We conducted a pragmatic, nonrandomized, pre-post evaluation of a QI pilot to examine the feasibility and effectiveness of low-carbohydrate nutrition counseling guided by continuous glucose monitoring (referred to hereafter as LC-CGM) among Veterans with T2D receiving insulin. The intervention was 24 weeks in duration, and all procedures, including patient recruitment, enrollment, education, and monitoring, were integrated into usual clinical care operations. The program evaluation was conducted from March 19, 2021, to May 3, 2024, with data analyzed from February 2024 through March 2025.

This program was implemented within the VA Ann Arbor Healthcare System, which includes the Lieutenant Colonel Charles S. Kettles VA Medical Center and 6 other freestanding VA outpatient health clinics serving a diverse patient population throughout southeast Michigan and northwest Ohio [26]. The program was delivered remotely using the VA Video Connect (VVC) platform or telephone, depending on patients’ preferences.

The QI team consisted of 1 clinical pharmacist practitioner (CPP), 2 registered dietitians (RD), 2 primary care physicians, and 1 endocrinologist. The CPP dedicated 0.1 full-time equivalent to 4 hours per week, and the RDs dedicated a combined 0.15 full-time equivalent to 6 hours per week to the program; the physicians did not have a dedicated program effort and devoted less than 1 hour per week, on average, to supporting the CPP and RDs. RDs and CPPs were trained to support Veterans’ adherence to a low-carbohydrate diet using materials previously developed by the study team [13,27], online resources [28], and case-based discussions during monthly team meetings.

For this pilot program, we aimed to recruit at least 20 Veterans meeting study inclusion criteria; we limited enrollment to 10 Veterans at one time due to staffing capacity. Inclusion criteria were: (1) Veterans with T2D, (2) use of ≥3 insulin injections per day (based on the VA criteria for use of CGM during the program evaluation period), (3) BMI≥25 kg/m², (4) desire to lose weight or improve metabolic health, (5) willingness and ability to use CGM technology, (6) willingness and ability to modify dietary intake to follow a carbohydrate-restricted eating pattern, (7) willingness to self-weigh at least once weekly and report weight information to the program team, and (8) willingness to engage in regular virtual visits with clinical team members, including a RD and clinical pharmacist. Exclusion criteria were (1) type 1 diabetes, (2) chronic kidney disease stage 4 or greater, (3) advanced heart failure, and (4) pregnant or breastfeeding.

Primary care clinicians or CPPs identified potentially eligible patients during routine clinical encounters, assessed their interest in LC-CGM program participation, and alerted QI team members via an electronic health record (EHR) message if patients desired to learn more about the opportunity. Within 2 weeks of receiving the EHR alert, a QI team member (DHG or KF) called patients by phone to discuss the program, address any questions or concerns, and initiate enrollment if desired; the enrollment phone script is shown in Multimedia Appendix 1.

Prior to initiating the LC-CGM program, patients completed baseline fasting laboratory testing, including a comprehensive metabolic panel, hemoglobin A_1c (HbA_1c), lipids, and c-peptide (to confirm endogenous insulin production). At 24 weeks, patients were asked to complete repeat laboratory testing for HbA_1c and fasting lipids.

Patients were advised to self-weigh at least once weekly [24]. Veterans without access to a home scale were provided with one, per usual clinical procedures. Patients reported weight change data to the RD during scheduled visits; these data were entered into patients’ medical records.

The following data were collected from the EHR: sociodemographic factors (age, sex, race, and ethnicity), anthropometric measures (height, weight, and BMI), laboratory test results (HbA_1c, total cholesterol, high-density lipoprotein [HDL], low-density lipoprotein [LDL], and triglycerides), and T2D medications prescriptions, including biguanides (eg, metformin), sulfonylureas, thiazolidinediones, meglitinides (eg, repaglinide), glucagon-like peptide 1 receptor agonists (GLP-1 RAs), dipeptidyl peptidase-4 (DPP-4) inhibitors, and SGLT2 inhibitors, and insulins. We also collected data on insulin type (eg, short-acting and long-acting) and total daily dosages (TDD) for each type.

Primary outcomes include measures of feasibility and acceptability:

Secondary outcomes include the following clinical effectiveness measures:

We evaluated weight and medication change outcomes for all program enrollees using data available in their final LC-CGM clinic note. Laboratory data were evaluated if available within 30 days of the final LC-CGM visit.

We assessed baseline characteristics of those who enrolled and compared baseline characteristics of those who enrolled and completed the program to those who enrolled but did not complete the program.

We computed descriptive statistics for collected variables, including means, medians, and IQRs for continuous variables and counts and proportions for categorical variables. To assess differences, we conducted intention-to-treat analyses and used the last observation carried forward method for missing data from noncompleters. We also assessed differences among those who completed the program. For continuous variables, we used the Shapiro-Wilk test to assess the normality of the distribution of differences. We compared baseline to 24-week changes using paired samples 2-tailed t tests or paired Wilcoxon exact signed rank tests with continuity correction. To compare differences in proportions, we constructed contingency tables and completed McNemar difference in paired proportions tests with continuity correction to improve the approximation of the discrete probability distribution with a continuous one [32]. To determine statistical significance, we used a 2-sided P<.05. All analyses were completed in R (version 4.1.3; R Core Team).

This program was reviewed by the VA Ann Arbor Healthcare System’s Quality Management Office, which reviews QI program applications to determine whether Institutional Review Board oversight is required. This project was approved as a nonresearch, QI initiative. As such, patients were informed that participation was voluntary and were also provided the opportunity to verbally accept or decline participation prior to data collection (Multimedia Appendix 1). To safeguard patient information, extracted data were deidentified and aggregated for analysis.

Overall, 43 Veterans were referred to the LC-CGM program during the evaluation period and 38 (88%) enrolled. Among program enrollees, 27/38 (71%) completed the 24-week program. Reasons for program nonenrollment and noncompletion are shown in Figure 1. The main reason for program noncompletion was Veterans’ preference not to follow a carbohydrate-restricted eating pattern (9/11, 81.8%).

Baseline sociodemographic and clinical characteristics for program enrollees, stratified by program completion status, are shown in Table 1. Most program enrollees were men (37/38, 97%), white (29/38, 76%), and the average age was 63.7 (SD 9.6) years. Mean BMI was 38.1 (SD 5.8) kg/m² and mean HbA_1c was 7.8% (SD 1.3%). There were no significant differences in baseline characteristics of program completers versus noncompleters. Overall, patients were prescribed an average of 3.5 (SD 0.8) glucose-lowering medications, and all were prescribed insulin.

The number of program enrollees and completers is shown in Figure 1. Among program enrollees, the mean number of weeks engaged in the LC-CGM was 20.7 (SD 5.6). As shown in Table 2, the mean number of program visits with the RD was 10.7 (SD 2.4) among program completers and 6.5 (SD 3.4) among program noncompleters (P<.001). The mean number of program visits with the CPP was 15.1 (SD 3) among program completers and 7.2 (SD 2.8) among program noncompleters (P<.001).

Table 3 shows weight change for the program at 4, 12, and 24 weeks among program enrollees, stratified by program noncompleters who discontinued the program at 4 and 12 weeks.

Overall, among 38 enrollees, mean weight change was −11.5 kg (SD 8.7) and mean percent weight change was 9.5% (SD 7.2%), with 68% (26/38) achieving ≥ 5% weight loss and 50% (19/38) achieving ≥ 10% weight loss at 24 weeks. Among 27 program completers, mean percent weight loss was 11.8% (SD 6.5%) at 24 weeks, with most achieving ≥5% weight loss (89%, 24/27) and approximately two-thirds achieving ≥10% weight loss (63%, 17/27). At week 4, weight change was significantly greater among those who completed the program compared to those who did not (−4.2% vs 1.4%; P=.03). A waterfall plot of weight change for each patient is shown in Figure 2.

Changes in cardiometabolic laboratory data at 24 weeks compared to baseline for program completers who had follow-up laboratory data are shown in Multimedia Appendix 2. Mean HbA_1c decreased by 0.7% (CI −0.9% to −0.3%), from 7.7% (SD 1.5%) to 7% (SD 0.9%), with 6 patients (22%) having an HbA_1c less than 6.5. On average, patients had significant reductions in total cholesterol (−18.9 mg/dL, SD 28) and triglyceride levels (−80.8 mg/dL, SD 103.2). Mean changes in LDL and HDL cholesterol levels were nonsignificant. In total, 42% (8/19) had a mean increase in LDL cholesterol of 15.8 mg/dL, and 58% (11/19) had a mean decrease in LDL cholesterol of −19.6 mg/dl. Waterfall plots of change in HbA_1c and LDL cholesterol for each patient are shown in Multimedia Appendices 3 and 4.

Table 4 shows changes in glucose-lowering medication use among program enrollees and program completers. At baseline, patients were prescribed an average of 3.5 (SD 0.8) glucose-lowering medications, which decreased to 2.4 (SD 0.9) at 24 weeks. Program completers were prescribed an average of 2.2 (SD 0.9) glucose-lowering medications at 24 weeks. Those who discontinued at 12 weeks were prescribed an average of 2.7 (SD 0.5) glucose-lowering medication, while those who discontinued the program after 4 weeks did not have any medication changes. Among those who completed the program, 21/27 (78%) were prescribed 2 or fewer medications at 24 weeks compared to 3 of 27 (11%) at baseline (Table 4 and Multimedia Appendix 5). Most patients discontinued use of short-acting insulin (24/27, 89%) and long-acting insulin (18/27, 67%) during the program period (Table 4). Changes in TDD of insulin are in Multimedia Appendix 6. Also, fewer patients used SGLT2 inhibitors, sulfonylureas, and meglitinides compared to baseline; however, these changes were nonsignificant (Table 4).

Notably, of the 27 participants who completed the program, 15 (56%) were prescribed a GLP-1 RA. Those who were prescribed a GLP-1 RA lost an average of 13% body weight (16.4 kg), compared to 10% (11.6 kg) for those (n=12) who were not prescribed a GLP-1 RA. Of the 24 participants who achieved at least 5% body weight loss, 14 (58%) were prescribed a GLP-1 RA at week 2. Of the 17 people who achieved at least 10% body weight loss, 11 (65%) were prescribed a GLP-1 RA at week 24. Figure 2 shows patient-level weight change stratified by prescription of GLP-1 RA for those who completed the program.

There were no serious adverse events during the evaluation period. Overall, 18% of patients (7/38) experienced mild (level 1) hypoglycemia (blood glucose level between 54 mg/dL and 70 mg/dL), which was managed as an outpatient using standard protocols and medication de-escalation [29]. Side effects included mild gastrointestinal upset, including belching (n=1), diarrhea with use of semaglutide, resulting in medication discontinuation (n=1), nausea due to metformin, resulting in dose de-escalation (n=1), and increased feelings of anger (n=1).

This QI pilot evaluated the feasibility, acceptability, and clinical effectiveness of tailored low-carbohydrate nutrition counseling supported by continuous glucose monitoring (LC-CGM) among Veterans with T2D using insulin. Most Veterans referred to the LC-CGM program enrolled (38/43, 88%) and, among enrollees, mean engagement was 20 weeks, with approximately 3-quarters (27/38, 71%) completing the 24-week program. Additionally, the LC-CGM program appeared to be an effective and safe intervention to support significant reductions in weight, HbA_1c, and use of glucose-lowering medications, including insulin.

Our team previously observed high levels of engagement (77%) in a local pilot evaluation of a group-based, VLC diabetes prevention program for Veterans with prediabetes [13]. In contrast, only 25% of Veterans who enroll in the VHA’s MOVE! Weight Management Program, a comprehensive lifestyle intervention that teaches Veterans to follow a conventional, calorie-restricted diet, complete 8 or more sessions within 6 months [33]. Taken together, these findings support prior literature suggesting that some Veterans may prefer a carbohydrate-restricted approach to dietary change [12,13,34].

Prior work has similarly demonstrated the effectiveness of VLC nutrition counseling among individuals who chose to participate in a continuous care intervention (Virta Health) consisting of individualized low-carbohydrate nutrition counseling focused on achieving nutritional ketosis, health coaching, and remote data monitoring (eg, weight and finger-stick blood glucose) [35,36]. A long-term (24-month) evaluation of this intervention demonstrated high levels of program retention (194/262, 74%), with sustained reductions in weight (−10%), HbA_1c (−0.9%), and prescriptions for T2D medications (−67%) compared to baseline (vs no change in the usual care group) [37]. In 2019, the Department of Veterans Affairs initiated a partnership with Virta Health to pilot test the intervention among Veterans with T2D and overweight or obesity to provide “a more comprehensive approach to care.” [38] At 24 months, weight loss was significantly greater among Veterans in the intervention versus waitlist control group (−12 kg vs 3 kg). Both groups had a reduction in HbA_1c (−0.79 % and −0.59% respectively). However, among intervention participants, this was achieved through dietary carbohydrate restriction with concomitant de-escalation of glucose-lowering medications; control participants required the use of more glucose-lowering medications to reduce HbA_1c [39]. Accordingly, monthly per-patient outpatient and pharmacy drug costs were significantly lower among intervention versus usual care participants [39]. Additional work is needed to evaluate the cost-effectiveness of low-carbohydrate nutrition counseling programs, including LC-CGM, which often require intensive support in the near term but may ultimately reduce health care and medication use.

Although overall use of glucose-lowering medications, including insulin, decreased during the study period, there was a significant increase in use of GLP-1 RAs, with 39% (15/38) of LC-CGM patients using a GLP-1 RA at baseline and 61% (23/38) using a GLP-1 RA at 24 weeks. This finding contrasts with recent literature demonstrating that carbohydrate-restricted diets can be used to de-escalate or avoid GLP-1 RA prescriptions [40] and support T2D remission, defined as HbA_1c <6.5% at least 3 months after stopping glucose-lowering medications [9,41]. This is due, at least in part, to 2 key factors. First, the LC-CGM program was intended to optimize Veterans’ glucose-lowering medication regimen to support weight loss and improved metabolic health; T2D remission was not an explicit goal of this program, though 22% (6/27) of program completers achieved HbA_1c <6.5% (Multimedia Appendix 7 for medication use among participants with HbA_1c <6.5% at 24 wk). Second, this was a highly pragmatic quality initiative with inclusion of patients who are less likely to achieve T2D remission (eg, older age, suboptimal glycemic control, use of multiple glucose-lowering medications, and the use of insulin) [42]. Notably, mean weight loss among Veterans prescribed a GLP-1 RA in the LC-CGM program (–10.2%) was greater than expected from clinical efficacy trials of GLP-1 RAs among patients with T2D, with the greatest weight loss (−9.6%) achieved with semaglutide [43]. Given the high cost of GLP-1 RAs [44] and prior reports of their attenuated effectiveness in general practice settings [45,46], further work is warranted to examine the clinical and cost-effectiveness of the LC-CGM program in a large-scale trial.

A key strength of the LC-CGM program is its use of existing VA infrastructure and resources, including clinical team members, and its alignment with VHA’s priorities to deliver team-based, coordinated, and patient-centered care to help Veterans achieve their personal health goals [47,48]. Due to the intensive nature of the program, inclusion of Veterans at high risk for adverse events (eg, hypoglycemia and hypotension), and need for proactive, early deprescribing of glucose-lowering medications [16,49], we were able to support a maximum of 10 Veterans at a time with 6 hours per week of dedicated RD effort and 4 hours per week of dedicated CPP effort. Additional work is needed to explore opportunities to increase LC-CGM capacity; this may include use of group-based and asynchronous education sessions, integration of digital health interventions (eg, text messages and artificial intelligence) [50], and inclusion of patients at lower risk for hypoglycemia (eg, individuals with prediabetes, new diagnosis of T2D, or T2D not requiring insulin) [51]. If offered as a group-based program, LC-CGM’s visit frequency and overall duration would align with other group-based comprehensive lifestyle interventions, including the Centers for Disease Control and Prevention’s Diabetes Prevention Program and the VHA’s MOVE! Weight Management Program. Notably, MOVE! is also offered as a one-on-one program, with 12 sessions more than 12 months with a dietitian. LC-CGM’s involvement of a CPP increases staffing requirements compared to MOVE! but it may not be required if the LC-CGM program was offered to patients at lower risk for hypoglycemia. Additional work is needed to understand if LC-CGM’s short-term, resource-intensive nature is offset by future reductions in health care use, including primary care visits, emergency department encounters, and inpatient hospitalizations.

This study had several limitations. First, this was a QI program evaluation from a single VA health care system; the results may not be generalizable to other VA sites or non-Veteran populations. Second, due to the pragmatic nature of the LC-CGM program referral process, we do not know the number of Veterans who were offered LC-CGM participation but declined referral; this information should be collected in a larger scale trial to further discern the program’s acceptability. Third, we did not evaluate outcomes beyond 6 months, and we are therefore unable to assess the program’s long-term clinical effectiveness. Prior literature demonstrates sustained clinical benefits of dietary carbohydrate restriction for up to 5 years among a non-Veteran population [52]. Fourth, we did not formally assess dietary carbohydrate intake using validated measures. Rather, we considered reductions in weight, HbA_1c, and medication use as surrogate indicators of dietary adherence. Fifth, due to the virtual nature of the program, most weight data were obtained through the Veteran’s self-report and may differ from in-clinic measured weight. Notably, prior literature suggests concordance between self-reported and measured weight [53], and objective favorable changes in HbA_1c and medication use among LC-CGM patients suggest accuracy of self-reported weight in this cohort. Sixth, we were unable to assess the program’s cost-effectiveness due to the limited, short-term nature of the available data and the nonpublic availability of the VA’s negotiated drug prices, which are often substantially lower than those in non-VA settings. Notably, other intensive lifestyle interventions focused on carbohydrate restriction and medication de-prescription have demonstrated cost savings [54].

This QI program evaluation demonstrates that individually tailored, LC-CGM can reduce body weight, HbA_1c, and use of glucose-lowering medications among Veterans with T2D on insulin. This approach aligns with clinical practice guideline recommendations that encourage preference-sensitive nutrition counseling using evidence-based approaches, including low- and VLC diets [11,55]. To inform approaches to T2D management in other VA Medical Centers nationwide and enhance Veterans’ access to guideline-adherent carbohydrate-restricted nutrition counseling, the LC-CGM program should be evaluated in a larger scale implementation initiative with examination of longer term clinical outcomes and program costs, including use of GLP-1 RAs.