Analytical accuracy is a critical characteristic of any measurement device, including CGM systems. However, an important distinction must be made compared with traditional self-monitoring of blood glucose (SMBG). In SMBG, only the instantaneous glucose value is evaluated. In contrast, CGM provides not only a glucose measurement but also directional trend information and integration into a continuous glucose profile illustrating the temporal dynamics of glucose levels. These additional data enable effective therapeutic decision-making even when point accuracy is lower than that required for SMBG.
For both manufacturers and users, achieving the highest feasible accuracy remains a primary objective. Crucially, accuracy must be sufficient to allow CGM to serve as a replacement for SMBG. To date, manufacturers have reported accuracy in terms of the Mean Absolute Relative Difference (MARD), which quantifies the deviation between CGM readings and a reference method (e.g., a laboratory analyser or a highly accurate SMBG device). Deviations for each CGM/reference value pair are summed and divided by the number of pairs to yield the MARD.
Over recent years, CGM systems have undergone substantial advancements—both in device technology and in production processes—leading to improved measurement reproducibility. In 2016, the U.S. Food and Drug Administration (FDA) authorised the first CGM system for use in all insulin-dosing decisions, thereby permitting replacement of SMBG. As a general benchmark, a MARD of ≤10% relative to a laboratory method is considered necessary for such approval [1].
It must be noted, however, that glucose concentrations in interstitial fluid and in blood are concordant only under conditions of glucose stability. Sensor calibration also influences accuracy when calibration is permitted. Devices such as Dexcom G6/G7 and Guardian 4 are factory-calibrated but may be user-calibrated if desired. This highlights the inherent limitations of MARD as a universal metric. In contrast, the factory-calibrated FreeStyle Libre 3 cannot be user-calibrated, leaving no possibility for user influence on its MARD.
Error Grid Analysis for Assessing CGM Accuracy
The most comprehensive and practically meaningful performance criteria to date were proposed by the FDA in defining an interoperable CGM, or iCGM. An iCGM is a real-time CGM system capable of reliably and securely transmitting data to multiple digitally connected devices, ensuring interoperability. The FDA established specific accuracy requirements for iCGM systems (illustrated in Fig. SK 6).
An iCGM must meet the following criteria, with values required to fall within ±15% of the reference:
The following combinations must not occur:
Additionally, no more than 1% of iCGM readings may display a positive glucose rate of change (>1 mg/dL/min) when blood glucose simultaneously shows a negative rate (<–2 mg/dL/min), and vice versa for negative versus positive rates. Throughout the entire sensor life, clinically significant data gaps must not occur.
All comparative data between iCGM and blood glucose measurements must be collected from parallel samples and analysed in an accredited laboratory.
These FDA-defined standards establish a rigorous and transparent methodology for performance assessment, ensuring validity and comparability of accuracy claims. The criteria help to close an important gap in the evaluation of CGM accuracy. An international working group, led by the Institute for Diabetes Technology in Ulm on behalf of the International Federation of Clinical Chemistry (IFCC), is currently working to establish a precise global standard for measurement conditions and analytical accuracy [5], including the development of new measurement and evaluation procedures [6].
Fig. SK 6: Definition of Sensor Accuracy Requirements for an iCGM
This definition is well substantiated but has not yet been adopted as a universal standard [4].
A fundamental difference between traditional self-monitoring of blood glucose (SMBG) and continuous glucose monitoring (CGM) lies in the fact that measurements are taken in different physiological compartments: SMBG in capillary blood and CGM in interstitial fluid. Glucose continuously diffuses between these compartments, moving from blood to tissue or vice versa, depending on which compartment has the higher concentration.
Following a meal, for example, blood glucose rises first, and glucose then diffuses into the interstitial space. If a blood glucose measurement at that moment is compared with the simultaneous CGM reading, the blood value will be higher than the interstitial value. The values therefore differ, yet both are physiologically correct. Under conditions of glucose stability—when only minimal fluctuations occur and overall glucose concentration remains nearly constant—glucose levels in blood and interstitial fluid converge and become essentially identical.
Fig. SK 7 illustrates this relationship using two simultaneously recorded glucose curves, one from blood and one from interstitial fluid. The temporal shift between the two curves, resulting from measurement in different compartments, appears as the time lag. This time lag represents the delay until the CGM registeres the same glucose concentration previously measured by SMBG. Its magnitude depends on the rate at which glucose levels are changing.
Consequently, the time lag is not a fixed value. It varies with the speed of glucose dynamics and thus with the type of food consumed, specifically the rate of carbohydrate absorption. For example, grapes cause a rapid increase in glucose levels, whereas pasta results in a slower rise. Data for different meal compositions indicate that, following a standard meal (55% carbohydrates, 30% fat, 15% protein), the time lag is approximately 13 minutes during rising glucose levels and approximately 17 minutes during falling levels [7].
In real-world conditions, the time lag following a mixed meal depends on the specific macronutrient composition (e.g., fat content), residual gastric contents, and, importantly, individual metabolic characteristics.
Fig. SK 7: Representation of a Glucose Curve Derived from Multiple SMBG Measurements (red).
The blue curve reflects the simultaneously recorded glucose profile obtained by the CGM sensor in interstitial fluid. The delay until the same glucose value is displayed—arising from measurement in different compartments—appears as the so-called time lag. When glucose levels are stable, blood and interstitial glucose concentrations are essentially identical. During rising or falling glucose levels, however, glucose concentrations in blood and interstitial fluid differ.
Factors influencing CGM performance can be divided into technical/environmental factors and user-related factors. Technical and environmental influences include [8]:
Defective glucose sensor
User-related influences may include [8]:
Calibration errors (if calibration is possible and required):
[1] Kovatchev BP, Patek SD, Ortiz EA, Breton MD. Assessing Sensor Accuracy for Non-Adjunct Use of Continuous Glucose Monitoring. Diabetes Technology & Therapeutics. 2015;17(3):117-186.
[2] Clarke WL, Cox D, Gonder-Frederick LA et al. Evaluating clinical accuracy of systems for self-monitoring of blood glucose. Diabetes Care. 1987;10(5):622-628.
[3] Kovatchev BP, Gonder-Frederick LA, Cox DJ et al. Evaluating the accuracy of continuous glucose-monitoring sensors: continuous glucose-error grid analysis illustrated by TheraSense Freestyle Navigator data. Diabetes Care. 2004;27(8):1922-1928.
[4] U.S. Food and Drug Administration. FDA authorizes first fully interoperable continuous glucose monitoring system… Publication 27 March 2018. Accessed 18 March 2025.
[5] IFCC Working Group on CGM. http://www.ifcc.org/… Accessed 18 March 2025.
[6] Eichenlaub M, Pleus S, Rothenbühler M et al. Comparator Data Characteristics and Testing Procedures for the Clinical Performance Evaluation of Continuous Glucose Monitoring Systems. Diabetes Technol Ther. 2024;26(4):263-275.
[7] Kovatchev BP, Shields D, Breton M. Graphical and numerical evaluation of continuous glucose sensing time lag. Diabetes Technol Ther. 2009;11(3):139-143.
[8] Schlüter S, Deiss D, Gehr B et al. Glukosemessung und -kontrolle bei Patienten mit Typ-1- oder Typ-2-Diabetes. Diabetol Stoffwechsel. 2023;18(Suppl 2):S114-S135.