Quick answer
A small high-fiber diet vs metformin trial, published on October 2, 2026, found better 84-day HbA1c and BMI results with a dietitian-guided dietary program. Participants were Chinese adults with newly diagnosed type 2 diabetes and BMI of at least 24. But the program combined two fiber formulas, dietary guidance, and substantially lower reported calorie intake. It did not isolate fiber, establish that short-chain fatty acids caused the benefits, or show that people should replace diabetes medication with a supplement. Original trial, Methods and Results.
The most interesting question is therefore: What did the whole dietary program achieve, and how much can we explain about why it worked?
Key points
- Researchers randomized 48 people; the reported clinical analyses included 46 completers, with only 12 in the metformin group.
- The adjusted HbA1c difference at day 84 was about 0.42 percentage points in favor of the dietary program. Several other glucose outcomes did not differ significantly at that final visit.
- Supplementary records describe unusually high fiber intake and a large calorie reduction. This was not an ordinary “add more vegetables” comparison.
- Stool SCFA measurements and the exploratory acetate subgroup do not establish a causal mechanism or a validated test for choosing treatment.
For the broader distinction between eating fiber and taking SCFAs, see our fiber, SCFA, and metabolic-health evidence review.
What did the high-fiber diet vs metformin trial compare?
The single-center GUT2DF trial enrolled adults aged 18–70 with newly diagnosed type 2 diabetes, HbA1c from 6.5% to below 9%, and BMI of at least 24 kg/m². That BMI threshold includes people who would not meet the usual U.S. obesity threshold. These were not participants without diabetes or people selected for advanced diabetes complications. Eligibility and trial design.
After a two-week run-in, researchers assigned 36 participants to the dietary program and 12 to metformin for 84 days. Two dietary participants withdrew, leaving 34 versus 12 in the reported clinical tables. Repeated stool collections produced 184 samples; that does not mean the study included 184 independent people. The open-label design meant investigators and outcome assessors knew treatment assignments. Trial flow and Methods.
The dietary arm received two standardized products. Formula 11 contained grain fibers and ingredients such as resistant dextrin, inulin, fructo-oligosaccharides, xylo-oligosaccharides, and konjac gum. Formula 3 provided a soluble-fiber blend. Dietitians individualized amounts and frequency and also advised participants about other foods. In contrast, the metformin arm did not receive this dietary program. Physicians titrated metformin according to clinical response and tolerance; we did not identify an achieved dose distribution in the materials checked. Interventions and Supplementary Table S2.
This matters when interpreting a headline about fiber “beating” a drug. The actual comparison was a supported dietary package versus metformin monotherapy, rather than the same meals and support with fiber as the only difference. It also did not compare diet plus metformin against metformin, or against a GLP-1 medicine.
What improved—and what did not?
The primary outcomes were HbA1c and BMI at day 84. Researchers used analysis of covariance, adjusting each outcome for its baseline value. The figures below are the paper’s baseline-adjusted group means, not raw before-and-after changes. Results section 3.1.
| Day-84 outcome | Dietary program | Metformin | Interpretation |
|---|---|---|---|
| HbA1c, adjusted mean | 6.32% (95% CI 6.11–6.53) | 6.74% (95% CI 6.39–7.09) | Between-group P = 0.042; about 0.42 percentage points lower, calculated from rounded means. |
| BMI, adjusted mean | 28.42 kg/m² (95% CI 28.07–28.78) | 30.22 kg/m² (95% CI 29.61–30.83) | Between-group P < 0.001; BMI units must not be described as kilograms lost. |
| Fasting glucose, meal-test glucose area, and HOMA-IR | No significant between-group difference at day 84 | Same comparison | Some earlier differences did not persist as statistically significant differences at the final visit. |
| CGM time in range | No significant between-group difference at day 84 | Same comparison | Several glucose-variability measures favored the diet; time in range did not. |
Sources: main Results and Supplementary Table S4. The group confidence intervals are not a confidence interval for the treatment difference; we have not reconstructed one.
A 0.42-percentage-point difference is not a 42% reduction. Nor is it the dietary arm’s total fall from baseline. Mixing those quantities makes treatment effects sound much larger than they are.
The result is encouraging for this particular program. Still, an 84-day HbA1c or BMI benefit does not establish long-term diabetes remission, prevention of complications, or sustained weight maintenance. Those outcomes need their own follow-up. The small comparator arm also leaves substantial uncertainty about how reproducible the difference would be.
The supplement changes how this study should be read
The dietary records are especially informative. Supplementary Table S3 reports the following group averages. These are reported intake values, not fixed prescriptions or independently measured energy expenditure. Original supplementary workbook, Table S3.
| Reported daily intake, rounded | Dietary group: baseline → day 84 | Metformin group: baseline → day 84 |
|---|---|---|
| Energy | 2,296 → 1,346 kcal | 1,894 → 1,640 kcal |
| Total fiber | 14.8 → 152.6 g | 13.1 → 12.7 g |
| Protein | 87.7 → 44.7 g | 62.5 → 55.5 g |
Why calorie intake matters
The paper reports lower energy and macronutrient intake in the dietary arm throughout intervention follow-up. Baseline energy averages also differed numerically, although the authors did not report that baseline difference as statistically significant. As a result, the tables do not justify assigning an exact percentage of the clinical benefit to calorie reduction, fiber, food substitution, or support. Dietary records and Results.
Importantly, lower calorie intake does not automatically invalidate a randomized dietary trial. It may be part of how the assigned program works. It does prevent us from treating the result as proof of a fiber-only effect independent of energy intake. An energy-matched comparison would answer that narrower question more directly.
An unusually high fiber intake
The reported 152.6 g/day of fiber is also unusually high. We checked this value in both the Word and Excel supplements; it is not a number inferred from the abstract. For perspective, the FDA’s dietary-fiber Daily Value is 28 g on a 2,000-calorie labeling basis. That labeling reference is not an individualized prescription, and the trial’s intake should not become one either. Supplementary Word tables; FDA label guidance.
Before anyone tried to reproduce this program, clearer information about achieved formula doses, dietary-record coding, nutritional adequacy, and tolerability would be valuable. We have not established that the reported intake is an error. We also cannot turn short-term liver and kidney laboratory findings into proof that this fiber amount is appropriate for routine use.
Did SCFAs explain the benefit?
Not conclusively. Dietary fiber can be fermented by gut microbes into short-chain fatty acids, including acetate, propionate, and butyrate. But demonstrating a clinical benefit from a fiber-containing program is a different task from proving that a particular SCFA mediated it.
In this trial, fecal SCFA concentrations did not differ significantly between treatment groups at matched follow-up time points. The authors nevertheless found differences in microbial functional potential and individual responses. However, gene abundance reflects potential capacity; it is not a direct measurement of metabolite production or absorption. The trial did not measure GLP-1. SCFA findings and Limitations.
A stool concentration is the material remaining after production, microbial use, absorption, and transit. It is not a direct readout of the total amount produced or delivered to tissues. Human tracer research supports treating production and systemic availability as separate measurements. It does not turn an unchanged stool result into proof of either no SCFA activity or a hidden benefit. Boets et al., human SCFA tracer study.
The acetate subgroup needs an extra qualification
The researchers classified 25 dietary participants as acetate-active because their stool acetate increased at least once after baseline; nine had no recorded increase. This classification used information collected after treatment began. In other words, researchers did not randomly assign an acetate response. Subgroup definition and Supplementary Tables S9–S10.
Table S10 makes the HbA1c claim more nuanced than a simple “responders did better” statement:
| HbA1c comparison: acetate-active versus inactive | Day 28 P value | Day 56 P value |
|---|---|---|
| Baseline-adjusted HbA1c comparison, ANCOVA | 0.055 | 0.100 |
| Percentage-change comparison | 0.021 | 0.040 |
Source: Supplementary Table S10, with P values rounded. These are different analyses, not interchangeable confirmations. The percentage-change comparisons crossed the conventional 0.05 threshold; the baseline-adjusted HbA1c comparisons did not.
The subgroup signal is worth following up. It does not show that raising acetate caused better HbA1c, that a stool test can select patients in advance, or that taking acetate would reproduce the effect. Candidate microbial response markers require prospective validation in larger, independent groups.
How does this fit the broader evidence?
The new finding does not appear in a vacuum. A 2020 systematic review included 42 controlled trials with 1,789 participants overall. Thirty-three trials contributed to the HbA1c analysis, which favored higher fiber by 2.00 mmol/mol (95% CI 0.71–3.30 lower), approximately 0.18 percentage points. Variation across studies limits any universal effect estimate. This supports fiber’s possible clinical value, not proof that every benefit is SCFA-mediated. Reynolds et al., original review.
The current trial also should not establish a blanket claim that metformin has no microbiome effect. A different double-blind randomized human study found substantial microbiome changes after four months of metformin. Its design and participants differed. The abstract supports that contrast; we did not fully re-extract that earlier trial for this article. Wu et al., 2017.
And none of these comparisons shows that sodium butyrate capsules, probiotics, or fermentation-derived SCFA mixtures outperform metformin. They are different interventions. Our direct butyrate and blood-sugar review addresses that separate evidence base.
What should a U.S. reader do with this information?
- Use the study to discuss a supported nutrition plan. Ask which foods, fiber types, and practical support fit your usual meals and health needs. NIDDK recommends working with a health care team on an individualized diabetes eating plan. NIDDK guidance.
- Do not copy the reported 153 g/day intake. The trial combined specialized formulas and supervision. Its dietary records are not a home recipe or a dose for a commercial supplement.
- Agree on how to assess benefit. HbA1c, weight, symptoms, nutritional adequacy, and the medication plan answer different questions. An attractive stool microbiome result is not a substitute for a clinical outcome.
- Coordinate major dietary changes with the clinician managing medication. The trial’s supervised medication arrangements do not establish that stopping a prescribed treatment is safe for someone else. NIDDK diabetes management guidance.
Frequently asked questions
Is a high-fiber diet better than metformin?
This 84-day trial favored a specific intensive dietary program for HbA1c and BMI. Its small sample, unequal groups, lower calorie intake, and missing achieved metformin-dose detail limit a broad superiority claim. It did not test every high-fiber diet against optimized usual diabetes care.
Does this prove that butyrate lowers blood sugar?
No. The intervention did not assign participants to take butyrate, and between-group fecal SCFA differences were not significant at matched time points. The clinical benefit and proposed microbial mechanism need separate judgments.
Could a stool test tell me whether the diet will work?
Not on this evidence. The acetate groups were defined by changes occurring during treatment. An exploratory classifier in a small cohort is not a prospectively validated clinical test.
Was the benefit sustained after the program ended?
This report does not establish that. The treatment comparison lasted 84 days; longer-term weight maintenance, medication needs, and clinical complications remain separate questions.
How we checked the evidence
On October 28, 2026, we checked the publisher’s full article and legal public XML, then downloaded the publisher’s Word and Excel supplements. Targeted checks covered the intervention description, participant flow, primary outcomes, dietary Tables S2–S3, clinical Table S4, and acetate subgroup Tables S9–S10. The unusually high fiber value was cross-checked between supplement formats. No patient-level data were reanalyzed.
We used the original Reynolds review, the Wu trial abstract, the Boets tracer source, and official FDA/NIDDK guidance for context. This is a targeted narrative evidence review, not a systematic search, independent duplicate extraction, or meta-analysis. Registry version history, a full protocol audit, achieved individual metformin doses, and all microbiome modeling details were not verified. No article-specific independent scientific review is claimed.
The trial authors declared no commercial or financial conflict, reported Central Plains Scholars funding, and disclosed one author’s journal editorial-board membership. These are author-reported disclosures, not an independent financial audit. Original disclosures. This article contains no affiliate purchase links or product recommendation.
Sources and further reading
- Li Y et al. High-fiber dietary program versus metformin, 2026. Full article; DOI 10.3389/fmicb.2026.1972983. PMID 42890753; trial registration ChiCTR-INR-17011752 as reported by the paper.
- Original Word supplement and original Excel supplement.
- Reynolds AN et al., 2020. Dietary fibre and whole grains in diabetes management. DOI 10.1371/journal.pmed.1003053.
- Wu H et al., 2017. Metformin and the human gut microbiome. DOI 10.1038/nm.4345. Abstract-level contextual check.
- Boets E et al., 2017. Systemic availability and metabolism of colon-derived SCFAs. DOI 10.1113/JP272613.
- FDA Nutrition Facts guidance and NIDDK healthy living with diabetes.
Written and edited by: Denny Jaeyong Choi (JD), INSEAD MBA
Written: October 28, 2026 · Last updated: October 28, 2026. Initial publication includes the supplementary intake and analysis-dependent acetate findings.
Medical information notice: This article provides general education and does not diagnose a condition or replace individualized medical care. Discuss medication and substantial dietary changes with your health care professional.