Fluoride Risk Hides in More Than Drinking Water, Review Warns

Fluoride is one of the most paradoxical trace elements in public health. At low doses it protects against tooth decay, yet chronic excessive intake causes dental and skeletal fluorosis and has been investigated in connection with kidney, liver, endocrine, and neurodevelopmental effects. A new open-access review in Environmental Earth Sciences argues that the field has been measuring the problem the wrong way for decades: by focusing almost exclusively on drinking water, risk assessors may be systematically underestimating how much fluoride people actually absorb, and why communities with nearly identical water chemistry can end up with wildly different disease burdens.

The review, led by Jie Gao and colleagues at the China University of Geosciences in Wuhan, synthesizes roughly 150 review articles identified through structured searches of Web of Science and PubMed, supplemented by primary studies supporting specific quantitative claims. Its central thesis is that fluoride exposure is inherently multi-pathway. Human intake arises not only from drinking water but from food grown in contaminated soils, inhalation of industrial and coal-combustion emissions, and fluoride-containing consumer products such as toothpaste. The relative weight of each pathway varies dramatically across regions, and assessments that treat water as the sole route can misidentify the true drivers of population risk.

The most striking evidence for this pathway heterogeneity comes from the Ethiopian Rift Valley. In three districts where drinking-water fluoride concentrations were essentially identical at around 5 milligrams per liter, well above the World Health Organization guideline of 1.5 mg/L, fluorosis outcomes diverged sharply. Moderate-to-severe dental fluorosis was nearly absent in Fentale, a dairy-consuming community, yet affected one in five children in Alaba and half the children in Adamitulu, where staple cereals contained less calcium and rainwater harvesting was rare. The divergence traced not to water fluoride itself but to dietary calcium intake, milk’s suppressive effect on fluoride bioavailability, and household water sources, a demonstration that diet can outweigh water chemistry in determining who develops disease.

Understanding where fluoride comes from in the first place is central to the framework. Naturally, fluoride enters the environment through weathering of minerals such as fluorite, fluorapatite, and cryolite. Groundwater typically carries more fluoride than surface water because of prolonged water-rock interaction: alkaline conditions, low calcium activity, long residence times, and high bicarbonate concentrations all favor mobilization. A global compilation of more than 400,000 groundwater measurements found that the probability of high-fluoride water drops sharply as calcium rises to about 80 mg/L, then climbs again above roughly 400 mg/L due to evaporative co-concentration in closed basins. In the Ethiopian Rift, terminal lakes fed by evaporation reach 200 to 264 mg/L, more than a hundred times the WHO guideline.

Human activity has profoundly reshaped this natural cycle. Coal combustion, aluminum smelting, brick and ceramic manufacturing, glass and steel production, and phosphate fertilizer industries all release gaseous and particulate fluoride that contaminates air, soil, water, and vegetation. Agricultural irrigation with fluoride-rich groundwater transfers the element from aquifers into crops, linking water and food pathways. Occupational studies illustrate the inhalation route vividly: aluminum smelter workers exposed to fluoride released from cryolite during electrolysis show significantly elevated urinary, serum, and nail fluoride, and personal air monitoring in a primary smelter recorded mean particulate and gaseous fluoride concentrations of 0.97 and 0.33 mg per cubic meter, with post-shift urinary fluoride significantly exceeding pre-shift levels.

Even consumer products matter more than commonly assumed. Children aged two to three are estimated to ingest 0.6 to 0.9 mg of fluoride per day from toothpaste alone at typical brushing frequency, and roughly ten percent of children in this age group swallow more than double that amount, approaching or exceeding the total recommended daily intake of 0.7 mg for the age group from all sources combined. An EPA-style multi-pathway assessment estimated cumulative fluoride intake for children at 0.06 to 0.23 mg per kilogram of body weight per day depending on scenario, with toothpaste contributing a non-trivial share of the reasonable-maximum estimate.

On health outcomes, the review ranks the evidence by strength. Dental fluorosis, arising from excessive intake during tooth development, and skeletal fluorosis, from long-term incorporation of fluoride into bone mineral, are the most consistently documented outcomes, backed by well-characterized dose-response relationships. Renal and hepatic effects are biologically plausible, since the kidney clears fluoride and is exposed to high concentrations during excretion, but the human evidence base remains limited and confounding by nutritional status is inadequately controlled. Neurodevelopmental associations between fluoride and children’s cognitive performance have attracted intense attention, yet most studies rely on area-level exposure rather than individual measurement, use cross-sectional designs that cannot establish temporal precedence, and fail to control for co-exposures such as arsenic and lead. The authors are explicit that this prevents causal conclusions, not because an effect is ruled out, but because the methodology cannot yet support one. Carcinogenicity remains unresolved, and fluoride is not currently classified as a confirmed human carcinogen.

Methodologically, the review traces an evolution across three traditions that have developed largely in parallel. Deterministic models based on the US EPA framework calculate chronic daily intake from fixed-point values of concentration, ingestion rate, exposure frequency, duration, and body weight, then compare against a reference dose via the hazard quotient. These are transparent and regulatory-friendly but blind to population variability. Monte Carlo simulation addresses this by representing each parameter as a probability distribution and resampling to generate full risk distributions; in Agra, India, where fluoride ranged from 0.15 to 2.50 mg/L, this approach revealed that the 95th-percentile hazard quotient for children exceeded 1 even though mean concentrations sat near the permissible limit, a risk that a single point estimate would have concealed. Machine learning, meanwhile, has demonstrated strong performance in mapping fluoride occurrence, integrating geology, hydrochemistry, climate, and land use at resolutions conventional monitoring cannot match, from a random forest model trained on over 12,000 wells in the western United States to CatBoost models in northern China.

The review’s key proposal is to connect these three approaches into a single workflow. Machine-learning-predicted concentrations across groundwater, surface water, and, where relevant, soil and atmospheric deposition would feed multi-pathway intake models, whose outputs would then pass through Monte Carlo simulation to characterize population-level risk distributions. Sensitivity analysis within that step would identify which parameters drive the most uncertainty, telling managers whether to invest in expanded water monitoring or in food fluoride surveys. The authors stress that this framework remains conceptual: it has not been empirically implemented or externally validated, machine-learning models perform worst in exactly the data-scarce regions that need them most, and comparable prediction tools for soil and atmospheric fluoride are largely undeveloped. Recent regional studies in China that combine hydrogeochemical machine learning with age-specific risk assessment are regarded as intermediate precedents rather than complete validations, since they remain drinking-water based.

The practical implications are significant. Because the dominant exposure pathway is not uniform across settings, interventions should be pathway-specific: defluoridation of drinking water yields the greatest benefit in groundwater-dependent communities, whereas emission controls, agricultural management, or dietary guidance may matter more where coal burning or contaminated irrigation dominates. Biomonitoring offers a direct check on exposure reconstructions, with urinary fluoride reflecting recent intake and nail clippings integrating exposure over three to four months; in Ethiopian Rift communities, fingernail fluoride averaged 5.1 mg/kg and correlated strongly with drinking-water fluoride. The review’s limitations are candidly acknowledged, including its narrative design, restriction to English-language literature, and the heterogeneity of exposure definitions across studies. But its core message is difficult to dismiss: cumulative intake across all relevant routes, not water concentration alone, is the metric that should define fluoride risk, and the tools to measure it that way finally exist.

Subject of Research: Multi-pathway fluoride exposure and integrated human health risk assessment

Article Title: Multi-pathway fluoride exposure and human health risk assessment: An integrated framework review

Article References: Gao, J., Yang, Y., Pi, K., Liang, Q., Wang, Y., Zhang, Y., & Dai, C. (2026). Multi-pathway fluoride exposure and human health risk assessment: An integrated framework review. Environmental Earth Sciences, 85(16), Article 420. https://doi.org/10.1007/s12665-026-13156-0

Image Credits: AI Generated

DOI: 10.1007/s12665-026-13156-0

Keywords: fluoride, fluorosis, drinking water, groundwater contamination, multi-pathway exposure, Monte Carlo simulation, machine learning, health risk assessment, dental fluorosis, skeletal fluorosis, neurodevelopment, Environmental Earth Sciences