Continuous Glucose Monitoring-Guided Insulin Administration in Long-Term Care Facilities: A Randomized Clinical Trial

Abstract

Objectives.

To evaluate the efficacy of real-time continuous glucose monitoring (rt-CGM) in adjusting insulin therapy in long-term care facilities (LTCF).

Design.

Prospective randomized clinical trial.

Settings and Participants.

Insulin-treated patients with type 2 diabetes (T2D) admitted to long term care facilities (LTCF).

Methods.

Participants in the standard of care wore a blinded CGM with treatment adjusted based on point of care capillary glucose results before meals and bedtime (POC group). Participants in the intervention (CGM group) wore a Dexcom G6 CGM with treatment adjusted based on daily CGM profile. Treatment adjustment was performed by the LTCF medical team, with a duration of intervention up to 60 days. The primary endpoint was difference in time in range (TIR 70–180 mg/dL) between treatment groups.

Results.

Among 100 participants (age 74.73 ±11 years, 80% admitted for subacute rehabilitation and 20% for nursing home care), there were no significant differences in baseline clinical characteristics between groups and CGM data was compared for a median of 17 days. There were no differences in TIR (53.38%±30.16 vs. 48.81%±28.03, p=0.40), mean daily mean CGM glucose (184.10±43.4 mg/dL vs. 190.0±45.82 mg/dL, p=0.71), or the percent of time below range (TBR) < 70 mg/dL (0.83%±2.59 vs. 1.18%±3.54; p=0.51 or TBR < 54 mg/dL (0.23%±0.85 vs. 0.56%±2.24; p=0.88 between rt-CGM and POC groups.

Conclusions and Implications.

The use of rtCGM is safe and effective in guiding insulin therapy in patients with T2D in LTCF resulting in a similar improvement in glycemic control compared to POC-guided insulin adjustment.

Summary:

Older adults in long term care facilities are a higher risk of hypoglycemia. Compared to point of care glucose testing, the use of CGM to guide insulin therapy was safe and detected more hypo- and hyperglycemia episodes.

Background:

The prevalence of diabetes increases with advancing age, and national estimates indicate that over 20% of individuals between the ages of 65 and 75, and more than 40% of individuals over 80 years old, are affected by diabetes ^1,2. It is anticipated that the prevalence of diabetes among older adults will continue to rise because of extended life expectancy and advancements in healthcare ^2,3. As the geriatric population continues to grow, there is an expected increase in the number of admissions to long-term care facilities (LTCF) ³, with a reported prevalence of diabetes in subacute and long-term skilled nursing care facilities between 20% to 34% ^4,5.

Management of older adults in LTCF is challenging due to the presence of multiple comorbidities such as chronic kidney disease ⁶, cognitive dysfunction ⁷, hypoglycemia unawareness ⁸, functional disability, cerebrovascular events, depression, and altered nutritional intake ^9–11. Although no prospective studies are available to determine optimal targets of glycemic control in older adults admitted to LTCF, professional organizations have recommended patient-centric individualized glycemic approaches, with an emphasis on potential harms of intensive glycemic control and hypoglycemia risk in older adults ^12,13. Subcutaneous insulin therapy use is common for the management of hyperglycemia in LTCFs ^14,15. Although effective in improving glycemic control, up to one-third of insulin treated residents in LTCFs experience hypoglycemia ^13,16,17 which has been associated with longer stay, emergency department visits and hospitalization ¹⁸.

Bedside capillary point-of-care (POC) glucose monitoring is the current standard of care to assess glycemic control in the in LTCFs ^5,19. POC testing is recommended by national guidelines before meals and at bedtime. This approach, however, provides limited evaluation of glycemic excursions and misses nocturnal hypoglycemia or prolonged hypoglycemia ^20–23 The frequency of hypoglycemia in home-dwelling older adults is reported to affect one-third to half participants using CGM in a Norway and the US. ^24–26

Several randomized trials have shown that the use of continuous glucose monitoring (CGM) provides a more accurate assessment of glycemic profile compared to POC testing in ambulatory and hospitalized patients ^{27–29 21,23,30–33}. In addition, compared to capillary POC glucose testing, the inpatient use of real-time CGM with remote glucose monitoring results in a similar improvement in glycemic control and in lower frequency of hypoglycemia in insulin-treated patients ^34,35, and to be safe in guiding insulin administration in general medicine and surgery patients hospitalized with T2D ^34,35.

Herein, we report the results of the first randomized control trial aiming to determine whether the use of Dexcom CGM with remote monitoring will facilitate diabetes treatment in insulin treated residents with T2D, compared to standard of care using capillary POC testing, in subacute and long-term care facilities.

Materials and methods

We performed a randomized controlled trial to assess the differences between POC testing (standard of care) and rt-CGM in guiding insulin therapy among LTCF residents. The study was conducted at the University affiliated subacute rehab and skilled nursing facilities. The study protocol was approved by the University Institutional Review Boards. Informed consent was obtained from all participants prior to enrollment. Consents from participants with mild dementia were collected from the power of attorney on file. This trial is registered with ClinicalTrials.gov.

Patients were screened by electronic medical records. We screened adult males and females with a known history of T2D and treated with insulin (glargine, detemir, degludec, NPH, premixed insulin) or sliding scale regular insulin, with or without additional oral antidiabetic agents (insulin secretagogues [sulfonylureas, repaglinide, nateglinide], thiazolidinedione, SGLT2- inhibitors, DPP4-inhibitors), short- and long-acting GLP1-RA (exenatide, liraglutide, dulaglutide, semaglutide). Participants were excluded if they had an expected length-of-stay for less than 1 week; need for MRI procedures; with clinically relevant hepatic disease with diagnosed liver cirrhosis and portal hypertension, corticosteroid therapy, end-stage renal disease and dialysis, or anasarca.

Participants were randomly assigned (1:1 ratio) to one of two groups following a computer-generated randomization table to either a standard of care POC group or to a real-time CGM group and were followed for a minimum of 10 days and up to 60 days of enrollment.

Participants in the standard of care group wore a blinded CGM and received POC testing before meals and bedtime, with providers adjusting oral or insulin therapy based on capillary POC results. Patients in the intervention CGM group had a single daily fasting POC testing and wore a real-time Dexcom G6 CGM, and providers adjusted insulin therapy based on daily CGM profile information. During admission, sensors were changed every 10 days by the research coordinator.

The CGM remote monitoring system included three main components i) Dexcom G6 (Dexcom^™, San Diego, CA) and CGM device (sensor, transmitter); ii) a smart phone, and iii) a tablet computer at the nursing station. As a first step, glucose values obtained from the CGM sensor were sent through the CGM transmitter to a smart phone that serves as an intermediate-transmitting (routing) device using Bluetooth^™ technology. The study phone was locked in a safe box in the subject’s room (Samsung on Verizon 4G LTE). For glucose values to be transmitted to the nursing station, we installed Dexcom G6 Follow digital software application https://www.dexcom.com/apps) on computer tablets (Apple iPad). The research team offered educational meetings and materials on CGM technology to nurses, providers, and staff.

Participants in the rt-CGM group had alarms for hypoglycemia set at 85 mg/dL. The nursing staff were advised to confirm the low blood glucose value with POC testing. Hyperglycemia alarms were set at 300 mg/dL and the nursing staff were instructed to perform a POC glucose testing to confirm the high glucose value. If confirmed, nursing staff informed the primary care team. Alarms were turned off in the control group who wore blinded CGM devices. The nursing staff were instructed to follow the facility protocol to manage hypoglycemia <80 mg/dL by providing 15 grams of carbohydrates to prevent clinically significant hypoglycemia.

Study Endpoints

The primary aim of this study was to determine difference between capillary POC testing and CGM-GTS in achieving glycemic targets, defined as percentage of time in range (TIR) between 70–180 mg/dL (primary endpoint). The international consensus³⁶ and the American Diabetes Association ³⁷ have recommended a TIR (70–180 mg/dL) of at least 70 percent of readings for most people with type 1 and type 2 diabetes. However, for older adults, a more relaxed target of around 50% is recommended, meaning that 50 percent of readings each day should be in range 70–180 mg/dL.³⁶ Secondary outcomes included the differences in hyperglycemia, defined as time above range (TAR) >180 mg/dL and >250 mg/dL, number of hypoglycemia events and time below range (TBR) < 70 and 54 mg/dL, and nocturnal hypoglycemia between 22:00 and 06:00 between groups. Glycemic variability was also measured by the mean amplitude of glycemic excursion (MAGE), coefficient of variation (CV), and standard deviation (SD).

Statistical Analysis

The objective of this clinical trial was to investigate whether the CGM with remote monitoring could improve glycemic control as measured by the TIR 70–180 mg/dL in patients with T2D in subacute and long-term skilled nursing facilities. The baseline and clinical characteristics are presented as mean (standard deviation) or median (interquartile range) for continuous variables and count (percentage) for discrete variables. We compared baseline clinical characteristics and clinical or CGM outcomes between the two study groups using non-parametric Wilcoxon tests for continuous variables and Chi-square tests (or Fisher’s exact tests) for discrete variables. A p value of less than 0.05 was considered significant. We performed all statistical analyses using SAS 9.4.

Results:

A total of one hundred participants were included in the study. Three (n= 3) participants were excluded from the CGM group because they had less than 7 days of data. Consequently, ninety-seven participants were included in the analysis: fifty participants randomly assigned to POC testing (control group), and forty-seven participants to CGM (intervention group).

Baseline characteristics are summarized in Table 1. The average age of all participants was 74.73 ± 11 years with 80% of participants enrolled in subacute rehabilitation units and 20% from long-term care. There were no significant differences in baseline clinical characteristics including age, sex, race, weight, body mass index (BMI), hemoglobin A1c (HbA1c), admission diagnosis, and total daily insulin dose (Table 1). Duration of CGM use was similar between groups, with a median of 17.0 (12,36) days in the real time-CGM group and 18.5 (10,28) days in the POC group (Table 2).

Table 1.

Baseline Clinical Characteristics by study group

	CGM Group (n= 47)	POC-group (n= 50)	P-value
Age	74.98 ± 11.7	74.50 ± 10.64	0.79
Sex, n (%)			0.28
Female	29 (62%)	36 (72%)
Male	18(38%)	14 (28%)
Age group, n (%)			0.86
< 65 years	11(23%)	10(20%)
65–69 years	5(11%)	8(16%)
70–74 years	7(15%)	6(12%)
≥ 75 years	24(51%)	26(52%)
Race, n (%)			0.17
Black	27 (57%)	36 (72%)
White	19 (40%)	14 (28%)
Hispanic	1(2.1%)	0(0.0%)
BMI	29.90 ± 8.57	29.39 ± 7.35	0.99
HbA1c (%)	7.97 ± 1.90	8.14 ± 2.49	0.57
Admission service, n (%)			0.31
Subacute Rehab	40 (85%)	37 (76%)
Long Term Care	7 (15%)	12 (24%)
Home DM treatment, n (%)			0.59
OAD	5(11%)	8(16%)
Insulin	20(43%)	24(48%)
OAD + Insulin	18(38%)	17(34%)
GLP1	3(6.4%)	1(2.0%)
GLP1+OAD	1(2.1%)	0(0.0%)
Treatment in LTCFs
Metformin, n (%)	20(43%)	12(24%)	0.05
Insulin secretagogues, n (%)	5(11%)	10(20%)	0.27
DPP4-i, n (%)	4(8.5%)	5(10%)	1.00
GLP1-ra, n (%)	4(8.5%)	4(8.0%)	1.00
Basal insulin, n (%)	36(92%)	32(76%)	0.07
Prandial insulin, n (%)	33(83%)	34(81%)	1.00
Insulin dose, Units/day	28.72 ± 20.56	30.74 ± 24.69	0.66
Insulin dose, Units/kg/day	0.33 ± 0.19	0.38 ± 0.28	0.53
Primary admission diagnosis
Cardiovascular, n (%)	7 (15)	6 (12)	0.77
Infectious, n (%)	3 (6.4%)	7 (14%)	0.32
Neurologic, n (%)	11(23%)	12(24%)	0.95
Pulmonary, n (%)	3(6.4%)	2(4.0%)	0.67
Dementia, n (%)	6(13%)	6(12%)	1.00
Primary diagnosis: Other, n (%)	25(46%)	23(46%)	0.98

BMI: body mass index; HbA1c: hemoglobin A1C; DM: diabetes mellitus; BG: blood Glucose; GFR: glomerular filtration rate; OAD: oral anti-diabetes

Table 2.

CGM Data by study group

	CGM Group (n=47)	POC Group (n=50)	P-value
Mean daily glucose by CGM, mg/dL	185.15 ± 44.08	191.02 ± 47.25	0.72
Mean daily glucose by POC, mg/dL	175.76 ± 41.79	167.26 ± 47.84	0.19
Mean nocturnal glucose by CGM, mg/dL	176.34 ± 44.56	181.91 ± 47.03	0.68
TIR 70–180 mg/dL, %	53.38 ± 30.16	48.81 ± 28.03	0.40
TAR >180 mg/dL, %	45.79 ± 30.74	49.13 ± 28.11	0.55
TAR >250 mg/dL, %	16.45 ± 19.27	17.80 ± 22.84	0.97
TBR < 70 mg/dL, %	0.83 ± 2.59	1.18 ± 3.54	0.51
TBR < 54 mg/dL, %	0.23 ± 0.85	0.56 ± 2.24	0.88
GV, MAGE	64.03 ± 27.47	62.17 ± 24.02	0.91
GV, SD	45.33 ± 15.35	47.59 ± 15.52	0.44
GV, CV	0.25 ± 0.06	0.25 ± 0.06	0.62
Nocturnal glucose < 70 mg/dL by CGM, %	0.28 ± 1.17	0.73 ± 2.43	0.75
Nocturnal glucose < 54 mg/dL by CGM, %	0.10 ± 0.42	0.33 ± 1.58	0.50
Glucose > 180 mg/dL among participants with hyperglycemia >180 mg/dL by CGM, %	45.79 ± 30.74	50.13 ± 27.48	0.43
Glucose > 250 mg/dL among participants with hyperglycemia >250 mg/dL by CGM, %	20.89 ± 19.47	23.42 ± 23.58	0.68
Glucose < 70 mg/dL among participants with hyperglycemia <70 mg/dL by CGM, %	1.77 ± 3.60	3.45 ± 5.47	0.64
Glucose < 54 mg/dL among participants with hyperglycemia <54 mg/dL by CGM, %	0.99 ± 1.59	3.06 ± 4.71	0.49
Number of days on CGM, Median (Q1, Q3)	17.0 (12, 36)	18.5 (10, 28)	0.74

GV: glycemic variability; MAGE: mean amplitude of glucose excursion; STD: standard deviation; CV: coefficient of variation

There were no significant differences regarding TIR 70–180 mg/dL between patients treated with rt-CGM compared to POC testing (53.3 ± 30.1 mg/dL vs. 48.8 ± 28.0 mg/dL; p= 0.40). The percent of TBR (<70 mg/dL and <54 mg/dL) was not different as well between the rt-CGM and the POC group (0.83% ± 2.59 vs 1.18% ± 3.54; p= 0.51 for BG < 70 mg/dL; 0.23% ± 0.85 vs 0.56% ± 2.24; p=0.88 for BG< 54 mg/dL).

The average daily glucose was 185.1 ± 44 mg/dL in the rt-CGM group and 191 ± 47.2 mg/dL in the control group using blinded CGM (p= 0.72). The average daily glucose assessed by POC testing was 175.7 ± 41.8 mg/dL in the rt-CGM group and 167.2 ± 47.9 mg/dL in the POC group (p= 0.19). We assessed the ability of CGM and capillary POC testing in detecting the frequency of hypoglycemia and hyperglycemia events (Table 3). CGM detected greater proportions of participants with hypoglycemia < 70 mg/dL (40.2% vs. 14%) and < 54 mg/dL (21% vs. 1.0%); as well as hyperglycemia >250 mg/dL (77% vs. 56%) compared to POC testing, all p<0.001.

Table 3.

Glycemic Control differences by continuous glucose monitoring and point-of-care testing.

	CGM Group	POC group	P value
Glycemic Control
Mean daily glucose, mg/dL	188± 45	171± 45	<0.001
Glucose > 180 mg/dL, n (%)	96 (99%)	77 (80%)	<0.001
Glucose > 250 mg/dL, n (%)	75 (77%)	54 (56%)	<0.001
Glucose < 70 mg/dL, n (%)	39 (40%)	13 (14%)	<0.001
Glucose < 54 mg/dL, n (%)	20 (21%)	1 (1.0%)	<0.001

The percent in time above range (TAR, >180 mg/dL) was not different between the two groups (45.79% ± 30.74 vs. 49.13% ± 28.11; p= 0.55). Similarly, the percentage in time spent below range (TBR) <70 mg/dL or 54 mg/dL did not differ between groups nocturnally (0.28% ± 1.17 vs. 0.73% ± 2.43; p=0.75, and 0.10% ± 0.42 vs. 0.33% ± 1.58; p= 0.50, respectively). In addition, there were nonstatistical reduction in overall TBR < 70 mg/dL (1.77% ± 3.60 vs. 3.45% ± 5.47; p= 0.64) or nocturnal hypoglycemia (1.17% ± 2.27 vs 3.31% ± 4.41; p= 0.17), as well as TBR < 54 mg/dL (0.99% ± 1.59 vs 3.06% ± 4.71; P= 0.49) and (0.88% ± 1.10 vs 2.36% ± 3.85, p= 0.87), respectively, (Table 2).

Analysis of glycemic variability between the two groups reported no differences in the coefficient of variation (CV) (0.25 ± 0.06 vs. 0.25 ± 0.06; p= 0.62), standard deviation (SD) (45.33 ± 15.35 vs. 47.59 ± 15.52; p= 0.44), or mean amplitude of glycemic excursion (MAGE) (64.03 ± 27.47 vs. 62.17 ± 24.02; p= 0.91) (Table 2).

Discussion

We present the results of the first randomized clinical trial investigating the safety and efficacy of rt-CGM to guide insulin therapy in older adults with T2D residing in LTCFs. The use of rt-CGM was found to be safe and effective in guiding insulin therapy and resulted in comparable improvement in glycemic control and a trend in reducing time spent in hypoglycemia compared to capillary POC testing. In addition, our results indicate that the implementation of rt-CGM technology detected higher rates of hypoglycemic and hyperglycemic events compared to standard care capillary POC testing.

The American Diabetes Association and the American Medical Directors Association recognize substantial difficulties associated with managing diabetes in older adults admitted to long-term skilled nursing facilities ^{5 38}. Older adults often have significant comorbidities that impact diabetes control and contribute to an increased risk of treatment-related complications. In a retrospective study of 11,531 long-term care residents who initiated a glucose-lowering medication, the mean age was 81.0 years ¹⁴ and had multiple comorbidities including hypertension, congestive heart failure, and ischemic heart disease. In addition, diabetes in older adults is accompanied by increased incidence of geriatric syndromes which can greatly impact diabetes care, quality of life, and hypoglycemia risk ³⁹.

Several observational and prospective RCT have reported on the efficacy and safety of diabetes treatment in older adults with T2D admitted to LTCFs ^16,40,41. These studies reported similar improvements in mean daily glucose and hospital complications between insulin and non-insulin regimens; however, insulin administration was associated with higher rate of hypoglycemia compared to non-insulin agents ¹⁶. A retrospective study of over 1400 residents found that 42% of patients in LTCFs had at least one episode of hypoglycemia < 70 mg/dL ^13,40,41. Hypoglycemia was most common in patients treated with sulfonylurea and/or insulin, with sulfonylureas accounting for 18.8% of hypoglycemic episodes and insulin responsible for 64% of events ^13,41. In addition, hypoglycemia was associated with significant morbidity including increased risk of falls and fractures ^42,43 and higher rates of emergency room visits or hospitalization than those without hypoglycemia ^13,41. Thus, it is important to develop strategies to minimize the risk of hypoglycemia including simplification of medication regimens, avoidance of insulin secretagogues, minimization of sole use of sliding scale insulin and avoidance of tight glycemic control in LTCFs ¹³.

An important aspect of diabetes care in LTCFs is the accuracy and safety of glucose monitoring in guiding treatment. The standard of care is to use POC capillary glucose testing before each meal and bedtime ⁵; however, POC testing does not provide a full picture of glucose control ^5,13,19, and on-site providers are left with few data points to make clinical judgments on insulin adjustment. Our group and others have reported on the efficacy and safety of using CGM with remote glucose monitoring in hospitalized medicine and surgery patients with T2D ^34,35. We reported that the use rt-CGM in insulin-treated hospitalized patients resulted in similar improvement in glycemic control but in a significant reduction in hypoglycemia rates and compared to POC glucose testing ³⁴. In addition, reduction in both hypoglycemia and hyperglycemia was reported in 2 studies that employed rt-CGM for diabetes management in community hospitals ^35,44. In addition, a CGM study detected hypoglycemia events in 79% of six geriatric care centers residents in France. ⁴⁵ Our results are in agreement with previous reports on the safety and efficacy of CGM with remote monitoring for the management of patients with T2D and support recent diabetes guidelines and consensus on hospital diabetes care, which reported that the use of rt-CGM may allow safer management of hyperglycemia in adults who are at high risk of hypoglycemia ⁴⁶, as well as in insulin dose adjustments ⁴⁷.

Our study has limitations including the small sample size of one hundred insulin-treated patients with T2D from a single academic institution, excluding patients with type 1 diabetes (T1D) and other forms of diabetes. We limited the observation to the use of real-time Dexcom G6 CGM; thus, the results may not be extrapolated to other intermittently scanned or rt-CGM devices. The diabetes management was carried out by internal medicine providers working in close collaboration with experienced endocrinology and research teams. The research protocol of adjusting insulin dose was shared with providers before the start of the study and was similar to the institution insulin protocol. Medical providers and staff adopted the new technology into the daily working flow without significant discernable increase in work burden; however, the acceptance and implementation in real-world settings need to be assessed objectively in future studies.

Conclusions and Implications:

Our results show that the use rt-CGM to guide insulin therapy in LTCF is safe, effective, and resulted in a similar improvement in glycemic control without increase in hypoglycemia. We have also shown that the use of CGM technology is superior to POC standard of care testing in detecting hyperglycemia and hypoglycemia events in LTCFs. The utilization of glucose telemetry system in LTCFs is feasible, effective, and should be considered to facilitate care and reduce the burden of frequent capillary POC testing in older adult populations in skilled nursing care facilities.