Sucralose: A Real Human Signal, In Vitro Alarms, and a Long-Term Benefit That Was Never Confirmed

Summary

Sucralose is approved, its acceptable daily intake was re-confirmed by European regulators in 2026, and no human outcome trial shows it causes disease — so a diet drink is still a genuine step off sugared drinks and this entry says so first. The human cancer and gut-barrier evidence is reassuring where it exists: in the one large cohort that measured the molecules themselves, sucralose showed no association with cancer while two other sweeteners in the same analysis did, and blood markers of gut leakiness measured in 572 adults showed no association with habitual sweetener intake. What has cha

Why Moderate

Moderate. The entry's actual recommendation — use it as a bridge off sugar, don't cook with it, prefer the drink to the sachet, aim at water — is supported at Moderate strength by three converging human randomised trials with real findings on surrogate measures, a regulatory refusal to clear high-heat use, and a global health body that could not confirm a long-term benefit. It is not a harm claim and is not graded as one. The strong version circulating publicly, that sucralose damages DNA and the gut lining in people, stays explicitly unsupported — and is now contradicted, not merely unsupported, by the human nulls on cancer and gut-leakiness markers.

Sub-claim tiers, which do not sit at the entry's level:
• The regulatory position (approved, intake limit re-confirmed 2026, no safety concern at authorised uses including carcinogenicity): Strong. A public risk assessor completed a full re-evaluation and published a conclusion. This is the highest-graded plank in the entry, and it runs against the entry's precautionary lean, which is exactly why it leads the controversy section rather than hiding in a footnote.
• The human glycaemic and insulin-sensitivity finding: Moderate, and now replicated. Three independent randomised trials, two to four weeks to a month, two of them with sucralose alone, effect sizes in the region of a 17 to 20 per cent fall in insulin sensitivity. Still surrogate endpoints, still first-exposure participants, still nothing longer than thirty days, and the trials disagree with each other about whether two weeks of sucralose alone is enough. The word "unreplicated" has been removed from this entry because it was wrong.
• The acute sweetener-alone null: Moderate. At least five independent human trials across three delivery routes, plus human gut-tissue receptor expression. Bounded strictly to acute, single-dose, sweetener-alone exposure.
• The transplant causal chain: Emerging, and rodent. The right experiment, correctly interpreted as sufficiency in germ-free mice with donors drawn from the extremes.
• The genotoxicity finding: Emerging, in vitro, and unreplicated. Millimolar concentrations, isolated cell layers, single laboratory, hazard identification rather than risk assessment, with published validity objections beyond concentration. Grading it low is not a statement that sucralose has been shown harmless — it is a statement that these experiments cannot bear the weight placed on them.
• The barrier findings: Emerging across three tiers, and the tiers must be named separately. In vitro at two to four hundred times a can in the 2023 screen; in vitro at roughly twenty-five times a can with a receptor mechanism and a knockdown control in independent 2020 work; rodent, in live animals with mucus and bacteria, at or above intake-limit-equivalent doses in chemically induced disease models. Human: one null on circulating leakiness markers in 572 adults, cross-sectional and surrogate — weak evidence of absence, not strong evidence of no effect.
• The immune finding: Emerging, and rodent. Mice, reversible on washout, at doses the authors state ordinary diets do not reach, in a paper proposing a therapeutic use. It changes the mechanism map rather than the risk assessment.
• Human cancer epidemiology: Moderate, human observational, and null. A within-study null for sucralose in a molecule-level cohort of 102,865 adults, and a pooled null across seventeen cohorts for the beverage category. Volunteer cohorts, self-reported intake, observational design — hence Moderate, not Strong.
• The rodent cancer bioassay: Emerging, rodent, and formally rejected. One lifespan mouse study reporting blood-cell tumours, assessed and rejected by the regulator on four stated grounds, with the corpus's reading of run-to-natural-death designs flagged as unsettled.
• The carbohydrate-pairing conditionality: Emerging, across three small trials, one formally disputed. Fifteen to forty-five participants each, a published re-analysis from authors carrying declared sweetener-industry support, a published rebuttal, and opposite gut-hormone directions by population. A live, internally consistent conditionality, not a receipt.
• Thermal degradation: Moderate for the regulator's non-confirmation, for the national 120 °C recommendation and for the identified chemistry; nothing at all for health outcomes, because no outcome study exists. The plank rests on absence of demonstrated safety, which is a legitimate precautionary position and an illegitimate harm claim.
• The broader human microbiome literature: Emerging and partly unverified. About two human trials at the time of the 2024 review, human and rodent findings pointing in opposite directions, no direct replication of the two-week protocol in either direction, and a prior pure-compound saccharin null at maximum intake whose own funding is unverified.
• The appetite route: Emerging, human, and a surrogate for behaviour. A 75-person publicly funded randomised crossover measuring brain blood flow and hunger ratings, with hunger not differing from water.
• The benefit question: Moderate, and it is the plank the entry's conclusion actually rests on. A conditional recommendation on low-certainty evidence is weak evidence for the recommendation, and the same review's short-term trials found weight and body mass index went down — but the failure to demonstrate any long-term body-fat benefit is enough to stop anyone assuming the durable benefit, which is all this entry needs it to do.

Practical takeaway

The bridge, stated first and without hedging. If someone is drinking sugared soft drinks daily, switching to the diet version is a genuine improvement and this platform recommends it without embarrassment. The sugar load being removed is a well-established metabolic problem; what replaces it carries open questions and no demonstrated harm. Do not let a set of open questions talk anyone back onto sugar. That trade is unambiguously worse.

And the bridge has a far side. The reason to treat it as a bridge rather than a destination is the benefit half: the durable benefit these products exist to deliver was reviewed and could not be confirmed beyond the short term. If the sweetener is not reliably buying long-term weight control, the open questions are being accepted for very little. Water is the finish line — plain, sparkling, or with something bitter or sour in it, since the practical difficulty of leaving diet drinks is usually the ritual and the mouthfeel rather than the sweetness.

A workable sequence for someone who wants off sugared drinks entirely. Sugared drink to diet drink is the first, largest and easiest step. Then dilute the ritual rather than removing it: sparkling water with citrus, cold tea, or a sugared drink kept as an occasional deliberate thing rather than a background default. Removing the ritual and the sugar in one move is the version people abandon.

The sachet is not the can, and this is the one place the evidence changes advice. Tabletop Splenda is roughly one per cent sucralose by weight; the rest is maltodextrin or dextrose, which is rapidly absorbed carbohydrate. So a sachet stirred into coffee delivers sucralose plus carbohydrate — the same pattern as the arm of the pairing trial where insulin sensitivity and the brain's sweet-taste response were impaired — while a diet drink delivers sucralose without a carrier. The evidence for that conditionality is thin: three small human trials, one of them formally disputed in print. It does not license a rule. But if the pairing finding holds, the sachet is where it would apply, and someone who wants to act on it can use the liquid sweetener drops or the plain drink rather than the packet, or simply take the coffee unsweetened. Nobody should be alarmed about sachets they have already used.

Do not bake or fry with it, and keep it below 120 °C. This is the one other genuinely specific, genuinely actionable line in this entry, and it comes from regulators rather than from a wellness claim: the European regulator has not cleared sucralose for prolonged high-temperature use and explicitly extended that concern to home baking and frying, and the German federal risk assessor recommends not heating sucralose-containing foods above 120 °C. Sweeten after cooking rather than before. This is a precaution against an unresolved question, not a warning that heated sucralose is known to be toxic, and no figure for any degradation product may be quoted alongside it.

Do not switch to fruit juice. Whole fruit and juice diverge sharply — in the same cohorts, whole fruit intake tracks lower type-2 diabetes risk while juice tracks higher. Juice is not the clean escape from a diet drink, and should never be offered as one.

Treat the "stevia is the clean one" assumption carefully, and update it in both directions. The human trial found stevia altered the microbiome and the plasma metabolome too — it simply did not move glycaemic response. "Sweetener X is inert" is not supported for any of them; what differs is which effects appeared. Two per-compound updates are worth carrying, and both are leads rather than receipts: in the receptor-dependent barrier system where sucralose, aspartame, saccharin and neotame were positive, a 2025 report has stevia returning a null — a genuine point in stevia's favour, printed here precisely because it cuts against this entry's precautionary lean — while monk fruit's sweet principle, mogroside V, was positive on that same barrier endpoint, which undercuts monk fruit's "least studied therefore cleanest" reputation. Both halves were located through a search summary rather than the primary paper: do not quote concentrations, tight-junction directions or assay details, and re-verify before either is treated as established. The durable rule is the one that survives regardless: less studied is not cleaner.

The pairing point, held at the weight the evidence allows. Three small human trials, one of them formally contested, suggest sucralose consumed with carbohydrate behaves differently from sucralose alone. That is more than the "single trial" this entry used to describe and still not strong enough to build a rule on. What it does do, if anything, is favour the bridge as usually practised — a diet drink on its own — while identifying the sachet as the everyday form where the pairing is built into the product.

If someone has inflammatory bowel disease or another established gut condition: the honest answer is a watch item rather than a warning, and both directions belong in it. In the one animal model of Crohn's-like disease, the microbiome shifted in healthy and disease-prone animals alike while the inflammatory marker rose only in the disease-prone host — and that study used the 1:99 sucralose-to-maltodextrin mixture, so it cannot attribute the effect to sucralose. In humans, the cohort signal for these diseases attaches to sugar, not to sweeteners. There is no basis here for telling anyone with inflammatory bowel disease to cut sweeteners, and there is a reasonable basis for someone with an established gut condition preferring to keep the count low while the question is open. Decisions about diet in active disease belong with their clinician.

If someone is already drinking several diet drinks a day and is anxious about this entry: the honest answer is that nothing here shows they have harmed themselves — the human cancer data specific to sucralose is null, the one human measurement of gut leakiness is null, European regulators re-confirmed the intake limit in 2026, and the trials that found something measured a short-term marker in first-time users over two to four weeks. The useful direction is reducing the count over months rather than treating it as urgent. Anxiety is not the intended output.

Evidence detail

Why This Entry Exists

Sucralose is one of the most widely used artificial sweeteners in the world and among the most argued-about, and the two loudest positions are both wrong in the same way: they treat a mixed, partially-resolved evidence file as though it were settled. One camp says approved-and-safe, decades of use, case closed. The other says it damages your gut and breaks your DNA. The first ignores three human trials; the second silently upgrades cell-culture experiments into human harm claims — which is the defining error on this topic and the one this entry exists to refuse.

The honest picture is narrower and less satisfying than either. There are now three independent human randomised trials, over two weeks to a month, all finding sucralose worsened blood-sugar handling at doses below the legal intake limit — on short-term surrogate measures, in participants meeting the molecule for the first time, with none running longer than thirty days. There is a striking cell-culture study reporting DNA damage and gut-barrier breakdown, run on isolated cells at concentrations far above realistic exposure — and, separately, an independent laboratory that reported a barrier effect three years earlier at roughly a tenth of those concentrations with a receptor mechanism and a genetic control, which is the sort of study this entry previously said would move it. There is live-animal work measuring the mucus and tight-junction proteins the argument is actually about. There is a mouse study describing a direct effect on immune cells, which does not fit the microbiome story this entry used to tell about how sucralose could matter at all. There is an unresolved question about what sucralose becomes when it is heated, which the approving regulator raised itself. And there is a large body of long-term human outcome evidence that still shows no established harm — a pooled null across seventeen prospective cohorts for artificially sweetened drinks and overall cancer, a within-study null for sucralose specifically in the one cohort that measured the molecule, a null on gut-permeability biomarkers in 572 adults — and, per the World Health Organization, no established long-term weight benefit either, though its short-term randomised evidence did show reduced weight and body mass index.

That last point is what makes the precautionary case legitimate rather than alarmist, and it is also where the case is weakest, so both halves are printed. The argument is not "this is proven dangerous." It is: the long-term reason these molecules exist is in doubt, while the short-term signal has now replicated and the questions keep multiplying. You are being asked to accept a set of open questions in exchange for a durable benefit that a global health body reviewed and could not confirm — while the short-term benefit it did find was itself graded low-certainty. That is a weak trade — and it is a completely different claim from "sucralose causes cancer," which nothing here supports and which the human cancer data actively argues against.

What bad advice this protects against, in all directions:
• "Artificial sweeteners are poison, go back to regular Coke, at least sugar is natural" → Wrong, and actively harmful. No human outcome trial shows sucralose causes disease; the largest pooled human cancer analysis on artificially sweetened drinks is null; European regulators re-confirmed its acceptable daily intake in 2026 after reviewing the disputed evidence. For someone drinking a litre of sugared cola a day, switching to the diet version is a real reduction in a real, well-established metabolic load. Scaring that person back onto sugar is a worse outcome achieved with a more exciting message.
• "A study found sucralose damages DNA and breaks the gut lining" → Still overstated as usually told, but the correction has to be more precise than it used to be. The DNA-damage finding is one cell-culture screen, unreplicated, and criticised in print on grounds sharper than concentration alone. The barrier finding is not unreplicated: an independent laboratory reported it in 2020 at roughly a tenth the concentration, receptor-dependent, with a knockdown control. Animal work with a real gut has reported reduced mucus and altered tight-junction proteins, mostly at or above intake-limit-equivalent doses and in chemically induced disease models. And when someone measured markers of gut leakiness in 572 people, there was no association with sweetener intake. So the honest sentence is: a real signal in dishes, a real signal in sick animals, and nothing yet in people. "Sucralose damages your DNA" and "sucralose breaks your gut lining" remain sentences this entry will not write.
• "It's approved, the ADI has huge safety margins, therefore it's fine and this is all noise" → The strongest pro-sweetener line, and it is genuinely true as far as it goes. But the acceptable daily intake is a hundredfold margin against reduced body-weight gain in rats — that is the endpoint the regulator's own reference point rests on — so it was never designed to detect a microbiome-mediated effect on blood-sugar handling. Approval at current uses and open questions at the edges are both true at once, and in the same 2026 opinion that re-confirmed the intake limit, the regulator could not conclude on the safety of sucralose in baked goods.
• "It's 99% water, a diet drink is basically water with flavour" → A category error worth retiring. Composition by mass says nothing about what the remaining fraction does. Sucralose is reported as roughly 385 to 650 times sweeter than sugar by weight, which is the entire reason it is used: a vanishing mass fraction carries the whole signal. The same arithmetic runs the other way and this entry now says so explicitly: tabletop Splenda is roughly one per cent sucralose by weight, the rest being maltodextrin or dextrose, so a sachet is mostly not the sweetener at all. "Less than one percent" is a statement about mass, never about consequence — in either direction. And "it contains a chemical" is likewise a statement about presence, not about dose.
• "Just drink fruit juice instead, it's natural sugar" → Whole fruit and fruit juice do not behave the same way. In the same cohorts, whole fruit intake tracks lower type-2 diabetes risk while fruit juice tracks higher. Juice is not the health drink in this comparison and must never be recommended as the clean escape from a diet drink.
• "Sweeteners help you lose weight, that's the point of them" → This is the claim the World Health Organization reviewed in 2023 and could not confirm over the long term: no long-term benefit on body fat in adults or children. Its short-term randomised evidence did show lower weight and body mass index, at low certainty and mostly in trials of three months or less. That review carries its own asterisks and they matter in both directions, but the durable benefit half of the trade can no longer be assumed.
• "Sucralose suppresses your immune system, there's a Nature paper" → There is a Nature paper, it is in mice, the effects reversed when the sucralose stopped, and the authors' own position is that ordinary food and drink would not reach the doses used — the paper is pitched as a possible treatment route for autoimmune disease, not as a consumer warning. Reporting it without the mice and without the authors' own dose statement turns a therapeutic hypothesis into a scare.

This entry OWNS: sucralose specifically — the three human randomised trials on glycaemic response and insulin sensitivity and the microbiome findings that travel with them, the carbohydrate-pairing conditionality across three small trials, the cell-culture genotoxicity and gut-barrier findings with their concentrations and their independent precedent, the live-animal barrier and mucus work, the human permeability and cancer nulls specific to sucralose, the rodent lifespan cancer bioassay and the regulator's rejection of it, the mouse immune finding, and thermal degradation together with the regulator's refusal to clear high-heat use and the 120 °C recommendation that sits underneath it.

This entry DEFERS: the class-level argument about non-sugar sweeteners as a category, including the full World Health Organization guideline and its criticism, to artificial_sweeteners_evidence; gut-barrier and intestinal-permeability biology to leaky_gut_intestinal_permeability; microbiome fundamentals, diversity measures and what they do and do not mean to diet_gut_microbiome; the aspartame-specific cancer-classification argument to aspartame_and_the_iarc_classification; sugar alcohols to erythritol_and_sugar_alcohols; and the general problem of reading observational sweetener studies to rct_vs_observational_evidence.

Evidence

Read the tiers, not the thesis. Every claim below is labelled by what kind of study produced it — human randomised trial, human observational, rodent, cell culture, or chemical analysis with no biological system at all — because on this topic the study type is the argument. Almost every dishonest sentence written about sucralose is produced by moving a finding up one rung without saying so. This version of the entry adds two rungs it was previously missing entirely: live animals for the gut lining, and humans for cancer and for permeability.

The human trial evidence, which is now three trials rather than one

1. A randomised controlled trial in 120 healthy adults tested four sweeteners at doses below the legal intake limit (Moderate Evidence — HUMAN RCT). Suez and colleagues randomised 120 healthy adults to six arms: saccharin, sucralose, aspartame, stevia, a glucose-vehicle control, and a no-supplement control, for two weeks, using sachets at doses below the acceptable daily intake. Per-arm numbers are approximately twenty, which follows directly from 120 participants across six arms and is stated in the same terms by independent commentators (Francisco Guarner, Vall d'Hebron University Hospital, and Ascensión Marcos, CSIC). Participants were healthy and had been avoiding these sweeteners at baseline, so this measures a first-exposure response, not the state of a habitual consumer. (Cui bono: industry bodies benefit from emphasising the small groups and short duration to file this as preliminary — a fair criticism, deployed selectively. The Weizmann group built its reputation on microbiome and personalised-nutrition work, and a positive finding advances that programme; wellness marketers benefit enormously from a headline saying a human trial found harm. Both incentives are real. It is still the best-designed study of its kind on this question.) Suez J, et al. Cell. 2022;185(18):3307-3328.e19.

2. In that trial, saccharin and sucralose significantly impaired glycaemic responses; aspartame and stevia did not (Moderate Evidence — HUMAN RCT). All four sweeteners altered the stool and oral microbiome and the plasma metabolome. Only saccharin and sucralose moved blood-sugar handling. Three limits travel with this and must never be dropped: glycaemic response is a surrogate measure — no diabetes, no cardiovascular event, no death was measured, and none could be in two weeks; the deviations were small, in the words of independent commentator Duane Mellor (Aston University); and the responses were highly person-specific, so the group result conceals participants who showed no effect at all. Statistician Kevin McConway (Open University) noted the human arm cannot by itself show that the microbiome changes caused the glucose changes. What has changed is that this is no longer a solitary finding — two independent trials in the two claims below point the same way with sucralose given alone, so the "one study" objection is no longer available to either side. (Cui bono: "only a surrogate, and the effect was small" is both a legitimate scientific point and the exact sentence industry-funded reviews use to neutralise every unfavourable finding. In the other direction, "sucralose impairs blood sugar control" as a bare headline sells supplements and natural-sweetener alternatives. The honest statement is narrower than either.) Suez J, et al. Cell. 2022;185(18):3307-3328.e19; Science Media Centre expert reaction, 19 August 2022.

3. An independent randomised trial four years earlier, in 66 healthy adults given sucralose alone, found insulin sensitivity fell by a median 17.7 per cent over fourteen days (Moderate Evidence — HUMAN RCT). Romo-Romo and colleagues randomised 66 healthy adults, 33 per group, to sucralose at fifteen per cent of the acceptable daily intake or control for fourteen days, with no carbohydrate carrier and no co-ingested sugar — sucralose on its own. Insulin sensitivity fell by a median 17.7 per cent, interquartile range −29.3 to −1.0, against −2.8 per cent in controls (p=0.04), and the acute insulin response rose from 577 to 671 in adherent participants (p=0.04). This trial was published in a leading nutrition journal in 2018, four years before the trial this entry used to call the only good one, and a previous version of this entry did not carry it at all — which is worth stating plainly, because "unreplicated" was the load-bearing word in the old framing and it was wrong. Limits, and they are the same ones that apply to every trial in this group: insulin sensitivity is a surrogate, fourteen days is short, participants were healthy and unaccustomed to sucralose, and the interquartile range crosses close to zero at one end, meaning some participants barely moved. Funding and conflict declarations for this trial were not extracted during the hardening pass and are unverified here — check them before citing this study in a context where funding matters. (Cui bono: this trial's existence removes the single most useful industry talking point on the glycaemic question, which was that one Israeli trial found something once. It equally removes a wellness talking point, since three trials converging on a 17 to 20 per cent shift in a surrogate marker is a much more boring sentence than "sucralose destroys your metabolism.") Romo-Romo A, Aguilar-Salinas CA, Brito-Córdova GX, Gómez-Díaz RA, Almeda-Valdes P. American Journal of Clinical Nutrition. 2018.

4. The same group's thirty-day trial reports a 20.3 per cent fall in insulin sensitivity, alongside reduced microbial diversity, reduced faecal butyrate and raised inflammatory markers, on public rather than industry funding (Moderate Evidence — HUMAN RCT). This study was carried in the previous version of this entry as an unverifiable lead whose numbers were forbidden; it verified, and it is promoted here to a receipt. Thirty days of sucralose at thirty per cent of the intake limit, randomised, placebo-controlled and triple-blind: glucose, insulin and gut-hormone areas under the curve all rose after a mixed meal, insulin sensitivity fell 20.3 per cent, microbial diversity fell, faecal butyrate — a short-chain fatty acid the gut uses as fuel and which the microbiome literature treats as favourable — fell, and inflammatory markers and branched-chain amino acids rose. Funded by Mexico's national science council under grant 316514, not by industry, which matters in a literature this contested. The participant count was not stated in the record retrieved during verification and must not be invented — no sample size may be printed for this trial until the primary text is read. The gut-hormone rise is worth flagging for a specific reason: the reassuring acute nulls further down this list all tested single doses, and a thirty-day trial finding the gut-hormone response moving is exactly the boundary those nulls do not cover. (Cui bono: public funding from a national science council with no product in the market is about as clean as this literature gets, and it cuts against industry. The counterweight is duration and endpoint, not money: thirty days, surrogate markers, first-exposure participants, and a dose at nearly a third of the legal limit rather than at ordinary consumption.) Romo-Romo A, Sánchez-Tapia M, Almeda-Valdes P, Tovar AR, Torres N, et al. Clinical Nutrition ESPEN. 2025;69:733-744.

Taken together, the honest summary of claims 1 to 4 is this: three independent human randomised trials, over two weeks to a month, converge on sucralose impairing glucose handling or insulin sensitivity at doses below the legal intake limit, two of them with sucralose given entirely alone. All three measured surrogate markers, none ran longer than thirty days, all studied people meeting sucralose more or less for the first time, and none measured a disease. That is a real, replicated signal and it is still not a harm.

5. Transplanting participants' gut bacteria into germ-free mice reproduced the donors' glucose responses, which is the causal test — and it is a rodent result (Emerging Evidence — rodent). Mice raised without any gut bacteria were colonised with microbiomes from the top and bottom responders in each sweetener group, and showed glycaemic responses "largely reflecting those noted in respective human donors," with sucralose given as the worked example. This is the strongest causal design available short of a mechanistic human experiment, and it is why this trial outranks the observational sweetener literature. What it establishes is that the microbiome change is sufficient to transfer the effect into a germ-free mouse. It does not establish that this produces disease in people. Two further limits: germ-free mice have abnormal immune and metabolic development, and donors were selected from the extremes of the response range, which maximises the chance of finding a transferable signal. (Cui bono: dismissing this as "just mouse data" would be unfair — it is the deliberate causal test and it worked. Quoting it as "proven to cause blood sugar problems" performs the exact upgrade this entry exists to refuse. Name the mice in the same sentence as the finding.) Suez J, et al. Cell. 2022;185(18):3307-3328.e19.

6. A thin but internally consistent conditionality runs across three independent small human trials: sucralose taken with or just before carbohydrate behaves differently from sucralose taken alone — and the best-known of the three is formally contested in print by authors with declared sweetener-industry support (Emerging Evidence — HUMAN RCTs, one disputed). The previous version of this entry called this a single contested trial. There are three, and naming only one understated it. First, Dalenberg and colleagues randomised 45 healthy participants to three groups — drinks sweetened with sucralose alone, with sucrose, or with sucralose plus maltodextrin — across seven beverages over roughly two weeks. Reduced insulin sensitivity and a blunted brain response to sweet taste appeared only in the sucralose-plus-maltodextrin group; sucralose alone produced no change, and neither did sucrose. That is fifteen participants per group as randomised, with fewer analysed on some measures — the widely repeated "thirteen per arm" figure could not be verified against the primary text and is deliberately not used here. It is disputed: Khan and Sievenpiper published a re-analysis in the same journal arguing the maltodextrin, itself a rapidly absorbed carbohydrate, plausibly did the damage on its own, and finding no difference in insulin indices between maltodextrin alone and the sucralose-maltodextrin combination; the original authors published a rebuttal. Second, Pepino and colleagues gave 17 obese, insulin-sensitive adults sucralose or water ten minutes before a five-hour modified glucose tolerance test in a randomised crossover, and reported roughly a twenty per cent increase in total plasma insulin with reduced insulin sensitivity, concluding sucralose is not inert in that setting; that trial drew its own published Comment and author Response, so it is contested in the same way. Third, Temizkan and colleagues reported sucralose enhancing gut-hormone release and lowering blood glucose in the presence of carbohydrate in healthy subjects but not in people with type 2 diabetes — a positive signal, conditional on both carbohydrate and metabolic status, verified only at abstract level. The honest reading, and it is genuinely awkward: acute single doses of sucralose alone reliably do nothing measurable; sucralose with or immediately before carbohydrate produced signals in three independent small trials, in opposite directions on the gut hormone depending on the population studied; and yet two-week-to-monthlong dosing with sucralose alone did move insulin sensitivity in the two trials above, while the pairing trial's own two-week sucralose-alone arm did not. Those last two facts contradict each other at similar durations, on fifteen participants against thirty-three. That inconsistency is printed rather than resolved, because resolving it would require a study nobody has run. (Cui bono, and this is verified rather than left open: the published re-analysis is a legitimate methodological critique and its authors' declared disclosures record support from the Calorie Control Council, a sweetener trade body, with Sievenpiper additionally reporting support from a fund established by Tate & Lyle, a major sucralose supplier. Both facts are true simultaneously — a correct methodological point from an interested party remains a correct point, and the interest still belongs on the page. On the other side, "sweeteners are fine alone but poison with food" is an unusually shareable framing that badly outruns three trials of fifteen to forty-five people.) Dalenberg JR, et al. Cell Metabolism. 2020;31(3):493-502; re-analysis Khan TA, Sievenpiper JL, Cell Metabolism 2020; author reply Cell Metabolism 2021; Pepino MY, Tiemann CD, Patterson BW, Wice BM, Klein S, Diabetes Care 2013;36(9):2530-2535 with Comment and Response, Diabetes Care 2014;37(6):e148/e149; Temizkan S, et al. European Journal of Clinical Nutrition. 2015.

7. Sucralose taken on its own, acutely, produces no insulin or gut-hormone response in humans — and this null rests on at least five independent trials across three delivery routes, not on one small study (Moderate Evidence — HUMAN, multiple trials; individual effect sizes confirmed at abstract level only). This is the reassurance plank the previous version of this entry under-used, and it deserves stating at its real strength. The receptors are not an inference from dishes and rodents: sweet-taste receptor components have been measured in human gastrointestinal tissue with co-localisation to gut-hormone-containing cells confirmed by immunohistochemistry, with a functional arm using a sweet-receptor blocker in 16 healthy subjects. And the consequence half is a consistent set: intragastric sucralose up to 800 mg produced no insulin or gut-hormone release and no change in gastric emptying; sucralose did not alter small-intestinal glucose absorption; oral sucralose produced no gut-hormone response in healthy normal-weight subjects; intraduodenal sucralose with or without glucose had no effect on blood glucose or gut hormone; artificial sweeteners had no effect on gastric emptying, gut hormone or glycaemia after an oral glucose load; and a dedicated human study concluded gut sweet-taste-receptor activation is of limited importance even for glucose-stimulated gut-hormone secretion. A diet drink does not spike your blood sugar, and this is the evidence for that sentence. The boundary is exact and must travel with it: these are acute, single-dose, sweetener-alone experiments. They say nothing about repeated exposure over weeks, which is where claims 3 and 4 found something, and nothing about co-ingestion with carbohydrate, which is where claim 6 found something. All of these studies were located at abstract or title level during verification — the direction is consistent across independent retrievals, but individual effect sizes must not be printed from this entry. (Cui bono: this null is the single most useful true thing the industry can say, and it is true. It is also the most commonly overstated into "sweeteners have no metabolic effect whatsoever," which the three repeated-exposure trials contradict.) Steinert RE, et al. Clinical Nutrition. 2011, with Gerspach AC, et al. American Journal of Physiology: Endocrinology and Metabolism. 2011; Ma J, et al. American Journal of Physiology: Gastrointestinal and Liver Physiology. 2009; Ma J, et al. British Journal of Nutrition; Ford HE, et al. European Journal of Clinical Nutrition. 2011; Wu T, et al. 2012; Brown AW, et al. Diabetes Care. 2013;36(12):e202; plus a 2017 human study on gut sweet-taste-receptor importance.

8. Beyond those trials the human microbiome evidence is thin, human and rodent findings point in opposite directions, and the replication question has answers in both directions that this entry previously left blank (Emerging Evidence — systematic review and meta-analysis of human and rodent studies, plus individual trials at the levels stated). A 2024 systematic review and meta-analysis located nine studies in total, of which only two were human trials, and reported higher relative abundance of Bacteroidetes in humans and the converse in rodents. That divergence is the important part: it undercuts cross-species extrapolation in both directions, which also weakens rodent-based reassurance, not just rodent-based alarm. On replication, three things are true at once and the previous version of this entry carried none of them. No study directly replicating the 2022 two-week protocol was located in either direction, so "not independently replicated" remains literally accurate for that specific trial. But the closest available comparator on the saccharin arm is null: a double-blind, placebo-controlled trial gave pure saccharin in capsules twice daily for two weeks at the maximum acceptable intake and found no microbiota, metabolite or glucose-tolerance change in humans or mice. Its design differs in a way that is itself a testable explanation rather than a wave-away — capsules bypassing the mouth and delivering pure compound, against oral sachets delivering sweetener in a glucose vehicle — and it predates rather than replicates the 2022 trial. On the concordant side, the independent thirty-day sucralose trial above reproduces the glycaemic and microbiome direction. And the largest and longest human microbiome dataset in this field, a one-year randomised trial of sweetener use inside a sugar-reduced diet, reported a shift toward short-chain-fatty-acid-producing taxa, which its investigators describe as a favourable direction rather than a harmful one — carried here for that microbiome direction only; its funding statement includes industry partners and was not verified, and its weight-outcome results belong to the class-level entry rather than to this one. Several individual sucralose studies also exist and point in a consistent direction — two twelve-week trials in Asian Indian adults finding diversity changes in people with type 2 diabetes but not in people with overweight or obesity, and a ten-week study reporting dysbiosis in healthy young adults. These were confirmed only to the level of title and journal and are carried as leads, not receipts — do not quote their numbers. (Cui bono: "multiple studies show sucralose wrecks your microbiome" overstates a literature this thin. "The evidence is thin" is true, and is also the permanent industry position on any unfavourable early signal. And note that the pure-compound saccharin null is promoted by an industry association and its own funding was not verified, so it must not be printed bare either.) Food Reviews International. 2024;40(5); Serrano J, Smith KR, Crouch AL, et al. Microbiome. 2021;9:11.

The gut-lining evidence, which has three tiers and used to be printed with only one

9. A 2023 study reported DNA damage from a sucralose impurity and gut-barrier breakdown from both the impurity and sucralose itself, in cell culture, at concentrations far above realistic exposure — and its reception is more substantive than trade-body press releases, including litigation (Emerging Evidence — IN VITRO). Schiffman and colleagues screened sucralose-6-acetate, a trace impurity, and sucralose itself in human lymphoblastoid cells and in a laboratory model of human intestinal lining. They reported the impurity was genotoxic and clastogenic, meaning it produced DNA strand breaks; that it raised expression of genes tied to inflammation, oxidative stress and cancer, most strongly metallothionein 1G; that electrical-resistance and permeability measurements indicated both the impurity and sucralose impaired barrier integrity; and that it inhibited two drug-metabolising enzymes. The concentrations are the confound the press coverage lost. As reported, barrier loss for the impurity began at 5 millimolar with total loss at 10 millimolar; sucralose itself was tested at 80 and 160 millimolar; genotoxicity appeared at 353 micrograms per millilitre with metabolic activation and 707 without. Stated as a comparison rather than as "orders of magnitude," which was too vague to be checked: a 355 millilitre can of diet drink containing roughly 60 milligrams of sucralose works out at about 0.4 millimolar, so the sucralose arms of this paper sit at roughly two to four hundred times the concentration of a can. That figure is an approximate calculation from stated serving sizes and molecular weight, not a measurement, and it is labelled as such. These are exposures on isolated cell layers, with no liver metabolism, no mucus layer, no microbiome and no clearance. That does not make the finding fake. It makes it hazard identification, not risk assessment. On replication, the previous version of this entry was wrong in one direction and right in the other, and the split matters: the genotoxicity finding has not been independently replicated; the barrier finding has independent precedent from a separate laboratory three years earlier at roughly a tenth of these concentrations (next claim). On reception, four things belong on the page and only one was here before. Two technical objections exist that are sharper than the concentration argument because they attack validity rather than relevance: high cytotoxicity may have confounded both genotoxicity assays, and the doses at which the impurity's genotoxicity appeared exceeded the assays' own recommended limits. There is a prior-record argument this entry can adopt: if the impurity were an in-vivo metabolite of sucralose capable of genotoxicity or carcinogenicity, the effect should have shown up in the many earlier in-vivo genotoxicity and carcinogenicity studies of sucralose, and it did not. The critique reached the peer-reviewed literature in a 2024 paper in Food and Chemical Toxicology by independent pathologists — whose authors and exact title could not be verified and which must not be cited by name until they are — and a 2025 class-level review on genotoxic and carcinogenic potential exists as a lead with funding unchecked. And cutting the other way: in 2023 the maker of Splenda commenced legal proceedings against the senior author alleging the study contained false and misleading scientific claims, and runs a rebuttal page asserting the product does not contain the impurity and that sucralose does not convert to it in the body. A manufacturer suing a critic is itself an incentive datum, and it is one reason an independent replication may never arrive. (Cui bono, and this one is unusual: the paper declares no conflict of interest and reports funding from the Engineering Foundation at North Carolina State University, so it is not industry-funded, which matters. But Susan Schiffman is a long-standing, publicly identified sucralose critic with prior work in that direction, one co-author works at a contract research organisation, and the paper's own closing line — that the findings "raise significant health concerns regarding the safety and regulatory status of sucralose itself" — is an advocacy conclusion drawn from screening assays. On the other side, the Calorie Control Council and the International Sweeteners Association issued rapid rebuttals; they exist to defend these products, and their criticism, however technically fair, is not disinterested. The cytotoxicity and assay-limit objections reach this entry partly through industry-adjacent summaries as well as the 2024 critique, and are attributed with that uncertainty attached.) Schiffman SS, Scholl EH, Furey TS, Nagle HT. Journal of Toxicology and Environmental Health, Part B. 2023.

10. An independent laboratory reported sucralose disrupting tight junctions and barrier function three years earlier, at roughly a tenth the concentration, with a receptor mechanism and a genetic control (Emerging Evidence — IN VITRO). Shil and colleagues, working in human intestinal epithelial cells with no live-animal arm, reported that sucralose and aspartame increased barrier permeability and down-regulated claudin-3 at the cell surface, that saccharin and aspartame caused cell death at higher concentrations, and — the part that makes this methodologically stronger than a screening assay — that knocking down the sweet-taste receptor T1R3 attenuated the effects, establishing receptor dependence. A mechanism and a genetic negative control are exactly the features a screening study lacks. The authors defend 10 millimolar as physiologically achievable and within acceptable intake levels. This entry contests that, and states the contest as arithmetic rather than as rhetoric: 10 millimolar is still roughly twenty-five times the sucralose concentration of a can, on the same approximate calculation given above, so "physiologically achievable" is the authors' claim and not settled fact. But twenty-five times is a completely different argument from two to four hundred times, and this entry previously made only the second one. The concentration gap has narrowed; it has not closed. The exact sub-10-millimolar values in this paper were not verified and must not be printed. The same line of work has since extended to neotame, reporting receptor-dependent effects on the epithelium plus pathogenic changes in model gut bacteria, and to monk fruit's sweet principle. (Cui bono: this is an academic paper whose finding cuts against industry, published in a journal industry also publishes in. Its authors have a stake in the receptor mechanism being important, which is the ordinary incentive of any research programme. This entry adopts its methodological strength and disputes its exposure claim, which is the correct way to handle a study you find half-convincing.) Shil A, Olusanya O, Ghufoor Z, Forson B, Marks J, Chichger H. Nutrients. 2020;12(6):1862.

11. Live-animal studies with a real gut have reported reduced mucus protein and disturbed tight junctions — mostly at or above intake-limit-equivalent doses, and in chemically induced disease models (Emerging Evidence — rodent). An entry whose organising instruction is "read the tiers" cannot skip a tier, and the previous version skipped this one entirely: it defended against cell culture by saying a dish is not a gut with mucus, a microbiome, blood flow and clearance, without telling the reader that studies in animals possessing all four have measured the exact proteins at issue. In a mouse model of colitis-associated colorectal cancer, sucralose in drinking water at 1.5 mg/mL for six weeks before chemical induction and throughout produced more severe weight loss and blood in the stool, tumour positivity rising from 50 to 87.5 per cent with more and larger tumours, reduced mucin 2, decreased occludin, and altered ZO-1 and claudin expression, alongside dysbiosis. Separately, six months of sucralose at 0.3 mg/mL — a dose that study states as equivalent to the American intake limit of about 5 mg/kg body weight per day — altered the gut microbiome and its functions and raised pro-inflammatory gene expression in the liver. The limits are load-bearing and cut hard. The barrier-and-tumour study is a chemically induced disease model, its own authors describe it as preliminary early-stage work with no direct translation to human doses, and its dose is roughly five times the one the other study treats as intake-limit-equivalent — so the barrier and mucus findings sit above the intake limit, not at it. That cross-study dose comparison is derived from two papers that state their doses differently; it comes from the hardening pass's own arithmetic and is unverified. The honest answer to "has anyone looked at the gut lining in a live animal" is therefore: yes; only the six-month microbiome study is at an intake-limit-equivalent dose; the barrier result is above it; and it is disease-model dependent. (Cui bono: this is the rung most useful to the alarm camp and it is why it is stated with its doses and its chemical colitis inducer attached. It is also the rung industry benefits from ignoring, because the concentration objection that answers cell culture does not answer an animal that has mucus and bacteria.) Li X, Liu Y, Wang Y, et al. Frontiers in Oncology. 2020;10:710; Bian X, et al. Frontiers in Physiology. 2017;8:487; also Frontiers in Nutrition. 2022;9:848392 on low-dose sucralose and the mouse microbiome.

12. Someone measured markers of gut leakiness in 572 people against habitual sweetener intake, and found no association (Emerging Evidence — human observational). This is the human rung of the barrier question and this entry previously did not have one, which meant its refusal to say "sucralose breaks your gut lining" rested entirely on an argument about concentrations when it could have rested partly on a measurement in people. In a cross-sectional substudy of 572 US adults inside a large cancer-prevention cohort, habitual intake of aspartame, sucralose, acesulfame-potassium and saccharin showed no association with antibodies to flagellin, antibodies to lipopolysaccharide, or total antibodies — circulating markers of bacterial products crossing the gut wall. Four limits keep this weak evidence of absence rather than strong evidence of no effect, and the authors themselves call for larger samples and randomised trials: it is cross-sectional and not a trial; intake was self-reported; 572 people is not large for this purpose; and antibody surrogates of bacterial translocation are not a direct permeability probe such as the urinary sugar test, which the literature notes is not feasible at cohort scale. The full author list could not be retrieved during verification and must be confirmed before this study is cited by name. The study that would actually settle the question — a direct permeability probe applied to habitual sweetener consumption — has not to our knowledge been done. (Cui bono: this is reassurance the previous version of this entry was throwing away for free, which is the mirror image of the error it usually guards against. Printing it is the same discipline as printing the millimolar concentrations.) "Association of Low-Calorie Sweeteners with Selected Circulating Biomarkers of Intestinal Permeability in the Cancer Prevention Study-3 Diet Assessment Substudy," Journal of Nutrition. 2025 (PubMed 40032143).

13. The 2023 paper's separate and more widely quoted argument — that a single drink can exceed a screening threshold for genotoxic compounds — is a trigger for assessment, not a demonstration of harm, and the two arguments must not be welded together (Emerging Evidence — IN VITRO paper plus exposure modelling). The paper notes a threshold of toxicological concern for genotoxicity of 0.15 micrograms per person per day, applied to chemicals present at low levels in the diet, and argues from an impurity content of up to 0.67% in commercial sucralose that single servings of sucralose-sweetened drinks may exceed that threshold by orders of magnitude. The specific multiple that circulates with this claim could not be re-verified against the primary text and must not be printed as a number — the arithmetic implied by the stated impurity fraction and ordinary serving sizes does not obviously support the largest figures in circulation. What this threshold is: a deliberately conservative screening value applied to compounds of unknown toxicity, to decide whether they need assessing. Exceeding it triggers evaluation. It does not describe a dose that damages DNA, and it is a completely different argument from the cell-culture concentrations above — an entry that presents them as one argument is misleading its reader. And the body whose threshold this is subsequently assessed the compound and reached the opposite conclusion (the regulatory claim below). (Cui bono: the exceedance sentence is the single most quotable and most dishonestly quotable line in this whole file, because it travels perfectly once the word "screening" falls off — and we are as capable of dropping that word as anyone, which is why the multiplier is not printed here at all. In the other direction, industry benefits from collapsing the paper into "wrong concentrations, ignore it," which skips the legitimate point that an impurity present at up to two-thirds of one percent was never itself toxicologically characterised before approval.) Schiffman SS, et al. Journal of Toxicology and Environmental Health, Part B. 2023, discussion of the threshold of toxicological concern.

The cancer evidence, which this entry previously had none of in either direction

14. A lifespan mouse bioassay reported blood-cell tumours from sucralose; the European regulator assessed it in a dedicated statement and concluded the data did not support that conclusion (Emerging Evidence — rodent, formally rejected by the regulator). This entry used to ban "a rodent finding presented without the word rodent" while carrying no rodent cancer finding to label — and a reader who meets "sucralose caused cancer in mice" online found nothing here to meet it with. The finding exists: a lifespan bioassay published in 2016 reported that sucralose in feed, from prenatal life through natural death, induced haematopoietic neoplasias — tumours of blood-forming tissue — in male Swiss mice. It was assessed and rejected in a dedicated 2017 statement by the European additives panel, on four stated grounds: no dose-response relationship between exposure and the reported effects; no known mode of action by which sucralose could cause carcinogenic effects; failure to meet the Bradford Hill criteria for cause and effect; and a treatment duration extending to natural death, which the panel argued compromises interpretation because tumour incidence rises with age. That rejection was carried forward into the 2026 re-evaluation, which concluded no safety concern for carcinogenicity or chronic toxicity — a conclusion this entry previously omitted while printing the genotoxicity one. Group sizes, doses and tumour incidences are unverified — the 2017 statement's full text was blocked and the primary paper was not read, so no numbers may be printed from either. One honest inconsistency, flagged rather than hidden: the sibling aspartame entry treats a run-to-natural-death bioassay design as arguably seeing more rather than less, and objects when a design difference is treated as a defect — while the regulator used that same objection to reject this study. Both readings cannot be right, and settling which one the corpus holds is an open question rather than something this entry resolves on its own authority. (Cui bono: this claim's absence was costing the entry a reassurance plank, so adding it helps the pro-sweetener side, which is exactly why it belongs. The Ramazzini Institute, which ran this class of bioassay, has a long-standing adversarial relationship with regulators in both directions.) Soffritti M, et al. 2016, International Journal of Occupational and Environmental Health; EFSA ANS Panel, "Statement on the validity of the conclusions of a mouse carcinogenicity study on sucralose (E 955) performed by the Ramazzini Institute," EFSA Journal. 2017;15(5):4784; carried forward in EFSA Journal. 2026;24:e9854.

15. The human cancer epidemiology specific to sucralose is null, and in the one cohort that measured the molecules themselves it is a within-study null (Moderate Evidence — human observational). In a French cohort of 102,865 adults followed a median 7.7 years with 3,358 incident cancers, artificial-sweetener intake was quantified at molecule level rather than as a beverage category. Sucralose showed no significant association with cancer — in the very same analysis that produced positive associations for aspartame (hazard ratio 1.15, 1.03 to 1.28) and acesulfame-K (1.13, 1.01 to 1.26). That makes it a within-study null rather than an underpowered absence, which is considerably harder to dismiss as the study simply being insensitive. Pooled across seventeen prospective cohorts, artificially sweetened beverages showed no significant association with overall cancer: relative risk 1.03, 95% confidence interval 0.96 to 1.11 — a category sucralose dominates in much of the market. The United States National Cancer Institute states directly that a range of studies have found no evidence that sucralose causes cancer in humans, naming that cohort among them. The limits of the source cohort are the authors' own and must travel: a volunteer cohort skewed female and highly educated, self-reported dietary records, intake lower than national estimates, residual confounding, reverse causality, and the authors' own statement that a single cohort cannot establish causation. The seventeen-cohort pooled analysis's full text was blocked during verification; only its overall-cancer estimate is carried here, and its site-specific results belong to the class-level and aspartame entries rather than to this one. (Cui bono: this is the cleanest human reassurance in the whole file and the entry previously did not carry it, which meant it spent four paragraphs defusing a cell-culture finding while ignoring the human null that defuses it in one sentence. Observational nulls are also the industry's favourite genre; the limits above are why this is graded Moderate and not Strong.) Debras C, Chazelas E, Srour B, et al. (Touvier M, senior). PLOS Medicine. 2022;19(3):e1003950; Yin S, et al. Nutrients. 2022;14(21):4445; National Cancer Institute summary on artificial sweeteners.

The immune evidence, which is a mechanism class this entry did not carry at all

16. A 2023 Nature paper reports sucralose blunting T-cell responses in mice — by a route that does not run through the gut bacteria — at doses the authors themselves say ordinary eating does not reach (Emerging Evidence — rodent). Zani, Blagih, Gronke and colleagues at the Francis Crick Institute gave mice sucralose in drinking water at 0.17 or 0.72 milligrams per millilitre, doses the authors relate to the acceptable daily intakes set by European (15 mg/kg body weight per day) and American (5 mg/kg per day) regulators. They reported reduced antigen-specific CD8 T-cell responses in subcutaneous tumour and bacterial infection models, and reduced T-cell function in models of T-cell-mediated autoimmunity. The mechanism is direct and not microbiome-mediated: altered T-cell membrane order, reduced efficiency of T-cell-receptor signalling, and reduced calcium mobilisation inside the cell. That matters beyond the finding itself, because it breaks the framing this entry used to open its mechanism section with — that the plausible route was contact with gut bacteria rather than anything systemic. An effect on a circulating immune cell requires the absorbed fraction to matter. Three limits are load-bearing and licence no harm claim whatsoever. Mice only; there is no human data. The effects were reversible once the sucralose stopped. And the authors' and the institute's own framing is that while these doses are achievable in principle they "would not normally be reached by people simply consuming food or drinks containing sweeteners as part of a normal diet" — with the paper pitched as a possible therapeutic route for autoimmune disease, not as a consumer warning. Printing the finding without that framing converts a paper about treating illness into a scare. Any per-kilogram conversion of the mouse dose is not carried here — allometric scaling appears to be what reconciles it to the human intake limit, but the citable statement is the authors' framing, not anyone's arithmetic. A review organising a "gut–immune axis" around this work exists at title level only and is a lead, not a receipt. (Cui bono: a Nature paper from a major institute is the most authoritative-sounding source in this file and the most abusable, which is why the mice, the reversibility and the authors' own dose statement are welded to it. In the other direction, industry benefits from filing it as irrelevant because it is mice — but a direct systemic mechanism in a mammal at intake-limit-related dosing is not nothing, and the honest position is that it is a hypothesis about people, not a finding about them.) Zani F, Blagih J, Gronke K, et al. Nature. 2023;615(7953):705-711; Nature Reviews Immunology research highlight, "Sucralose: not sweet enough for T cells"; Francis Crick Institute press release, 15 March 2023.

The heat question, where the regulator is the one raising the flag

17. In 2026 European regulators confirmed sucralose safe at current authorised uses but could not conclude on its safety in baked goods, citing uncertainty about chlorinated compounds forming across the range of baking processes, and extended that concern to home baking and frying; the German federal risk assessor separately recommends not heating sucralose-containing food above 120 °C (Moderate Evidence — regulatory assessment, supported by chemical analysis in heated model systems). This is the strongest precautionary line available on the topic, and it is not an advocacy source: it is the approving regulator, clearing the molecule for cold use and refusing to clear it for prolonged heat in the same document. The European Food Safety Authority could not conclude on the safety of the proposed extension into fine bakery wares because of uncertainties about the potential formation of chlorinated compounds under the wide range of baking processes that could apply, and stated that formation during home preparation requiring high temperatures — frying and baking — cannot be excluded, since domestic temperature, time and quantity vary far more widely than authorised industrial processing. The supporting chemistry is separate and must be attributed separately. From the German federal risk-assessment body: heating sucralose alone, with protein, and as a food ingredient produced "remarkable instability and discolouration after heating at 85–90 °C for 1 h," with chlorinated furan-3-one and chlorinated dicarbonyl compounds identified for the first time, and 3-chlorotyrosine formed in the presence of protein — indicating sucralose can chlorinate other biomolecules. That chlorine-transfer mechanism belongs to the chemistry paper, not to the regulator's stated reasoning. The same body has also published a consumer recommendation not to heat sucralose-containing foods above 120 °C, stating that sucralose decomposes when heated above that temperature for a prolonged period and naming polychlorinated dibenzo-p-dioxins, dibenzofurans and chloropropanols as compound classes of concern. Three honesty clauses travel with all of this permanently. First, these are analytical and precautionary statements about compounds that may form — no human or animal outcome study links heated sucralose to any disease. Second, the European regulator's position is precautionary non-confirmation: that is an absence of evidence of safety, not evidence of harm, and the entry must say it that way. Third, no quantified measurement of dioxin formation from sucralose in real food was located — the compound classes are named by a risk assessor as plausible products, so the 120 °C recommendation may be cited, but no numeric yield or exposure figure may ever be attached to it. (Cui bono: "don't bake with Splenda, it makes toxic chlorinated compounds" is clean, actionable and shareable, and the evidence supports "a national risk assessor recommends staying under 120 °C and the European regulator will not clear high heat," not "it is proven toxic hot." On the other side, an industry-associated paper argues heated sucralose in manufacturing poses no risk — it is promoted as reassurance by the sweetener trade body, its funding was not verified here, and note that it addresses industrial heat, precisely the scenario the regulator distinguished from domestic frying and baking. The manufacturer-side interest in confining the question to industrial processing is obvious.) EFSA FAF Panel, EFSA Journal. 2026;24:e9854, adopted 10 December 2025, published 17 February 2026; Eisenreich A, Gürtler R, Schäfer B. Food Chemistry. 2020;321:126700; German Federal Institute for Risk Assessment consumer notification on heating sucralose above 120 °C.

The regulatory position, which is stronger and more careful than either camp admits

18. Regulators reviewed this evidence and concluded there is no need to change the intake limit — and the intake limit itself is a hundredfold margin against reduced body-weight gain in rats, which is the sharpest version of this entry's own precautionary argument (Strong Evidence — regulatory assessment). The 2026 European re-evaluation concluded that there is no need to revise the acceptable daily intake of 15 mg per kg of body weight per day, that European dietary exposure sits below that level across all population groups, that there is no safety concern at the reported uses and use levels, no safety concern regarding genotoxicity of sucralose, its impurities and its degradation products, and no safety concern for carcinogenicity or chronic toxicity. The Panel also recommended amendments to the additive's specifications, which the coverage ties to purity and manufacturing questions; that recommendation is a partial concession to the impurity argument and belongs on the page. The part this entry previously asserted without evidence, and can now state with the regulator's own numbers: the relevant endpoint behind the reference point is a decrease in body weight in rats — not cancer, not organ damage, not a metabolic endpoint. Effects on body-weight gain appeared at 0.3 per cent in the diet, reported as roughly 150 mg per kg body weight per day, in the chronic toxicity and carcinogenicity study and in parents and offspring in the reproductive toxicity study, with reductions in food consumption of around ten per cent or less and not dose-related; the no-observed-adverse-effect level of 1500 mg per kg came from a 26-week gavage study, and the 15 mg per kg limit follows from an assessment factor of 100. So the safety margin is a hundredfold margin against reduced weight gain in rats, which is precisely why it is silent on a microbiome-mediated change in glucose handling. That is the legitimate core of the precautionary case, stated as arithmetic rather than as assertion. On how the disputed 2023 study was handled: the account in circulation is that the Panel was aware of an ongoing complaint about the publication, contacted the journal to establish whether any action had been taken, found none publicly communicated as of adoption, and included the data in the assessment anyway. That would be a regulator reading a study and disagreeing with it rather than ignoring it — but this specific account could not be confirmed against the primary text or any accessible source and is carried as UNVERIFIED. Do not present it as established. On the United States: no responsive statement was located. Sucralose remains approved under 21 CFR 172.831 with no pending restriction. The accurate sentence is that the Food and Drug Administration has not publicly revised its position — never that it rejected or responded to the study. (Sourcing caveat: the full 2026 opinion was paywalled and access-blocked during research; the intake limit, the exposure conclusion, the bakery non-conclusion, the home-cooking extension and the reference-point figures above are corroborated across independent retrievals with consistent wording, but the primary text was not read directly. Verify exact phrasing before publishing any direct quotation, and treat the reference-point numbers as inheriting that limitation.) (Cui bono: "approved by both agencies, intake limit re-confirmed in 2026" is the industry's strongest and most honest line and belongs in this entry. "Regulators are captured" is the reflexive dismissal and is refuted by the fact that the same opinion refused to clear high-temperature use.) EFSA FAF Panel, EFSA Journal. 2026;24:e9854; 21 CFR 172.831.

The benefit side, which is the half that makes the trade weak

19. The World Health Organization concluded in 2023 that non-sugar sweeteners should not be used for weight control, on evidence it graded low-certainty — and its trial evidence and its cohort evidence disagreed (Moderate Evidence — human RCTs and prospective cohorts, conditional recommendation). The guideline suggests non-sugar sweeteners not be used as a means of achieving weight control or reducing the risk of non-communicable disease. The two evidence streams must be reported separately, and reporting only one of them is the standard way this guideline is misused. In randomised trials — most lasting three months or less, graded low certainty — sweetener use produced reduced sugar and energy intake, lower body weight and lower body mass index. In prospective cohorts, with roughly thirteen years of average follow-up and graded very low to low certainty, higher sweetener intake was associated with higher body mass index and increased risk of incident obesity, and the guideline suggested possible undesirable effects of long-term use including raised risk of type 2 diabetes, cardiovascular disease and mortality. The conclusion that follows is narrow: no long-term benefit in reducing body fat in adults or children was demonstrated, while a short-term benefit was. Four qualifiers are load-bearing and dropping any of them makes this dishonest. It is a conditional recommendation, not a strong one. The overall certainty is low by the guideline's own grading. The disease-risk associations are observational and carry severe reverse-causation risk: people already gaining weight or already dysglycaemic are the ones who switch to diet drinks, which makes the sweetener look causal when it is a marker. And guidance for people with pre-existing diabetes was out of scope, with studies conducted exclusively in that population excluded. Most importantly for this entry: the guideline is about sweeteners as a long-term weight-control strategy. It is not a finding that substituting a diet drink for a sugared drink is harmful. The guideline was also publicly criticised in the peer-reviewed literature, and that criticism is acknowledged rather than hidden. (Cui bono: "the WHO says don't use sweeteners" is a maximally authoritative-sounding line produced by deleting "conditional," "low certainty," and the short-term trial result — do not produce it. The sugar and beverage industries benefit from the fact that the guideline did not say diet drinks are worse than sugared drinks, and that distinction is real and protects the honest position here.) World Health Organization. Use of non-sugar sweeteners: WHO guideline. Geneva; 15 May 2023.

A reasoning correction, not a finding

20. The "it's 99% water" defence is a category error, and so is its mirror image — and the same arithmetic tells you that a Splenda sachet is about 99% not-sucralose (not empirical — a reasoning point, labelled as such). Composition by mass is independent of biological activity: potency and mass fraction are unrelated quantities. Sucralose is the clean illustration — the 2023 paper's introduction gives it as approximately 385 to 650 times sweeter than sugar by weight, a figure carried here on that paper's authority rather than independently re-verified. A molecule at a vanishing mass fraction delivers the entire sweet signal, which is the whole commercial point of the product. The same applies to the trace impurity at up to 0.67%: "less than one percent" describes mass, not consequence. And the same arithmetic identifies something about the product that this entry previously never told its reader: tabletop Splenda is roughly one per cent sucralose by weight, the remaining ninety-nine per cent being maltodextrin or dextrose — rapidly absorbed carbohydrate acting as bulking agent. That has two consequences. It means a sachet stirred into coffee delivers sucralose plus carbohydrate, which is the exposure pattern of the pairing conditionality above, while a diet drink delivers sucralose without a carrier. And it means the most-cited animal study on sucralose and gut inflammation — a Crohn's-model study that administered Splenda specified as sucralose to maltodextrin at 1:99 by weight — cannot attribute its findings to sucralose at all, since ninety-nine per cent of the administered mass was the carrier. That carrier question recurs across three separate disputes in this file and is a real structural observation rather than a debating point: the 2014 origin microbiome study used commercial formulations with glucose bulking agents while the pure-compound replication attempt did not; the pairing trial's published re-analysis argues the maltodextrin did the work; and the Crohn's-model study cannot separate them. In several of the places where a signal appeared, a carbohydrate was in the mixture. Primary receipts for maltodextrin's own effects on mucosal defence were not confirmed and must not be asserted. And the correction constrains both sides equally. "A small fraction can matter" does not mean any particular small fraction does matter — establishing that requires dose-response data, which at realistic exposures for this impurity does not exist. The argument defeats a bad rebuttal. It proves nothing about harm, and an entry that uses it to imply harm has committed the mirror-image fallacy.

Mechanism

Why a molecule that is barely absorbed can still do something — and why that is no longer the only route described. Most ingested sucralose passes through the small intestine unabsorbed and reaches the colon, where the bacterial community lives. The best-supported route for the glycaemic finding is that colonic contact with the microbiome, which then alters how the body handles a glucose load, and it is best-supported because that is what the human trials measured: sweeteners altered the stool and oral microbiome in every group, glycaemic handling moved in two of them, and transferring the altered community into germ-free mice transferred the phenotype. But the previous version of this entry said the plausible route "is not systemic toxicity but contact with the gut microbiome," and that sentence is now too narrow — narrow enough that it is what kept an entire mechanism class off this page. Two direct host mechanisms have since been described, neither of which runs through the bacteria: an effect on immune-cell membrane order and receptor signalling in mice, which requires the absorbed fraction to matter; and sweet-taste-receptor-dependent effects on epithelial tight junctions in cell culture, attenuated by knocking the receptor down. Microbiome fundamentals belong to diet_gut_microbiome; what this entry owns is that sucralose is one of the sweeteners where the human glycaemic signal actually appeared, and that the map of how it could matter is now larger than the colon.

Why the microbiome mechanism sits outside the framework that approved it. The acceptable daily intake was built from rodent toxicology, and the regulator's own reference point rests on reduced body-weight gain in rats, with a hundredfold assessment factor applied. A microbiome-mediated shift in glucose tolerance is not the kind of endpoint that framework was designed to catch. This is not an accusation of negligence — it is a structural point about what the safety assessment was looking for. It is why "the intake limit has a huge safety margin" and "there is an unresolved question here" can both be true.

Why the immune finding is a different shape of mechanism. An effect on T-cell membrane order, receptor-signalling efficiency and intracellular calcium is systemic: it acts on a circulating cell, not on a bacterial community, and it cannot be reasoned about using the "barely absorbed" argument at all. It is in mice, it reversed on washout, and its authors say ordinary diets do not reach the doses used. It is carried here as a mechanism that exists rather than a risk that has been demonstrated — and specifically as the reason the old framing sentence had to change.

Why the pairing finding, if it holds, has a coherent story. The proposed explanation for the sucralose-with-carbohydrate results is a mismatch between sweet taste and delivered calories: the body uses sweetness as a forecast of incoming energy, and repeatedly breaking that forecast may degrade the anticipatory metabolic response. The pairing trial found the blunting was specific to sweet taste rather than sour, salty or savoury, which is consistent with that story. Three small trials point the same way with a live dispute about whether the added carbohydrate did the work in the best-known of them, so treat the mechanism as a hypothesis that fits, not as an explanation of something established.

The appetite route, which is no longer untested. A 75-person randomised crossover in young adults across body weights, publicly funded by the United States national institutes of health rather than by industry, measured brain blood flow after sucralose, sucrose and water: sucralose increased hypothalamic blood flow against both sucrose and water, increased connectivity between the hypothalamus and regions involved in motivation and somatosensory processing against both comparators, and produced greater hunger than sucrose; only sucrose raised blood glucose, and that rise correlated with reduced blood flow in the medial hypothalamus. Two limits cut hard the other way and must travel with it. Hunger did not differ significantly between sucralose and water — so this is not evidence that a diet drink makes you hungrier than drinking nothing sweet. And brain perfusion and appetite ratings are surrogates for behaviour, not behaviour. The honest net statement is that the appetite mechanism is real and measurable in the brain, and that the taste-set-point argument for treating sweeteners as a bridge is a design preference and a plausible mechanism rather than an outcome finding. Chakravartti SP, Jann K, Veit R, et al. Nature Metabolism. 2025.

Why heating is a different mechanism entirely. Sucralose is sucrose with three chlorine atoms substituted in, and the chlorine is why it is not metabolised as a sugar. Heating it degrades the molecule and can transfer chlorine to other organic molecules present in food — including, in the presence of protein, chlorination of tyrosine. This has nothing to do with the microbiome pathway and nothing to do with the intake limit. It is a chemical stability question, and it is the one the regulator declined to close.

Where the barrier findings do and do not connect, stated as a ladder rather than a dismissal. A barrier-integrity effect on intestinal lining would be a mechanistically coherent way for a gut-contact compound to matter systemically, which is why the in-vitro findings were reported so widely. Concentration is still the break in the chain, but the break is smaller than this entry used to claim: the 2023 paper's sucralose arms sit at roughly two to four hundred times the concentration of a can, while the independent 2020 receptor-dependent work sits at roughly twenty-five times, on an approximate calculation from serving size and molecular weight. And the "a dish is not a gut with mucus, a microbiome, blood flow and clearance" argument, which this entry has leaned on, must now be followed by its concession: animal studies possessing all four have reported reduced mucin 2 and decreased occludin, at or above intake-limit-equivalent doses, in chemically induced disease models. What is still missing is a barrier or mucus result in a healthy animal at a genuinely intake-limit-equivalent dose, and any barrier result in a person — where the one available measurement, circulating markers of gut leakiness in 572 adults, was null. Barrier biology itself belongs to leaky_gut_intestinal_permeability; what this entry owns is the refusal to treat a dish result as a person result, and the refusal to pretend the intermediate rungs are empty.

Risks And Contraindications

• Never state or imply that sucralose causes cancer, damages DNA in people, or breaks the gut lining. On cancer the anchor is now human rather than merely argumentative: in the one large cohort measuring the molecule, sucralose showed no association while two other sweeteners in the same analysis did, the seventeen-cohort pooled estimate for artificially sweetened drinks and overall cancer is null (1.03, 0.96 to 1.11), and the regulator concluded no safety concern for genotoxicity or carcinogenicity. On the gut lining, the human measurement that exists — circulating leakiness markers in 572 adults — is null; the positive findings are cell culture and diseased animals. The moment "possibly, in cell culture" becomes "causes," this entry is worthless.
• State the barrier concentrations as a comparison, not as "orders of magnitude." The 2023 paper's sucralose arms sit at roughly two to four hundred times the concentration of a can; the independent 2020 receptor-dependent work sits at roughly twenty-five times. Both figures come from an approximate calculation using stated serving sizes and molecular weight and must be labelled as such. Twenty-five times is a much weaker objection than four hundred, and pretending otherwise is the kind of imprecision this entry exists to refuse.
• Never say the barrier finding is unreplicated. The genotoxicity finding is unreplicated. The barrier finding has independent precedent from a separate laboratory three years earlier, receptor-dependent, with a knockdown control, at roughly a tenth the concentration.
• Never present the germ-free mouse transplant result as human proof. It is the causal test and it worked — in mice, using donors selected from the extremes of the response range. Name the mice in the same sentence as the finding, every time.
• Never render the immune finding as "sucralose suppresses your immune system." It is mice, the effects reversed on washout, and the authors state ordinary food and drink would not reach the doses used; the paper is pitched as a possible treatment for autoimmune disease. Never separate the finding from the word mice, and never separate it from the authors' own dose statement. No per-kilogram conversion of the mouse dose may be printed.
• Never present the rodent cancer bioassay without the regulator's rejection of it, and never with numbers. A lifespan mouse study reported blood-cell tumours; the European panel assessed it in a dedicated statement and concluded the data did not support that conclusion, on four stated grounds. Group sizes, doses and tumour incidences are unverified and must not be printed. Flag, rather than resolve, that the sibling aspartame entry reads run-to-natural-death designs more favourably than the regulator does.
• Never say the Food and Drug Administration rejected or responded to the 2023 study. No such statement was located. The accurate sentence is that sucralose remains approved and the agency has not publicly revised its position.
• Never quote the screening-threshold exceedance without the word screening, and never print a multiple. That threshold is a conservative trigger for assessment applied to compounds of unknown toxicity, not a harm level; the assessment was subsequently done and reached the opposite conclusion; and the specific multiple in circulation could not be verified against the primary text.
• Never present the heating finding as demonstrated toxicity. The European regulator could not conclude on safety for baked goods and could not exclude formation during home frying and baking. That is an absence of demonstrated safety. No human or animal outcome study links heated sucralose to disease. The German risk assessor's recommendation to stay below 120 °C may be cited as a recommendation; no numeric yield or exposure figure for dioxins, furans or chloropropanols may be attached to it, because no quantified measurement from sucralose in real food was located.
• Never quote the World Health Organization position without "conditional" and "low certainty," and never without the fact that its short-term randomised trials showed lower weight and body mass index. What was not demonstrated is a long-term body-fat benefit. It does not say substitution is harmful, and guidance for people with pre-existing diabetes was out of scope.
• Never use this entry to push anyone back toward sugar, or toward fruit juice. Both are worse trades on better-established evidence, and juice specifically tracks higher type-2 diabetes risk where whole fruit tracks lower.
• Do not print a sample size for the thirty-day trial — it was not stated in the record retrieved during verification. Its other numbers are verified and may be used. Do not print funding claims for the 2018 trial, whose funding and conflict declarations were not extracted.
• Do not quote numbers from the remaining microbiome leads — the two twelve-week Indian trials and the ten-week study were confirmed only to title and journal level. Do not quote numbers from the stevia and monk-fruit barrier report, the gut–immune-axis review, or the 2024 critique of the screening paper, and do not name that critique's authors until they are verified.
• People with diabetes: the World Health Organization recommendation explicitly does not cover them, and glycaemic management decisions belong with their clinician. Nothing here is a reason to abandon a sweetener that is helping someone control blood glucose — and the one conditional trial that included people with type 2 diabetes found the gut-hormone effect present in healthy subjects and absent in that group.
• People with inflammatory bowel disease or another established gut condition: a watch item, not a warning. The animal evidence says the host's disease state is what decides the outcome; the human cohort signal for these diseases attaches to sugar; and patients with these conditions preferentially choose artificial sweeteners and diet products, which is why observational data in this population is especially prone to reverse causation.
• Anyone with an eating disorder history: framing any everyday food or drink as contaminating is a live harm in that population, and this entry's precautionary structure must never be delivered as a purity message.

Controversy

Nature: a genuinely open file, in which the pro side's strongest claims are true, the anti side's strongest claim has now replicated on a surrogate marker while its most quotable claim still sits in a dish, and the argument that actually decides the matter is about the durable benefit, not the risk.

Position A — "Approved, well within the intake limit, decades of use, this is settled."
• Best evidence: European regulators completed a full re-evaluation in 2026, confirmed the acceptable daily intake of 15 mg per kg of body weight per day, found European exposure below that across all population groups, and concluded no safety concern at reported uses — including for genotoxicity, carcinogenicity and chronic toxicity, and having assessed and rejected the one lifespan mouse cancer bioassay in a dedicated statement. Sucralose remains approved in the United States. Human cancer epidemiology is null: a within-study null for sucralose in the one cohort that measured the molecules, and a pooled null across seventeen cohorts. The one human measurement of gut-leakiness markers, in 572 adults, is null. At least five independent human trials across three delivery routes show sucralose taken alone acutely produces no insulin or gut-hormone response, and the pooled acute data show no glucose spike. The World Health Organization's own randomised-trial evidence showed sweetener use reduced energy intake and lowered body weight and body mass index over the short term. No human outcome trial shows harm. All of that is true and this entry prints it.
• Where it overreaches: it treats an intake limit derived from rodent toxicology as though it settled a question that framework was never designed to ask — the limit is a hundredfold margin against reduced weight gain in rats, by the regulator's own reference point. It can no longer say the glycaemic finding is a one-off: three independent randomised human trials over two weeks to a month converge on impaired glucose handling or insulin sensitivity below the legal limit, two of them with sucralose given alone. It skips the fact that in the very opinion it cites, the regulator could not conclude on safety for baked goods, extended that to home cooking, and recommended the specifications be amended. It has nothing to say about live-animal work reporting reduced mucus protein and disturbed tight junctions. And it quietly assumes a long-term benefit that the World Health Organization reviewed and could not confirm.

Position B — "It damages your gut and breaks your DNA."
• Best evidence: three independent human randomised trials found sucralose impaired glucose handling or insulin sensitivity at doses below the legal limit, with microbiome and metabolome changes, and the transplant experiment showed the microbiome change was sufficient to carry the effect into germ-free mice. An independent laboratory reported barrier disruption and tight-junction loss in human intestinal cells at roughly a tenth the concentration of the famous 2023 screen, receptor-dependent and confirmed by knockdown. Live animals with mucus, bacteria and blood flow have shown reduced mucin 2 and decreased occludin. A 2023 Nature paper describes sucralose blunting T-cell responses by a route that does not run through the bacteria at all. An impurity present at up to 0.67% of commercial sucralose was never itself toxicologically characterised before approval. And the heat-stability question is real enough that the European regulator refused to close it and a national risk assessor issued a temperature recommendation.
• Where it overreaches: the DNA finding is a single unreplicated cell-culture screen, criticised in print on validity grounds sharper than concentration — possible cytotoxic confounding in both assays, doses above the assays' own recommended limits, and the absence of any such effect across the many prior in-vivo studies. The barrier concentrations are still roughly twenty-five to four hundred times a can depending on the study. The live-animal barrier findings are above intake-limit-equivalent dosing and come from chemically induced disease models. The immune work is mice, reversible on washout, at doses its own authors say ordinary diets do not reach, in a paper pitched as a treatment for autoimmune disease. The pairing conditionality is three small trials, one formally disputed, and the trials disagree with each other about whether sucralose alone does anything over two weeks. The most-cited animal study of "Splenda and gut inflammation" administered 99% maltodextrin. Human cancer epidemiology specific to sucralose is null, gut-leakiness markers in people are null, and there is no human outcome evidence of harm at all — none, in decades of use.

The funding and bias dimension — cui bono, both ways. On the defending side the incentives are obvious and enormous: the beverage industry, the sweetener manufacturers, and the trade bodies that exist to rebut exactly these studies. Their criticisms of the 2023 paper were technically fair and entirely self-interested at the same time, which is the normal condition of this literature — and the published re-analysis of the pairing trial comes from authors whose declared disclosures include support from the Calorie Control Council and, for one author, a fund established by a major sucralose supplier. New to this version, and it is a fact rather than an inference: in 2023 the maker of Splenda commenced legal proceedings against the 2023 paper's senior author. A manufacturer suing a critic is an incentive datum in its own right, and it is one reason an independent replication of the genotoxicity finding may never arrive — which has to be said in the same breath as "it has not been replicated," because those two sentences interact. On the attacking side, the incentive structure is less obvious but real: the 2023 paper declares no conflict and was funded by a university engineering foundation, not by industry — which matters and is stated plainly — but its senior author is a long-standing public critic of sucralose, one co-author works at a contract research organisation, and the paper's closing sentence is advocacy rather than assay. Two of the studies that cut hardest against industry in this version are publicly funded and not industry-linked: the thirty-day insulin-sensitivity trial by Mexico's national science council, and the 75-person brain-imaging trial by the United States national institutes of health. Two of the most reassuring items carry unresolved funding questions in the other direction: the pure-compound saccharin null is promoted by an industry association with its funding unverified, and the one-year microbiome data comes from a consortium publicly described as including industry partners, unverified in the paper. A structural point worth stating once rather than repeating: across three separate disputes in this file the live question is whether the sweetener or the bulking carbohydrate did the work — the origin microbiome study used commercial formulations with glucose bulking agents while the pure-compound null did not, the pairing trial's re-analysis argues the maltodextrin did the damage, and the most-cited animal inflammation study administered a 1:99 sucralose-to-maltodextrin mixture. Beyond the researchers, an entire wellness economy monetises "sweeteners are poison" through natural-sweetener brands, detox and gut-healing products, and scanner apps. The least conflicted anchors in the file are the academic and publicly funded human trials, whose findings cut against industry, and the 2026 regulatory opinion, which is a public risk assessor and whose findings cut against the alarm camp — and note that these disagree in emphasis without contradicting each other. The regulator did not test glycaemic response; the trials did not test genotoxicity. And this platform's own incentive belongs on the ledger: a service positioned on honest evidence has a structural pull toward discovering that the mainstream is wrong, because an open question is more interesting to publish than a closed one. Every asterisk in this entry exists because that pull was resisted deliberately — including printing the short-term trial result that weakens our own conclusion, and printing the human nulls on cancer and permeability that weaken our most quotable sources.

Realised Position: Sucralose is not proven harmful and a diet drink is a genuine step off a sugared one — say that first, and never say otherwise. But it is a bridge, not a destination, and the reason is still the benefit rather than the risk: a global health body reviewed whether these products deliver long-term weight control and could not confirm that they do, while the short-term signal that used to be one trial is now three, cell-culture and animal work raise flags that remain unresolved at realistic doses, a mouse study describes an immune route nobody expected, and the approving regulator itself declined in 2026 to clear the molecule for prolonged heat. Against all of that sit human nulls on cancer and on gut leakiness, and no demonstrated harm in decades of use. Accepting a growing set of open questions in exchange for a long-term benefit nobody can confirm is a weak trade — so use it to get off sugar, don't cook with it, prefer the drink to the sachet if you are going to act on anything here, and treat water as the finish line. Every hedge in that sentence is doing work and none of them may be removed.

Cross-Pillar Connections

• Diet (artificial_sweeteners_evidence): the parent entry for the class-level argument, the full World Health Organization guideline with both its trial and cohort streams and its published criticism, and the comparison across sweeteners. This entry defers all of that and owns only what is sucralose-specific.
• Diet (aspartame_and_the_iarc_classification): the sibling case study in how a hazard classification gets read as a harm claim — the same failure mode this entry guards against with cell-culture findings. One live inconsistency between the two files, flagged rather than papered over: that entry reads a run-to-natural-death rodent bioassay as arguably seeing more rather than less, while the regulator used that same design objection to reject the sucralose mouse study. Whatever the settled reading is, it has to be the same in both files.
• Diet (erythritol_and_sugar_alcohols): the other direction people go when leaving sucralose, with its own distinct evidence file. Findings from that class — including its genetic instrumental-variable work — must never be imported here, and vice versa.
• Diet (diet_gut_microbiome): owns microbiome fundamentals, what diversity measures do and do not mean, and why "altered microbiome" is not automatically "harm." The mechanism here plugs into that entry, including the butyrate reduction reported in the thirty-day trial.
• Diet (leaky_gut_intestinal_permeability): owns barrier biology and the extremely contested question of what intestinal permeability means clinically — the correct destination for anyone who arrives here via the in-vitro barrier findings, and the right place for the general problem that antibody surrogates of bacterial translocation are not a direct permeability measurement.
• Diet (blood_sugar_regulation) and (postprandial_glucose_spikes): the glycaemic-response endpoint the human trials actually measured, and what a change in it is and is not worth — including the distinction between no acute spike from the drink and a two-to-four-week shift in insulin sensitivity.
• Diet (insulin_resistance_and_metabolic_dysfunction): where the repeated-exposure insulin-sensitivity finding would matter if it persisted past a month, and the outcome nobody has yet linked to sucralose.
• Diet (food_additives_to_avoid): the additive-triage entry. Keep the two consistent — that entry's position that meta-analyses show modest or no harm for weight management stands, and its preference for stevia should carry two qualifiers: the human trial found stevia altered the microbiome and metabolome too, just not glycaemic response, and a 2025 report (carried as a lead only) has stevia returning a null on intestinal barrier function where several other sweeteners were positive.
• Cross-pillar (rct_vs_observational_evidence): the reasoning tool this topic needs most, since reverse causation is the single largest confound in the observational sweetener literature — with the sharpest available illustration being that patients with inflammatory bowel disease preferentially choose artificial sweeteners and diet products over table sugar, which is the confound observed directly in a named population rather than argued in the abstract.

What would change our mind

Industry bias note

Sources (34)

• **Suez J, Cohen Y, Valdes-Mas R, Mor U, Dori-Bachash M, Federici S, Zmora N, Leshem A, Heinemann M, Linevsky R, Zur M, Ben-Zeev Brik R, Bukimer A, Eliyahu-Miller S, Metz A, Fischbein R, Sharov O, Malitsky S, Itkin M, Stettner N, Harmelin A, Shapiro H, Stein-Thoeringer CK, Segal E, Elinav E. Personalized microbiome-driven effects of non-nutritive sweeteners on human glucose tolerance. Cell. 2022;185(18):3307-3328.e19. doi:10.1016/j.cell.2022.07.016↗ — HUMAN RCT, 120 healthy adults, six arms, two weeks, doses below the acceptable daily intake; saccharin and sucralose significantly impaired glycaemic responses, aspartame and stevia did not; all four altered stool and oral microbiome and plasma metabolome; gnotobiotic mouse colonisation reproduced donor responses, exemplified by sucralose (rodent). Per-arm figure of approximately twenty follows from 120 across six arms. Re-verified against the primary abstract.** (Academic; findings cut against industry interest.)
• **Romo-Romo A, Aguilar-Salinas CA, Brito-Córdova GX, Gómez-Díaz RA, Almeda-Valdes P. Sucralose decreases insulin sensitivity in healthy subjects: a randomized controlled trial. American Journal of Clinical Nutrition. 2018. — HUMAN RCT, 66 healthy adults, 33 per group, sucralose alone at fifteen per cent of the acceptable daily intake for fourteen days; insulin sensitivity fell a median 17.7% (interquartile range −29.3 to −1.0) against −2.8% in controls, p=0.04; acute insulin response rose 577 to 671 in adherent participants, p=0.04. This trial predates the 2022 trial by four years and was absent from earlier versions of this entry — its absence is why the entry wrongly described the finding as unreplicated.** (Funding and conflict declarations were not extracted and are unverified — check before citing in any funding-sensitive context.)↗
• **Romo-Romo A, Sánchez-Tapia M, Almeda-Valdes P, Tovar AR, Torres N, et al. Sucralose consumption modifies glucose homeostasis, gut microbiota, Curli protein, and related metabolites in healthy individuals: a randomized placebo-controlled, triple-blind trial. Clinical Nutrition ESPEN. 2025;69:733-744. — HUMAN RCT, thirty days at thirty per cent of the intake limit; glucose, insulin and gut-hormone areas under the curve all rose after a mixed meal; insulin sensitivity down 20.3%; microbial diversity down; faecal butyrate down; inflammatory markers and branched-chain amino acids up. Funded by Mexico's national science council under grant 316514, not industry. Promoted in this version from a title-level lead to a receipt.** (Sample size was not stated in the retrieved record — no participant count may be printed.)↗
• Science Media Centre expert reaction to Suez et al., 19 August 2022 — Francisco Guarner (Vall d'Hebron University Hospital) and Ascensión Marcos (CSIC) on per-arm sample size; Kevin McConway (Open University) on what the human arm can and cannot show about causation; Duane Mellor (Aston University) on effect size. (Secondary commentary; not re-verified line by line.)↗
• **Dalenberg JR, Patel BP, Denis R, Veldhuizen MG, Nakamura Y, Vinke PC, Luquet S, Small DM. Short-Term Consumption of Sucralose with, but Not without, Carbohydrate Impairs Neural and Metabolic Sensitivity to Sugar in Humans. Cell Metabolism. 2020;31(3):493-502. doi:10.1016/j.cmet.2020.01.014↗ — HUMAN RCT, 45 participants randomised to three groups (fifteen per group as randomised, fewer analysed on some measures; the "thirteen per arm" figure in circulation could not be verified and is not used); insulin sensitivity and brain sweet-taste response impaired only in the sucralose-plus-maltodextrin arm. Contested:** Khan TA, Sievenpiper JL, re-analysis, Cell Metabolism 2020, arguing the comparator choice confounds the result and finding no difference between maltodextrin alone and sucralose-maltodextrin; author reply, Cell Metabolism 2021. Note the tension with the two trials above, which found sucralose alone did move insulin sensitivity at similar or longer duration. (Disclosure verified: the re-analysis authors' declared disclosures include support from the Calorie Control Council, a sweetener trade body, and Sievenpiper additionally reports support from a fund established by Tate & Lyle, a major sucralose supplier. The critique may be methodologically sound and the interest still must be printed.)
• **Pepino MY, Tiemann CD, Patterson BW, Wice BM, Klein S. Sucralose affects glycemic and hormonal responses to an oral glucose load. Diabetes Care. 2013;36(9):2530-2535 — HUMAN RCT, 17 obese insulin-sensitive adults, randomised crossover, sucralose or water ten minutes before a five-hour modified glucose tolerance test; roughly a twenty per cent rise in total plasma insulin with reduced insulin sensitivity. Contested in the same way:** published Comment and author Response, Diabetes Care 2014;37(6):e148/e149. (Verified at abstract level.)↗
• Temizkan S, et al. Sucralose enhances GLP-1 release and lowers blood glucose in the presence of carbohydrate in healthy subjects but not in patients with type 2 diabetes. European Journal of Clinical Nutrition. 2015 — HUMAN, conditional on carbohydrate and on metabolic status. Verified at abstract level only.↗
• The acute sweetener-alone human null, carried as a set rather than as one study. Steinert RE, et al. Clinical Nutrition 2011, with Gerspach AC, et al. American Journal of Physiology: Endocrinology and Metabolism 2011 — sweet-receptor components measured in human gastrointestinal tissue with co-localisation to gut-hormone cells by immunohistochemistry, plus a sweet-receptor blocker arm in 16 subjects. Ma J, et al. American Journal of Physiology: Gastrointestinal and Liver Physiology 2009 — intragastric sucralose up to 800 mg, seven subjects, no insulin, GLP-1 or GIP response and no change in gastric emptying. Ma J, et al. British Journal of Nutrition — no change in small-intestinal glucose absorption. Ford HE, et al. European Journal of Clinical Nutrition 2011 — no gut-hormone response to oral sucralose. Wu T, et al. 2012 — intraduodenal sucralose with or without glucose, no effect. Brown AW, et al. Diabetes Care 2013;36(12):e202 — no effect on gastric emptying, GLP-1 or glycaemia after oral glucose. Plus a 2017 human study (PubMed 28441725) concluding gut sweet-taste-receptor activation is of limited importance even for glucose-stimulated incretin release. All located at abstract or title level: directions are consistent across independent retrievals, individual effect sizes were not extracted and must not be printed. (This is the strongest reassurance plank in the file on the acute question and earlier versions of this entry rested it on one seven-person study.)↗
• **Schiffman SS, Scholl EH, Furey TS, Nagle HT. Toxicological and pharmacokinetic properties of sucralose-6-acetate and its parent sucralose: in vitro screening assays. Journal of Toxicology and Environmental Health, Part B. 2023. doi:10.1080/10937404.2023.2213903↗ — IN VITRO: TK6 human lymphoblastoid cells and a RepliGut human intestinal epithelium model. Sucralose-6-acetate genotoxic and clastogenic; raised inflammation, oxidative-stress and cancer-associated gene expression, greatest for MT1G; both compounds impaired barrier integrity by electrical resistance and permeability; inhibited CYP1A2 and CYP2C19. Concentrations as reported: barrier loss for the impurity from 5 mM with total loss at 10 mM; sucralose tested at 80 and 160 mM; genotoxicity at 353 µg/ml with metabolic activation and 707 µg/ml without — roughly two to four hundred times the ~0.4 mM of a 355 ml can, an approximate calculation and not a measurement. Separate exposure argument from a threshold of toxicological concern for genotoxicity of 0.15 µg/person/day. Sweetness potency figure of approximately 385 to 650 times sucrose by weight is from this paper's introduction. Full text inaccessible in verification (publisher returned 403 on repeated attempts): concentration figures, impurity fraction and potency range are carried on the paper's authority and were NOT independently re-verified, and the exceedance multiple often quoted with the threshold argument is UNVERIFIED and must not be printed. Reception:** two validity objections beyond concentration — possible cytotoxic confounding in both genotoxicity assays, and doses exceeding the assays' own recommended limits — plus the prior-in-vivo-record argument; a 2024 critique in Food and Chemical Toxicology by independent pathologists whose authors and exact title are unverified and must not be named until confirmed; a 2025 class-level review on genotoxic and carcinogenic potential (likely Advances in Nutrition 2025, funding unchecked, lead only); and legal proceedings commenced in 2023 by the maker of Splenda against the senior author, alongside a manufacturer rebuttal page. (Declares no conflict of interest; funded by the Engineering Foundation at North Carolina State University — not industry-funded. Senior author is a long-standing public sucralose critic; one co-author at a contract research organisation.)
• **Shil A, Olusanya O, Ghufoor Z, Forson B, Marks J, Chichger H. Artificial Sweeteners Disrupt Tight Junctions and Barrier Function in the Intestinal Epithelium through Activation of the Sweet Taste Receptor, T1R3. Nutrients. 2020;12(6):1862. doi:10.3390/nu12061862↗ — IN VITRO, human intestinal epithelial cells, no in-vivo arm; sucralose and aspartame increased permeability and down-regulated claudin-3 at the cell surface; saccharin and aspartame caused apoptosis at higher concentrations; receptor knockdown attenuated the effects, establishing sweet-taste-receptor dependence. Authors defend 10 mM as physiologically achievable and within acceptable intake levels; this entry contests that, since 10 mM is roughly twenty-five times a can. This is the independent precedent that makes "the barrier finding is unreplicated" false — the genotoxicity finding remains unreplicated. Paper not read in full: exact sub-10-millimolar values are unverified and must not be printed.** Same line of work extended to neotame (Frontiers in Nutrition 2024;11:1366409) and to monk fruit's sweet principle.
• **Li X, Liu Y, Wang Y, et al. Sucralose Promotes Colitis-Associated Colorectal Cancer Risk in a Murine Model Along With Changes in Microbiota. Frontiers in Oncology. 2020;10:710 — RODENT, in vivo, chemically induced colitis-cancer model, sucralose 1.5 mg/mL in drinking water six weeks before induction and throughout; more severe weight loss and blood in stool, tumour positivity 50% to 87.5% with more and larger tumours, mucin 2 reduced, occludin decreased, ZO-1 and claudin altered, with dysbiosis. Authors describe it as preliminary early-stage work with no direct translation to human doses. Its dose is roughly five times the intake-limit-equivalent dose used in the study below — that cross-study comparison is derived arithmetic, not stated by either paper, and is unverified.**↗
• **Bian X, et al. Gut Microbiome Response to Sucralose and Its Potential Role in Inducing Liver Inflammation in Mice. Frontiers in Physiology. 2017;8:487 — RODENT**, six months at 0.3 mg/mL, stated as equivalent to the American intake limit of about 5 mg/kg body weight/day; altered gut microbiome and function, raised pro-inflammatory gene expression in liver. Also Frontiers in Nutrition 2022;9:848392 on low-dose sucralose and the mouse microbiome.↗
• **Association of Low-Calorie Sweeteners with Selected Circulating Biomarkers of Intestinal Permeability in the Cancer Prevention Study-3 Diet Assessment Substudy. Journal of Nutrition. 2025 (PubMed 40032143) — HUMAN OBSERVATIONAL, 572 US adults, cross-sectional; habitual aspartame, sucralose, acesulfame-potassium and saccharin intake against antibodies to flagellin, antibodies to lipopolysaccharide and total antibodies: no association. Limits: cross-sectional, self-reported intake, 572 participants, antibody surrogates of bacterial translocation rather than a direct permeability probe (which the literature notes is not feasible at cohort scale); authors call for larger samples and randomised trials. Full author list could not be retrieved and must be confirmed before citing by name. The direct human method has not to our knowledge been applied to habitual sweetener consumption — that is the study that would settle the barrier question.**↗
• **Zani F, Blagih J, Gronke K, et al. The dietary sweetener sucralose is a negative modulator of T cell-mediated responses. Nature. 2023;615(7953):705-711 (Francis Crick Institute) — RODENT, sucralose in mouse drinking water at 0.17 or 0.72 mg/mL, doses the authors relate to the European (15 mg/kg/day) and American (5 mg/kg/day) intake limits; reduced antigen-specific CD8 T-cell responses in tumour and bacterial infection models, reduced T-cell function in autoimmunity models; mechanism direct rather than microbiome-mediated — altered T-cell membrane order, reduced T-cell-receptor signalling efficiency, reduced intracellular calcium mobilisation; effects reversible on washout. The authors' and institute's own framing is that these doses "would not normally be reached by people simply consuming food or drinks containing sweeteners as part of a normal diet," and the paper is pitched as a possible therapeutic route for autoimmune disease.** Supporting commentary: Nature Reviews Immunology research highlight, "Sucralose: not sweet enough for T cells," doi 10.1038/s41577-023-00873-x↗; institute press release 15 March 2023. No per-kilogram conversion of the mouse dose may be printed — allometric scaling appears to reconcile it to the human limit but the citable statement is the authors' framing. A review on a sucralose gut–immune axis exists at title level only: lead, not receipt.
• **Chakravartti SP, Jann K, Veit R, Liu H, Yunker AG, Angelo B, Monterosso JR, Xiang AH, Kullmann S, Page KA. Non-caloric sweetener effects on brain appetite regulation in individuals across varying body weights. Nature Metabolism. 2025 — HUMAN RCT, 75 young adults across body weights, randomised crossover, registration NCT02945475, funded by the United States national institutes of health (R01DK102794, F31DK137584), not industry; sucralose increased hypothalamic cerebral blood flow against sucrose (p<0.018) and water (p<0.019), increased hypothalamic connectivity with motivation and somatosensory regions against both, and produced greater hunger than sucrose (p<0.001); only sucrose raised peripheral glucose, and that rise correlated with reduced medial hypothalamic blood flow (p<0.007). Hunger did NOT differ significantly between sucralose and water** — this is not evidence a diet drink makes you hungrier than drinking nothing sweet, and brain perfusion and appetite ratings are surrogates for behaviour.↗
• **Debras C, Chazelas E, Srour B, et al. (Touvier M, senior). Artificial sweeteners and cancer risk: Results from the NutriNet-Santé population-based cohort study. PLOS Medicine. 2022;19(3):e1003950 — HUMAN OBSERVATIONAL, 102,865 French adults, median 7.7 years, 3,358 incident cancers, molecule-level exposure; sucralose showed no significant association, in the same analysis where aspartame was 1.15 (1.03-1.28) and acesulfame-K 1.13 (1.01-1.26) — a within-study null**. Authors' own limits: volunteer cohort skewed female and highly educated, self-reported 24-hour dietary records, intake below national estimates, selection bias, residual confounding, reverse causality, and no causal inference from a single study. (The aspartame and acesulfame-K results belong to aspartame_and_the_iarc_classification and to the class-level entry; only the sucralose null and its within-study character are owned here.)↗
• **Yin S, et al. Artificially Sweetened Beverage Consumption and Cancer Risk: A Comprehensive Dose-Response Meta-Analysis of Prospective Studies. Nutrients. 2022;14(21):4445 — HUMAN OBSERVATIONAL, 14 articles across 17 cohorts; overall cancer, highest vs lowest intake: relative risk 1.03, 95% CI 0.96-1.11, p=0.407. Full text was blocked in verification: only the overall-cancer estimate is carried here, and site-specific results are deferred to the class-level and aspartame entries rather than being restated on sucralose's page.**↗
• National Cancer Institute summary on artificial sweeteners — states that a range of studies have found no evidence that sucralose causes cancer in humans, naming the French cohort among them. (Secondary summary of primary literature; useful as a plain-language anchor, not as a primary source.)↗
• **Soffritti M, et al. 2016, International Journal of Occupational and Environmental Health — RODENT lifespan bioassay: sucralose in feed from prenatal life through lifespan reported to induce haematopoietic neoplasias in male Swiss mice. Assessed and rejected:** EFSA ANS Panel, "Statement on the validity of the conclusions of a mouse carcinogenicity study on sucralose (E 955) performed by the Ramazzini Institute," EFSA Journal. 2017;15(5):4784, on four stated grounds — no dose-response; no known mode of action; failure to meet the Bradford Hill criteria; and treatment to natural death compromising interpretation because tumour incidence rises with age. Carried forward in EFSA Journal 2026;24:e9854, which concluded no safety concern for carcinogenicity or chronic toxicity. Group sizes, doses and tumour incidences are UNVERIFIED — the 2017 statement's full text was blocked and the primary paper was not read. No numbers may be printed from either. Corpus inconsistency flagged: the aspartame entry reads run-to-natural-death designs more favourably than this rejection does.↗
• **EFSA FAF Panel. Re-evaluation of sucralose (E 955) as a food additive and evaluation of a new application on extension of use of sucralose (E 955) in fine bakery wares. EFSA Journal. 2026;24:e9854. doi:10.2903/j.efsa.2026.9854↗, adopted 10 December 2025, published 17 February 2026 — acceptable daily intake of 15 mg/kg body weight/day retained, European exposure below it across all population groups, no safety concern at reported uses and use levels, no safety concern for genotoxicity, carcinogenicity or chronic toxicity; safety of the fine bakery wares extension not concluded, citing uncertainties about potential formation of chlorinated compounds across the wide range of applicable baking processes; formation during home preparation at high temperature, including frying and baking, cannot be excluded; amendments to the specifications recommended. The reference point behind the intake limit: a decrease in body weight in rats — effects on body-weight gain at 0.3% in the diet, reported as roughly 150 mg/kg body weight/day, in the chronic toxicity and carcinogenicity study and in parents and offspring in the reproductive toxicity study, with food-consumption reductions of about ten per cent or less and not dose-related; a no-observed-adverse-effect level of 1500 mg/kg from a 26-week gavage study; an assessment factor of 100. The circulating account that the Panel chased the journal over a complaint about Schiffman et al. 2023 and included the data anyway remains UNVERIFIED.** (Public risk assessor, not an interested party. Primary text not directly accessed — paywalled, open-access mirror access-blocked; the intake limit, exposure conclusion, bakery non-conclusion, home-cooking extension and the reference-point figures are corroborated across independent retrievals with consistent wording. Verify exact phrasing before quoting directly; the reference-point numbers inherit this limitation.)
• **Eisenreich A, Gürtler R, Schäfer B. Heating of food containing sucralose might result in the generation of potentially toxic chlorinated compounds. Food Chemistry. 2020;321:126700. doi:10.1016/j.foodchem.2020.126700↗ — CHEMICAL ANALYSIS IN HEATED MODEL SYSTEMS — no cells, no tissue, no organism at any point; this is not an in-vitro biological study and must not be labelled as one. From the German Federal Institute for Risk Assessment; sucralose showed remarkable instability and discolouration after heating at 85–90 °C for 1 hour; chlorinated furan-3-one and chlorinated dicarbonyl compounds identified for the first time; 3-chlorotyrosine formed in the presence of protein, indicating chlorination of other biomolecules. Re-verified against the primary abstract.** (Identifies compounds formed. No health-outcome data.)
• German Federal Institute for Risk Assessment (BfR), consumer notification on heating sucralose — states that sucralose decomposes when heated above 120 °C for a prolonged period, names polychlorinated dibenzo-p-dioxins, dibenzofurans and chloropropanols among the chlorinated organic compounds that can form, and recommends not heating sucralose-containing foods above 120 °C. Re-verified. (No quantified measurement of dioxin or chloropropanol yield from sucralose in real food was located, so the recommendation may be cited and no figure may ever be attached to it.)↗
• **World Health Organization. Use of non-sugar sweeteners: WHO guideline. Geneva: WHO; 15 May 2023 (executive summary via NCBI Bookshelf NBK592245) — conditional recommendation on low-certainty evidence that non-sugar sweeteners not be used for weight control or to reduce non-communicable disease risk. Two streams, both of which must be reported: randomised trials (low certainty, most ≤3 months) found reduced sugar and energy intake, lower body weight and lower BMI; prospective cohorts (very low to low certainty, ~13 years mean follow-up) found higher intake associated with higher BMI and incident obesity, plus suggested undesirable effects including type 2 diabetes, cardiovascular disease and mortality — human observational, severe reverse-causation risk. No long-term body-fat benefit demonstrated in adults or children. Guidance for people with pre-existing diabetes was out of scope and studies exclusively in that population were excluded. Re-verified against the executive summary.** Publicly criticised in European Journal of Clinical Nutrition.↗
• **Serrano J, Smith KR, Crouch AL, et al. High-dose saccharin supplementation does not induce gut microbiota changes or glucose intolerance in healthy humans and mice. Microbiome. 2021;9:11 — HUMAN RCT plus rodent; double-blind, placebo-controlled, parallel-arm, pure saccharin in capsules twice daily for two weeks at the maximum acceptable intake; null on microbiota, metabolites and glucose tolerance in both species. It predates rather than replicates the 2022 trial, and its design differs — capsules of pure compound rather than oral sachets in a glucose vehicle — which is a testable explanation for the discrepancy rather than a wave-away.** (Promoted by an industry association; its own funding was not verified and it must not be printed bare.)↗
• **Rodriguez-Palacios A, et al. The Artificial Sweetener Splenda Promotes Gut Proteobacteria, Dysbiosis, and Myeloperoxidase Reactivity in Crohn's Disease-Like Ileitis. Inflammatory Bowel Diseases. 2018;24(5):1005-1020 — RODENT, Crohn's-prone mice plus healthy controls, six weeks of Splenda specified as sucralose to maltodextrin at 1:99 by weight; ileitis severity was not increased, but ileal myeloperoxidase activity rose in the disease-prone mice (p<0.022), Proteobacteria expanded in all mice, and E. coli overgrew with increased bacterial infiltration into the ileal lamina propria of the disease-prone animals. Because ninety-nine per cent of the administered mass was maltodextrin, this study cannot attribute its findings to sucralose — it is carried here as the source of the susceptible-population shape and of the Splenda composition point, not as sucralose evidence. Primary receipts for maltodextrin's own effects on mucosal defence were not confirmed and must not be asserted.**↗
• Human inflammatory bowel disease context: a meta-analysis of prospective cohorts associates sugar intake with ulcerative colitis (relative risk 1.59) and Crohn's disease (1.90) while finding no significant association for sugar-sweetened beverages across four estimates; and "Sweets and Inflammatory Bowel Disease: Patients Favor Artificial Sweeteners and Diet Foods/Drinks Over Table Sugar and Consume Less Fruits/Vegetables," Inflammatory Bowel Diseases. 2023;29(11):1751 — the reverse-causation confound observed directly in a named population. The artificially-sweetened-beverage hazard ratio from the 121,490-person cohort (doi 10.1111/apt.17149↗) was paywalled and is not carried.
• One-year human microbiome data: a randomised trial of sweetener use inside a sugar-reduced diet reported a shift toward short-chain-fatty-acid and methane-producing taxa, described by the investigators as a favourable direction. Carried here for that microbiome direction only. Its funding statement includes industry partners and was not verified; its weight and cardiometabolic outcomes belong to artificial_sweeteners_evidence and are not restated here.↗
• Effect of Sucralose Intake on Human and Mouse/Rat Gut Microbiota Composition: A Systematic Review and Meta-Analysis. Food Reviews International. 2024;40(5). doi:10.1080/87559129.2023.2212045↗ — systematic review and meta-analysis; nine studies total, only two human; humans showed higher relative Bacteroidetes abundance, rodents the converse. (Not re-verified beyond citation level.)
• Formation of Chlorinated Carbohydrate Degradation Products and Amino Acids during Heating of Sucralose in Model Systems and Food. Journal of Agricultural and Food Chemistry. doi:10.1021/acs.jafc.4c08059↗ — newer primary chemistry on heated sucralose, plausibly the recent work the 2026 regulatory opinion refers to. Located at title and journal level only — lead, not receipt. Do not quote its numbers.
• Remaining sucralose microbiome leads: two twelve-week trials in Asian Indian adults finding diversity changes in type 2 diabetes but not in overweight or obesity; a ten-week study in healthy young adults (PubMed 35208888, 2022). Confirmed only to title and journal level — leads, not receipts. Do not quote their numbers. (The thirty-day trial formerly on this list has been promoted to a receipt above.)↗
• "The non-synthetic sweeteners, miraculin and mogroside V, but not stevia, disrupt the intestinal epithelial barrier function through a sweet taste receptor-dependent mechanism," Scientific Reports 2025, doi 10.1038/s41598-025-28759-z↗ — reports stevia null and monk fruit's mogroside V positive on barrier function in the same receptor-dependent system. Located through a search summary only; paper not read. Concentrations, tight-junction directions and assay system are UNVERIFIED. Carried as a lead in both directions and must be re-verified before either half is treated as established.
• 21 CFR 172.831 — United States approval of sucralose as a food additive. No responsive statement from the Food and Drug Administration to Schiffman et al. 2023 was located; no pending restriction identified. The accurate statement is that the agency has not publicly revised its position.↗
• Gujral J, Carr J, Tonucci D, Darwen C, Grotz VL (2021) — argues heated sucralose in manufacturing does not pose a risk to human health. Funding and affiliations not verified; promoted by the sweetener trade body as confirmation that sucralose is safe when heated, and it addresses industrial manufacturing heat, which is the scenario the 2026 regulatory opinion explicitly distinguished from domestic baking and frying. Check before citing in either direction.↗
• Funding notation: the load-bearing anchors on the concern side are now three human randomised trials, two of them publicly funded (Mexico's national science council; the United States national institutes of health) and one with funding unverified, none industry-linked. The load-bearing anchor on the reassurance side is a public regulator, joined in this version by a large academic cohort whose sucralose result is a within-study null and by a pooled analysis across seventeen cohorts. The 2023 cell-culture paper is not industry-funded and is limited by concentration and by validity objections rather than by funding — and its senior author is being sued by a manufacturer, which is a fact about the field, not about the science. The published re-analysis of the pairing trial is not industry-funded in origin but its authors declare sweetener-industry support, which is printed rather than hidden; the pure-compound saccharin null is promoted by an industry association with funding unverified; the one-year microbiome data comes from a consortium including industry partners, unverified. No claim of human harm is made anywhere in this entry, because no source supports one.*↗