Commentary Neurodevelopmental Disorders

NTP study does not support fluoride as a neurotoxin

Publication reviewed:

An Evaluation of Neurotoxicity Following Fluoride Exposure from Gestational Through Adult Ages in Long-Evans Hooded Rats

McPherson CA, Zhang G, Gilliam R et al. — Neurotoxicity Research. 2018 February 5. Epub ahead of print

Section of dorsal lobe from NTP study not supporting fluoride as a neurotoxin

Background on fluoride as a neurotoxin review: The National Toxicology Program (NTP) released a systematic review, the Effects of Fluoride on Learning and Memory in Animal Studies, in 2016 in response to reported association between high levels of naturally occurring fluoride in water and lower IQ. Of 68 studies reviewed in detail, only 32 studies were available for the analyses to generate conclusions due to serious risk of bias and incomparable measurements and designs used across the rest of the studies. These studies showed low-to-moderate confidence for a pattern of findings suggestive of fluoride’s effect on learning and memory. The NTP review group identified the following issues in the available literature and declared an intent to fill data gaps by conducting laboratory studies in rodents in the near future.

  • Very few studies assessed learning and memory effects in experimental animals at exposure levels near 0.7ppm and had information on alternative sources of fluoride (i.e. food, water supply) available, thus relevance of the findings to human exposure levels in the optimally fluoridated communities (0.7ppm fluoride concentration) is unknown.
  • The outcome endpoint in the majority of studies was a simple latency measurement of learning or memory in the final training session rather than an evaluation of the acquisition of the task to demonstrate learning. Thus, interpretation of the data is hindered by inability to exclude alterations from baseline levels or differences in motor-related performance over the training session as contributing factors.
  • In many studies, there was a lack of reporting of 1) randomization and blinding, 2) specification of test methodologies to assess the outcomes, and/or 3) controlling of confounders such as litter effects, sex, life-stage at exposure, and duration of exposure.  Studies also appeared statistically underpowered to detect a <10% or <20% change from controls for most behavioral endpoints.

Methods: The current study led by the Neurotoxicology Group of the NTP Laboratory addressed the previously identified methodological/design issues in the literature successfully. Specifically, the methodological improvements are notable in the following areas:

  • The fluoride exposures simulated human fluoride exposures by the use of equivalent fluoride doses and establishing a separate route of exposures from diet (20.5ppm vs. 3.24ppmF) and drinking water (0, 10, or 20ppm). Fluoride concentration of 20ppm in rat’s drinking water was equivalent to 4ppm, the US EPA’s current maximum contaminant level, based on the conventional wisdom that a 5-fold increase in dose is required in animals to achieve comparable human serum levels. The exposure levels were validated by assessing the fluoride deposition and accumulation in brain and bone (femur) in addition to fluoride levels in plasma and urine.
  • The experiment (exposure to fluoridated food and water, available ad libitum) began on gestational day 6 and continued throughout lactation. Male pups were observed through adulthood (postnatal day >90).
  • The neurobehavioral endpoint in male pups was measured in various domains: Learning, memory, motor, sensory function, depression, and anxiety. Learning and memory were also evaluated across different tests and in reversal trials and demonstrated acquisition over sessions examining a number of different aspects of performance.
  • Additional effects reported for fluoride exposure that may influence behavior were examined (i.e. thyroid hormone levels, kidney, liver, reproductive system histopathology, and neuronal and glia morphology in the hippocampus) to obtain a better understanding of observed effects.
  • To minimize biases, randomizations and blinding are sufficiently implemented and documented along with detail description of test procedures. Many of behavioral tests were video captured for detail analysis.
  • The authors statistically determined group sizes to sufficiently detect significant differences (p<0.05) between experimental and control groups.

Summary Findings and Public Health Implications:

  • Developmental exposure to fluoride from drinking water and diet beginning on gestational day 6 were associated with elevated internal fluoride levels in brain and femur as well as plasma and urine of male rat offspring. A differential absorption of fluoride between water and food was also demonstrated.
  • Fluoride exposure at the levels examined in this study was not found to alter motor performance or learning and memory in the test paradigms assessed or alter thyroid hormone (T3, T4, or TSH) levels or produce neuronal damage or glia reactivity in the hippocampus, or histological damage in heart, kidney, or liver. The only exposure-related effect that they found was mild hyperanalgesia and mild inflammatory response in the prostate.
  • This latest research on fluoride and neurobehavioral health overcame many limitations and weaknesses of previous studies and demonstrated 1) relationships between developmental fluoride exposures from water & diet and fluoride levels in various tissues and specimens in offspring and 2) no exposure-related differences in motor, sensory, or learning and memory performances in rats.
  • When the 2006 NRC report suggested a need for more research on neurotoxicity and neurobehavioral effects of fluoride, the committee was basing this on available data from human studies conducted in fluoride-endemic regions showing high-risk of bias but some consistencies in the findings. Meanwhile data from available molecular and cellular studies could be interpreted to suggest potential changes in nervous system functions but only a few animal studies reported unsubstantial magnitude of alterations in the behavior of rodents after fluoride treatment. Since then, more research (including epidemiological and animal studies) were published, yet the majority of studies still suffer from various sources of risk for bias and the accumulated evidence remains mixed.
  • The findings of this well-controlled animal study directly address previous concerns regarding potential biological plausibility of fluoride as a neurotoxin. The findings provide valuable information and assurance that low-level fluoride exposures from water and diet that are equivalent to the levels allowed in the US does not result in clinically adverse neurobehavioral function or pathological effects in various organs.
Appraisal Neurodevelopmental Disorders

Appraisal of prenatal fluoride IQ study

Publication reviewed:

Prenatal fluoride exposure and cognitive outcomes in children at 4 and 6-12 years of age in Mexico

Morteza Bashash, Deena Thomas, Howard Hu et al. — Environmental Health Perspectives


The authors analyzed data from the Early Life Exposures in Mexico to Environmental Toxicants (ELEMENT) project to examine if prenatal exposure to fluoride is associated with declined childhood intelligence.

Subjects: The ELEMENT project recruited women who were 14 or less weeks pregnant and free of medical, mental disorders, high-risk pregnancy as well as use of recreational alcohol and drugs use at three clinics of the Mexican Institute of Society Security in Mexico City that serve low-to-moderate income populations.

Exposure measure: Prenatal F exposure was measured as an averaged value of maternal creatinine-adjusted urinary fluoride concentrations (maximum three and minimum one spot urine sample[s] were archived for each woman).

Outcome measure: Offspring’s neurocognitive outcomes were measured as the General Cognitive Index (GCI) score at 4 years and IQ score at 6-12 years.

Covariates: Maternal age, education, marital status, birth order, birth weight, gestational age at delivery, maternal smoking, maternal IQ (estimated using selected subtests of the WAIS-Spanish measured at 6-12 months after birth), and cohort ID. The specific-gravity adjusted urinary fluoride values obtained from offspring at 6-12 years of age were included in the model for prenatal F exposure and IQ.

The study found:

  • Significant correlation between GCI and IQ scores.
  • No significant correlation between prenatal creatinine-adjusted urinary fluoride and offspring’s specific-gravity adjusted urinary fluoride levels at 6-12 years of age.
  • Prenatal creatinine-adjusted urinary fluoride level and GCI at 4 years of age showed mild linear relationship: 0.5mg/L increase in prenatal urinary fluoride was associated with 3.15-point drop in GCI scores (p=0.01, N=287).
  • Prenatal urinary fluoride level and IQ at 6-12 years of age showed mild curvilinear relationship: 1) no clear association between prenatal urinary fluoride and IQ scores below approximately 0.8mg/L urinary fluoride levels, and 2) a negative association above prenatal urinary fluoride 0.8mg/L. The authors found 0.5 mg/L increase in prenatal urinary fluoride was associated with -2.5 points in IQ scores (p=0.01, N=211).
  • Sensitivity analyses conducted for the subsets of data (N<200) indicated the following:
  • The negative associations between prenatal urinary fluoride and GCI or IQ persisted with further adjustment for other potential confounders (family possession, maternal bone lead and blood mercury levels). The effect estimates were attenuated when family possession (SES proxy) and maternal blood mercury values were adjusted in the models relative to unadjusted models, while all of the effect estimates were higher in the subset of subjects with available data of SES, maternal bone lead and blood mercury levels.
  • There was no clear, statistically significant, association between contemporaneous children’s urinary fluoride and IQ at 6-12 years of age either unadjusted or adjusted for maternal urinary fluoride during pregnancy.
Prenatal fluoride IQ study plot


  • A – Strong methodology and unbiased, appeared in peer-reviewed in respected science journal
  • B – Strong methodology and unbiased, not in peer-reviewed journal
  • C – Weak methodology and/or biased
  • F – Not a scientific finding


  • High – All the peer-reviewed research to date support these findings, and a significant amount of research has been done in this area.
  • Medium – Most, but not all, peer-reviewed research to date support these findings, and a significant amount of research has been done in this area.
  • Low – Not a lot of research has been done in this area, or some, but not most, other peer-reviewed research supports these findings.
  • Not Supported – No other studies support this study’s conclusions.
  • Contradicted – Most studies contradict this study’s conclusions.


• Data of childhood neurocognitive outcomes collected in the longitudinal birth cohort research project with various maternal and perinatal covariates data including maternal IQ, education, smoking, and birth outcomes.
• Although sample size of subset data was small, the authors were able to check the effect of SES (although proxy), maternal lead and mercury in the investigated association.
• Urinary fluoride data were adjusted by creatinine and specific gravity for variation in urinary dilution.
• The authors report detail methods and results.


• Limitation of urinary fluoride as a biomarker of fluoride exposure: Urinary fluoride level fluctuates during the day and reflects only recent exposures, and it is unknown if fluoride level measured in spot urine samples during pregnancy is a good measure of prenatal fluoride exposure for fluoride’s neurotoxic effect in children.
• No fluoride data other than urinary fluoride levels were collected or available, thus we do not know the source of fluoride exposure (i.e. fluoride in water, salt, toothpaste, environmental or industrial F exposure etc.) or how such external dose of exposures reflected internal F dose (in urine).
• Lack of data on iodine in salt, other nutritional intake and dietary practices that could influence pregnancy, urinary excretion, and fetus-child cognitive development, and environmental neurotoxicants such as arsenic.


This study had an advantage of using the data from the Early Life Exposures in Mexico to Environmental Toxicants (ELEMENT) project, which collected data longitudinally from pregnancy to childhood on the exposures to environmental toxicants such as lead and mercury and childhood neurocognitive outcomes. However, this study on fluoride was not planned prior to the ELEMENT data collection, therefore the authors had limited ability to validate fluoride exposures and relied solely on fluoride concentrations in spot urine samples.

Biomarkers of fluoride exposure such as urinary and serum fluoride are considered a marker of recent fluoride exposures. Urinary fluoride fluctuates, thus the value can be influenced by the timing of exposure and sampling, and we do not know if the level captured in a spot urine sample reasonably reflects the usual and/or long-term exposures to fluoride during the prenatal period. In this study, some of the subjects had three spot (second morning void) urine samples obtained from each of trimesters, but approximately 80% of subjects provided only one or two spot urine samples during pregnancy. While the authors adjusted fluoride concentration in spot urine samples with creatinine and specific gravity for dilution factor, there are a number of factors that affect fluoride uptake, retention, and excretion. It is anticipated that fluoride metabolisms would change with gestation, yet we do not know how it changes during the different phases of pregnancy. The authors reported a mean prenatal urinary fluoride value of 0.9 mg/L among the study subjects and thought the value was within normal range, however there are limited population-based data available to determine the reference value of urinary fluoride concentrations during pregnancy. EPA considers urinary fluoride as Group I biomarker for fluoride-related neurotoxicity because there is a lack of established methodology of sampling (i.e. first morning vs. second morning void, spot urine vs. 24-hour urine sampling), analytic strategies, and established relationship between external dose (i.e. supplemental fluoride dose, fluoride concentration in water), internal dose (i.e. in urine), and biological endpoint (i.e. neurotoxicity).

The negative association between prenatal urinary fluoride level and cognitive ability found at 4 and 6-12 years of age in the offspring, no association found between children’s urinary fluoride and IQ at 6-12 years of age, and no significant effect of prenatal urinary fluoride below 0.8 mg/L on childhood IQ in non-linear relationship found in this study all corroborate with a portion of the published literature. A largely spread scatter plot distribution suggests that prenatal fluoride exposure may be a small portion of variations that explain the relationship. We agree with the authors on that additional studies are needed to examine if the association found in this study are replicated in other study populations and if fluoride exposure during pregnancy is indeed a critical window of susceptibility for population’s neurocognitive health. There are only a few studies of relatively small observational studies from Mexico that looked at the fluoride exposure in pregnant women and its association with neurobehavioral outcomes in their offspring (Bashash et al 2017, Valdez Jiménez et al. 2017, and unpublished thesis of Thomas 2014). We also desperately need to learn more about fluoride metabolism during pregnancy and how prenatal urine fluoride concentrations are related to external fluoride doses such as fluoride in drinking water.