Can a High-Carb Diet Reverse Type 2 Diabetes?

Blog · Nutrition & Diet

Can a High-Carb Diet Reverse Type 2 Diabetes?

By Vitae Team ·

A 2026 trial cut liver fat by a third on 75% carbohydrate. What it shows about fibre, remission and why the macronutrient argument misses the point.

The usual argument about type 2 diabetes is whether carbohydrate is the problem or the solution. A trial published in 2026 is an awkward fact for both camps: people eating three quarters of their calories as carbohydrate cut their liver fat by a third and dropped a full point of HbA1c. What that does and does not prove is the interesting part.

TL;DR

  • "Reverse" has a formal definition and this is not it. Remission means HbA1c below 48 mmol/mol, held for at least three months, with no glucose-lowering medication.
  • The evidence for what produces remission points at weight loss, not at a particular carbohydrate level. In the DiRECT trial, remission was 46% at one year and 10% at five.
  • A high-carbohydrate, high-fibre diet improved things substantially in a 2026 trial. HbA1c fell by a full percentage point and liver fat by 33% on a diet that was 75% carbohydrate.
  • That trial had no control group and ran 16 weeks, and its authors say the results cannot be separated from the weight loss that came with them.
  • Higher-fibre patterns did well in both comparisons here, though neither study can separate fibre from the other things that differed alongside it.
  • In a small 2026 trial of two very low-calorie formulas, the higher-fibre, higher-protein, lower-carbohydrate one produced more short-term remission by the study's own fasting-glucose definition. The design cannot say which component was responsible.
  • Time since diagnosis also mattered. Remission reached 82% in people diagnosed within four years and 50% in those diagnosed eight or more years ago.

What "reverse" means, and why the word is doing too much work

Remission has an agreed definition in the UK, and it is stricter than most people assume. Diabetes UK describes it as HbA1c returning below 48 mmol/mol, or 6.5%, and staying there for at least three months, with the person off glucose-lowering medication. It is not a cure. Diabetes can return, and people in remission still need at least annual monitoring, including retinal screening.

That definition matters for reading any claim about diet. A diet that improves blood glucose, reduces medication or lowers HbA1c by a point has done something worthwhile, and it has not produced remission. Most of what circulates online as "reversing" diabetes is the first thing described using the language of the second. One point cannot be skipped either. Remission requires coming off glucose-lowering medication, and in every trial discussed here that was done under clinical supervision with monitoring. Substantial changes to what you eat can shift blood glucose and the medication you need, even when the diet is not extreme, so anyone on glucose-lowering treatment should speak to their GP or diabetes team before making one.

What the trials show produces remission

The clearest evidence comes from DiRECT, which used a low-calorie soups-and-shakes programme rather than a particular macronutrient ratio. Remission reached 46% at one year and 36% at two. At five years it was 10%, or 12 of 118 people, against 5% in the control group. Among those who kept GP support through years three to five, 13% were in remission.

The pattern underneath those numbers is about weight. Participants still in remission at five years had lost an average of 8.9 kg, against 6.1 kg across their group as a whole, and regaining weight was what commonly ended remission. Over the full five years the intervention group spent 27% of the time in remission, compared with 4% for controls, which is a more honest summary than any single-year figure.

What DiRECT does not tell you is anything about carbohydrate. It restricted energy rather than a macronutrient, and was never designed to settle an argument about ratios. So whether you can arrive at the same place while eating plenty of carbohydrate remains open, which is what makes the newer work worth a close look.

The case for fibre, and the problem with it

The most direct attempt to settle high against low carbohydrate is a review by Reynolds, Lang, Brand and Mann, published in Obesity Reviews in 2025. It pooled 10 randomised trials covering 499 people, comparing higher-fibre, higher-carbohydrate diets with lower-carbohydrate, lower-fibre ones, and excluded trials that also restricted energy so that the diets themselves were being compared.

The higher-fibre arms came out ahead. HbA1c was 0.50 percentage points lower, about 5.5 mmol/mol, at moderate certainty. Fasting insulin was lower at high certainty, which is the firmest finding in the review. There was no overall difference in body weight. The authors' own conclusion is that carbohydrate amount is relatively unimportant and that fibre is what matters.

Two things should temper that. The most recent trial eligible for inclusion was published in 2012, so a 2025 review is summarising a literature that stopped more than a decade ago. And the "lower carbohydrate" comparison arms ran at 34 to 47% of energy, which the authors acknowledge may not meet common definitions of low carbohydrate. The review therefore compares high carbohydrate with moderate carbohydrate, not with anything a low-carb advocate would recognise.

The 2026 trial that cut liver fat by a third

The new evidence is a 16-week trial published in BMJ Open Diabetes Research & Care. Participants ate a low-fat vegan diet of roughly 75% carbohydrate, 15% protein and 10% fat, with no calorie target and no meals provided. Fibre intake rose from 24.0 to 41.4 grams a day. Twenty-four adults with type 2 diabetes completed it, and the researchers measured fat inside the liver and the muscle directly rather than inferring it from blood markers.

The results were substantial. HbA1c fell by a full percentage point, from 7.6% to 6.6%. Liver fat fell by 33%. Weight fell by 6.0 kg and visceral fat by 290 cm³. Measures of insulin sensitivity and beta cell function improved. Of the 17 participants taking metformin, five reduced their dose and two stopped, and nobody started or increased a glucose-lowering drug.

The limitations are serious and the authors state them clearly. There was no control group, the sample was small, and the trial ran 16 weeks. Despite no calorie target, reported intake fell by about 346 kcal a day, so the effects cannot be separated from eating less and losing weight. Of the 6.0 kg lost, 4.2 kg was fat mass, so roughly a third was not.

It is also worth knowing who paid. The trial was funded by the Physicians Committee for Responsible Medicine, an organisation that advocates plant-based eating and whose unpaid president is the study's senior author, with five authors compensated by it for the work. That is direct trial funding by an advocacy body, a stronger conflict than an advisory role.

Note also what the trial did not establish. Mean HbA1c of 6.6% remained above the remission threshold, most of those taking metformin stayed on it, and the study was not designed to test sustained medication-free remission. What the diet produced was a sizeable improvement in control over four months, which is a different claim from reversal and is worth reporting as one.

The trial that put diet composition to the test

One recent study did test composition with remission as its outcome, and it does not support the high-carbohydrate case. Published in Diabetes Research and Clinical Practice in February 2026, it randomised 47 adults to one of two very low-calorie formula diets of 600 to 800 kcal a day for three months, with antidiabetic medication withdrawn under supervision. That is a clinically supervised intervention, not a level of energy restriction to attempt on your own.

Using its own definition, fasting plasma glucose below 126 mg/dL at three months, the trial reported remission in 34 of the 47 participants. The higher-fibre, higher-protein, lower-carbohydrate, lower-fat formula did better, at 91% against 56%. Two things limit what that shows. The formulas differed in several nutrients at once, so the design cannot identify which component was responsible, and a fasting-glucose endpoint at three months is not comparable with the sustained HbA1c-defined remission reported in DiRECT.

Time since diagnosis also mattered. Participants diagnosed within the previous four years met the study's endpoint 82% of the time, against 50% of those diagnosed eight or more years ago, despite similar average weight loss. The researchers read that as beta cell function being more recoverable early on. With 47 people the trial cannot rank duration against composition, but both were associated with the outcome.

What This Actually Means

Put the three studies side by side and the macronutrient argument starts to look like the wrong argument. DiRECT cut energy hard and liver fat followed. The vegan trial cut fat and raised fibre, people spontaneously ate less, and liver fat fell by a third. Liver fat is the mechanism researchers propose sits underneath remission. Two very different diets arrive at the same place, which suggests the useful question is not which macronutrient to cut but which approach you can sustain long enough to lose the weight and keep it off.

On fibre, the honest position is narrower than the headline. Higher-fibre patterns performed well in both comparisons here, but neither study can separate fibre from the other things that changed alongside it. The 2025 review's diets differed in carbohydrate as well as fibre, and the 2026 formulas differed in several nutrients at once. Eating more fibre remains a defensible thing to do, and most people in the UK fall well short of recommended intakes, but the evidence does not isolate it as the active ingredient.

What has not been shown is remission. The 2026 high-carbohydrate trial did not assess sustained medication-free remission, so it cannot establish whether that approach produces it. High-carbohydrate, high-fibre eating looks like a reasonable way to manage type 2 diabetes on the evidence available. The case that it reverses the condition has not been made.

Frequently Asked Questions

Is type 2 diabetes remission permanent?

No. Diabetes UK is explicit that remission is not a cure, that symptoms can return, and that people in remission need ongoing monitoring at least annually, including retinal screening. In DiRECT, regaining weight was what commonly ended remission.

Does cutting carbohydrate work better than cutting fat?

For remission, the comparison has barely been tested. For day-to-day control, a 2025 review found higher-fibre, higher-carbohydrate diets slightly ahead on HbA1c, though its newest trial was from 2012 and its comparison arms were not low-carb by most definitions.

How much fibre should I be eating?

UK recommendations are 30 grams a day for adults and most people get nowhere near it. In the 2026 vegan trial, participants went from 24 to 41 grams a day, which shows the scale of change involved.

Can I come off my diabetes medication if my numbers improve?

That is a decision for your GP or diabetes team, not something to do on your own. In every trial described here, medication changes were made under clinical supervision with monitoring.

Does it matter how long I have had diabetes?

It appears to matter. In the 2026 formula-diet trial, 82% of people diagnosed within four years met the study's remission endpoint, against 50% of those diagnosed eight or more years ago, despite similar weight loss.

The Bottom Line

A diet of 75% carbohydrate cut liver fat by a third and HbA1c by a full point in 2026, which is a real result and not a reversal. The people in that trial also ate 346 fewer calories a day and lost 6 kg, their trial had no control group, and it was funded by an organisation that campaigns for the diet being tested.

Set beside DiRECT, where energy restriction produced remission that lasted five years in a minority of people, and a 2026 trial where the better short-term result came from a higher-fibre, lower-carbohydrate formula, the pattern that holds is about energy and liver fat rather than about carbohydrate itself. Higher-fibre eating did well in both comparisons, though neither can prove fibre was the reason. Eat more of it anyway, since almost nobody in the UK gets enough. Expect better management rather than reversal, and expect how long you have had the condition to count for a good deal.

This article is for general information and is not medical advice. Do not change or stop diabetes medication without speaking to your GP or diabetes team.

Sources

  • Kahleova H, Petersen K, Shulman GI, et al. Changes in ectopic lipids, insulin sensitivity, and beta cell function after a low-fat vegan diet in adults with type 2 diabetes: a 16-week single-arm clinical trial. BMJ Open Diabetes Research & Care, 2026;14(5):e006472. (Funded by the Physicians Committee for Responsible Medicine.) https://drc.bmj.com/content/14/5/e006472
  • Reynolds AN, Lang J, Brand A, Mann J. Higher fiber higher carbohydrate diets better than lower carbohydrate lower fiber diets for diabetes management: rapid review with meta-analyses. Obesity Reviews, 2025;26(1):e13837. https://pmc.ncbi.nlm.nih.gov/articles/PMC11611436
  • Schuppelius B, Lalama E, Zhang J, et al. Remission of type 2 diabetes depends on time since diagnosis and low-calorie diet composition: results of a randomized controlled trial in individuals with overweight and obesity. Diabetes Research and Clinical Practice, 2026. https://www.diabetesresearchclinicalpractice.com/article/S0168-8227(26)00078-1/fulltext
  • Lean MEJ, Leslie WS, Barnes AC, et al. Primary care-led weight management for remission of type 2 diabetes (DiRECT): an open-label, cluster-randomised trial. The Lancet, 2018;391(10120):541-551. https://doi.org/10.1016/S0140-6736(17)33102-1
  • Lean MEJ, Leslie WS, Barnes AC, et al. 5-year follow-up of the randomised Diabetes Remission Clinical Trial (DiRECT) of continued support for weight loss maintenance in the UK: an extension study. The Lancet Diabetes & Endocrinology, 2024;12(4):233-246. https://doi.org/10.1016/S2213-8587(23)00385-6
  • Diabetes UK. Type 2 diabetes remission: position statement for health professionals. https://www.diabetes.org.uk/for-professionals/improving-care/clinical-recommendations-for-professionals/statement-remission-type2