Can Keto Reverse Type 2 Diabetes? What Virta's 2-Year Trial Found

After a type 2 diabetes diagnosis, most people get handed the same familiar path: start metformin, raise the dose a few years later, add another drug, and eventually maybe insulin. Most patients hear some version of "diabetes is a progressive disease" at least once. But a series of clinical trials published since 2018 suggests that progression isn't the only road — in some people, a low-carbohydrate eating pattern can genuinely push the disease back. This article looks at the study with the most comprehensive data on the topic: the continuous remote care model from Virta Health in the US. We cover the 1-year and 2-year results, how medication was actually reduced, and where the study falls short.

The Mechanism: Why Does Cutting Carbs Lower Blood Sugar?

Type 2 diabetes centers on insulin resistance: cells no longer respond to insulin as readily, and the liver keeps producing more glucose than it should. Cutting dietary carbohydrate reduces the main input that raises blood sugar, which eases both the immediate glucose load and, over time, the burden on the pancreas to produce insulin. The body starts meeting a meaningful share of its energy needs from ketone bodies produced in the liver — researchers call this "nutritional ketosis." Virta's model doesn't actually use the term "ketogenic diet"; it's described as a low-carbohydrate care model targeting nutritional ketosis. In practice, though, it usually brings daily carbohydrate down to roughly 20-50 grams, which overlaps substantially with a ketogenic approach in principle.

One important difference: this wasn't just a diet change. Virta's model also included continuous remote medical supervision, a health coach, and regular medication adjustments under a doctor's guidance. That distinction explains why the safety note further down matters so much.

The 1-Year Trial: Design and Numbers

Published in Diabetes Therapy in 2018, the trial enrolled 262 people in the continuous care intervention (CCI) group and 87 in a usual-care (UC) comparison group. This wasn't a randomized assignment — participants chose which group to join themselves, one of the study's most important limitations. By the end of year one, 218 CCI participants (83%) remained in the study.

In the CCI group, HbA1c fell on average from 7.6% to 6.3% (p<1.0×10⁻¹⁶), while the UC group showed no significant change. Weight loss in the CCI group averaged 13.8 kg, roughly 12% of starting body weight. Use of diabetes medications other than metformin fell from 56.9% to 29.7%, and sulfonylurea use was eliminated entirely, dropping from 23.7% to 0%. Of 78 completers who had been on insulin, 31 (40%) stopped it completely, and the remaining 47 reduced their average daily dose from 105.2 to 53.8 units — meaning 94% of insulin users either reduced their dose or discontinued it.

Lipid values shifted too: triglycerides dropped 24%, and HDL cholesterol rose 18%. Calculated LDL cholesterol rose 10% (p=5.1×10⁻⁵), but ApoB — often considered a more reliable marker of LDL particle count — didn't change significantly (p=0.37). This mixed lipid picture is its own separate topic worth understanding, and why LDL doesn't move the same direction in everyone on keto; we cover it in more depth in our article on LDL cholesterol.

The 2-Year Follow-Up: Did the Effect Hold, or Fade?

The real test of any dietary intervention usually comes after the first year — does the effect last, or do people drift back to old habits? The 2-year follow-up data, published by Athinarayanan and colleagues in Frontiers in Endocrinology in 2019, partly answers that. By the end of 2 years, 194 CCI participants (74%) and 68 UC participants (78%) remained in the study.

HbA1c in the CCI group fell from 7.7% to 6.7% (about a 12% drop, p=1.8×10⁻¹⁷), while the UC group's HbA1c actually rose from 7.5% to 7.9% — a highly significant between-group difference (p=1.3×10⁻⁹). Weight loss in the CCI group held at an average of 11.9 kg (~10%); 74% of CCI completers maintained at least a 5% weight loss, versus only 14% in the UC group.

The most striking number on diabetes reversal: in the intention-to-treat analysis, 53.5% of the CCI group met the "reversal" criterion (HbA1c under 6.5% with no diabetes medication other than metformin). 17.6% met the remission criterion (glycemic control with no diabetes medication at all), of which 6.7% was complete remission. In the UC group, these figures showed no improvement and stayed at just 2.4%. Among completers, use of medications other than metformin fell from 55.7% to 26.8%, sulfonylurea use was eliminated entirely, insulin use fell 62%, and among those on insulin at baseline, the average daily dose dropped roughly 81% (from 81.9 to 15.5 units).

A Critical Safety Note on Reducing Medication

These numbers shouldn't be read as "if I go keto, I can cut my own medication." In Virta's model, medication changes happened gradually, under continuous remote blood-sugar monitoring and physician oversight — a reminder that the trial's results rest as much on the "continuous medical care" component as on the diet itself. For anyone on insulin or a sulfonylurea in particular, cutting carbohydrate sharply while keeping the same medication dose carries a real risk of hypoglycemia (dangerously low blood sugar), which can be serious. If you have type 2 diabetes and take medication, talk to your doctor before switching to a low-carb way of eating, monitor your blood sugar closely, and only adjust medication doses under medical supervision.

What Does This Mean in Practice?

• The effect is real and held for two years, but the study wasn't randomized. Participants self-selected, which could be an independent factor in the results — people already motivated to try low-carb eating may simply have done better.
• "Reversal" holds only as long as the diet continues. The study doesn't show what happens after stopping; clinical experience and the physiology of diabetes both suggest blood sugar typically rises again after returning to a prior diet.
• Never attempt this on your own if you're on medication. Unsupervised carbohydrate restriction carries a real hypoglycemia risk for anyone on insulin or a sulfonylurea.
• Don't skip LDL monitoring. If your lipid panel shows a rise in LDL, discuss it with your doctor; we go into this in more depth in our article on LDL cholesterol.
• If you want to better understand the link between diabetes and insulin resistance, see our article on insulin resistance.
• To track other changes in your bloodwork, our article on blood test effects may help.

Limitations

In the authors' own words, this study used an "open-label, non-randomized, controlled" design; participants self-selected, and the two groups were recruited separately. The study was conducted at a single site, and most participants were white, which limits how far the results generalize to other populations. By year two, 26% of the CCI group and 22% of the UC group had dropped out; people who left may differ systematically from those who stayed. The most important point, though, is funding: the study was sponsored by Virta Health Corp, and a significant number of authors are company employees, founders, or stock-option holders. That doesn't make the data wrong, but it's a conflict of interest worth keeping in mind when weighing the results. Independent, randomized controlled trials would help clarify how far these findings generalize.

Sources: Hallberg, McKenzie et al., "Effectiveness and Safety of a Novel Care Model for the Management of Type 2 Diabetes at 1 Year: An Open-Label, Non-Randomized, Controlled Study", Diabetes Therapy (2018) · Athinarayanan 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", Frontiers in Endocrinology (2019)