Allulose vs Stevia

Commonly cited as approximately 70% of sucrose. Convergent across Xie 2024, Han 2018 and the comments FDA received on the 2014 proposed rule — but this is a repeated figure, not a specification band or a sensory measurement, and no concentration dependence has been characterised the way it has for erythritol.

A MULTIPLE, not a fraction. FDA's figure for steviol glycosides as a class. Individual glycosides differ: stevioside is the least potent and rebaudioside A among the most, so a reformulation that swaps Reb A for Reb M changes potency without changing the word "stevia" on the label.

ENFORCEMENT DISCRETION, not a rule. FDA states that under 21 CFR 101.9(c)(1)(B) "a caloric value of 4 kcal/g must currently be used for allulose because it is a carbohydrate", and that it merely intends not to enforce that pending rulemaking. Two citizen petitions asked for 0.4 and 0.2; FDA chose the higher figure because it "would not underestimate the caloric contribution".

FDA: allulose "is virtually unmetabolized in the human body", produces "only a negligible increase in glycemic and insulinemic responses", and provides no more than 0.4 kcal/g.

Applies to the PURIFIED GLYCOSIDES ONLY, used at milligram levels. As with monk fruit, a retail "stevia" product is mostly bulking agent by weight — commonly erythritol, dextrose or maltodextrin — and dextrose and maltodextrin are NOT low-glycaemic. Compute the carbohydrate from the carrier, and note that the carrier differs between brands.

UPDATED 2026-08-31 with the field's current best synthesis, previously flagged in this backlog as bot-blocked — it is Open Access and loads cleanly in a real browser; only automated fetchers (WebFetch, server curl) were gated, the same JS bot-check pattern already on file for EUR-Lex and the ADA Standards of Care. Osborn, DuPuis, Liu, Della Corte & Della Corte 2026 (AJCN 123(6):101314) is a PROSPERO-registered (CRD42024567433) systematic review and meta-analysis of 20 controlled human intervention trials (12 allulose, 8 tagatose; 1,033 participants total), searched through 30 April 2025 — broader than either single study previously cited here. ALLULOSE ARM: significantly reduced postprandial glucose (incremental AUC SMD -0.66, 95% CI -0.92 to -0.39, moderate certainty) and postprandial insulin (SMD -1.27, 95% CI -2.14 to -0.40, moderate certainty) — but NO significant pooled effect on HbA1c, fasting glucose, fasting insulin, blood lipids, uric acid, or body composition (very low to moderate certainty across those outcomes). Read the postprandial finding and the null findings together: the acute glucose/insulin-blunting effect this record has documented since Tani 2023 holds up under a larger, more rigorous synthesis, but it does not extend to weight, fasting markers or long-term glycaemic control on the evidence gathered so far — allulose blunts a spike, it has not been shown to improve the numbers that matter for longer-run metabolic health. No third-party funding; the review's only disclosed conflict (SL) is unrelated consulting/editorial work with no rare-sugar-industry tie, a materially cleaner disclosure than tani2023's Kagawa University rare-sugar-group authorship (still flagged on that source below). Tani 2023 (145 participants/group, healthy adults, mechanism-establishing) and Ayesh 2024 (126 people with type 2 diabetes, first T2D-specific signal) remain cited in full below as the specific studies this broader 2026 synthesis draws on and updates, not as duplicated or superseded claims.

Zare et al. 2024, meta-analysis of 26 RCTs, 1,439 participants (PROSPERO CRD42023414411). Stevia supplementation significantly reduced fasting glucose, with the effect most consistent in trials shorter than 120 days and in participants with higher BMI, diabetes or hypertension. NO significant effect on insulin concentration or HbA1c — the fasting-glucose finding does not generalise to those two measures. Authors rate the overall certainty of evidence LOW and call for more RCTs at higher stevia doses. Read alongside health_claim_status below: a real, specific, but modest and short-duration signal — not evidence of durable metabolic benefit.

Expressed as STEVIOL EQUIVALENTS, not as grams of product — a conversion step every formulator has to do and many skip. Set by JECFA at its 69th meeting in 2008 (corrected 2026-08-31 from an earlier "2009" taken second-hand; the JECFA database itself says 2008, when the temporary designation was removed) on a NOEL of 970 mg/kg bw/day from a long-term stevioside study with a 100-fold safety factor, and independently by EFSA in 2010 from a 2-year rat carcinogenicity study. In November 2024 EFSA assessed an applicant request to raise it to 6 or 16 mg/kg bw/day and refused: "there is insufficient justification to increase the current ADI."

Han 2018, two non-randomised dose-escalation experiments (n=30 and n=19), Korean adults aged 21-30 with BMI 18.5-23. Severe diarrhoea appeared at 0.5 g/kg single dose. The authors explicitly warn that "considerably more symptoms may appear in individuals who are old, obese, or with certain metabolic and GI diseases" and flag the small sample and a possibly inadequate one-week washout. Not a regulatory ADI — no authority has set one.

Browns MORE than sucrose, not less. In sponge cake, lightness fell and redness/yellowness rose with allulose level, with the browning index rising sharply above 20% sucrose replacement and antioxidant activity nearly tripling at the highest level. Practical consequence: the window between caramelised and burnt is narrower than with sugar.

Used at sugar-like weights, but above 20% sucrose replacement in sponge cake hardness and chewiness rose significantly, and specific volume fell from 3.36 to 2.96 cm3/g at 50% replacement. A 1:1 swap also under-sweetens at ~70% relative sweetness.

High-intensity; contributes no volume, structure, browning or mouthfeel. Anything cup-for-cup is mostly carrier.

SOURCED 2026-08-31, previously flagged unsourced. Runs OPPOSITE to erythritol's mechanism, not the same direction weaker: erythritol recrystallises because it is relatively LOW-solubility for a bulk sweetener (see that record's own recrystallisation_risk note); allulose is markedly MORE soluble, which several independent manufacturer/patent sources describe as the reason it is less prone to gritty recrystallisation than erythritol at typical use concentrations in finished products such as caramel and frozen desserts. That is NOT the same claim as "allulose never crystallises" — a Roquette Frères patent's own storage-stability data (Example 2) found allulose SYRUP at 77% dry solids crystallising out during 4 weeks' storage at 4-15°C, which is exactly why a whole patent literature exists on deliberately crystallisable vs. non-crystallisable D-allulose syrup formulations. Read the two together: the CONCENTRATED-SYRUP ingredient form is a genuine crystallisation risk manufacturers actively manage; a FINISHED product at typical (much lower) allulose concentrations is not. DECLARED INTEREST: Roquette Frères is a sweetener/ingredient manufacturer with a commercial stake in allulose syrup formulations — a different company to the Kagawa/Matsutani interest disclosed on cooling_effect above, which is useful independent corroboration on the general solubility-driven mechanism even though both are industry sources.

Multiple "no questions" letters: GRN 400 (18 June 2012), Matsutani (12 June 2014), GRN 693 (Samyang, closed 28 August 2017), and later notices. FDA is explicit that such a letter "is not an affirmation that D-psicose is GRAS under 21 CFR 170.35".

HARD PURITY THRESHOLD. FDA has evaluated many GRAS notices for "high purity (95% minimum purity) steviol glycosides" including Reb A, stevioside and Reb D. But: "The use of stevia leaf and crude stevia extracts is not considered GRAS, and their import into the U.S. is not permitted for use as sweeteners" — Import Alert 45-06. A whole-leaf or crude-extract product is not a lower grade of the same legal thing; it is a different regulatory category and it is not permitted.

READ THE MECHANISM BEFORE REPEATING THIS. EFSA concluded on 6 May 2025 that "the safety of the NF D-allulose cannot be established" — but the reason is that the applicant stopped responding to EFSA after February 2020, leaving gaps on identity, production process, genotoxicity, stability, batch data and human data. This is an abandoned dossier, NOT a finding of harm. Allulose is a novel food under Regulation (EU) 2015/2283 and cannot lawfully be placed on the EU market without authorisation.

Authorised across 32 food categories. CORRECTED 2026-08-31: this note previously omitted E 960b, one of four group members confirmed directly against the EU Food Additives register. E 960a is steviol glycosides obtained by water extraction from stevia leaf; E 960b is steviol glycosides from fermentation; E 960c covers enzymatically produced variants including rebaudioside M, D and AM; E 960d is glucosylated steviol glycosides. Route of manufacture therefore changes the E number while the front label still says stevia.