Minimizing Glucose Excursions (GEM) With Continuous Glucose Monitoring in Type 2 Diabetes: A Randomized Clinical Trial

Abstract This study aimed to compare conventional medication management of type 2 diabetes (T2D) to medication management in conjunction with a lifestyle intervention using continuous glucose monitoring to minimize glucose excursions. Thirty adults (63% female; mean age, 53.3 years) who were diagnosed with T2D for less than 11 years (mean, 5.6 years), had glycated A1c (HbA1c) ≥ 7.0% (51 mmol/mol) (mean 8.8%, [73 mmol/mol]), and were not using insulin, were randomly assigned in a 1:2 ratio to routine care (RC) or 4 group sessions of glycemic excursion minimization plus real-time CGM (GEMCGM). Assessments at baseline and 5 months included a physical exam, metabolic and lipid panels, a review of diabetes medications, and psychological questionnaires. For the week following assessments, participants wore a blinded activity monitor and completed 3 days of 24-hour dietary recall. A subgroup also wore a blinded CGM. GEMCGM participants significantly improved HbA1c (from 8.9% to 7.6% [74-60 mmol/mol] compared with 8.8% to 8.7% [73-72 mmol/mol] for RC (P = .03). Additionally, GEMCGM reduced the need for diabetes medication (P = .01), reduced carbohydrate consumption (P = .009), and improved diabetes knowledge (P = .001), quality of life (P = .01) and diabetes distress (P = .02), and trended to more empowerment (P = .05) without increasing dietary fat, lipids, or hypoglycemia. Confirming our prior research, GEMCGM appears to be a safe, effective lifestyle intervention option for adults with suboptimally controlled T2D who do not take insulin.

Pharmacological and lifestyle interventions are both important components of managing type 2 diabetes (T2D), but compared to the wide variety of pharmaceutical interventions with several different mechanisms of action, there are few lifestyle intervention options that are offered and they are often perceived to be difficult and complicated. The American Diabetes Association recommends 150 minutes per week of moderate physical activity plus 7% weight loss through caloric restriction [1]. Weight loss is not a viable option for patients who do not need or want to lose weight, are unable to lose weight, or cannot maintain weight loss. Consequently, lifestyle interventions with a different approach are needed. More recently, the American Diabetes Association stated: "Reducing overall carbohydrate intake for individuals with diabetes has demonstrated the most evidence for improving glycemia and may be applied in a variety of eating patterns" [2].
Previously, we reported on a new lifestyle intervention for T2D that, instead of focusing on reducing weight, shifts the focus to reducing postprandial glucose [3], which is a primary contributor to glycated hemoglobin (HbA 1c ) [4]. This program is called glycemic excursion minimization (GEM), which reflects its goal of minimizing postnutrient (meals, snacks, drinks) blood glucose (BG) excursions, or area under the curve (AUC). This initial study had 5 sessions of face-to-face GEM instruction and employed systematic BG monitoring [5]; it measured BG both before and 2 hours after meals and before and 30 minutes following exercise. BG feedback was used to educate participants regarding how different amounts and kinds of food and exercise affected their BG and to motivate them to repeat those choices that led to desirable BG levels. Participants were also encouraged to use BG feedback to activate them to make choices consistent with their current BG level and personal BG targets. For example, if participants discovered their BG was higher than desired before a meal, they could choose to postpone eating or take a brisk walk to lower their BG. However, the use of systematic BG was problematic because some people did not like pricking their finger, seeing blood, always having BG monitoring supplies available before and after eating and exercising, or trying to remember to monitor BG at specific times.
Continuous glucose monitoring (CGM) is an attractive alternative to systematic BG monitoring. CGM reduces the frequency of finger pricks, the need to remember equipment and when to take BG measurements, and it provides a more comprehensive BG record than systematic BG monitoring. These advantages could enhance the benefits of GEM [6]. Therefore, we hypothesized that relative to routine care (RC), GEM with CGM feedback (GEM CGM ) would 1) significantly improve the primary outcome variables of HbA 1c and need for diabetes medication; 2) achieve these results through the mechanisms of increasing knowledge concerning food and activity choices, reducing consumption of carbohydrates, increasing routine physical activity, and diminishing BG excursions as reflected in CGM profiles; 3) have the secondary benefits of improving quality of life, empowerment, and diabetes distress; and 4) not have side effects like episodes of hypoglycemia [7] or increased compensatory fat consumption.

A. Participants
Thirty adults with T2D were recruited through radio and print advertisements and the University of Virginia Hospital patient registry between July 2018 and August 2019. GEM CGM training was conducted between December 2018 and September 2019, and 3-month follow-up assessments were performed between April 2019 and January 2020. Participants were between ages 30 and 80 years, had T2D for less than 11 years, had an HbA 1c greater than or equal to 7.0% (53 mmol/mol), were not on insulin or nondiabetic medications that could affect BG control (eg, prednisone), were able to walk for 30 minutes, and were interested in CGM. They were randomly assigned to RC or GEM CGM . Incentives to participate were free blood tests, blood glucose meters and supplies (Bayer Contour), activity monitors (Fitbit Charge 2), $100 on completion of postassessment, and CGM supplies if assigned to GEM CGM . The demographics of this sample appear in Table 1. The 2 groups did not significantly differ on any demographic variables.

B. Procedures
After being thoroughly informed, participants signed a University of Virginia institutional review board-approved Abbreviations: BMI, body mass index; HbA 1c , glycated hemoglobin; T2D, type 2 diabetes. consent form. Next, they participated in a baseline assessment that included a brief physical, blood tests for HbA 1c and lipids, a review of their current medications, and psychological questionnaires to assess attitudes toward glucose monitoring [8], quality of life [9], diabetes empowerment [10], diabetes distress [11], and depressive symptoms [12].
Using the Medication Effect Scale (MES) [13], the dose and type of each diabetes medicine in a participant's medication regimen was converted to a common denominator-that medication's average HbA 1c -lowering potential. These potentials were then summed over the diabetes medications the individual was taking to estimate the potential HbA 1clowering effect of a participant's medication regimen. The following week participants wore a blinded activity monitor (Fitbit Charge 2), and were interviewed over the telephone on 2 work days and 1 weekend day to complete the automated self-administered 24-hour dietary recall dietary recall [14]. Ten RC and 12 GEM CGM participants also wore a blinded CGM (Dexcom Platinum G4). This assessment was repeated a second time 5 months later-3 months after the conclusion of GEM CGM . Following the baseline assessment, participants were randomly assigned, by a flip of a coin in blocks of 3 by the data manager, who was unfamiliar with the individual, to either GEM CGM or RC in a 2:1 ratio; if "heads" came up during the first 2 participants, they were assigned to RC, and if "heads" did not come up then the third person was assigned to RC. All participants continued their usual care in consultation with their treating physician, who adjusted medication as clinically indicated throughout the 5-month study. One RC patient dropped out before postassessment to pursue bariatric surgery.
The 2-month GEM CGM intervention period involved meeting in groups of 8 to 10 for 90 minutes on 4 occasions, with 1 week between sessions 1 and 2 and 3 weeks between sessions 2 and 3 and 3 and 4 ( Fig. 1). At each session, participants were given a 7-day Dexcom G5 sensor, and 1 month after session 4, a fifth sensor was given. This timing was intended to diminish reliance on CGM and group support and to encourage autonomy following the conclusion of the intervention. Follow-up assessment occurred three months after session 4.
With the assistance of an instructor's manual, a diabetes nurse educator (A.D.) led group sessions. GEM CGM participants were given a 98-page manual and a diary that focused on the following:  • Session 4, Continuing and enhancing optimal food and activity choices over a lifetime and managing relapses.
At the beginning of each session, participants were asked to report the number of hypoglycemic experiences they had since the previous session, regarding level 1 (low enough for treatment), level 2 (sufficiently low to indicate serious, clinically important hypoglycemia), and level 3 (associated with severe cognitive impairment requiring external assistance for recovery) hypoglycemic events [7]. GEM CGM participants were given 5 G5 sensors, 1 to insert at each session and 1 to be inserted 8 weeks before the 3-month follow-up assessment.
Three-month follow-up assessments were conducted between February 2019 and December 2019. Pre and post change scores for RC and GEM CGM were compared in analyses of covariance, with baseline measures serving as the covariant. SPSS version 25 was used to perform the analyses.  Table 2). Table 2 displays the pre and post means, SDs, F tests, and probabilities of the primary outcomes, mechanisms, secondary, and side effect variables.

B. Mechanism Variables
As hypothesized, GEM CGM did produce a significant increase in knowledge and led to a reduction in carbohydrate ingestion relative to RC. GEM CGM did not significantly differ from RC in regards to increased physical activity or decreased glucose excursions (as quantified by AUC or time in range). However, change in the following variables did correlate with change in HbA 1c : Fitbit calories burned (r = -0.40, P = .01) and minutes of daily moderate to vigorous activity (r = -0.37, P = .04); CGM AUC (r = 0.61, P = .009), and CGM time in range (r = -0.62, P = .006). Additionally, responsiveness to GEM CGM was not limited by depressive symptoms because baseline Patient Health Questionnaire-9 (PHQ9) did not correlate with change in HbA 1c (r = -0.13, P = .50). This negative outcome was further confirmed when comparing participants with and without significantly elevated baseline PHQ9 scores (> 9) in regard to change in HbA 1c , MES, or TTE (P = .51, .31, and .82, respectively).

C. Secondary Outcome Variables
As hypothesized, GEM CGM significantly improved psychological function relative to RC, including World Health Organization (WHO)-Quality of Life (Psychological subscale), Diabetes Empowerment, Diabetes Distress Scale (Emotional and Regimen subscales), and the Glucose Monitor Satisfaction Survey. It did not significantly improve WHO-Quality of Life (Physiological subscale).

D. Side Effects
GEM CGM did not lead to an increase in caloric, fat, or protein consumption nor a worsening in lipids relative to RC. Self-report experiences with hypoglycemia were not collected among RC participants. GEM CGM participants reported 12 episodes of level 1 symptomatic hypoglycemia: 2 at session 2, 8 at session 3, and 2 at session 4. These were defined as "trembling, sweaty, pounding heart, unusual fatigue, pale skin, anxiety, irritability, with BG usually in the 60s, but it could be slightly higher or lower." During the blinded CGM period, using the criterion that hypoglycemia events are defined as BG of 70 mg/dL or less [7], GEM and RC did not significantly differ regarding the pre and post change in the number of hypoglycemic events recorded.

Discussion
This study replicates our previous report [3] that found, relative to RC, that GEM led to a significant reduction in HbA 1c in adults with T2D. The present study extends those findings by demonstrating that when CGM was used instead of systematic BG monitoring, GEM CGM reduced HbA 1c by a mean of 1.3% instead of the previous 1.0%, and did so with 20% fewer contacts (4 vs 5 training sessions) in a more diverse participant sample (longer diabetes duration, taking more varied diabetes medications). Further, the present study demonstrated that GEM CGM reduced HbA 1c and MES more than RC (1.1% and 0.83, respectively), for a TTE difference of 1.94% HbA 1c equivalent between the 2 groups (see Fig. 2).
Compared to RC, GEM CGM increased physical activity and decreased hyperglycemic excursions (AUC). However, these were not significantly different from RC. This was in part because the Dexcom G4 platinum captured only a maximum of 6.5 days of CGM data, the BG sampled in the HbA1c and blinded CGM measures did not overlap temporally (CGM followed the A 1c ), and the small sample size and large CGM SDs (see Table 3) restricted statistical power. However, changes in these parameters did significantly correlate with improvement in HbA 1c , as did physical activity in a previous study [15]. Similarly to the previous study, GEM CGM was associated with multiple psychological benefits compared to RC, including improvements in quality of life, greater sense of empowerment, and reduction in diabetes distress.
These benefits were achieved without worsening lipids or increasing hypoglycemia. It is interesting to note that no GEM CGM participants experienced level 2 hypoglycemia, whereas 2 RC participants, either having gone on insulin or glipizide, experienced 3 level 2 hypoglycemia events per patient during follow-up. This is reflected by an area within the curve under 70 mg/dL. RC showed a pre and post increase of 639 ± 1204, whereas GEM CGM demonstrated an increase of 7 ± 22. These changes were not significantly different because of RC's large variance.
Sample size is a significant limitation of the present study, potentially limiting the external validity of these findings and precluding investigation of individual differences. However, given replication of a previous study [3] with a different sample, any threat to external validity in the present study is diminished. The sample size was too small for the 2 groups to be matched on all variables. Although not significantly different, the RC group contained more female and fewer white participants. Furthermore, the sustainability of these findings is not certain given that the present study employed only a 3-month follow-up. Additionally, given that we recruited individuals interested in using CGM, these findings cannot be extrapolated to the broader T2D community, which includes people uninterested in CGM for a variety of reasons as noted previously. Finally, as in our previous study, which did not separate the effects of BG monitoring from the GEM didactic information, the present study does not separate any additive benefits of CGM from GEM didactics. This will be addressed in a future study.
These findings indicate that GEM CGM is a viable option for adults with T2D interested in CGM, and gives diabetes clinicians an additional tool to aid consumers with T2D. It will be up to future studies to determine whether GEM CGM is equivalent or superior to conventional weight loss therapy, which intervention is more sustainable, and which option best matches which patient characteristics.