Sugar Substitutes
"Diet" sweeteners are not biologically inert, and the two dominant "keto" sugar alcohols carry a real cardiovascular signal. The least troubling replacements are stevia for everyday sweetness and allulose for baking, with monk fruit reasonable but barely tested in people; non-nutritive sweeteners are useful for getting off sugar but are not a long-term default.
Aspartame, sucralose, saccharin, and acesulfame-K were licensed on the assumption that anything calorie-free couldn't matter metabolically. That assumption has not held up: they shift the microbiome, carry observational harm signals, and do not produce sustained weight loss on their own. The reconciliation is comparator-dependent — these sweeteners look clearly helpful when they displace sugar-sweetened drinks, and the long-term cohort data raise a caution against water that the short substitution trials are not long enough to settle. Erythritol and xylitol, marketed as the "natural" sugar-free option, are the larger surprise: both associate with major cardiovascular events at the population level. For the caloric sugars these substitutes replace, see Sweeteners.
What the evidence says
Strong (for biological activity; surrogate endpoints only):
- Non-nutritive sweeteners (NNS) are not biologically inert: below the acceptable daily intake (ADI) they alter the human stool and oral microbiome and the plasma metabolome, and germ-free mice given the strongest and weakest responders' microbiomes inherited their donors' glycemic responses[1].
Moderate:
- Non-nutritive sweeteners do not produce sustained weight loss in long-term trials and associate with type 2 diabetes, cardiovascular disease and all-cause mortality in cohort data; the World Health Organization (WHO) issued a 2023 conditional recommendation against their use for weight control or to reduce non-communicable-disease risk[2]. Two limits on that guideline: WHO's own review found no association with cancer overall, and it assessed no dementia or cognitive outcome in adults at all — the stroke and cognition findings below come from individual cohorts, not from WHO.
- Substituting low/no-calorie sweetened beverages for sugar-sweetened beverages produces modest reductions in body weight, body mass index (BMI), body fat, and liver fat in randomized controlled trials (RCTs) — around a kilogram of body weight across the pooled trials[3].
- Erythritol and xylitol — the two dominant "keto" sugar alcohols — associate with major adverse cardiovascular events at the population level and enhance platelet reactivity in mechanistic experiments[4][5].
- Allulose lowers the glucose and insulin rise after a meal in human RCTs — a moderate-sized effect on both, pooled across twelve trials — while leaving HbA1c (average blood sugar over months), fasting glucose, blood lipids and body composition unchanged[6].
Weak / preliminary:
- Non-nutritive sweeteners do not reliably increase sweet-taste preference or total energy intake in the balance of human evidence — contradicting the popular "they make you crave more" claim[7].
- Xylitol may reduce dental caries (~13% versus fluoride-only toothpaste), low certainty[8].
- Allulose, monk fruit, and stevia activation of AMPK / SIRT1 longevity pathways — preclinical.
- Intergenerational microbiome and gene-expression effects of sweeteners — mouse data only.
Caution:
- Erythritol and xylitol used as routine bulk sweeteners in adults with established cardiovascular disease, prior stroke, atrial fibrillation, or otherwise elevated thrombotic risk.
Non-nutritive sweeteners are not inert
The WHO 2023 guideline (Moderate). A systematic review found no long-term body-fat benefit from non-sugar sweeteners and signals of increased risk for type 2 diabetes, cardiovascular disease, and all-cause mortality with habitual use. The certainty rating in the underlying evidence is low overall, and very low to low for the diabetes, cardiovascular and mortality signals specifically — the conditional recommendation against use reflects weighing potential harm against the absence of demonstrated benefit, not strong direct evidence of damage[9]. Three points are easy to miss: WHO frames this as a dietary recommendation, not a toxicological safety assessment, so it does not override the acceptable daily intakes below; the recommendation applies to everyone except people with pre-existing diabetes; and it does not reach the sugar alcohols at all, which fall outside WHO's definition and are handled separately further down this page. WHO also declines to rank individual sweeteners against one another, saying the evidence is insufficient to do so. The 2025–2030 US Dietary Guidelines, released in January 2026, went further in one respect: they group low-calorie non-nutritive sweeteners with added sugars and highly processed foods in a single instruction to limit them[10].
Cohort signals (Moderate). The Framingham Offspring Study found daily artificially-sweetened soft-drink intake associated with roughly threefold higher rates of incident ischemic stroke over a decade of follow-up, and that result held up when the authors adjusted for everything they had. The dementia half did not: it fell to about twofold and could no longer be told apart from chance in the same paper's fully adjusted model[11]. A later pooled analysis of six cohorts — one of them Framingham itself, and with 2,445 cases of dementia against Framingham's few dozen — found no association between artificially sweetened drinks and dementia at all[12].
The Brazilian ELSA-Brasil cohort — 12,772 adults followed across three waves over about 8 years — found the highest intake of low- and no-calorie sweeteners associated with faster decline in global cognition, memory, and verbal fluency. The pattern is only partly graded — on global cognition the middle third of intake was indistinguishable from the lowest — and the effects are small: the largest is about four-hundredths of a standard deviation over eight years. The association also held only in adults under 60, with no measurable effect in those aged 60 and older[13]. And the paper has been corrected twice in under a year, the second time after the authors' own audit led them to rerun the whole analysis from the raw data[14] — so the direction is more trustworthy than any single number in it.
The largest quantitative estimates come from the French NutriNet-Santé cohort, which measures sweetener intake from all dietary sources using repeated 24-hour records with brand-level product data. Comparing higher consumers to non-consumers, total artificial-sweetener intake was associated with about 13% higher overall cancer risk (HR 1.13, 95% CI 1.03–1.25 — the bracketed range is where the true effect most plausibly lies; because it stays above 1.0, chance alone is an unlikely explanation, whereas a range crossing 1.0 would be compatible with no effect), driven by aspartame (~15% higher) and acesulfame-K (~13% higher), with sucralose not significantly associated and breast cancer specifically tied to aspartame (~22% higher; HR 1.22, 1.01–1.48)[15]. The same cohort found about 9% higher cardiovascular disease risk (HR 1.09, 1.01–1.18) — a signal carried mainly by strokes and other cerebrovascular events rather than by heart disease[16] — and, most strongly, roughly 69% higher type 2 diabetes risk (HR 1.69), concentrated in aspartame (about 1.6×) and acesulfame-K (about 1.7×) more than sucralose (about 1.3×)[17].
That diabetes figure is the largest in the literature and should be read as its upper end: three later syntheses covering millions of participants put the pooled association nearer 14–26% higher risk[18][19][20]. All of these are observational, and NutriNet-Santé is a single, predominantly female, health-conscious cohort. Reverse causation — people switching to diet products because of pre-existing metabolic disease — is plausible across all of them and is partially controllable but not eliminable. A UK Biobank replication is the most informative piece of the picture: about 70% of the cardiovascular association ran through newly diagnosed diabetes rather than around it, which is what a shared-pathway or reverse-causation story would look like[21].
Microbiome and glucose tolerance (Strong for biological activity; surrogate endpoints only). A 2022 randomised trial assigned 120 healthy, NNS-avoiding adults to saccharin, sucralose, aspartame, or stevia (or glucose/no-supplement controls) at below-ADI doses for two weeks. All four sweeteners distinctly altered the stool and oral microbiome and the plasma metabolome, and saccharin and sucralose significantly impaired glycemic responses. Transplanting the microbiomes of the strongest and weakest responders into germ-free mice reproduced their donors' glycemic responses — moving this from correlation toward causation, though the transplant deliberately tested the extremes rather than a cross-section, and effects were strongly person-specific[22]. This is the cleanest causal evidence that NNS are biologically active, but the long-term clinical significance is unclear. One specific safety question has been raised and answered since: a 2023 paper reported that sucralose-6-acetate, an impurity formed in manufacturing, damaged DNA and intestinal barrier integrity in cell studies[23]. The European Food Safety Authority (EFSA) reviewed it in February 2026, acknowledged one of the two positive signals but judged it of limited relevance against a negative standard mutagenicity test, and left the sucralose limit unchanged[24].
Intergenerational data. A 2026 mouse study reported that parental sucralose and stevia consumption produced microbiome and gene-expression changes in offspring out to the second (F2) generation[25]. Interesting but mouse-only; transgenerational extrapolation to humans is speculative, and rodent coprophagy complicates microbiome-mediated inheritance specifically.
A candidate mechanism for aspartame. In mice and monkeys, dietary aspartame triggered an insulin release through a vagal reflex, and that insulin drove inflammation in the artery wall and accelerated plaque[26]. This is the aspartame counterpart of the platelet work that makes the erythritol case plausible, and it carries the same limitation: an animal mechanism is a reason to keep looking, not a finding about people.
Aspartame: the 2023 hazard-versus-risk split (Caution / regulatory). In July 2023 the two WHO bodies issued seemingly opposite verdicts on the same day. The International Agency for Research on Cancer (IARC) classified aspartame as "possibly carcinogenic to humans" (Group 2B) on the basis of limited evidence for hepatocellular carcinoma. Simultaneously, the Joint FAO/WHO Expert Committee on Food Additives (JECFA) reaffirmed the acceptable daily intake of 0–40 mg/kg body weight, concluding the evidence of an association with cancer in humans "is not convincing"[27]. The reconciliation is that Group 2B is a hazard classification — on IARC's own list it holds aloe vera whole-leaf extract and pickled vegetables — not a measure of real-world risk at dietary doses; the classification was published in full as IARC Monographs Volume 134 in 2024, unchanged from the 2023 announcement. By JECFA's own illustration, a 70-kg adult drinking diet soda with 200–300 mg of aspartame per can would need more than 9–14 cans a day to exceed the ADI — assuming no aspartame from anything else, which the same WHO page notes is unrealistic, since it also turns up in chewing gum, yogurt, cereal, toothpaste and cough drops.
Acceptable daily intakes. The cohort signals above sit alongside regulatory limits — set by the US Food and Drug Administration (FDA), EFSA, and JECFA — that adult intake surveys consistently fall below. Children are the exception: JECFA's 2026 review estimated that high-consuming children in the EU could exceed the steviol-glycoside limit (a conservative estimate it judged not likely to be a health concern), and projected that children would exceed the monk-fruit limit under the proposed uses and maximum levels. These are toxicological ceilings, not targets:
| Sweetener | ADI | Body / notes |
|---|---|---|
| Aspartame | 40 mg/kg/day | JECFA and EFSA; FDA sets 50 mg/kg/day. EFSA reconfirmed 40 mg/kg/day in September 2026 as part of its re-evaluation of the aspartame-acesulfame salt (E 962), finding no new studies that would move the reference point and no exceedance in any age group[28] |
| Acesulfame-K | 15 mg/kg/day | FDA; EFSA raised its own figure from 9 to 15 in April 2025[29] |
| Sucralose | 5 mg/kg/day (FDA); 15 mg/kg/day (EFSA) | EFSA reconfirmed 15 in February 2026[30] |
| Saccharin | 15 mg/kg/day (FDA); 9 mg/kg/day (EFSA) | The widest disagreement between the two bodies in this table[31] |
| Steviol glycosides | 4 mg/kg/day (steviol equivalents) | EFSA / JECFA. High-consuming children in the EU are estimated to reach up to 4.8 mg/kg/day, which JECFA judged a conservative estimate and not likely a health concern |
| Monk fruit (mogrosides) | 10 mg/kg/day, temporary (as mogroside V) | JECFA set this in June 2026 and estimated that, under the proposed uses and maximum levels, children's exposure would be about four times it — a potential safety concern pending further data. EFSA judged the toxicity package insufficient to conclude on safety in 2019[32]. FDA GRAS (generally recognized as safe) |
| Allulose | GRAS; no ADI | FDA excludes it from "added sugars" labeling; ~0.4 kcal/g. Not authorised in the EU — EFSA could not establish safety on the dossier filed[33] |
Practical reading. Non-nutritive sweeteners are better than the sugar versions they replace if the alternative is a daily sugar-sweetened beverage (SSB), and they're useful as a transitional tool when tapering off sugar — RCTs show modest weight and cardiometabolic benefit when they displace sugar-sweetened drinks[34]. The comparison with plain water is thinner than it sounds: the water arm of the same review rested on three small trials and showed nothing either way, and where the two were tested against each other directly the sweetened drinks came out slightly worse on long-term blood sugar. Two twelve-month randomised trials have since come out mildly in favour of the sweeteners: one found better maintained weight loss against sugar with no adverse cardiometabolic markers[35], and an industry-funded one found sweetened drinks slightly ahead of plain water on maintained weight loss[36]. What no trial can address is the multi-year cohort data on cognitive and vascular endpoints, which runs far longer than any of them — so the case against settling on diet drinks long term rests on that, not on the trial evidence.
Do sweeteners make you crave more sugar?
Weak / preliminary. A common claim is that non-nutritive sweeteners "train" a sweet tooth and drive up later intake. The balance of human evidence does not support it: a 2022 narrative review concluded that NNS do not reliably increase sweet-taste preference or total energy intake versus sugar or water comparators[37], and a randomised crossover trial has since reached the same conclusion for two of the newer sweeteners over two weeks[38]. The notable counterpoint is a brain-imaging trial in which sucralose, compared with sucrose, changed how the brain responded to pictures of food — in the medial frontal and orbitofrontal cortex, and concentrated in females and people with obesity[39]. Two things temper it: the group as a whole showed no difference once covariates were accounted for, and the hypothalamus — the region usually invoked in this argument — was not among the areas that responded. So: no robust population-level signal that NNS increase craving or intake, but appetite responses may be person- and context-specific.
Sugar alcohols: the polyol cardiovascular signal
The most consequential dietary-sweetener finding of the last several years concerns erythritol and xylitol — the two polyols that dominate "keto," "diabetic," and sugar-free product categories.
Erythritol (Moderate). A Cleveland Clinic group analysed a discovery cohort plus a 2,149-adult US replication cohort and an 833-adult European replication cohort. Plasma erythritol in the highest quartile predicted roughly double the three-year risk of major adverse cardiovascular events (MACE — cardiovascular death, heart attack, or stroke) — adjusted HR 1.80 (95% CI 1.18–2.77) in the US cohort and 2.21 (1.20–4.07) in the European cohort.[40] Mechanistically, the same group showed that erythritol enhances platelet reactivity at physiologically achievable plasma concentrations, and that a single sweetener-typical dose (around 30 g) raises plasma levels above the platelet-activation threshold for more than 48 hours. A third cohort has since tested the same question in 4,006 older adults and replicated only part of it: erythritol tracked with heart failure, cardiovascular death and total mortality, but showed no association with heart attack or ischemic stroke — the two events that carry most of the weight in that combined endpoint[41].
One important caveat the headlines often miss. Circulating erythritol is partly endogenous: the body produces it from glucose via the pentose phosphate pathway, and production rises in metabolic stress. The cohort data therefore mixes a biomarker effect (erythritol as a marker of underlying cardiometabolic disease) with a direct-toxicity effect. Genetic studies, which are designed to cut through exactly this kind of confounding, have not settled it: the best-powered of them found no causal effect of erythritol on coronary artery disease[42], while two smaller 2025 analyses reported effects that were statistically positive but very small. The mechanistic platelet work is what makes the dietary-intake interpretation plausible — and is the load-bearing piece for the recommendation, not the hazard ratio alone. Two 2026 results are consistent with the marker reading without settling it: in the two-year POUNDS Lost weight-loss trial, analysed observationally, falls in plasma erythritol (but not in other polyols) tracked small falls in estimated 10-year cardiovascular risk (about 0.16 percentage points per standard deviation at six months)[43], and a meta-analysis of six blood-level studies found about 26% more cardiovascular events at the highest versus lowest polyol levels, with high heterogeneity and suspected publication bias, while the intake studies — of artificial sweeteners generally, not sugar alcohols specifically — gave mixed results, most estimates leaning positive[44]. A 2026 Mendelian randomization paper reporting an erythritol–stroke link uses the same small set of genetic instruments as earlier positive analyses, so it is not independent confirmation.
Xylitol (Moderate). The same group ran the same analysis for xylitol and found about 1.6× higher three-year MACE risk in the highest tertile of plasma xylitol (adjusted HR 1.57, 95% CI 1.12–2.21), with mechanistic platelet-reactivity data again confirming biological plausibility[45]. At the ESC Congress in August 2026 the same group reported the association in two general-population cohorts, the Canadian Longitudinal Study on Aging (57% higher event risk, top vs bottom quartile, six years) and EPIC-Norfolk in England (18% higher, 30 years). It is a conference presentation with no published hazard ratios or confidence intervals, from the same laboratory, so treat it as pending rather than as independent replication[46].
A second, independent line of evidence points the same way. The ELSA-Brasil cognitive-decline cohort described earlier measured seven sweeteners, and three of them were polyols — erythritol, xylitol and sorbitol — each individually associated with faster decline in adults under 60[47]. That is a different population, a different endpoint and a different research group from the cardiovascular work above, which makes it worth more than its small effect size suggests.
No outcome trial yet. Both signals rest on observational cohorts plus mechanistic platelet work and small human challenge studies — an eight-person single-arm erythritol pilot and a ten-person xylitol challenge, the latter run without a placebo arm and with a single blood draw thirty minutes after the dose, which a later editorial in the same journal argued cannot separate the platelet change from ordinary short-term fluctuation[48]. The one randomised trial of sustained intake — 36 g of erythritol or 24 g of xylitol daily for five weeks in 42 adults with obesity — found no effect on arterial stiffness, abdominal fat or glucose tolerance[49]. There is still no randomized controlled trial on hard cardiovascular outcomes for either polyol — the caution is graded on plausibility, not demonstrated harm.
Sorbitol now has a signal of its own. In a nested case-control study within the Nurses' Health Study, the highest quarter of combined mannitol and sorbitol levels carried about 40% higher coronary heart disease risk — and unlike erythritol's association in the same paper, it survived adjustment for diabetes[50]. The assay measures the two compounds together, so this is not yet a sorbitol-specific finding. Maltitol and isomalt have not generated comparable signals.
Practical: treat erythritol- and xylitol-sweetened products (keto baking goods, sugar-free candies, low-carb protein bars) as occasional, not staples — particularly for adults with established cardiovascular disease, prior stroke, atrial fibrillation, or those on antiplatelet therapy. The xylitol gum used in dentistry is small-dose and doesn't reach the plasma levels that drive the platelet effect.
Glycemic and tolerance profile
Polyols are "diabetic-friendly" because they are absorbed slowly and incompletely, so their glycemic indices sit far below sucrose (GI 65). The same incomplete absorption is why they ferment in the colon and cause gas, bloating, and an osmotic laxative effect in bulk; the glycemic indices in the table below come from the same source[51]. EU labelling rules require a laxative warning on any food containing more than 10% added polyols — a single threshold that applies to all of them, with no exemption for erythritol[52]. EFSA refused an industry request for exactly that exemption in 2023, and its exposure assessment found that high consumers — mainly children and adolescents — may already exceed the intake limit it set, while calling its own exposure estimates an overestimation[53].
| Polyol | Glycemic index | Notes |
|---|---|---|
| Erythritol | 0 | Mostly absorbed and excreted unchanged; best gastrointestinal tolerance. EFSA sets an intake limit of 0.5 g/kg body weight per day — about 35 g for a 70-kg adult |
| Mannitol | 0 | Strong laxative effect |
| Lactitol | 6 | — |
| Isomalt | 9 | Better tolerated in bulk |
| Sorbitol | 9 | Common laxative threshold issues |
| Xylitol | 13 | Toxic to dogs |
| Maltitol | 35 | Highest glycemic impact of the polyols |
| (Sucrose, reference) | 65 | — |
Dental health
Non-cariogenic benefit Strong; xylitol's active anti-caries claim Weak / low certainty. Polyols and high-intensity non-nutritive sweeteners are non-cariogenic: oral bacteria cannot ferment them into the acids that demineralise enamel, so they do not promote tooth decay the way sucrose does. Xylitol is widely marketed for an active anti-caries benefit beyond simply not feeding decay; the evidence is weaker than the marketing. A Cochrane review found low-certainty evidence that a 10% xylitol fluoride toothpaste may reduce caries about 13% versus fluoride-only toothpaste over 2.5–3 years, with insufficient evidence for other xylitol products; that 13% rests on two trials from one research group in one population, which the review's own authors flag[54]. Later systematic reviews are equivocal in a specific way: xylitol gum does show a real effect in people at moderate or high caries risk, while xylitol sweets show none[55]. Allulose and the high-intensity sweeteners also do not promote decay — though for the high-intensity sweeteners this rests on the chemistry rather than on trials: no clinical trial has ever measured caries outcomes for sucralose, saccharin or aspartame[56]. Choosing non-cariogenic sweeteners over sugar is genuinely better for teeth; xylitol's added caries-reducing claim is the part to discount.
Stevia and monk fruit
Both are plant-derived, zero-calorie high-intensity sweeteners, and both are reasonable replacements for synthetic NNS when sweetness is wanted without calories — stevia on the strength of a larger record that leans mildly favourable, monk fruit on the strength of a thin one that is mildly favourable. The ordering below is this page's judgement; WHO counts stevia as a non-sugar sweetener and declines to rank individual sweeteners at all.
Stevia (Weak / preliminary). Steviol glycosides, 200–400× sweeter than sucrose. Across four meta-analyses of short-term human trials the direction is mildly favourable rather than merely neutral — small reductions in fasting glucose and blood pressure at low certainty, with HbA1c and body weight unchanged[57]; some animal data suggests mild antioxidant and anti-inflammatory effects. Two qualifications: in laboratory culture it interferes with the chemical signalling bacteria use to coordinate, though this was shown in an engineered strain of an environmental pathogen rather than in gut bacteria[58], which leaves the question of microbiome effects at high chronic intake open rather than answered; and the mouse study cited above included stevia, where the inflammatory changes seen in the first generation had normalised by the second while a different liver signal appeared instead — a different pattern, not a weaker version of the same one. For the average adult, stevia in modest amounts is fine; it should not be assumed perfectly inert.
Monk fruit (Weak / preliminary). Mogrosides, ~150–250× sweeter than sucrose. Mogrosides are largely undigested in the upper gastrointestinal (GI) tract, though mogroside V is absorbed to some extent and is systemically bioavailable[59]; colonic bacteria cleave the glucose moiety and use the rest as a substrate, with possible prebiotic effects. In vitro and animal data suggest mogroside V has antioxidant activity. There is no human cardiovascular, microbiome, or cognitive harm signal to date — but that is because there is almost nothing to find one in. No randomised trial has tested monk fruit on its own in people, on any outcome, and the one human study that includes it uses it blended with stevia over four hours.
Monk fruit carries the fewest findings against it of any high-intensity sweetener — but that is an absence of evidence, not evidence of safety, which is why stevia is the better-supported first choice. The regulatory file is also still open: EFSA judged its toxicity package insufficient to conclude on safety in 2019, with unresolved effects in animal testing it said it could not dismiss[60], and JECFA's limit above is only temporary. Use in small amounts is reasonable; treating it as the demonstrated-safe choice is not.
Rare sugars: allulose and tagatose
Rare sugars are naturally occurring monosaccharides whose structural variations change how the human body handles them. The two with meaningful human data are D-allulose and D-tagatose. Both are technically caloric (~0.4 kcal/g for allulose, ~1.5 kcal/g for tagatose) but mostly not metabolized for energy.
Allulose (Moderate). Roughly 70% is absorbed in the small intestine and excreted unchanged in urine; the remaining 30% reaches the colon and resists fermentation — FDA's own summary of the evidence, and the reason it treats allulose as contributing almost no calories[61]. In human RCTs, pre-meal allulose lowers postprandial glucose and insulin, particularly with carbohydrate-heavy meals, without moving HbA1c, fasting glucose, lipids or body composition[62]. A 25 g dose, given by nasogastric tube to 18 lean young adults, does trigger gut release of glucagon-like peptide-1 (GLP-1) in people, alongside cholecystokinin and peptide YY[63]. The rest of the mechanism is rodent: in mice and rats that GLP-1 release activates vagal afferents signalling satiety to the brainstem and hypothalamus, and vagotomy or loss of the GLP-1 receptor blunts — the paper's own word — the anti-obesity and glucose-lowering effects[64]. The vagal step has not been shown in people.
Allulose is not a "natural Ozempic": the GLP-1 signal it produces is tiny beside a dose of a prescription GLP-1 receptor agonist. It is a useful sugar substitute that doesn't raise blood glucose, and nothing more.
Tagatose has FDA GRAS status and a low glycemic index. Pooled small trials show a modest HbA1c improvement — about a quarter of a percentage point — but no effect on blood lipids; the placebo-controlled trial states outright that it did not lower triglycerides or raise HDL[65]. No safety signals beyond the gut: in one pilot, ten of twelve participants reported gastrointestinal side effects.
Practical. For an adult who wants a sugar substitute that bakes and tastes close to sucrose without raising glucose, allulose is currently the best-evidenced option. The gating factors are cost and availability; GI tolerability (loose stools, gas) is the most common limiting issue — severe diarrhoea appeared at 0.5 g per kilogram of body weight in a single dose, and the study's authors put their recommended ceiling below that, at 0.4 g/kg in one sitting and 0.9 g/kg across a day[66].
Chicory syrup and fibre syrups
"Chicory syrup" names two products with almost nothing in common, and the label is the only way to tell them apart.
Fibre syrups (oligofructose / FOS). Chicory root stores inulin, a fructose polymer with chains of up to about 60 units. Partial enzymatic hydrolysis cuts it into short chains (degree of polymerisation about 2–8) called oligofructose or fructo-oligosaccharides (FOS). A pure commercial grade is about 95% oligofructose and 5% free glucose, fructose and sucrose on a dry basis, about 35% as sweet as sucrose[67]; manufacturers also sell lower-purity syrup grades with more free sugar and correspondingly more sweetness. Because the fructans escape small-intestinal digestion and are fermented in the colon, their energy value is modelled at 0–2.5 kcal/g, with 1.5 kcal/g recommended for labelling (EU labels use the general fibre value of 2 kcal/g)[68]. Retail fibre syrups typically declare about 160–180 kcal, 55–70 g fibre and 6–15 g sugars per 100 g, and those figures reconcile with 4 kcal/g for sugars plus 2 kcal/g for fibre.
Fructose syrups ("inulin syrup"). Full hydrolysis of inulin yields mostly fructose. EU law defines "inulin syrup" as the immediate product of hydrolysing inulin or oligofructose, with at least 10% fructose (free or as sucrose) in the dry state; syrups above 50% fructose fall under a separate tariff subheading (CN 1702 60 80). Because chicory inulin is chains of fructose capped by a single glucose, full hydrolysis is mostly fructose: a US Customs ruling on a product sold as "chicory syrup" and "chicory honey" records the importer's description of solids as 85% fructose and 15% glucose[69], and a 2004 ruling on another chicory fructose syrup gives 83% fructose and 12% glucose. Nutritionally that is a free-fructose syrup, and everything on the Sweeteners page about fructose dose applies. Free fructose is absorbed better when glucose accompanies it: abnormal breath tests occurred in 68% of people after fructose alone against 26% after high-fructose corn syrup[70], and more than half of 15 healthy adults malabsorbed a 25 g fructose dose[71]. An 85:15 syrup should therefore be harder on the gut than corn syrup — an inference from these studies, not a direct test. Fructose does lower the immediate glucose and insulin peak when it replaces glucose or sucrose[72], which is how such syrups get sold as "diabetic-friendly"; that acute effect says nothing about the load on the liver.
Reading the label. Per 100 g: roughly 55–70 g fibre and under ~15 g sugars is a fibre syrup; 60 g or more of sugars with little or no fibre is a fructose syrup, whatever the front of the pack says.
Glycaemic and metabolic effects of the fibre type. In two double-blind randomised crossover trials in about 40 healthy adults each, replacing 20% of the sugar in a yogurt drink with oligofructose lowered the glucose response by about 14%, and replacing 30% of the sugar in a fruit jelly with inulin lowered it by about 16%, with insulin falling in parallel and the reduction tracking the share of sugar replaced. Authors from a chicory-fibre manufacturer co-wrote the paper[73]. Longer term, a pooled analysis of 55 randomised trials (2,518 participants) found inulin-type fructans lowered body weight by about 1 kg (low certainty), LDL cholesterol by 0.14 mmol/L (very low certainty) and triglycerides by 0.06 mmol/L (low certainty), with little or no effect on fasting glucose, HbA1c or blood pressure; effects were larger with at least six weeks' use and in people with overweight[74]. In prediabetes and type 2 diabetes the picture is better: 33 trials (1,346 people) found fasting glucose 0.60 mmol/L and HbA1c 0.58 percentage points lower, at about 10 g a day for six weeks or more[75]. The best single weight trial gave 48 overweight adults 21 g of oligofructose or maltodextrin daily for 12 weeks; the oligofructose group lost about 1 kg while the placebo group gained about 0.5 kg, with lower ghrelin, higher PYY and lower self-reported intake[76] — though 21 g a day is more than most people tolerate from a syrup.
Microbiome. The bifidogenic effect is robust: across 50 studies (2,525 participants) using 3–20 g a day of chicory fructans, Bifidobacterium abundance rose substantially (standardised mean difference 0.83), in a review with industry-affiliated authors[77]. Whether more bifidobacteria translates into a health outcome in healthy adults is not established. The EU-authorised bowel-function claim ("contributes to normal bowel function by increasing stool frequency", Regulation (EU) 2015/2314) applies to native long-chain chicory inulin at 12 g a day or more, not to the short-chain oligofructose in syrups.
Tolerance is the binding constraint. In single doses, up to 10 g of native inulin and 5 g of oligofructose were well tolerated, but 10 g of oligofructose clearly raised flatulence and bloating[78]. Over a week of daily short-chain FOS, excess flatus appeared at 20 g, and 10 g was the well-tolerated dose that still raised bifidobacteria[79]. A 15 g serving of fibre syrup carries about 9–10 g of FOS, so the syrup can replace only a few teaspoons of sugar a day. Fructans are FODMAPs: in a placebo-controlled rechallenge in IBS patients with fructose malabsorption, symptoms were inadequately controlled in 77% on fructans against 14% on glucose (the 25 patients had already responded to a low-FODMAP diet)[80], and a meta-analysis of prebiotics in IBS found no overall benefit, with inulin-type fructans worsening flatulence[81].
Safety signals. In mice with disturbed gut bacteria, and in ordinary mice on an inulin-enriched high-fat diet, dietary inulin caused microbiota-dependent cholestatic liver cancer; germ-free status, antibiotics or a bile-acid binder prevented it[82]. No human counterpart has been reported. It argues for treating refined inulin as a food ingredient rather than something to take in escalating doses. Inulin anaphylaxis exists but is confined to case reports. For teeth, oral streptococci ferment FOS in vitro, with plaque formation similar to sucrose[83]; there are no human caries trials, so the fibre type should not be assumed tooth-safe, and the fructose type is simply a fermentable sugar.
Practical. A fibre syrup is a reasonable way to cut some sugar in small amounts (at most about a tablespoon a day, split across meals, since a tablespoon holds roughly the 10 g single dose that raised symptoms), but not if you have IBS. A fructose-type chicory syrup is sugar and belongs on the same limit as honey and corn syrup.
Practical guidance
- Default beverages: water, coffee, tea, sparkling water. Unsweetened. More impactful over a lifetime than choosing between non-nutritive alternatives.
- Use NNS as a tool, not a default. A short-term taper from full-sugar SSBs to diet versions is a reasonable use — RCTs show real benefit when they displace sugar, though not that they match plain water. The reason not to settle there is the multi-year cohort data on cognitive and vascular endpoints, which no trial has run long enough to test.
- For zero-calorie sweetness when wanted: stevia first, then monk fruit, over synthetic NNS, in modest amounts. Stevia has the larger and mildly favourable human record; monk fruit's is close to empty, and its regulatory file is still open.
- For sugar substitutes in baking: allulose if cost permits, keeping single servings under about 0.4 g per kilogram of body weight; otherwise small amounts of regular sugar in an otherwise Mediterranean-pattern diet.
- Avoid erythritol and xylitol as routine bulk sweeteners if you have established cardiovascular disease, atrial fibrillation, prior stroke, or are on antiplatelet therapy.
- Mind polyol tolerance. Sugar-free candies and keto bars in bulk cause gas and an osmotic laxative effect; erythritol is the best-tolerated polyol, maltitol the worst on both glycemic and GI grounds.
- Check what a "chicory syrup" actually is. Mostly fibre: acceptable in small amounts (at most about a tablespoon a day, split across meals; avoid with IBS). Mostly sugars: it is a fructose syrup — count it as added sugar.
What's overrated
- "Allulose is a longevity mimetic." Based on AMPK/SIRT1 activation in adipocytes and C. elegans lifespan extension. Real preclinically; not a human longevity claim.
- "Stevia is biologically inert because it's natural." It alters bacterial signalling in culture and showed intergenerational signals in mice. Probably fine; not zero.
- "Monk fruit is the proven-safe choice." It has the fewest findings against it because it is the least studied in people; its regulatory file is still open.
- "Diet sodas are safe replacements." Better than SSBs in the short term, and the stroke signal in long-term cohort data is hard to dismiss. The dementia signal is weaker than it is usually reported to be — the largest pooled analysis finds none.
- "Erythritol and xylitol are the safe natural choice." The "natural sugar alcohol" framing is exactly backwards on cardiovascular grounds — these are the substitutes with the clearest harm signal.
- "Xylitol prevents cavities." Non-cariogenic, yes; as an active caries-reducing agent, only weakly, at low certainty, and mainly as gum in people at raised risk.
- "Chicory syrup is a natural, diabetic-friendly fibre sweetener." True of some products and false of others that are 85% fructose. The fibre kind is prebiotic but gassy above about 10 g; the sugar kind is sugar.
- "Sweeteners make you crave more sugar." The balance of human trials does not support a reliable increase in sweet-taste preference or intake.
Further reading
- World Health Organization. Use of non-sugar sweeteners: WHO guideline. 2023.[84]
- WHO / IARC / JECFA. Aspartame hazard and risk assessment results released. 2023.[85]
- Debras C et al. Artificial sweeteners and cancer risk: results from the NutriNet-Santé population-based cohort study. PLOS Medicine 2022.[86]
- Debras C et al. Artificial sweeteners and risk of cardiovascular diseases: results from the prospective NutriNet-Santé cohort. BMJ 2022.[87]
- Debras C et al. Artificial sweeteners and risk of type 2 diabetes in the prospective NutriNet-Santé cohort. Diabetes Care 2023.[88]
- Suez J et al. Personalized microbiome-driven effects of non-nutritive sweeteners on human glucose tolerance. Cell 2022.[89]
- McGlynn ND et al. Association of low- and no-calorie sweetened beverages as a replacement for sugar-sweetened beverages with body weight and cardiometabolic risk: a systematic review and meta-analysis. JAMA Network Open 2022.[90]
- Witkowski M et al. The artificial sweetener erythritol and cardiovascular event risk. Nature Medicine 2023.[91]
- Witkowski M et al. Xylitol is prothrombotic and associated with cardiovascular risk. European Heart Journal 2024.[92]
- Gonçalves NG et al. Association between consumption of low- and no-calorie artificial sweeteners and cognitive decline: an 8-year prospective study (ELSA-Brasil). Neurology 2025.[93]
- Pase MP et al. Sugar- and artificially sweetened beverages and the risks of incident stroke and dementia. Stroke 2017.[94]
- Chen H et al. Sweetened beverages and incident all-cause dementia among older adults. JAMA Psychiatry 2025.[95]
- Abushamat LA et al. Erythritol, erythronate, and cardiovascular outcomes in older adults in the ARIC study. JACC Advances 2025.[96]
- Heianza Y et al. Plasma levels of polyols erythritol, mannitol, and sorbitol and incident coronary heart disease. European Journal of Preventive Cardiology 2025.[97]
- Younes M et al. Re-evaluation of erythritol (E 968) as a food additive. EFSA Journal 2023.[98]
- Osborn L et al. Glycemic and cardiometabolic effects of rare sugars allulose and tagatose: a systematic review and meta-analysis. American Journal of Clinical Nutrition 2026.[99]
- Pang MD et al. Effect of sweeteners and sweetness enhancers on weight management and gut microbiota composition. Nature Metabolism 2025.[100]
- Iwasaki Y et al. GLP-1 release and vagal afferent activation mediate the beneficial metabolic and chronotherapeutic effects of D-allulose. Nature Communications 2018.[101]
- Lightowler H et al. Replacement of glycaemic carbohydrates by inulin-type fructans from chicory (oligofructose, inulin) reduces the postprandial blood glucose and insulin response to foods: report of two double-blind, randomized, controlled trials. European Journal of Nutrition 2018.[102]
- Talukdar JR et al. The effects of inulin-type fructans on cardiovascular disease risk factors: systematic review and meta-analysis of randomized controlled trials. American Journal of Clinical Nutrition 2024.[103]
- Bonnema AL et al. Gastrointestinal tolerance of chicory inulin products. Journal of the American Dietetic Association 2010.[104]
- Riley P et al. Xylitol-containing products for preventing dental caries in children and adults. Cochrane Database of Systematic Reviews 2015.[105]
- Livesey G. Health potential of polyols as sugar replacers, with emphasis on low glycaemic properties. Nutrition Research Reviews 2003.[106]
- Wilk K et al. The effect of artificial sweeteners use on sweet taste perception and weight loss efficacy (review). Nutrients 2022.[107]
- Concha Celume F et al. Artificial and natural non-nutritive sweeteners drive divergent gut and genetic responses across generations. Frontiers in Nutrition 2026.[108]
- Yunker AG et al. Obesity and sex-related associations with differential effects of sucralose vs sucrose on appetite and reward processing. JAMA Network Open 2021.[109]