick meters: a glucose oxidase enzyme catalyzes the oxidation of glucose to gluconolactone, producing hydrogen peroxide which is detected amperometrically (the current generated is proportional to glucose concentration). The difference is that the sensor is implanted in subcutaneous tissue and runs continuously.

A typical sensor consists of:

1. Insertion needle — spring-loaded, typically < 5 mm depth, retracted immediately after sensor placement.
2. Filament — a flexible 5 mm long, 0.4 mm diameter wire of platinum or other inert material coated with immobilized glucose oxidase and a permselective membrane to exclude interfering substances.
3. Transmitter housing — sits on the skin surface, contains the analog-to-digital converter, microcontroller, battery, and Bluetooth radio.
4. Smartphone app or dedicated receiver — displays the current glucose value, trend arrow, and 24-hour curve. Most modern systems also send the data to a cloud service for healthcare provider review.

The filament measures glucose in interstitial fluid, not blood. There is an inherent 5-15 minute lag between blood glucose changes and interstitial fluid glucose changes, which becomes clinically meaningful during rapid changes (early after meals, during exercise, or correcting a hypoglycemic episode). Modern algorithms partially correct for this lag using trend extrapolation, but the lag cannot be eliminated entirely.

Sensor lifespan ranges from 7 days (older-generation sensors) to 10-15 days (current 10-day and 14-day sensors) to 365 days (the current implanted sensor; earlier versions lasted 90 and then 180 days). All require periodic replacement and most require warm-up periods of 30 minutes to 24 hours after insertion before producing usable readings.

Back to Table of Contents

The Device Landscape (10-Day, 14-Day, Over-the-Counter and Implanted Sensors)

The current US market (2024-2026):

• 10-day sensor — 1-minute readings, factory-calibrated (no fingerstick calibration required), mean absolute relative difference (MARD, the average reading error; see Accuracy below) approximately 8.2%. Widely used in Type 1 and intensively-managed Type 2 diabetes. Predictive low-glucose alerts. Pairs with smartwatches.
• 14-day sensor — 1-minute readings, factory-calibrated, MARD approximately 7.9%. Lower cost per day than the 10-day sensor. Optional alerts. The most widely prescribed CGM type in Type 2 diabetes.
• Over-the-counter CGM, 15-day (U.S. Food and Drug Administration-cleared 2024) — no prescription required, aimed at non-diabetic adults and Type 2 diabetics not on insulin. No alarms. The FDA's March 2024 clearance covers adults 18 and older who do not use insulin; the FDA states it is not for people with problematic hypoglycemia (low blood sugar), because the system is not designed to alert the user to this potentially dangerous condition. Approximately $89/month direct-to-consumer subscription. The first CGM marketed at the metabolic-curiosity rather than diabetic-management user.
• Implanted sensor — surgically placed in the upper arm under local anesthesia; the current version lasts 365 days (earlier versions lasted 90 and then 180 days). In its one-year accuracy study (ENHANCE, Bailey 2025), 110 adults with diabetes were implanted: the average error (MARD) was 8.8% with mostly one fingerstick calibration per week, alerts caught 96.6% of confirmed readings at 70 mg/dL and 97.9% at 180 mg/dL, 90% of sensors lasted the full 365 days, and no related serious adverse events were reported. A niche option used for long-duration monitoring without repeated insertions.
• Over-the-counter CGM, 14-day — a second over-the-counter product built on the 14-day sensor platform, no alarms.

The clinical-medical CGMs (the 10-day and 14-day prescription sensors) are cleared by the U.S. Food and Drug Administration (FDA) as Class II devices ("integrated continuous glucose monitoring systems", Class II with special controls under 21 CFR 862.1355) for diabetes management, including dosing of insulin from the CGM reading. The over-the-counter CGMs are explicitly not for insulin dosing and are positioned as "general wellness" devices.

Insurance coverage in the United States: Medicare and most commercial insurers cover CGM for Type 1 diabetes and for Type 2 diabetes on intensive insulin therapy (3+ injections per day). Coverage for Type 2 on basal insulin or oral medications is increasingly common as of 2024 following the MOBILE trial evidence. Coverage for non-diabetic use remains zero; cash-pay only.

Back to Table of Contents

Time-in-Range as the New Gold Standard

The Battelino 2019 international consensus on time in range (TIR) established a new framework for evaluating glycemic control that captures information hemoglobin A1c cannot. The consensus targets for Type 1 and Type 2 diabetes:

• Time in Range (TIR) — percentage of CGM readings in 70-180 mg/dL. Target: >70% (corresponds approximately to hemoglobin A1c < 7%).
• Time Below Range (TBR), Level 1 — percentage in 54-69 mg/dL. Target: <4%.
• Time Below Range (TBR), Level 2 — percentage <54 mg/dL. Target: <1%.
• Time Above Range (TAR), Level 1 — percentage 181-250 mg/dL. Target: <25%.
• Time Above Range (TAR), Level 2 — percentage >250 mg/dL. Target: <5%.
• Glucose Variability (CV) — coefficient of variation. Target: <36%.
• Glucose Management Indicator (GMI) — an estimated hemoglobin A1c calculated from average CGM glucose using the Bergenstal 2018 formula. Useful for comparing CGM-derived control to lab hemoglobin A1c.

For non-diabetic adults interested in metabolic optimization, a tighter target range is proposed by some practitioners (it is not part of the Battelino consensus): 70-140 mg/dL with target TIR >90%. The Hall 2018 glucotypes paper showed that even non-diabetic adults frequently exceed 140 mg/dL after common foods, often without symptoms.

The Lu 2018 Diabetes Care study of 3,262 people with type 2 diabetes found that diabetic retinopathy (eye damage from diabetes) became less common with each higher quarter of TIR, and TIR stayed linked to every stage of retinopathy after adjustment for age, sex, body mass index, diabetes duration, blood pressure, blood fats and HbA1c. It was a one-time (cross-sectional) comparison, so it shows an association, not cause and effect; adding glucose-variability measures partly weakened the links, although the link between TIR and any retinopathy remained. TIR adds information beyond hemoglobin A1c, particularly because TIR captures the glucose variability and postprandial excursion patterns that drive microvascular damage.

Back to Table of Contents

Why Hemoglobin A1c Hides What Matters

Hemoglobin A1c reflects the percentage of hemoglobin molecules that have been non-enzymatically glycated by exposure to circulating glucose over the prior 90-120 days (the lifespan of erythrocytes). It is a useful integrated measure of average glucose exposure, but it has three significant blind spots:

1. Average does not capture variability — two patients with identical hemoglobin A1c of 6.5% can have radically different glucose profiles. One can be steady throughout the day with no excursion above 160 mg/dL; the other can spike to 250 mg/dL three times a day and rebound to 60 mg/dL at night, with the highs and lows averaging to the same value. The second pattern produces far more vascular damage despite identical hemoglobin A1c.
2. Hemoglobin variants and turnover affect the measurement — hemoglobin A1c is falsely low in conditions of accelerated red cell turnover (hemolytic anemia, recent blood loss, pregnancy, hereditary spherocytosis), and falsely high in conditions of slowed turnover (iron deficiency anemia, vitamin B12 deficiency, splenectomy). Sickle cell trait, hemoglobin C, hemoglobin E, and hemoglobin F all interfere with hemoglobin A1c assays to varying degrees depending on the specific assay method.
3. Hemoglobin A1c lags behavior changes by 8-12 weeks — a patient who makes a major dietary change today will not see the effect on hemoglobin A1c for two to three months. CGM shows the effect within hours.

For patients who are already at goal hemoglobin A1c but having unrecognized variability, or for non-diabetic patients trying to understand their own glucose dynamics, hemoglobin A1c is uninformative. CGM is the only available tool that surfaces this information.

Back to Table of Contents

Individual Variation (Zeevi, Hall Glucotypes)

Two influential papers established the individual-variation case for personalized rather than population-average dietary advice:

Zeevi et al. (Cell 2015) — the Weizmann Institute Personalized Nutrition Project. 800 healthy and prediabetic Israeli adults wore CGM for one week each while logging meals via smartphone; the study measured responses to 46,898 meals. Findings:

• High between-subject variability in postprandial glucose response to identical foods. The same standardized banana could produce a flat curve in one subject and a 100 mg/dL spike in another.
• A machine-learning algorithm trained on subject characteristics (microbiome composition, blood biomarkers, anthropometrics, food log) accurately predicted each person's postprandial glucose response to real-life meals. The predictions were then confirmed in a separate 100-person cohort.
• In a blinded randomized controlled dietary intervention, diets based on the algorithm produced significantly lower postprandial glucose responses and consistent changes in the gut microbiota. The authors concluded that universal dietary recommendations may have limited usefulness.

Hall et al. (PLoS Biology 2018) — the Stanford Glucotypes study. 57 non-diabetic adults wore CGM for 2-4 weeks and were challenged with standardized meals (cornflakes + milk; bread + peanut butter; protein bar). Findings:

• Identified three distinct "glucotypes" based on glucose variability patterns: low (stable), moderate (intermediate variability), and severe (high variability with frequent excursions >140 mg/dL).
• Even people considered normoglycemic (normal blood sugar) by standard tests showed high glucose variability on CGM, with glucose in the prediabetic range 15% of the time and in the diabetic range 2% of the time — excursions that the conventional tests had missed.
• The cornflakes + milk meal produced postprandial spikes >200 mg/dL in many subjects with otherwise normal labs — demonstrating that "healthy" non-diabetic adults often have substantial postprandial dysglycemia.

The take-home from both papers: population-average dietary advice is a starting point, but the actual postprandial response to specific foods varies enough between individuals that personalized measurement adds clinically meaningful information. CGM is the practical tool for that measurement.

Back to Table of Contents

CGM in Non-Diabetic Adults

The case for CGM in non-diabetic adults is still emerging, with both proponents and skeptics. The proponent argument:

1. Postprandial excursions above 140 mg/dL drive endothelial dysfunction and advanced glycation end-product (AGE) formation regardless of diabetic status.
2. People with normal standard test results spent 15% of the time in the prediabetic glucose range and 2% in the diabetic range on CGM (Hall 2018) — patterns invisible to standard testing.
3. Individual variation in response to specific foods is large enough that personalized measurement adds value beyond glycemic index (GI) tables.
4. CGM provides immediate feedback on the impact of exercise timing, meal sequencing, sleep, and stress — supporting behavior change with quantitative evidence.
5. Early identification of subtle dysglycemia could enable interventions a decade before hemoglobin A1c becomes diagnostic.

The skeptic argument:

1. No randomized controlled trials demonstrate that CGM use in non-diabetic adults improves hard endpoints (cardiovascular events, mortality).
2. CGM in healthy adults can drive obsessive food restriction and disordered eating in susceptible individuals.
3. The cost is non-trivial ($1,000+/year cash) and unreimbursed.
4. Most actionable insights (less added sugar, exercise after meals, losing excess weight) are already known without CGM.
5. Sensor lag and accuracy limitations may produce misleading data, particularly for rapid changes around exercise.

A 2023 review of CGM use by people without diabetes (Klonoff 2023) found four groups of users: people with metabolic diseases related to diabetes in which the insulin-glucose system is the main problem, people with other metabolic diseases, health and wellness users, and elite athletes. The review describes the use of these devices outside diabetes as growing as supporting data accumulate, and expects clinicians to learn more about CGM patterns and glucotypes that predict a higher risk of diabetes.

A middle position common among self-trackers: a one-month CGM self-experiment, which they describe as revealing individual food responses, exercise effects, and unrecognized postprandial spikes — without lifelong wear. Many users report finding one or two months sufficient to develop a personalized understanding of their glucose responses, then discontinue. Since the U.S. Food and Drug Administration clearance of over-the-counter CGMs in 2024, a CGM can be obtained without a prescription.

Back to Table of Contents

Accuracy, Lag, and Interference

Modern CGM accuracy is described by mean absolute relative difference (MARD), the average percentage difference between the CGM reading and a reference blood glucose measurement (typically a laboratory glucose analyzer). Current best-in-class CGMs achieve MARD around 8-9%, meaning the typical reading is within ~8% of the true value. For comparison, fingerstick meters achieve MARD 5-7%.

In practical terms, a MARD of 8% means that for a true glucose of 100 mg/dL, the CGM reading will typically fall in the range of 92-108 mg/dL. At higher glucose values (e.g., true 200 mg/dL), the absolute error is correspondingly larger (typically 184-216 mg/dL). The accuracy is best in the middle range and degrades somewhat at the extremes.

Key accuracy considerations:

• Lag time — 5-15 minutes between blood glucose and interstitial glucose. Becomes clinically significant during rapid changes (early after meals, during exercise, during hypoglycemia correction). Modern algorithms partially compensate via trend extrapolation but cannot eliminate the lag.
• Acetaminophen interference — older-generation sensors of the 10-day family gave falsely high readings when patients took acetaminophen. The two most recent generations have largely resolved this through improved membrane chemistry.
• Vitamin C (ascorbate) — high-dose vitamin C (>1 g/day) interferes with the first two generations of the 14-day sensor family, giving falsely high readings. The third generation has improved performance.
• Mannitol, salicylates, hydroxyurea — documented interferences in specific sensor models; each model's package insert lists its known interfering substances.
• Compression artifacts — sleeping on the sensor or wearing tight clothing over it can produce false low readings.
• Day 1 inaccuracy — most sensors show reduced accuracy in the first 12-24 hours after insertion while inflammation around the filament resolves.
• End-of-life drift — sensors typically become less accurate in the final 24-48 hours before scheduled replacement.

Common practice among CGM users is to confirm clinically critical decisions with a fingerstick, to rely on the CGM trend for routine management, and to re-check a suspicious individual reading. The FDA states that users of the over-the-counter sensor should not make medical decisions based on its output without talking to their healthcare provider. Non-diabetic self-trackers, who are not dosing medication, commonly use the readings for behavioral self-experimentation without fingerstick confirmation.

Back to Table of Contents

Practical CGM Self-Experiment Protocol

A protocol described for non-diabetic adults using a CGM (an over-the-counter CGM, or a prescription sensor used off-label) for one month of metabolic self-experimentation, as people who self-track commonly run it:

1. Days 1-3: Baseline observation. The person eats their normal diet with no changes, watches the trace for patterns, and notes the foods that drive the largest excursions.
2. Day 4-7: Standardized food challenges. On consecutive mornings, after an overnight fast, each of the following is eaten on its own and the 2-hour response is watched:
   • 50 g rice, white or brown (or 1 slice of bread, or 1 medium banana — whichever the person most commonly eats)
   • Steel-cut oats with no fruit or sweetener (1/2 cup dry)
   • 2 eggs with no carbohydrate
   • Greek yogurt (full-fat, plain, no fruit) with 10 berries
3. Days 8-14: Sequencing experiments. A starchy meal is eaten three ways on three consecutive days:
   • Starch first, then protein, then vegetables (typical Western order)
   • All mixed together
   • Vegetables and protein first, starch last
   The postprandial peaks and the total incremental area under the curve (iAUC) are compared.
4. Days 15-21: Exercise effects.
   • A normal lunch followed by 2 hours of sitting; the curve is recorded.
   • The next day, the same lunch followed by a brisk 10-15 minute walk starting 30 minutes after eating; the two curves are compared.
   • Morning resistance training, with glucose effects observed through the following day.
5. Days 22-28: Time-restricted eating. The eating window is compressed to 10 hours (e.g., 8 AM to 6 PM) for one week while the overnight glucose, dawn phenomenon, and morning fasting values are observed.
6. Documentation. Self-trackers take screenshots of the daily trace with notes on what was eaten, when, and what activity preceded or followed. By the end of the month they describe having a personalized record of what raises their glucose, by how much, and for how long.

Self-trackers commonly describe one month of disciplined CGM self-experimentation as enough to make durable dietary and lifestyle changes informed by their own physiology, with continued wear optional — many users report finding a quarterly "tune-up" two-week wear sufficient to verify behaviors remain effective.

Back to Table of Contents

Key Research Papers

1. Battelino T, Danne T, Bergenstal RM et al. (2019). Clinical targets for continuous glucose monitoring data interpretation: recommendations from the international consensus on time in range. Diabetes Care, 42(8):1593-1603. — PubMed PMID: 31177185
2. Beck RW, Riddlesworth T, Ruedy K et al. (2017). Effect of continuous glucose monitoring on glycemic control in adults with type 1 diabetes using insulin injections: the DIAMOND randomized clinical trial. JAMA, 317(4):371-378. — PubMed PMID: 28118453
3. Martens T, Beck RW, Bailey R et al. (2021). Effect of continuous glucose monitoring on glycemic control in patients with type 2 diabetes treated with basal insulin: a randomized clinical trial (MOBILE). JAMA, 325(22):2262-2272. — PubMed PMID: 34077499
4. Lu J, Ma X, Zhou J et al. (2018). Association of time in range, as assessed by continuous glucose monitoring, with diabetic retinopathy in type 2 diabetes. Diabetes Care, 41(11):2370-2376. — PubMed PMID: 30201847
5. Hall H, Perelman D, Breschi A et al. (2018). Glucotypes reveal new patterns of glucose dysregulation. PLoS Biology, 16(7):e2005143. — PubMed PMID: 30040822
6. Zeevi D et al. (2015). Personalized nutrition by prediction of glycemic responses. Cell. — PubMed PMID: 26590418
7. Bergenstal RM, Beck RW, Close KL et al. (2018). Glucose management indicator (GMI): a new term for estimating A1C from continuous glucose monitoring. Diabetes Care, 41(11):2275-2280. — PubMed PMID: 30224348
8. Beck RW, Bergenstal RM, Riddlesworth TD et al. (2019). Validation of time in range as an outcome measure for diabetes clinical trials. Diabetes Care, 42(3):400-405. — PubMed PMID: 30352896
9. Vigersky RA, McMahon C (2019). The relationship of hemoglobin A1C to time-in-range in patients with diabetes. Diabetes Technol Ther, 21(2):81-85. — PubMed PMID: 30575414
10. Monnier L, Lapinski H, Colette C (2003). Contributions of fasting and postprandial plasma glucose increments to the overall diurnal hyperglycemia of type 2 diabetic patients: variations with increasing levels of HbA1c. Diabetes Care, 26(3):881-885. — PubMed PMID: 12610053
11. Aleppo G, Ruedy KJ, Riddlesworth TD et al. (2017). REPLACE-BG: a randomized trial comparing continuous glucose monitoring with and without routine blood glucose monitoring in adults with well-controlled type 1 diabetes. Diabetes Care, 40(4):538-545. — PubMed PMID: 28209654
12. Bailey TS, Liljenquist DR, Denham DS et al. (2025). Evaluation of accuracy and safety of the 365-day implantable Eversense continuous glucose monitoring system: the ENHANCE study. Diabetes Technol Ther, 27(5):407-411. — PubMed PMID: 39869118
13. Klonoff DC, Nguyen KT, Xu NY et al. (2023). Use of continuous glucose monitors by people without diabetes: an idea whose time has come? J Diabetes Sci Technol, 17(6):1686-1697. — PubMed PMID: 35856435

PubMed Topic Searches

1. PubMed: CGM in diabetes
2. PubMed: Time in range
3. PubMed: CGM in non-diabetic adults
4. PubMed: Glucose variability and endothelial
5. PubMed: Postprandial glucose and CV risk

Back to Table of Contents

Connections

• All Remedies
• Blood Sugar (Main)
• Glycemic Index & Load
• Insulin Resistance
• Exercise & Meal Timing
• Lab Tests
• Endocrinology
• Ketogenic Diet
• Intermittent Fasting
• Berberine
• Cinnamon
• Chromium
• Magnesium
• Food

Back to Table of Contents

Featured Videos