Xylitol and cardiovascular risk

What the 2024 study measured

Witkowski and colleagues at the Cleveland Clinic ran untargeted metabolomics on overnight fasting plasma from a discovery cohort of 1,157 patients undergoing elective diagnostic cardiac evaluation, and found that a polyol tentatively identified as xylitol tracked with three-year major adverse cardiovascular events. They then built a stable-isotope dilution LC-MS/MS assay able to separate xylitol from its structural isomers and applied it to an independent cohort of 2,149 similar patients.¹

In that validation cohort the top tertile carried an adjusted hazard ratio of 1.57 (95 percent confidence interval 1.12 to 2.21, P < .01) against the bottom tertile, after adjustment for traditional risk factors. The isomer-specific assay matters, because pentitols such as arabitol and ribitol are chemically close and are produced by human metabolism and by microbes; earlier polyol epidemiology has not always separated them.

The population is worth holding in view. Both cohorts were people arriving for cardiac evaluation: 76 to 78 percent had a history of cardiovascular disease, 22 percent had diabetes, around 70 percent had hypertension, and roughly three quarters were on aspirin. This is a high-risk referral population, not a cross-section of the public.

The ingestion experiment, and how fast xylitol clears

Ten healthy volunteers drank water containing 30 grams of xylitol, an amount the paper compares to a xylitol-sweetened baked good or several sweets. At baseline, after an overnight fast, median plasma xylitol was 0.30 micromolar. Thirty minutes after the drink it was 312 micromolar, a thousandfold increase. By four hours it had fallen to 1.87 micromolar and by six hours to 0.67 micromolar. The reported plasma half-life was about 13 minutes, plus or minus 4.¹

Two things follow from that curve. The first is that oral xylitol does reach the bloodstream, which nobody disputes; only about half of an oral dose is absorbed in the small intestine, with the rest passing to the colon.² The second is that the exposure is a spike, not a plateau. Restoring an elevated concentration requires continual redosing.

How much of an oral dose is absorbed at all is contested. A double-blind dose-ranging study in seventeen participants gave 7, 17 or 35 grams of xylitol and reported very low absorption across the whole range.³ The 2024 authors dispute that study's analytical method on isomer-specificity grounds, which is a fair objection to raise and an unresolved one.

Why the authors attribute their own signal to endogenous production

This is the part of the paper that most coverage omits, and it comes from the paper itself. The authors write that the associations observed in the discovery and validation cohorts reflect endogenous xylitol levels, a steady state between production and excretion, and not recent dietary exposure. Their reasoning is the clearance data above: with a half-life of roughly a quarter of an hour and a return to near-baseline within six hours, a sample drawn after more than twelve hours of fasting cannot be reporting on the previous day's food.

Xylitol is produced in human cells as an intermediate of the glucuronic acid pathway, an alternative route of glucose utilization. The 2024 paper puts endogenous output at around 15 grams a day; a 2025 review in Cardiovascular Research gives a range of 5 to 15 grams.^1,2 Either figure exceeds any plausible oral-care exposure by two to three orders of magnitude.

A correspondence from oral-health researchers at Washington and Turku pushed this further, arguing that the pentose phosphate pathway is upregulated during ischaemia-reperfusion, so elevated fasting xylitol may be a benign marker of metabolic dysregulation in people who already have vascular disease. Their suggested test is a good one: check whether xylose, xylulose, glucuronate and other upstream metabolites carry the same association.⁴ The reply from Witkowski and Hazen concedes that related intermediates may well co-associate, while noting that co-association does not rule out a causal contribution from any single one.⁵

The platelet and thrombosis findings

Xylitol at 30 micromolar had no effect on platelet-rich plasma on its own. Added alongside a submaximal dose of a known agonist, adenosine diphosphate or the thrombin receptor activator peptide TRAP6, it markedly enhanced the response, with a dose-dependent effect from the fasting range up to post-prandial concentrations. In animal models of thrombosis, plasma xylitol as low as 19 micromolar increased thrombosis potential. Blood drawn from the ten volunteers thirty minutes after their drink showed enhanced aggregation in every subject.¹ In their later reply the authors report effects from 10 micromolar.⁵

This mechanistic work is what lifts the paper above a metabolite association study, and it is the part that an accompanying editorial in the same issue engaged with.⁶

The 2026 conference analysis

Presented at the ESC Congress on 29 August 2026, this analysis covers 17,710 participants across the Canadian Longitudinal Study on Aging and EPIC-Norfolk, with more than 8,600 events. In the Canadian cohort the highest quartile of blood xylitol carried a 57 percent higher risk of death, myocardial infarction or stroke over six years than the lowest. In EPIC-Norfolk, followed for thirty years, the difference was 18 percent. Adjustments covered age, sex, smoking, lipids, diabetes and hypertension.⁸

Moving from cardiac-clinic patients to general-population cohorts is a real methodological gain, and the size is substantial. Four caveats sit against it. It has not been published or peer reviewed, so the methods available are a press release and a conference abstract. Dietary xylitol was again not recorded. The concentrations defining each quartile have not been reported, which makes it impossible to place the top group against the 2024 tertiles or against the concentrations at which platelet effects were observed.

The fourth concerns the samples. Neither cohort collects fasting blood: the CLSA protocol describes approximately 60 mL of non-fasting blood per participant, and EPIC-Norfolk's baseline samples from 1993 to 1997 were non-fasting.^9,10 The abstract does not state which samples were assayed. If they are the standard baseline draws, then the 2024 paper's central argument for interpreting its levels as endogenous, an enforced twelve-hour fast, does not apply here, and the measurement becomes a mixture of endogenous production and whatever was recently consumed. That makes the 2026 result harder to attribute in either direction, not easier.

Putting the exposures in order

Almost all of the confusion in this topic dissolves once the doses are written next to each other. The doses studied for effects in the mouth are covered in xylitol and oral health. The European Commission's Scientific Committee on Consumer Safety models adult toothpaste exposure at 2.75 g of paste applied per day with a retention factor of 0.05, giving 0.138 g of paste actually ingested daily.¹¹ Multiply that by the xylitol fraction of a formula and the answer is hundredths of a gram. The same chain of quantity, swallowed fraction and systemic dose is worked through for fluoride in how much toothpaste to use.

The gap between the two ends of that table is roughly a thousandfold. A toothpaste contribution of a few hundredths of a gram, spread across two brushings and cleared with a thirteen-minute half-life, is not a plausible mechanism for a raised twelve-hour fasting level. Products designed to dissolve slowly in the mouth are a genuinely different case, since essentially all of the dose is swallowed and the delivery is deliberately prolonged.

The authors of the 2024 paper draw the line in the same place, and said so directly. In their published reply they wrote that they would speculate the use of xylitol in swish-and-spit oral-care products should be safe from a cardiovascular perspective because the amount ingested would be minimal, while arguing that xylitol's use as a one-for-one sugar replacement in processed foods is where the exposure lies.⁵ They also argued in the same reply that the 5 to 6 g used for cavity prevention would exceed the concentrations at which they observed platelet effects, so the lozenge and gum question is contested rather than settled.

Confounding that has not been ruled out

Disease driving the metabolite. Chronic hyperglycaemia recruits alternative glucose-handling routes, including the polyol and pentose phosphate pathways, which would raise polyol levels in exactly the people who already carry elevated cardiovascular risk.² The 2024 analyses adjusted for diabetes as a diagnosis, which is not the same as adjusting for the metabolic state.

Renal clearance. Xylitol is excreted renally. The 2024 paper notes its validation cohort had largely preserved renal function, with a median eGFR above 90. An aging general-population cohort will not, and declining kidney function raises circulating metabolites while independently predicting cardiovascular events. Whether the 2026 analysis adjusted for eGFR is not stated in the material released so far.

Confounding by indication. The heaviest users of xylitol lozenges are people with xerostomia, which is most often caused by medication or by a systemic condition. Anticholinergics, diuretics, antihypertensives and antidepressants all reduce salivary flow, and the conditions they treat carry cardiovascular risk of their own. A cohort that recorded xylitol intake without recording why people were taking it could reproduce this association with no causal contribution from xylitol at all. No study has yet tested this, so it is a hypothesis, not a finding.

The oral-disease pathway runs the other way. Caries and periodontal disease are themselves associated with incident stroke, coronary heart disease and death.¹³ Correspondents used this to argue that reducing oral disease should reduce cardiovascular risk; Witkowski and Hazen replied that improving a risk marker does not reliably improve outcomes, citing diabetes drugs that lowered HbA1c without lowering cardiovascular events.⁵ Both points stand. The direction of the net effect is unknown.

A balanced interpretation

Well supported: higher fasting plasma xylitol is associated with three-year cardiovascular events in cardiac-referral cohorts; higher blood xylitol is associated with events in two general-population cohorts, pending publication; xylitol at micromolar concentrations enhances agonist-stimulated platelet reactivity and increases thrombosis in animal models; a 30 g oral dose raises plasma xylitol roughly a thousandfold for a few hours.

Not established: that consuming xylitol causes cardiovascular events. No study has measured intake. The clinical work is associative, the mechanistic work uses concentrations at or above the top of the observed human range, and the human intervention arm was ten people without a placebo group. Whether ordinary consumption sustains a raised concentration long enough to matter is the open question, and the paper's own clearance data make it a hard case to argue.

What would settle it: a cohort with recorded xylitol intake alongside blood levels; repeated rather than single baseline measurements; a Mendelian randomization analysis for xylitol specifically, of the kind already performed for erythritol; and a placebo-controlled crossover platelet study at realistic doses with a full time course. Until then the honest statement is that people with more xylitol in their blood have more cardiovascular events, that fasting levels appear to be largely self-produced, and that very high concentrations make platelets more reactive.

Our own interest should be on the record: we sell a toothpaste that contains xylitol, so we have a commercial stake in how this question resolves. The senior author of the 2024 study discloses cardiovascular diagnostics patents and industry royalties; one correspondent opposing it discloses ownership of a pharmaceutical company. None of that decides who is right, and the arithmetic above can be checked independently of all of it.