This study was performed in accordance with the principles of the Declaration of Helsinki. Ethical approval for this study was obtained from the University of Virginia Institutional Review Board for Health Sciences Research (protocol HSR200250).

To maximize external validity, one-third of the participants were recruited from each of three diverse centers: University of Virginia (n=5), University of Colorado (n=6), and West Virginia University (n=6). Inclusion criteria were as follows: age of 35-85 years, HbA_1c levels of 6.5%-11.5%, diagnosed with T2D within the past 12 months, not taking diabetes medication, no medical condition or medication that precludes reducing carbohydrates or walking 120 steps per minute for 10 minutes (eg, prednisone, severe neuropathy, cardiovascular disease, chronic obstructive pulmonary disease or emphysema, osteoarthritis, stroke, severe gastroparesis, ulcers, or food allergies), and ablilty to read English. Our age criteria aimed to select individuals most likely to be in control of their daily routine, HbA_1c criteria aimed to ensure the diagnoses of T2D but to avoid individuals whose condition was so progressive that immediate medication management was indicated, the no diabetes medication criterion was essential to test the hypothesis that GEM would prevent or diminish the need for diabetes medication, and the remaining criteria aimed to ensure the feasibility of engaging in the comprehensive GEM lifestyle.

After signing a University of Virginia IRB-approved consent, each participant’s primary care physician or clinician was contacted to affirm that the participant met eligibility criteria and to provide written approval for participation. Next, participants were sent a weblink to complete a series of questionnaires (Baseline: routine consumption of high and low glycemic load foods [19]; psychological questionnaires to assess diabetes empowerment [20], diabetes distress (emotional and regimen) [21], and depressive symptoms [22]; and diabetes knowledge as it relates to GEM principles [23].

After completing questionnaires, participants were mailed a CGM reader and 4.5 months of sensor supplies, (FreeStyle Libre 2 CGM system), a Fitbit Charge 3 activity monitor, and the GEM pocket guide (hard copy). This was followed by a telephone call introducing them to the CGM and activity monitor technology, inserting the CGM sensor, registering and activating the technology, and selecting seven personalized text messages, for example, “Food choices are life choices, Exercise is my friend,” which would be delivered at a time and frequency selected by the participant, to encourage GEM engagement [18]. Three days later, they received the second and final call to review the GEM pocket guide to initiate treatment.

GEM is neither a behavior modification nor a prescription program. Rather, GEM is an empowerment program [24,25] that provides information that a patient can choose to employ, to identify food and activity choices that either exacerbate or diminish postnutrient glucose excursions (represented by the area under the curve in Figure 1). The GEM pocket guide is a 4.25×5.50-inch booklet with four units. In unit 1, with CGM feedback, participants spend 5 days learning which of their routine food and physical activity choices have major and minor impacts on their glucose excursions. In unit 2 participants spend 14 days focusing on reducing, substituting, replacing, or eliminating high impact carbohydrates to diminish glucose excursions; for example, replacing breakfast oatmeal with unsweetened Greek yogurt and fresh fruit or substituting cauliflower rice for white rice. In unit 3, participants spend 14 days learning how to hasten recovery of glucose excursions by changing the type, intensity, duration, and timing of routine physical activity, such as walking the dog after supper instead of sitting in front of the television or computer. In unit 4, participants learn to continue experimenting with new nutrients and activities, sustain support of significant others, and prevent, anticipate, and recover from relapses resulting from fatigue, life stress, or change in routine. GEM was executed in the context of self-monitoring with diary entries, personal feedback from the CGM and the activity monitor, and automated daily motivational text messages. Following unit 1, there were 5 diary pages, and following units 2-4, there were 14 days of diaries where participants recorded their food and activity choices and how these impacted their glucose excursions.

Figure 1 displays 2 days of pre- and post-CGM data for 1 participant, illustrating how GEM reduces glucose excursions.

After this 6-week GEM program, for the next 3 months, participants continued exploring the principles of GEM with the assistance of their CGM and Fitbit feedback. A 3-month posttreatment assessment involved downloading their CGM data, and repeating the web-based questionnaires. Following completion of these questionnaires, participants rated how “helpful” different elements of the program were on a Likert scale (0=not helpful at all, 1=slightly helpful, 2=somewhat helpful, 3=very helpful, and 4=extremely helpful).

Investigators then extracted participants’ diagnostic and post-GEM HbA_1c levels, BMI, and prescribed diabetes medications from their medical records (Figure 2).

SPSS (version 27, IBM Corp) was used to perform 2-tailed paired t tests to test for differences in means between GEM participants’ diagnostic (baseline, preintervention) and 3-month postintervention data among the following outcome variables: HbA_1c levels, metformin dose, number of CGM readings that were >140 mg/dL, CGM glucose variability expressed as SD values [26,27], BMI, and weight. Pre- and postintervention mean scores were also compared for the following questionnaires or scales: high-carbohydrate food intake, low-carbohydrate food intake, diabetes knowledge, diabetes empowerment, diabetes distress (emotional subscale), diabetes distress (regimen subscale), and depressive symptoms. The Benjamini-Hochberg procedure [28] was used to control for multiple comparisons, correcting for all P values in Table 1. Exploratory Pearson correlations were performed between HbA_1c levels and both baseline and postintervention variables. There were no missing data, and none of the data sets violated the assumptions of a normal distribution.

Our primary hypotheses were confirmed. Mean HbA_1c levels were reduced by 1.8% among all participants, with 67% being classified as having diabetes remission and 80% being classified as responders with a mean HbA_1c level reduction of –2.3% among responders. This was achieved with only one participant needing to start taking medication. This participant’s HbA_1c level decreased 3.6%, from 12.7% to 9.1%, which was a clinically important improvement that subsequently required the additional introduction of metformin.

Regarding secondary hypotheses, there was a posttreatment decrease in diabetes distress, depression symptoms, and BMI, and a trend toward increased diabetes empowerment.

The strengths of this study are that it was a multicenter, brief, self-administered intervention, which recruited a diverse sample by diverse investigators. The mean change in HbA_1c levels of –1.8% by these GEM participants incorporating CGM compares favorably to the –1.0% change in HbA_1c levels by a similar group in a randomized controlled trial delivered in a face-to-face format with adults having T2D for an average of 5.3 years of taking diabetes medication [12].

Limitations of our study include a small sample size, no control group, and a limited follow-up duration. Additionally, pre-post change in physical activity, a presumed primary mechanism of GEM, was not monitored in this study. Despite these limitations and multiple positive findings, this is still a pilot study in need of replication with a larger and more diverse sample, which could tease out whether CGM alone would lead to such benefits.

This compares favorably to a mean reduction in HbA_1c levels of 1.5% and no psychological benefits with maximum dosage of metformin [30].

Despite being a self-directed program, these results were better than any of our previous efforts [12,15]. This could have resulted from our subject sample consisting of recently diagnosed adults who had not yet begun diabetes medication. This speculation is supported by the UK Prospective Diabetes Study [31], which initiated treatment for newly diagnosed T2D with 3 months of “dietary counseling” alone, with no medication, which led to a reduction in HbA_1c levels by ~1%. It could also be because of all participants in this study wanting this intervention—no random assignment. However, it may also be due to the multidimensional nature of the intervention: feedback from CGM and activity monitors, a structured and brief pocket guide and diary, and daily text messages, all of which were considered helpful by participants. It could have been that a reduction in HbA_1c levels was highly associated with baseline HbA_1c levels, since those with a higher HbA_1c level had the possibility of lowering it further, or that greater diabetes distress was associated with greater reduction, as this could reflect greater motivation. Likewise, the greater knowledge acquired about the impact of diet and activity on diabetes, and the greater reduction in routine carbohydrate ingestion were associated with more reduction in HbA_1c levels, as these were the hypothesized mechanisms. Reduction in BMI was a secondary benefit and was not correlated with improvement in HbA_1c levels.

GEM augmented with CGM feedback may be an effective initial intervention for adults newly diagnosed with T2D. A self-administered version of GEM may provide primary care physicians and their patients with a new tool to help people recently diagnosed with T2D to achieve remission independent of medication and without weight loss as the primary focus. Future research is needed with a larger and more diverse sample.