After the discovery of fluoride as a caries-preventing agent in the mid-twentieth century, fluoridation of community water has become a widespread intervention, sometimes hailed as a mainstay of modern public health. However, this practice results in elevated fluoride intake and has become controversial for two reasons. First, topical fluoride application in the oral cavity appears to be a more direct and appropriate means of preventing caries. Second, systemic fluoride uptake is suspected of causing adverse effects, in particular neurotoxicity during early development. The latter is supported by experimental neurotoxicity findings and toxicokinetic evidence of fluoride passing into the brain.
An integrated literature review was conducted on fluoride exposure and intellectual disability, with a main focus on studies on children published subsequent to a meta-analysis from 2012.
Fourteen recent cross-sectional studies from endemic areas with naturally high fluoride concentrations in groundwater supported the previous findings of cognitive deficits in children with elevated fluoride exposures. Three recent prospective studies from Mexico and Canada with individual exposure data showed that early-life exposures were negatively associated with children’s performance on cognitive tests. Neurotoxicity appeared to be dose-dependent, and tentative benchmark dose calculations suggest that safe exposures are likely to be below currently accepted or recommended fluoride concentrations in drinking water.
The recent epidemiological results support the notion that elevated fluoride intake during early development can result in IQ deficits that may be considerable. Recognition of neurotoxic risks is necessary when determining the safety of fluoride-contaminated drinking water and fluoride uses for preventive dentistry purposes.
In 2006, the U.S. National Research Council (NRC) evaluated the fluoride standards of the Environmental Protection Agency (EPA) and concluded that fluoride can adversely affect the brain through both direct and indirect means, that elevated fluoride concentrations in drinking-water may be of concern for neurotoxic effects, and that additional research was warranted . At the time, and continuing through today, the EPA’s Maximum Contaminant Level Goal (MCLG) for fluoride was 4.0 mg/L that aimed at protecting against crippling skeletal fluorosis, which is still considered to be the critical adverse health effect from fluoride exposure . Following the NRC review, evidence has accumulated that the developing human brain is inherently much more susceptible to injury from neurotoxic agents, such as fluoride, than is the adult brain . A review and meta-analysis published in 2012  assessed a total of 27 research reports, all but two of them from China, on elevated fluoride exposure and its association with cognitive deficits in children. All but one study suggested that a higher fluoride content of residential drinking water was associated with poorer IQ performance at school age. Only a couple of these studies had been considered by regulatory agencies [1, 5]. As much additional evidence has emerged since then, it seems appropriate to update the assessment of potential human neurotoxicity associated with elevated fluoride exposure, especially during early development.
The present review first outlines the importance of drinking water as a source of fluoride exposure, followed by the toxicokinetics of fluoride absorbed into the body, including passage through the placenta and the blood-brain barrier, and finally a brief summary of the experimental evidence of developmental neurotoxicity. All of this evidence supports the plausibility that elevated fluoride exposure in early life may cause adverse effects on the brain. The main part of this review addresses the epidemiological studies of fluoride neurotoxicity, with a focus on children and the dose-dependent impact of prenatal and early postnatal exposures.
Potential sources of fluoride exposure
Fluoride occurs in many minerals and in soil , thus also in groundwater; the average concentration in the U.S. is 0.26 mg/L . Since the mid-1940s, fluoride has been added to many community water supplies with the aim of preventing tooth decay . In the U.S., fluoridation is recommended at a concentration of 0.7 mg/L . Water fluoridation is applied in several other countries as well, such as Australia, Brazil, Canada, Chile, Ireland, New Zealand, and the United Kingdom. For adults in the U.S., fluoridated water and beverages contribute an average of about 80% of the daily total fluoride intake (estimated to average 2.91 mg) in fluoridated communities . In a Canadian study of pregnant women , water fluoridation was the major predictor of urinary fluoride excretion levels, with creatinine-adjusted concentrations of 0.87 mg/L and 0.46 mg/L in fluoridated (0.6 mg/L water) and non-fluoridated (0.12 mg/L) communities.
In addition to fluoridated water and other forms of caries prevention, tea is an important source of fluoride exposure, even if prepared with deionized water [12, 13]. Additional sources of fluoride intake include certain foods (such as sardines), industrial emissions, supplements, pesticide residues, and certain pharmaceuticals that can release fluoride . Few studies provide population-based data on fluoride exposure, although national data on plasma-fluoride concentrations are available from a recent NHANES study in the U.S. .
Uptake, distribution and retention
Approximately 75–90% of ingested fluoride is absorbed and readily distributed throughout the body, with approximately 99% of retained fluoride being bound in calcium-rich tissues such as bone and teeth [6, 15] as well as the calcified parts of the pineal gland . Fluoride also crosses the placenta and reaches the fetus [1, 6] and the amnionic fluid . The fluoride concentration in breast milk is low, generally less than 0.01 umg/L [1, 18], and formula can therefore contribute much higher intakes, especially when prepared with high-fluoride water [19,20,21]. Children and infants retain higher proportions of absorbed fluoride compared to adults, i.e., about 80–90%, as compared to about 50–60% in adults [6, 15].
As drinking water is usually the major source of exposure, the community water-fluoride concentration has often been used as an exposure parameter in ecological studies. For individual exposure assessment, the total fluoride intake can be calculated from daily water consumption and the intakes of other major sources, such as tea. Analyses of biological samples, i.e., urine and blood (generally in the form of plasma or serum) provide information on fluoride circulating in the body . In adults, the fasting plasma-fluoride concentration, when expressed in micromoles per liter [umol/L], is approximately equal to the concentration in the drinking water or in the urine expressed in mg/L . Fluoride excretion is mainly via urine, and the concentration represents both recent absorption and releases from long-term accumulation due to continuous bone tissue remodeling . Pregnant women may show lower urinary fluoride levels than non-pregnant controls, perhaps due to fetal uptake and storage in hard tissues , although the urinary fluoride excretion tends to increase from the first to the third trimester [11, 24]. Children have lower urine-fluoride concentrations, most likely due to fluoride incorporation in the growing skeleton .
As indicator of daily intake [25, 26], urinary fluoride excretion is often assessed in spot urine samples, although morning urine or 24-h samples may provide better precision, as may be the case with timed excretion . To adjust for temporal differences in urine production, fluoride concentrations in spot samples are usually standardized according to the creatinine concentration and/or relative density. These considerations are important when evaluating the validity of exposure assessments in epidemiological studies.
While the blood-brain barrier may to some extent protect the adult brain from many toxic agents, this protection is less likely in the fetus and small child with an incompletely formed barrier . As indication that fluoride passes the blood-brain barrier, fluoride concentrations in human cerebrospinal fluid approach those occurring in serum . Further, imaging studies of radioactive fluoride used in cancer treatment document that circulating fluoride reaches the brain [30,31,32,33]. Within the brain, fluoride appears to accumulate in regions responsible for memory and learning [34, 35].
As fluoride can pass both the placental barrier and the blood-brain barrier, it reaches the fetal brain . Accordingly, autopsy studies in endemic areas in China have shown elevated fluoride concentrations in aborted fetal tissues, including brain [37, 38]. Also, fluoride concentrations in maternal and cord serum correlate well , cord blood showing slightly lower concentrations, apparently about 80% of the concentrations in maternal serum , though depending on gestational age . Fetal blood sampling techniques have allowed documentation of elevated fluoride concentrations in the fetal circulation after administration of sodium fluoride to the mother . Accordingly, assessment of fluoride in maternal samples during pregnancy may be used as indicator of fetal exposure.
Due to a well-established dose-response relationship between early-life fluoride exposure and the degree of dental fluorosis [6, 20, 42], this abnormality can serve as a useful biomarker of developmental fluoride exposure. When water fluoridation was first introduced in the middle of the twentieth century, U.S. health authorities estimated that less than 10% of children in fluoridated communities (at 1 mg/L water) would develop dental fluorosis, and only in its mildest forms . Subsequent epidemiological studies have demonstrated prevalence and severity of fluorosis much higher than predicted [9, 44, 45]. Increased occurrence of dental fluorosis has also been recorded in fluoridated areas in the United Kingdom . This increase may be related to the widened use of fluoridated water for beverages and food products for general consumption and for formula preparation for infants [19, 21], as well as increased usage (and ingestion) of fluoride-containing toothpastes among preschoolers .
In vitro studies have documented fluoride toxicity to brain cells, most of the studies using high fluoride concentrations, though some effects have been demonstrated at lower, more realistic levels [48, 49]. In the low-dose studies, 0.5 umol/L (10 ug/L) was sufficient to induce lipid peroxidation and result in biochemical changes in brain cells , while 3 umol/L (57 ug/L) induced inflammatory reactions in brain cells . These concentrations are similar to the upper ranges of serum-fluoride levels reported in the general population . In addition, fluoride can negatively affect brain development in rats at levels below those that cause dental lesions .
Utilizing computerized surveillance of rat behavior, a landmark study showed signs of neurotoxicity at elevated fluoride exposure , and more recent studies have reported fluoride-induced neurochemical, biochemical, and anatomic changes in the brains of treated animals, although often at doses much above human exposure levels. Among possible mechanisms of developmental neurotoxicity is toxicity to the thyroid gland , a mechanism relevant in regard to several neurotoxicants [53, 54]. Thus, the NRC concluded that fluoride is an endocrine disrupter that can affect thyroid function at intake levels as low as 0.01 to 0.03 mg/kg/day in individuals with iodine deficiency .
A 2016 review by the National Toxicology Program (NTP) focused on fluoride neurotoxicity in regard to learning and memory . At water concentrations higher than 0.7 mg/L, NTP found a low-to-moderate level of evidence. The evidence was the strongest (moderate) in animals exposed as adults and weaker (low) in animals exposed during development, where fewer studies were available at relevant exposure levels. Most experimental studies had used concentrations exceeding the levels added to water in fluoridation programs, but the NTP recognized that rats require about five times more fluoride in their water to achieve the same serum-fluoride concentrations as humans .
Subsequently, several additional developmental studies have been published, including two that reported impaired learning/memory in rats consuming water with fairly low fluoride concentrations [56, 57]. However, not all studies have reported adverse effects , perhaps due in part to strain or species-related differences in vulnerability to fluoride. In addition, most animal studies used subchronic exposure scenarios and, due to the lack of fluoride transfer into milk, neonatal exposure was not considered, thereby likely underestimating the effect from early-life exposure. Overall, the experimental evidence of developmental neurotoxicity appears to be strengthened and to provide plausibility to the potential occurrence of neurodevelopmental effects in humans.
Publications on fluoride neurotoxicity in humans were identified from the PubMed data base by using “fluoride” along with search terms “neurotoxic*”, “neurologic”, and “intelligence”. The searches were narrowed by limiting to “human,” “most recent 10 Years,” and “English.” Additional searches using “fluoride” also included search terms “prenatal exposure delayed effects”[MeSH] or “neurotoxicity syndrome”[MeSH]. Secondary searches used combinations of fluoride with “maternal exposure” or “academic disorder, developmental”.
Supporting literature from earlier years was obtained by using the terms “occupational exposure” or “endemic disease”. References cited in the publications and in recent review reports [55, 59,60,61] were also retrieved, as were publications listed by PubMed under “Similar articles”. Because these articles may not represent an exhaustive list of relevant studies, separate searches included the web site of the journal Fluoride (http://www.fluorideresearch.org/) and the site (http://oversea.cnki.net/kns55/default.aspx) that covers many Chinese-language journals not included in PubMed. Full-text copies of all relevant studies were obtained, and studies were disregarded if no more than an abstract in English was available.
For the purpose of identifying safe exposure levels, regulatory agencies routinely use benchmark dose calculations . While such calculations would normally require access to the original data, approximate BMD and BMDL results can be generated from descriptive data on associations between maternal urinary fluoride concentrations and the child’s IQ . The benchmark dose (BMD) is the dose leading to a pre-defined change (denoted BMR) in the response (in this case, an IQ loss), when compared to comparable, but unexposed individuals. The BMR must be defined before the analysis , and recent practice suggests that a decrease in IQ of one point is an appropriate BMR [64,65,66,67].
In the above framework, the difference between the expected IQ level at the unexposed background (E [Y (0)]) and at the BMD (E [Y (BMD)]) is equal to the BMR:
In a linear model (Y(d) = a + Bd + e), we get BMD = BMR/B. The main result of the benchmark analysis is the benchmark dose level (BMDL), which is defined as a lower one-sided 95% confidence limit of the BMD. In the linear model
where Blower is the one-sided lower 95% confidence limit for B . Thus, in this model the benchmark results are a function of statistics routinely calculated in regression analysis.
For a linear dose-response model, epidemiological studies that report developmental fluoride exposure in regard to IQ will allow computation of BMD and BMDL based only on the regression coefficient and its uncertainty, assuming a Gaussian distribution.
Occupational and endemic area studies
The neurotoxicity of chemicals is often first discovered from workplace exposures , later followed by case reports and small studies of highly-exposed children or pregnant women, then confirmed in population studies that are later complemented by prospective studies . The same seems to be true of fluoride. A brief summary is therefore presented on the progress of this evidence before focusing on developmental exposures.
In connection with his seminal studies of occupational fluoride poisoning in the 1930s, Kaj Roholm reported evidence of nervous system effects in the Copenhagen cryolite workers : “The marked frequency of nervous disorders after employment has ceased might indicate that cryolite has a particularly harmful effect on the central nervous system.” (p. 178). Later on, the Manhattan Project in the 1940s recorded neurological effects in workers exposed to uranium hexafluoride gas (UF6), and the “rather marked central nervous system effect with mental confusion, drowsiness and lassitude as the conspicuous features” was attributed to the fluoride rather than uranium .
Subsequent occupational health studies are somewhat harder to interpret, as fluoride exposure usually occurs as part of a mixture, e.g., in aluminum production . Nonetheless, industrial fluorosis (a.k.a. osteosclerosis) was found to be associated with gradually progressive effects on the normal function and metabolism of the brain and other aspects of the nervous system , and a review highlighted difficulties with concentration and memory accompanied by general malaise and fatigue . More recent studies have applied neuropsychological tests to assess cognitive problems associated with occupational fluoride exposures [75, 76]. The present literature search did not reveal any recent publications on neurotoxicity from occupational fluoride exposure. While Roholm  described unusually serious dental fluorosis in a son of a female cryolite worker, none of the occupational studies identified referred to adverse neurobehavioral effects in the progeny of female workers.
Opportunities for epidemiological studies of the general population depend on the existence of comparable groups exposed to different and stable amounts of fluoride, e.g., from drinking water. Such circumstances are difficult to find in many industrialized countries, as water-fluoride concentrations may not be well defined, residents may consume beverages from a variety of sources, and exposures are affected by residences changing over time. Multiple epidemiological studies of developmental fluoride neurotoxicity have been conducted in countries such as China where elevated water-fluoride concentrations may exceed 1 mg/L in many rural communities. In these settings, families typically remain at the same residence, with a well-defined water source that has provided fairly constant fluoride exposures.
Studies from high-fluoride endemic areas in China have reported on abnormal neuropathology findings from aborted fetuses  and lower nerve cell numbers and volumes in fetal brain tissue at the elevated exposures . Deviations observed in neurotransmitters and receptors have suggested neural dysplasia , as later replicated along with decreased excitatory aspartic acid and elevated inhibitory taurine in comparison to controls . Although these studies are in agreement with the notion that fluoride from the mother’s circulation can pass into the fetal brain with subsequent anatomic and biochemical changes, the studies related to elevated fluoride exposure originate primarily from coal burning, which may have contributed other, undocumented contaminants.
Additional community studies in adults have focused on cognitive problems and neurological symptoms in subjects with skeletal fluorosis. Using neuropsychological tests, including the Wechsler scale, 49 adult fluorosis patients were compared with controls and showed deficits in language fluency, recognition, similarities, associative learning, and working memory . Further, cognitive impairment in elderly subjects from a waterborne fluorosis area was found to be much more common than in less-exposed controls . Dementia diagnosis in North Carolina was more common at higher water-fluoride concentrations , and similar findings for fluoride (and aluminum) have recently been reported from Scotland . Excess occurrence of neurological symptoms (i.e., headaches, insomnia, and lethargy) have also been recorded in both adults and children from waterborne fluorosis areas . However, these studies are hard to evaluate due to uncertainty about past fluoride exposure levels and the possible influence of other risk factors. The literature search did not reveal any other recent studies that added important evidence in this regard.
Cross-sectional studies of children in exposed communities
Most studies that have investigated fluoride’s impact on childhood IQ are from locations in China with elevated exposure to fluoride, within and outside of known endemic areas [1, 4, 84]. When water supplies derive from springs or mountain sources, small or large pockets of increased exposures may be created near or within similar areas of lower exposures, thus representing useful epidemiology settings. The fluoride exposure from the household water would then represent the only or major difference between nearby neighborhoods. At the time, children in rural China had very little exposure to fluoridated dental products . The local water-fluoride concentration can then serve as a feasible and appropriate exposure parameter, and some studies emphasized that the children were born in the particular study area, and/or had been using the same water supply since birth. Reliable exposure assessment then becomes possible when rural families remain for a long time at the same residence. Any deviation from stable exposure would result in exposure misclassification and thereby a likely underestimation of the toxicity . Thus, the consistency of study findings supports the likelihood that developmental fluoride exposure causes cognitive deficits . Although the study designs are technically cross-sectional, many of the settings allowed consideration of the current exposure as an indicator also of a longer-term exposure level.
Most study reports have not been widely disseminated and considered in literature reviews. Four studies from China that were published in English [87,88,89,90] were cited in the 2006 NRC report , while the World Health Organization (WHO) considered only two [87, 90] in its revised Environmental Health Criteria document on fluoride from 2002 . A meta-analysis from 2007 included five studies , four of which were not in a subsequent review . The latter review was cited by the EU Scientific Committee on Health and Environmental Risks (SCHER) working group in 2010  in support of a conclusion that the evidence of neurotoxicity was insufficient.
A meta-analysis from 2012 was based on a collaboration with Chinese experts on fluoride toxicity and covered 27 cross-sectional studies reporting associations between children’s intelligence and their fluoride exposure . Overall, children who lived in areas with high fluoride exposure had lower IQ scores than those who lived in low exposure or control areas, the average difference being close to 7 IQ points. These findings were consistent with an earlier review , but included nine more studies and more systematically addressed study selection, exclusion information, and bias assessment.
Two of the 27 studies that we included in the analysis were conducted in Iran [92, 93], while all other study populations were from China. Two cohorts were exposed to fluoride from coal burning [94, 95], but otherwise the study populations were exposed to fluoride through drinking water contaminated from soil minerals. Due to the use of different cognitive tests, normalized data were used to estimate the possible effects of fluoride exposure on intelligence. The results were materially unchanged in various sensitivity analyses, as were analyses that excluded studies with possible concerns about co-factors, such as iodine deficiency and arsenic toxicity, or non-water fluoride exposure from coal burning .
Among the 27 studies, all but one showed random-effect standardized mean difference (SMD) estimates that indicated an inverse association, ranging from -0.95 to -0.10 (one study showed a slight, non-significant effect in the opposite direction). The overall random-effects SMD estimate (and 95% confidence interval, CI) was -0.45 (-0.56, -0.34). Given that the standard deviation (SD) for the IQ scale is 15, an SMD of -0.45 corresponds to a loss of 6.75 IQ points. Although substantial heterogeneity was present among the studies, there was no clear evidence of publication bias . Given the large number of studies showing cognitive deficits associated with elevated fluoride exposure under different settings, the general tendency of fluoride-associated neurotoxicity in children (p < 0.001) seems robust.
Recent cross-sectional studies of children
The present study presents an updated literature search that revealed 14 new studies on the association between early-life fluoride exposure and IQ in children (Table 1). All 14 studies reported apparent associations between elevated fluoride exposure and reduced intelligence, although one did not reach statistical significance. The several new Chinese-language studies showed similar associations between fluoride exposure and reduced IQ [96, 101,102,103, 105, 107, 108], although often published as short reports in national journals and according to the standards of science at the time. Similar findings were reported from India [98, 100, 110] and Africa [104, 106]. As with the previous reports, most of these newer studies suffer from limitations of covariate reporting, which limited the opportunity to assess possible bias. Also, a variety of outcomes have been employed, such as neuropsychological tests and Raven-based intelligence scales. Of note, fluoride exposure was accompanied by other contaminants from coal burning in some studies [96, 99, 101, 102]. Four studies used the degree of dental fluorosis as exposure parameter, and three of them reported a clear negative association with IQ [100, 103, 107], although statistical significance was not reached in one study . The water-fluoride concentrations tended to be somewhat lower than in previous studies and thus more relevant to exposures occurring outside of endemic areas.
|Reference||Study location, year||No. in high-exposure group||No. in reference group||Age range (or mean), years||Fluoride exposure||Outcome measure||Results|
|Assessment||Mean or range (mg/L)|
|||China, 2014||123||42||8–12||Urine||3.03 (urine, short-term); 2.33 (urine, long-term); 1.34 (urine, ref)||RSPMa||Fluoride exposure was negatively associated with children’s IQ|
|||China, 2015||26 (moderate/severe dental fluorosis)||8 (normal/questionable dental fluorosis)||6–8||Drinking water, urine||2.66 (water, moderate/severe); 1.0 (water, normal/questionable); 2.44 (urine, moderate/severe); 0.45 (urine, normal/questionable)||WRAMLb; WISC-IVc||Moderate and severe fluorosis were significantly associated with deficits in digit span scores.|
|||India, 2015||215||214||6–12||Drinking water||2.41 (water, high); 0.19 (water, ref)||RCPMd||IQs of highly exposed children were significantly lower than those with low-level exposure|
|||China, 2015||84||96||7–13||Drinking water, urine||1.40 (water, high); 0.63 (water, ref); 2.40 (urine, high); 1.10 (urine, ref)||CRT-RCe||Fluoride exposure was negatively associated with children’s IQ|
|||India, 2016||23 (severe dental fluorosis)||4 (normal dental fluorosis)||6–18||Groundwater and urine||2.11 (water); 0.45–17.00 (range, urine)||CRT-RCe||IQ was negatively correlated with degree of dental fluorosis|
|||China, 2017||68||50||3–12 months||Coal burning vs. control||Mothers in exposed group had dental fluorosis||MDI & PDI (CDCC)f||MDI & PDI in exposed group were significantly lower than those in the control group|
|||China, 2017||167||120||8–12||Coal burning vs. control||Dental fluorosis index 53.9% in exposed group||RSPMa||IQ was lower in children with high fluoride exposure (not significant)|
|||China, 2018||221||100||8–12||Drinking water||1.2 (water, high); 0.25 and 0.78 (water, controls)||CRT-RCe||IQ was lower in children from endemic areas and in those with dental fluorosis|
|||Sudan, 2018||775 (total)||N/a||6–14||Drinking water||0.01–2.07 (water)||School performance based on method adopted by MoE||Inverse relationship between fluoride in drinking water and average school performance|
|||China, 2018||134||134||8–12||Dental fluorosis||N/a||CRT-RCe||IQ was lower in children from endemic areas|
|||Egypt, 2018||186||814||4.6–11||Drinking water||0.92–3.75 (water)||DAPg||Decreased scores in children from areas with elevated fluoride in drinking water|
|||China, 2018||1250||1636||7–13||Drinking water and urine||2.00 + 0.75 (water, high); 1.37 + 1.08 (urine, high); 0.50 + 0.27 (water, ref); 0.41 + 0.49 (urine, ref)||CRT-RC2e||IQ was lower in children at higher fluoride in water and urine and at greater severity of dental fluorosis|
|||China, 2019||25||27||8–12||Drinking water||N/a||CRT-RC2e||IQ was lower at elevated fluoride exposure|
|N/a||7–13||Drinking water, urine||1.39 + 1.01 (water); 1.28 + 1.30 (urine)||CRT-RC2e||Low to moderate fluoride exposure is associated with alterations in thyroid function and lower IQ|