Abstract
Aim
To evaluate the long‐term effectiveness and safety of a remote, medically supervised ketogenic nutrition therapy intervention for Veterans with type 2 diabetes (T2D).
Materials and Methods
This retrospective observational analysis examined de‐identified medical records of Veterans with T2D who engaged in a remote continuous care intervention. Outcomes were HbA1c, weight, diabetes medication use and cardiometabolic markers (lipids, liver enzymes, kidney function) among those who remained enrolled for 2 or 3 years. Separately, we assessed weight, glucose and medication changes at programme departure among Veterans who discontinued before 2 years. Outcomes were analysed across subgroups. Linear mixed‐effects models and paired t‐tests evaluated changes over time.
Results
Among 640 enrolled Veterans (mean age: 59 years, BMI: 35 kg/m2, HbA1c: 8.4%), 310 (49%) remained engaged at 2 years and 197 (33%) at 3 years. At both time points, HbA1c was reduced by approximately 0.8%, alongside decreases in diabetes medication use. Weight decreased by approximately 9% at both 2 and 3 years. Overall, reductions in HbA1c and weight were seen across subgroups. Veterans who discontinued before 2 years experienced metabolic improvements, with greater benefits among those enrolled at least 6 months.
Conclusions
For US Veterans, long‐term participation in a remote ketogenic nutrition therapy intervention was associated with sustained improvements in glycaemic control, weight, medication use and select cardiometabolic markers. A 0.8% HbA1c reduction is associated with meaningful reductions in diabetes‐related complications, highlighting the potential clinical relevance of these findings.
Keywords: diabetes mellitus type 2, glycaemic control, ketogenic diet, telemedicine, veterans
1. INTRODUCTION
Type 2 diabetes (T2D) is a growing public health concern, disproportionately affecting Veterans in the United States. 1 Veterans often present with unique health issues, including higher rates of comorbidities and specific psychosocial stressors 2 , 3 , 4 that complicate the treatment of this already challenging disease.
Managing T2D requires blood glucose control through medications and lifestyle interventions, but adherence is often poor due to factors such as limited disease understanding, lack of support, financial burdens from medications and medical care, limited healthcare access and medication side effects. 5 Medication intensification is common 6 but does not always align with patient preferences. 7 These challenges can be even greater among Veterans, who may face additional barriers stemming from their service and post‐service experiences. 8
Telemedicine helps address some barriers by offering remote access to care and support. However, interventions that are challenging to implement and sustain in person may have similar outcomes when delivered via episodic virtual care. 9 , 10 For example, the Veteran Health Administration's (VHA) MOVE! and TeleMOVE weight‐loss programmes have similar attendance (approximately 30% attending all 8 weekly sessions), with average weight loss below the 5% clinical significance threshold in both groups (−1.8% MOVE! and −3.6% for TeleMOVE). 9 Like standard in‐person care, most telemedicine services lack the comprehensive, continuous support needed to sustain adherence to diabetes regimens. These challenges underscore the need for innovative approaches that provide continuous support and address the multifaceted needs of individuals with T2D. For example, the virtual Advanced Comprehensive Diabetes Care (ACDC) programme, which includes regular remote monitoring and more comprehensive support, has improved HbA1c and other health outcomes among Veterans with persistently poorly controlled diabetes. 11 , 12
Virta Health is a commercial digital health solution designed to address the complex challenges of managing T2D through medically supervised ketogenic diet nutrition therapy, delivered via continuous remote care. Multiple randomized controlled trials have demonstrated the effectiveness of ketogenic diet nutrition therapy for improving glycaemic control, weight and other cardiometabolic outcomes in individuals with T2D relative to other dietary approaches. 13 Extending these findings, Virta's 2‐year clinical trial showed sustained reductions in HbA1c, significant weight loss, decreased diabetes medication use and improvements in inflammatory and cardiovascular markers, compared to usual care control. 14
Virta's continuous remote care model is well suited to the challenges Veterans face in managing T2D. By providing digital access to a care team, real‐time biomarker monitoring and personalized feedback, the programme ensures ongoing, high‐touch support regardless of location. This structured approach enhances adherence through individualized nutrition guidance and behavioural coaching, while the ketogenic diet may further support sustainability by reducing hunger and cravings. Finally, Virta's model aligns with patient preferences by offering opportunities to reduce or discontinue diabetes medications, addressing concerns about polypharmacy and side effects while supporting long‐term metabolic health.
Recognizing the significant health and economic burden of T2D among Veterans, the VHA partnered with Virta Health to pilot this intervention. Strombotne et al. published 5‐month and 2‐year outcomes from this pilot, employing a quasi‐experimental, difference‐in‐differences design to compare outcomes among Veterans enrolled in Virta's programme versus a waitlisted usual care group. 15 , 16 At 2 years, both groups achieved clinically significant HbA1c reductions, but between‐group differences were not statistically significant. However, Veterans in the Virta group experienced greater weight loss and reduced reliance on diabetes medications and had significant cost savings compared to the usual care group (see Data S1 for a detailed commentary on Strombotne et al.).
The current study builds on these findings in several ways. First, we extend follow‐up to 3 years for a longer‐term assessment of effectiveness and sustainability among those who remained enrolled in the programme. Second, we incorporate a broader range of health and safety markers—kidney, liver and cardiovascular function—as well as a more detailed evaluation of diabetes medication changes. Third, unlike Strombotne et al., who utilized an intent‐to‐treat approach without accounting for approximately 50% programme retention at 2 years, we focus specifically on outcomes among Veterans who remained enrolled over time. This approach offers critical insights into the clinical and cost value of long‐term participation. For those who discontinued the programme, we report outcomes at the time of departure to assess any meaningful clinical improvements achieved during participation. Finally, this study explores who benefits most from this intervention by analysing HbA1c and weight outcomes by enrolment demographic and medical characteristics—age, race, socioeconomic factors and enrolment HbA1c and diabetes medication use—as well as adherence during the first 6 months (ketone levels).
In sum, this study uses Virta medical record data to address critical gaps in understanding the long‐term impact of the VHA‐Virta pilot programme by focusing on three key objectives: (1) evaluating 2‐ and 3‐year outcomes, including HbA1c, weight, diabetes medication use and additional safety and health markers such as kidney, liver and cardiovascular function for Veterans who remained active in the programme; (2) assessing health changes at the time of programme departure for Veterans who discontinued the intervention; and (3) exploring how enrolment demographic and clinical characteristics, as well as adherence levels, influence outcomes to better understand for whom this intervention is most effective. These findings provide insights into the programme's long‐term effectiveness, sustainability and impact across diverse Veteran subgroups, informing strategies to optimize care and better meet the needs of this high‐risk population.
2. MATERIALS AND METHODS
2.1. Study population and design
This retrospective observational analysis evaluated outcomes from the VHA pilot of a commercial digital health programme (Virta Health) providing continuous remote care and ketogenic nutrition therapy for T2D management. The analysis includes cohorts of Veterans who enrolled and were followed over 2‐ or 3‐year periods, with separate analyses for a cohort that discontinued before 2 years.
The VHA pilot was conducted nationwide, with all care delivered via telemedicine through a mobile app. Veterans used the app to communicate regularly with a dedicated health coach and medical provider, receiving continuous support and individualized guidance on nutritional ketosis. The programme incorporated a low‐carbohydrate, high‐fat, moderate‐protein ketogenic diet, biometric monitoring (e.g., blood ketones, glucose and weight), and personalized care adjustments. Veterans could message their care team at any time, with providers proactively monitoring biometric data and responding as needed. This real‐time feedback allowed individualized treatment modifications and timely intervention. This telemedicine model enhanced accessibility, overcoming geographic and logistical barriers common in in‐person care.
Data were obtained from Virta's de‐identified medical records only. Outcomes were assessed at enrolment, 2 and 3 years to evaluate effectiveness and safety in a real‐world Veteran population. The study was exempt from ethics committee approval under Health Insurance Portability and Accountability Act (HIPAA) standards as it did not involve identifiable human subjects.
2.2. Study population
This analysis included Veterans with T2D who participated in the VHA‐Virta pilot. Veterans were eligible for the pilot if they were enrolled in VHA healthcare, had a T2D diagnosis (defined as HbA1c ≥6.5%) and used at least one diabetes medication other than metformin. Veterans also had to be ≥18 years old and free of advanced renal, cardiac, hepatic or other chronic diseases.
2.3. Outcomes and study measures
All data were extracted from Virta's medical records. The primary outcomes of the study were changes in HbA1c and weight at 2 and 3 years. The secondary outcomes included diabetes medication use and additional cardiometabolic markers including lipids (total cholesterol, triglycerides, HDL cholesterol, LDL cholesterol), liver enzymes (alanine transaminase, aspartate transaminase and alkaline phosphatase) and kidney markers (creatinine, estimated glomerular filtration rate (eGFR) and urinary albumin‐to‐creatinine ratio). The eGFR was calculated using the most recent Chronic Kidney Disease Epidemiology Collaboration (CKD‐EPI) equation. 17 Non‐HDL cholesterol was calculated by subtracting HDL cholesterol from total cholesterol, and LDL cholesterol was recalculated using the 2020 National Institutes of Health equation. 18 Outcomes were assessed according to the study's three objectives:
2.3.1. Objective 1
Among Veterans who remained enrolled in the programme, we assessed primary and secondary outcomes at 2–3 years.
2.3.2. Objective 2
Among Veterans who discontinued before 2 years, we assessed:
Percent change in weight (calculated as the average of weights collected via cellular‐connected scale 30 days before the last logged weight).
Percent change in glucose (average of glucose values collected via Bluetooth‐connected meter 30 days before the last logged glucose).
Percent change in diabetes medication prescriptions.
2.3.3. Objective 3
To evaluate who benefited most from the intervention, we analysed HbA1c and weight by:
Age: <65 or ≥65 years.
Race/ethnicity: White or non‐White.
Socioeconomic conditions: Area Deprivation Index (ADI). 19
Enrolment HbA1c: <8 or ≥8%.
Enrolment insulin use: Yes or No.
Enrolment sulphonylurea use: Yes or No.
Ketone adherence: Average ketone biomarkers entered in the app during the first 6 months of care.
2.4. Statistical analysis
All analyses were conducted using R version 4.2.2 or SPSS Statistics version 30. A P value <0.05 was considered statistically significant.
Enrolment characteristics were described and compared between Veterans who remained enrolled and those who returned to usual care at 2 and 3 years. Group differences were assessed using t‐tests or Kruskal–Wallis tests for continuous variables (mean [SD]) and Chi‐squared or Fisher's exact tests for categorical variables (n [%]). Normality was assessed using a skewness cutoff of ±4 and kurtosis cutoff of ±7, following Kline's guidelines. 20 When these criteria were not met, the top 1% of outliers were removed to improve distributional fit.
2.4.1. Objective 1
For each cohort (2‐year and 3‐year), on‐care effects over time were evaluated using linear mixed‐effects models (LMMs) for continuous outcomes. LMMs accounted for fixed and random effects, handling repeated measures and within‐subject correlations. Time was modelled as a fixed effect at enrolment, 6, 12, 18 and 24 months for the 2‐year analysis, with additional time points at 30 and 36 months for the 3‐year analysis. Covariates included age, sex, race/ethnicity, enrolment diabetes medication use and enrolment BMI (kg/m2). LMMs utilize maximum likelihood estimation to handle missing data, enabling an analysis of all Veterans enrolled for 2 or 3 years.
Diabetes medication data were complete for all Veterans. McNemar's test (with continuity correction when needed) assessed changes in the percentage of Veterans prescribed each diabetes medication class. Paired t‐tests assessed: (1) changes in the total number of diabetes medications prescribed and (2) changes in insulin dose among those taking it at both enrollment and follow‐up.
2.4.2. Objective 2
We were unable to ascertain the precise reasons for patient discontinuation, as they could not be contacted at that time. Therefore, paired t‐tests assessed changes in weight, glucose levels, and the average number of diabetes‐specific medications (excluding metformin) from enrollment to programme departure among Veterans who discontinued before 2 years.
2.4.3. Objective 3
For the 2‐year cohort, LMMs were used to assess differential effects on HbA1c and weight by enrollment characteristics. Stratification groups were modelled as fixed effects, with time × group interaction terms included to test differential effects over time. Subgroup definitions are described in the Outcomes and Study Measures section.
3. RESULTS
3.1. Veteran characteristics
Among 640 eligible Veterans enrolled from 2017 to 2022, 310 (49%) remained enrolled at 2 years and 197 (33%) at 3 years. Table 1 presents cohort characteristics. Veterans who remained enrolled were, on average, 2 years older, had lower enrollment HbA1c (by 0.3–0.4) and had greater ketone adherence in the first 6 months. In the 3‐year cohort, 8% more Veterans ≥65 years remained enrolled. No other baseline characteristics differed between enrolled and non‐enrolled Veterans.
TABLE 1.
Veteran characteristics at enrollment.
| Characteristic | 2‐year cohort (n = 310) | 3‐year cohort (n = 197) |
|---|---|---|
| Age (years) | 58.8 (8.5) | 59.2 (8.5) |
| Age <65 | 226 (72.9) | 137 (69.5) |
| Age ≥65 | 84 (27.1) | 60 (30.5) |
| Female | 43 (13.9) | 25 (12.7) |
| Race/ethnicity | ||
| White | 194 (62.6) | 132 (67.0) |
| Non‐White | 94 (30.3) | 50 (25.4) |
| American Indian/Alaska Native | 3 (1.0) | 1 (0.5) |
| Asian | 2 (0.6) | 2 (1.0) |
| Black | 41 (13.2) | 20 (10.2) |
| Hispanic | 33 (10.6) | 19 (9.6) |
| Pacific Islander | 1 (0.3) | 1 (0.5) |
| Multiple | 14 (4.5) | 7 (3.6) |
| Unknown | 22 (7.1) | 15 (7.6) |
| ADI quintile | ||
| 1 | 63 (20.5) | 43 (21.8) |
| 2 | 85 (27.7) | 52 (26.4) |
| 3 | 65 (21.2) | 44 (22.3) |
| 4 | 60 (19.5) | 40 (20.3) |
| 5 | 34 (11.1) | 18 (9.1) |
| BMI (kg/m2) | 34.7 (6.7) | 34.6 (6.7) |
| Weight (lbs) | 240.0 (51.5) | 238.1 (49.0) |
| HbA1c (%) | 8.4 (1.5) | 8.4 (1.5) |
| HbA1c <8 | 144 (46.5) | 90 (45.7) |
| HbA1c ≥8 | 166 (53.5) | 107 (54.3) |
| LDL‐C (mg/dL) | 89.2 (37.0) | 87.0 (35.0) |
| HDL‐C (mg/dL) | 41.7 (14.6) | 41.9 (16.5) |
| Total cholesterol (mg/dL) | 172.2 (74.7) | 168.8 (66.3) |
| Triglycerides (mg/dL) | 196.2 (143.9) | 193.4 (145.7) |
| Alanine aminotransferase (U/L) | 34.2 (19.5) | 34.6 (21.1) |
| Aspartate aminotransferase (U/L) | 26.3 (13.4) | 27.3 (15.0) |
| Alkaline phosphatase (U/L) | 79.9 (24.7) | 79.8 (24.1) |
| Creatinine (mg/dL) | 1.02 (0.26) | 1.01 (0.27) |
| UACR (mg/g) | 72.1 (200.5) | 73.7 (223.7) |
| No. diabetes medications without metformin | 1.7 (0.8) | 1.66 (0.81) |
| No. diabetes medications | 2.38 (0.90) | 2.38 (0.91) |
| Taking insulin | 164 (52.9) | 105 (53.3) |
| Taking sulfonylurea | 117 (37.7) | 75 (38.1) |
| Ketone adherence category | ||
| 0.00–0.29 mM | 42 (13.5) | 17 (8.6) |
| 0.30–0.49 mM | 88 (28.4) | 54 (27.4) |
| 0.50–0.99 mM | 137 (44.2) | 96 (48.7) |
| ≥1.0 mM | 43 (13.9) | 30 (15.2) |
Note: Characteristics of Veterans who remained enrolled at 2 or 3 years. Values are presented as mean (standard deviation) for continuous variables and n (%) for categorical variables.
Abbreviations: ADI, area deprivation index; BMI, body mass index; HbA1c, hemoglobin A1c; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; UACR, urine albumin‐to‐creatinine ratio.
3.2. Objective 1: Two‐ and three‐year outcomes
Changes in primary and secondary outcomes are shown in Table 2 (2‐year cohort) and Table 3 (3‐year cohort). In both cohorts, HbA1c, weight, HDL‐C, alanine transaminase and aspartate transaminase improved significantly by 1 year and remained stable through 2–3 years. Total cholesterol, LDL cholesterol, triglycerides, non‐HDL cholesterol, alkaline phosphatase, eGFR, creatinine and UACR remained unchanged. Glycaemic, weight and other metabolic improvements coincided with reduced diabetes medication use. The average number of diabetes medications decreased at 2 and 3 years, along with the percentage of Veterans prescribed insulin and other diabetes medications. Among those prescribed insulin at both enrolment and follow‐up, the average dose decreased significantly (2 years: −68% or −50.9 units from 75.3; 3 years: −61% or −42.6 units from 69.6).
TABLE 2.
Two‐year cohort outcomes.
| Outcomes | Mean (SD) or % (n/N) | Mean change | ||
|---|---|---|---|---|
| Enrollment | 1 year | 2 years | ||
| Hemoglobin A1c (%) | 8.3 (0.1) | 7.3 (0.1) | 7.5 (0.1) | −0.8*** |
| Weight (lbs) | 229.8 (3.9) | 210.8 (3.9) | 209.9 (3.9) | −19.9*** |
| Medications prescribed | ||||
| Insulin | 53% (163/310) | 34% (105/310) | −19.0%*** | |
| Sulfonylureas | 37% (115/310) | 10% (32/310) | −27.0%*** | |
| DPP4i | 18% (56/310) | 9% (28/310) | −9.0%*** | |
| Thiazolidinediones | 8% (25/310) | 4% (13/310) | −4.0%* | |
| Incretin mimetics | 25% (76/310) | 28% (88/310) | 3.0% | |
| SGLT2i | 25% (76/310) | 21% (65/310) | −4.0% | |
| Metformin | 73% (227/310) | 64% (198/310) | −9.0%*** | |
| No. DM meds exc metformin | 1.7 (0.0) | 1.1 (0.1) | −0.6%*** | |
| Cardiovascular | ||||
| Total cholesterol (mg/dL) | 176.1 (4.3) | 176.8 (4.8) | 172.5 (4.6) | −3.6 |
| LDL‐C (mg/dL) | 100 (3.9) | 97.7 (4.4) | 95.8 (4.2) | −4.3 |
| HDL‐C (mg/dL) | 44.4 (0.9) | 49.3 (0.9) | 47.9 (0.9) | 3.5*** |
| Non‐HDL‐C (mg/dL) | 132.5 (4.3) | 127.8 (4.8) | 124.9 (4.6) | −7.6 |
| Triglycerides (mg/dL) | 175.5 (7.5) | 157.6 (8.3) | 159.5 (8.1) | −16.0 |
| Kidney | ||||
| eGFR (mL s−1 m−2) | 83.4 (1.4) | 84.0 (1.5) | 82.6 (1.5) | −0.8 |
| Creatinine (μmol L−1) | 0.94 (0.02) | 0.93 (0.02) | 0.95 (0.02) | 0.01 |
| UACR (mg/g) | 44.8 (13.6) | 57.9 (16.8) | 56.0 (14.9) | 11.2 |
| Liver | ||||
| Alanine aminotransferase (Units/L) | 29.6 (1.1) | 24.9 (1.2) | 25.2 (1.2) | −4.4*** |
| Aspartate aminotransferase (Units/L) | 24.2 (0.7) | 20.9 (0.8) | 21.6 (0.8) | −2.6* |
| Alkaline phosphatase (Units/L) | 81.8 (2.0) | 77.4 (2.2) | 79.5 (2.2) | −2.3 |
Note: Mean change is from Enrollment to 2 years. For medications prescribed, mean change is mean change in percent taking the medication. All changes that were significant at 2 years were also significant at 1 year. *Statistical significance: *p < 0.05; **p < 0.01; ***p < 0.001 for difference from enrollment. p value for medications is based off change in the proportion.
Abbreviations: eGFR, estimated glomerular filtration rate; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; UACR, urine albumin‐to‐creatinine ratio.
TABLE 3.
Three‐year cohort outcomes.
| Outcomes | Mean (SD) or % (n/N) | Mean change | |||
|---|---|---|---|---|---|
| Enrollment | 1 year | 2 years | 3 years | ||
| Hemoglobin A1c (%) | 8.3 (0.1) | 7.1 (0.1) | 7.5 (0.1) | 7.4 (0.1) | −0.8*** |
| Weight (lbs) | 228.6 (4.8) | 210.1 (4.8) | 209.8 (4.8) | 206.9 (4.8) | −21.8*** |
| Medications prescribed | |||||
| Insulin | 53% (104/197) | 36% (70/197) | −17.0%*** | ||
| Sulfonylureas | 38% (75/197) | 11% (21/197) | −27.0%*** | ||
| DPP4i | 20% (40/197) | 11% (22/197) | −9.0%** | ||
| Thiazolidinediones | 10% (19/197) | 4% (8/197) | −6.0%** | ||
| Incretin mimetics | 22% (44/197) | 33% (65/197) | 11.0%** | ||
| SGLT2i | 21% (42/197) | 27% (53/197) | 6.0% | ||
| Metformin | 72% (142/197) | 66% (131/197) | −6.0% | ||
| No. DM meds without metformin | 1.7 (0.1) | 1.2 (0.1) | −0.44*** | ||
| Cardiovascular | |||||
| Total cholesterol (mg/dL) | 171.5 (4.9) | 176.2 (5.6) | 172.3 (5.2) | 164.5 (5.3) | −6.9 |
| LDL‐C (mg/dL) | 94.7 (4.4) | 96.8 (5.1) | 94.1 (4.6) | 88.1 (4.8) | −6.6 |
| HDL‐C (mg/dL) | 44.8 (1.2) | 50.0 (1.2) | 48.2 (1.2) | 48.6 (1.2) | 3.9*** |
| non‐HDL‐C (mg/dL) | 126.8 (5.0) | 126.8 (5.6) | 124.1 (5.1) | 116.9 (5.3) | −9.9 |
| Triglycerides (mg/dL) | 173.5 (9.7) | 168.0 (10.9) | 166.1 (10.1) | 153.2 (10.3) | −20.3 |
| Kidney | |||||
| eGFR (mL s−1 m−2) | 85.2 (1.9) | 85.7 (2.0) | 84.4 (1.9) | 83.2 (1.9) | −2.0 |
| Creatinine (μmol L−1) | 0.91 (0.03) | 0.90 (0.03) | 0.93 (0.03) | 0.95 (0.03) | 0.03 |
| UACR (mg/g) | 33.4 (20.4) | 69.6 (25.5) | 53.6 (21.7) | 42.6 (22.9) | 9.3 |
| Liver | |||||
| Alanine aminotransferase (Units/L) | 29.5 (1.4) | 23.8 (1.6) | 24.4 (1.5) | 22.7 (1.6) | −6.7*** |
| Aspartate aminotransferase (Units/L) | 24.7 (0.9) | 20.6 (1.0) | 20.9 (1.0) | 20.0 (1.0) | −4.7*** |
| Alkaline phosphatase (Units/L) | 81.7 (2.6) | 78.5 (2.8) | 79.3 (2.7) | 76.3 (2.8) | −5.4 |
Note: Mean change is from Enrollment to 3 years. For medications prescribed, mean change is mean change in percent taking the medication. All changes that were significant at 3 years were also significant at 1 and 2 years. *Statistical significance: *p < 0.05; **p < 0.01; ***p < 0.001 for difference from enrollment. p value for medications is based off change in the proportion.
Abbreviations: eGFR, estimated glomerular filtration rate; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; UACR, urine albumin‐to‐creatinine ratio.
In addition to the primary analyses, sensitivity analyses were conducted using multiple imputation (100 imputations) to address potential missing data. The results of these analyses were consistent with the primary findings and did not substantially alter the interpretations. Therefore, these results are not presented.
3.3. Objective 2: Health changes at the time of programme departure for veterans who discontinued the intervention
Among Veterans who discontinued before 2 years, weight, glucose and medication use were assessed at programme departure (Table 4). Metabolic improvements were observed regardless of tenure, with greater benefits among those enrolled longer. Veterans participating for at least 6 months lost an average of 6% of their body weight, with those continuing longer achieving greater weight loss.
TABLE 4.
Outcomes at program departure.
| Days in program at departure | n | Weight Change (lbs), mean (SD) | % Weight change | n | Glucose change (mg/dL), mean (SD) | % Change in glucose | n | Change in medications, mean (SD) | % Change in medications |
|---|---|---|---|---|---|---|---|---|---|
| 1–180 | 68 | −7.7 (17.1)*** | −3.2% | 68 | −26.1 (91.8)* | −13.8% | 72 | −0.4 (0.6)*** | −25.5% |
| 181–365 | 95 | −14.4 (29.2)*** | −6.0% | 94 | −29.1 (93.5)** | −15.2% | 95 | −0.8 (0.8)*** | −45.5% |
| 366–545 | 91 | −14.0 (22.8)*** | −6.1% | 89 | −32.0 (73.8)*** | −17.4% | 91 | −0.7 (0.9)*** | −38.9% |
| 546–730 | 62 | −17.6 (35.1)*** | −7.3% | 62 | −26.3 (66.6)** | −15.1% | 62 | −0.5 (1.0)*** | −32.6% |
Note: n refers to the number of participants in each group and varies by outcome due to data availability. Medication refers to the number of diabetes medications excluding metformin. *Statistical significance: *p < 0.05; **p < 0.01; ***p < 0.001 for difference from enrollment.
3.4. Objective 3: Two‐year HbA1c and weight outcomes by veteran characteristics
Two‐year HbA1c and weight outcomes were analysed by enrolment characteristics (see Methods for subgroup definitions). Results are summarized below, in Figure 1, and in Data S2 for outcomes where differences were detected.
HbA1c outcomes differed only by enrolment HbA1c, with significant reductions seen in those starting at HbA1c ≥8% (Figure 1A). No significant differences were found by age, race, socioeconomic conditions, insulin or sulphonylurea use, or ketone adherence.
Weight outcomes differed by race, insulin use and ketone adherence, with all subgroups achieving clinically and statistically significant weight loss at 2 years. White Veterans lost more weight (−9.0% from 233.7 lbs., p < 0.001) than non‐White Veterans (−7.8% from 222.1 lbs., p < 0.001), a significant difference of approximately 1%. Veterans on insulin at enrolment lost more weight (−10.0% from 239.8 lbs., p < 0.001) than those not on insulin (−6.9% from 219.7, p < 0.001). Greater weight loss was also observed in those with higher ketone adherence during the first 6 months (Figure 1B). No significant differences in weight loss were found by age, socioeconomic conditions, enrolment HbA1c or sulphonylurea use.
FIGURE 1.
Changes in HbA1c and weight from enrollment to 2 years in key subgroups. (A) Mean HbA1c (%) at enrollment and 2 years for the full cohort and by baseline HbA1c category (<8% vs ≥8%). (B) Mean body weight (lbs) at enrollment and 2 years for the full cohort and by ketone adherence group, defined by average blood ketone levels during the first 6 months of care. Value labels at each time point indicate the mean for each group. Values displayed to the right of each line represent the absolute change in HbA1c (panel A) or the percent change in body weight (panel B) from enrollment to 2 years. The full set of subgroup analyses is provided in Data S2. HbA1c indicates hemoglobin A1c.
4. DISCUSSION
This study provides real‐world evidence that long‐term participation in a remote ketogenic nutrition therapy programme can lead to sustained and clinically meaningful metabolic improvements for Veterans with T2D. At 2 and 3 years, Veterans who remained enrolled experienced clinically significant reductions in HbA1c and weight while maintaining cardiovascular, liver and kidney markers at enrolment levels or within the normal range. Some of these markers also showed evidence of potential benefit, consistent with emerging evidence suggesting broader cardiometabolic advantages of carbohydrate‐restricted nutritional interventions. 14 , 21 , 22 , 23 By 2 and 3 years, Veterans lost an average of 9% of their initial body weight, similar to some pharmacologic treatments, including GLP‐1 receptor agonists 24 , 25 and approaching the threshold for bariatric surgery eligibility. HbA1c was reduced by nearly 1% over 3 years, even as diabetes medication use declined, reinforcing that glycaemic control was achieved through dietary changes rather than medication intensification typical of usual care. 6 A 1% reduction in HbA1c, even if transient, has been associated with a 21% lower risk of diabetes‐related death, a 14% lower risk of myocardial infarction, and a 37% lower risk of microvascular complications. 26 The long‐term participation rate of approximately 50% at 2 years is notable for an intensive lifestyle intervention and suggests many Veterans found the programme feasible and effective. This retention rate is particularly significant given that other VHA programmes, such as MOVE!, have reported much lower long‐term participation, with most participants dropping out within the first few sessions. 27
Even among those who discontinued participation before 2 years, meaningful metabolic improvements were observed at the time of departure, particularly among those enrolled for at least 6 months. This provides an important perspective on programme participation—while sustained participation is ideal, shorter‐term participation may still confer lasting benefits. These findings complement Strombotne et al.'s quasi‐experimental analysis of the same VHA‐Virta pilot, which included all enrolled Veterans regardless of tenure. Their study found significant overall health and cost benefits at 2 years, suggesting some Veterans who discontinued may have maintained or even built upon their health improvements after leaving the programme.
Across subgroups, the programme was generally effective, with clinically significant improvements observed in all groups. The only difference in glycaemic outcomes was that significant HbA1c reductions were seen only in those with enrolment HbA1c ≥8% (−1.3%). This is important because uncontrolled diabetes is strongly associated with macrovascular complications, such as cardiovascular disease, and microvascular complications, including neuropathy, retinopathy and nephropathy. 28 , 29 Persistently poor glycaemic control (A1c ≥8.5%) remains a significant concern among Veterans, affecting nearly 12% of those with T2D. 30 This result aligns with prior research demonstrating that individuals with higher initial HbA1c see the largest reductions in response to lifestyle and pharmacologic interventions. 31 In contrast, HbA1c remained stable (7.0% at 2 years) in those with baseline HbA1c <8%, though diabetes medication use decreased, showing that glycaemia could be maintained with less pharmacologic intervention. Importantly, this pattern provides insight into why Strombotne et al. did not find a statistically significant difference‐in‐difference in HbA1c between the Virta and usual care control groups; the higher baseline HbA1c in the control group likely contributed to greater reductions, attenuating the ketogenic nutrition programme's intervention effect. Beyond enrolment HbA1c, the programme was effective across all other clinical and demographic subgroups, including socioeconomic conditions (ADI). The consistency of impact reinforces the intervention's potential for broad applicability and equitable diabetes care. These findings suggest that targeting Veterans with higher HbA1c may yield the greatest glycaemic benefits. However, given that weight loss was observed across enrolment HbA1c subgroups, a broader approach that considers clinical benefits beyond glycaemic control may be most appropriate.
Weight outcomes were also clinically significant across all groups, though small differences were observed. Veterans on insulin at enrolment, who had a higher enrolment weight, experienced greater weight loss. This aligns with expectations, as insulin use is associated with weight gain. 32 In this ketogenic nutrition programme, some Veterans who started on insulin were able to discontinue it entirely, while others significantly reduced their dosage, which may have contributed to their weight loss. Greater weight loss was also observed in those with higher ketone adherence during the first 6 months, reinforcing prior research showing that early dietary adherence predicts long‐term success. 33 Mirroring common findings in weight loss interventions, including bariatric surgery, 34 other lifestyle interventions, 35 and GLP‐1 treatments, 36 a modest difference in weight loss was observed by race, with White Veterans losing 9.0% of enrolment weight versus 7.8% for non‐White Veterans. Although both groups achieved meaningful weight loss, future research should explore factors contributing to differential weight outcomes and strategies to enhance equitable results.
These findings, together with Strombotne et al.'s analysis, highlight that this intervention provides Veterans with an effective alternative to usual VHA diabetes care, where medication intensification is common. Results underscore the potential value of an approach that prioritizes carbohydrate restriction rather than medication intensification. While the lack of a control group in this study limits causal inference, Strombotne et al.'s matched‐control quasi‐experimental design helps address this limitation. However, differences in baseline characteristics, particularly the higher starting HbA1c in their control group, may have influenced their findings.
Additionally, although longer participation was associated with greater improvements, outcome analyses were limited to Veterans who remained enrolled at 2 and 3 years (49% and 33% of the original cohort, respectively), which may introduce bias. However, the risk is mitigated by the nature of the programme: participation was not time‐limited, and discontinuation may reflect a variety of factors—goal completion or perceived success— rather than disengagement or failure. Indeed, results from Objective 2 showed that many Veterans experienced meaningful improvements prior to earlier departure. Strombotne et al.'s intent‐to‐treat analysis of this same cohort, which included all enrolled Veterans regardless of tenure, supports this interpretation—finding significant improvements in health and cost outcomes at 2 years. Importantly, remaining enrolled in the programme does not necessarily equate to sustained engagement or adherence over time, as adherence behaviours beyond the initial phase of care were not assessed. However, early ketone data indicate that approximately 90% of Veterans restricted carbohydrates during the first 6 months (Table 1), and prior research shows early ketone adherence predicts long‐term outcomes. 37 Future research should explore reasons for programme exit, post‐exit trajectories and strategies to support sustained benefits beyond active participation.
Observed improvements in cardiovascular, liver and kidney markers align with prior research suggesting potential broader cardiometabolic benefits 14 , 21 , 22 , 23 and warrant further prospective investigation, particularly among Veterans at high risk for cardiometabolic complications. However, we did not assess clinical cardiovascular or renal outcomes, which limits the ability to draw conclusions about long‐term event risk. Although racial disparities in weight loss were small, identifying potential barriers—differences in socioeconomic factors, cultural dietary preferences or access to resources—and developing targeted strategies to ensure equitable outcomes will be important in optimizing this intervention for all Veterans.
This study provides real‐world evidence that a ketogenic nutrition therapy intervention, delivered via continuous remote care, is effective and safe for Veterans with T2D who participate, producing clinically meaningful and durable metabolic improvements comparable to leading pharmacologic therapies. Many Veterans face healthcare access barriers, high medication costs and a significant comorbidity burden, making a dietary intervention that supports diabetes management while reducing medication reliance particularly beneficial. This intervention's effectiveness across diverse demographic and clinical subgroups, including socioeconomic conditions, underscores its potential as an accessible treatment option. Expanding comprehensive virtual, evidence‐based interventions—both pharmacologic and lifestyle‐based—may further enhance patient‐centred care. Future work should explore how to best integrate such approaches within the VHA system to optimize long‐term diabetes management, expand patient choice and address barriers to sustained engagement.
CONFLICT OF INTEREST STATEMENT
RNA, SJA, and ARZ are current employees of Virta Health Corp. and have been offered stock options. ALM and RER are former employees of Virta Health Corp. and have been offered stock options.
PEER REVIEW
The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/dom.16525.
Supporting information
Data S1. Supporting Information.
Data S2. Supporting Information.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the Virta Health care team and support team members for their dedication in monitoring and caring for our patients. Thank you Drs. Matthew Crowley and Priya Shanmugam for your thoughtful review of drafts of this manuscript.
Adams RN, Athinarayanan SJ, Zoller AR, McKenzie AL, Ratner RE. Sustained metabolic improvements in a remotely delivered ketogenic nutrition programme for veterans with type 2 diabetes: A 3‐year observational study. Diabetes Obes Metab. 2025;27(9):4825‐4835. doi: 10.1111/dom.16525
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
REFERENCES
- 1. Liu Y, Sayam S, Shao X, et al. Prevalence of and trends in diabetes among veterans, United States, 2005–2014. Prev Chronic Dis. 2017;14:170230. doi: 10.5888/pcd14.170230 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Campbell SB, Gray KE, Hoerster KD, Fortney JC, Simpson TL. Differences in functional and structural social support among female and male veterans and civilians. Soc Psychiatry Psychiatr Epidemiol. 2021;56(3):375‐386. doi: 10.1007/s00127-020-01862-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Boersma P, Cohen RA, Zelaya CE, Moy E. Multiple chronic conditions among veterans and nonveterans: United States, 2015–2018. Natl Health Stat Report. 2021;153:1‐13. [PubMed] [Google Scholar]
- 4. SAMHSA, Center for Behavioral Health Statistics and Quality . National Survey on Drug Use and Health, 2008–2012.
- 5. García‐Pérez LE, Álvarez M, Dilla T, Gil‐Guillén V, Orozco‐Beltrán D. Adherence to therapies in patients with type 2 diabetes. Diabetes Ther. 2013;4:175‐194. doi: 10.1007/s13300-013-0034-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Turner RC, Cull CA, Frighi V, Holman RR, for the UK Prospective Diabetes Study (UKPDS) Group . Glycemic control with diet, sulfonylurea, metformin, or insulin in patients with type 2 diabetes mellitus: progressive requirement for multiple therapies (UKPDS 49). JAMA. 1999;281(21):2005‐2012. doi: 10.1001/jama.281.21.2005 [DOI] [PubMed] [Google Scholar]
- 7. Hayes RP, Bowman L, Monahan PO, Marrero DG, McHorney CA. Understanding diabetes medications from the perspective of patients with type 2 diabetes. Diabetes Educ. 2006;32(3):404‐414. doi: 10.1177/0145721706288182 [DOI] [PubMed] [Google Scholar]
- 8. Sheng T, Fairchild JK, Kong JY, et al. The influence of physical and mental health symptoms on veterans' functional health status. J Rehabil Res Dev. 2016;53(6):781‐796. doi: 10.1682/JRRD.2015.07.0146 [DOI] [PubMed] [Google Scholar]
- 9. Rutledge T, Skoyen JA, Wiese JA, Ober KM, Woods GN. A comparison of MOVE! Versus TeleMOVE programs for weight loss in veterans with obesity. Obes Res Clin Pract. 2017;11(3):344‐351. doi: 10.1016/j.orcp.2016.11.005 [DOI] [PubMed] [Google Scholar]
- 10. Al‐Badri M, Kilroy CL, Shahar JI, et al. In‐person and virtual multidisciplinary intensive lifestyle interventions are equally effective in patients with type 2 diabetes and obesity. Ther Adv Endocrinol Metab. 2022;13. doi: 10.1177/20420188221093220 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Crowley MJ, Tarkington PE, Bosworth HB, et al. Effect of a comprehensive telehealth intervention vs telemonitoring and care coordination in patients with persistently poor type 2 diabetes control: a randomized clinical trial. JAMA Intern Med. 2022;182(9):943‐952. doi: 10.1001/jamainternmed.2022.2947 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Kobe EA, Lewinski AA, Jeffreys AS, et al. Implementation of an intensive telehealth intervention for rural patients with clinic‐refractory diabetes. J Gen Intern Med. 2022;37(12):3080‐3088. doi: 10.1007/s11606-021-07281-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Rafiullah M, Musambil M, David SK. Effect of a very low‐carbohydrate ketogenic diet vs recommended diets in patients with type 2 diabetes: a meta‐analysis. Nutr Rev. 2022;80(3):488‐502. doi: 10.1093/nutrit/nuab040 [DOI] [PubMed] [Google Scholar]
- 14. Athinarayanan SJ, Adams RN, Hallberg SJ, et al. Long‐term effects of a novel continuous remote care intervention including nutritional ketosis for the management of type 2 diabetes: a 2‐year non‐randomized clinical trial. Front Endocrinol. 2019;10:348. doi: 10.3389/fendo.2019.00348 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Strombotne KL, Lum J, Ndugga NJ, et al. Effectiveness of a ketogenic diet and virtual coaching intervention for patients with diabetes: a difference‐in‐differences analysis. Diabetes Obes Metab. 2021;23(12):2643‐2650. doi: 10.1111/dom.14515 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Strombotne KL, Lum J, Pizer SD, Figueroa S, Frakt AB, Conlin PR. Clinical effectiveness and cost‐impact after 2 years of a ketogenic diet and virtual coaching intervention for patients with diabetes. Diabetes Obes Metab. 2024;26(3):1016‐1022. doi: 10.1111/dom.15401 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Inker LA, Eneanya ND, Coresh J, et al. New creatinine‐ and cystatin C‐based equations to estimate GFR without race. N Engl J Med. 2021;385(19):1737‐1749. doi: 10.1056/NEJMoa2102953 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Sampson M, Ling C, Sun Q, et al. A new equation for calculation of low‐density lipoprotein cholesterol in patients with normolipidemia and/or hypertriglyceridemia. JAMA Cardiol. 2020;5(5):540‐548. doi: 10.1001/jamacardio.2020.0013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Kind AJH, Buckingham WR. Making neighborhood disadvantage metrics accessible: the neighborhood atlas. N Engl J Med. 2018;378:2456‐2458. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Kline RB. Principles and Practice of Structural Equation Modeling. 3rd ed. The Guilford Press; 2011. [Google Scholar]
- 21. Athinarayanan SJ, Roberts CGP, Vangala C, et al. The case for a ketogenic diet in the management of kidney disease. BMJ Open Diabetes Res Care. 2024;12(2):e004101. doi: 10.1136/bmjdrc-2024-004101 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Watanabe M, Tozzi R, Risi R, et al. Beneficial effects of the ketogenic diet on nonalcoholic fatty liver disease: a comprehensive review of the literature. Obes Rev. 2020;21(8):e13024. doi: 10.1111/obr.13024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Dyńka D, Kowalcze K, Charuta A, Paziewska A. The ketogenic diet and cardiovascular diseases. Nutrients. 2023;15(15):3368. doi: 10.3390/nu15153368 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Wong HJ, Sim B, Teo YH, et al. Efficacy of GLP‐1 receptor agonists on weight loss, BMI, and waist circumference for patients with obesity or overweight: a systematic review, meta‐analysis, and meta‐regression of 47 randomized controlled trials. Diabetes Care. 2025;48(2):292‐300. doi: 10.2337/dc24-1678 [DOI] [PubMed] [Google Scholar]
- 25. Ryan DH, Lingvay I, Deanfield J, et al. Long‐term weight loss effects of semaglutide in obesity without diabetes in the SELECT trial. Nat Med. 2024;30(7):2049‐2057. doi: 10.1038/s41591-024-02996-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Stratton IM, Adler AI, Neil HA, et al. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ. 2000;321(7258):405‐412. doi: 10.1136/bmj.321.7258.405 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Spring B, Sohn MW, Locatelli SM, Hadi S, Kahwati L, Weaver FM. Individual, facility, and program factors affecting retention in a national weight management program. BMC Public Health. 2014;14:363. doi: 10.1186/1471-2458-14-363 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Colayco DC, Niu F, McCombs JS, Cheetham TC. A1C and cardiovascular outcomes in type 2 diabetes: a nested case‐control study. Diabetes Care. 2011;34(1):77‐83. doi: 10.2337/dc10-1318 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Zoungas S, Chalmers J, Ninomiya T, et al. Association of HbA1c levels with vascular complications and death in patients with type 2 diabetes: evidence of glycaemic thresholds. Diabetologia. 2012;55(3):636‐643. doi: 10.1007/s00125-011-2404-1 [DOI] [PubMed] [Google Scholar]
- 30. Alexopoulos AS, Jackson GL, Edelman D, et al. Clinical factors associated with persistently poor diabetes control in the veterans health administration: a nationwide cohort study. PLoS One. 2019;14(3):e0214679. doi: 10.1371/journal.pone.0214679 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Giugliano D, Maiorino M, Bellastella G, Chiodini P, Esposito K. Relationship of baseline HbA1c, HbA1c change and HbA1c target of <7% with insulin analogues in type 2 diabetes: a meta‐analysis of randomised controlled trials. Int J Clin Pract. 2011;65(5):602‐612. doi: 10.1111/j.1742-1241.2010.02619.x [DOI] [PubMed] [Google Scholar]
- 32. Russell‐Jones D, Khan R. Insulin‐associated weight gain in diabetes—causes, effects and coping strategies. Diabetes Obes Metab. 2007;9(6):799‐812. doi: 10.1111/j.1463-1326.2006.00686.x [DOI] [PubMed] [Google Scholar]
- 33. McKenzie AL, Athinarayanan SJ, Adams RN, Volk BM, Phinney S, Ratner RE. Mean blood beta‐hydroxybutyrate predicts clinically significant weight loss following 90 days carbohydrate restricted nutrition therapy. Diabetes. 2021;70(Suppl 1):307‐OR. doi: 10.2337/db21-307-OR [DOI] [Google Scholar]
- 34. Zhao J, Samaan JS, Abboud Y, Samakar K. Racial disparities in bariatric surgery postoperative weight loss and co‐morbidity resolution: a systematic review. Surg Obes Relat Dis. 2021;17(10):1799‐1823. doi: 10.1016/j.soard.2021.06.001 [DOI] [PubMed] [Google Scholar]
- 35. West DS, Dutton G, Delahanty LM, et al. Weight loss experiences of African American, Hispanic, and non‐Hispanic white men and women with type 2 diabetes: the look AHEAD trial. Obesity (Silver Spring). 2019;27(8):1275‐1284. doi: 10.1002/oby.22522 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Rubino D, Angelene H, Fabricatore A, Ard J. Efficacy and safety of semaglutide 2.4 mg by race and ethnicity: a post hoc analysis of three randomized controlled trials. Obesity (Silver Spring). 2024;32(7):1268‐1280. doi: 10.1002/oby.24042 [DOI] [PubMed] [Google Scholar]
- 37. Adams RN, McKenzie A, Athinarayanan SJ, Hallberg S, McCarter JP, Phinney S. Early engagement in a continuous care intervention predicts 1‐year improvements in weight and HbA1c among adults with T2D. Diabetes. 2019;68(Suppl 1):889‐P. doi: 10.2337/db19-889-P [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1. Supporting Information.
Data S2. Supporting Information.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.