How Accurate Are CGMs? and What the Data Really Means

It is a fair question. If you are going to base coaching decisions on CGM data, you need to understand what the data represents, how closely it aligns with the "true" blood glucose, and under what circumstances the readings are most and least reliable. Getting this wrong in either direction creates problems. Overconfidence in CGM numbers leads to coaching decisions based on readings that do not accurately reflect what is happening in the bloodstream. Excessive skepticism about accuracy leads professionals to dismiss data that is genuinely informative.

This post covers how CGM accuracy is measured, what the current accuracy standards mean in practical terms, the specific conditions that reduce accuracy, and how personal trainers, health coaches, physical therapists, chiropractors, and registered dietitians should think about accuracy when using CGM data within the framework of the Glucose Pattern Recognition Methodology™ (GPRM™).

How CGM Accuracy Is Measured: Understanding MARD

CGM accuracy is expressed using a metric called MARD, or Mean Absolute Relative Difference. MARD measures the average percentage difference between CGM readings and simultaneous reference measurements of blood glucose, taken either from a calibrated laboratory analyzer or a clinical-grade fingerstick glucometer.

The calculation is straightforward: for each paired reading, you take the absolute difference between the CGM value and the reference value, divide by the reference value, and express it as a percentage. The mean of all those paired percentages across a study population and time period is the MARD. A MARD of 10 percent means that, on average, the CGM reads within 10 percent of the true blood glucose value.

Most modern CGMs used in wellness contexts fall within the 8-10% MARD range. The Dexcom Stelo, the first over-the-counter CGM cleared by the FDA in March 2024, achieved an overall MARD of approximately 8.7 percent in its pivotal study.¹ Abbott Lingo, the second OTC-cleared device, achieved comparable performance. For context, the ISO standard for clinical-grade blood glucose meters requires a MARD of 15 percent or better. Consumer CGMs are performing well within that threshold.

What an 8 to 10 percent MARD means in practical terms: if a client's true blood glucose is 100 mg/dL, a CGM with 10 percent MARD will typically read somewhere between 90 and 110 mg/dL. At 180 mg/dL, the typical range is roughly 162 to 198 mg/dL. The absolute numerical difference increases at higher glucose values, but the relative accuracy remains constant.

For the pattern-based coaching application of CGM data, this level of accuracy is entirely sufficient. The question a coach is answering is not "is this person's glucose exactly 103 mg/dL?" It is "what direction is glucose trending, how does this post-meal response compare to yesterday's, and does this overnight trace look stable or variable?" Trends, patterns, and relative comparisons are robust to the small numerical differences introduced by MARD.

When CGM Accuracy Is Most Reliable

CGM readings are most accurate under stable glucose conditions. When glucose is relatively flat or changing slowly, the interstitial fluid glucose that the sensor reads closely tracks blood glucose, and the correlation between CGM values and true blood glucose is strongest.

This is the majority of a typical client's day. Most of the overnight trace, most of the between-meal periods, and stable portions of post-meal recovery all fall within conditions where CGM accuracy is at its best. For the patterns that matter most in fitness and wellness coaching, the data is reliable.

Accuracy is also influenced by proper sensor placement and wear. The upper arm, the standard placement site for most OTC CGM devices, produces consistent readings when the sensor is placed correctly according to manufacturer instructions. Sensors that are poorly placed, partially detached, or worn beyond their recommended wear period produce less reliable data.

When CGM Accuracy Is Reduced: Four Specific Conditions

There are four specific conditions under which CGM readings are less reliable. Understanding each allows a professional to interpret data with appropriate context rather than either dismissing all CGM data or treating every reading as precise.

Rapid glucose changes. The physiological lag between blood glucose and interstitial glucose, typically five to fifteen minutes, becomes most clinically significant when glucose is changing rapidly. During the acute rise of a post-meal spike, the CGM is reading a value that reflects where blood glucose was a few minutes ago, not where it is right now. This means the peak on the CGM trace may appear slightly delayed and slightly lower than the true blood glucose peak. During a rapid decline, the CGM may read higher than true blood glucose for the same reason. For the pattern-based coaching application of the GPRM™, this lag is a known feature of the data rather than a problem, because the pattern shape and relative comparisons remain meaningful even with a small time offset.

Compression artifacts during sleep. When a client sleeps directly on the arm where the CGM sensor is placed, body weight can compress the tissue around the sensor, temporarily reducing interstitial fluid flow and producing an artificially low reading. These compression artifacts typically appear as a sudden, sharp drop in the overnight CGM trace, sometimes reaching values well below any physiologically plausible overnight glucose level, followed by an equally abrupt return to normal once the pressure is relieved. They look distinctive on the trace once you know what they are: too sharp, too low, and too brief to reflect a real physiological event. Teaching clients to identify these artifacts prevents unnecessary alarms and prevents coaching responses to non-events.

Sensor warm-up period. All CGM devices require a sensor initialization period after placement, typically 1 to 2 hours, during which the sensor equilibrates with interstitial fluid and readings are less reliable. Most devices alert users not to rely on readings during warm-up. Coaching decisions should not be based on readings from this period.

Dehydration. Because CGMs measure glucose concentration in interstitial fluid, the volume and composition of that fluid affects readings. Significant dehydration, which reduces interstitial fluid volume and can concentrate solutes, may cause CGM readings to appear higher than true blood glucose. This is particularly relevant for clients who exercise intensely, train in heat, or have inadequate fluid intake. When a client's CGM shows unexpectedly elevated readings alongside other signs of dehydration, rehydration, and retesting before drawing coaching conclusions is the appropriate response.

What CGM Accuracy Means for Coaching Decisions

The most important principle for coaching with CGM data is one that the GPRM™ makes explicit: decisions are based on patterns over time, not on individual readings.

A single CGM reading of 155 mg/dL might reflect true blood glucose, a compression artifact, a reading taken during rapid change, or a minor sensor inaccuracy. Any of those explanations could be correct. A consistently elevated pattern of readings in the 150 to 170 mg/dL range across multiple post-meal events over two weeks, independent of the accuracy variation of any single point, is meaningful.

This pattern-first orientation is also what protects professionals and clients from the clinical errors that can arise from over-reliance on individual CGM numbers. A CGM reading that appears elevated should not trigger a clinical conclusion. It should trigger a review of the surrounding context, a check for the four accuracy-reducing conditions described above, and, if the pattern persists consistently, a referral conversation rather than an independent coaching response.

For the specific referral threshold, any CGM pattern showing consistently elevated fasting or post-meal glucose that lacks a clear coaching explanation and does not improve with sustained lifestyle intervention warrants physician evaluation. The coach's role is to observe the pattern and initiate the referral. Diagnosis is outside the scope of every fitness and wellness professional, regardless of how clearly the data appears to point in a particular direction.

CGM Accuracy and Clinical Diagnosis: A Hard Boundary

This point deserves its own section because it consistently arises in professional practice. CGM devices used in over-the-counter wellness contexts are not cleared for clinical diagnosis. The FDA authorization for the Dexcom Stelo, for example, specifies that it is intended for use by adults 18 and older without diabetes who want to understand how lifestyle factors affect their glucose. It is not indicated for diagnosis, and its performance specifications are not evaluated against the diagnostic use case.

A clinical diagnosis of prediabetes or diabetes requires a venous fasting blood glucose draw, an HbA1c test, or an oral glucose tolerance test, interpreted by a licensed physician or qualified healthcare provider. No CGM reading, regardless of what it shows, substitutes for that clinical evaluation. This is true even when a CGM trace looks clearly abnormal to a trained observer.

A personal trainer, health coach, physical therapist, chiropractor, or registered dietitian who observes CGM patterns that appear clinically significant should say clearly: "I am seeing a pattern in your data that I think your physician should evaluate." They should not say: "Based on your CGM data, I think you have prediabetes." The first is professional and protective of the client. The second is a diagnostic claim that falls entirely outside the scope of practice for every fitness and wellness professional.

The Bottom Line on CGM Accuracy

Modern consumer CGMs are accurate enough for the pattern-based coaching application described by the GPRM™. The 8 to 10 percent MARD typical of current devices is well within the range needed to observe trends, compare patterns across days, and identify coaching-relevant responses to exercise, meals, sleep, and stress. The four conditions that reduce accuracy are all identifiable, and knowing what they look like prevents misinterpretation of the artifacts they produce.

What CGM data is not accurate enough for is clinical diagnosis. That boundary is absolute, and it is the most important accuracy-related concept for any fitness or wellness professional to internalize before working with this data in their practice.

Everyone has the data. Be the one who can read it.

Build your complete CGM literacy from the ground up. Watch the Free BioFit Masterclass.

References

1  Dexcom Stelo Glucose Biosensor System.

2  FDA De Novo Authorization. March 2024.

3  ISO 15197:2013 — In vitro diagnostic test systems requirements for blood-glucose monitoring systems. International Organization for Standardization.

4  Rebrin K et al. Subcutaneous glucose predicts plasma glucose independent of insulin. Journal of Physiology. 1999;277(3):E561-E569.

5  Danne T et al. International consensus on use of continuous glucose monitoring. Diabetes Care. 2017;40(12):1631-1640.

6  ADA Standards of Medical Care in Diabetes. Diabetes Care. 2024