Burden and risk factors of CGM-detected hypoglycemia in older adults with type 2 diabetes

Abstract

Background:

Comprehensive data on the burden for biochemical hypoglycemia in older adults with type 2 diabetes are lacking. We seek to characterize the burden and risk factors for hypoglycemia detected by continuous glucose monitoring (CGM) in older adults with type 2 diabetes.

Methods:

A cross-sectional analysis of 315 older adults with type 2 diabetes who attended visit 9 of the Atherosclerosis Risk in Communities Study (2021-22). We examined rates of Level 1 hypoglycemia (<70 mg/dL) and percentage meeting recommended targets for hypoglycemia (<1% of time). We identified risk factors for CGM-detected hypoglycemia by comparing the median time spent in hypoglycemia across subgroups. All analyses were stratified by high-risk medication use (insulin/sulfonylureas vs not).

Results:

Of the 315 participants with type 2 diabetes (mean age 83 years, 55% women, 41% Black adults), 32.4% were using insulin or sulfonylureas. Among individuals using these high-risk medications, the median time spent in hypoglycemia was 3.4%, and ~66% of participants spent more than 1% of the time with CGM glucose <70 mg/dL. Hypoglycemic episodes typically occurred overnight and lasted a median of ~1.5 hours among individuals using insulin or sulfonylureas. CGM-detected hypoglycemia was low in participants not using high-risk medications (median time spent in hypoglycemia: 0.7%). In unadjusted analyses, cardiovascular disease, cognitive impairment, poor physical functioning, and chronic kidney disease were associated with increased time spent in hypoglycemia, regardless of high-risk medication use.

Conclusions:

There may be a substantial burden of unrecognized CGM-detected hypoglycemia in very old adults with type 2 diabetes using high-risk medications in the general population. Hypoglycemia may also be present among persons not on high-risk medication, but the burden is much lower. Further research is needed to clarify the clinical significance of these hypoglycemic episodes, especially in persons not on high-risk medication.

Introduction

Older adults with type 2 diabetes have an elevated risk of severe hypoglycemia, particularly when treated with insulin or sulfonylureas. In the US, approximately half of adults with diabetes over 65 use these hypoglycemia-causing medications.(1) Clinical guidelines recommend routine hypoglycemia screening for older patients using insulin or sulfonylureas, with frequent mild episodes suggesting potential overtreatment and a need for deintensification.(2, 3)

Prior studies suggest that the rates of biochemical hypoglycemia (glucose <70 mg/dL) are ~35 to 40% in older adults with type 2 diabetes using high risk medications.(4, 5) However, existing research has relied on self-report or data from traditional glucose meters, potentially resulting in misclassification and underestimation.(6) Comprehensive data on the burden for biochemical hypoglycemia in older adults with type 2 diabetes are lacking, and little is known about major risk factors.

Continuous glucose monitoring (CGM) systems typically record glucose every 1 to 15 minutes and are recommended to identify biochemical hypoglycemia.(3) Trials of older adults with type 2 diabetes using basal insulin have found a high burden of CGM-defined hypoglycemia, especially during nighttime hours.(4, 7, 8) However, community-based studies of CGM-defined hypoglycemia in older patients are lacking, particularly for people with type 2 diabetes using high risk medications.(9)

Some data suggest older patients with type 2 diabetes may experience biochemical hypoglycemia, even without the use of high-risk medication. One pilot study found that in a community-based population of very old adults, low readings on CGM were common among persons with and without diabetes.(10) However, few studies have examined biochemical hypoglycemia in older adults with diabetes who are not treated with high-risk medications.

Our primary objective was to characterize the burden, timing, and duration of CGM-detected hypoglycemia in older adults with type 2 diabetes treated and not treated with high-risk glucose-lowering medications. Our secondary goal was to identify potential risk factors for biochemical hypoglycemia in this population. To accomplish these goals, we collected and analyzed CGM data from a large, community-based cohort of very old adults with type 2 diabetes in the Atherosclerosis Risk in Communities (ARIC) study.

Methods

Study population

The ARIC study recruited participants in 1987-1989 from four U.S. communities: Forsyth County, North Carolina; Jackson, Mississippi; suburbs of Minneapolis, Minnesota; and Washington County, Maryland. Over the past three decades, ARIC participants have undergone 11 study visits consisting of clinical examinations, medical interviews and laboratory tests. Study visit 12 is ongoing, and further details about the ARIC study are available elsewhere.(11)

For this study, we used data from visit 9 (2021-22), when all participants were offered a CGM, regardless of diabetes status. Of the 1,787 eligible participants, 1,319 (74%) consented to wear a CGM and 1,241 (94%) participants returned their sensors to the clinic. Of these participants, 324 participants had diabetes, and we excluded participants with less than three days of CGM data (n=9). The final analytical sample included 315 very old adults (76-97 years old).

Diabetes status and treatment

We defined diabetes as a self-report of a diagnosis of diabetes within the past 5 years, use of glucose-lowering medication at any time during the prior 10 years, or HbA1c≥6.5%. We further classified participants as using or not using insulin or sulfonylureas.

Specifically, diabetes medication use was defined as the use of any diabetes medication in the four weeks prior to the study clinic visit and was based on review of medication containers brought to the study visit by participants. The class of diabetes medications was defined as any use of that specific medication type in the previous four weeks assessed by container review.

Continuous glucose monitoring

Trained technicians placed the Abbott Libre Freestyle Pro on participants during clinical study visits. Sensors were placed on the back of the right arm, per manufacturer recommendations.(12) The Libre Pro was factory calibrated (did not require finger sticks) and recorded interstitial glucose every 15 minutes for up to 14 days. Glucose readings were masked (participants were unaware of the glucose readings during the data collection period). After the completion of the wear period, participants returned sensors with a prepaid mailer.

CGM-detected hypoglycaemia

Consistent with clinical guidelines, we defined Level 1 hypoglycemia as CGM-glucose <70 mg/dL. Level 1 hypoglycemia is an alert threshold(3) and the clinical recommendation for Level 1 hypoglycemia is <1% of time with CGM-glucose <70 mg/dL.

We defined a hypoglycemic episode as having at least two consecutive (>15 minutes) CGM-glucose <70 mg/dL. A hypoglycemic episode was considered “resolved” when CGM glucose was ≥70 mg/dL. While guidelines do not define a threshold for the frequency or duration of hypoglycemic episodes, they do recommend minimizing the occurrence of these episodes, especially in older adults.(2)

We also explored Level 2 hypoglycemia, defined as CGM-glucose <54 mg/dL (considered clinically significant and requiring immediate attention) to examine how patterns of severe hypoglycemia may differ by high-risk medication use.

Risk factors

We selected potential risk factors for CGM-detected hypoglycemia based on clinical guidelines and prior studies focused on severe hypoglycaemia.(3, 13)

Demographic information (age, sex, race) was self-reported. Prevalent cardiovascular disease was defined as a history of coronary heart disease, stroke, or heart failure at or prior to visit 9. Cardiovascular events were identified through active surveillance of hospitalizations, annual telephone interviews, and review of hospital discharge records. All potential cardiovascular events were adjudicated by a panel of experts.

History of severe hypoglycemia was defined using International Classification of Diseases (ICD) 9/10 codes for severe hypoglycemia from hospital records or CMS claims through December 31, 2020.(13) Frailty was defined as having three or more of the following clinical features obtained via self-report or observation: low energy, low physical activity, low strength, slowed gait speed, and unintentional weight loss or low BMI.(14) Functional status was assessed using the Short Physical Performance Battery. Poor physical functioning was defined as a Short Physical Performance Battery score <7. Dementia status and mild cognitive impairment (MCI) were defined by an expert adjudication panel using a combination of neurocognitive testing, annual or semiannual telephone interviews and in-person interviews with participants and informants, and diagnostic codes in hospital records.(15) Chronic kidney disease (CKD) was defined as an estimated glomerular filtration rate (eGFR) based on creatinine <60 mL/min/1.73 m² or albumin-to-creatinine ratio <30 mg/g.

Statistical analyses

All analyses were stratified by high-risk medication use (insulin or sulfonylureas). We summarized demographic and clinical risk factors. We characterized the distribution of time spent with Level 1 hypoglycemia and calculated the percentage of participants that met recommended targets for hypoglycemia (<1% of time with CGM glucose <70 mg/dL). We examined Level 1 hypoglycemia during each hour of the day (i.e., over a 24-hour period) and calculated the median time spent in hypoglycemia during the day (6:00am-9:59pm) vs. night (10:00pm-5:59am). Sleep and wake times were not ascertained during the study visit. Consistent with prior CGM studies, we assumed sleep time ranged from 10pm-5:59am.(16) We also summarized the number and duration of hypoglycemic episodes (<70 mg/dL).

In exploratory analyses, we examined differences in the median percent time spent in hypoglycemia by potential risk factors, including cardiovascular disease, dementia or MCI, poor physical functioning, frailty, history of severe hypoglycemia, and CKD.

In sensitivity analyses, we repeated all analyses using the threshold for Level 2 hypoglycemia (<54 mg/dL). We also excluded episodes of hypoglycemia that may be erroneous (so called “compression lows” or other erroneous readings), defined as a sustained glucose ≤40 mg/dL for ≥ 1 hour.(17) We characterized the distribution and timing of percent time spent in Level 1 hypoglycemia, after excluding these likely erroneous episodes of hypoglycemia.

Results

We included 315 participants with type 2 diabetes, 32.4% (n=102) treated with insulin or sulfonylureas (Table 1). Among these 102 participants, 59.8% were on sulfonylurea only, 36.3% were on insulin only and 3.9% were on both. Among insulin users, 13.5% were using rapid acting insulin (Supplemental Table 1). Among participants not treated with insulin or sulfonylureas, 60.1% of them were on one or more other diabetes medication (Supplemental Table 2). Mean age was similar across both groups (~83 years), but participants using insulin or sulfonylureas were less likely to have HbA1c <7% and more likely to have dementia. Median CGM wear time was high in both groups (13.9 out of a maximum of 14 days).

Table 1.

Participant characteristics according to use of high-risk glucose-lowering agents, older adults with a history of diabetes, ARIC Visit 9 (2021-22).

	Insulin or sulfonylurea use (n=102)	No insulin or sulfonylurea use (n=213)
Age, years (SD)	82.6 (0.4)	82.5 (0.3)
Female, %	51.0	57.3
Black adults	33.3	45.5
HbA1c<7%, %	33.3	71.4
Duration of diabetes ≥10 years	21.6	15.1
History of cardiovascular disease	24.5	23.9
Dementia or mild cognitive impairment	20.6	19.7
Poor physical functioning	24.5	19.2
Frailty	9.2	6.8
Diabetes Medication use, %
Insulin and sulfonylurea	3.9	-
Insulin only	36.3	-
Sulfonylurea only	59.8	-
Other	-	60.1
None	-	39.9

Displaying mean (SD) or %, unless otherwise noted

Poor physical functioning was defined as a Short Physical Performance Battery score <7.

Frailty was defined as the have three or more of the following clinical features: low energy, low physical activity, low strength, slowed motor performance, and unintentional weight loss.

The median time spent with hypoglycemia (CGM-glucose <70 mg/dL) was significantly higher in persons using (versus not using) insulin or sulfonylureas (3.4% versus 0.7% time spent with hypoglycemia) (Table 2). More than half of the participants not treated with these medications spent <1% of the time with hypoglycemia, compared to 35% of those using these medications (Figure 1).

Table 2.

Descriptive statistics of Level 1 hypoglycemia (CGM glucose <70 mg/dL) by high-risk medication use (insulin/sulfonylurea).

	Insulin or sulfonylurea use (n=102)	No insulin or sulfonylurea use (n=213)
Level 1 hypoglycemia (CGM glucose <70 mg/dL)
Percent time, median (p25, p75)	3.4 (0.4, 9.0)	0.7 (0, 3.3)
Number of episodes*/2 weeks, mean (SD)	8.1 (8.6)	6.4 (10.2)
Median (p25, p75) number of hypoglycemic episodes*/2 weeks	5 (1, 12)	2 (0, 7)
% with ≥1 episode /2 weeks	78.4	70.4
% with ≥2 episode /2 weeks	71.6	55.9
% with ≥3 episode /2 weeks	63.7	46.0
Median (p25, p75) duration of episodes*, minutes [no.]	90 (45, 195) [n=822]	60 (30, 120) [n=1,356]

^*A hypoglycemic episode is defined as having at least two consecutive (>15 minutes) CGM-glucose <70 mg/dL.

Figure 1.

Distribution of percent time spent in Level 1 hypoglycemia (CGM glucose <70 mg/dL) by high-risk medication use. Dashed red line indicates 1% recommended target for time spent in hypoglycemia. Level 1: 35% of those on insulin or sulfonylurea and 56% of those not on insulin or sulfonylurea meeting recommendation

Time spent in hypoglycemia was low during daytime but increased substantially after 10:00PM. Hypoglycemia (<70 mg/dL) peaked between the hours of 2:00 AM and 6:00AM; the median time spent in hypoglycemia during daytime vs. nighttime was ~0.5% vs. ~0.8% for those not using insulin or sulfonylureas and ~2% vs. ~4.5% for those using these medications (Figure 2, Supplemental Table 3).

Figure 2.

Timing of percent time spent in Level 1 hypoglycemia over a 24-hour period by high-risk medication use. The solid line indicates the median time spent in hypoglycemia and the shading indicates the 25^th to 75^th percentiles.

More hypoglycemic episodes occurred in participants using (versus not using) insulin or sulfonylurea (median number of episodes: 5 versus 2) (Table 2). Over 60% of participants using high-risk medications had three or more episodes during the 2-week wear period compared to 45% of participants not using these medications. The typical episode lasted 1.5 hours among participants using high-risk medications, compared to 1 hour among participants not using these medications (Table 2).

Participants with cardiovascular disease, cognitive impairment, poor physical functioning, frailty, or chronic kidney disease, spent more time in hypoglycemia, regardless of insulin or sulfonylurea use (Figure 3, Supplemental Table 4). Associations were similar when examining hypoglycemia that occurred during daytime and overnight.

Figure 3.

Association of risk factors of hypoglycemia with CGM defined time spent in Level 1 hypoglycemia (<70 mg/dL), (a) Insulin or sulfonylurea use, (b) No insulin or sulfonylurea use.

When using a lower CGM-glucose cut point (<54 mg/dL), participants using insulin or sulfonylureas experienced more hypoglycemic episodes for a longer period than persons not using these medications (Supplemental Table 5). Only 15% of the participants not treated with these medications spent more than the recommended 1% of the time with hypoglycemia, compared to 37% of those using high-risk medications (Supplemental Figure 1). Time spent in Level 2 hypoglycemia was restricted to nighttime and early morning hours among persons not using insulin or sulfonylurea (12:00 AM – 6:00 AM) and extended slightly longer into the morning among persons using insulin or sulfonylurea (11:00 PM – 9:00 AM) (Supplemental Figure 2).

The results of our sensitivity analyses after excluding potentially erroneous hypoglycemic episodes (140 episodes, median length of 2 hours) were very similar to our main findings (Supplemental Figures 3 and 4).

Discussion

In this community-based study of very old adults with type 2 diabetes, two-thirds of participants using insulin or sulfonylureas spent more than 1% of their time with CGM-glucose <70 mg/dL. Hypoglycemic episodes generally occurred overnight and typically lasted ~1.5 hours among persons on these high-risk medications. Overall, these results suggest there may be a substantial burden of potentially unrecognized hypoglycemia in older adults with type 2 diabetes using high-risk medications in the general population.

Rates of CGM-detected hypoglycemia in our analyses were higher than those reported in prior research. In studies of individuals with type 2 diabetes, the median time with CGM glucose <70 mg/dL ranged from ~0.3 to 0.8%.(7, 18, 19) In contrast, the median time in hypoglycemia was 3.4% in participants using insulin or sulfonylureas in our study. This difference is likely related to differences in age. Participants in ARIC were significantly older (mean age: 83 years) than participants in prior CGM studies of individuals with type 2 diabetes (mean age ~60 years). Even in patients using less intensive insulin therapy or sulfonylureas, our results suggest there may be a major burden of unrecognized hypoglycemia.

The clinical relevance of CGM-detected hypoglycemic episodes in patients with type 2 diabetes remains unclear. Clinical guidelines suggest that frequent episodes of biochemical hypoglycemia are an indication of overtreatment and may warrant less intensive therapy.(3) However, CGM sensors are less accurate at the low range and are designed to over-detect low glucose values for patients’ safety. Data artifacts, including compression lows, may occur frequently.(12, 20) Emerging evidence suggests that CGM-detected hypoglycemia is predictive of severe hypoglycemia in persons with type 1 diabetes.(21) More research is needed to determine if this is also true in persons with type 2 diabetes. In clinical practice, it is recommended to confirm CGM-detected hypoglycemia with fingerstick glucose or symptoms before changing the treatment plan in older adults with type 2 diabetes.(3, 12) Longitudinal studies are needed to determine whether CGM-detected hypoglycemia at current thresholds predicts adverse outcomes and whether alternative thresholds may be more appropriate for older adults.

Consistent with prior research, we found that hypoglycemia occurred predominately overnight. (22) This may partly reflect skipping or delaying meals in the evening, evening exercise, or impaired counterregulatory responses which occur during sleep. Meal skipping is highly prevalent in older age; one study of ~1,200 older adults (ages 65-75 years old) found that more than one third regularly skipped meals (self-reported regularly skipping one meal a day).(23) Overnight low glucose may also be due to direct pressure on the CGM sensor (e.g., laying on the sensor) can cause artificial “compression lows”.(24)

Among persons on insulin or sulfonylureas, we found that CGM-detected hypoglycemia was more common in participants with cardiovascular disease, cognitive impairment, poor physical functioning, or chronic kidney disease. These findings support the American Diabetes Association (ADA) Standards of Care which note that cardiovascular disease, cognitive impairment, and end-stage kidney disease are major risk factors of hypoglycemia in individuals treated with insulin or sulfonylureas.(2, 3) This highlights the importance of closely monitoring older adults with these conditions for hypoglycemia. Furthermore, among persons on insulin or sulfonylureas, participants with a history of severe hypoglycemia (n=3) spent a median of 0% of time with glucose below 70 mg/dL (vs. 3.7% for participants without a history of severe hypoglycemia, n=99). This difference may reflect survival bias.

We found that participants not using insulin or sulfonylureas may experience some biochemical hypoglycemia, though the burden was low (median time spent in Level 1 hypoglycemia: 0.7%). On average, these participants spent 135 minutes under 70 mg/dL over the 2-week wear period and experienced a median of 2 episodes of hypoglycemia. Most of these participants (60%) were on other glucose-lowering medications suggesting that glucose-lowering medications other than insulin and sulfonylureas or possibly changes in diet may pose a low but non-zero risk of biochemical hypoglycemia. However, it is unclear how clinicians should use CGM data to guide management in these patients. Furthermore, while CGM glucose under 70 mg/dL is an established threshold for hypoglycemia, its clinical significance in individuals not using medications known to cause hypoglycemia remains uncertain. Additional research is needed to determine whether this threshold represents pathologic hypoglycemia in this population. Current guidelines focus on using CGM to inform treatment decisions in patients using high-risk medications. Recommendations for interpreting CGM data for patients with type 2 diabetes not treated with insulin/sulfonylureas is a timely issue, as the ADA recently broadened its recommendation for CGM use, stating that “anyone who could benefit from its use in diabetes management” should be offered these devices.(25) This new recommendation may result in increased CGM uptake for new populations, highlighting the need for guidance in noninsulin treated patients.

ARIC is one of the largest community-based studies to collect ~14 days of data from a blinded CGM sensor in the very old (age range 76-97 years), using standardized CGM data collection. We had a high rate of return of CGM sensors (only 1 participant with diabetes did not return their sensor). The study population included black and white adults and participants with type 2 diabetes who did and did not use high-risk medications. Hypoglycemia risk factors were systematically assessed in all participants.

Our study had several limitations. CGM systems are less accurate at low glucose levels(12) and the Abbott FreeStyle Libre Pro sensor is known to overestimate the degree of hypoglycemia compared with venous glucose.(12) Additionally, the Abbott Libre Pro is an older generation CGM system, however, the underlying biochemical technology of the more modern devices is in the same as the one used in this older generation device.

In summary, we found that most very old adults with type 2 diabetes in our community-based study who use insulin or sulfonylureas exceed the recommended target of <1% time with CGM-glucose <70 mg/dL. Further research is needed to clarify the clinical significance of these hypoglycemic episodes.

Key Points.

• Among individuals using insulin or sulfonylureas, the median time spent in hypoglycemia was 3.4% and ~66% of participants exceeded the target of <1% time with CGM glucose <70 mg/dL.

• Cardiovascular disease, cognitive impairment, poor physical functioning, and chronic kidney disease were associated with increased time spent in hypoglycemia.

Why does this paper matter?

We characterize the burden and risk factors for CGM-detected hypoglycemia in older adults with type 2 diabetes, demonstrating a substantial burden of unrecognized CGM-detected hypoglycemia among those using high-risk medications. We also show that while CGM-detected hypoglycemia occurs in persons not using high-risk medications, the burden is much lower and the clinical significance of CGM glucose <70 mg/dL in this population remains uncertain.