The Use of Continuous Glucose Monitoring to Diagnose Stage 2 Type 1 Diabetes

Abstract

This consensus report evaluates the potential role of continuous glucose monitoring (CGM) in screening for stage 2 type 1 diabetes (T1D). CGM offers a minimally invasive alternative to venous blood testing for detecting dysglycemia, facilitating early identification of at-risk individuals for confirmatory blood testing. A panel of experts reviewed current evidence and addressed key questions regarding CGM’s diagnostic accuracy and screening protocols. They concluded that while CGM cannot yet replace blood-based diagnostics, it holds promise as a screening tool that could lead to earlier, more effective intervention. Metrics such as time above range >140 mg/dL could indicate progression risk, and artificial intelligence (AI)-based modeling may enhance predictive capabilities. Further research is needed to establish CGM-based diagnostic criteria and refine screening strategies to improve T1D detection and intervention.

Introduction

Systems for continuous glucose monitoring (CGM) are increasingly being used by individuals with type 1 diabetes (T1D) instead of traditional blood glucose (BG) measurements or laboratory methods. ¹ Recently, a new therapy, teplizumab, has received US Food and Drug Administration (FDA) approval to help to delay the onset of symptomatic T1D. ¹ The FDA recommends that teplizumab should be used for stage 2 T1D, a stage in the natural history of progression of T1D which can be defined based on the presence of multiple islet autoantibodies and dysglycemia without overt hyperglycemia. ² In this expert consensus, we consider the evidence supporting the use of CGM as a screening test to identify people with stage 1 T1D at high risk of stage 2, as an adjunctive test leading to definitive venous blood testing. If the CGM tracing shows glucose values are abnormal, then the at-risk individual tested can undergo formal testing with measurements of venous glucose levels (either fasting or after an oral glucose load) or measurement of hemoglobin A_1c (HbA_1c). This use of CGM is intended to help overcome barriers to traditional approaches to screening for stage 2 T1D, such as transportation problems to reach a clinical laboratory or failure by children or others to tolerate regular phlebotomy. In this article, we discuss the current evidence about:

• which specific CGM measurements should be made to suggest a diagnosis of stage 2 T1D,

• available treatments if stage 2 T1D is diagnosed,

• why one should use CGM for screening and how often to repeat CGM testing,

• whether screening should be blinded,

• thresholds for CGM testing accuracy,

• the use of modeling tools using CGM data to predict stage 2 T1D,

• psychosocial aspects of CGM screening, and

• future research priorities in screening for this condition.

What are the Stages of Type 1 Diabetes?

• A family history of T1D or increased genetic risk alone can be called pre-stage 1 T1D.

• The three stages of T1D are characterized by the presence of two or more islet cell antibodies plus progressively higher glucose levels (normal, slightly elevated, and frankly elevated) and progressively greater hyperglycemic symptoms.

• Traditional tests for stages of T1D are performed by way of venous phlebotomy at a clinical laboratory and measure fasting glucose, 2-hour glucose during an oral glucose tolerance test (OGTT), or HbA_1c, but CGM measurements are not used for diagnosing diabetes, prediabetes, or any stage of T1D.

Autoimmune T1D is caused by immune-mediated destruction of pancreatic beta cells. As such, markers of the disease can be detected early in the natural history of the condition before the onset of symptoms. This allows for classification of T1D into 3 stages as presented in Table 1. ³ In stage 1 T1D, two or more types of diabetes-related autoantibodies are present, including antibodies against insulin (IAA), glutamic acid decarboxylase (GADA), insulinoma-associated protein 2 (IA-2, also called ICA512, islet cell antigen 512, or islet cell cytoplasmic antigen 512), and zinc transporter 8 (ZnT8A). ⁴ These individuals are asymptomatic and have normoglycemia. ⁴ In stage 2 T1D, two or more diabetes-related autoantibodies are present, and these individuals are asymptomatic, but have mildly hyperglycemic metrics, known as dysglycemia, but they lack laboratory findings of true hyperglycemia that would be diagnostic of diabetes or indicate a requirement for insulin therapy. ⁵ They have mild elevation of glycemic metrics in the prediabetes range, based on one of three venous blood tests: (1) elevated fasting plasma glucose (100-125 mg/dL or 5.55-6.94 mmol/L), (2) impaired glucose tolerance based on an elevated 2-hour plasma glucose during a 75-gram OGTT (140–199 mg/dL or 7.77-11.05 mmol/L), or (3) elevated HbA1c (5.7–6.4%, or 42-47 mmol/mol). ⁵ In stage 3 T1D, again two or more diabetes-related autoantibodies are present. These patients often have symptoms of hyperglycemia and laboratory evidence of definitely abnormal or frank hyperglycemia based on one of four venous blood tests: (1) elevated fasting plasma glucose (at least 126 mg/dL or 7 mmol/L), (2) a 2-hour plasma glucose of at least 200 mg/dL or 11.1 mmol/L during an OGTT, (3) a random plasma glucose level of at least 200 mg/dL or 11.1 mmol/L, or (4) an elevated HbA_1c of at least 6.5%. It is important to note that in the absence of symptoms, confirmation of stage 3 T1D typically is only made after confirmation with a second test on a different day. Currently, CGM measurements are not used for diagnosing diabetes, prediabetes, or any stage of T1D.

Table 1.

Evolution of the Risks and Stages of T1D.

Evolving risk for type 1 diabetes			The stages of type 1 diabetes
Genetic risk	Immune activation	Immune response	Stage 1	Stage 2	Stage 3
Risk is ↑ 8-15x if a first-degree relative has T1D	Beta cells are attacked	Single autoantibody is detected	Normal BG and ≥ 2 Ab	Dysglycemia and ≥ 2 Ab	Hyperglycemia and ≥ 2 Ab

Upon development of two or more islet cell autoantibodies without symptoms or abnormal glycemia (stage 1), the disease progresses to development of dysglycemia (stage 2) to overt diabetes values (ADA definition of diabetes), which can include insulin requirements, severe hyperglycemia, or diabetic ketoacidosis (stage 3).

Abbreviations: Ab, autoantibodies; BG, blood glucose; T1D, type 1 diabetes. Table modified with permission from Ayers et al. ³

Standard diagnostic testing for stages of T1D requires venous blood phlebotomy, and in some cases, a glucose tolerance test. These procedures are frequently poorly tolerated by patients and their families because of pain or poor vascular access from venous blood testing, anxiety or gastrointestinal symptoms from the large glucose load in an OGTT, and the time-consuming nature of an OGTT. ⁶ Traveling to a laboratory for testing can be a burden for many families. Furthermore, there is significant day-to-day variability in glucose profiles. Both OGTT and HbA_1c are also of unclear value in very young children. ⁷ The use of multiple approaches to testing can lead to conflicting results, and while OGTT is most often considered the gold standard, it is unclear which blood test is most indicative of progression to insulin requirement and which test correlates most closely to CGM testing.

What is the Burden of Stage 2 Type 1 Diabetes, at a Population and an Individual Level?

• In stage 1 T1D, the lifetime risk of evolution toward stage 3 T1D is nearly 100%.

• In stage 1 T1D, the five-year risk of developing stage 3 T1D is 44%. In stage 2 T1D, the five-year risk of developing stage 3 T1D is 75%.

• Failing to identify stage 2 T1D can lead to diabetic ketoacidosis (DKA) and hospitalization at the onset of stage 3 T1D.

It is important to understand the significance of the stages of T1D at a population level. One study in Germany that screened 154,462 children for islet autoantibodies found that 0.19% of children (293) had stage 1 T1D, and 0.019% of children (30) had stage 2 T1D. ⁸

In stage 1 T1D, the lifetime risk of evolution toward clinical stage 3 T1D approaches 100%, but the rate of progression differs. The 5-year and 10-year risks of developing symptomatic T1D average 44% and 70%, respectively. ⁵ In stage 2 T1D, the five-year and lifetime risks of developing symptomatic T1D average 75% and 100%, respectively. ⁹ The rate of progression is influenced by a greater burden of autoimmunity (the number and titers of islet cell autoantibodies), an early age of seroconversion, and a high genetic predisposition.^2,10 Beta-cell destruction is more rapid in children than in adults and can often be precipitated by intercurrent infection and other stressors.

Identification of T1D before stage 3 T1D is crucial because it not only allows physicians to implement therapies that could delay the onset of stage 3 T1D, but it also prevents potentially life-threatening outcomes such as DKA at the onset of stage 3 T1D. ³ DKA, often requiring hospital admission, is a major concern when the identification of T1D is delayed. If there is more frequent screening for the early stages of T1D, DKA could be avoided or greatly reduced.

What Treatments are Available if Stage 2 Type 1 Diabetes is Diagnosed?

• Teplizumab, a monoclonal antibody, is currently the only FDA-approved drug for delaying the onset of frank T1D if it is initiated during stage 2 T1D.

• Trials of immunomodulators, small molecules, and cell-based therapies to preserve beta-cell function and delay the onset of stage 3 T1D are underway.

• Diet and exercise regimens have also been associated with delayed onset of frank T1D in observational studies and anecdotal cases.

Identification of stage 2 T1D through screening is crucially important. Once stage 2 T1D is diagnosed, strategies to delay progression to stage 3 T1D, and the need for exogenous insulin have aimed to improve insulin sensitivity or preserve beta cell function. Much of the evidence supporting possible therapies comes from studies in newly diagnosed, stage 3 T1D, with very few interventions studied in robust clinical trials of people with stage 2 T1D. For example, based on studies in early-onset stage 3 T1D, it has been suggested that exercise could reduce insulin need and delay initiation of insulin therapy, ¹¹ but this has yet to be tested in stage 2 T1D. Observational data from The Environmental Determinants of Diabetes in the Young (TEDDY) study showed an association of more daily minutes of moderate-to-vigorous physical activity with reduced risk of progression to stage 3 T1D in children aged 5 to 15 years with multiple autoantibodies (hazard ratio [HR] = 0.92, 95% confidence interval [CI] = 0.86-0.99, per 10-min increase). ¹² In addition, diet changes have also been hypothesized to slow the onset of stage 3 T1D, based on observational data from the Diabetes Autoimmunity Study in the Young (DAISY), in which progression to stage 3 T1D was associated with higher dietary glycemic index (HR = 2.20, 95% CI = 1.17-4.15) ¹³ and higher total sugar intake (HR = 1.75, 95% CI = 1.07-2.85). ¹⁴ However, prospective clinical trials of physical activity or diet changes are needed to definitively understand the effect of these lifestyle modifications to slow progression to stage 3 T1D.

Several pharmacologic therapies that target the autoimmune response have been proposed to preserve beta cell function. Only one such agent is FDA-approved and recognized to delay progression to stage 3 T1D. Various immunotherapies have been tested in newly diagnosed stage 3 T1D, with varying capabilities to preserve beta cell function. Of these, teplizumab, a humanized anti-CD3 monoclonal antibody, is the only therapy tested in stage 2 T1D and shown to significantly delay the onset of stage 3 T1D. ¹⁵ In a randomized controlled trial of 76 people with stage 2 T1D (ages 8 and older), 44 received a 14-day course of teplizumab and 32 received a placebo. The median duration of follow-up was 742 days, with a median time of progression to stage 3 T1D of 48.4 months with teplizumab compared to 24.4 months with placebo (HR = 0.41, 95% CI = 0.22-0.78; P = .006). ¹⁵ In an extension of this cohort with a median duration of follow-up of 923 days, a median time of progression to stage 3 T1D was 59.6 months with teplizumab compared to 27.1 months with placebo (HR = 0.457, P = .01). ¹⁶ These results led the FDA to approve teplizumab in November 2022 to delay the onset of stage 3 T1D, and teplizumab to become the only pharmacologic agent recommended by the American Diabetes Association (ADA) for those with stage 2 T1D. ¹⁷ In a pre-clearance administration program, the medication is already available for treatment in more countries, eg, Germany. ¹⁸ Active clinical trials of other therapies, such as other immunomodulators, small molecules, and cell-based therapies to preserve beta cell function and delay the onset of stage 3 T1D are currently underway. Understanding CGM metrics in individuals with stage 2 T1D would be helpful to understand the natural progression of T1D and to monitor progression in such clinical trials.

Why Should One Use Continuous Glucose Monitoring for Screening, and How Often Should One Repeat Continuous Glucose Monitoring Testing?

• CGM screening of children for stage 2 T1D performed at home would decrease the need for screening by routine phlebotomy and often poorly tolerated glucose loads necessary for an OGTT that are currently performed at a clinical laboratory.

• For adults, CGM screening should be repeated every 6-12 months based on the presence of multiple antibodies, and every 12-36 months based on the presence of a single antibody or only a family history of T1D.

• For children with stage 1 T1D, CGM screening should be performed every three months if less than 3 years old, at least every six months if 3 to 9 years old, and at least every 12 months if greater than 9 years old.

CGM is a potential screening test to identify people with stage 1 T1D at high risk of stage 2, as an adjunctive test to traditional venous blood testing. If the CGM tracing shows glucose values are abnormal, then the at-risk individual tested can undergo formal testing with measurements of venous glucose levels (either fasting or after an oral glucose load) or measurement of HbA_1c. This use of CGM is intended to help overcome barriers to traditional approaches to screening for stage 2 T1D. In asymptomatic older children and young adults with positive diabetes-related autoantibodies, glucose profiles recorded by metabolic monitoring of interstitial fluid (ISF) by way of CGM screening have been shown to be useful in providing evidence of early dysglycemia ¹⁹ in individuals with normal fasting or 2-hour glucose OGTT results ²⁰ or individuals who are at increased risk of progression to stage 3 T1D.^21,22

In adults with risk for T1D, CGM screening should be repeated every 6 to 36 months depending on the number of antibodies, and whether there is a family history of T1D.^23,24 The more frequent monitoring schedule is intended for those with a higher immune burden and/or higher risk of progression. In adults, annual metabolic monitoring should also be considered for those with a single positive antibody if additional risk factors are present, including having a first-degree relative with T1D, elevated genetic risk for T1D if tested, impaired fasting glucose, impaired glucose tolerance, and/or if there is a history of stress hyperglycemia. For adults with only a single antibody and no other risk factors, metabolic status should be monitored every 36 months. ²⁵ Differences in glycemic control may be found across the menstrual cycle in women. ²⁶

For children with stage 1 T1D, metabolic monitoring should be performed every three months if less than 3 years old, at least every six months if 3 to 9 years old, and at least every 12 months if greater than 9 years old. ²³ The frequency will depend on the immune burden and/or risk of progression. ²⁷ For children who have multiple antibodies, education should also be provided in case symptoms and/or hyperglycemia develop. Importantly, CGM in this population must be interpreted by trained professionals, with education provided to the user and their family regarding the implications of the profiles and plans for future surveillance.

Guidance for the practical use of CGM as a tool has been published, and this is also appropriate for future studies of the use of CGM in presymptomatic individuals at risk of T1D. ²³ However, it is noteworthy that participants in previous studies of CGM in asymptomatic, antibody-positive children and adults have been overwhelmingly non-Hispanic white individuals. It remains to be determined whether the value of CGM for screening and monitoring is similar in those individuals who identify as other races and ethnicities, as in type 2 diabetes (T2D) and prediabetes, there appear to be discrepancies between estimated glycemia from HbA_1c values and glycemia assessed using CGM, by race and ethnicity. ²⁸

Should Continuous Glucose Monitoring Screening Be Blinded?

• CGM screening for stage 2 T1D may be conducted blinded or unblinded.

• Unblinded CGM might encourage a wearer to influence results of what is intended to be a retrospective screening test and not a test for treating glucose concentrations.

• A standardized home meal test can be developed to be used as part of a blinded CGM evaluation.

Since the results of the CGM monitoring period with screening for dysglycemia are interpreted retrospectively and not used for managing glycemia during wear, the preferred type of CGM monitoring is blinded which will not affect the wearer’s behavior. ²⁹ The drawback of unblinded CGM testing is that in seeing the results, the individual might change dietary or other behavior and thereby reduce sensitivity of the test. Prior studies have also emphasized the need to use blinded CGM because the collected information should not influence a patient’s behavior. ³⁰ The impact of real-time CGM, where users have immediate access to their glucose profiles, on trajectories of dysglycemia in populations being screened for T1D, has not been determined. Similarly, as the rate of progression within these stages is heterogeneous, the use of personalized risk calculators may help to improve clinical implementation. ³¹ Recently, studies of CGM used to screen for diabetes have suggested that the application of machine learning (ML), based on home use of CGM accompanied by standardized home meals, can be used to predict T1D risk.^32,33

As more data become available, it is likely that additional metrics derived from CGM profiles may shed additional light on the heterogeneity of risk and progression of T1D. Such metrics have been described that have the potential to allow stratification into subgroups of risk. ³⁴ Furthermore, as over-the-counter CGM devices become more ubiquitous, it is likely that professionals—both expert and non-expert—will encounter increasing numbers of individuals with concerns about their personal risk of developing T1D based on their CGM profiles in the absence of antibody screening.^35,36

What Level of Accuracy is Required for Continuous Glucose Monitoring to Become a Diagnostic Test for Stage 2 Type 1 Diabetes, Rather Than an Adjunctive Test?

• No CGM system is currently considered sufficiently accurate by a major professional organization or regulatory body to diagnose stage 2 T1D with robust specificity and sensitivity.

• The mean glucose concentration measured by a CGM, creates an estimated HbA_1c called the glucose management indicator (GMI), and this metric is not sufficiently accurate to diagnose stage 2 T1D.

• In setting cut points for CGM measurements as a screening test for stage 2T1D, specificity will take priority over sensitivity which will prioritize missed opportunities for diagnosis over excessive false alarms.

Accuracy of Diagnostic Tests

Given the importance of differentiating normal glycemia (stage 1), dysglycemia (stage 2), and overt diabetes (stage 3) for treatment decisions, both precise and accurate measures will be needed. For instance, if a person has the lowest fasting glucose value to establish stage 2 T1D (100 mg/dL), then the only way a CGM sensor could accurately differentiate stage 1 and stage 2 is if the lowest fasting value measured was 99.5 mg/dL (which would round up to 100 mg/dL). This edge case reading would require a relative error of ≤0.5% for fasting glucose. Therefore, it would be necessary to accept stringent CGM error measures as satisfactory to match venous blood testing. It should be noted, however, that no CGM system has been evaluated, and no accuracy criteria have been established to determine whether CGM can be used for this purpose.

A potential method using CGM for screening that has not been reported in the medical literature is to use a combination of CGM measurements for the time above euglycemia (percent of time > 140 mg/dL, or TAR140) along with a series of capillary BG tests for calibration. This combination could enhance the accuracy of the CGM measurements and increase detection accuracy using the percent of TAR140.

Glucose Management Indicator

The GMI estimates HbA_1c based on mean glucose readings obtained from CGM. It is important to note that the formula used to derive GMI was developed and validated using clinical trial data in patients with diabetes (456 with T1D and 72 with T2D).^{37 -40} In individuals without diabetes, GMI may overestimate HbA_1c. ⁴¹ The GMI can be an inaccurate predictor of HbA_1c, particularly at lower HbA_1c concentrations where the cut-off for diagnosing diabetes would occur. ⁴² Moreover, even among studies involving patients with diabetes, there has been significant discordance of ≥0.5% between GMI and HbA_1c, with discrepancies reported in 26 to 68% of cases. ⁴³ While longitudinal data collection may help reduce individual variations, changes in HbA_1c and changes in GMI are only moderately correlated in individuals with chronic kidney disease (CKD), for example. ⁴⁴ Consequently, GMI should not be relied upon to determine glycemic categories. ⁴³

Accommodating Error

Error will always exist in measurement; the question is how much is tolerable. The direction of sensor bias appears to vary by manufacturer and comparator and could misclassify stage 2 T1D as a result. ⁴⁵ As dysglycemia progresses, this bias could eventually be picked up with continued wear, but this approach also carries the cost and burden of repeated wear for someone who may have normal glycemia.

Given the current accuracy of CGM, it might be necessary to prioritize specificity to avoid unnecessary repeat diagnostic testing at the price of missing some cases of stage 2 T1D that will eventually be discovered through regular blood testing. This approach is probably preferable to prioritizing sensitivity in order to avoid missing a case of early stage 2 T1D, especially when it would result in excessively frequent, often misleading, and usually inconvenient repeat phlebotomy testing at a laboratory—particularly for children, who are often averse to phlebotomy.

Which Continuous Glucose Monitoring Systems are Suitable for Diagnosis of Stage 2 Type 1 Diabetes? Which Performance Parameters Must Be Met by the Continuous Glucose Monitoring System?

• Diagnostic thresholds based on venous blood cannot be exactly transferred to CGM readings because of differences in blood and ISF glucose, and calibration differences of CGM systems.

• CGM performance testing requires comprehensive guidelines for study procedures, and in particular, the collection and quality of comparator data.

• Accuracy of current CGM systems must be confirmed by additional research to screen for this purpose, and no regulatory clearance is available to use only CGM for screening.

Current methods used for defining dysglycemia and diagnosing diabetes including stage 2 T1D rely on the measurement of venous HbA_1c and plasma glucose. ² In contrast, CGM systems measure glucose levels in the ISF in subcutaneous tissue. CGM readings are often in between capillary and venous glucose concentrations. Because of physiologic differences in glucose concentrations, established diagnostic thresholds based on venous blood cannot be exactly transferred to CGM readings. In addition, CGM systems lack standardization, leading to variability of CGM readings between systems. Currently, there are no quality control measures available for CGM users apart from parallel BG monitoring.

The accuracy of available CGM systems has increased over recent years, but even current-generation devices can show considerable discrepancies with each other. A head-to-head study comparing two flagship integrated CGM (iCGM) systems last year reported a mean bias of nearly 10% against each other. ⁴⁶ However, a more recent study could not reproduce this large difference. ⁴⁵ An important factor leading to discrepancies between different CGM systems is the definition of the “true” glucose concentrations that are used as targets when CGM systems are (factory-)calibrated. There is currently no standard that defines a “reference,” i.e., glucose concentration in a designated compartment (venous or capillary blood) and a sufficiently accurate device to measure it. ⁴⁷

This does not generally mean that CGM systems might not be used for stage 2 T1D diagnosis or that no current CGM systems would be suitable; however, systematic differences (biases) between different CGM systems’ readings can lead to discrepant CGM-based metrics, such as time in different glucose ranges.^{48 -51} If diagnostic thresholds based on absolute glucose concentrations, such as the percentage of time at CGM values of TAR140,^3,23 are used, then current state-of-the-art CGM systems could provide divergent diagnoses. Most studies investigating the relationship between CGM parameters and the precursor states of frank T1D have used older generations of a given CGM system. ³

Similar to discussions about device-specific GMI equations in the last few years,^43,52,53 device-specific diagnostic thresholds would need to be established or diagnostic thresholds independent from absolute glucose concentrations. A clear definition of performance criteria, including a maximum permitted bias of CGM systems, may be required irrespective of whether the system is used for diagnoses or therapy. ⁵⁴

Which Continuous Glucose Monitoring Metrics Indicate Normoglycemia?

• Time in euglycemia is the percentage of time spent between 70 and 140 mg/dL.

• Populations defined as being normoglycemic have usually reported less than 10% time spent above euglycemia and almost no time at all in percentage time above 180 mg/dL (TAR180).

• Time in euglycemia decreases in persons over 60 years old.

Studies Defining Normoglycemia

A number of studies have identified CGM metrics that indicate normoglycemia. However, there is no agreement on what constitutes a normal CGM report. Table 2 shows CGM metrics reported in populations identified as having normoglycemia using blinded CGM. It is important to note that definitions of normoglycemia varied considerably across these studies. Many of these studies assessed the percentage of time >140 mg/dL or 7.8 mmol/L, which is known as TAR140.

Table 2.

Populations Defined as Having Normoglycemia^*.

Reference	Participants (n) country	Age and BMI	CGM System	<70 mg/dL (% time) [IQR]	70-140 mg/dL (% time) [IQR]	TAR >140 mg/dL (% time) [IQR]	TAR >180 mg/dL (% time) [IQR]	Mean Glucose (±SD), [IQR], min-max
Rodriguez-Segada et al.55a	254 Spain	≥18yo, 42yo, BMI 26.3	iPro2, 7d, calibrate TID	0 [0,1.2]	97.4 [94.2,99.4]	1.2 [0.2, 3.7]	–	102.6±7.2
Shah et al.56b	153 T1D Exchange	≥6 yo, BMI <30 if ≥18 yo	DexCom G6 (Blind), 10d, calibrate QD	1.1 [0.3,2.9]	96 [93, 98]	2.1 [0.9, 3.9]	0 [0, 0.2]	99±7
Shah et al. 56 (subset) b	26 T1D Exchange	≥60 yo	DexCom G6 (Blind), 10d, calibrate QD	1.4 [0.2, 3.4]	93 [89, 96]	4.1 [1.3, 8.6]	0.1 [0.0.5]	104±9
Dimova et al.57c	33 Bulgaria	30-65 yo, 45.2 yo, BMI 29.9	FreeStyle Libre Pro, 14d	–	99 [97,100]**	1 [0,2]	–	86 [85, 92]
Barua et al.58b	32 Hispanic	≥18 yo, 54.6 yo, BMI 31.2	FreeStyle Libre Pro, 14d	2 [0.3, 4.4]	95.5 [91.7, 97.4]	2.5	0 [0,0]	99.5 [91.8, 105.1]
Rizos et al.59a	18 Greece	37-74 yo, 49.4 yo, BMI 26.9	Medtronic EnvisonPro, 7d	0	96.8	2.8	0	108.8 (min-max, 93.5-115.5)
Spartano et al.60a	560 Framingham, USA	58.8 yo, BMI 26.5	DexCom G6 Pro, ≥7d	0.9	86.8	12.3	1.3	114.5±11.5 111.6 [103.8,119.3]
Spartano et al. 60 (subset) a	223 Framingham, USA	>60 yo, 51.9 yo, BMI<30, BMI 24.7	DexCom G6 Pro, ≥7d	0.7	89	10.3	1	112.7±10.7 109.9 [102.8,116.8]

Abbreviations: BMI, body mass index; CGM, continuous glucose monitoring; d, days; mg/dL, milligrams per deciliter; QD, each day; SD, standard deviation; TAR, time above range; TID, three times a day; T1D, type 1 diabetes; yo, years old.

^*Defined as having normoglycemia by ^aboth FPG <100 mg/dL and HbA_1c <5.7%, ^bonly HbA_1c <5.7%, and ^cOGTT (FPG <110 mg/dL, 2hPG <140 mg/dL).

^**Target range defined as 55 to 140 mg/dL. Data reported as mean, or when italicized, median [interquartile range], unless otherwise noted.

Glucose Thresholds

Nearly every study in Table 2 defined CGM-based glucose threshold differently, although most reported a percent TAR140, and some reported a percent of time above range for glucose levels >180 mg/dL (TAR180). While TAR180 occurred infrequently in populations with normoglycemia, TAR140 varied considerably depending on the population. Older individuals exhibited a higher percentage of TAR140. For instance, Spartano et al ⁶⁰ investigated the grandchildren of participants in the original Framingham Heart Study, with a mean age of 58.8 years. Even when stratified to include only those under 60 years of age and non-obese (body mass index [BMI] <30 kg/m²), the percent TAR140 was found to be 10%. In contrast, younger cohorts shown in Table 2 demonstrated a median TAR140 of less than 5%.^{55 -59} Although glucose levels increase with age, if TAR140 is utilized as a criterion to avoid laboratory glucose testing, then lower cut points for TAR140 (<5%) should be considered, particularly for individuals under 40 years of age.

Carbohydrate Intake

While an OGTT involves the consumption of a standardized 75 g of glucose, carbohydrate intake is not standardized when using CGM. Studies have been conducted in home settings with participants instructed to consume the same 75 g of glucose as in an OGTT ³³ or standardized liquid meal replacements, such as Boost. ³² It is important to note that when CGM systems are worn during an OGTT, the interstitial glucose levels following glucose ingestion are often higher than the corresponding plasma glucose levels.

Exceeding Normoglycemia

Another approach besides TAR to defining stage 2 T1D could be an elevated mean ISF glucose concentration (above normoglycemia) as measured by CGM. Mean glucose levels during CGM monitoring varied among populations classified as having normoglycemia, as shown in Table 2. Younger cohorts had average glucose levels <110 mg/dL (<6.105 mmol/L). In the Framingham study, a mean glucose level of <117 mg/dL (<6.494 mmol/L)—corresponding to an HbA_1c of 5.7%—included 75% of participants in that study who were <60 years old and non-obese. Thus, a mean glucose of 117 mg/dL (6.94 mmol/L) during CGM use could be considered the upper limit for individuals classified as having normoglycemia.

Further studies are needed in healthy populations that are rigorously defined as having normoglycemia using both HbA_1c and OGTT results to establish CGM criteria for normoglycemia. In populations at risk for T1D, particularly among younger individuals, stricter CGM criteria for normoglycemia should be utilized if the goal is to minimize reliance on laboratory testing. Established and not established (potential) criteria to define normoglycemia are presented in Table 3.

Table 3.

Established and Not Established (Potential) Criteria to Define Normoglycemia.

	Normal
HbA1c (%)	<5.7
OGTT (mg/dL)
FPG	<100
2hPG	<140
CGM (mg/dL) a
Fasting Glucose	<100
Average Glucose	<117
TAR >140	<5%
TAR >180	0%

Abbreviations: CGM, continuous glucose monitor; FPG, fasting plasma glucose; HbA_1c, hemoglobin A_1c; OGTT, oral glucose tolerance test; TAR, time above range; TBR, time below range; 2-h PG, 2-hour post-load glucose.

^aNot established.

What are the Cut-offs for Continuous Glucose Monitoring Metrics in Individuals at Risk of Type 1 Diabetes With Dysglycemia?

• An advantage of using CGM technology compared to traditional BG testing is its ability to fully capture postprandial glucose excursions.

• An elevated percentage of TAR140 (ie, above 5%-10%) is a good metric for assessing the risk of stage 2 T1D, but more evidence is needed.

• An elevated mean ISF glucose concentration is an additional potential diagnostic target for screening.

Determining a test and its cut-off should depend on benefits and risks (medical, social, and economic). Hence, diagnosis of stage 2 T1D should be defined based on: (1) risk (probability) of developing T1D, (2) clinical effectiveness of a potential intervention that may change the course of the disease, (3) cost-effectiveness of the test, and (4) potential harms that tests may cause to an individual being diagnosed with the condition and risks associated with a false positive test.

To our knowledge, none of the published studies using different variables (either glucose-based, C-peptide, or mathematical modeling using various clinical data) to predict progression from stage 2 to stage 3 T1D have evaluated these four points together. Moreover, just as CGM is not generally recommended to diagnose prediabetes, CGM is currently also not used to define or screen for stage 2 T1D. ⁶¹ The definition of prediabetes was based on the eventual development of type 2 diabetes and risk for retinopathy progression, but defining stage 2 T1D should be aimed at prognosticating development of stage 3 T1D (and not the risk for retinopathy) and the opportunity to intervene at stage 2 T1D that may alter the disease course.

Studies Defining Dysglycemia

Many studies have evaluated the utility of CGM-based metrics for progression of stage 2 to stage 3 T1D, and these have been summarized in a recent publication. ³ The most consistent observation in these studies is that the higher the TAR, the higher the likelihood of developing stage 3 disease. Hence, defining lower cut-off percentages of time above any range (eg, >5% instead of >10%) would increase sensitivity and decrease specificity, whereas increasing the cut-off to 10% or higher would be associated with higher specificity and lower sensitivity (see Table 4). An elevated percentage of time above 140 mg/dL (ie, above 5%-10%) appears to be a good metric for assessing the risk of stage 2 T1D, but more evidence is needed.

Table 4.

CGM Metrics to Predict the Course of Progression Toward Stage 3 T1D.

First author	Year	Subject demographics	Conclusions
Steck et al. 62	2014	N = 14 IAb+ children and nine IAb– children, matched for age and sex. Mean age = 13.7	IAb+ progressors spent 31% of time above 140 mg/dL compared with 12% of time in IAb+ nonprogressors (P = 0.04)
Van Dalem et al. 21	2015	N = 22 IAb+ participants with FDRs with T1D. Control groups: N = 20 age-matched healthy volunteers and N = 9 individuals with T1D. Mean age = 18.8	CGM measures above reference ranges, compared with elevated SMBG measures from healthy controls, were better at detecting impending dysglycemia (77%-82% vs 73% diagnostic efficiency)
Helminen et al. 63	2016	N = 10 IAb+ children and N = 10 IAb– controls matched for age and sex	Time spent ≥ 140 mg/dL was 5.8% in IAb+ (two or more IAbs) individuals compared with 0.4% in IAb– individuals (P = 0.04)
Steck et al. 64	2019	N = 23 IAb+ children. Mean age = 15.7	≥ 18% of CGM time spent at 140 mg/dL predicts progression to diabetes in IAb+ children, with 75% sensitivity, 100% specificity, and a 100% predictive value for diabetes prediction
Kontola et al. 19	2022	N = 46 . Mean age = 11.7 (range: 3.9-25.4)	The mean percentage of time spent in the 70 to 140 mg/dL range decreased progressively from 94% ± 2.7% to 68% in subjects with 0 IAb to S3 T1D
Steck et al. 27	2022	N = 91 IAb+ children. Median age = 11.5	Progressors spent 21% of time > 140 mg/dL and 8% of time > 160 mg/dL, compared with 3% and 1% for nonprogressors (P < .0001)
Wilson et al. 65	2023	N = 105 relatives of people with T1D. Median age = 16.8 years	Spending ≥ 5% time with glucose levels ≥ 140 mg/dL (P = 0.01), ≥ 8% time with glucose levels ≥ 140 mg/dL (P = 0.02), ≥ 5% time with glucose levels ≥ 160 mg/dL (P = 0.0001), and ≥ 8% time with glucose levels ≥ 160 mg/dL (P = 0.02) were all associated with progression to S3
Ylescupidez et al. 66	2023	N = 93, all with ≥ 2 IAb	CGM metrics with the highest AUC for prediction of subsequent T1D included mean average glucose, percent TIR (60-140 mg/dL), GRADE, and one measure of variability
Calhoun et al. 67	2023	N = 218 IAb+ participants. Median age = 14 (range: 2-50), 61% participants with FDR T1D.	CGM-only model selected time >140 mg/dL achieved an AUROC of 0.70. The AUROC of a full model (CGM metrics and participant characteristics) for predicting stage 3 T1D by year 2 was 0.79

Abbreviations: AUC, area under the curve; AUROC, area under the receiver operating characteristic curve; CGM, continuous glucose monitoring; IAb+, positive islet autoantibodies; IAb–, negative islet autoantibodies; milligram/deciliter, mg/dL; SMBG, self-monitored blood glucose; S2, stage 2; S3, stage 3; T1D, type 1 diabetes mellitus; TIR, time in range.

Areas Where Additional Data Are Needed

Currently, additional experimental data are needed to establish consensus metrics (e.g., percent of time above a defined cut-off level of glycemia or delta glucose concentrations after standardized meal tolerance tests). Data are also needed to identify rapid and slow progressors who might require different schedules of CGM screening. Eventually, CGM will probably be used as an adjunctive test to alert a patient when to go for traditional diagnostic tests or (with regulatory approval) even to be eligible for immunotherapy without a need for traditional blood testing. Previously published TrialNet data should be reanalyzed in this effort to develop glucose concentration cut-offs and corresponding CGM metrics to develop CGM-based criteria for defining stage 2 T1D.

When Should Continuous Glucose Monitoring Data Be Avoided Because of Interfering Factors?

• Interfering substances can degrade the accuracy of CGM glucose measurements.

• Current CGM technologies show a significant decline in accuracy at low glycemic values.

• Measurement of glycemic variability (GV) with a CGM does not appear to be a useful metric for identifying stage 2 T1D.

CGM holds promise as a tool for monitoring the gradual deterioration of glucose levels as people at risk transition between stages 1 and 2. Because these devices operate continuously, they can enhance the likelihood of timely treatment initiation, such as teplizumab, to delay the progress to stage 3 T1D and the subsequent need for insulin therapy. However, it is essential to exercise caution, because known interfering factors (e.g., some endogenous/exogenous substances and intercurrent conditions)^{68 -70} might increase the occurrence of false positives.

Table 5 provides a list of current CGM systems and substances that the respective manufacturer’s labeling describes as affecting sensor glucose concentrations. Evidence for the impact of interfering substances is often gained from clinical trials with human participants. These trials tend to be limited in the substance concentrations that can be achieved, especially for exogenous substances. In vitro bench tests may be a useful complement. ⁷¹ However, substance concentrations, like those in the CLSI (Clinical and Laboratory Standards Institute) EP37 standard, ⁷² are often defined for blood rather than ISF, so the identification of adequate concentrations may limit the interpretation of such bench tests’ results.

Table 5.

Continuous Glucose Monitoring System Interfering Substances According to Manufacturing Labeling.

CGM System	Interfering Substance
FreeStyle Libre 2, Abbott 73	#Ascorbic acid
FreeStyle Libre 3, Abbott 73	#Ascorbic acid
Dexcom G6 74	*Acetaminophen, **Hydroxyurea
Dexcom G7 74	*Acetaminophen, **Hydroxyurea
Guardian 3, Medtronic75,76	**Acetaminophen, **Hydroxyurea
Guardian 4, Medtronic 77	**Acetaminophen, **Hydroxyurea
Simplera, Medtronic 78	**Acetaminophen, **Hydroxyurea
Eversense E3, Senseonics 79	**Mannitol, **Sorbitol, ***Tetracycline
Eversense 365, Senseonics 80	**Mannitol, **Sorbitol, ***Tetracycline

All of the listed substances are declared to lead to falsely elevated sensor glucose readings if taken in sufficiently high doses.

Abbreviation: CGM, continuous glucose monitoring.

^#Over 500 mg per day; *Over 1000 mg every 6 hours in adults; **Dose not specified by the manufacturer ***Maximum physiologic or therapeutic plasma concentration 0.5 (0.0113) (mg/dL [mmol/L]). ⁸¹

While known substances and intercurrent conditions/comorbidities are also expected to impact CGM readings in people with stage 2 T1D,^{68 -70,82,83} a thorough analysis of interfering factors in CGM systems in this population demands a more contextual approach. Current CGM technologies show a significant decline in accuracy at both low glycemic values (40-70 mg/dL or 2.2-3.9 mmol/L) ⁸⁴ and periods of rapid BG concentration changes, with the former being more problematic since most individuals with preserved beta cell mass stay close to 70 mg/dL (3.9 mmol/L) in fasting and post-absorptive states and may have normal BG values as low as 50 mg/dL (2.8 mmol/L). ⁸⁵

Although CGM can detect increased GV, which is a sign of dysglycemia, the magnitude of any increase in GV is expected to be significantly less pronounced in stage 2 than in frank stage 3 T1D where overt hyperglycemia is present. No metrics of increased GV have been proposed as diagnostic tests for stage 2 T1D. Novel near real-time data processing and classification techniques based on emerging technologies (time series analysis or artificial intelligence, AI) will be needed to automate and enhance the detection of changing glycemic patterns in response to known disturbances.

What Type of Modeling Data Can Predict Stage 2 Type 1 Diabetes?

• The AI and ML models using CGM-derived glycemic features have been successful in classifying individuals at risk for T1D progression.

• The CGM-based dynamic markers, incorporating temporal glycemic patterns, offer an alternative approach to predicting stage 2 T1D progression.

• Validation studies are needed to assess the accuracy and generalizability of both AI-driven models and CGM-based dynamic markers in diverse populations.

The key advantage of CGM compared to BG monitoring lies in its ability to measure glucose values continuously and reflect glycemic regulation over time. One promising avenue involves integrating CGM data into AI/ML models, which has shown potential in advancing predictive tools for diabetes screening and treatment. For example, a one-week CGM home test combined with a linear Support Vector Machine (SVM) model, which is applied in classification analysis to find the optimal hyperplane that best separates different classes of data points in a high-dimensional space, ⁸⁶ has been used to classify participants’ autoantibody status (positive vs negative). ⁸⁷ In addition, glycemic features from CGM data following standardized liquid mixed meals (SLMM) have been used to predict an individual’s risk of progression to stage 3 T1D (low risk vs high risk) via an SVM model with Recursive Feature Elimination (RFE). ³² Furthermore, combining CGM-derived features with the T1D Genetic Risk Score (GRS) enables the classification of individuals with one vs multiple autoantibodies, aiding in the prediction of immunological risk for T1D. ⁸⁸

Beyond static measures, incorporating temporal dynamics into CGM-derived metrics such as a semi-Markov chain framework (a probabilistic method that accounts for how long a process has been in the current state when calculating the probability of transitioning to a new state) ⁸⁹ may improve early detection of glycemic deviations and disease progression. This approach involves (1) encoding CGM data into sequences of glycemic states; (2) calculating a transition probability matrix between these states; (3) deriving the stationary distribution of time spent in each state; and (4) computing the entropy rate of the encoded glycemic process. ⁹⁰

This method maps CGM profiles into symbolic sequences, which serve as markers of disease progression or treatment response. By combining the transition probability matrix with traditional CGM metrics like the Ambulatory Glucose Profile (AGP), additional insights into GV and the timing of clinically relevant events are obtained. These insights provide a more dynamic and comprehensive understanding of T1D progression, as presented in Figure 1.

Figure 1.

Transition probability matrices for two individuals: autoantibody-negative (a) vs autoantibody-positive (b). (a) Weighted average transition probability matrix for an autoantibody-negative individual with a low entropy rate (0.08). The x- and y-axes represent eight glycemic states based on mean CGM glucose (MG): (a) MG <54 mg/dL, (b) 54≤MG<69 mg/dL, (c) 70≤MG<120 mg/dL, (d) 121≤MG<140 mg/dL, (e) 141≤MG<160 mg/dL, (f) 161≤MG<180 mg/dL, (g) 181≤MG<250 mg/dL, and (h) MG >250 mg/dL. All glycemia remain in low-risk zones (c and d), indicating stable glycemic control. (B) Weighted average transition probability matrix for an autoantibody-positive individual (with two or more autoantibodies) and a high entropy rate (0.96). Glycemia spans across zones (a to e), with progression to higher-risk zones (f and g), reflecting increased GV and risk. Colored zones indicate the probability of transitioning between glycemic states.

Validation of these approaches in diverse populations is critical to ensure broad applicability and equity in clinical care. By developing CGM-based diagnostic tools tailored to identifying stage 2 T1D, earlier interventions and improved disease management for individuals at high risk become increasingly achievable.

What are Psychosocial Aspects of Blood Test Screening and Continuous Glucose Monitoring Screening?

• Individuals who learn that they or a loved one have screened positive for stage 2 T1D often experience heightened anxiety.

• Monitoring someone at high risk of T1D might help reduce the unpredictability of clinical disease onset.

• Concerns about CGM monitoring may include fears about device insertion, explaining the device to others, wear time, activity limitations, and meaning of CGM results, particularly when results from different sources provide conflicting information about disease progression.

Humans are stressed when faced with situations that are threatening, unpredictable, and uncontrollable. T1D is clearly threatening—a life-long disease requiring daily insulin replacement and glucose monitoring, a shortened life-expectancy, numerous serious health complications, and no cure. For the patient, a diagnosis of early-stage (stage 1 or 2) T1D provides neither predictability nor controllability. The patient is told they are likely to develop clinical T1D but when this will occur remains unclear—it could be months or even years away. Furthermore, there is almost nothing they can do to stop the disease—the possibility of delaying disease onset is currently their only option.

Monitoring someone diagnosed with early-stage T1D may help reduce the unpredictability of clinical disease onset. CGM could be useful in this regard if it can add clarity about disease progression, timing of clinical disease onset, or possible access to treatments to delay T1D. However, CGM has been primarily used unblinded in patients with T1D; far less is known about its acceptability and utility among blinded persons without clinical T1D. ³⁶ Studies that have used CGM with healthy individuals often focus on adults,^55,91,92 who may find it more acceptable than children. However, in one of the largest studies of CGM with healthy adults, Rodriguez-Segade et al ⁵⁵ reported that only 58% of 1065 eligible participants agreed to undergo a six-day period of CGM.

The global incidence of stage 3 T1D is consistently highest among individuals aged 5 to 24 years, peaking at ages 10 to 14 years. ⁹³ Research indicates that those in this age group often face challenges with device adherence because of various lifestyle factors, social stigma, and forgetfulness. ⁹⁴ As a result, it is reasonable to expect a greater reluctance to wear a device like a CGM system, even intermittently, and even if it is used by people with stage 1 T1D who are asymptomatic and apparently euglycemic to screen for the onset of stage 2 T1D, that might not appear to the subject to provide immediate benefits. Therefore, support from an interdisciplinary expert team is advised to increase adherence. ⁹⁵ Six issues that will need to be addressed if CGM is to be included as part of monitoring those with early-stage T1D are presented in Table 6.

Table 6.

Psychosocial Issues That Will Need to Be Addressed Before CGM Can Be a Routine Screening Test for Stage 2 T1D.

1	Acceptance of an invasive procedure, particularly in children who may find the device insertion scary. Older children and teens may be concerned with questions from their peers about the device and how they should respond.
2	Education about the purpose of CGM and how the data can be used to inform the patient and family about the possible timing of progression to clinical T1D.
3	Determination of the length of time needed to wear the CGM as well as any activity limitations or considerations.
4	Recognition of the importance of blinded CGM collection must be part of the educational process because some individuals or parents may want to know the CGM results in real-time.
5	Provision of resources to respond to participant questions and concerns about CGM readings. Individuals should be monitored for concerning behavior changes—such as severe dietary restrictions in an effort to get “better” CGM readings.
6	Establishment of a plan to provide consistent messaging about multiple data streams from CGMs and other sources to provide predictability to the individual and family. Avoid delivering conflicting information to the family.

What are Future Research Priorities in the Area?

• More information is needed about what amount of time in range and what amount of time above range should trigger an OGTT.

• Greater accuracy of CGM technology will be needed for CGM screening to replace blood testing for identifying stage 2 T1D.

• At this time, the best role for CGM screening for stage 2 T1D is as a screening test leading to definitive blood testing.

The staging of early-stage T1D has allowed us a clear platform for potential treatments to delay or even stop the progression of beta cell failure. ⁵ Much has been learned about the limitations of HbA_1c for making diagnoses, and therefore, some believe only glucose should be used for the diagnosis of diabetes or prediabetes. ⁹⁶ With that being the case, are we at the point in the evolution of CGM that we can use this technology to diagnose stage 2 T1D?

Several concerns are immediately obvious. First, CGM measures ISF glucose levels while the dysglycemia in stage 2 T1D is currently diagnosed with standardized plasma glucose measurements derived from the OGTT. Furthermore, while the accuracy of CGM systems has dramatically improved over the past 20 years, it is unclear if the accuracy of today’s technology is adequate for the diagnosis of stage 2 T1D.

It should also be noted that the traditional OGTT has its own limitations, especially for a “gold standard.” This has been well-known for over six decades. ⁹⁷ The amount of carbohydrate consumed in the days prior to the test impacts postprandial glycemia.^98,99 Animal models have shown that longer-term low-carbohydrate diets result in decreased beta-cell mass ¹⁰⁰ that appears to be reversible. ¹⁰¹ Therefore, the current recommendation is three days of 150 grams of carbohydrate consumed prior to the test.^2,102

The fundamental question now is how CGM should be used to diagnose stage 2 T1D. Should CGM be used as a screening tool to decide when a formal OGTT should be performed? The alternative is to randomly perform an OGTT without specific timing. For the use of CGM as a screening tool, we would require a better understanding of what time in which range and, more importantly, how much time above range should trigger the OGTT.

The other option is to use CGM alone to make the diagnosis of stage 2 T1D without the OGTT. The problem with this option is that CGM glucose metrics are completely dependent on dietary carbohydrate intake; thus, diagnosing dysglycemia will be difficult. However, if it can be determined what time in ranges and time above ranges are correlated with the dysglycemia from the OGTT seen in stage 2 T1D, then it would be possible to use CGM. Realistically, this would need to be accomplished with a normal diet (especially in children), making the correlation for dysglycemia for individuals challenging. It seems that in the short-term, CGM could best be verified as a screening tool for stage 2 T1D.

In the European IHI project, EDENT1FI (European action for the Diagnosis of Early Non-clinical Type 1 diabetes For disease Interception), 200 000 children and adolescents will be screened for antibodies by 2028. Enrollees with at least two antibodies will be assessed for progression to stage 2 T1D and stage 3 T1D. ¹⁰³ Metabolic monitoring, including optional CGM will be offered. This study should reveal additional information about the natural history of the stages of T1D and the value of CGM screening. ¹⁰⁴

Conclusion

CGM appears to be a promising tool to screen for stage 2 T1D. This approach will allow identification of individuals who meet consensus criteria for definitive traditional venous BG testing to screen for stage 2 T1D as required presently by the FDA prior to prescribing an immune protective drug. Preliminary data suggest that 5% to 10% TAR140 looks like a promising CGM metric to predict progression to symptomatic stage 3 T1D. The same metric might turn out to be useful for identifying progression from stage 1 T1D to stage 2 T1D. This means that the more time spent with ISF glucose ≥140 mg/dL, the greater the chance of progressing from stage 1 T1D to stage 2 T1D. It is important to identify the onset of stage 2 T1D because if it is missed, then not only is the patient at risk of progressing to frank T1D, but the opportunity to treat with an immune drug at stage 2 T1D will have been bypassed. The CGM-based dynamic markers, along with AI/ML tools, show promise in using CGM data to identify stage 2 T1D. In developing these tools and metrics, it is crucial to balance sensitivity with specificity. If stage 2 T1D is identified, then it appears that teplizumab, and potentially exercise and/or dietary modifications, could delay progression to stage 3 T1D. At this time, we recommend additional research on topics related to CGM screening for stage 2 T1D, including the accuracy of CGM measurements needed to identify/diagnose stage 2 T1D, any CGM metrics that should trigger the need for more definitive tests for T1D, and/or dietary guidelines for those using CGM to identify stage 2 T1D. The panel developed 16 conclusions about CGM screening for stage 2 T1D. These conclusions are presented in Table 7. CGM has a promising future in screening for stage 2 T1D, and this test might eventually replace plasma glucose testing and HbA_1c testing as a standalone method for identifying people who can benefit from immunotherapy.

Table 7.

Sixteen Conclusions About CGM Screening for Stage 2 T1D.

The current status of screening for stage 2 T1D

(1) The three stages of T1D are characterized by two or more islet cell autoantibodies plus progressively higher glucose levels (normal, slightly elevated, and frankly elevated) and progressively greater hyperglycemic symptoms.

(2) Teplizumab, a monoclonal antibody, is currently the only FDA-approved drug for delaying the onset of frank T1D if it is initiated during stage 2 T1D, which is currently diagnosed by fasting plasma glucose, 2-hour plasma glucose on an oral glucose tolerance test, or hemoglobin A1c.

(3) No CGM system is currently considered sufficiently accurate by a major professional organization or regulatory body to diagnose stage 2 T1D with robust specificity and sensitivity.

(4) Concerns about CGM monitoring may include fears about device insertion, explaining the device to others, wear time, activity limitations, and meaning of CGM results, particularly when results from different sources provide conflicting information about disease progression.

(5) Interfering substances can degrade the accuracy of CGM glucose measurements.

The current status of CGM screening for stage 2 T1D

(6) Accuracy of current CGM systems must be confirmed by additional research to screen for this purpose, and no regulatory clearance is available to use only CGM for screening.

(7) Populations defined as being normoglycemic have usually reported less than 10% time spent above euglycemia (>140 mg/dL) and almost no time at all above 180 mg/dL.

(8) An elevated percentage of time (ie, above 5%-10%) above euglycemia (ie, >140 mg/dL) is a good metric for assessing the risk of stage 2 T1D.

(9) AI and ML models using CGM-derived glycemic features have been successful in classifying individuals at risk for T1D progression.

(10) In research studies, CGM-based dynamic markers have successfully identified and staged individuals at risk for T1D progression.

(11) CGM screening of children for stage 2 T1D performed at home would decrease the need for routine phlebotomy screening and often poorly tolerated glucose loads necessary for an oral glucose tolerance test, both of which are currently performed at a clinical laboratory.

The future status of CGM screening for stage 2 T1D

(12) At this time, the best role for CGM screening for stage 2 T1D is as a screening test leading to definitive blood testing.

(13) CGM screening for stage 2 T1D should use blinded CGM if possible.

(14) For children with stage 2 T1D, CGM screening should be performed every three months if less than 3 years old, at least every six months if 3-9 years old, and at least every 12 months if greater than 9 years old.

(15) For adults, CGM screening should be repeated every 6-12 months based on the presence of multiple antibodies, and every 12-36 months based on the presence of a single antibody or only a family history of T1D.

(16) For CGM screening to be adopted, more data will be needed to understand this method’s clinical effectiveness to minimize unnecessary phlebotomy, its risks to patients of inaccurate findings, and its economic impact.

Abbreviations: AI, artificial intelligence; CGM, continuous glucose monitoring; FDA, Food and Drug Administration; mg/dL, milligrams per deciliter; ML, machine learning; T1D, type 1 diabetes.