Nov 5, 2026 · @Yoon Hang Kim MD

Abstract

Carbohydrate restriction has emerged as one of the most discussed dietary strategies for managing type 2 diabetes, yet conflicting claims about "reversal" and "cure" obscure what the evidence actually supports. This review examines the published clinical data from Virta Health's continuous-care intervention (one-year through five-year outcomes), the foundational Duke University studies under Eric Westman, MD (including the 2008 randomized trial comparing ketogenic and low-glycemic-index diets), and the Stanford Keto-Med crossover trial. Short-term glucose lowering and medication reduction are well supported by randomized and nonrandomized studies. Sustained benefit at two to five years is demonstrated in selected participants, though with significant attrition. Medication-free remission — defined by international consensus as HbA1c below 6.5% maintained at least three months without glucose-lowering medication — occurs in a meaningful but modest subset, reaching approximately 20% of Virta completers at five years. Comparative trial data and meta-analyses suggest that the metabolic benefits of carbohydrate restriction may not require sustained ketosis, and that the independent contribution of ketone production remains uncertain. The most clinically transferable finding across these studies is that carbohydrate restriction, delivered as a structured medical intervention with proactive medication adjustment, behavioral support, and long-term follow-up, can meaningfully improve glycemic outcomes and reduce medication burden in type 2 diabetes.

Few topics in metabolic medicine generate as much enthusiasm — or as much confusion — as low-carbohydrate diets for type 2 diabetes. Scroll through any physician forum or patient community and you will find confident claims on every side: that carbohydrate restriction "reverses" diabetes, that it is dangerous, that ketosis is essential, that it is unnecessary.

The truth, as usual, lives in the data — and the data tell a more nuanced story than any headline. In this post, I walk through the published clinical evidence from Virta Health, the foundational Duke University studies under Eric Westman, MD, and the comparative trial data that challenge the assumption that ketosis itself is the indispensable ingredient. I also address why the words we use — reversal, remission, cure — matter far more than most practitioners realize.

My goal is not to advocate for or against a ketogenic approach. It is to give you the evidence as it actually stands, so you can make informed decisions for your patients — or for yourself.

Virta Health: What the Clinical Evidence Actually Shows

Virta is not simply a ketogenic diet prescription. The studied intervention combines carbohydrate restriction (initially below 30 g/day), home glucose and ketone monitoring, health coaching, telemedicine support, and active medication adjustment. Understanding that distinction matters — because the outcomes reflect the entire package, not diet alone.

The Original Study and Two-Year Follow-Up

The principal study enrolled 262 adults who chose Virta's continuous-care intervention and 87 who chose usual care. It was prospective and controlled, but not randomized. Participants had established disease with an average diabetes duration of approximately eight years and baseline HbA1c of roughly 7.6%.

At one year, approximately 83% of participants remained enrolled — good engagement for an intensive dietary intervention. About 60% of completers achieved an HbA1c below 6.5% while taking no glucose-lowering medication or metformin alone. That is an impressive clinical outcome, but it is important to note: this is not the same as medication-free remission.

At two years, 194 intervention participants had follow-up data — approximately 74% of the original 262. The proportion using glucose-lowering medication other than metformin fell from 55.7% to 26.8%. Insulin use declined substantially, and sulfonylureas were eliminated in the reported analysis. Using intention-to-treat modeling, the investigators estimated 53.5% "reversal" and 17.6% remission by the definitions employed at the time.

A critical denominator issue: Virta's own subsequent analysis gives approximately 55% reversal among those remaining enrolled, versus 38% of everyone who originally started under a more conservative calculation. These figures differ based on how missing follow-up data are handled — and it matters which number you cite.

The Five-Year Extension: Durable Benefit, but More Modest Than Early Headlines

The five-year extension, published in Diabetes Research and Clinical Practice in 2024, provides a useful longer-term reality check. Of the 262 originally enrolled, 122 participants completed five-year follow-up — about 47% of the original cohort.

Among those five-year completers, approximately 20% met international-consensus remission criteria. Approximately 32.5% met the broader "reversal" definition allowing metformin. Average weight remained about 7.6% below baseline. Triglycerides and fasting insulin remained lower, HDL remained higher, and diabetes medication use remained reduced.

The average HbA1c reduction at five years was much smaller than the early response — approximately 0.3 percentage points below baseline — despite sustained medication reduction. Clinically, maintaining similar or somewhat better glycemia with fewer medications remains meaningful, even when full remission is not achieved.

My Assessment of Virta

Virta offers persuasive evidence of feasibility and sustained benefit for a subset of patients with established type 2 diabetes. Its most clinically compelling finding is improved glycemia alongside medication deintensification — not simply weight loss.

Its main limitations deserve candid acknowledgment: self-selection rather than randomization, a multicomponent treatment package that makes the independent contribution of ketosis uncertain, substantial attrition by five years, company funding and author affiliations, and outcomes centered on biomarkers and medication use rather than definitive reductions in cardiovascular events, kidney failure, or mortality.

These limitations qualify the claims. They do not negate the observed improvements.

Duke University and Eric Westman, MD: The Foundational Studies

Westman's work with William Yancy and colleagues is important because it includes both an early clinical pilot and a randomized comparison — not just practice testimonials.

The 2005 Pilot: 16 Weeks, No Control Group

Published in Nutrition & Metabolism, this uncontrolled, single-arm study enrolled 28 adults with type 2 diabetes and prescribed less than 20 g of carbohydrate per day. Twenty-one completed the study.

HbA1c dropped from 7.5% to 6.3% — an absolute reduction of 1.2 percentage points. Among completers, diabetes medications were discontinued in seven participants, reduced in ten, and unchanged in four. Glucose control improved even as medications came down — the same pattern later seen in the Virta data.

The absence of a control group and the small sample size prevent strong comparative conclusions, but the signal was clear and laid the groundwork for what came next.

The 2008 Randomized Trial: Keto vs. Low-Glycemic-Index

The more rigorous study, also published in Nutrition & Metabolism, randomized participants to either a ketogenic diet (less than 20 g carbohydrate/day, no explicit calorie restriction) or a low-glycemic-index diet (approximately 55% of calories from carbohydrate, with a prescribed 500-kcal/day deficit). Both groups received regular group visits, exercise recommendations, supplements, and physician-directed medication adjustment.

At 24 weeks, the ketogenic arm showed a 1.5 percentage-point HbA1c reduction compared with 0.5 in the low-glycemic-index group. Weight loss was 11.1 kg versus 6.9 kg. Among ketogenic-arm completers, 95.2% had diabetes medications reduced or eliminated, compared with approximately 62% in the comparator arm. HDL cholesterol rose 5.6 mg/dL in the keto group with approximately no change in the other.

The between-group differences favored the ketogenic arm on these outcomes. However, the frequently quoted 95.2% medication-reduction result comes from 20 of 21 completers — not 95% of everyone randomized. About 50 of the original participants completed follow-up, and after adjustment for weight change, the between-group HbA1c difference was borderline at p = 0.06.

This is a positive randomized trial demonstrating clinically meaningful short-term benefit. But substantial dropout and short follow-up make it inadequate evidence for a broad claim of durable "diabetes cure."

Duke's Broader Clinical Experience

In a 2020 Duke report, Westman stated that approximately 5,000 patients had participated in the program since 2006, emphasizing real foods generally below 20 g total carbohydrate daily, together with counseling and follow-up.

That is substantial practice experience, and it deserves respect. But it is not equivalent to a prospectively followed 5,000-patient research cohort. The Duke report does not provide a standardized all-patient remission rate, long-term retention rate, or controlled cardiovascular outcomes.

Is Ketosis the Magic Ingredient?

This is the key unresolved scientific distinction, and it deserves honest treatment.

The Virta and Duke interventions change several things simultaneously: carbohydrate exposure, calorie intake, body weight, medication use, and behavioral support. Their results establish the effectiveness of the overall treatment strategy more clearly than they isolate a specific benefit from ketone production itself.

The Stanford Keto-Med Trial

The Stanford Keto-Med randomized crossover trial offers useful perspective. Forty adults with prediabetes or type 2 diabetes were assigned to ketogenic and Mediterranean-plus diets for 12 weeks each, in random order. Both emphasized nonstarchy vegetables and avoided added sugar and refined grains. The Mediterranean-plus diet retained legumes, fruit, and intact whole grains.

HbA1c improved on both diets, without a statistically significant difference in the primary analysis. The ketogenic diet lowered triglycerides more, but LDL cholesterol increased approximately 10% — compared with a 5% decrease during Mediterranean-plus eating. Fiber and several nutrient intakes were lower on the ketogenic arm, and follow-up suggested lower sustainability.

This was a small, short study with crossover and medication-adjustment limitations. But it argues against claiming that every patient needs sustained ketosis to obtain substantial glycemic benefit.

What the Meta-Analyses Say

A meta-analysis of eight randomized trials involving 648 participants reached a compatible conclusion: very-low-carbohydrate ketogenic diets produced roughly 0.6 percentage points greater HbA1c reduction than comparator diets at three to six months, but were not superior for HbA1c or weight at 12 months. Medication reduction and triglyceride/HDL benefits could persist longer.

The pattern across these studies suggests that carbohydrate restriction is a powerful metabolic lever — but that the additional benefit of deep ketosis, beyond what moderate carbohydrate restriction achieves, may be smaller and less durable than early proponents assumed.

Reversal, Remission, and Cure: Why the Words Matter

Much of the apparent disagreement in this field comes down to terminology — and using these terms interchangeably causes real confusion for both patients and clinicians.

Improved control means lower HbA1c, with or without medication reduction. Meaningful? Absolutely. But it is not the same as the disease going away.

Reversal, as Virta uses the term, means HbA1c below 6.5% without diabetes medication other than metformin. This is a strong clinical result — but it still allows a core diabetes medication.

Consensus remission, per the 2021 international expert panel, requires HbA1c below 6.5% persisting at least three months without any usual glucose-lowering medication. This is a stricter standard and the one most aligned with what patients hear when someone says their diabetes has been "reversed."

Cure is not the preferred term in the scientific literature, because hyperglycemia may recur and complication surveillance remains necessary.

Here is the practical takeaway: someone who achieves an HbA1c of 6.2% while continuing metformin has an excellent clinical outcome, but does not meet the standard medication-free remission definition. And conversely, stopping useful treatment solely to qualify for a remission label should never become the clinical objective.

What This Means at the Bedside

Both the early Duke studies and the Virta program explicitly adjusted medications when carbohydrate restriction began. That medication management is integral to the intervention — not an optional add-on. If you take nothing else from this post, take this: prescribing a low-carbohydrate diet without proactive medication adjustment is not evidence-based practice.

The most important clinical considerations:

Insulin and sulfonylureas require immediate review. Glucose can improve quickly on carbohydrate restriction, and hypoglycemia can occur if doses are left unchanged. Proactive dose reduction on day one is the standard of care in every published protocol.

SGLT2 inhibitors deserve special attention. Ketogenic eating can increase the risk of euglycemic diabetic ketoacidosis — a dangerous condition precisely because a normal or moderately elevated glucose can mask it. The medication and dietary plan require clinician review before initiation.

Lipid response varies. Follow the individual patient rather than assuming improved triglycerides and HDL guarantee an improved overall cardiovascular risk profile. LDL cholesterol can rise, as the Keto-Med trial demonstrated.

Adherence and nutritional quality are inseparable from outcomes. Assess whether the patient can sustain the restriction while maintaining adequate fiber and nutrient intake. Long-term adherence is the central limitation across the entire trial literature.

Continued diabetes surveillance is non-negotiable. Even patients who achieve remission still require ongoing glycemic testing and routine monitoring for diabetes-related complications. Remission is not graduation from medical care.

The Bottom Line

For a clinician considering this approach, here is what the evidence supports and what it does not.

Short-term glucose lowering and medication reduction are supported by randomized trials and consistent clinical studies. The effect is real, reproducible, and clinically meaningful.

Sustained benefit over two to five years has been demonstrated in selected participants — most prominently in Virta's nonrandomized program. The benefits attenuate over time but do not disappear.

Medication-free remission is achievable, but substantially less common than the frequently advertised "reversal" rate. At five years in the Virta data, approximately 20% of completers met consensus remission criteria.

Ketosis as the uniquely necessary mechanism, universal superiority over other high-quality dietary patterns, or permanent cure: none of these are established by the current evidence base.

The most transferable lesson from both Westman and Virta is not "tell patients to do keto." It is to deliver carbohydrate restriction as a structured medical intervention — with medication adjustment, behavioral support, and long-term follow-up. The diet is one component. The medical supervision is what makes it safe. And the sustained relationship is what makes it work.

Conclusion

The conversation around low-carbohydrate diets and type 2 diabetes has too often been framed as a binary: either carb restriction cures diabetes, or it is a dangerous fad. Neither framing serves patients well.

What the Virta, Duke, and comparative trial data tell us — taken together, with their strengths and limitations weighed honestly — is something more useful than either extreme. Carbohydrate restriction is a legitimate, evidence-supported metabolic intervention that can produce clinically meaningful improvements in glycemic control and reduce the burden of diabetes medications. For some patients, it produces remission. For many more, it produces meaningfully better control with fewer drugs. And for a significant number, adherence over years remains the limiting factor.

The evidence does not support the claim that ketosis is the uniquely necessary mechanism, that this approach is universally superior to other high-quality dietary patterns, or that it constitutes a permanent cure. What it does support is that when carbohydrate restriction is delivered with the same rigor we apply to any other medical intervention — careful patient selection, proactive medication management, behavioral support, and sustained follow-up — it earns its place in the therapeutic toolkit.

As clinicians, our job is not to champion a diet. It is to match the right intervention to the right patient, manage the transition safely, and stay engaged for the long haul. The patients who do best in these studies are not the ones who simply "went keto." They are the ones whose physicians treated this as medicine.

Yoon Hang Kim MD is a board-certified preventive medicine physician practicing integrative and functional medicine. He is the author of eight books and more than 25 peer-reviewed articles. Learn more at www.yoonhangkim.com.