Allulose: The FDA Allows It, EFSA Would Not — What the Human Trials Show
What Happened
Two regulators looked at the same sweetener and reached opposite answers, and the reason is procedural rather than scientific. In May 2025 the European Food Safety Authority's panel on nutrition, novel foods and food allergens adopted an opinion concluding that the safety of D-allulose could not be established. The opinion was published in June 2025 in the EFSA Journal (volume 23, issue 6, article e9468).
The stated reason is narrow and worth reading in full: EFSA identified gaps in the data on the ingredient's identity, production process, proposed uses and use levels, genotoxicity and human data. It asked the applicant for additional information several times. The applicant did not reply. EFSA then concluded that it could not establish safety on the record before it.
In the United States, allulose has moved the other way. The FDA accepted GRAS (generally recognized as safe) notices for allulose starting in 2012, with further notices in 2014 and 2020. In 2019 the agency issued labeling guidance allowing allulose to be excluded from the total sugars and added sugars lines on a Nutrition Facts panel while still counting inside total carbohydrate, at an energy value of 0.4 kcal per gram. China's National Health Commission approved D-allulose as a new food raw material on July 2, 2025, with a stated maximum daily intake of 20 g and a recommendation that infants, pregnant women and breastfeeding mothers not consume it.
Canada has not listed allulose among permitted sweeteners. The European Union has not authorized it. So the practical map in late 2026 is a patchwork, and the patchwork does not map onto how strong the evidence is in either direction.
Why Two Regulators Can Disagree Without Either Being Wrong
The two agencies were answering different questions, which is why the contradiction is smaller than it reads.
EFSA's question was whether a specific application dossier was complete enough to authorize a novel food — a category for ingredients not widely consumed in the EU before May 1997. That process asks the applicant to demonstrate safety, and when the applicant stops answering, the file stops moving. EFSA's conclusion says the safety of this ingredient could not be established on the available data. It does not say the ingredient is harmful, and it does not claim the applicant's product is unsafe. It says the file was incomplete and nobody filled it in.
The FDA's pathway is structurally different. A manufacturer assembles its own evidence and files a GRAS notice. If the agency has no questions, the substance is treated as generally recognized as safe. That is a review of a company's evidence package rather than an agency-run safety assessment, and it produces a different kind of answer to a differently framed question.
Anyone reading a headline that says one regulator "banned" allulose while another "approved" it should treat both verbs as imprecise. Nothing was banned, and the EU default of no data and no approval is not the same as a finding of harm.
The Acute Effect Has Consistent Evidence
Where the human data do line up is on what happens in the two hours after a meal, and the effect is real and repeatable.
The broadest recent synthesis is a 2026 systematic review and meta-analysis in the American Journal of Clinical Nutrition covering 20 controlled trials and 1,033 participants, 12 of them allulose trials. Pooled across those trials, allulose significantly reduced postprandial glucose (standardized mean difference −0.66) and postprandial insulin (SMD −1.27), both graded moderate certainty.
A 2024 meta-analysis in Clinical Nutrition ESPEN narrowed the question to people with type 2 diabetes: six randomized controlled trials, 126 participants. Postprandial glucose area under the curve fell significantly (SMD −0.6662, 95% CI −1.1360 to −0.1964, p = 0.0054), and time above range fell significantly (mean difference −8.8204, 95% CI −14.4101 to −3.2307, p = 0.0020).
An earlier 2023 systematic review in PLOS ONE looked at eight experiments in healthy adults and found that both 5 g and 10 g doses reduced the post-meal glucose curve, with small effect sizes (SMD −0.26 to −0.28).
Read together, these three point the same direction: taken with a meal, allulose blunts the glucose response to that meal. The effect size is modest, the trials are small, and the outcome is measured in hours rather than months — but the direction is consistent across independent groups.
Where the Evidence Stops
The same 2026 review that found the acute effect also looked at longer endpoints, and this is the part that tends not to appear in product copy.
| Endpoint | Pooled result | Source |
|---|---|---|
| Postprandial glucose | Significant reduction (SMD −0.66) | AJCN 2026 meta-analysis |
| Postprandial insulin | Significant reduction (SMD −1.27) | AJCN 2026 meta-analysis |
| HbA1c | No significant effect | AJCN 2026 meta-analysis |
| Fasting glucose | No significant effect (p = 0.4288) | AJCN 2026; Clinical Nutrition ESPEN 2024 |
| Fasting insulin | No significant effect | AJCN 2026 meta-analysis |
| Blood lipids | No significant effect | AJCN 2026 meta-analysis |
| Body composition | No significant effect pooled; one 12-week RCT reported a reduction | AJCN 2026; Nutrients 2018 |
Two entries in that table deserve more than a row each, because they show how uneven this literature is.
Body composition has a genuine conflict. A 2018 double-blind randomized trial in Nutrients assigned 121 adults with a BMI of 23 or above to 7 g of allulose twice daily or placebo for 12 weeks and reported a significant reduction in body fat percentage and fat mass. The 2026 pooled analysis found no significant effect on any measure of body composition. A pooled estimate across many small trials and a single larger trial can disagree like this, and when they do, the honest reading is that the question is not settled in either direction.
There is one chronic trial in people with diabetes, and it was null. A 2023 double-blind randomized crossover trial at Siriraj Hospital in Bangkok gave 16 patients with type 2 diabetes 7 g of allulose twice daily — 14 g a day — for 12 weeks, then switched them to aspartame after a washout. It found no significant effect on glucose homeostasis, incretin levels or body composition. Two secondary findings moved unfavorably: HDL cholesterol fell from 51 mg/dL to 41 mg/dL (p < 0.001) and the inflammatory marker MCP-1 rose from 259 to 297 pg/mL (p = 0.002). The trial was small, it was registered retrospectively, and the lipid finding has not been replicated. It is also the only chronic human trial that measured incretins, and it did not find what the marketing implies.
What This Means If You Use Allulose on a Low-Carb Pattern
Three practical points follow from the data above, and none of them require a position on whether the sweetener is good or bad.
The carbohydrate accounting is a labeling question, not a metabolic one. The FDA permits allulose to be listed outside total and added sugars at 0.4 kcal per gram, but it still appears inside total carbohydrate on the panel. For comparison, table sugar carries 99.98 g of net carbs per 100 g in the USDA record (SR Legacy, FDC #169655). Products count allulose differently, so the number on the panel is the one that matters, and our net carb methodology covers how to read it.
The dose that causes digestive trouble is documented. Dose-escalation work puts the maximum single tolerated dose near 0.4 g per kg of body weight, with severe diarrhea reported around 0.5 g per kg, and a maximum daily intake near 0.9 g per kg. Children appear more sensitive. For a 70 kg adult, the single-dose threshold sits around 28 g. That is well above a typical drink or dessert, and well below what someone might use in a day of home baking.
The evidence supports one claim and not several others. The claim that allulose does not spike blood glucose is supported by pooled trial data. Claims about fat loss, HbA1c improvement or appetite suppression are not supported by the same body of evidence, and one of them is contradicted by the only chronic trial run in people with diabetes.
How to Read a Sweetener Claim
The allulose file is a good training case, because the same set of trials supports a strong claim and undercuts a weaker one. Five questions separate them.
- What endpoint was measured? A two-hour glucose curve and a 12-week HbA1c are different questions with different answers. Both can be true at once.
- Who was studied? Healthy adults, people with type 2 diabetes and children do not respond identically, and a trial in one group does not transfer automatically.
- What dose was used? Many trials dose at 5 g or 10 g alongside a controlled meal. That is a tested quantity, not a daily habit.
- What kind of source is it? A regulatory opinion, a randomized trial, a pooled meta-analysis and a preprint carry different weight, and they should not be quoted with the same confidence.
- Is the outcome acute or chronic? Almost all of the positive allulose evidence is acute. Almost all of the null evidence is chronic. That pattern is the story.
We applied the same five questions to a different controversy recently, when we looked at a mouse study on sucralose and stevia that ran across two generations. The animal-versus-human distinction there is the same discipline as the acute-versus-chronic distinction here.
What This Does Not Prove
EFSA's conclusion is not a finding of harm, and it should not be reported as one. The opinion states that safety could not be established on an incomplete dossier. Those are different statements, and the gap between them is where most of the misleading coverage lives.
Nothing in the trial evidence above shows allulose causes metabolic damage. A pooled analysis that finds no significant effect on HbA1c is not evidence that HbA1c rises. A single small trial reporting an HDL drop is a signal worth following, not a settled finding, and it has not been reproduced.
The reverse also holds. None of this establishes that allulose is a weight-loss tool or a long-term glucose solution. The acute effect is documented. The chronic benefits are not, and at least one chronic trial in the population that would care most about them came back empty.
What the record does show is a sweetener that behaves as advertised in the short term and remains unproven in the long term — with two regulators looking at that same record and writing two different conclusions, because they were asked two different questions.
Is allulose approved in the United States?
Yes, through the GRAS pathway. The FDA accepted GRAS notices for allulose beginning in 2012, with additional notices in 2014 and 2020, and its 2019 labeling guidance allows the ingredient to be excluded from total and added sugars while remaining inside total carbohydrate at 0.4 kcal per gram. That is a different route from the EU's novel food authorization, which is why the two jurisdictions reached different outcomes on similar evidence.
Why did EFSA decline to approve allulose?
Because the application dossier was incomplete and the applicant did not supply the missing pieces. EFSA's panel adopted its opinion in May 2025 and published it in the EFSA Journal in June 2025 (23(6):e9468, DOI 10.2903/j.efsa.2025.9468), concluding that safety could not be established. The panel had requested additional data on the ingredient's identity, production process, proposed uses and use levels, genotoxicity and human data, and the requests went unanswered. This is a data-gap outcome, not a finding that allulose is harmful.
Does allulose raise blood sugar?
The pooled human evidence says it does not, and several trials found it lowers the glucose response to a meal eaten alongside it. A 2026 meta-analysis in the American Journal of Clinical Nutrition covering 20 controlled trials found significant reductions in postprandial glucose (SMD −0.66) and postprandial insulin (SMD −1.27) at moderate certainty. An earlier analysis restricted to people with type 2 diabetes found a significant reduction in postprandial glucose area under the curve. The same reviews found no significant effect on fasting glucose or HbA1c.
How does allulose count toward net carbs on a keto diet?
It depends on the label rather than the chemistry, because brands handle it differently. US guidance permits allulose to be excluded from the total sugars and added sugars lines at 0.4 kcal per gram while still appearing inside total carbohydrate. For scale, granulated table sugar measures 99.98 g of net carbs per 100 g in the USDA record (SR Legacy, FDC #169655). If you are tracking, use the panel in front of you rather than a rule of thumb. The same label-versus-chemistry gap shows up with sweeteners at the coffee shop, where the syrup decides the number.
This page summarizes published research and regulatory documents, with sources named throughout. It is not medical advice. Whether a sweetener belongs in your diet depends on your own health history, your medications and your goals, and that is a conversation to have with your own clinician rather than a decision to make from a product label.
Related reading
Filed under newsresearchsweeteners