Limitations of 14-Day Continuous Glucose Monitoring Sampling for Assessment of Hypoglycemia and Glycemic Variability in Type 1 Diabetes

Abstract

Continuous glucose monitoring (CGM) has become the standard of care in diabetes management with the recent advances in technology and accessibility in the last decade. An International Consensus was established to define CGM metrics and its goals in diabetes care. The 2019 International Consensus suggested 14 days of CGM sampling for the assessment of CGM metrics stating the limitations that may occur for hypoglycemia and glycemic variability metrics. Since then, several studies assessed the correlation between CGM metrics and duration of the sampling period. This review summarized the studies that investigated the relationship between 14-day CGM sampling to 90-day CGM data in >70% CGM users for all CGM metrics and highlighted possible solutions for more accurate CGM sampling durations in type 1 diabetes (T1D). Accumulating evidence showed that 14-day CGM sampling correlates well with 90-day CGM data for mean glucose, time in 70–180 mg/dL, and hyperglycemia metrics; however, it correlates weakly for hypoglycemia and glycemic variability metrics. In the studies included in this review, in adults with T1D, minimum sampling duration was 14 days for mean glucose, time in 70–180 mg/dL, and time in hyperglycemia (>180 and >250 mg/dL); however, minimum sampling duration varied between 21 to 30 days for time <70 mg/dL, 30 to 35 days for time <54 mg/dL, and 28 to 35 days for coefficient of variation. Longer than 14 days of CGM, sampling was required to properly assess hypoglycemia and glycemic variability in T1D.

Introduction

A1c has been the gold standard of glycemic control for many decades in diabetes management; however, its use has been limited due to various reasons.^1,2 A1c does not provide information about glycemic variability or risk of hypoglycemia and is affected by many pathologic conditions such as anemia, chronic kidney disease.³ Introduction of continuous glucose monitoring (CGM) opened a new era in diabetes management. CGM has the advantage of showing not only glucose value but also its trends.⁴ CGM use has been shown to improve glycemic control and decrease risk of hypoglycemia compared to self-monitoring blood glucose in patients with type 1 (T1D) and type 2 diabetes (T2D).^5

–8 Use of CGM has increased in the last decade in parallel with the advancements in CGM technology and increasing insurance coverage in developed countries.⁹ CGM has been endorsed by several medical associations as the standard of care for monitoring glycemic changes for adults, children, and pregnant women with T1D and T2D with multiple daily injections (MDI), hybrid closed-loop (HCL) users, and basal insulin users with T2D.^10,11 However, its use is still limited due to cost and access in developing countries.⁴ Another group of experts suggested CGM use in patients with T1D who are not meeting glycemic targets, have hypoglycemia unawareness, and/or episodes of hypoglycemia.¹²

The International Consensus defined CGM metrics in 2017¹³ and suggested standardization of data interpretation with suggested targets for CGM metrics in 2019.¹⁴ CGM metrics such as time in range (TIR), time above and below ranges, and coefficient of variation (CV) were described with their targets in different types of diabetes (T1D, T2D, and pregnancy, elderly, etc.)^13
–15 The International Consensus also recommended using CGM metrics as core endpoints in diabetes clinical trials in a third consensus in 2023.¹⁶ The International Consensus recommended 14-day CGM sampling for >70% of CGM users for data interpretation citing two studies (Xing et al.¹⁷ and Riddlesworth et al.¹⁸) that found a good correlation between 14-day sampling and 90-day CGM sampling. However, the International Consensus also acknowledged the limitations of 14-day sampling in certain situations in the 2019 consensus.¹⁴ In addition to the limitations of 14-day sampling for hypoglycemia metrics, it has been highlighted that reporting of hypoglycemia as a percentage can also be challenging in clinical practice.¹⁹ This review aims to compare studies in T1D investigated optimal sampling duration of CGM metrics to reflect 90 days of CGM data.

Methods

A comprehensive search of databases (MEDLINE, EMBASE, Cochrane Database of Systematic Reviews, and Scopus) for articles in English, published from January 2000 to December 2023, was conducted based on the terms in titles and abstracts: “optimal sampling duration, minimum sampling duration, sampling duration” AND “CGM, continuous glucose monitoring,” AND “type 1 diabetes, OR CGM metrics.” Inclusion criteria were studies that investigated the correlation of CGM metrics, TIR (70–180 mg/dL), time >180 and 250 mg/dL, time <54 and 70 mg/dL, and CV between 14-day sampling and 90-day CGM data in CGM users >70% in patients use CGM and/or automated insulin delivery systems. Exclusion criteria was any optimal sampling duration less or up to 14 days. All data reported in this review were obtained from previous publications. Sampling duration days for each CGM metric for the coefficient of determination, R ² of 0.88 (or closest), were compared. R ² of 0.88 was chosen because most of the studies used this cutoff as detailed in the following sections.

Statistics

Statistical analyses in the studies summarized in this review used r or R ² for comparisons. The correlation coefficient (r) is a measure that determines the direction and strength of a linear relationship between two variables. The coefficient of determination (R ²) is a measure that explains the proportion of the variance in the dependent variable that is predictable from the independent variable(s). It is the square of the correlation coefficient (r), (r × r = R ²) for univariate correlations. This represents the proportion of the total variance in the dependent variable explained by the independent variable(s).

Results

The literature search yielded seven studies. After reading titles and abstracts, six full-text studies were reviewed, and four of them met the inclusion criteria to be included (three adults and one pediatric).^18,20
–22 Two studies were excluded; one compared no more than 15-day optimal sampling days,¹⁷ other one calculated optimal sampling duration for glycemia risk index.²³ Three studies used R ², and one study used r, therefore, R ² was chosen for comparison. R ² was obtained by multiplying r value times itself in one study,²¹ other R ² results were obtained directly from the studies.^18,20,22 In all published studies to reach the same level of R ² required more days for hypoglycemia metrics and CV.^18,20
–22 Table 1 shows a summary of published studies that compared 14-day sampling to 90-day CGM data in >70% CGM users.

Table 1.

Minimum Continuous Glucose Monitoring Sampling Duration of CGM Metrics in the Studies That Investigated the Correlation of CGM Metrics of 14-Day CGM Sampling to 90-Day CGM Data in >70% CGM Users in Type 1 Diabetes

Mean glucose	TIR 70–180 mg/dL	T > 180 mg/dL	T > 250 mg/dL	T < 70 mg/dL	T < 54 mg/dL	CV	CGM	Sample size	R²	Insulin delivery	Study population	Study
14 Days	14 Days	14 Days	14 Days	25 Days	30 Days	30 Days	Freestyle Libre Pro, Dexcom G4	257	0.84–0.86	MDI and CSII (no HCL)	Adults	Riddlesworth et al.^18,a
14 Days	14 Days	14 Days	14 Days	21–28 Days	35 Days	28 Days	Dexcom G6	336	0.88	MDI and CSII (no HCL)	Adults	Akturk et al.^20,b
∼14 Days	∼14 Days	∼14 Days	∼14 Days	∼25–30 Days	∼30–35 Days	∼35 Days	Freestyle Navigator II, Guardian 3	56	0.88	HCL	Adults	Leelarathna et al.^21,c
∼21 Days	∼21 Days	∼21 Days	∼21 Days	∼28 Days	∼35 Days	∼28 Days	Libre 1, Dexcom G5 and G6	654	0.88	MDI and CSII (no HCL)	Pediatric	Piona et al.^22,d

R ²: Squared value of the Spearman's correlation coefficient. For each CGM metric, the association between the value from each sampling period and the value using all 3 months of data was determined using the R ². This represents the proportion of the total variance explained by the sample. Riddlesworth and Akturk et al. studies used “reaching a plateau” method for R ². A plateau was defined as flattening of R ² curve with consideration of incremental changes in R ². Leelarathna et al. study used a cut off r.

Riddlesworth et al.¹⁸ study, data were obtained from Supplementary Files and Supplementary Table 3. R ² was 0.84 for mean glucose, and T > 180 mg/dL, 0.85 for T > 250 mg/dL and 0.86 for TIR (70–180 mg/dL).

Akturk et al.²⁰ study, data were obtained from Table 1. Optimal duration for CGM sampling was 21 days for CGM use ≥80% and 28 days for CGM use 70%–80%.

Leelarathna et al.²¹ study, data were projected for R ² of 0.88 from Figure 1. R ²: 0.88 was chosen for comparison purposes with other studies in the table. Results are published for r: 0.95.

Piona et al. study,²² data were projected from Table 2. R ²: 0.88 was chosen for comparison purposes with other studies in the table.

CSII, continuous subcutaneous insulin infusion; CV, coefficient of variation; HCL, hybrid-closed loop; TIR, time in range (70–180 mg/dL), MDI, multiple daily injections; T, time; ∼: about, used for projected data.

Table 2.

R ² for Continuous Glucose Monitoring Metrics with the 3-Month CGM Data Compared to Number of Sampling Days in Riddlesworth Et Al. Study

Sampling days	14 Days	25 Days	30 Days
Mean glucose	0.84	0.87	0.92
TIR	0.86	0.92	0.94
T > 180	0.84	0.90	0.92
T > 250	0.85	0.92	0.93
T < 70	0.76	0.85	0.87
T < 54	0.68	0.80	0.83
CV	0.70	0.82	0.84

R ²: squared value of the Spearman's correlation coefficient. For each CGM metric, the association between the value from each sampling period and the value using all 3 months of data was determined using the R ². This represents the proportion of the total variance explained by the sample. Supplementary table 3 data were used to make this table from Riddlesworth et al.¹⁸ The table is recreated with the published data from Riddlesworth et al.¹⁸

TIR, time in range (70–180 mg/dL); T, time.

14-Day CGM sampling correlates strongly for mean glucose, TIR, and hyperglycemia metrics

As quoted from the 2019 International Consensus¹⁴ “>70% use of CGM over the most recent 14 days correlates strongly with 3 months of mean glucose, TIRs, and hyperglycemia metrics. In individuals with T1D, correlations are weaker for hypoglycemia and glycemic variability. However, these correlations have not been shown to increase with longer sampling periods. Longer CGM data collection periods may be required for individuals with more variable glycemic control (e.g., 4 weeks of data to investigate hypoglycemia exposure).”

As seen above, in the first and second sentences, the International Consensus agrees there is a strong correlation between 14 days of CGM sampling and 3 months of mean glucose, TIR and hyperglycemia metrics, and a weak correlation of hypoglycemia and glycemic variability metrics.¹⁴ However, in the following sentences, suggestions contradict each other as the first sentence mentions that “correlations have not been shown to increase with longer sampling periods” then the second sentence mentions that “longer CGM sampling may be required for people with more variable glycemic control,” and they gave an example of possible 4 weeks of sampling for hypoglycemia.¹⁴ The International Consensus cited the study by Riddlesworth et al.¹⁸ for these recommendations and adapted this paragraph from the same study conclusion section. If we review that study, we see high R ² for mean glucose, TIR and hyperglycemia metrics, and lower R ² for hypoglycemia metrics and CV for 14-day sampling (Table 2). However, the authors interpreted the results as “correlations have not been shown to increase with longer sampling durations for hypoglycemia metrics” despite the numerical and statistical differences showing that longer duration of CGM sampling increased the correlation between 14-day sampling and 90-day data for hypoglycemia metrics and CV (Table 2). Authors interpreted R ² levels of 0.84–0.86 for mean glucose, TIR, hyperglycemia metrics as a strong correlation, and R ² levels of 0.68–0.70 for hypoglycemia and glycemic variability metrics as a weak correlation for 14-day sampling (Table 2), but then contradicted with the conclusion that longer sampling duration does not improve the correlation for hypoglycemia and glycemic variability metrics despite R ² having reached 0.84–0.87; the same cutoff values they considered as a strong correlation for mean glucose, TIR, and hyperglycemia metrics¹⁸ (Table 2). Therefore, Riddlesworth et al.'s¹⁸ study found that the optimal duration for CGM sampling was 14 days for mean glucose, TIR, and hyperglycemia metrics (time >180 and >250 mg/dL); 30 days for CV and time <54 mg/dL and 25 days for time <70 mg/dL.

14-Day CGM sampling correlates weakly for hypoglycemia metrics and CV

Xing et al.¹⁷ conducted the first study to investigate the correlation between the sampling duration and 30–60–90-day CGM data. They showed that R ² correlation increased from 3 to 15 days gradually with longer sampling duration for all CGM metrics, 15-day sampling showed a better correlation with the 90-day CGM data. However, this study did not investigate sampling for more than 15 days therefore, the evidence from this study was limited and excluded from this review.

Akturk et al.²⁰ investigated the minimum duration of CGM sampling to represent 90-day CGM data in suboptimal CGM users (<70%) with T1D. CGM use was grouped into 5% increments between 45% and 95% over 90 days. When comparing the suboptimal (<70% CGM use) and optimal users (≥70% CGM use) in 336 CGM users, it was found that the minimum duration for CGM sampling was 14 days for mean glucose, TIR, and hyperglycemia metrics (time >180 and >250 mg/dL); 28 days for CV; and 35 days for time <54 mg/dL for CGM users >70%; however, for time <70 mg/dL, it was 21 days for CGM use ≥80% and 28 days for CGM use 70%–80%.²⁰ For patients use CGM >45% in the last 90 days; 14 days of CGM sampling was enough for mean glucose, TIR, time >180 and >250 mg/dL; in contrast, longer duration of CGM sampling was required to assess time <70 and <54 mg/dL and CV to reflect 90-day CGM data.

Leelarathna et al.²¹ investigated the duration of HCL therapy to achieve representative glycemic outcomes in adults with T1D. In this study, CGM data of 56 adults using two different forms of HCL for 12 weeks in a multinational clinical trial was reviewed. The r between CGM data over the first 60 days of HCL use and the respective values collected over the entire 12-week study was compared. Optimal duration to reach r of 0.95 for mean glucose was 27 days, for TIR it was 26 days, time >180 mg/dL was 24 days, time >250 mg/dL was 33 days, time <70 mg/dL was 38 days, time <54 mg/dL was 49 days, and CV was 47 days.²¹ To compare to other studies, r values from the study were converted to R ² of 0.88. The estimated optimal duration for CGM sampling was 14 days for mean glucose, TIR, and hyperglycemia metrics (time >180 and >250 mg/dL); 35 days for CV, 30–35 days for time <54 mg/dL, and 25–30 days for time <70 mg/dL.

Piona et al.²² investigated optimal sampling duration in 654 pediatric patients with T1D. They found that 4 weeks of CGM data better represents 90-day CGM data for glucose variability in patients using MDI and intermittently scanned CGM; on the other hand, 2 weeks of CGM data may be enough in CSII and real-time CGM users. In this study, R ² values for 4-week sampling period for CV and time in hypoglycemia metrics were significantly higher than 2-week period.²² The estimated optimal duration for CGM sampling was 21 days for mean glucose, TIR, and hyperglycemia metrics (time >180 and >250 mg/dL); 28 days for CV, 35 days for time <54 mg/dL and 28 days for time <70 mg/dL.

Discussion

There is a confusion for the CGM sampling terminology. The terms that were used in the studies with T1D described the sampling duration as either “optimal”^17,21,22 or “minimum.”²⁰ One study used both terms interchangeable throughout the manuscript.¹⁸ In fact, the duration that is calculated in studies is not “optimal” but “minimum” CGM sampling reflects 90-day CGM data. In all studies, the CGM metrics achieved a higher R ² with longer duration of CGM sampling^18,20
–22; therefore, the longer the duration, the better the correlation for “optimal” sampling. However, it is unknown if there is any clinical correlation of that increasing statistical correlation. Therefore, “minimum” CGM sampling should be the correct term.

There is no consensus for “strong” and “weak” correlation for R ² in CGM sampling. R ² is the squared value of the Spearman's correlation coefficient. For each CGM metric, the association between the value from each sampling period and the value using all 3 months of data was determined using R ². This represents the proportion of the total variance explained by the sample. Studies used different R ² to define minimum sampling duration. For example, Riddlesworth et al.¹⁸ and Akturk et al.²⁰ studies used the same methodology, “reaching a plateau” method for R ², plateaued R ² levels were between 0.84 and 0.88 for different CGM metrics. A plateau was defined as the flattening of the R ² curve with consideration of incremental changes in R ².^18,20 However, the Leelarathna et al.²¹ study used a cutoff r (Table 1). One way to solve the problem is either finding a meaningful R ² cutoff for each CGM metric or using the “plateau” method for all CGM metrics. For example, some correlations (e.g., TIR and A1c or mean glucose and A1c) do not ever reach a high R ² (e.g., >0.8). Using the plateau method can be more useful considering there may be no clinical correlation after a reached R ² level.

Ambulatory glucose profile (AGP) reports all CGM metrics as percentages. Reporting hypoglycemia metrics as percentages can be problematic in 14-day sampling.¹⁹ AGP can calculate hypoglycemia and variability metrics using longer than 14-day sampling or developing different hypoglycemia and variability metrics from the 14-day sampling that reflects 90-day CGM data accurately can be helpful.

The strengths of this review are methodological and systematic review of the literature, including the studies after the International Consensus publication date (2019), obtaining the data from the published manuscripts and supplementary files, and highlighting the need for an update for the minimum sampling recommendations. The limitations of this review are the limited number of published studies in T1D and may not apply to other types of diabetes (especially for people with T2D that are not on insulin), using different types of CGM and insulin delivery methods in those studies. Further larger studies are needed to better understand the minimum sampling duration for each CGM metric that represents the long-term glycemic control.

Conclusion

This review summarized that 14-day CGM sampling can strongly reflect 90-day CGM data for mean glucose, TIR, and time in hyperglycemia (>180 and >250 mg/dL); however, more than 14 days of CGM sampling is required to appropriately assess hypoglycemia (<70 and <54 mg/dL) and glycemic variability. Minimum sampling duration varies between 21 to 30 days for time <70 mg/dL, 30 to 35 days for time <54 mg/dL, and 28 to 35 days for CV according to the published studies in adults with T1D. Minimum sampling duration may require longer days in pediatric population.

Footnotes

Acknowledgment

I thank Janet Snell-Bergeon PhD for statistical consultation for this review.

Author Disclosure Statement

Dr. Akturk received research support through the University of Colorado from Dexcom, Medtronic, Tandem Diabetes, Eli Lilly, Mannkind, IM Therapeutics, REMD Biotherapeutics; consultation fees through the University of Colorado from Dexcom, Medtronic, and Tandem Diabetes.

Funding Information

No funding was received for this article.

References

Wright

, Hirsch

. Metrics beyond hemoglobin A1C in diabetes management: Time in range, hypoglycemia, and other parameters. Diabetes Technol Ther, 2017; 19(S2):S16–S26.

Beck

, Bergenstal

. Beyond A1C-standardization of continuous glucose monitoring reporting: Why it is needed and how it continues to evolve. Diabetes Spectr, 2021; 34(2):102–108.

Vigersky

, McMahon

. The relationship of hemoglobin A1C to time-in-range in patients with diabetes. Diabetes Technol Ther, 2019; 21(2):81–85.

Garg

, Akturk

. The future of continuous glucose monitoring. Diabetes Technol Ther, 2017; 19(S3):S1–S2.

Pratley

, Kanapka

, Rickels

, et al. Effect of continuous glucose monitoring on hypoglycemia in older adults with type 1 diabetes: A randomized clinical trial. JAMA, 2020; 323:2397–2406.

Laffel

, Kanapka

, Beck

, et al. Effect of continuous glucose monitoring on glycemic control in adolescents and young adults with type 1 diabetes: A randomized clinical trial. JAMA, 2020; 323:2388–2396.

Lind

, Polonsky

, Hirsch

, 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, 2017; 317:379–387.

Beck

, Riddlesworth

, Ruedy

, et al. Effect of continuous glucose monitoring on glycemic control in adults with type 1 diabetes using insulin injections: The DIAMOND randomized clinical trial. JAMA, 2017; 317:371–378.

Akturk

, Garg

. Technological advances shaping diabetes care. Curr Opin Endocrinol Diabetes Obes, 2019; 26(2):84–89.

10.

ElSayed

, Aleppo

, Aroda

, et al. on behalf of the American Diabetes Association. 7. Diabetes Technology: Standards of care in diabetes-2023. Diabetes Care, 2023; 46(Suppl 1):S111–S127.

11.

Blonde

, Umpierrez

, Reddy

, et al. American Association of Clinical Endocrinology Clinical Practice Guideline: Developing a Diabetes Mellitus Comprehensive Care Plan-2022 Update. Endocr Pract, 2022; 28(10):923–1049.

12.

Sacks

, Arnold

, Bakris

, et al. Guidelines and recommendations for laboratory analysis in the diagnosis and management of diabetes mellitus. Diabetes Care, 2023; 46(10):e151–e199.

13.

Danne

, Nimri

, Battelino

, et al. International consensus on use of continuous glucose monitoring. Diabetes Care, 2017; 40:1631–1640.

14.

Battelino

, Danne

, Bergenstal

, 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.

15.

Monnier

, Colette

, Wojtusciszyn

, et al. Toward defining the threshold between low and high glucose variability in diabetes. Diabetes Care, 2017; 40(7):832–838.

16.

Battelino

, Alexander

, Amiel

, et al. Continuous glucose monitoring and metrics for clinical trials: An international consensus statement. Lancet Diabetes Endocrinol, 2023; 11(1):42–57.

17.

Xing

, Kollman

, Beck

, et al. Optimal sampling intervals to assess long-term glycemic control using continuous glucose monitoring. Diabetes Technol Ther, 2011; 13(3):351–358.

18.

Riddlesworth

, Beck

, Gal

, et al. Optimal sampling duration for continuous glucose monitoring to determine long-term glycemic control. Diabetes Technol Ther, 2018; 20(4):314–316.

19.

Shah

. Limitations of reporting time below range as a percentage. Diabetes Technol Ther, 2023; 25(11):822–825; doi: 10.1089/dia.2023.0247

20.

Akturk

, Sakamoto

, Vigers

, et al. Minimum sampling duration for continuous glucose monitoring metrics to achieve representative glycemic outcomes in suboptimal continuous glucose monitor use. J Diabetes Sci Technol, 2023:19322968231200901. [Epub ahead of print]; doi: 10.1177/19322968231200901

21.

Leelarathna

, Thabit

, Willinska

, et al. Duration of hybrid closed-loop insulin therapy to achieve representative glycemic outcomes in adults with type 1 diabetes. Diabetes Care, 2020; 43:e38–e39.

22.

Piona

, Marigliano

, Mozzillo

, et al. Long-term glycemic control and glucose variability assessed with continuous glucose monitoring in a pediatric population with type 1 diabetes: Determination of optimal sampling duration. Pediatr Diabetes, 2020; 21(8):1485–1492.

23.

Shah

, Sakamoto

, Pyle

. Optimal sampling duration for continuous glucose monitoring for the estimation of glycemia risk index. Diabetes Technol Ther, 2023; 25(2):140–142