Abstract
Background
Continuous glucose monitoring (CGM) performance is characterized by technical metrics that are inconsistently standardized across regulatory frameworks and variably connected to clinically meaningful outcomes, complicating interpretation and cross-device comparison.
Objectives
To develop an exploratory framework translating CGM performance characteristics and derived metrics into clinically interpretable domains while distinguishing empirical evidence from expert judgment.
Methods
Candidate metrics were extracted from contemporary clinical, technical, and regulatory sources and organized into six functional domains. Targeted PubMed searches in ambulatory adults with diabetes evaluated domain-specific technical-to-clinical questions. Evidence was classified as direct device-performance, direct between-device comparative, or indirect/contextual. Five experts participated in two private voting rounds using a virtual Nominal Group Technique and prespecified RAND/UCLA agreement criteria. The study was conducted from November 2025 to January 2026.
Results
More than 30 metrics and attributes were mapped across analytical accuracy and performance, glycemic exposure, glycemic variability, clinical safety, adherence and device use, and device reliability. Direct evidence was most clearly identified for device-level analytical performance and between-device differences in analytical accuracy and glycemic exposure; evidence in other domains was predominantly indirect, contextual, or device-specific. Twenty-four of 25 statements achieved consensus after round 1 and all 25 after round 2. The panel considered dynamic accuracy, alert functionality when available, and monitoring continuity potentially clinically meaningful, but these ratings did not establish category-wide superiority of FDA-designated integrated CGM systems.
Conclusion
The framework provides an exploratory structure for interpreting CGM technical performance and derived metrics. Regulatory designation and individual device characteristics should be considered separately, and purpose-designed comparative studies are required to determine whether specific attributes produce reproducible clinical benefits.
Keywords: continuous glucose monitoring, iCGM, eCGM, CGM performance metrics, regulatory standards, nominal group technique, analytical accuracy, diabetes technology
Plain Language Summary
Continuous glucose monitors (CGMs) help people with diabetes track glucose levels and make treatment decisions. However, CGM systems differ in accuracy, alarms, connectivity, wear characteristics, and data availability. The technical measures used to evaluate these features can be difficult to interpret, and regulatory frameworks do not assess every aspect of performance in the same way. This study developed an exploratory framework to translate CGM technical characteristics and glucose-derived measures into clinically understandable concepts. More than 30 metrics and attributes were organized into six areas: analytical accuracy and performance, glycemic exposure, glycemic variability, clinical safety, adherence and device use, and device reliability and data continuity. Targeted literature searches showed that direct comparisons between individual CGM systems were limited and were most clearly available for analytical accuracy and glycemic exposure. Evidence concerning safety, adherence, and reliability was mainly indirect, contextual, or device-specific. Five experts evaluated 25 statements in two voting rounds. They considered performance during rapid glucose changes, low-glucose accuracy, alert performance when alerts are available, and monitoring continuity potentially important for treatment decisions and patient safety. However, expert agreement does not prove that one CGM category is clinically superior to another. The FDA integrated CGM (iCGM) designation indicates that a device has met specified regulatory requirements, but it does not by itself guarantee better clinical outcomes. Clinicians should also consider individual device characteristics and how performance data were generated. Further direct comparisons are needed to determine which technical features produce meaningful clinical benefits.
Introduction
Continuous glucose monitoring (CGM) technologies have reshaped contemporary diabetes management by enabling real-time assessment of glucose dynamics, improving the detection of glycemic excursions, and supporting more precise therapeutic decision-making. CGM use has been associated with improved glycemic control, increased time in range, reduced risk of hypoglycemia, and enhanced quality of life across diverse patient populations. The performance of a CGM system depends on device-level characteristics—including measurement agreement, sensor stability, rate-of-change performance, calibration, signal processing, and data continuity—that are grounded in sensor engineering and analytical evaluation. These characteristics should be distinguished from downstream CGM-derived metrics, such as time in range, mean glucose, GMI, and coefficient of variation, which summarize the measured glucose profile and reflect the combined influence of physiology, treatment, behavior, and the measurement process. Translating device-performance characteristics into their potential effects on these derived metrics and on clinical decisions remains challenging because both the metrics and the conditions under which they are generated require specialized methodological interpretation.
Regulatory frameworks have attempted to translate this technical complexity into defined performance expectations. In the United States, FDA-designated integrated continuous glucose monitoring (iCGM) systems are CGM systems that have demonstrated compliance with specified regulatory performance and interoperability criteria. In this manuscript, CGM systems without this designation include both real-time and intermittently scanned systems; the absence of designation should not be interpreted as evidence that an individual device lacks any particular technical capability. 1 Under 21 CFR 862.1355, the iCGM special controls require robust clinical data demonstrating device accuracy in the intended-use population through comparison with parallel blood glucose measurements obtained using an FDA-accepted laboratory-based method that is precise, accurate, and traceable to a higher order. The analytical requirements include glucose-range-specific agreement proportions whose lower one-sided 95% confidence bounds must exceed prespecified thresholds, restrictions on clinically discordant rate-of-change results, and demonstration of performance throughout the labeled sensor-wear period. The lower confidence bound reflects the statistical uncertainty surrounding the observed agreement proportion; it should not be interpreted as a measure of sensor “stability.” Neither MARD nor error-grid classification constitutes an FDA iCGM acceptance criterion, although both may be reported as complementary descriptive measures in CGM performance studies.
Importantly, the iCGM designation should be understood as a regulatory qualification rather than as an independent device characteristic or a guarantee of superior clinical outcomes. In addition to analytical requirements, the special controls address clinically significant gaps in sensor data, secure and reliable data transmission, interfering substances, usability, manufacturing controls, and performance labeling. They do not require uninterrupted data availability; rather, the submitted data must demonstrate that gaps do not prevent connected devices from achieving their intended use. Likewise, the special controls do not require every iCGM system to provide glucose alerts or establish numerical acceptance thresholds for alert performance. When alerts are applicable, labeling must describe the available settings and the observed true, false, missed, and correct alert rates. Clinical performance therefore depends on the characteristics of the individual device, including sensor chemistry, calibration approach, sampling frequency, algorithmic filtering, lag compensation, alert availability and accuracy, connectivity, interoperability, user interface, and wear characteristics. These features may vary among devices within the same regulatory category and may independently influence CGM-derived metrics and clinical outcomes. Accordingly, this framework considers regulatory status and individual device characteristics as related but distinct sources of information.
A complementary perspective is provided by the proposed guideline from the International Federation of Clinical Chemistry and Laboratory Medicine Working Group on CGM (IFCC WG-CGM). 2 This guideline emphasizes that CGM performance cannot be interpreted independently of the conditions under which the data were generated. In addition to proposing performance metrics and acceptance criteria, it addresses study design and procedures, comparator measurement characteristics, measurement pairing and synchronization, sampling across clinically relevant static and dynamic glucose ranges, and performance characterization throughout sensor wear. From a clinical perspective, these methodological characteristics are essential because apparently similar accuracy estimates may not be comparable when derived using different comparator methods, glucose distributions, rates of change, sampling procedures, or temporal-alignment methods. The translation of CGM performance data should therefore consider both what was measured and how the supporting study was conducted.
In Europe, the absence of CGM-specific criteria within the Conformité Européenne (CE) regulatory framework has motivated the proposal of a new eCGM compliance status, intended to establish minimum expectations for clinical evidence, safety verification, and analytical performance before market authorization. This initiative emerged partly in response to recent reports of technically CE-marked devices whose real-world accuracy in critical glucose ranges proved insufficient, prompting concerns regarding patient safety and unintended therapeutic risk. The proposed eCGM framework emphasizes the need for harmonized criteria linking technical performance metrics to clinically meaningful reliability. Still, its structure remains under development and has yet to converge on a definitive set of priority metrics for clinical decision-making. 3
In Latin America, the regulatory landscape is even more heterogeneous. The rapid expansion of the CGM market and the availability of devices with widely varying accuracy have highlighted the lack of standardized performance expectations across the region. A recent regional consensus underscored the need for harmonized evaluation metrics aligned with international best practices, emphasizing analytical accuracy, performance consistency, safety evaluation, and post-market vigilance. 4 Nevertheless, the consensus also revealed a central challenge: most recommended indicators, such as MARD, accuracy stratified by glycemic range or glucose-change rate, hypoglycemia performance, device reliability, or sensor-to-sensor variability, are inherently technical constructs that lack immediate clinical interpretability. As a result, clinicians and decision-makers face substantial difficulty translating engineering-based performance metrics into judgments about safety, appropriateness, and therapeutic confidence in routine practice.
Addressing this gap requires an interpretive framework that links upstream device-performance characteristics and performance-study conditions with downstream CGM-derived summaries and their clinical consequences. Metrics and attributes identified in the foundational documents were organized into six domains according to the construct directly assessed and its principal position in this technical-to-clinical pathway. The framework comprises three interpretive levels: upstream device-performance domains—analytical accuracy and performance, and device reliability; downstream CGM-derived clinical summaries—glycemic exposure and glycemic variability; and clinical or implementation domains—clinical safety, and adherence and device use. Glycemic exposure and variability were retained as separate domains because exposure metrics characterize average glucose and the magnitude or duration of glucose values within defined ranges, whereas variability metrics characterize dispersion, fluctuation, and temporal instability. However, neither is considered an intrinsic device-performance domain. Rather, their inclusion allows examination of how measurement bias, filtering, sampling, lag, and missing data may propagate into clinically interpreted CGM outcomes. Because several metrics have implications across more than one level, primary domain assignment was used for organization without implying that the domains are mutually exclusive.
The present framework and evidence search focus primarily on ambulatory adults. Although the six domains may also be conceptually relevant in other populations, their relative importance and interpretation may differ. In pediatric populations, rapid glucose changes, smaller insulin doses, caregiver involvement, remote monitoring, school-related use, and alarm burden may increase the importance of dynamic accuracy, connectivity, and usability. Pregnancy requires assessment within tighter glycemic targets and may involve distinct physiological glucose dynamics. In older adults, hypoglycemia vulnerability, cognitive or functional limitations, multimorbidity, and caregiver support may place greater emphasis on alarms, simplicity of use, and monitoring continuity. Hospitalized patients represent a separate context in which acute illness, perfusion abnormalities, interfering treatments, rapidly changing glucose concentrations, and clinical workflow integration may affect device performance and interpretation. Therefore, extrapolation beyond ambulatory adults requires population- and setting-specific validation.
Accordingly, this study aims to develop an exploratory expert-consensus framework linking CGM device-performance characteristics and evaluation methods to downstream glycemic metrics and clinical interpretation. The framework draws on metrics and concepts represented in the FDA iCGM pathway, the proposed European eCGM framework, and the emerging Latin American model, and is further contextualized using the IFCC WG-CGM recommendations for standardized performance assessment. Rather than testing the superiority of a regulatory category, the study seeks to identify which technical attributes, study conditions, and device-specific characteristics may influence the interpretation, safety, and clinical use of CGM data, while distinguishing empirical comparative evidence from expert-derived clinical judgment.
Methods
Conceptual Basis and Development of the Technical-To-Clinical Domain Framework
We conducted an exploratory extraction of candidate CGM metrics and device attributes described in three contemporary documents outlining clinical, technical, and regulatory approaches to CGM evaluation.1,3,4 The purpose was to identify the constructs considered relevant across these frameworks rather than to assume that all extracted variables represented equivalent measures of intrinsic device performance. During manuscript revision, the proposed guideline from the International Federation of Clinical Chemistry and Laboratory Medicine Working Group on CGM was additionally incorporated as a contemporary technical standard for interpreting performance-study design, comparator methods, testing conditions, and dynamic glucose ranges. Its incorporation refined the technical description and interpretation of the framework but did not retrospectively alter the original consensus ratings.
Domain development followed an iterative, construct-based mapping process conducted by the academic committee. First, extracted metrics and attributes were reviewed to identify the construct directly measured, such as sensor–comparator agreement, glucose exposure, temporal variability, data continuity, or user interaction. Second, each construct was positioned within a technical-to-clinical pathway extending from device measurement and data generation to derived glycemic summaries and potential clinical consequences. Third, conceptually related metrics were grouped into a primary domain, while clinically relevant relationships with other domains were documented separately. The process was qualitative and consensus-based; no statistical clustering or independent external validation of the domain architecture was performed.
The resulting six domains were organized into three interpretive levels. Upstream device-performance domains comprised analytical accuracy and performance, and device reliability. Downstream CGM-derived clinical summaries comprised glycemic exposure and glycemic variability. Clinical and implementation domains comprised clinical safety, and adherence and device use. This organization distinguishes intrinsic device-performance characteristics from metrics that summarize the glucose profile and from outcomes influenced by clinical or behavioral context.
Glycemic exposure and glycemic variability were retained as separate downstream domains because they describe related but non-equivalent characteristics of the measured glucose profile. Exposure metrics—including mean glucose, time in range, time below range, time above range, and GMI—summarize average glucose or the magnitude and duration of glucose values within defined ranges. Variability metrics—including standard deviation, coefficient of variation, and excursion- or risk-based indices—describe dispersion, fluctuation, and temporal instability. Neither domain was considered an independent regulatory-performance requirement. Their inclusion was intended to examine how measurement bias, filtering, lag, sampling frequency, and missing data may propagate into clinically interpreted CGM summaries.
For each metric or attribute, a standardized structure documented: (i) its name; (ii) primary domain; (iii) technical or clinical definition; (iv) position within the technical-to-clinical pathway; (v) potential cross-domain implications; and (vi) relevant scientific commentary. Primary domain assignment provided an organizational structure and did not imply that domains were mutually exclusive.
In this manuscript, an integrated continuous glucose monitoring (iCGM) system refers specifically to a CGM system that had received FDA iCGM designation under the applicable US regulatory pathway at the time of the review. “CGM systems without FDA iCGM designation” includes both real-time and intermittently scanned CGM systems that had not received this designation. This classification is regulatory, jurisdiction-specific, and time-dependent; it does not imply technological homogeneity within either group or establish comparative clinical superiority. Regulatory designation and individual device characteristics were therefore considered separately whenever permitted by the available evidence.
Domain Assignment and Cross-Domain Relationships
Metrics were assigned to a primary domain according to the technical or clinical construct they directly measured. Primary assignment was used to provide a reproducible organizational structure and did not imply that metrics were exclusive to a single domain. When a metric had clinically relevant implications for additional domains, these were recorded as cross-domain relationships. For example, MARD was assigned primarily to analytical accuracy because it summarizes sensor–comparator disagreement, while its potential consequences for hypoglycemia detection and insulin-dosing decisions were cross-referenced under clinical safety. Similarly, data gaps and sensor interruptions were assigned primarily to reliability and continuity but were also linked to adherence and safety.
Domain assignments were developed and refined by the academic committee and were subsequently considered during the expert-consensus process. No independent external validation of the domain architecture was performed; accordingly, the framework is presented as exploratory and subject to future validation and refinement.
Purpose of the Targeted Domain-Mapped Evidence Search
The targeted evidence search was not conducted to define the metrics or to provide an exhaustive review of all CGM performance studies. Its purpose was to identify empirical evidence addressing whether, and under what conditions, device-performance characteristics could influence downstream CGM-derived summaries, safety, adherence, reliability, or other clinically interpretable outcomes. Each domain was therefore mapped to a predefined technical-to-clinical question, such as whether measurement disagreement affects hypoglycemia assessment, whether device-specific bias influences exposure metrics, or whether interruptions in monitoring alter clinical use. The search was also intended to identify where the proposed pathways were supported by direct device-level evidence and where they remained dependent predominantly on indirect evidence or expert interpretation.
Development of the Search Strategy
A modular, block-based strategy was developed to ensure consistency across domains. The following structure was applied:
(Population) AND (CGM technology) AND (Domain-specific terms) [AND (Outcome terms, when appropriate)] NOT (Pediatric terms).
The population was restricted to ambulatory adults (≥18 years) with diabetes. Studies that exclusively involved hospitalized patients, individuals with chronic kidney disease, or pregnant populations were excluded to maintain a consistent physiological and clinical context across the metric domains. The technology block included rtCGM, isCGM/FGM (intermittently scanned Continuous Glucose Monitoring/Flash Glucose Monitoring), and iCGM systems. Domain-specific blocks reflected the terminology associated with each metric category, and outcome terms were included only when clinically relevant to the domain. All phrases were validated against the PubMed index to ensure compatibility. The complete search strategies for all domains are available in Supplementary Material 1.
Eligibility Criteria and Study Selection
Studies were eligible when they: (i) included ambulatory adults aged 18 years or older with diabetes; (ii) evaluated an identifiable CGM system or a clearly defined technical or functional CGM feature; (iii) reported at least one metric relevant to a predefined domain; and (iv) provided an analytical, clinical, safety, adherence, reliability, or health-economic outcome capable of informing the corresponding technical–clinical question. Eligible designs included randomized and nonrandomized comparative studies, analytical performance studies, prospective or retrospective observational studies, real-world studies, surveys, and health-economic analyses when relevant to the domain.
Studies were excluded when they were conducted exclusively in pediatric populations, pregnancy, hospitalized patients, or individuals with chronic kidney disease; combined eligible and ineligible populations without separately interpretable data for ambulatory adults; did not identify the CGM system or relevant device feature; or did not report outcomes relevant to the predefined domain question. English-language publications were retained. Reviews, editorials, commentaries, and clinical, regulatory, or technical documents could be used to provide conceptual context but were not counted as eligible comparative clinical evidence.
Regulatory designation was not used as an inclusion or exclusion criterion. Studies comparing two or more identifiable CGM systems were eligible regardless of whether the comparison involved two FDA-designated iCGM systems, two systems without FDA iCGM designation, or systems with different regulatory designations. Regulatory status was recorded as a descriptive device characteristic and was not assumed to explain observed between-device differences. Studies directly assessing an identifiable CGM system against a specified comparator method were also eligible as device-performance evidence even when no second CGM system was evaluated.
Records were assessed in two stages: title and abstract screening followed by full-text evaluation. Reasons for exclusion at the full-text stage were recorded by domain. Because the eligibility requirements were tailored to each domain-specific technical–clinical question, a study could be relevant for contextual discussion without meeting the criteria for inclusion in the corresponding domain-level evidence synthesis.
Information Sources and Search Scope
Searches were conducted in PubMed/MEDLINE without date restrictions. The selection of PubMed was deliberate and reflected the clinically focused purpose of the evidence assessment. The study was not designed as a systematic or exhaustive review of engineering, regulatory, and device-performance literature. Rather, the initial inventory of CGM metrics was derived separately from contemporary clinical, technical, and regulatory documents, after which the PubMed search was used to identify clinically oriented evidence linking those metrics or device characteristics to interpretable outcomes in ambulatory adults with diabetes.
Other multidisciplinary and technical databases were not included because the literature component was intended as a targeted, domain-mapped clinical assessment rather than a comprehensive systematic review. Consequently, evidence counts describe the availability of eligible studies within the predefined PubMed search and should not be interpreted as exhaustive estimates of the global evidence base. The complete domain-specific search strategies and search dates are provided in Supplementary Material 1.
The IFCC WG-CGM proposed guideline was used as a technical interpretive standard rather than counted as comparative clinical evidence. In particular, it informed the description of comparator-method characteristics, measurement synchronization and pairing, glucose-range and rate-of-change coverage, sampling procedures, wear-period assessment, and other study conditions that may influence reported CGM performance. Consequently, the interpretation of an accuracy estimate considered not only the reported numerical metric but also, when available, the comparator and the procedures through which the estimate was generated.
Domain-specific Search Objectives
Each domain had a predefined conceptual objective that guided the selection of search terms. These included exploring links among accuracy and safety, exposure metrics and long-term outcomes, variability and hypoglycemia risk, and device adherence/reliability and clinical benefit. This ensured terminological alignment between the metric classification and the search queries.
Evidence was classified according to the relationship between the study design and the corresponding domain-specific question rather than according to the regulatory designation of the evaluated devices.
Direct device-performance evidence comprised studies that directly assessed an identifiable CGM system against a specified comparator method under defined clinical or experimental conditions. These studies could directly inform the analytical performance of the evaluated device but did not, in the absence of another CGM system, establish between-device differences.
Direct between-device comparative evidence comprised studies evaluating two or more identifiable CGM systems under sufficiently comparable conditions to support device-level comparison. This category included concurrent within-participant studies, randomized crossover studies, and appropriately designed parallel-group comparisons, regardless of the regulatory designation of the systems involved.
Indirect or contextual evidence comprised studies linking CGM use, device features, derived glycemic metrics, adherence, or monitoring continuity to clinical outcomes without a directly interpretable comparison of the relevant device characteristic. Regulatory, technical, consensus, and methodological documents were used to provide conceptual context but were not treated as comparative clinical evidence.
A study could contribute to more than one domain or evidence category when it directly addressed more than one predefined question. FDA iCGM designation was recorded separately as a regulatory characteristic and was not used as a proxy for device performance, as an indicator of evidentiary directness, or as the presumed causal explanation for an observed difference.
The design required for a clinically interpretable comparison differed across domains. Concurrent within-participant wear was considered particularly informative for analytical performance and CGM-derived exposure or variability metrics because the systems evaluated the same underlying glucose profile. However, simultaneous wear does not permit two systems to independently guide therapy in the same participant. For adherence, user experience, and therapy-dependent clinical outcomes, randomized crossover studies with separate active-use periods or pragmatic parallel-group comparisons were considered more appropriate. Concurrent blinded wear could still inform device-level retention, signal availability, data gaps, and technical interruptions. Evidence directness was therefore determined relative to the domain-specific question and not by requiring a single head-to-head design across all domains.
Because the literature assessment was targeted rather than systematic, no formal certainty-of-evidence grading or risk-of-bias framework was applied. Evidence was therefore categorized according to its directness and availability rather than labeled as high, moderate, or low certainty.
Nominal Group Technique Process
A structured Nominal Group Technique (NGT) was employed to achieve expert consensus on the clinical interpretation of continuous glucose monitoring (CGM) performance metrics, incorporating analytical elements derived from the RAND/UCLA Appropriateness Method to assess agreement and disagreement. Five experts were purposively selected based on a combined profile of clinical and academic expertise in diabetes technology and CGM interpretation. Selection considerations included specialist experience in diabetes care, sustained clinical management of people using CGM, routine interpretation of CGM data for therapeutic decision-making, and relevant experience in research, publication, teaching, technology assessment, or consensus development. Publication experience was considered an indicator of expertise but was not used as an isolated or mandatory numerical threshold. 5
The panel size was selected to support structured participation and in-depth deliberation within the Nominal Group Technique. It was not intended to provide statistical representation of the broader professional community. Because the panel was geographically concentrated, its judgments may have been influenced by the device availability, access conditions, reimbursement structures, clinical workflows, and regulatory priorities of the region represented.
The process followed established methodological frameworks for NGT, including adaptations for virtual environments, 6 and integrated principles commonly used in consensus studies—particularly the use of a 9-point Likert scale and predefined quantitative criteria for agreement. 7
Initial statements were generated by mapping the extracted metrics and domain-specific technical–clinical questions to potential clinical interpretations. The purpose of the statements was to assess the clinical plausibility and relevance of these interpretations rather than to establish the superiority of a regulatory category. The evidence framework accompanying statement development distinguished direct comparative evidence from indirect or conceptual support. Potential confirmation bias was addressed procedurally through independent private voting, prespecified agreement and disagreement criteria, the absence of a requirement to achieve consensus, and the opportunity to rate statements as uncertain or inappropriate.
The consensus process consisted of two structured virtual sessions and two rounds of private voting, conducted as follows:
Preparatory Session and First Voting Round
A preparatory virtual meeting was convened before the first voting round. During this session, the study objectives, consensus methodology, rating scale, and predefined criteria for agreement and disagreement were explained, and procedural questions were addressed. Following the meeting, panelists independently evaluated each statement in a first round of private voting to minimize social influence. Ratings reflected the perceived appropriateness and clinical relevance of each statement, using a 1–9 Likert scale.
Feedback, Structured Discussion, and Statement Refinement
Aggregated results from the first voting round, including median scores and dispersion measures, were presented to the panel during a second virtual meeting. In accordance with principles of the Nominal Group Technique, this session incorporated structured turn-taking, opportunities for clarification, and moderator-facilitated discussion. Based on this deliberative process, statements were refined to improve clarity, define scope more precisely, and enhance clinical applicability, without forcing consensus or requiring convergence of individual opinions.
Final Voting Round and Consensus Determination
Panelists subsequently independently re-rated the revised statements in a second private voting round. Final consensus was assessed using established RAND/UCLA analytical criteria. For each statement, the median rating and the presence of disagreement were calculated. Disagreement was defined by comparing the interpercentile range (IPR) with the interpercentile range adjusted for symmetry (IPRAS); disagreement was considered present when the IPR exceeded the IPRAS. For the final consensus, statements with a median score ≥7 and no disagreement were classified as demonstrating high agreement and high clinical relevance.
This hybrid approach combined the interactive strengths of the Nominal Group Technique with the analytical rigor of the RAND/UCLA method, ensuring a transparent and reproducible assessment of expert agreement.
Results
Study Period
The study was conducted from November 2025 to January 2026, encompassing framework development, the targeted literature searches, two expert-consensus voting rounds, structured discussion, and final analysis.
Construction of a Unified Framework for CGM Performance Metrics
The metric extraction process yielded more than 30 distinct CGM evaluation metrics referenced in contemporary clinical and regulatory documents. These metrics encompassed a broad range of device performance characteristics, from analytical accuracy to long-term reliability.
Each metric was documented according to the predefined structure established in the Methods section, enabling consistent comparison across sources. The complete set of extracted metrics is summarized in Table 1.
Table 1.
Technical-To-Clinical Classification of CGM Evaluation Metrics and Attributes. The Table Distinguishes Upstream Device-Performance Attributes From Downstream CGM-Derived Summaries and Clinical or Implementation Outcomes. Primary-Domain Assignment Identifies the Construct Measured Directly; Regulatory Provenance and Secondary Implications are Reported Separately
| Metric or attribute | Interpretive level | Primary functional domain | Metric type or regulatory/technical source | Brief technical definition | Cross-domain clinical implications | Evidence classification | Interpretive caution |
|---|---|---|---|---|---|---|---|
| MARD | Upstream device performance | Analytical Accuracy and Performance | Aggregate sensor–comparator error metric | Mean absolute relative difference across paired CGM and comparator values. | May affect confidence in dosing, exposure metrics, and event detection. | DDP | Distribution-, comparator-, pairing-, and study-design dependent; not an FDA iCGM acceptance criterion. |
| Absolute-difference summaries (MAD/MedAD) | Upstream device performance | Analytical Accuracy and Performance | Sensor–comparator error metrics | Mean or median absolute difference, usually emphasized at lower glucose concentrations. | Relevant to low-glucose interpretation and treatment decisions. | DDP | Units, glucose range, comparator, and pairing window must be reported. |
| Signed bias | Upstream device performance | Analytical Accuracy and Performance | Sensor–comparator agreement metric | Mean or median signed CGM-minus-comparator difference, expressed absolutely or relatively. | Systematic bias can shift TIR/TBR/TAR, mean glucose, and GMI. | DDP + DBC | Direction and magnitude are device- and study-specific. |
| Agreement proportions (e.g., 15/15 and 20/20) | Upstream device performance | Analytical Accuracy and Performance | Range-based sensor–comparator agreement metric | Percentage of paired values within prespecified absolute or relative limits. | Complements aggregate error and helps characterize clinically relevant disagreement. | DDP | Threshold, denominator, confidence interval, and glucose distribution should be stated. |
| FDA iCGM range-specific agreement criteria (A–K) | Upstream device performance | Analytical Accuracy and Performance | FDA iCGM special-control performance criteria | Range- and condition-specific agreement, extreme-discrepancy, and rate-of-change criteria, including specified lower one-sided 95% confidence bounds. | Supports regulatory assessment of accuracy and interoperability; may inform confidence in connected use. | RT | These are analytical-performance criteria; regulatory designation alone does not establish superior clinical outcomes. |
| Large or extreme sensor–comparator discrepancies | Upstream device performance | Analytical Accuracy and Performance | Tail-error distribution metric; included in FDA iCGM assessment | Frequency of paired values with large absolute or relative disagreement. | Rare errors may have disproportionate consequences for insulin dosing or missed events. | DDP + RT | Report direction, glucose range, and confidence bounds. |
| Clinical error-grid classification | Upstream device performance | Analytical Accuracy and Performance | Clinical-risk classification (e.g., Clarke or Consensus Error Grid) | Classifies paired measurements by the potential clinical consequence of disagreement. | Provides a risk-oriented complement to numerical accuracy summaries. | DDP | It estimates potential decision risk; it is not an FDA iCGM acceptance criterion and does not measure observed outcomes. |
| Rate-of-change/trend accuracy | Upstream device performance | Analytical Accuracy and Performance | Dynamic sensor-performance metric; included in FDA iCGM assessment | Agreement in glucose direction and rate of change between CGM and comparator data. | Relevant to trend-arrow interpretation, impending hypoglycemia, and dose timing. | DDP + RT | Requires synchronized measurements and prespecified dynamic ranges. |
| Physiological and algorithmic lag | Upstream device performance | Analytical Accuracy and Performance | Temporal-performance attribute | Delay attributable to interstitial physiology, sensing, filtering, and signal processing. | May delay recognition of rapid rises or falls and alter apparent trend accuracy. | DDP + RT | Estimation method and temporal alignment materially affect results. |
| Accuracy by glucose range | Upstream device performance | Analytical Accuracy and Performance | Stratified analytical-performance metric | Sensor–comparator agreement evaluated separately in low, target, and high glucose ranges. | Low-range performance is particularly relevant to hypoglycemia detection and response. | DDP | Adequate observations and confidence intervals within each range are essential. |
| Accuracy across labeled wear | Upstream device performance | Analytical Accuracy and Performance | Wear-period analytical-performance metric | Accuracy characterized by sensor day or predefined phases of the wear period. | Identifies early- or late-wear degradation that may affect clinical interpretation. | DDP + RT | Distinct from the proportion of sensors that survive to end of wear. |
| Accuracy by insertion site | Upstream device performance | Analytical Accuracy and Performance | Site-specific analytical-performance metric | Accuracy and precision assessed at each labeled anatomical site. | Site-dependent behavior may affect real-world interpretation and user choice. | DDP | Conclusions apply only to studied and labeled sites. |
| Sensor-to-sensor precision | Upstream device performance | Analytical Accuracy and Performance | Reproducibility metric | Agreement between simultaneously worn sensors or across sensors/sites under comparable conditions. | Poor reproducibility may reduce confidence in individual readings and derived patterns. | DDP + DBC | Simultaneous wear and site effects must be considered. |
| Calibration-related performance | Upstream device performance | Analytical Accuracy and Performance | Device-design and performance attribute | Effect of factory or user calibration, calibration frequency, timing, and errors on sensor output. | Can influence accuracy, burden, and risk of user-induced error. | DDP + RT | Calibration design should not be conflated with regulatory category. |
| Interference/analytical robustness | Upstream device performance | Analytical Accuracy and Performance | Interference testing; FDA iCGM special-control area | Performance in the presence of specified endogenous or exogenous interferents and relevant conditions. | Undetected interference may cause clinically consequential erroneous readings. | DDP + RT | Interferents, concentrations, and direction of bias are device-specific. |
| Comparator quality and traceability | Upstream device performance | Analytical Accuracy and Performance | Performance-study design attribute; IFCC-aligned | Precision, accuracy, traceability, and operating characteristics of the comparator method. | Determines how confidently apparent sensor error can be attributed to the CGM system. | RT | Analytical accuracy is agreement with a specified comparator, not error-free physiological truth. |
| Measurement pairing and synchronization | Upstream device performance | Analytical Accuracy and Performance | Performance-study design attribute; IFCC-aligned | Rules for temporal matching, interpolation, and inclusion of CGM–comparator pairs. | Misalignment can distort lag, dynamic accuracy, and error estimates. | RT | Procedures must be prespecified and reported. |
| Static and dynamic glucose-range coverage | Upstream device performance | Analytical Accuracy and Performance | Performance-study design attribute; IFCC-aligned | Distribution of reference observations across clinically relevant concentrations and rates of change. | Determines whether reported performance applies to hypoglycemia, hyperglycemia, and rapidly changing glucose. | RT | Overall summaries are not comparable when sampling conditions differ materially. |
| TIR | Downstream derived metric | Glycemic Exposure | CGM-derived exposure metric | Percentage of valid CGM time within a prespecified target range, commonly 70–180 mg/dL. | Associated with overall glycemic control and affected by sensor bias and missing data. | DBC | This is an outcome/summary metric, not an intrinsic sensor-performance metric. |
| TBR | Downstream derived metric | Glycemic Exposure | CGM-derived exposure metric | Percentage of valid CGM time below prespecified thresholds, commonly <70 and <54 mg/dL. | Informs hypoglycemia burden but depends on low-range accuracy and data continuity. | DBC + IC | Device-derived differences may reflect analytical bias rather than true exposure. |
| TAR | Downstream derived metric | Glycemic Exposure | CGM-derived exposure metric | Percentage of valid CGM time above prespecified thresholds, commonly >180 and >250 mg/dL. | Informs hyperglycemia burden and may influence treatment intensification. | DBC | Device bias, sampling, and missing-data handling should be considered. |
| Mean/median sensor glucose | Downstream derived metric | Glycemic Exposure | CGM-derived distribution summary | Average or median of valid CGM readings during the analysis period. | Drives GMI and can shift apparent glycemic exposure across devices. | DBC | Sensitive to systematic bias, data sufficiency, and sampling procedures. |
| GMI | Downstream derived metric | Glycemic Exposure | CGM-derived estimated glycemic marker | Glucose management indicator calculated from mean sensor glucose. | Supports clinical communication but may be discordant with laboratory HbA1c. | DBC | Device-specific bias may change GMI without a change in underlying glycemia. |
| Standard deviation (SD) | Downstream derived metric | Glycemic Variability | Dispersion metric | Absolute dispersion of CGM glucose values around their mean. | May help characterize glucose fluctuation and risk beyond mean exposure. | DBC + IC | Influenced by mean glucose, sensor error, sampling, and missingness. |
| Coefficient of variation (%CV) | Downstream derived metric | Glycemic Variability | Normalized dispersion metric | SD divided by mean sensor glucose, expressed as a percentage. | Used to describe relative glycemic variability and hypoglycemia susceptibility. | DBC + IC | A downstream outcome metric, not a direct measure of device performance. |
| MAGE | Downstream derived metric | Glycemic Variability | Excursion-based variability metric | Mean amplitude of qualifying glycemic excursions according to a specified algorithm. | Describes large excursions that may be obscured by aggregate exposure measures. | IC | Results depend on implementation, sampling frequency, smoothing, and sensor error. |
| GVP | Downstream derived metric | Glycemic Variability | Path-length variability metric | Glycemic variability percentage derived from the length of the glucose trace relative to elapsed time. | Captures temporal fluctuation and may complement SD and CV. | IC | Algorithm and preprocessing should be reported. |
| LBGI/HBGI | Downstream derived metric | Glycemic Variability | Transformed glucose-risk indices | Low and high blood glucose risk indices calculated after nonlinear transformation of glucose values. | May summarize asymmetric hypo- and hyperglycemic risk. | IC | Sensitive to low/high-range measurement error and not a sensor-performance criterion. |
| ADRR | Downstream derived metric | Glycemic Variability | Composite daily risk index | Average daily risk range derived from daily low- and high-risk glucose components. | May support risk stratification across days. | IC | Requires transparent algorithm, sufficient days, and cautious cross-device comparison. |
| GRI | Downstream derived metric | Glycemic Variability | Composite CGM risk index | Weighted summary of time spent in clinically relevant hypo- and hyperglycemic ranges. | Integrates exposure to clinically important extremes. | IC | Closely related to exposure metrics and should not be interpreted as independent device performance. |
| Alert performance | Clinical/implementation outcome | Clinical Safety | Device-feature performance and labeling metric | True, false, missed, and correct alert rates, sensitivity, timing, and threshold-specific performance where alerts are available. | May affect event detection, alarm burden, trust, and response behavior. | RT + IC | iCGM designation does not require every device to provide alerts or impose a universal alert-accuracy threshold. |
| Low-glucose event detection | Clinical/implementation outcome | Clinical Safety | Event-level detection metric | Detection of independently defined or confirmed hypoglycemic events within a prespecified window. | Directly relevant to prevention and treatment of hypoglycemia. | IC | Event definition, confirmation method, alert availability, and timing must be specified. |
| Treatment-decision discrepancy/confirmatory testing | Clinical/implementation outcome | Clinical Safety | Clinical-decision process metric | Frequency with which CGM information conflicts with comparator data, symptoms, or a prespecified treatment decision and prompts confirmation. | Links analytical disagreement to potential therapeutic consequences and burden. | No direct evidence identified; proposed comparative outcome | Requires prospective, decision-focused study design; regulatory status alone cannot predict it. |
| Device-related adverse events and insertion-site tolerability | Clinical/implementation outcome | Clinical Safety | Safety outcome | Frequency and severity of skin reactions, insertion events, device-related harms, and associated discontinuation. | May affect safety, acceptability, and sustained use. | IC | Device-specific evidence may also apply. Standardized definitions and exposure denominators are needed. |
| Sensor wear/active-use time | Clinical/implementation outcome | Adherence and Device Use | Use/adherence metric | Time the sensor is worn and actively used relative to the expected observation period. | Determines the amount of clinically interpretable data and can influence glycemic outcomes. | IC | Use may be confounded by device access, education, treatment modality, and user preference. |
| Scanning or data-interaction frequency | Clinical/implementation outcome | Adherence and Device Use | User-interaction metric | Frequency of scans, display checks, uploads, or other required interactions. | May influence exposure to information, engagement, and timely response. | IC | Interaction requirements differ by device modality and interface. |
| Persistence/discontinuation | Clinical/implementation outcome | Adherence and Device Use | Longitudinal use outcome | Continuation, interruption, or discontinuation of CGM over a prespecified period, with reasons. | Reflects acceptability, access, burden, adverse events, and perceived benefit. | IC | Category attribution is inappropriate without adjustment for patient, care-system, and device-specific factors. |
| Calibration and user-task burden | Clinical/implementation outcome | Adherence and Device Use | Workflow/usability attribute | Frequency and complexity of calibration, insertion, replacement, charging, scanning, and troubleshooting tasks. | Can influence sustained use, user error, and treatment burden. | RT + IC | Burden is device-specific and should be measured directly. |
| Alarm burden | Clinical/implementation outcome | Adherence and Device Use | User-experience metric | Frequency, timing, actionability, and perceived burden of alerts and alarms. | May affect sleep, alarm fatigue, trust, response behavior, and discontinuation. | IC | Depends on alert availability, settings, glucose exposure, and user behavior. |
| Usability/human factors | Clinical/implementation outcome | Adherence and Device Use | Human-factors and user-experience construct | Ease, learnability, errors, comprehension, and successful performance of intended user tasks. | May affect safe operation, adherence, equity, and real-world effectiveness. | RT + IC | Requires population- and task-specific assessment. |
| Data availability | Upstream device performance | Device Reliability and Data Continuity | Continuity metric; FDA iCGM special-control area | Percentage of the intended observation period with valid, retrievable CGM data. | Affects metric completeness, event detection, connected-device function, and clinical confidence. | RT + IC | High overall availability can still coexist with clinically important gaps. |
| Clinically significant data gaps | Upstream device performance | Device Reliability and Data Continuity | Gap-distribution metric; FDA iCGM special-control area | Frequency, duration, timing, and clustering of missing data capable of impairing intended connected use. | May interrupt alerts, obscure events, and bias derived metrics. | RT | FDA controls do not require uninterrupted data or prescribe one universal numerical gap threshold. |
| Signal loss/transmission reliability | Upstream device performance | Device Reliability and Data Continuity | Communication-performance metric | Frequency and duration of failures to transfer sensor data to a receiver, application, or connected device. | May impair monitoring continuity, alerts, and automated insulin-delivery function. | RT + IC | Distinguish sensor failure from communication failure. |
| Sensor survival to end of labeled wear | Upstream device performance | Device Reliability and Data Continuity | Reliability/survival metric | Proportion of sensors that remain functional through the manufacturer-labeled wear period. | Unexpected loss shortens monitoring coverage and increases burden and cost. | DDP + RT | Report censoring, reasons for removal, and wear-day distribution. |
| Premature sensor failure/replacement | Upstream device performance | Device Reliability and Data Continuity | Reliability and resource-use metric | Frequency and timing of sensor failures before end of labeled wear and resulting replacements. | May affect continuity, treatment burden, direct costs, and trust. | IC | Device-specific evidence may also apply. Standard definitions and exposure denominators are required. |
| Warm-up/start-up latency | Upstream device performance | Device Reliability and Data Continuity | Device-operational attribute | Time from insertion or activation to the first clinically usable reading. | Creates predictable monitoring gaps and can affect transition between sensors. | RT | Labeled duration and real-world completion should be distinguished. |
| Connectivity, interoperability, and secure transmission | Upstream device performance | Device Reliability and Data Continuity | System-integration attribute; FDA iCGM special-control area | Ability to exchange CGM data reliably and securely with intended receivers and compatible connected devices. | Supports data continuity and safe integrated use, including automated insulin delivery where applicable. | RT | Compatibility, cybersecurity, and intended-use claims are device- and ecosystem-specific. |
Note. Primary functional domain corresponds exclusively to one of the six domains defined in the framework. It identifies the construct most directly measured and does not imply that the metric has consequences only within that domain. Regulatory or technical provenance is shown separately. The FDA iCGM A–K criteria are therefore assigned primarily to Analytical Accuracy and Performance because they characterize sensor–comparator agreement, extreme discordance, and rate-of-change performance; FDA origin is not a seventh functional domain.
Interpretive levels: Upstream device performance = a property of the sensor/system or of the study used to characterize it; downstream derived metric = a summary calculated from the CGM trace that may be influenced by both physiology and device behavior; clinical/implementation outcome = safety, use, or user-experience construct observed in practice.
Evidence classification: DDP = direct device-performance evidence; DBC = direct between-device comparative evidence; IC = indirect/contextual evidence; RT = regulatory/technical evidence. More than one category may apply when a metric is informed by different types of evidence. These classifications describe the type and directness of information used in this framework and should not be interpreted as formal certainty-of-evidence ratings. Because the literature assessment was targeted rather than systematic and no formal risk-of-bias or certainty-of-evidence assessment was performed, evidence was not graded as high, moderate, or low certainty.
Abbreviations. ADRR, average daily risk range; CGM, continuous glucose monitoring; DBC, direct between-device comparative evidence; DDP, direct device-performance evidence; FDA, US Food and Drug Administration; GMI, glucose management indicator; GRI, glycemia risk index; GVP, glycemic variability percentage; HBGI, high blood glucose index; HbA1c, hemoglobin A1c; IC, indirect/contextual evidence; IFCC, International Federation of Clinical Chemistry and Laboratory Medicine; iCGM, integrated continuous glucose monitoring system; LBGI, low blood glucose index; MAD, mean absolute difference; MAGE, mean amplitude of glycemic excursions; MARD, mean absolute relative difference; MedAD, median absolute difference; RT, regulatory/technical evidence; SD, standard deviation; TAR, time above range; TBR, time below range; TIR, time in range.
All extracted metrics were classified into six functional domains: (i) analytical accuracy and performance, (ii) glycemic exposure, (iii) glycemic variability, (iv) clinical safety, (v) adherence and device use, and (vi) device reliability. The six domains represent primary construct-based assignments rather than mutually exclusive categories. Several metrics have downstream implications across domains. Analytical error metrics, for example, may influence clinical safety, while data continuity may affect reliability, adherence, and the ability to detect acute glycemic events. Cross-domain relationships are identified in Table 1.
This classification reflects the major thematic areas through which CGM systems influence clinical decision-making and patient outcomes. The domain structure also served as the organizational basis for the subsequent structured literature searches. All domain assignments and metric definitions are presented in Table 1.
Results From the Review
Analytical Accuracy and Performance
Analytical accuracy describes the agreement between CGM readings and glucose measurements obtained using a specified comparator method under defined study conditions. It should not be interpreted as agreement with an error-free estimate of “actual physiological glucose.” The observed disagreement may reflect CGM measurement error, comparator uncertainty, differences between interstitial and blood glucose compartments, physiological lag, measurement timing, pairing procedures, and other characteristics of the study design. Its interpretation therefore depends not only on the evaluated sensor but also on the quality, precision, traceability, sampling procedures, and temporal alignment of the comparator measurements.
Two related constructs were considered within this domain: (i) the magnitude and distribution of sensor–comparator disagreement; and (ii) the consistency of device behavior across glucose ranges, rates of change, wear periods, anatomical sites, and clinically relevant testing conditions. These characteristics may influence confidence in CGM-guided insulin dosing and pattern interpretation, particularly when decisions are made without confirmatory capillary testing.
MARD remains a useful aggregate descriptor of sensor–comparator disagreement, but it does not provide a complete characterization of CGM performance. Its value is influenced by the glucose distribution of the study sample, comparator methodology, sampling procedures, and study design. Moreover, similar overall MARD values may coexist with different performance in hypoglycemia, during rapid glucose changes, across wear periods, or in the frequency of large and potentially consequential errors. MARD should therefore be interpreted alongside glucose-range-specific agreement, confidence bounds, dynamic and rate-of-change performance, clinical-risk classifications, comparator quality, and data continuity.
Findings From Included Studies
The targeted search yielded 354 records relevant to analytical accuracy and performance. Many of these studies evaluated the analytical performance of a single CGM system against a specified comparator and were therefore recognized as direct device-performance evidence under the revised evidence framework. Three studies provided directly interpretable between-device evidence that also addressed dynamic testing conditions, device-specific differences, or downstream clinical implications and were selected for detailed synthesis below. Thus, the number of studies described in detail should not be interpreted as the total number of available CGM analytical-accuracy studies. Direct device-performance evidence, direct between-device comparisons, and contextual evidence are reported separately in the revised supplementary evidence summary.
Analytical Deviations in Daily-Life Use
An evaluation of two widely used CGM systems, one real-time and one intermittently scanned, conducted under both home and clinic conditions, showed that moderate analytical deviations were common. 8 At time points relevant for therapeutic decisions, approximately one in four readings differed from the BGMS reference by more than ±15 mg/dL or ±15% (24.3% and 25.0% for the two systems, respectively), and around 5% of values exceeded ±30 mg/dL or ±30%. Although considerable deviations (>±50 mg/dL/%) were uncommon, the overall frequency of moderate analytical error is clinically meaningful, particularly in scenarios requiring precise interpretation for insulin dosing or hypoglycemia prevention.
Sensor-to-sensor precision also differed: FreeStyle Libre showed higher site-to-site variability, reflecting less reproducible signal behavior across wear locations. Accuracy further deteriorated during rapid glycemic changes, when both systems showed higher MARD and larger absolute errors, precisely the periods when users rely most on trend information to guide dose adjustments or respond to emerging hypoglycemia.
Despite these deviations, clinical accuracy remained high (>99.5% of pairs in Clarke zones A+B), indicating that most sensor–comparator pairs were classified within lower-risk Clarke Error Grid zones. This classification estimates the potential clinical consequence of measurement disagreement but does not directly demonstrate the treatment decisions or outcomes that occurred. Still, the magnitude of moderate deviations has implications for fine-tuning bolus dosing and hypoglycemia management.
Comparison Across Two Generations of the Same CGM Platform
A comparative evaluation of two generations of the same CGM platform—one without iCGM designation and its iCGM-designated successor—showed that point accuracy was broadly similar between systems. Mean absolute relative difference values were 9.0% for the earlier system without FDA iCGM designation and 9.9% for the iCGM version, with nearly identical agreement within the ±20%/20 mg/dL threshold (93.1% vs. 92.5%). 9
Despite similar MARD and overall 20%/20 mg/dL agreement, the successor system satisfied the applicable range-specific FDA accuracy criteria, whereas the earlier generation did not satisfy all required lower one-sided 95% confidence-bound thresholds. This finding illustrates that MARD and overall agreement alone do not fully characterize analytical accuracy. Range-specific agreement, the frequency of large discordant pairs, rate-of-change performance, wear-period performance, and the statistical uncertainty surrounding agreement estimates provide additional analytical information. Separately, FDA iCGM designation also incorporates controls beyond analytical accuracy, including data availability, transmission, interference testing, usability, manufacturing controls, and labeling requirements.
In real-world use among 10,000 adults, switching from G5 to G6 was associated with: 34% reduction in time spent <55 mg/dL, 16% reduction in time <70 mg/dL, and a modest increase in time in target range (57.3% vs 56.0%). 9
Because this comparison involved successive generations of a single manufacturer’s platform, it cannot establish general differences between iCGM and non-iCGM categories. Shared technological lineage may reduce some sources of between-device heterogeneity, but generational changes in sensor design, calibration, algorithms, and functionality remain potential explanations for the observed differences. The comparison is therefore used to illustrate the limitations of relying on MARD or overall 20/20 agreement alone, rather than as evidence of category-wide superiority.
Analytical Performance Under Exercise-Induced Hypoglycemia
In a controlled clamp experiment designed to reproduce the rapid glucose declines and hypoglycemic conditions commonly encountered during physical activity, two CGM systems, one designated as an iCGM and the other an intermittently scanned system without FDA iCGM designation, demonstrated markedly different behaviors when exercise was introduced. 10 The intermittently scanned system showed an apparent deterioration in analytical performance: mean absolute relative difference increased during both the plasma glucose decline phase (DMARD +5.3%, 95% CI 1.6–9.1) and the early hypoglycemic period (DMARD +13.5%, 6.4–20.5), indicating that exercise amplified the magnitude of analytical error. Correspondingly, Clarke Error Grid analysis revealed a shift of sensor-reference pairs from clinically acceptable zones into zone D, reflecting a greater likelihood that treatment decisions would be misinformed under these conditions.
In contrast, the real-time system evaluated in the same study exhibited the opposite pattern. During exercise-related glucose decline, accuracy improved (DMARD −6.2%, −11.2 to −1.2), and a similar improvement was observed during sustained hypoglycemia (DMARD −8.4%, −12.4 to −4.3). Rather than compromising accuracy, exercise resulted in a higher proportion of values within Clarke zones A+B, suggesting improved clinical reliability precisely when physiological stress would be expected to challenge sensor performance. These contrasting responses should be interpreted as device-specific findings. The evaluated systems differed in modality, sensor and algorithm characteristics, data presentation, and regulatory status; the study design cannot isolate FDA iCGM designation as the causal explanation for their different behavior during exercise-induced glucose decline. Nevertheless, the findings demonstrate the clinical importance of evaluating individual systems under rapid rates of change and hypoglycemic conditions rather than relying exclusively on aggregate accuracy estimates obtained under more stable conditions.
Across the studies synthesized in detail, analytical performance varied according to the individual device, comparator methodology, glucose range, rate of glucose change, wear context, and physiological testing conditions. The findings also demonstrate why a single aggregate measure such as MARD cannot capture the full distribution or potential clinical relevance of sensor–comparator disagreement. Range-specific agreement, confidence bounds, large discordant errors, rate-of-change performance, and performance under dynamic conditions provide complementary information.
The G5–G6 comparison showed that similar aggregate accuracy estimates do not imply identical range-specific performance or equivalent satisfaction of the complete regulatory requirements. The exercise study demonstrated different responses between the individual systems evaluated during rapid glucose decline and hypoglycemia. However, neither comparison isolated regulatory designation from device generation, modality, sensor design, calibration, signal processing, or other technological differences. These studies therefore support device-level evaluation under clinically relevant conditions but do not establish category-wide superiority.
Glycemic Exposure
Glycemic exposure metrics derived from CGM—including mean glucose, time in range (TIR), time below range (TBR), time above range (TAR), and the glucose management indicator (GMI)—summarize the glucose profile recorded by the system. These are downstream clinical metrics rather than intrinsic measures of device performance. Their values reflect the combined influence of the participant’s underlying glycemia, treatment and behavior, sensor–comparator bias, calibration, signal processing, sampling frequency, lag, data gaps, and missing-data procedures.
Systematic measurement bias provides the most direct pathway through which analytical performance can affect glycemic exposure. For example, a device that reports systematically lower glucose values may produce higher TBR, lower TAR and mean glucose, and a lower GMI than another system measuring the same underlying glucose profile. Differences in filtering, interpolation, sampling, and data completeness may further influence the duration and classification of glucose excursions. Consequently, apparently identical exposure metrics should not automatically be assumed to be interchangeable across CGM systems.
Findings From Included Studies
The targeted search retrieved 772 records relevant to glycemic exposure. Most evaluated exposure metrics within a single device platform, included mixed-device cohorts without device-stratified results, or compared clinical interventions rather than the measurements generated by different CGM systems. The evidence most directly informing the technical-to-clinical question comprised one concurrent within-participant comparison of three CGM systems and contextual evidence regarding the transportability of GMI equations across devices.
Direct Between-Device Comparison of CGM-Derived Glycemic Metrics
Freckmann et al evaluated 23 adults who wore FreeStyle Libre 3, Dexcom G7, and Medtronic Simplera systems concurrently for 14 days. 11 Because the three systems recorded glucose in the same participants during the same observation period, the study permitted direct device-level comparison of the resulting glycemic metrics while substantially reducing variation attributable to differences between participants.
The systems generated clinically relevant differences in mean glucose, GMI, and times within, below, and above specified glucose ranges. These findings demonstrate that device selection can influence the glycemic exposure profile reported for an individual, even when the underlying physiological period is shared. The observed differences may reflect systematic measurement bias, calibration, algorithmic filtering, lag compensation, sampling characteristics, data availability, or other device-specific features. The study was not designed to isolate FDA regulatory designation as the causal explanation, and its findings should therefore be interpreted at the level of the individual systems evaluated.
Contextual Evidence Regarding Device-Specific GMI Relationships
In a single-centre observational cohort of 120 adults with type 1 diabetes using Medtronic Guardian 3 or Guardian 4, device-specific equations relating mean CGM glucose to laboratory HbA1c were compared with the original GMI equation, which was developed using data substantially derived from earlier-generation Dexcom systems. 12 The original equation showed systematic discrepancies across clinically relevant HbA1c values, whereas application of device-specific equations reduced the magnitude of the difference between GMI and HbA1c.
These findings illustrate that the relationship between mean CGM glucose and laboratory HbA1c may not transport uniformly across device platforms or populations. They do not demonstrate that one regulatory category provides a more valid GMI. Rather, they show that the derivation dataset, device-specific measurement behavior, population characteristics, and biological determinants of HbA1c should be considered when interpreting discordance between GMI and laboratory HbA1c.
Taken together, the identified evidence supports glycemic exposure as a downstream interpretive domain through which device-specific measurement and processing characteristics may become clinically visible. The concurrent comparison by Freckmann et al provides direct evidence that different CGM systems can generate different exposure metrics during the same observation period. However, the available studies do not establish that FDA iCGM designation independently determines the direction or magnitude of these differences. Comparisons of CGM-derived exposure should therefore identify the device model and software generation and, when possible, account for systematic bias, sampling frequency, data completeness, and missing-data procedures.
Glycemic Variability
Glycemic variability (GV) captures the amplitude and frequency of glucose fluctuations and includes both distribution-based metrics, such as SD and CV, and risk-oriented indices, such as LBGI, HBGI, ADRR, and GRI. These metrics are clinically relevant because greater variability is associated with hypoglycemia risk, treatment burden, and impaired quality of life.
Variability metrics calculated from CGM data may inherit device-specific measurement and processing characteristics. Sampling frequency, algorithmic smoothing, interpolation, physiological and computational lag, data gaps, and missing-data rules may influence the number, amplitude, duration, and rate of detected glucose excursions. Metrics such as GVP, MAGE, GRI, LBGI, and ADRR may be particularly sensitive to these factors, although SD and CV should not be assumed to be fully interchangeable across devices without comparative evaluation.
Findings From Included Studies
The targeted search identified a broad body of literature evaluating GV in adults using CGM. However, most studies evaluated variability within a single device platform or included mixed-device cohorts without reporting device-stratified results. No eligible concurrent within-person study was identified that was specifically designed to determine whether two CGM systems produced different GV estimates while measuring the same underlying glucose profile.
One randomized study comparing real-time CGM (rtCGM) with intermittently scanned CGM (isCGM) in adults at high risk of hypoglycemia provided direct between-device evidence regarding the clinical modification of GV during active device use. 13 In that trial, rtCGM was associated with greater improvements in several variability and risk-related metrics, including CV, LBGI, and ADRR, as well as reductions in biochemical hypoglycemia. The isCGM group showed improvements in selected fluctuation-over-time measures.
These outcomes were secondary, and the study was not specifically powered for between-group differences in GV metrics. Moreover, because each system actively informed treatment and user behavior, the observed differences cannot isolate the effects of measurement performance from those of real-time streaming, alarm availability, user interaction, engagement, data processing, or other device-specific characteristics. The findings therefore represent a direct comparison of the clinical effects of two device modalities, but not a direct comparison of their measurement of the same glycemic variability.
The available evidence supports distinguishing between measurement of variability and clinical modification of variability. Concurrent within-person studies are needed to determine whether devices produce different GV estimates under the same physiological conditions, whereas randomized crossover or parallel-group trials are more appropriate for evaluating whether device use modifies variability and hypoglycemia risk. The identified evidence does not establish that regulatory classification independently determines either type of outcome.
Clinical Safety
Clinical safety in CGM encompasses the potential consequences of measurement error, delayed or missed detection of hazardous glucose values, clinically significant data gaps, alert performance when alerts are available, and the user’s ability to respond appropriately to the information provided.
The iCGM special controls include requirements related to accuracy across glucose ranges, specified rate-of-change mismatches, and the absence of clinically significant data gaps that could prevent connected devices from achieving their intended use. When alerts are applicable, labeling must report relevant alert-performance characteristics. However, alarm availability is not itself required for iCGM designation, and alarm configuration and performance remain individual device characteristics.
The objective of this domain was to identify studies evaluating the relationship between device characteristics—particularly low-glucose measurement performance, monitoring continuity, and alarm functionality—and safety-related outcomes such as hypoglycemia, time below range, and treatment burden.
Findings From Included Studies
Of 78 records screened for this domain, one study meeting the domain-specific eligibility criteria evaluated the clinical effect of introducing alarms within the same sensor product family. 14 The study included 100 adults with type 1 diabetes who transitioned from FreeStyle Libre without alarms to FreeStyle Libre 2 with alarms in routine practice.
After three months, time spent at 70–54 mg/dL decreased from 4.3% to 3.6%, and time below 54 mg/dL decreased from 1.5% to 0.6%. Time in range increased, hyperglycemic exposure and CV decreased, and users interacted more frequently with the system. No modification of the insulin regimen or basal dose was reported.
These findings are compatible with a clinically relevant contribution from alarm availability and subsequent user response. However, the before–after observational design cannot exclude behavioral changes, increased engagement, temporal effects, regression to the mean, or other unmeasured factors. Furthermore, the study did not evaluate iCGM designation as the exposure of interest and therefore does not provide evidence of a regulatory-category effect.
The potential clinical benefit of alarms depends not only on their availability but also on their sensitivity, timing, false-alert and missed-alert rates, correct-alert rate, alarm burden, and the user’s ability and willingness to respond. The available evidence supports the clinical relevance of evaluating these device-level characteristics but does not establish that iCGM designation independently reduces hypoglycemia or treatment burden.
Adherence and Device Use
This domain examined how sensor wear, data availability, monitoring interruptions, and persistence with CGM use may affect clinical outcomes. Because the benefits of CGM depend on the availability and use of sensor data, premature sensor failure, detachment, medically required removal, connectivity problems, supply interruptions, or user discontinuation may reduce the capacity of the system to support therapeutic decisions.
Adherence and device use were operationalized through several related but distinct measures, including intended wear time, proportion of available sensor data, scanning or interaction frequency, interruption of monitoring, premature sensor discontinuation, and persistence with CGM use. Because these measures do not represent an interchangeable construct, they were considered separately and were not pooled.
Within the targeted search, no eligible randomized crossover or parallel-group study was identified that directly compared adherence-related or treatment-guided outcomes between devices while adequately separating device characteristics from patient, healthcare-system, and access-related factors. Simultaneous wear may permit comparison of technical data availability, but it cannot provide an unbiased comparison of adherence or treatment-guided outcomes when only one device can ordinarily be used to guide therapeutic decisions.
One survey-based study provided contextual evidence regarding the consequences of CGM interruption. 15 Interruptions in sensor wear were frequently reported, and only a small proportion of respondents reported achieving the full intended wear time throughout the preceding 12 months. Reported causes included device malfunction, insertion problems, adhesive failure, and removal for medical procedures. These interruptions occurred among users of multiple CGM brands.
More than one-third of respondents reported at least four hyperglycemic events per year during periods without CGM data, and more than one in ten reported recurrent hypoglycemia. However, these outcomes and their attribution to CGM interruption were self-reported and cross-sectional. The study therefore does not provide independently verified event rates or establish a causal relationship between interruption and adverse glycemic outcomes.
The findings demonstrate that monitoring interruption is a clinically relevant implementation problem, but they do not establish differences according to regulatory designation. Adherence and continuity of use are multifactorial and may be influenced by sensor durability, adhesive performance, connectivity, alarm burden, user preference, education, digital literacy, reimbursement, access to replacement sensors, and socioeconomic conditions. Consequently, regulatory classification should not be treated as the primary determinant of adherence without direct comparative evidence.
Device Reliability
Within the targeted search, no eligible study directly compared the technical reliability of multiple CGM systems under comparable conditions while also linking sensor survival, signal loss, or technical failure to clinical or health-economic outcomes. This domain was therefore informed primarily by contextual evidence rather than purpose-designed comparative studies.
A survey-based analysis of adults using several contemporary CGM systems reported that approximately 10–15% of sensors did not reach their intended wear duration, despite labeled wear periods of 7–14 days. 16 Sensor failure was the most frequently reported reason, followed by heat exposure, adhesive or patch-related problems, perceived accuracy concerns, and physical activity. Premature discontinuation was reported across several CGM brands.
These findings suggest that failure to complete the labeled wear period is relevant to real-world monitoring continuity. However, the results were self-reported, were not confirmed through device records or manufacturer data, and were not stratified in a manner that permitted robust comparison among individual systems or regulatory categories. The reported proportions should therefore not be interpreted as independently verified device-failure rates.
A cost–utility analysis comparing Dexcom G6 rtCGM with alarms and first-generation FreeStyle Libre 1 isCGM without alarms in adults with type 1 diabetes in Belgium provided complementary economic evidence. 17 Using the IQVIA CORE Diabetes Model, the analysis projected lower lifetime costs and greater quality-adjusted survival with Dexcom G6 under its specified assumptions. These projections were driven primarily by differences in HbA1c, severe hypoglycemia, and fear of hypoglycemia derived from the underlying clinical evidence.
The economic model did not explicitly incorporate sensor-failure rates, premature sensor loss, replacement costs, or the duration of monitoring gaps. Moreover, the compared systems differed in alarm availability, monitoring modality, user interaction, measurement characteristics, and other device-specific features. Its findings therefore cannot be interpreted as direct evidence of greater technical reliability or as an effect of iCGM designation.
Overall, the available evidence indicates that premature sensor discontinuation and monitoring gaps may have clinical and economic consequences, but it does not establish category-specific differences in technical reliability. Future comparative studies should separately quantify completion of labeled wear, signal loss, sensor failure, replacement requirements, duration of monitoring gaps, associated adverse events, and direct and indirect costs.
Expert Perspective
Evolution of Statements Across Consensus Rounds
The consensus process comprised two structured voting rounds with an intermediate phase of discussion and refinement. Across rounds, the statements remained conceptually stable, with revisions primarily reflecting semantic refinement, increased clinical caution, and clarification of scope rather than substantive changes in expert judgment.
Eight of the 25 statements underwent wording changes between rounds, whereas 17 retained the same formulation. Revisions attenuated deterministic or causally attributive language, clarified the clinical construct being assessed, or removed unsupported explanations based solely on regulatory category. Some statements underwent broader conceptual reframing rather than simple semantic editing. No statements were removed, and no new conceptual domains were introduced. The statement-level wording used in each round and the rationale for each refinement are presented in Supplementary Table 1.
No statements were removed, and no new conceptual domains were introduced between rounds, supporting the internal coherence and stability of the initial consensus framework.
Overall Consensus Outcomes
The expert consensus, structured into six predefined functional domains, demonstrated high agreement across the statement set. After the first voting round, 24 of 25 statements met criteria for appropriateness (median ≥7) without disagreement. One statement, addressing differential dependence on user adherence in non-iCGM systems, was initially classified as uncertain because disagreement was present according to the predefined RAND/UCLA criterion.
Following structured discussion and statement refinement, all statements achieved consensus in the second round, with median ratings ≥7 and no disagreement according to RAND/UCLA criteria (IPR ≤ IPRAS). Median scores remained consistently high across domains—frequently reaching values of 8 or 9—while dispersion measures decreased between rounds, indicating convergence of expert judgment. Detailed final-round ratings and their evidence-aligned interpretive context are presented in Table 2.
Table 2.
Final-Round Expert Ratings of Statements Concerning the Clinical Interpretation of CGM Performance Characteristics. The Table Presents Faithful English Translations of the Statements Rated in Round 2. The Final Column Contextualizes the Panel Ratings Against the Evidence Assessment and was Not Itself Rated by the Panel
| Statement group | Statement rated in round 2 | Median | IPR | IPRAS | Disagreement | Consensus | Evidence-aligned interpretation |
|---|---|---|---|---|---|---|---|
| Analytical Accuracy and Performance | iCGM systems provide more consistent analytical accuracy under dynamic conditions, particularly during rapid glucose changes. | 9.0 | 0.0 | 4.00 | No | Achieved | Device-specific comparisons support the relevance of dynamic testing but do not establish a category-wide difference. |
| Analytical Accuracy and Performance | For therapeutic decisions without capillary confirmation, accuracy stratified by glycemic range and rate of change is more reliable in iCGM than in non-iCGM systems. | 9.0 | 0.0 | 4.00 | No | Achieved | Range- and rate-stratified performance is clinically relevant; universal superiority by regulatory category was not demonstrated. |
| Analytical Accuracy and Performance | The iCGM designation represents a clinically meaningful standard that ensures superior analytical performance compared with non-iCGM systems. | 9.0 | 0.0 | 4.00 | No | Achieved | iCGM designation confirms compliance with specified FDA requirements; it does not guarantee superiority over every system without the designation. |
| Glycemic Exposure | Glycemic exposure metrics (TIR, TBR, TAR) derived from iCGM systems are more stable and less susceptible to device-dependent bias than those from non-iCGM systems. | 8.0 | 0.8 | 3.40 | No | Achieved | Concurrent evidence shows device-level differences in derived metrics, but no consistent direction according to iCGM status. |
| Glycemic Exposure | Clinical interpretation of TIR is more reliable when derived from iCGM sensors due to greater underlying analytical consistency. | 9.0 | 0.8 | 3.60 | No | Achieved | TIR interpretation depends on device accuracy, bias, sampling, processing, and missing data; a category-wide advantage was not established. |
| Glycemic Exposure | Discrepancies between GMI and HbA1c are smaller with iCGM systems than with non-iCGM systems. | 8.0 | 0.8 | 2.60 | No | Achieved | Device-specific bias can alter GMI, but the targeted evidence did not establish systematically smaller GMI–HbA1c discordance by category. |
| Glycemic Variability | iCGM systems allow more precise clinical interpretation of glycemic variability due to higher accuracy. | 9.0 | 0.8 | 3.60 | No | Achieved | Variability estimates may be affected by accuracy and data processing; no direct concurrent evidence demonstrated category-wide superiority. |
| Glycemic Variability | Variability metrics derived from non-iCGM systems may over- or underestimate hypoglycemia risk compared with those derived from iCGM. | 8.0 | 0.8 | 3.40 | No | Achieved | Device effects are plausible, but neither their direction nor attribution to regulatory status was established. |
| Glycemic Variability | Glycemic risk indices (LBGI, ADRR) are more clinically reliable when derived from iCGM systems than from non-iCGM systems. | 8.0 | 0.0 | 3.00 | No | Achieved | These indices depend on low-range accuracy, sampling, preprocessing, and missingness; comparative category-level evidence was not identified. |
| Clinical Safety | iCGM systems provide greater protection against hypoglycemia than non-iCGM systems, due in part to regulatory alarm requirements and better performance under physiological stress. | 9.0 | 0.0 | 4.00 | No | Achieved | Alerts are not required for iCGM designation. Potential benefit depends on device-specific alert availability, timing, and true, false, and missed-alert rates. |
| Clinical Safety | Non-iCGM systems have a higher risk of failing to detect rapid or nocturnal hypoglycemia compared with iCGM systems. | 9.0 | 0.0 | 4.00 | No | Achieved | The targeted evidence did not demonstrate a category-wide difference; low-range, dynamic, continuity, and alert performance should be evaluated per device. |
| Clinical Safety | Hypoglycemia accuracy is significantly higher in iCGM sensors than in non-iCGM sensors and translates into greater clinical safety. | 9.0 | 0.0 | 4.00 | No | Achieved | Low-range performance is clinically important, but category-wide analytical or clinical superiority was not established. |
| Adherence and Device Use | iCGM systems may contribute to greater continuity of use, due in part to lower rates of failures and disconnections compared with non-iCGM systems. | 8.0 | 2.4 | 2.60 | No | Achieved | Continuity is influenced by device, user, access, and care-system factors; no category-level comparative evidence was identified. |
| Adherence and Device Use | Interruptions are lower in iCGM systems, which could lead to lower discontinuation rates than in users of non-iCGM systems. | 8.0 | 0.8 | 3.40 | No | Achieved | Neither lower interruption nor lower discontinuation according to iCGM status was established in the targeted evidence. |
| Adherence and Device Use | The clinical benefit of CGM depends more on adherence in non-iCGM systems due to greater variability in performance. | 8.0 | 1.6 | 3.00 | No | Achieved | Adherence is important for every CGM system; differential dependence according to regulatory category remains unproven. |
| Device Reliability and Data Continuity | iCGM systems are more reliable in achieving full sensor survival through the labeled wear period compared with non-iCGM systems. | 8.0 | 0.8 | 3.40 | No | Achieved | No direct comparative evidence established category-specific differences in completion of labeled wear. |
| Device Reliability and Data Continuity | Signal loss or premature sensor failures occur more frequently in non-iCGM systems than in iCGM systems. | 8.0 | 0.8 | 3.40 | No | Achieved | Signal loss and sensor failure should be distinguished and measured directly; a category-wide difference was not established. |
| Device Reliability and Data Continuity | Monitoring continuity may be greater in iCGM than in non-iCGM systems, with potentially important clinical implications for users. | 9.0 | 0.8 | 3.60 | No | Achieved | Continuity has plausible clinical implications, but superiority according to iCGM designation remains hypothesis-generating. |
| Cross-domain Integration | Overall, iCGM systems provide more technically robust information for clinical decision-making than non-iCGM systems. | 9.0 | 0.0 | 4.00 | No | Achieved | iCGM criteria define specified performance expectations; overall clinical robustness still requires device- and outcome-specific evidence. |
| Cross-domain Integration | Clinical interpretation of technical metrics is more reliable when derived from iCGM devices due to regulatory standardization. | 9.0 | 0.8 | 3.60 | No | Achieved | Regulatory standardization informs covered attributes but does not guarantee the reliability of every metric or clinical outcome. |
| Cross-domain Integration | Non-iCGM systems require more cautious interpretation of metrics. | 9.0 | 0.0 | 3.00 | No | Achieved | All devices require device-specific interpretation. Absence of FDA iCGM designation does not itself demonstrate inferior performance. |
| Cross-domain Integration | From a clinical perspective, differentiating between iCGM and non-iCGM is essential to appropriately evaluate CGM performance metrics. | 9.0 | 0.0 | 4.00 | No | Achieved | Regulatory status is one contextual variable and should be considered separately from individual device characteristics and study methods. |
| Clinical Priority and Regulatory Outlook | Hypoglycemia and rapid-decline performance metrics should be evaluated with greater caution in non-iCGM systems. | 9.0 | 0.0 | 4.00 | No | Achieved | These conditions should be evaluated carefully for every system; higher risk cannot be inferred solely from absence of iCGM designation. |
| Clinical Priority and Regulatory Outlook | Differences between iCGM and non-iCGM systems directly impact patient safety and the precision of therapeutic decision-making. | 9.0 | 0.0 | 4.00 | No | Achieved | Device-characteristic differences may affect safety and decisions, but regulatory designation was not isolated as the causal factor. |
| Clinical Priority and Regulatory Outlook | Regulatory harmonization should be based on iCGM performance standards, given evidence of greater clinical reliability. | 9.0 | 0.0 | 4.00 | No | Achieved | Harmonization may draw on explicit FDA performance expectations and standardized IFCC-aligned study procedures; greater clinical reliability by category was not established. |
Panel and scale. All five panelists rated every statement in Round 2 using a 1–9 scale: 1–3 = inappropriate or low clinical relevance; 4–6 = uncertain; and 7–9 = appropriate or high clinical relevance.
Consensus criteria. Consensus was classified as achieved when the median was ≥7 and disagreement was absent. Disagreement was defined as IPR > IPRAS according to the prespecified RAND/UCLA analytical rule.
Interpretation. Consensus strength represents convergence of expert judgment and should not be interpreted as evidence strength, proof of causality, or confirmation of category-wide superiority. Several statements retain categorical comparisons from the original consensus framework. FDA iCGM designation and individual device characteristics should be interpreted as related but distinct constructs.
Abbreviations. ADRR, average daily risk range; CGM, continuous glucose monitoring; FDA, US Food and Drug Administration; GMI, glucose management indicator; HbA1c, hemoglobin A1c; iCGM, integrated continuous glucose monitoring system; IPR, interpercentile range; IPRAS, interpercentile range adjusted for symmetry; LBGI, low blood glucose index; TAR, time above range; TBR, time below range; TIR, time in range. Non-iCGM, CGM system without FDA iCGM designation at the time of the review.
These findings indicate that the panel considered regulatory and technical distinctions clinically relevant across multiple domains. The ratings reflect expert interpretation and should not be construed as comparative empirical evidence or as confirmation of category-wide clinical superiority.
Domain 1. Analytical Accuracy and Performance
The panel judged that meeting stringent range-specific and rate-of-change performance requirements may increase confidence in the interpretation of CGM values, particularly when therapeutic decisions are made without confirmatory capillary testing. However, the panel ratings should be interpreted as judgments about the potential clinical relevance of these requirements rather than as evidence that every iCGM-designated device is more accurate than every device without this designation. The comparative evidence identified in this domain remained device-specific.
Domain 2. Glycemic Exposure
The panel considered it plausible that systematic differences in analytical performance could influence mean glucose, TIR, TBR, TAR, and GMI. Direct concurrent evidence supports the existence of device-level differences in CGM-derived exposure metrics. However, the identified evidence does not establish that iCGM systems consistently produce less biased exposure estimates or smaller GMI–HbA1c discrepancies as a regulatory category. The consensus ratings therefore represent expert interpretation of the potential downstream consequences of device-specific measurement bias.
Domain 3. Glycemic Variability
The panel considered analytical consistency and data processing relevant to the interpretation of variability metrics, including CV, LBGI, and ADRR. Experts considered it plausible that measurement bias, smoothing, sampling, missing data, and rate-of-change performance could result in different variability estimates across devices. Nevertheless, no direct concurrent evidence was identified demonstrating that devices without iCGM designation systematically overestimate or underestimate glycemic variability or hypoglycemia risk.
Domain 4. Clinical Safety
All safety-related statements achieved strong agreement. The panel identified low-range accuracy, response during rapid glucose changes, monitoring continuity, and alert performance when alerts are available as clinically relevant safety characteristics. However, the identified empirical evidence does not demonstrate that devices without iCGM designation have a category-wide higher risk of missing rapid or nocturnal hypoglycemia. The consensus findings should instead be interpreted as emphasizing the importance of evaluating these characteristics for each individual device.
Domain 5. Device Use and Adherence
The panel considered monitoring continuity and sustained device use important determinants of the clinical benefit of CGM. Experts also considered it plausible that interruptions could have greater consequences when combined with less predictable measurement or data availability. However, no comparative evidence by regulatory category was identified, and adherence is influenced by multiple device-, user-, healthcare-system-, and access-related factors. The consensus therefore does not establish that adherence is systematically lower or more clinically consequential in devices without iCGM designation.
Domain 6. Device Reliability and Continuity
Experts considered sensor survival, signal stability, and continuity of monitoring clinically relevant aspects of CGM performance. The panel hypothesized that stringent technical and performance expectations may support greater reliability. However, no direct comparative evidence was identified demonstrating superior sensor survival, lower failure rates, or fewer monitoring gaps according to iCGM status. These ratings should therefore be regarded as hypothesis-generating expert judgments.
Global Integration of the Framework
Integrative statements achieved high agreement. The panel identified hypoglycemic ranges and rapid glucose changes as priority conditions for evaluating CGM performance and considered low-range accuracy, rate-of-change behavior, monitoring continuity, and alert performance potentially relevant to patient safety and therapeutic decision-making. These ratings represent expert interpretation and do not constitute comparative empirical evidence or confirmation of category-wide clinical superiority. The panel also supported the potential value of using clearly defined performance expectations to inform future regulatory harmonization.
Clinical Priority
The panel emphasized that performance during hypoglycemia and rapid glucose change should be examined carefully for every CGM system, irrespective of regulatory designation. A higher risk of error should not be assumed solely from the absence of iCGM designation but should be established using device-specific evidence generated with appropriate comparator methods and standardized study procedures. Figure 1 summarizes the design classification of studies selected for detailed synthesis across the six domains alongside the final median expert ratings. The coexistence of high expert agreement with limited direct comparative evidence in several domains highlights the distinction between consensus strength and evidentiary support.
Figure 1.
Evidence Base (Direct and Indirect) and Final Expert Agreement Across CGM Performance Domains. Evidence represented and final expert agreement across CGM domains. Panel A shows the number of studies selected for detailed domain-level synthesis, classified according to the most direct design applicable to the corresponding technical to-clinical question. Panel B shows the median expert rating in Round 2. Counts do not represent all records retrieved or all available CGM performance studies. Expert ratings reflect convergence of panel judgment and should not be interpreted as certainty-of evidence estimates or confirmation of category-wide superiority. Studies contributing to more than one domain were classified separately within each relevant domain
Discussion
The findings of this review reveal a marked mismatch between the conceptual complexity of CGM performance constructs and the limited and uneven empirical evidence linking them to clinically relevant outcomes. Although more than 30 evaluation metrics were identified across contemporary clinical, technical, and regulatory sources, the type and strength of supporting evidence varied substantially by domain. Direct device-performance and between-device evidence was identified most clearly for analytical accuracy, while a concurrent comparison also demonstrated device-level differences in glycemic exposure metrics. A randomized comparison of rtCGM and isCGM informed the clinical modification of glycemic variability but could not isolate measurement effects from differences in alarms, data presentation, user interaction, and other device characteristics. Clinical safety, adherence, and device reliability were informed predominantly by observational, device-specific, modeled, or contextual evidence. This fragmentation highlights the difficulty of deriving clinically meaningful inferences from isolated metrics and provides a central rationale for organizing CGM performance constructs within a structured, domain-based framework.18,19
An important interpretive boundary should be recognized when applying this framework. iCGM designation indicates that a device has satisfied defined regulatory criteria; it does not, by itself, establish superiority in clinical outcomes or user experience. Differences between individual CGM systems may arise from multiple device-specific characteristics, including sensor chemistry, calibration requirements, signal-processing algorithms, sampling frequency, alarm configuration, connectivity, software, and user interface. Accordingly, where direct head-to-head evidence is unavailable, the distinctions identified in this review should be regarded as expert-derived and hypothesis-generating rather than as established category-wide advantages of iCGM systems.
Across domains, a consistent interpretive pattern emerged: the clinical meaning of CGM metrics depends not only on the metric selected but also on device-specific characteristics and the conditions under which performance was evaluated. Algorithmic filtering, sampling frequency, physiological and computational lag, responsiveness to rapid glucose changes, alarm functionality, data completeness, and real-world reliability may all influence reported performance and CGM-derived outcomes. Consequently, apparently similar metrics should not be assumed to be interchangeable across devices.
The IFCC Working Group proposal reinforces this interpretive principle by standardizing not only which aspects of CGM performance should be assessed but also the comparator methods, glucose ranges, rates of change, sampling procedures, and study conditions under which they should be evaluated. 2 For clinicians, interpretation of a reported accuracy or performance estimate should therefore consider the comparator method, synchronization and pairing procedures, distribution of reference measurements across clinically relevant dynamic glucose ranges, study setting, participant characteristics, and handling of missing data. Device-level differences in glycemic exposure and GMI further illustrate how measurement behavior and derivation procedures may affect downstream interpretation even when metrics share the same definitions.11,12 These considerations are particularly relevant in regions such as Latin America, where heterogeneous device availability and regulatory pathways increase clinicians’ reliance on transparent, device-specific performance information.
Within this landscape, the six-domain structure proposed in this work offers a clinically oriented framework for organizing and interpreting CGM performance metrics. It is essential to acknowledge, however, that this classification is inherently conceptual and reflects expert judgment regarding how technical constructs map onto clinically meaningful processes. To our knowledge, no prior publication has proposed an explicitly clinical-domain architecture for CGM evaluation. As such, the framework is novel but necessarily interpretive. Its boundaries, internal organization, and metric prioritization within each domain are shaped by expert consensus and remain open to refinement as additional comparative evidence and real-world data become available. Accordingly, the proposed domains should be viewed as a flexible structure intended to support clinical reasoning and facilitate more systematic assessment and regulatory harmonization, rather than as a fixed or exhaustive taxonomy.
The consensus process adds an expert interpretive layer to the evidence synthesis by identifying device attributes that the panel considered potentially relevant to clinical interpretation. The high level of agreement across the 25 statements demonstrates internal consistency within this panel but should not be interpreted as an indicator of the strength of the underlying comparative evidence or of the magnitude of any clinical benefit. In domains supported predominantly by indirect or device-specific evidence, the consensus statements represent informed clinical judgments requiring prospective validation.20,21 Importantly, the consensus did not merely reiterate technical specifications identified in the literature; instead, it translated heterogeneous and incomplete evidence into a coherent clinical interpretation applicable to real-world decision-making, particularly in dynamic, safety-critical scenarios.
The evolution of the statement addressing adherence and monitoring continuity illustrates the sensitivity of consensus outcomes to framing and wording. The statement initially showed disagreement when it attributed differential dependence on adherence specifically to non-iCGM systems. Following structured discussion, its wording was refined to clarify the clinical relevance of monitoring continuity and to attenuate the regulatory-category attribution. The subsequent convergence indicates that the panel reached a shared interpretation of the revised statement; it should not be interpreted as empirical resolution of the underlying uncertainty or as evidence that adherence has systematically different effects according to iCGM designation. This distinction is particularly important because adherence and continuity of CGM use reflect interactions among user behavior, device reliability, education and digital literacy, socioeconomic conditions, reimbursement and access, healthcare-system factors, and analytical performance, while direct comparative evidence remains limited.
Several limitations warrant consideration. The literature searches were intentionally targeted, clinically focused, and restricted to PubMed/MEDLINE rather than designed as a comprehensive systematic review. Consequently, relevant clinical, engineering, regulatory, technical, or device-performance studies indexed exclusively in Embase, Scopus, Web of Science, the Cochrane Library, engineering databases, or grey-literature sources may not have been captured. Evidence counts should therefore be interpreted as the yield of the predefined searches rather than as exhaustive estimates of the available literature. Publication bias is also possible because studies reporting favorable, statistically significant, or clinically relevant device performance may be more likely to be published. Direct comparative evidence was sparse and unevenly distributed across domains, and several domains relied predominantly on observational, device-specific, modeled, or contextual evidence. Furthermore, regulatory designation does not capture all differences among individual devices, including sensor chemistry, calibration, algorithms, sampling frequency, alarm functionality, connectivity, software, and user interface. The designation itself should therefore not be interpreted as evidence of clinical superiority.
The consensus process also has limitations. The initial statements were generated by the academic committee and framing or confirmation bias toward the selected regulatory and technical constructs cannot be excluded. Independent private voting, prespecified agreement and disagreement criteria, structured discussion without forced convergence, and attenuation of categorical wording were used to reduce this risk; nevertheless, the final ratings represent expert judgments regarding clinical plausibility and relevance rather than confirmation of comparative benefit. The panel was small, purposively selected, and geographically concentrated. Although its composition was relevant to the Latin American context motivating the study, regional perspectives regarding device availability, access, reimbursement, clinical workflows, and regulatory priorities may limit international generalizability. In addition, the evidence search focused primarily on ambulatory adults and does not directly establish the applicability of the framework to pediatric populations, pregnancy, older adults with specific vulnerabilities, or hospitalized patients. Finally, the proposed domains are conceptual, may overlap, and have not undergone independent external validation. External evaluation involving additional regions, disciplines, health systems, patient representatives, and clinical populations is therefore warranted.
Finally, search-syntax limitations may also have affected retrieval. The use of pediatric exclusion terms may have excluded mixed-age studies even when adult results were reported separately. In selected domains, exclusions related to pregnancy, chronic kidney disease, hospitalization, or inpatient care may likewise have excluded mixed-population or ambulatory studies mentioning these characteristics or outcomes. In the Clinical Safety search, the mandatory device-name block and the separate use of Guardian and Medtronic may have reduced both sensitivity and device specificity. The reported evidence should therefore be interpreted as the yield of the predefined searches rather than as evidence that no additional relevant studies exist.
Within these constraints, the findings indicate that the panel considered several device-performance characteristics clinically relevant, particularly low-range accuracy, behavior during rapid glucose change, alarm performance when alarms are available, data continuity, and reliability. Direct comparisons demonstrated device-specific differences in analytical performance and glycemic exposure metrics, while randomized evidence comparing CGM modalities suggested that alarms, real-time data availability, and user interaction may influence clinical outcomes. However, these studies do not establish that regulatory designation independently explains the observed differences. Evidence for clinical safety, adherence, and device reliability remained predominantly observational or contextual. The proposed interpretations should therefore be regarded as plausible but provisional and should be evaluated through purpose-designed comparisons of individual CGM systems, irrespective of their regulatory classification.
Beyond device-to-device comparison, future CGM interpretation frameworks may increasingly integrate sensor performance characteristics and longitudinal glucose patterns with complementary clinical and biological information. Exploratory studies have shown that combining CGM-derived features with genetic risk information may contribute to risk classification in early type 1 diabetes, 22 while machine-learning approaches using longitudinal CGM features have shown potential for identifying adults with type 1 diabetes at risk of incident diabetic retinopathy. 23 These applications illustrate how CGM data could contribute to individualized risk assessment beyond isolated technical performance metrics. However, they do not establish comparative superiority of any CGM category, and their clinical use will require external validation, evaluation of calibration and transportability across populations and devices, and explicit consideration of the quality and completeness of the underlying sensor data.
To move from expert-derived plausibility toward comparative evidence, future CGM studies should report a standardized minimum dataset and select the study design according to the question being addressed. Concurrent within-participant comparisons are preferable when the objective is to determine whether devices produce different measurements or derived metrics under the same physiological conditions. Randomized crossover or parallel-group designs are more appropriate when the objective is to compare how active device use, alarms, data presentation, or user interaction modifies behavior and clinical outcomes. Studies should identify the device model, sensor generation, algorithm and software version, calibration requirements, regulatory status, labeled and observed wear duration, sensor placement, clinical setting, and whether each device was blinded or used to guide treatment.
Consistent with the IFCC Working Group proposal, analytical assessment should use an appropriately characterized and traceable comparator method, synchronized measurements, predefined pairing procedures, and sufficient reference observations across clinically relevant glucose concentrations and dynamic rates of change. 2 Reports should include range-stratified agreement with confidence intervals, error distributions, rate-of-change or trend accuracy, physiological and algorithmic lag, and performance under relevant study conditions. Where alarms are available, studies should report alert thresholds, sensitivity, timing, false-alert, missed-alert, and correct-alert rates. Data availability, clinically significant gaps, signal loss, completion of labeled wear, premature sensor failure, replacement requirements, and reasons for discontinuation should also be reported.
Recent international clinical opinion has further emphasized that meaningful cross-system comparison requires not only standardized study procedures and reference methods but also transparency of the supporting clinical evidence and attention to calibration alignment. 24 Differences in calibration approaches and alignment with reference methods may affect cross-system comparability and should therefore be considered when interpreting reported glucose values and performance estimates. These considerations reinforce the need to report both how device performance was evaluated and how calibration and reference-method alignment may influence cross-device comparisons.
At the downstream clinical-interpretation level, dietary exposures, metabolic regulation, and interindividual metabolic physiology shape the glucose patterns from which CGM-derived metrics are calculated. 25 These biological influences should be distinguished from analytical sensor error: they may alter the measured glucose profile and its clinical interpretation without necessarily indicating impaired device accuracy. Future interpretive frameworks may therefore benefit from integrating standardized device-performance information with explicit characterization of the physiological and clinical context in which CGM data are generated.
Comparisons of CGM-derived metrics should use standardized definitions and disclose sampling frequency, algorithmic smoothing when known, interpolation, minimum data sufficiency, and missing-data procedures. Relevant clinical and user-centered outcomes should include independently confirmed hypoglycemia, treatment-decision discrepancies, adherence, alarm burden, usability, adverse events, confirmatory capillary testing, and direct and indirect costs. Participant characteristics and clinically relevant subgroups, including age, pregnancy status, comorbidities, treatment modality, skin characteristics when relevant, and care settings should be reported to permit assessment of generalizability.
Conclusions
This study proposes an exploratory six-domain framework for translating CGM performance characteristics and derived metrics into clinically interpretable constructs. The framework distinguishes upstream device-performance attributes from downstream glycemic summaries and clinical or implementation-related outcomes. The targeted evidence search identified direct device-level comparisons most clearly for analytical accuracy and glycemic exposure, while randomized evidence comparing CGM modalities informed the clinical modification of glycemic variability. Evidence concerning clinical safety, adherence, and technical reliability remained predominantly observational, modeled, device-specific, or contextual. The expert consensus adds a structured interpretive perspective but should not be considered a substitute for comparative empirical evidence.
iCGM designation indicates that a device has satisfied defined regulatory special controls, but it does not independently establish superior clinical outcomes or user experience. The clinical meaning of CGM data may be influenced by individual device characteristics, including range-specific and dynamic measurement performance, algorithms, sampling and data-processing procedures, monitoring continuity, connectivity, alarm functionality and performance when applicable, and user interaction. Clinicians should therefore interpret CGM metrics using device-specific evidence and information about how performance data were generated rather than assuming equivalence or superiority based solely on regulatory classification.
The proposed framework should be regarded as hypothesis-generating and adaptable. Future studies should use standardized comparator methods and study procedures, conduct purpose-designed comparisons of individual devices irrespective of regulatory designation, and report the proposed minimum dataset across clinically relevant glucose ranges, dynamic conditions, populations, and care settings. Such evidence is needed to determine which device characteristics produce reproducible differences in clinical interpretation, therapeutic decision-making, safety, patient experience, and health outcomes.
Supplemental Material
Supplemental material for Translating CGM Regulatory Performance Metrics into Clinical Interpretation: An Exploratory Expert Consensus Framework by Alex Ramírez-Rincón, Ana M. Gómez, Carlos A. Yepes-Cortés, Karen Feriz-Bonelo, Alejandro Marín-Sánchez, and Carlos E. Builes-Montaño in Clinical Medicine Insights: Endocrinology and Diabetes.
Acknowledgements
Festina Lente provided methodological support and medical writing and editorial assistance. The authors retained complete control over the design, evidence selection, analysis, interpretation, conclusions, and decision to submit the manuscript and take full responsibility for its content.
Author Contributions: Alex Ramírez-Rincón: Conceptualization; Methodology; Investigation; Data curation; Validation; Writing – review and editing.
Ana M. Gómez: Investigation; Validation; Writing – review and editing.
Carlos A. Yepes-Cortés: Investigation; Validation; Writing – review and editing.
Karen Feriz-Bonelo: Investigation; Validation; Writing – review and editing.
Alejandro Marín-Sánchez: Investigation; Validation; Writing – review and editing.
Carlos E. Builes-Montaño: Conceptualization; Methodology; Formal analysis; Investigation; Project administration; Supervision; Writing – original draft; Writing – review and editing.
All authors participated as subject-matter experts in the construction and refinement of the analytical framework, critical appraisal of the regulatory and scientific evidence, interpretation of the findings, and revision of the manuscript. All authors approved the final manuscript and accept responsibility for the integrity of the work.
Funding: The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Asociación Colombiana de Endocrinología, Diabetes y Metabolismo through an unrestricted educational grant from Abbott. The funder had no role in study conception, evidence selection, expert voting, statement development, data analysis, interpretation, manuscript-drafting decisions, or the decision to submit the manuscript for publication.
The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Carlos E. Builes-Montaño has received consulting and speaking fees from Sanofi, Novo Nordisk, Novartis, and Boehringer Ingelheim and is a shareholder of Festina Lente. All other authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Declaration of Generative AI and AI-Assisted Technologies: OpenAI and Grammarly were used to support language and style review and refinement of analytical programming code. These tools were not used to generate study data, make methodological or consensus decisions, or autonomously interpret the findings. All AI-assisted outputs were reviewed and verified by the authors, who take full responsibility for the accuracy and integrity of the manuscript.
Supplemental Material: Supplemental material for this article is available online.
ORCID iD
Carlos E. Builes-Montaño https://orcid.org/0000-0002-2418-6159
Ethical Considerations
This study comprised a targeted literature assessment and a professional expert-consensus exercise involving the coauthors. It did not enroll patients, use patient-level clinical data, involve animals, or include any clinical intervention. Institutional Review Board or Research Ethics Committee approval was therefore not sought.
Consent to Participate
All panelists participated voluntarily and agreed to the use of the aggregated consensus results.
Consent for Publication
The manuscript does not contain identifiable individual-level participant information, patient images, videos, or case data.
Data Availability Statement
The full search strategies are provided in the Supplementary Material. Deidentified statement-level consensus data, aggregated results, and analysis scripts are available from the corresponding author upon reasonable request, subject to protection of panelist confidentiality.*
References
- 1.Klonoff DC, Gabbay M, Moon SJ, Wilmot EG. Importance of FDA-Integrated Continuous Glucose Monitors to Ensure Accuracy of Continuous Glucose Monitoring. Journal of diabetes science and technology. 2025;19(5):1392-1399. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pleus S, Eichenlaub M, Dabla PK, et al. Clinical assessment and acceptance criteria for continuous glucose monitoring (CGM) system performance: A proposed guideline by the IFCC Working Group on CGM. Clin Chim Acta. 2026;580:120728. [DOI] [PubMed] [Google Scholar]
- 3.Mathieu C, Irace C, Wilmot EG, et al. Minimum expectations for market authorization of continuous glucose monitoring devices in Europe-'eCGM' compliance status. Diabetes, obesity & metabolism. 2025;27(3):1025-1031. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Scharf M, Feriz K, Yepes C, et al. Consensus statement on standardizing CGM evaluation metrics in Latin America: an expert approach. Diabetol Metab Syndr. 2025;17(1):291. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Nair R, Aggarwal R, Khanna D. Methods of formal consensus in classification/diagnostic criteria and guideline development. Semin Arthritis Rheum. 2011;41(2):95-105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Harb SI, Tao L, Peláez S, Boruff J, Rice DB, Shrier I. Methodological options of the nominal group technique for survey item elicitation in health research: A scoping review. J Clin Epidemiol. 2021;139:140-148. [DOI] [PubMed] [Google Scholar]
- 7.Fitch K, Bernstein SJ, Aguilar MD, et al. RAND/UCLA appropriateness method user's manual. Santa Monica, CA: RAND corporation; 2000. [Google Scholar]
- 8.Freckmann G, Link M, Pleus S, Westhoff A, Kamecke U, Haug C. Measurement Performance of Two Continuous Tissue Glucose Monitoring Systems Intended for Replacement of Blood Glucose Monitoring. Diabetes technology & therapeutics. 2018;20(8):541-549. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Welsh JB, Gao P, Derdzinski M, et al. Accuracy, Utilization, and Effectiveness Comparisons of Different Continuous Glucose Monitoring Systems. Diabetes technology & therapeutics. 2019;21(3):128-132. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Maytham K, Hagelqvist PG, Engberg S, et al. Accuracy of continuous glucose monitoring during exercise-related hypoglycemia in individuals with type 1 diabetes. Front Endocrinol (Lausanne). 2024;15:1352829. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Freckmann G, Wehrstedt S, Eichenlaub M, et al. A Comparative Analysis of Glycemic Metrics Derived From Three Continuous Glucose Monitoring Systems. Diabetes care. 2025;48(7):1213-1217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Jávorfi T, Kocsis G, Svébis MM, et al. Glucose management indicator: Do we need device-specific equations? Diabetes Metab. 2025;51(4):101661. [DOI] [PubMed] [Google Scholar]
- 13.Avari P, Moscardo V, Jugnee N, Oliver N, Reddy M. Glycemic Variability and Hypoglycemic Excursions With Continuous Glucose Monitoring Compared to Intermittently Scanned Continuous Glucose Monitoring in Adults With Highest Risk Type 1 Diabetes. Journal of diabetes science and technology. 2020;14(3):567-574. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Gutiérrez-Pastor A, Quesada JA, Soler-Martínez MM, Carratalá Munuera C, Pomares-Gómez FJ. Effect of switch from flash glucose monitoring to flash glucose monitoring with real-time alarms on hypoglycaemia in people with type 1 diabetes mellitus. Prim Care Diabetes. 2024;18(3):333-339. [DOI] [PubMed] [Google Scholar]
- 15.Markov AM, Krutilova P, Cedeno AE, McGill JB, McKee AM. Interruption of Continuous Glucose Monitoring: Frequency and Adverse Consequences. Journal of diabetes science and technology. 2024;18(5):1096-1101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Ehrmann D, Olesen B, Roos T, Kulzer B, Hermanns N, Heinemann L. Real-Life Wear Time and Reasons for Reduced Wear Time of Glucose Sensors of Continuous Glucose Monitoring Systems: Findings from the DiaLink Panel. Journal of diabetes science and technology. 2025;19(2):584-586. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Visser MM, Van Muylder A, Charleer S, et al. Cost-utility analysis of Dexcom G6 real-time continuous glucose monitoring versus FreeStyle Libre 1 intermittently scanned continuous glucose monitoring in adults with type 1 diabetes in Belgium. Diabetologia. 2024;67(4):650-662. [DOI] [PubMed] [Google Scholar]
- 18.DeSalvo D, Buckingham B. Continuous glucose monitoring: current use and future directions. Curr Diab Rep. 2013;13(5):657-662. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Lind M, Polonsky W, Hirsch IB, et al. Continuous Glucose Monitoring vs Conventional Therapy for Glycemic Control in Adults With Type 1 Diabetes Treated With Multiple Daily Insulin Injections: The GOLD Randomized Clinical Trial. JAMA : the journal of the American Medical Association. 2017;317(4):379-387. [DOI] [PubMed] [Google Scholar]
- 21.Stone JY, Bailey TS. Benefits and limitations of continuous glucose monitoring in type 1 diabetes. Expert Rev Endocrinol Metab. 2020;15(1):41-49. [DOI] [PubMed] [Google Scholar]
- 22.Montaser E, Farhy LS, Rich SS. Enhancing Type 1 Diabetes Immunological Risk Prediction with Continuous Glucose Monitoring and Genetic Profiling. Diabetes technology & therapeutics. 2025;27(4):292-300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Montaser E, Shah VN. Prediction of Incident Diabetic Retinopathy in Adults With Type 1 Diabetes Using Machine Learning Approach: An Exploratory Study. Journal of diabetes science and technology. 2026;20(3):743-750. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Pemberton JS, Andrews RC, Barnard-Kelly K, et al. International clinical opinion on transparency, standardisation, and calibration alignment in the performance evaluation of systems for continuous glucose monitoring. Diabetes, obesity & metabolism. 2026;28(4):2551-2565. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Belani JD. Interpreting Continuous Glucose Monitoring Through Metabolic Physiology: A Framework for Precision Nutrition in Type 2 Diabetes. Nutrients. 2026;18(15):2578. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental material for Translating CGM Regulatory Performance Metrics into Clinical Interpretation: An Exploratory Expert Consensus Framework by Alex Ramírez-Rincón, Ana M. Gómez, Carlos A. Yepes-Cortés, Karen Feriz-Bonelo, Alejandro Marín-Sánchez, and Carlos E. Builes-Montaño in Clinical Medicine Insights: Endocrinology and Diabetes.
Data Availability Statement
The full search strategies are provided in the Supplementary Material. Deidentified statement-level consensus data, aggregated results, and analysis scripts are available from the corresponding author upon reasonable request, subject to protection of panelist confidentiality.*