Glucose Abnormalities Detected by Continuous Glucose Monitoring in Very Old Adults With and Without Diabetes

Continuous glucose monitoring (CGM) has revolutionized the lives of people with diabetes. The latest generation of CGM devices are minimally invasive and no longer require calibration with fingerstick glucose. CGM systems report glucose measurements every minute or every few minutes and provide the wearer with comprehensive information on glucose patterns. In people with diabetes, CGM reports can provide important insights into glucose abnormalities not reflected in a hemoglobin A_1c (HbA_1c) test (1). In March 2024, the U.S. Food and Drug Administration approved the first over-the-counter CGM system (2). While there is growing interest in using CGM systems in individuals with diabetes, the value of CGM systems in those without diabetes is unknown (3).

Age is one of the most important risk factors for type 2 diabetes, and older adults have a high burden of diabetes and prediabetes (4). Even in the absence of diabetes, older adults are less able to maintain normal glucose homeostasis than their younger counterparts (5). It is possible that older adults experience glucose abnormalities that are undetected by standard diagnostic tests (3,6–8). To identify undetected cases of diabetes, the 2019 Endocrine Society guidelines for the treatment of diabetes in older adults recommend screening with a 2-h oral glucose tolerance test (OGTT) all adults aged ≥65 years who have an HbA_1c or fasting glucose in the prediabetes range (9). CGM systems provide an opportunity to determine the prevalence of undetected hyperglycemia or other abnormal glucose patterns in older adults without diabetes.

We undertook this study to examine the prevalence of CGM-defined glucose abnormalities in a large, community-based population of very old adults (>75 years) who participated in the Atherosclerosis Risk in Communities (ARIC) Study. We characterized glucose patterns and the frequency of CGM-defined abnormalities in individuals with prediabetes (HbA_1c 5.7% to <6.5%) and compared the rates of glucose abnormalities with those detected in individuals with normoglycemia (HbA_1c <5.7%) or diabetes (prior diagnosis or HbA_1c ≥6.5%). We evaluated the ability of the CGM system to detect hyperglycemia not reflected in HbA_1c in older adults without diabetes.

Participants from the ARIC Study were recruited from four communities in the U.S.: Forsyth County, NC; Washington County, MD; Minneapolis, MN; and Jackson, MS. For this study, we used data from visit 9 (2021–2022), when all participants were offered CGM regardless of diabetes status. Among the 1,787 eligible participants, 1,319 (74%) consented to wear a CGM for up to 2 weeks. At the end of the 2-week period, 1,241 participants (94%) returned their sensors to the clinic. We were able to successfully download data from 1,216 sensors. We excluded participants missing HbA_1c (n = 45) and participants who had <3 days of CGM data (n = 18). We excluded three participants with possibly faulty sensors because their CGM glucose readings were unusually high given no other evidence of diabetes (i.e., repeated tests of normal HbA_1c, glucose, fructosamine, and glycated albumin at the current and past visits, no history of diabetes, and no history of diabetes medication use). Our final study population consisted of 1,150 participants.

We used the Abbott FreeStyle Libre Pro (Abbott Diabetes Care) CGM system to measure glucose for up to 14 days. This blinded system is factory calibrated (no finger stick needed) and records interstitial glucose every 15 min.

We calculated the percentage of CGM readings for each participant defined as the number of sensor readings over the number of potential readings given the time the CGM was worn. For each participant, we calculated CGM mean glucose, the coefficient of variation (CV), SD, percentage of time >140 mg/dL, percentage of time >180 mg/dL, percentage of time spent in range (70–180 mg/dL), and percentage of time spent in tight range (70–140 mg/dL) (10,11). We examined hyperglycemic excursions, defined as the number of episodes of sustained glucose >140 mg/dL or 180 mg/dL for >15 min (at least two consecutive readings). We derived the total glucose area under the curve (AUC) >140 mg/dL, which integrates the amplitude and duration of hyperglycemia. We also calculated AUC >180 mg/dL and the mean amplitude of glycemic excursions using the Baghurst algorithm (12).

We calculated percentage of time <70 mg/dL and excluded hypoglycemic episodes ≤40 mg/dL for ≥1 h as we these may reflect unrealistic hypoglycemic episodes (e.g., compression lows).

We defined diagnosed diabetes based on a self-reported diagnosis of diabetes by a health care provider, or use of diabetes medication, and undiagnosed diabetes based on an HbA_1c ≥6.5%. In participants without diabetes, we defined prediabetes as an HbA_1c of 5.7% to <6.5% and normoglycemia as an HbA_1c of <5.7%.

We examined the characteristics and CGM metrics in participants with normoglycemia, prediabetes, and diabetes (diagnosed or undiagnosed). We visually displayed the distributions of CGM metrics using kernel density plots. We also generated summary 24-h ambulatory glucose profiles (AGP) with Loess smoothing and plotted the 5th, 25th, 50th, 75th, and 95th percentiles of the glucose distribution. The summary AGP combined all daily glucose profiles from participants within each diabetes category (normoglycemia, prediabetes, and diabetes) into a single 24-h period.

We calculated the differences and used t tests to compare CGM metrics and the prevalence of abnormalities among participants with normoglycemia, prediabetes, and diabetes.

We evaluated the associations of HbA_1c and CGM metrics (mean glucose, time above range, and CV) in participants without diabetes (i.e., normoglycemia or prediabetes). Currently, no cut points have been established for CGM metrics for individuals without diabetes. We defined glucose abnormalities for participants without diabetes using lower cut points than the CGM international consensus recommended cut points for individuals with diabetes. We defined normal CGM mean glucose as ≤120 mg/dL, normal percentage time spent >140 mg/dL as <20%, and normal CV as ≤30%. These cut points were consistent with prior studies that characterized the CGM profiles in individuals without diabetes (3). We compared the prevalence of CGM abnormalities among those without diabetes using the above definitions with the prevalence of CGM abnormalities among participants with diagnosed diabetes using the CGM international consensus recommended cut points for percent time spent >180 mg/dL (<10%) and CV (≤36%) for the latter. We defined a well-controlled CGM mean glucose for participants with diagnosed diabetes as ≤150 mg/dL because there is no recommendation for this metric.

All analyses were conducted using Stata 18.0 software (StataCorp). We developed and used a Stata package, “cgmstats” to output all CGM metrics.

Among 1,150 very old adults (mean age 83 years, SD 4; 59% women; 26% who are Black), 35.1% had normoglycemia, 34.5% had prediabetes, and 30.4% had diabetes (Table 1). Among participants with diabetes, 27% were on insulin or sulfonylurea, 50% were taking other diabetes medication, and 23% were not taking any diabetes medication. The mean HbA_1c among those with normoglycemia, prediabetes, and diabetes were 5.4% (SD, 0.22), 5.9% (SD, 0.20), and 6.8% (SD, 0.90), respectively. Because the number of participants with undiagnosed diabetes was small (n = 29) and their glucose profiles were more similar to those with diagnosed diabetes than prediabetes, we combined undiagnosed and diagnosed diabetes into one category (Supplementary Table 1).

The overall mean percentage of CGM readings was 99.6% (SD 1.9) and was similar by diabetes status. On average, the glucose profiles for participants with prediabetes were similar to those with normoglycemia. Participants with diabetes had higher median glucose and a more variable glucose profile than participants with prediabetes (Fig. 1).

The CGM metrics comparing participants with prediabetes to those with normoglycemia were generally similar, with some small differences (Table 1 and Supplementary Table 2). The median (25th percentile–75th percentile) CGM mean glucose among participants with prediabetes was 102 mg/dL (93–110), compared with 97 mg/dL (90–105) among participants with normoglycemia (P < 0.001) (Fig. 2A). The median variability in CGM glucose was similar among those with normoglycemia and those with prediabetes (CV of 20% for both groups, P = 0.25) (Fig. 2B). Participants with prediabetes spent very little time in hyperglycemia (median of 5.8% of time with CGM glucose >140 mg/dL), as did those with normoglycemia (median of 3.4% of time with CGM glucose >140 mg/dL) (P < 0.001) (Fig. 2C). Participants with prediabetes had a median number of 17 excursions >140 mg/dL compared with 10 excursions >140 mg/dL among those with normoglycemia (P < 0.001) (Supplementary Fig. 1). AUC >140 mg/dL was higher among participants with prediabetes (mean 1.40 mg/dL/day, SD 1.93) compared with those with normoglycemia (mean 0.83 mg/dL/day, SD 1.40; P < 0.001) (Supplementary Fig. 2). Participants with normoglycemia and those with prediabetes spent, on average, 87.2% and 86.8% of their time in tight range (70–140 mg/dL), respectively (P = 0.64).

Differences in CGM metrics between participants with diabetes and those with prediabetes were statistically significant and large relative to the differences between participants with normoglycemia and prediabetes (Table 1 and Supplementary Table 2). No participants with normoglycemia or prediabetes had a CGM mean glucose >140 mg/dL, while 32.7% of participants with diabetes had a CGM mean glucose >140 mg/dL (P < 0.001).

The prevalence of glucose abnormalities detected by CGM in very old adults without diabetes was low. Among participants without diabetes, 10.5% had one or more CGM glucose abnormalities. Specifically, among participants with normoglycemia (n = 404), 3.0% had a mean glucose >120 mg/dL, 3.0% spent ≥20% of time with CGM glucose >140 mg/dL, and 3.5% had a CV >30% (Supplementary Fig. 3). Among participants with prediabetes (n = 397), 7.3% had a mean glucose >120 mg/dL, 7.1% spent ≥20% of time with CGM glucose >140 mg/dL, and 5.8% had a CV >30% (Supplementary Fig. 3). Among participants with diagnosed diabetes, 46.9% had one or more CGM glucose abnormalities. Specifically, 23.8% had a mean glucose >150 mg/dL, 44.4% spent ≥10% of time with CGM glucose >180 mg/dL, and 13.4% had a CV >36% (Supplementary Fig. 3).

In this community-based study of 1,150 adults >75 years of age, we examined glucose patterns of older adults over a 14-day period and found that the percentage of time spent in hyperglycemia (>140 mg/dL) and the number of hyperglycemic excursions was low in very old adults with prediabetes or normoglycemia defined by HbA_1c. Older adults with prediabetes had CGM glucose profiles and measures of glycemic variability that were much more similar to their counterparts with normoglycemia than to those with diabetes.

In older adults without diagnosed diabetes, we did not find strong evidence that CGM provides information on hyperglycemia beyond HbA_1c. For participants without diabetes, we used lower cut points than the CGM international consensus recommendations for defining hyperglycemia and time in range. Despite this, among individuals with normoglycemia or prediabetes, 6.7% and 14.4%, respectively, CGM detected one or more glucose abnormalities, compared with 46.9% among participants with diagnosed diabetes. For most individuals in our study, 2 weeks of CGM wear does not appear to detect significant hyperglycemia that would not already be identified using current clinical cut points for HbA_1c.

Reference ranges and CGM targets in individuals without diabetes are not established (6). We found that participants with normoglycemia and prediabetes spent 87.2% and 86.8% of their time, respectively, with CGM glucose 70–140 mg/dL. These results are consistent with a prior study which found that adults aged ≥60 years without diabetes (n = 25), spent a median of 93% of their time with CGM glucose 70–140 mg/dL (6). We found that older adults in ARIC with normoglycemia and prediabetes spent 5% and 7.8% of their time, respectively, with CGM glucose >140 mg/dL, and both groups had an average CV of 21%. Additional studies in individuals without diabetes, using different CGM systems and examining associations with clinical outcomes relevant to older adults (e.g., quality of life, frailty, mortality), will be helpful to establish and validate CGM reference ranges in normoglycemia and prediabetes (13,14).

Participants with prediabetes did not experience significant time in hyperglycemia in this study. Our findings support prior research which found that the additional screening in older adults, such as OGTT testing in all older adults with prediabetes recommended in 2019 by the Endocrine Society, may offer little if any benefit (15). OGTT is cumbersome because it requires an 8-h fast, commitment of nursing staff, and is poorly reproducible (16). Endocrine Society guidelines recommend 2-h OGTT screening in all adults aged ≥65 years who have an HbA_1c or fasting glucose in the prediabetes range. The rationale provided is that in older adults, an OGTT may detect diabetes earlier than the standard measures of HbA_1c and fasting glucose (9). While we did not conduct OGTT assessments in ARIC, the CGM data provided an opportunity to determine the prevalence of undetected hyperglycemia or other abnormal glucose patterns in participants with prediabetes over the span of 14 days. Our findings suggest that HbA_1c adequately captures the burden of hyperglycemia and that OGTT testing is unlikely to reveal substantial numbers of people with undiagnosed diabetes. Our results do not support additional screening with OGTT in older adults with prediabetes defined by HbA_1c.

Our study has several strengths. This study is one of the largest to characterize and compare CGM patterns among a population of community-dwelling older adults (age range, 76–98 years old), with and without diabetes, using standardized CGM data collection. Half of adults aged ≥65 years in the U.S. have prediabetes based on fasting glucose and/or HbA_1c (4), but information on CGM-defined glucose in this population is scarce (17). Few prior CGM studies have included older adults, especially those without diabetes. Our findings from participants during late life are novel and provide a unique insight into an understudied population. Lastly, we obtained ∼14 days of CGM data from a blinded CGM sensor on nearly all participants. Prior studies have had much shorter (3–7 days) CGM wear periods (18–20).

This study has a few limitations. While no lifestyle modifications or interventions were implemented and we used blinded CGM sensors, participants may have altered their eating and/or activities, knowing that glucose levels were being monitored. No fasting glucose or 2-h OGTT assessments were available (21). The Abbott Libre Pro is no longer commonly used in practice; however, we do not believe the relative differences between the groups based on diabetes status, would materially change using a different CGM system. Lastly, the Abbott Libre Pro device may inaccurately indicate hypoglycemia (22). To address this, we excluded unrealistic hypoglycemic episodes from our analyses; however, there may still be false readings of hypoglycemia. Again, we believe this bias would affect the three groups similarly and thus would not materially change our conclusions.

In conclusion, we found the prevalence of CGM abnormalities among very old adults with prediabetes was low and similar to adults with normoglycemia. Very old adults without diabetes (including those with prediabetes) spent a small percentage of their time in hyperglycemia and had significantly lower CGM mean glucose compared with those with diabetes. Overall, our findings suggest that HbA_1c will typically provide adequate information on the burden of hyperglycemia in older adults.

Acknowledgments. The authors thank the staff and participants of the ARIC study for their important contributions.

E.S. and J.B.E.-T. are editors of Diabetes Care but were not involved in any of the decisions regarding review of the manuscript or its acceptance.

Funding. The ARIC study has been funded in whole or in part with funds from the National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health (NIH), and Department of Health and Human Services (contract nos. 75N92022D00001, 75N92022D00002, 75N92022D00003, 75N92022D00004, and 75N92022D00005). N.R.D. was supported by the NIH/NHLBI grant T32 HL007024. M.F. was supported by the NIH/National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant K01 DK138273. G.W. was supported by the NIH/National Institute on Aging (NIA) grant R01 AG054787. J.B.E.T. was supported by the NIH/NHLBI grant K23 HL153774. E.S. was supported by NIH/NHLBI grants K24 HL152440 and R01 HL158022. This research was also supported by NIH/NIDDK grants R01 DK128837 and R01 DK128900, and NIH/NIA grant RF1 AG074044 to E.S.

Abbott Diabetes Care provided CGM systems for this investigator-initiated research.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. N.R.D. drafted the initial manuscript. N.R.D., M.F., D.W., A.V., B.G.W., J.C., J.B.E-T., and E.S. contributed to the study concept and design and interpretation of the data. All authors approved the final version of the manuscript. N.R.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the American Heart Association Epidemiology, Prevention, Lifestyle & Cardiometabolic Health Scientific Sessions 2023, Boston, MA, 28 February–3 March 2023.

Handling Editors. The journal editor responsible for overseeing the review of the manuscript was M. Sue Kirkman.