Sugar Substitutes

"Diet" sweeteners are not biologically inert, and the two dominant "keto" sugar alcohols carry a real cardiovascular signal. The cleanest replacements for routine use are monk fruit and allulose; 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 helpful when they displace sugar-sweetened drinks and worse than water over a lifetime. 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:

Non-nutritive sweeteners are not biologically inert: at below-ADI doses they distinctly alter the human stool and oral microbiome and the plasma metabolome, and germ-free-mouse transplant reproduces the donors' altered glycemic responses — moving this from correlation toward causation^[1]00919-9).

Moderate:

Non-nutritive sweeteners do not produce sustained weight loss in long-term trials and associate with type 2 diabetes, cardiovascular disease, stroke, dementia, and all-cause mortality in cohort data; WHO 2023 issued a conditional recommendation against their use for weight control or to reduce non-communicable-disease risk^[2].
Substituting low/no-calorie sweetened beverages for sugar-sweetened beverages produces modest reductions in body weight, BMI, body fat, and liver fat in RCTs — benefits comparable to switching to water^[3]. The cohort harm signals and the trial benefits are reconciled by what NNS replace: helpful versus sugar, not versus water.
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 postprandial glucose and insulin in human RCTs by stimulating endogenous GLP-1 release.

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^[6]^[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

Aspartame, sucralose, saccharin, and acesulfame-K were originally licensed on the assumption that anything that didn't deliver calories couldn't matter metabolically. That assumption has not held up.

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 — the conditional recommendation against use reflects weighing potential harm against the absence of demonstrated benefit, not strong direct evidence of damage^[9]. Two points are easy to miss: WHO frames this as a dietary recommendation, not a toxicological safety assessment, so it does not override the JECFA acceptable daily intakes below; and the recommendation applies to everyone except people with pre-existing diabetes.

Cohort signals (Moderate). The Framingham Offspring Study found daily artificially-sweetened soft-drink intake associated with roughly threefold higher rates of incident ischemic stroke and Alzheimer's dementia over a decade of follow-up.^[10] The 2024 Brazilian ELSA-Brasil cohort — 12,772 adults followed for about 8 years — reported that the highest intake of low- and no-calorie sweeteners was associated with measurably faster declines in global cognition, memory, and verbal fluency, equivalent to about 1.3 years of accelerated cognitive aging across the study period.^[11]

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 sweeteners associated with a higher rate of overall cancer (HR 1.13, 95% CI 1.03–1.25; aspartame 1.15, acesulfame-K 1.13; sucralose not significantly associated), with breast cancer specifically tied to aspartame (HR 1.22, 1.01–1.48)^[12]; with overall cardiovascular disease (HR 1.09, 1.01–1.18)^[13]; and most strongly with type 2 diabetes (total HR 1.69, 1.45–1.97; aspartame 1.63, acesulfame-K 1.70, sucralose 1.34)^[14]. All of these are observational from 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.

Microbiome and glucose tolerance (Strong). A 2022 Cell RCT randomized 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 human microbiomes into germ-free mice reproduced the donors' glycemic responses — moving this from correlation toward causation — though effects were strongly person-specific^[15]00919-9). This is the cleanest causal evidence that NNS are biologically active, but the long-term clinical significance is unclear.

Intergenerational data. A 2026 Frontiers in Nutrition mouse study reported that parental sucralose and stevia consumption produced microbiome and gene-expression changes in offspring out to F2. Interesting but mouse-only; transgenerational extrapolation to humans is speculative, and rodent coprophagy complicates microbiome-mediated inheritance specifically.

Aspartame: the 2023 IARC/JECFA split (Caution / regulatory). In July 2023 the two WHO bodies issued seemingly opposite verdicts on the same day. IARC classified aspartame as "possibly carcinogenic to humans" (Group 2B) on the basis of limited evidence for hepatocellular carcinoma. Simultaneously, 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"^[16]. The reconciliation is that Group 2B is a hazard classification — the same category as aloe vera whole-leaf extract and pickled vegetables — not a measure of real-world risk at dietary doses. 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.

Acceptable daily intakes. The cohort signals above sit alongside regulatory limits that population intake surveys consistently fall below. These are toxicological ceilings, not targets:

Sweetener	ADI	Body / notes
Aspartame	40 mg/kg/day	JECFA and EFSA; FDA sets 50 mg/kg/day
Acesulfame-K	15 mg/kg/day	FDA / EFSA
Sucralose	5 mg/kg/day (FDA); 15 mg/kg/day (EFSA)	—
Saccharin	15 mg/kg/day	FDA
Steviol glycosides	4 mg/kg/day (steviol equivalents)	EFSA / JECFA
Monk fruit (mogrosides)	No numerical ADI	FDA GRAS; EFSA set no ADI as none was deemed necessary
Allulose	GRAS; no ADI	FDA exempts from "added sugars" labeling; ~0.4 kcal/g

Practical reading. Non-nutritive sweeteners are better than the sugar versions they replace if the alternative is a daily 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, comparable to switching to water^[17]. They are not a long-term default; water, coffee, tea, and sparkling water are. The apparent conflict between the WHO guideline and these trials is mostly a question of the comparator: NNS look harmful in cohorts of habitual high users (confounded by reverse causation) and helpful in trials where they replace sugar.

Do sweeteners make you crave more sugar?

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 Nutrients review and a 2024 Frontiers in Nutrition meta-analysis both concluded that NNS do not reliably increase sweet-taste preference or total energy intake versus sugar or water comparators^[18]^[19]. The notable counterpoint is a brain-imaging trial in which sucralose, compared with sucrose, increased hypothalamic blood flow and self-reported hunger — an effect concentrated in females and people with obesity^[20]. The honest read: 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. A 2023 Nature Medicine paper from the 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 cardiovascular events (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.^[21] 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.

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. 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.

Xylitol. A 2024 European Heart Journal paper from the same group ran the same analysis and found ~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^[22].

No outcome trial yet. Both signals rest on observational cohorts plus mechanistic platelet work and small human challenge studies (a single-arm erythritol pilot and a 10-person xylitol challenge). There is no randomized controlled trial on hard cardiovascular outcomes for either polyol — the caution is graded on plausibility, not demonstrated harm.

Sorbitol is less studied, but the same metabolic family. 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^[23]. EU labelling rules require a laxative warning above roughly 10 g/100 g for sorbitol, maltitol, mannitol, and xylitol (the better-tolerated erythritol and isomalt trigger it nearer 25 g/100 g).

Polyol	Glycemic index	Notes
Erythritol	0	Mostly absorbed and excreted unchanged; best GI tolerance
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

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 (gums, lozenges); later systematic reviews remain equivocal^[24]. Allulose and the high-intensity sweeteners also do not promote decay. The practical reading: choosing non-cariogenic sweeteners over sugar is genuinely better for teeth, but xylitol's added caries-reducing claim is modest and low-certainty.

Stevia and monk fruit

Both are plant-derived, zero-calorie high-intensity sweeteners. They are reasonable replacements for synthetic NNS when sweetness is wanted without calories, with monk fruit currently the cleaner profile.

Stevia (steviol glycosides, ~200–300× sweeter than sucrose). In short-term human studies, no consistent harm signal on glucose, blood pressure, or weight; some animal data suggests mild antioxidant and anti-inflammatory effects. Two qualifications: it has been shown to interfere with bacterial quorum sensing, raising the question of microbiome effects at high chronic intake; and the 2026 Frontiers in Nutrition mouse study (cited above) included stevia and found milder but nonzero intergenerational signals. For the average adult, stevia in modest amounts is fine; it should not be assumed perfectly inert.

Monk fruit (mogrosides, ~150–250× sweeter than sucrose). Mogrosides pass intact through the upper GI tract; 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 of meaningful size to date — though the human evidence base is thinner than for sucralose or aspartame, and "no signal yet" is not the same as "demonstrated safe at high lifetime intake."

For the healthy adult choosing between high-intensity sweeteners, monk fruit is currently the most defensible default.

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. Roughly 70% is absorbed in the small intestine and excreted unchanged in urine; the remaining 30% reaches the colon and resists fermentation. In human RCTs, pre-meal allulose lowers postprandial glucose and insulin, particularly with carbohydrate-heavy meals. The mechanism is endogenous GLP-1 release: allulose triggers GLP-1 secretion in the gut, activating vagal afferents that signal satiety to the brainstem and hypothalamus^[25]. In animal models, GLP-1-receptor knockout or vagotomy abolishes the anti-obesity and glucose-lowering effects, confirming the GLP-1–vagal axis is doing the work.

A practical caveat: allulose is not a substitute for prescription GLP-1 receptor agonists. The endogenous GLP-1 signal it produces is small in absolute terms compared to a semaglutide dose. Frame it as a useful sugar substitute that doesn't raise blood glucose, not as a "natural Ozempic."

The longevity-pathway claims sometimes attached to allulose — AMPK and SIRT1 activation in adipocytes, C. elegans lifespan extension — are real preclinical findings but should not be over-interpreted; they are not a basis for human longevity recommendations.

Tagatose has FDA GRAS status and a low glycemic index. Small trials show modest HbA1c improvements and lipid changes; no significant safety signals.

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 at >0.5 g/kg per single dose.

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 benefit comparable to switching to water when they displace sugar. Multi-year cohort data on cognitive and vascular endpoints is worth taking seriously, so don't settle there long term.
For zero-calorie sweetness when wanted: monk fruit > stevia > NNS.
For sugar substitutes in baking: allulose if cost permits; 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. The xylitol gum used dentally is too small-dose to matter.
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.

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 quorum sensing and showed mild intergenerational signals in mice. Probably fine; not zero.
"Diet sodas are safe replacements." Better than SSBs in the short term; the long-term cohort data on stroke and dementia is hard to dismiss.
"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; an active caries-reducing agent worth paying for, only weakly and at low certainty.
"Sweeteners make you crave more sugar." The balance of human trials does not support a reliable increase in sweet-taste preference or intake.
Transgenerational sweetener inheritance from mouse studies as a near-term human concern. Worth tracking; not currently a basis for recommendations.