Abstract
BACKGROUND:
The glucose management indicator (GMI) is an estimated measure of hemoglobin A1c (HbA1c) recommended for the management of persons with diabetes using continuous glucose monitoring (CGM). However, GMI was derived primarily in young adults with type 1 diabetes, and its performance in patients with type 2 diabetes is poorly characterized.
METHODS:
We conducted a prospective cohort study in 144 adults with obstructive sleep apnea and type 2 diabetes not using insulin (mean age: 59.4 years; 45.1% female). HbA1c was measured at the study screening visit. Participants simultaneously wore 2 CGM sensors (Dexcom G4 and Abbott Libre Pro) for up to 4 weeks (2 weeks at baseline and 2 weeks at the 3-month follow-up visit). GMI was calculated using all available CGM data for each sensor.
RESULTS:
Median wear time was 27 days (IQR: 23–29) for the Dexcom G4 and 28 days (IQR: 24–29) for the Libre Pro. The mean difference between HbA1c and GMI was small (0.12–0.14 percentage points) (approximately 2 mmol/mol). However, the 2 measures were only moderately correlated (r = 0.68–0.71), and there was substantial variability in GMI at any given value of HbA1c (root mean squared error: 0.66–0.69 percentage points [7 to 8 mmol/mol]). Between 36% and 43% of participants had an absolute difference between HbA1c and GMI ≥0.5 percentage points (≥5 mmol/mol), and 9% to 18% had an absolute difference >1 percentage points (>11 mmol/mol). Discordance was higher in the Libre Pro than the Dexcom G4.
CONCLUSIONS:
GMI may be an unreliable measure of glycemic control for patients with type 2 diabetes and should be interpreted cautiously in clinical practice.
Clinicaltrials.gov Registration Number: NCT02454153.
Introduction
Glycated hemoglobin, or hemoglobin A1c (HbA1c) reflects average glucose over the past 2 to 3 months and is the standard of care for monitoring glycemic control in clinical practice. However, HbA1c does not characterize variability or acute excursions in glucose and may be less reliable in the setting of altered red cell turnover and certain comorbidities, such as some types of anemia or chronic kidney disease (1).
There is increasing use of integrated continuous glucose monitoring (CGM) systems in the management of diabetes. The glucose management indicator (GMI) is an estimated measure of HbA1c based on CGM data (2). Because GMI is unaffected by red blood cell turnover, there is growing interest in using this measure as a complement or replacement to HbA1c. GMI was first recommended for clinical use in 2017 and is included in all standardized CGM data reports (3). However, GMI was derived primarily in younger adults with type 1 diabetes (2), and its performance in patients with type 2 diabetes is unclear. Differences in the performance of GMI across CGM devices and patient populations remain poorly characterized.
As the use of CGM continues to grow in patients with type 2 diabetes, addressing these knowledge gaps is critical. In this study, we used 2 different CGM systems to evaluate the performance of GMI in a cohort of middle-age and older adults with type 2 diabetes and untreated obstructive sleep apnea (OSA). We assessed differences overall and across a range of patient demographic characteristics and comorbidities.
Materials and Methods
STUDY POPULATION
The Hyperglycemic Profiles in Obstructive Sleep Apnea (HYPNOS) study was a single-center, randomized clinical trial designed to assess the effect of positive airway pressure therapy on glycemic control. From 2017 to 2019, adults ages 21 to 75 with type 2 diabetes and untreated OSA were recruited from the Baltimore–Washington region of Maryland. Exclusion criteria included HbA1c <6.5% or use of insulin therapy. Participants were interviewed and received clinical examinations and laboratory testing at a screening visit, a baseline visit (approximately 2 to 3 weeks after the screening visit), and a follow-up visit (3 months after the baseline). The institutional review board at the Johns Hopkins School of Medicine approved the study protocols, and all participants provided written informed consent (IRB #00093188). The main trial results were null (4). Further details about the HYPNOS study design are available elsewhere (5). We included all participants with valid measurements of HbA1c who wore CGM devices at both the baseline and 3-month follow-up study visit (n = 144 for Dexcom G4 and n = 138 for Libre Pro).
HBA1C ASSESSMENT
At the study screening visit, trained technicians collected venous blood samples from participants. These samples were analyzed for HbA1c using the Dimension Vista 1500 (Siemens Healthcare Diagnostics) at the University of Maryland. This is an immunoassay-based National Glycohemoglobin Standardization Program–certified method with little evidence of interference from medications or hemoglobinopathies.
CGM ASSESSMENT
Participants simultaneously wore a Dexcom G4 Platinum and Abbott Freestyle Pro Libre for up to 14 days following the baseline and 3-month follow-up (maximum wear time: 28 days). Devices were placed on recommended sites on the body by trained technicians (right abdominal wall for the Dexcom G4 and back of the right arm for the Libre Pro). Following manufacturer recommendations, the Dexcom G4 was calibrated at least twice a day using blood glucose values measured with a FreeStyle InsuLinx glucometer. The Dexcom G4 measured glucose every 5 minutes for up to 7 days, and participants were masked to all glucose readings. Dexcom G4 sensors were replaced after 7 days to collect up to 14 days of continuous information. The Freestyle Libre measured glucose every 15 minutes for up to 14 days. All Freestyle Libre sensors are factory calibrated and did not require manual calibration with fingersticks. CGM devices that were dislodged or mal-functioned before the completion of the wear period were replaced by technicians. We calculated all CGM metrics, including GMI, by combining all 4 weeks of CGM data together.
BASELINE DEMOGRAPHIC CHARACTERISTICS AND COMORBIDITIES
Sociodemographic characteristics (age, sex, race/ethnicity), health behaviors (smoking status), and medical history were collected during face-to-face interviews using a standardized questionnaire. Blood pressure was measured 3 consecutive times on the same arm using a mercury sphygmomanometer. We calculated mean systolic and diastolic blood pressure using all 3 readings and defined hypertension as mean systolic blood pressure ≥130 mm Hg, mean diastolic blood pressure ≥80 mm Hg, or use of blood-pressure lowering medication. A prior history of high cholesterol or cardiovascular disease (congestive heart failure, stroke, coronary artery bypass surgery, myocardial infarction, or angina) was defined by self-report. Weight and height were measured using standardized procedures. We calculated body mass index (BMI; the weight in kilograms divided by the square of the height in meters) and classified participants as normal (BMI of <25 kg/m2), overweight (BMI of 25 to <30 kg/m2), or obese (BMI of ≥30 kg/m2).
STATISTICAL ANALYSES
We examined baseline characteristics and assessed differences in CGM parameters across the 2 sensors. Consistent with the original GMI derivation study (2), we used linear regression models to identify values of HbA1c corresponding to different concentrations of CGM-measured mean glucose. We used linear regression, scatterplots, Pearson correlations, and Bland–Altman plots to characterize the agreement between HbA1c and GMI. We calculated the root mean squared errors from regression models to characterize the typical difference between GMI and HbA1c. In sensitivity analyses, we reexamined agreement (1) after excluding potential outliers, defined as values that were >3 SDs from the mean difference between HbA1c and GMI (6). We examined the mean absolute difference between HbA1c and GMI and calculated the proportion of participants with discordance <0.5 percentage points, 0.5 to 1.0 percentage points, and >1.0 percentage points (<5 mmol/mol, 5–11 mmol/mol, and >11 mmol/mol) (7, 8). We used chi-square and t-tests to assess differences in discordance across participant characteristics and comorbidities. We conducted all analyses using Stata version 17.0 (StataCorp). A 2-sided P value <0.05 was considered statistically significant.
Results
We included 144 participants from the HYPNOS study (mean age: 59.4 years; 45.1% female) (Supplemental Appendix 1). Mean HbA1c was 7.2%, and over two-thirds of participants had obesity, high blood pressure, or high cholesterol. Median CGM wear time was 27 days (IQR 23–29) for the Dexcom G4 and 28 days (IQR 24–29) for the Libre Pro. Mean glucose (approximately 156 mg/d [8.7 mmol/L]), glycemic variability (CV approximately 27%–29%), and mean GMI (approximately 7.0% [53 mmol/mol]) were similar across both devices (Table 1).
Table 1.
CGM parameters during up to 4 weeks of wear time by CGM device (Dexcom G4 or Abbott Libre Pro).
| Dexcom G4 (n = 144) | Libre Pro (n = 138) | |
|---|---|---|
| CGM wear time, days, median (p25, p75) | 27.0 (23.0, 29.0) | 28.0 (24.0, 29.0) |
| Mean glucose, mg/dL, mean (SD) | 155.7 (36.8) | 156.9 (43.0) |
| Median glucose, mg/dL, mean (SD) | 149.9 (36.2) | 150.9 (43.4) |
| SD, mg/dL, median (p25, p75) | 40.2 (33.1, 49.1) | 42.3 (35.4, 54.1) |
| CV, %, mean (SD) | 26.8 (5.6) | 28.9 (6.5) |
| IQR, mg/dL, median (p25, p75) | 52.0 (42.5, 64.5) | 56.0 (44.0, 70.0) |
| % of time glucose ≥200 mg/dL, median (p25, p75) | 11.8 (3.8, 25.1) | 12.3 (4.5, 27.4) |
| % of time glucose ≥180 mg/dL, median (p25, p75) | 19.4 (9.2, 38.1) | 21.5 (9.5, 39.6) |
| % of time glucose ≥140 mg/dL, median (p25, p75) | 50.7 (35.5, 71.0) | 51.2 (32.9, 71.7) |
| % of time glucose 70–180 mg/dL, median (p25, p75) | 79.8 (62.1, 89.9) | 76.7 (58.6, 87.7) |
| % of time glucose 70–140 mg/dL, median (p25, p75) | 49.7 (29.8, 63.4) | 48.8 (28.9, 65.2) |
| % of time glucose <70 mg/dL, median (p25, p75) | 0.4 (0.1, 1.3) | 0.6 (0.0, 2.5) |
| % of time glucose <54 mg/dL, median (p25, p75) | 0.1 (0.0, 0.3) | 0.0 (0.0, 0.4) |
| Glucose Management Indicator, %, mean (SD) | 7.0 (0.9) | 7.1 (1.0) |
To convert mean glucose from mg/dL to mmol/L, multiply by 0.0555.
Mean glucose levels ranged from 120 to 381 mg/dl for the Dexcom G4 and 114 mg/dl (6.3 mmol/L) to 423 mg/dl (23.5 mmol/L) for the Libre Pro (Table 2). For every 1 percentage point increase in HbA1c (e.g.,7% [53 mmol/mol] to 8% [64 mmol/mol]), mean glucose rose by an average of approximately 52 mg/dl (2.9 mmol/L) for the Dexcom G4 and approximately 62 mg/dl (3.4 mmol/L) for the Libre Pro. Mean glucose concentrations corresponding to HbA1c of 7.0%, 7.5%, and 8.0% (53, 58, and 64 mmol/mol) were similar for the Dexcom G4 (146 mg/dl, 172 mg/dl, and 198 mg/dl [8.1, 9.5, and 11.0 mmol/L]) and Libre Pro (145 mg/dl, 176 mg/dl, and 207 mg/dl [8.0, 9.8, 11.5 mmol/L]).
Table 2.
Corresponding values between HbA1c and CGM-derived mean glucose by CGM device (Dexcom G4 or Abbott Libre Pro).
| HbA1c (%) | CGM-derived mean glucose from Dexcom G4a (95% CI), mg/dL | CGM-derived mean glucose from Libre Prob (95% CI) mg/dL |
|---|---|---|
| 6.5 | 120 (112, 128) | 114 (104, 124) |
| 7.0 | 146 (140, 152) | 145 (137, 152) |
| 7.5 | 172 (166, 179) | 176 (168, 184) |
| 8.0 | 198 (190, 207) | 207 (196, 218) |
| 8.5 | 224 (212, 237) | 238 (222, 253) |
| 9.0 | 250 (235, 266) | 269 (248, 289) |
| 9.5 | 277 (257, 296) | 300 (274, 325) |
| 10.0 | 303 (279, 326) | 331 (300, 361) |
| 10.5 | 329 (301, 356) | 361 (326, 397) |
| 11.0 | 355 (323, 386) | 392 (352, 433) |
| 11.5 | 381 (345, 416) | 423 (378, 469) |
To convert HbA1c from percent to mmol/mol, multiply by 10.93 and subtract 23.5.
To convert mean glucose from mg/dL to mmol/L, multiply by 0.0555.
Dexcom G4 linear equation: A1c = 4.196164 + 0.0191796 * (CGM-derived mean-glucose in mg/dl).
Freestyle Libre linear equation: A1c = 4.657133 + 0.0161631* (CGM-derived mean-glucose in mg/dl).
HbA1c was moderately correlated with GMI derived from the Dexcom G4 (r = 0.71) (Fig. 1, A) and Libre Pro (r = 0.68) (Fig. 1, B). The root mean squared errors from linear regression models ranged from 0.66 to 0.69 percentage points (7 to 8 mmol/mol). The mean difference between HbA1c and GMI was low (0.12–0.14 percentage points [approximately 2 mmol/mol]), and bias remained constant across all levels of HbA1c and GMI values (Fig. 1, C and D). Results were similar after excluding values >3 SDs from the mean difference (Supplemental Appendix 2).
Fig. 1.
Scatterplots and Bland–Altman plots of HbA1c and GMI by CGM device (Dexcom G4 or Abbott Libre Pro). To convert HbA1c and GMI from percent to mmol/mol, multiply by 10.93 and subtract 23.5. For the scatterplots, the solid black lines are the lines of identity, and the dashed black lines are based on linear models that regress HbA1c on GMI. For the Bland–Altman plots, the blue dashed lines are the mean difference between HbA1c and GMI, the red dashed lines are the mean difference ±1.96 SD of the differences, the dashed black lines are based on linear models that regress the difference between HbA1c and GMI onto the average of HbA1c and GMI, and the solid black lines are when mean difference between HbA1c and GMI are zero.
Abbreviations: RMSE, root mean squared error.
The mean absolute difference between HbA1c and GMI was higher in the Libre Pro (0.57 percentage points) (9 mmol/mol) compared to the Dexcom G4 (0.49 percentage points) (8 mmol/mol) (Supplemental Appendix 3). However, results did not differ significantly across age, sex, race, or different comorbidities for either device. Results were similar when examining the categories of discordance between GMI and HbA1c (Fig. 2). For the Dexcom G4, 36% of participants had discordance ≥0.5 percentage points (≥5 mmol/mol), and 9% had discordance >1.0 percentage points (>11 mmol/mol). For the Libre Pro, 43% had an absolute difference ≥0.5 percentage points (≥5 mmol/mol), and 18% had a difference >1.0 percentage points (>11 mmol/mol).
Fig. 2.
Frequency histogram of the absolute difference between HbA1c and GMI, by CGM device (Dexcom G4 or Abbott Libre Pro). To convert HbA1c and GMI from percent to mmol/mol, multiply by 10.93 and subtract 23.5.
Discussion
In this study of adults in with type 2 diabetes, GMI was moderately correlated with HbA1c. Between 36% and 43% of participants had clinically significant discordance (≥0.5 percentage points) (≥5 mmol/mol) (7, 8) between HbA1c and GMI, and 9% and 18% had discordance >1 percentage points (>11 mmol/mol). Discordance was higher in the Libre Pro compared to the Dexcom G4, but results were consistent across patient characteristics and comorbidities. Overall, these findings suggest that GMI may be unreliable in patients with type 2 diabetes and should be used with caution in clinical practice.
GMI demonstrated good agreement with HbA1c in the original derivation study (2), which combined data from trials of mostly younger patients with type 1 diabetes (mean age: 44–46) (9–11). Indeed, only 28% of participants had an absolute difference between HbA1c and GMI >0.5 percentage points (≥5 mmol/mol)—a common cutpoint used by to signify clinically significant discordance (7, 8). Moreover, only 3% had a difference >1 percentage points (>11 mmol/mol). However, 2 recent studies failed to replicate these findings. A clinical trial of older adults with type 1 diabetes found that 46% of participants had an absolute difference between GMI and HbA1c >0.5 percentage points (>5 mmol/mol), and 16% had a difference >1 percentage points (>11 mmol/mol) (7). An analysis of middle-age adults with type 1 diabetes in a diabetes clinic also found that 50% had discordance >0.5 percentage points (>5 mmol/mol) and 22% had discordance >1 percentage points (>11 mmol/mol) (8). Extending this literature, we also found much greater discordance between GMI and HbA1c among adults with type 2 diabetes than the derivation study. Considered together, these results suggest that the current GMI equation may not generalize to populations beyond those in the derivation study sample.
HbA1c and GMI are fundamentally different measures of glycemic control. HbA1c is the proportion of all hemoglobin that has glucose attached. Since red blood cells turn over every approximately 90 to 120 days, HbA1c represents an average measure of glycemia over the past 2 to 3 months. HbA1c values are weighted more heavily toward glucose concentrations in the most recent 2 to 4 weeks. On the other hand, GMI is based on mean glucose from CGM sensors. These devices are typically worn for 10 to 14 days and estimate glucose concentration in interstitial fluid every 5 to 15 minutes. These measurement differences likely contribute to some of the observed discordance between GMI and HbA1c.
The correlation between GMI and HbA1c in our study (r = 0.68–0.71) was similar to prior work. For example, Beck et al. (12) reported a correlation of 0.71 to 0.78 between CGM mean glucose and HbA1c. However, the A1c-Derived Average Glucose (ADAG) study reported a much stronger relationship between estimated HbA1c and measured HbA1c (r = 0.92) (13). This analysis relied heavily on fingerstick measurements to calculate mean glucose and included participants with a wider range of HbA1c values, potentially explaining some of the difference. The ADAG study also excluded persons with health conditions (e.g., anemia or hemoglobinopathies) that may affect the mean glucose–HbA1c relationship. Their results suggest that translating CGM measures into accurate estimates of HbA1c requires accounting for interfering factors that may alter the glucose–HbA1c association.
The poor performance of GMI may be related to the equation used to calculate this metric. The GMI equation included CGM-measured mean glucose as the only predictor of HbA1c. However, nonglycemic factors such as red blood cell lifespan are important determinants of HbA1c, and including these in estimation equations may significantly enhance prediction accuracy (14–16). The GMI equation also assumes that the association between CGM-measured mean glucose and HbA1c is the same for all patients, though prior studies have documented substantial heterogeneity across patients (7, 14). These issues can produce substantial prediction error and may account for much of the discordance between HbA1c and GMI.
The goal of GMI is to express CGM-derived mean glucose in terms of HbA1c. However, putting GMI on the same scale as HbA1c may create the impression that the 2 are equivalent, interchangeable measures. Indeed, some have made have this mistake and based important clinical decisions solely on using GMI (17). To clarify this issue, it may be most useful to emphasize mean glucose rather than GMI and to consider HbA1c and mean glucose as separate and complementary tests when communicating to patients and clinicians. Mean glucose is expressed on a different scale than HbA1c, highlighting that the 2 are distinct measures and eliminating the confusion arising from having both a measured and estimated HbA1c. The link between CGM-measured mean glucose and clinical endpoints is still an emerging area of research, with few studies in patients with type 2 diabetes. Therefore, pairing CGM-measured mean glucose with HbA1c—which is strongly linked to clinical outcomes (18, 19)—may be the most useful approach for monitoring both short-term glycemic control and long-term risk of complications (2).
Because mean CGM glucose and HbA1c are measured on different scales, we derived roughly equivalent values of both measures to facilitate comparison. These values were lower than those reported in prior research. For example, HbA1c of 7% (53 mmol/mol) corresponded to a CGM-measured mean glucose of approximately 145 mg/dl (8.0 mmol/L) in our analyses and 154 mg/dl (8.5 mmol/L) in the ADAG and GMI derivation studies (2, 13). Difference in the age of study populations and other clinical characteristics, in particular the use of insulin, may have contributed to this discrepancy. While we examined middle-age and older adults with type 2 diabetes, the derivation study assessed mostly younger patients with type 1 diabetes. Several studies have noted larger differences between HbA1c and CGM-measured mean glucose in older populations, possibly due to differences in red blood cell turnover (7, 16, 20). The ADAG study also excluded persons with health conditions that may interfere with the mean glucose–HbA1c relationship, possibly contributing to differences between studies.
Our study had several limitations. First, all participants had OSA, which may limit the generalizability of our findings. Studies in other populations are needed to validate GMI more comprehensively in type 2 diabetes. Second, our study used an older Dexcom device (G4). Nonetheless, GMI was developed using the Dexcom G4, making our analyses directly comparable to the original derivation study. Third, we may have lacked the power to detect small differences in discordance across subgroups due to our sample size. Fourth, we did not measure nonglycemic factors that might account discordance between HbA1c and GMI. Fifth, CGM data were collected approximately 11 weeks apart due to the design of the trial.
Our study had notable strengths. To our knowledge, this is the largest assessment of the performance of GMI in adults with type 2 diabetes and untreated OSA not using insulin. Participants were mostly middle-age or older adults and came from diverse racial/ethnic backgrounds. Participants wore 2 different types of CGM devices (Dexcom G4 and Libre Pro) simultaneously for up to 4 weeks, allowing for comparison of performance across devices.
In conclusion, GMI was only moderately correlated with measured HbA1c in middle-age and older adults with type 2 diabetes. There was clinically significant discordance in over one-third of participants. These results suggest that GMI may not be a reliable metric of glycemic control and should be interpreted cautiously in clinical practice. We recommend that clinicians rely on mean glucose directly from CGM, not GMI, an estimated HbA1c value.
Supplementary Material
Research Funding:
E. Selvin was supported by National Institutes of Health/National Heart, Lung, and Blood Institute (NIH/NHLBI) Grant K24 HL152440. This work was also supported by NIH/NIDDK Grants R01 DK128837 and R01 DK128900 and NIH/NIA Grant RF1 AG074044. The HYPNOS Trial was supported by NIH/NHLBI Grant R01HL117167. J.B. Echouffo-Tcheugui was supported by NIH/NHLBI Grant K23 HL153774. N.M. Punjabi was supported by NIH/NHLBI Grant R01 HL117167, R01 HL146709, and HL118414. R.N. Aurora was supported by NIH/NHLBI Grant K23 HL118414. Abbott provided CGM and self-monitoring of blood glucose supplies. Dexcom provided CGM systems at a discount.
Role of Sponsor:
The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, preparation of manuscript, or final approval of manuscript.
Nonstandard Abbreviations:
- HbA1c
hemoglobin A1C
- CGM
continuous glucose monitoring
- GMI
glucose management indicator
- OSA
obstructive sleep apnea
- BMI
body mass index
Footnotes
Supplementary Material
Supplementary material is available at Clinical Chemistry online.
Authors’ Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:
Employment or Leadership: R.N. Aurora, American Academy of Sleep Medicine, American Academy of Sleep Medicine—Foundation.
References
- 1.Sacks DB. A1c versus glucose testing: a comparison. Diabetes Care 2011;34:518–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Bergenstal RM, Beck RW, Close KL, Grunberger G, Sacks DB, Kowalski A, et al. Glucose management indicator (GMI): a new term for estimating A1C from continuous glucose monitoring. Diabetes Care 2018;41:2275–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Danne T, Nimri R, Battelino T, Bergenstal RM, Close KL, DeVries JH, et al. International consensus on use of continuous glucose monitoring. Diabetes Care 2017;40:1631–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.ClinicalTrials.gov. Hyperglycemic profiles in obstructive sleep apnea: effects of PAP therapy (HYPNOS)—study results. https://clinicaltrials.gov/ct2/show/results/NCT02454153 (accessed August 2022).
- 5.Rooney MR, Aurora RN, Wang D, Selvin E, Punjabi NM. Rationale and design of the Hyperglycemic Profiles in Obstructive Sleep Apnea (HYPNOS) trial. Contemp Clin Trials 2021;101:106248. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Healy M Outliers in clinical chemistry quality-control schemes. Clin Chem 1979;25:675–7. [PubMed] [Google Scholar]
- 7.Toschi E, Slyne C, Sifre K, O’Donnell R, Greenberg J, Atakov-Castillo A, et al. The relationship between CGM-derived metrics, A1C, and risk of hypoglycemia in older adults with type 1 diabetes. Diabetes Care 2020;43:2349–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Perlman JE, Gooley TA, McNulty B, Meyers J, Hirsch IB. Hba1c and glucose management indicator discordance: a real-world analysis. Diabetes Technol Ther 2021;23:253–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Heinemann L, Freckmann G, Ehrmann D, Faber-Heinemann G, Guerra S, Waldenmaier D, et al. Real-time continuous glucose monitoring in adults with type 1 diabetes and impaired hypoglycaemia awareness or severe hypoglycaemia treated with multiple daily insulin injections (HypoDE): a multicentre, randomised controlled trial. Lancet 2018;391: 1367–77. [DOI] [PubMed] [Google Scholar]
- 10.Aleppo G, Ruedy KJ, Riddlesworth TD, Kruger DF, Peters AL, Hirsch I, et al. REPLACE-BG: a randomized trial comparing continuous glucose monitoring with and without routine blood glucose monitoring in adults with well-controlled type 1 diabetes. Diabetes Care 2017;40:538–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Beck RW, Riddlesworth T, Ruedy K, Ahmann A, Bergenstal R, Haller S, 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;317:371–8. [DOI] [PubMed] [Google Scholar]
- 12.Beck RW, Bergenstal RM, Cheng P, Kollman C, Carlson AL, Johnson ML, et al. The relationships between time in range, hyperglycemia metrics, and HbA1c. J Diabetes Sci Technol 2019;13:614–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Nathan DM, Kuenen J, Borg R, Zheng H, Schoenfeld D, Heine RJ, et al. Translating the A1C assay into estimated average glucose values. Diabetes Care 2008;31:1473–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Malka R, Nathan DM, Higgins JM. Mechanistic modeling of hemoglobin glycation and red blood cell kinetics enables personalized diabetes monitoring. Sci Transl Med 2016;8:359ra130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Xu Y, Hirota Y, Ajjan RA, Yamamoto A, Matsuoka A, Ogawa W, et al. Accurate prediction of HbA1c by continuous glucose monitoring using a kinetic model with patient-specific parameters for red blood cell lifespan and glucose uptake. Diab Vasc Dis Res 2021;18:14791641211013734. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Xu Y, Dunn TC, Ajjan RA. A kinetic model for glucose levels and hemoglobin A1C provides a novel tool for individualized diabetes management. J Diabetes Sci Technol 2021;15:294–302. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Pluchino KM, Wu Y, Silk AD, Yi J, Lias CH. Comment on Bergenstal et al. Glucose management indicator (GMI): a new term for estimating A1C from continuous glucose monitoring. Diabetes Care 2018;41: 2275–80 (Letter). Diabetes Care 2019;42: e28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Diabetes Control Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. NEJM 1993;329:977–86. [DOI] [PubMed] [Google Scholar]
- 19.UK Prospective Diabetes Study Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet 1998;352:837–53. [PubMed] [Google Scholar]
- 20.Munshi M, Segal A, Slyne C, Samur A, Brooks K, Horton E. Shortfalls of the use of HbA1C-derived eAG in older adults with diabetes. Diabetes Res Clin Pract 2015;110:60–5. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.