CGM vs A1C — when continuous glucose data overrides A1C

Continuous glucose monitors reveal what A1C cannot: the shape of your glucose day, not just its three-month average. Two people can have identical A1Cs of 7.0% while living completely different glycemic lives — a problem compounded by the engineering limits of CGM sensors that add their own uncertainty layer — one with steady glucose between 80–140 mg/dL all day, the other oscillating between nocturnal lows of 55 mg/dL and post-meal spikes of 220 mg/dL. The A1C average obscures the oscillation. CGM data from the Dexcom G7 or Abbott Libre 3, combined with the time-in-range (TIR) metric — the percentage of time glucose sits between 70 and 180 mg/dL — gives clinicians and patients a far more granular picture of actual glycemic burden. Per ADA Standards of Care 2024 §7, a TIR of ≥70% is now a primary therapeutic target equivalent in clinical weight to achieving an A1C below 7.0%. The shift from A1C to CGM-derived metrics is not incremental — it is a redefinition of what “good” diabetes control means. This guide explains exactly when CGM data should override A1C as the primary decision-making input, what the clinical thresholds are, and how to present CGM data to your diabetes care team.

What A1C measures — and what it systematically misses

HbA1c is a glycated hemoglobin measurement — the percentage of hemoglobin molecules that have glucose attached to them after weeks of circulating in blood plasma. Because red blood cells live approximately 90–120 days, the A1c reflects the average plasma glucose concentration over that period, weighted toward the most recent 30 days (because newer red blood cells are proportionally more abundant than older ones). An A1c of 7.0% corresponds to an estimated average glucose of approximately 154 mg/dL.¹

The structural limitation of A1c is precisely what makes it useful for some purposes and misleading for others. It is a mean — mathematically incapable of capturing variance, distribution, or the shape of the glucose curve across the day. Two glucose profiles that produce identical 90-day averages can have completely different clinical meanings and completely different complication risks.

Consider the “glycaemic roller coaster” scenario: fasting glucose of 65 mg/dL (mildly hypoglycaemic) at 6 a.m., post-breakfast spike to 240 mg/dL (significantly hyperglycaemic) by 8:30 a.m., return to 90 mg/dL by noon. This patient’s A1c may be 6.5% — technically below the diagnostic threshold for diabetes — while they are experiencing both hypoglycaemia and significant hyperglycaemia daily. The A1c would pass a clinical review; the glucose profile would not.²

Additionally, A1c is falsified in several physiological conditions. Haemolytic anaemia, iron-deficiency anaemia, haemoglobin variants (HbS, HbC, HbE), and recent blood transfusions all alter red blood cell turnover in ways that make A1c values unreliable. A patient with sickle cell trait may have a spuriously low A1c that does not reflect actual glucose control. In these cases, CGM-derived average glucose is the only reliable mean glucose metric. The A1C to eAG conversion explains how to translate between these different measurement systems in practice.³

A1c also gives no information about hypoglycaemia frequency or severity. A patient who experienced three severe nocturnal lows (below 54 mg/dL) in the past month has a serious safety issue that their A1c of 6.8% will not flag. CGM time below range (TBR) — specifically the percentage of time below 70 mg/dL and below 54 mg/dL — is the metric that captures this risk.

The time-in-range framework — the CGM metric that matters most

The 2019 International Consensus on Time in Range, developed by 43 diabetes experts from 22 countries and published in Diabetes Care, established standardized TIR targets as the primary CGM-derived clinical metrics.⁴ These targets are now incorporated into ADA Standards of Care 2024.

For most non-pregnant adults with Type 1 or Type 2 diabetes, the targets are:

• Time in range (TIR): ≥70% of time between 70–180 mg/dL (≥16.8 hours/day)
• Time below range (TBR) — Level 1: <4% of time below 70 mg/dL (<58 minutes/day)
• Time below range (TBR) — Level 2: <1% of time below 54 mg/dL (<15 minutes/day)
• Time above range (TAR) — Level 1: <25% of time above 180 mg/dL
• Time above range (TAR) — Level 2: <5% of time above 250 mg/dL

These targets function as a set — achieving TIR ≥70% while violating TBR Level 1 is not “good” diabetes control. The TBR targets are safety constraints; TIR and TAR targets are efficacy measures.

Epidemiological data linking TIR to outcomes comes primarily from the DCCT/EDIC cohort (Type 1 diabetes) and the T2D-REAL CGM substudy. Each 10-percentage-point increase in TIR is associated with a 40% reduction in proliferative diabetic retinopathy risk and a 25% reduction in microalbuminuria incidence — complication risk reductions comparable to those achieved by reducing A1c from 8% to 7%.⁴

For pregnant women with diabetes, TIR targets are tighter: ≥70% time between 63–140 mg/dL (lower upper threshold), with TBR Level 1 below 4% and TBR Level 2 below 1%. The tighter targets reflect the teratogenic risk of maternal hyperglycaemia during organogenesis and fetal growth.

When CGM data should drive clinical decisions over A1C

CGM data should become the primary management driver in five specific clinical scenarios, based on evidence-based clinical guidelines:

1. Frequent nocturnal hypoglycaemia. Nocturnal lows below 70 mg/dL — particularly Level 2 lows below 54 mg/dL — are a safety emergency that A1c cannot flag. If a CGM report shows TBR Level 2 exceeding 1% of time, treatment adjustment is indicated regardless of A1c. Typical interventions: reducing basal insulin dose, changing basal injection timing, adjusting pre-dinner carbohydrate, or transitioning to a closed-loop insulin delivery system.¹

2. High glucose variability despite normal A1c. Coefficient of variation (CV) greater than 36% indicates high glucose variability that produces equivalent complication risk to elevated mean glucose, even when the mean (and A1c) appears normal. High-amplitude glucose oscillations generate more oxidative stress than equivalent steady-state hyperglycaemia. If a patient’s A1c is 6.8% but their CV is 42% and their TBR Level 1 is 8%, the clinical picture is worse than the A1c suggests.²

3. Haemoglobin disorders and conditions that falsify A1c. Sickle cell trait, thalassaemia, haemolytic anaemia, iron deficiency, and post-transfusion states can all produce A1c values that do not correlate with actual glucose control. In these patients, CGM-derived glucose management indicator (GMI) — a formula that estimates A1c from mean CGM glucose — is the reliable alternative: GMI (%) = 3.31 + 0.02392 × mean glucose (mg/dL).³

4. Pregnancy. Gestational diabetes management requires tighter targets and faster response to out-of-range glucose than quarterly A1c reviews can support. CGM provides real-time data that allows dietary and insulin adjustments within days rather than weeks. The ADA recommends CGM use in all pregnancies complicated by pre-existing Type 1 diabetes; its benefit in gestational diabetes and Type 2 is increasingly supported by evidence.¹

5. Intensive insulin therapy. Any patient on multiple daily injections (MDI) or continuous subcutaneous insulin infusion (CSII/pump) benefits from CGM as the primary management tool. The dose decisions — basal rate adjustment, correction factor, insulin-to-carb ratio — all depend on time-series glucose data that A1c cannot provide. A1c is used for quarterly review and as a cross-check against CGM-derived GMI; CGM data drives daily management.¹

Reading a CGM ambulatory glucose profile — the 14-day snapshot

The Ambulatory Glucose Profile (AGP) is the standardized CGM output format, developed by the International Diabetes Center and endorsed by the ADA.⁴ A 14-day AGP compresses two weeks of continuous glucose data — typically 2,000–4,000 readings from a 5-minute-interval CGM — into a single visual display used for clinical review.

The AGP contains several layers. The median glucose line shows the 50th percentile glucose at each time point across all 14 days — the “middle” glucose pattern. The interquartile band (25th–75th percentile, shown in light shading) represents the range within which glucose falls on most days. The 10th and 90th percentile overlays (dotted lines) show the variability range — the wider this band, the more variable the glucose.

A narrow IQR band with the median tracking near 100 mg/dL indicates consistent, near-normal glucose patterns throughout the day — exactly what good CGM management looks like. A wide band — especially a wide band that dips below 70 mg/dL at night (suggesting hypoglycaemia) while spiking above 180 mg/dL post-meal — indicates glycaemic variability that needs addressing regardless of where the median line sits.

The AGP also includes a statistics summary: TIR, TBR Level 1 and 2, TAR Level 1 and 2, mean glucose, CV, and GMI. These six statistics, read together, give a complete picture of the 14-day glucose pattern in less than a minute.

To bring your AGP to a clinical appointment: most CGM apps (Dexcom CLARITY, LibreView, Libre LinkUp) generate a printable AGP PDF from 14 days of data. Download the PDF from the web portal and print it or share it digitally before your endocrinology visit. Walk in with the AGP in hand — it is the single document that most efficiently communicates your glucose pattern to a clinician.

How CGM data reshapes meal-timing and dietary choices

The post-meal glucose trace on a CGM provides individual metabolic feedback that no population-average GI table can replicate. When you eat a meal and observe your CGM trace over the following two hours, you see your glucose response to that food at that portion size — information that is physiologically personalized in a way that published GI data is not.

A healthy post-meal glucose curve peaks between 45–60 minutes after eating, reaches no higher than 180 mg/dL (with 140 mg/dL as an aspirational upper limit for well-controlled patients), and returns to fasting baseline within 2 hours. Deviation from this pattern — a spike above 180, a peak that persists past 90 minutes, or a late dip into hypoglycaemia 2–3 hours post-meal — indicates either the meal composition or the insulin dose (in insulin-using patients) needs adjustment.

Running a structured “food experiment” with CGM: eat the same meal (same food, same portion, same time of day) on three separate days. Compare the three CGM traces. Consistency across traces confirms the meal-glucose relationship is reproducible and not confounded by activity, stress, or prior glucose levels. Once you have a reproducible response, you know whether that meal is within your target range or needs modification — reduce the rice portion, add fiber, adjust the protein-to-carbohydrate ratio.

Pairing CGM with CalEye’s meal log creates the most informative dataset for this analysis: CalEye records what you ate (with portion estimates and GL per component), and the CGM records your glucose response. The postprandial glucose variability guide explains how to interpret the curve shapes that emerge from this paired analysis. Over 4–6 weeks of paired data, the pattern of which meals produce above-target responses becomes visible — often revealing that one or two specific foods or portion sizes account for the majority of out-of-range readings.

Choosing between CGM devices — accuracy, cost, and data access

The Dexcom G7 and Abbott FreeStyle Libre 3 dominate the CGM market as of 2025, with similar accuracy profiles but meaningfully different feature sets.

Accuracy (MARD): The G7 reports a MARD of 8.2% in its FDA submission pivotal trial; the Libre 3 reports 7.9%.⁵ Both meet the ISO 15197:2013 accuracy standard (≥95% of readings within 15% of reference glucose or within 15 mg/dL in the hypoglycaemic range). At glucose values below 70 mg/dL, both sensors show slightly higher MARD — approximately 10–12% — which is why hypoglycaemia alarms are set conservatively at 70 mg/dL rather than 54 mg/dL.

Calibration: The G7 does not require fingerstick calibration (factory calibrated); the Libre 3 is also factory calibrated. This represents an improvement over older models that required twice-daily calibration and improves real-world accuracy by eliminating user-calibration error.

Wear time and waterproofing: G7 wears for 10 days; Libre 3 wears for 14 days. Both are waterproof to 1 meter for 30 minutes. For active users, the Libre 3’s 14-day wear reduces sensor change frequency and cost.

Alarms: The G7 provides real-time Bluetooth alarms for hypoglycaemia, hyperglycaemia, and urgent low glucose to a smartphone and to the G7 receiver. The Libre 3 provides optional Bluetooth alarms to a smartphone (via the LibreLink app) but not to a dedicated reader. For patients who require reliable hypoglycaemia alarming — particularly those with hypoglycaemia unawareness — the G7’s alarm system is more robust.

Cost and insurance: Both require a prescription for insurance coverage in the US. Out-of-pocket cost for uninsured patients is approximately $400–500/month for sensors alone, which is a significant barrier. Abbott’s Lingo (OTC, US) and Libre Without Prescription (UK) are entering the consumer market at lower price points for non-diabetic metabolic health users.

References

1. American Diabetes Association Professional Practice Committee. “Diabetes Technology: Standards of Care in Diabetes—2024.” Diabetes Care 47, Supplement 1 (2024): S126–S144. Section 7.

2. Hirsch IB. “Glycemic Variability and Diabetes Complications: Does It Matter? Of Course It Does!” Diabetes Care 38, no. 8 (2015): 1610–1614.

3. Bergenstal RM, Beck RW, Close KL, et al. “Glucose Management Indicator (GMI): A New Term for Estimating A1C From Continuous Glucose Monitoring.” Diabetes Care 41, no. 11 (2018): 2275–2280.

4. 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 42, no. 8 (2019): 1593–1603.

5. Dexcom G7 CGM System User Guide. San Diego: Dexcom, 2023. Accuracy data from pivotal clinical trial (FDA 510k submission K221745).