The standardized and clinical metrics (Table 1) comprise a core set of measures that have achieved widespread consensus and adoption. These metrics form the foundation of modern diabetes management, prioritizing simplicity, actionability, and interpretability for real-world diabetes management. They are designed to be easily understood by both health care professionals and people with diabetes, optimized for rapid interpretation during patient visits to enable real-time clinical decision-making. Overall, these metrics align well with established diabetes management practices and can be readily incorporated into existing clinical workflows and electronic health records. Standardized and clinical CGM metrics can be broadly categorized into several distinct categories described below, each serving specific purposes in diabetes care.

The ambulatory glucose profile (AGP) provides a standardized, single-page visual report of CGM data [18,22], summarizing key metrics, including composite indices like the glucose management indicator (GMI), facilitating the clinical interpretation of overall glycemic quality [17,58]. Daily glucose profiles are now typically presented in the AGP for several days [12,63] to aid in identifying typical 24-hour glucose patterns across treatment periods [18,63].

These metrics quantify the percentage of time spent within defined glucose thresholds [18], and are widely used in clinical care [18,45]. Key metrics include TIR, time below range TBR; low <70 mg/dL and very low <54 mg/dL, and time above range TAR; high >180 mg/dL and very high >250 mg/dL [18]. A primary limitation is the inherent bias toward hyperglycemia due to the asymmetrical distribution of blood glucose values, which gives disproportionately heavier weight to high excursions in statistical calculations [29]. Consequently, metrics impacted by both low and high measures of glycemia, such as TIR, may lack sensitivity to changes in low values, such as longitudinal improvements in TBR. This limitation highlights the inadequacy of time-based measures in capturing the full clinical risk associated with hypo- or hyperglycemia as compared with composite indices [58].

These metrics characterize overall and long-term glycemic exposure, reflecting average glucose and stability over periods typically ranging from weeks to months [23]. These measurements include mean glucose, glucose variability (commonly %CV), and the estimated average measure GMI. This category also includes critical data sufficiency metrics such as percent wear time and days of wear. A significant limitation of summary metrics, like mean glucose and GMI, is that they do not reflect glucose fluctuations or glycemic variability [22,29,45,63], meaning significantly varied glucose profiles can yield identical mean values [63]. GMI often shows clinically significant discordance with measured HbA_1c (≥0.5%, 6 mmol/mol) due to nonglycemic factors influencing red blood cell turnover, reinforcing caution against interpreting GMI as interchangeable with HbA_1c [18,30,31]. Studies introduced an updated calculation of the GMI, known as the updated glucose management indicator (uGMI). The uGMI is derived from a kinetic model, and it is the expected HbA_1c under a steady glucose level [32]. Furthermore, estimates of glycemic variability often have a larger relative error compared with estimates of mean glucose [23] due to their sensitivity to noise and extreme values.

These metrics focus on discrete episodes of acute glycemic extremes, providing actionable data crucial for assessing patient safety and guiding immediate therapeutic adjustments [17,18,33,63]. Unlike time-based metrics, these metrics include quantifying dangerous excursions such as hypoglycemic events (levels 1 and 2, eg, <70 mg/dL and <54 mg/dL) and hyperglycemic events (levels 1 and 2, eg, >180 mg/dL and >250 mg/dL) [18], alongside critical metabolic incidents (eg, euglycemic diabetic ketoacidosis). A primary limitation of event-based metrics is that CGM accuracy is often lower in the hypoglycemic range [17,18,34]. Also, the precise calculation of the number of events requires strict adherence to standardized definitions regarding the specific glucose thresholds and the required minimum duration (≥15 min) that must be maintained below or above the threshold [45], and subsequent events must be separated by a return to the target range for a specified time (eg, ≥15 min above 70 mg/dL) [17]. Without these standards, definitions can lead to over- or underestimation of event incidence [17]. For diagnosing acute events during periods of apparent euglycemia, such as euglycemic diabetic ketoacidosis, relying solely on glucose monitoring is insufficient, necessitating separate monitoring of metabolic acidosis and ketones [35]. Evaluating the frequency of hypoglycemic and hyperglycemic events is often considered more clinically meaningful than solely reporting the percentage of time spent in the respective ranges [45].

Mean average glucose excursions (MAGE) focuses on the minimum to maximum span of blood glucose levels and is widely considered the gold standard for assessing short-term glycemic variability [36]. It was developed as a measure to characterize glycemic variability specifically for postprandial glucose excursions above the mean glucose level. MAGE calculates the mean of glycemic excursions by identifying nadir to peak blood glucose levels. The amplitude from peak to nadir glucose averages that exceed 1 SD of glucose are included in the final MAGE calculation.

Previously, researchers evaluated MAGE at multiple 2-day periods within a 14-day CGM sensor lifespan, finding these equivalent to MAGE calculated from longer (3‐4 day) periods [36]. The study suggested that short-term periods are sufficient. MAGE lacks standardization in its calculation. There is no standard defining whether the amplitude should be calculated using ascending limbs, descending limbs, or a weighted average of both limbs. It requires specialized software programs for accurate calculations, limiting the metric’s ability to be used in real-time practice or integrated into electronic health records.

The continuous overall net glycemic action (CONGA) was originally developed for data from continuous interstitial glucose monitoring [37,38], and can be described as an index of the amount of time spent in glycemic excursions and the degree of glycemic variation that an individual experiences [37]. CONGA is particularly useful for helping clinicians and patients understand the magnitude and timing of blood glucose fluctuations in order to understand and optimize the patient’s degree of glycemic control. The normative reference range for CONGA (mean +/− 2 SD) for healthy patients is 3.6‐5.5, but may vary slightly by different ethnicities [39]. CONGA includes all data points, which means it may fail to effectively separate meaningful glycemic excursions from data noise [37].

The mean of daily differences (MODD) was originally developed as a measure of intraday glycemic variability [12]. MODD is calculated using pairs of blood glucose measurements from CGM that are 24 hours apart. MODD was originally introduced as a metric to be used to better understand the magnitude of glycemic variability through CGM data. MODD is highly correlated with other measures of variability, such as CONGA and SD. The average normative reference range for MODD for healthy patients is 0.0‐3.5 (SD 2), varying slightly for different ethnicities [39].

The lability index (LI) is designed to quantify glycemic variability by assessing fluctuation in glucose levels over time. It is particularly useful for identifying periods of significant glucose variability, which can be critical for managing diabetes effectively. The LI formula assesses variability in 3 consecutive glucose values at a time [4]. High LI indicates greater variability and potential instability in glucose levels, which can be associated with increased risks of hypoglycemia and hyperglycemia. Researchers compared the outcome of the HYPO score and the LI in patients living with type I diabetes who were undergoing islet transplant therapy [41]. The HYPO score is a measurement based on the frequency, severity, and degree of unawareness of a hypoglycemic event. Posttransplant patients experienced a significant reduction in the HYPO score and LI, indicating clinical relevance of LI as an objective measure for assessing the severity of glycemic lability [41]. The LI also had a high correlation with the metric, MAGE. This correlation makes it challenging to statistically isolate the unique and independent contribution of liability to long-term risk assessment [58]. The LI also requires specialized software to address complex calculation methods.

Standard CGM metrics are inadequate in representing the shape of glucose variability [42]. The glucose color index (GCI) is a new CGM metric that captures the shape of glucose variability. GCI uses periodogram signal decomposition. The metric was created through a multistep process involving the calculation of log-periodograms for each CGM time series, summarization using disjoint piece-wise linear models in 3 frequency regions, and refinement into a single value using canonical correlation analysis. The GCI is calculated by applying a linear combination of weights to the 6-parameter log-periodogram summaries, with these weights derived to maximize the test-retest correlation within participants over a 3-month period. Raw GCI values are then standardized to a scale of 0‐100 using a reference population for interpretability. In the study, the GCI was used to assess its reliability by comparing its 3-month test-retest correlation (R=.75) with standard CGM metrics and evaluate associations with diabetes comorbidities in older adults with type 2 diabetes [42]. The GCI showed significant associations (P<.05) with impaired physical functioning, frailty or prefrailty, cardiovascular disease, chronic kidney disease, and dementia or mild cognitive impairment after adjustment for confounders [42]. Overall, the study explains how the GCI captures clinically relevant information about glucose dysregulation that is not fully captured by standard CGM metrics. The GCI metric creates space for the development of CGM metrics that more fully incorporate time series information.

Overall, while metrics of glycemic variability provide granular insights into glucose stability and fluctuations that standard averages miss, their application is currently hindered by a lack of computational standardization and the frequent need for specialized software.

The glycemic risk index (GRI) aims to assist with basic clinical interpretation of CGM data. To develop the GRI, researchers compiled a dataset of 14-day CGM tracings from 225 insulin-treated adults with diabetes [58]. Using a balanced incomplete block design, 330 clinicians with extensive experience in CGM analysis and interpretation ranked these tracings based on the quality of glycemia [58]. Principal component analysis and multiple regressions were used to create a model predicting clinician rankings using the following 7 standard metrics from an AGP: very low-glucose and low-glucose hypoglycemia, very high-glucose and high-glucose hyperglycemia, TIR, mean glucose, and CV. The resulting GRI provides a single-number summary of glycemic quality, with its hypoglycemia and hyperglycemia components offering actionable scores. Additionally, the GRI grid, a graphical display, enables clinicians and researchers to visualize and determine the glycemic effects of prescribed and investigational treatments, enhancing its utility in both clinical practice and research settings [43]. The GRI is based on clinician rankings rather than clinical outcomes. Additional studies are necessary to determine how accurately GRI predicts long-term complications. Furthermore, the GRI does not inherently distinguish glycemic risks by time of day (eg, daytime vs nighttime), which is crucial for making precise therapeutic adjustments [58].

The M-value provides valuable insights into glycemic control and variability. It is defined as the difference between the observed blood sugar and normal blood sugar [37]. Lower M-values indicate better glycemic control, as demonstrated in a study comparing insulin glargine and insulin degludec in patients with type 1 diabetes using CGMs [44]. The researchers found that M-values were significantly smaller in patients treated with insulin degludec compared with those treated with insulin glargine, suggesting improved glycemic variability with insulin degludec. This study highlights the utility of the M-value in assessing the effectiveness of different insulin treatments and its role in evaluating glycemic control in clinical research [34]. A fundamental limitation is that the M-value is a hybrid measure that reflects both glucose variability and mean glycemia. It does not isolate variability itself [37]. Additionally, calculating the M-value requires complex calculation programs, which limits immediate application in acute settings like the ICU [46].

The J-index was proposed as an improvement over the existing M-value, integrating both the mean blood glucose (MBG) and SD of blood glucose to provide a comprehensive representation of glycemic status [59]. By incorporating these 2 components, the J-index effectively captures both the mean level and variability of glycemia, addressing key aspects of glucose control. Moreover, 1 study applied the J-index to evaluate glycemic control in a group of patients with type 1 diabetes undergoing intensive insulin therapy [59]. Researchers used the J-index to assess the effectiveness of different insulin regimens and found it to be a valuable tool for comparing glycemic control between patient groups and monitoring individual progress over time [24,46,60]. Similar studies demonstrated the practical application of the J-index in clinical settings and its potential to provide meaningful insights into glucose management strategies for diabetic patients. Similar to the M-value, the J-index requires the use of specialized software or programming, thereby hindering its immediate application in clinical care settings.

The GRADE score is an innovative metric designed to comprehensively evaluate glycemic control in patients with diabetes [22]. GRADE combines 2 risk quantities to summarize the overall risk associated with both hypoglycemia and hyperglycemia. The score’s calculation involves a log-log transformation, with risk increasing as glucose levels deviate from an assigned nadir of 90 mg/dL. This approach provides a clinically meaningful measure of glycemic risk, applying different weights to hyperglycemic and hypoglycemic values.

The development of GRADE stemmed from the increasing practice of multiple daily glucose assessments in patients with diabetes and the need for a more comprehensive evaluation of glycemic control [61]. To create this methodology, Hill et al [50] collaborated with 50 diabetes professionals who assigned risk values to a range of 40 blood glucose concentrations. From these responses, a generic function of glycemic risk was devised and applied to patient glucose profiles to generate an integrated risk score.

Interpreting GRADE scores is straightforward, with values below 5 corresponding to euglycemia. The simplicity of generating GRADE scores from any blood glucose profile makes it a valuable tool for clinicians. It can be used in conjunction with HbA_1c to provide a more comprehensive assessment of glycemic control, offering insights into the degree of risk associated with glycemic variability that HbA_1c alone cannot capture. The GRADE score does not directly measure glucose fluctuations. The derivation results in an expression that often reflects quasi-mean glycemia or a frequency distribution [37]. A high score can be generated by either hyperglycemia or hypoglycemia. Therefore, the formula does not explicitly specify the nature of the associated risk [37].

The average daily risk range (ADRR) was created to serve as a single metric to summarize blood glucose variability [51]. The ADRR is a diabetes-specific measure designed to assess the risk of both hyperglycemia and hypoglycemia. It is calculated using mathematically transformed glucose levels to give equal weight to high and low blood glucose excursions [52]. ADRR requires at least 14 days of blood glucose data with a minimum of 3 checks per day, although some studies have used fewer days. The metric is scored based on risk categories: low risk (0‐19), moderate risk (20-40), and high risk (40 and above) [62].

A study used ADRR to assess glycemic variability in children aged 2‐6 years with type 1 diabetes [33]. The researchers found that 72% of the children’s ADRR values fell into the “high-risk” range when using adult guidelines, suggesting a need for pediatric-specific ADRR standards [33]. The study also revealed that ADRR was highly correlated with indicators of hyperglycemia but only weakly correlated with measures of hypoglycemia in this young population [33]. Some articles in this review highlighted ADRR’s broader applications in clinical settings [62]. It emphasized ADRR’s ability to provide meaningful data on patients’ risk for both hyperglycemia and hypoglycemia, offering insights not available from HbA_1c values alone. The review also noted that research has shown ADRR to be a reliable predictor of extreme blood glucose values across different diabetes types and patient ages.

Low blood glucose index (LBGI) was developed specifically to measure the risk associated with low blood glucose events. There are four distinct risk categories of the LBGI: (1) minimal risk for hypoglycemia (LBGI≤1.1), (2) low risk (1.1<LBGI≤2.5), (3) moderate risk (2.5<LBGI≤5), and (4) high risk (LBGI>5) [29]. The LBGI has been demonstrated to predict severe hypoglycemia, while the high blood glucose index (HBGI) is associated with HbA_1c levels and predicting hyperglycemic excursions. Additionally, the ADRR serves as a measure of overall glucose variability, capturing the risk of both hypoglycemia and hyperglycemia.

LBGI's primary strength is its ability to quantify the frequency and extent of low blood glucose excursions, placing more emphasis on the hypoglycemic risk [44]. It has been shown to be a robust predictor of symptomatic hypoglycemic episodes [54]. Because LBGI correlates well with the occurrence of severe hypoglycemia, it is a valuable tool for identifying people at higher risk of hypoglycemia episodes. However, LBGI only focuses on hypoglycemia and does not provide information about overall glycemic control or hyperglycemic events. Additionally, interpreting LBGI values may require a higher level of clinical experience as well as clinical context to translate into meaningful treatment decisions.

High blood glucose index (HBGI) was designed to assess the risk associated with hyperglycemic events. Like LBGI, the HBGI quantifies glycemic variability focusing on hypoglycemic excursions. The cutoff points for the HBGI (4.5 and 9) are used to categorize the risk levels low, moderate, and high for hyperglycemia [56]. Risk scores above 0 are defined as HBGI [4]. In a study evaluating the improvement in glucose variability and its correlation with baseline HbA_1c and hypoglycemic events, the HBGI was found to have the largest proportional reduction (65.5%) among several glucose variability metrics after 24 weeks of treatment intensification. Furthermore, the HBGI was identified as the metric most predictive of improvements in HbA_1c levels [56].

This metric can be used to quantify the frequency and extent of high blood glucose excursions and can provide insights into the risk of hyperglycemia-related complications. HBGI has been shown to positively correlate with HbA_1c, making it useful for estimating overall glycemic control and potential long-term outcomes [57]. Similar to LBGI, a limitation of HBGI is that it provides a unidimensional view of glycemic control. Furthermore, the interpretation of HBGI values may not be intuitive for all people with diabetes and health care professionals, which may require additional education for effective use in clinical practice.

Index of glycemic control (IGC) assesses the overall quality of glycemic control [23,24]. It is calculated as the sum of the hypoglycemia index and the hyperglycemia index, assigning more weight to severe hypo- or hyperglycemic values [22,24]. One study concluded the IGC to be one of the most sensitive criteria for detecting responses to therapeutic intervention, such as the unmasking of data [24]. A disadvantage of using IGC as a composite measure is that it relies solely on hypoglycemia and hyperglycemia components, which possibly excludes other valuable metrics [24].

This metric is primarily based on a change in severe hypoglycemic events. Hypoglycemia score (HAS) aims to express overall glycemic control by simultaneously accounting for 2 key factors: the incidence of hypoglycemia and the change in HbA_1c [22]. HAS provides a method for comprehensive assessment and does not require complex data or equations. Its current framework uses subjective weightings and lacks CIs, meaning the results are arbitrary and there is no assurance of reproducibility from one observer to the next [22].

The Q-score was developed to automate objective evaluations of CGM profiles and serve as a screening tool to identify patients who require therapeutic intervention [60]. It is calculated using MBG, TBR, TAR, within-day variability, and MODD for between-day variability. Q-score values range from poor (≥12.0) to very good (<4.0) [22]. The examination of components that contribute to the Q-score allows clinicians to identify weak points in glycemic control. The Q-score is similar to a difficulty rating for a task. The score significantly increases with the complexity of treatment. A limitation of the Q-score is that there is no evidence on its use of MODD offering a better insight into glycemic variability than standard measures, such as SD or CV [60].

Personal glycemic state (PGS)is designed to provide a simple measure for quickly identifying problem areas in glycemic control [22]. It is calculated using TIR, mean glucose, glycemic variability percentage, and the number of hypoglycemic episodes at <70 mg/dL and <54 mg/dL thresholds [22]. The score is used to complement HbA_1c by measuring glycemic control. However, a limitation is that it does not include hyperglycemia as a component, and considers the frequency of glucose oscillations, making it partially a measure of frequency stability rather than exclusively magnitude [22].

The hypotriad uses 3 components within a 3D space model to assess the severity and risk associated with hypoglycemia. Its components are area under the curve (AUC), time spent in hypoglycemia, and the frequency of hypoglycemic episodes per day [62]. These core metrics are integrated to calculate two related composite metrics known as (1) the intensity of hypoglycemia (IntHypo), defined as a vector of AUC and time, and (2) the hypoglycemia risk volume, which is the product of IntHypo and the episode rate [62]. A major limitation of the hypotriad is that it focuses exclusively on hypoglycemia and does not include any components describing hyperglycemia, which is an important factor for assessing overall glycemic control [22].

The glucose pentagon and comprehensive glucose pentagon (CGP) are composite metrics designed to provide a simultaneous numerical and visual representation of glycemic control [22]. The original glucose pentagon integrated 5 axes, including MBG, metrics for hyperglycemic AUC, time in hyperglycemia, SD, and HbA_1c [22]. The CGP modified this by eliminating HbA_1c and introducing mean glucose, %CV, time out of range, the intensity of hyperglycemia (IntHyper), and the intensity of hypoglycemia (IntHypo), with %CV replacing SD to account for abnormal glucose distributions and independence from mean glucose. The core output of the CGP is the area of the resulting pentagon, which increases as glycemic control worsens relative to a person without diabetes (area normalized to 1) [22]. A key limitation is the complexity of the nonstandard metrics like IntHyper and IntHypo, which use quadratic averages of AUC and duration, making them difficult for most clinicians to visualize and interpret [22].

Continuous glucose monitoring index (COGI) evaluates the quality of glucose control in patients with type 1 diabetes [22]. It is calculated using 3 weighted components: TIR, TBR, and SD, with TIR carrying the most weight [22]. COGI provides a focused view on the 3 components, also allowing for %CV to be used in replacement of SD to measure variability. A potential limitation of COGI is that its component weighting was determined by expert opinion and it does not account for TBR very low (levels below 54 mg/dL) [22].

Overall, composite indices of glycemic quality and risk attempt to simplify multidimensional CGM data into unified risk scores for easier comparison, yet their widespread clinical adoption is limited by complex, nonstandardized definitions and a frequent reliance on expert opinion rather than hard clinical outcomes.