Concerns Experts Shared About the 2019 Green, et al. Fluoride/IQ Study

Some explanations on why the Green, et al. fluoride - IQ study (which makes the claim that "A 1-mg/L increase in Maternal Urinary Fluoride level was associated with a 4.49-point lower IQ score in boys, but there was no statistically significant association with IQ scores in girls") is not likely to change the scientific consensus that community water fluoridation is a safe and effective public health measure to reduce the risk of dental decay and related health issues.  Correlation is another term often used when changes in variables are presented as related to one another.
       Additional Fluoridation Resources

Statistician, Rene Najera’s discussion of the Green, et al. paper:
In consultation with friends and colleagues, we found a lot to be worried about in the epidemiological design of the study and the biostatistical analysis of the resulting data… And, of course, of the conclusions reached by the authors and the press (with some help from the authors). In the [second] part, I will lay out the epidemiological design of the study and how it is flawed both in terms of its internal and external validity. In the third part, I will lay out the biostatistical analysis issues we observed.
The discussion section is standard jargon that researchers include in their papers where they basically acknowledge that their study is based on a limited sample of the entire population and that more research is needed. (That’s the standard line for asking for more cash to do another study.) They also state clearly: ‘Nonetheless, despite our comprehensive array of covariates included, this observational study design could not address the possibility of other unmeasured residual confounding.''

Expert reaction to study looking at maternal exposure to fluoride and IQ by eight international experts:
In summary it is not correct to imply that the data here show evidence of a link between maternal fluoride exposure and IQ. The average change in IQ is not statistically significant.”, “I don’t think there is a need to worry.”, “There are many limitations which could affect the findings reported in this paper and until these are addressed, I regard the results as interesting, but not something to alter fluid or food intake by mothers.”, “There are numerous other complicating factors that limit the interpretability of the results…”, “The reduction in average IQ scores in young children, associated with higher fluoride intake during their mothers’ pregnancies, is really fairly small, and there are reasons to doubt whether the reduction is entirely caused by fluoride intake anyway.”, “I would be very cautious about over interpreting this data. Statistical significance does not equal ‘importance’.”

If at first you don't succeed . . . statistical manipulation might help:
Details of why anti-F interpretations are faulty: “… the study itself shows no significant difference in children whose mothers lived in fluoridated or unfluoridated areas during pregnancy. Here is the relevant data from Table 1 in the paper..:

No, fluoride doesn't lower IQ. It fails to satisfy Hill's Criteria Of Causality, American Council on Science and Health:
While the authors conclude that a 1-mg increase in fluoride detected in the mother's urine is linked to an IQ drop of about 4.5 points in boys, there is no statistically significant IQ difference among girls. (Actually, it's worse than that. The point estimate shows an IQ increase of 2.4 points for girls.) Obviously, that is incoherent. There is no sensible biochemical reason for why fluoride would harm the brains of boys but not those of girls.

A request from 30 science and health experts to the National Institute of Environmental Health Sciences for the release of the Green et al. study data for independent analysis based on the fact that “In recent weeks, a number of experts in epidemiology, psychology, statistical methodology and other fields have raised numerous concerns about the Green article, including the following:
1. Focusing on a subgroup analysis amid “noisy data”:
2. Modeling and variable anomalies:
3. Lacking data on relevant factors that can impact children’s intelligence and cognitive ability:|
4. Omitting crucial findings:
5. Using invalid measures to determine individual exposures:
6. Defining the final study group:
7. Assessing the impact of fluoride exposure:
8. Reporting anomalies:
9. Internal inconsistency of outcomes:
10. Overlooking research that conflicts with the authors’ conclusions:

CADTH RAPID RESPONSE REPORT: Community Water Fluoridation Exposure: A Review of Neurological and Cognitive Effects (10/23/19) “Key Findings: This review identified one prospective birth cohort study13 examining the association between fluoride exposure of mothers during pregnancy and subsequent children’s intelligence quotient scores at age 3 to 4 years. Both unadjusted and adjusted estimates showed no significant association between an increase of 1 mg/L in mother urine fluoride and Full Scale intelligence quotient score in the total sample of boys and girls, or in girls. Adjusted estimates also showed no statistically significant association between an increase of 1 mg/L in mother urine fluoride and performance intelligence quotient or verbal intelligence quotient in all children. In boys, every 1 mg/L increased in mothers’ urine fluoride levels was associated with a 4.49 point lower intelligence quotient score. Every 1 mg increase in daily fluoride intake of mothers corresponded with 3.66 points lower in total children’s intelligence quotient score. The interaction between child sex and maternal fluoride intake was not statistically significant. The evidence is weak due to multiple limitations (e.g., non-homogeneous distribution of data, potential errors and biases in the estimation of maternal fluoride exposure and in IQ measurement, uncontrolled potential important confounding factors); therefore, the findings of this study should be interpreted with caution.

American Dental Association News: Responses to fluoride study flood in from all over the globe :
The American Dental Association remains committed to fluoridation of public water supplies as the single most effective public health measure to help prevent tooth decay,”  One of the study’s authors, Dr. Angeles Martinez-Mier, chair of cariology, operative dentistry and dental public health at the Indiana University School of Dentistry, said while she stands “fully behind our study’s conclusions, as an individual, I am happy to go on the record to say that I continue to support water fluoridation.”

American Academy of Pediatrics continues to recommend fluoride following new study on maternal intake and child IQ
The Academy continues to recommend children drink fluoridated tap water despite a new study linking fluoride intake among pregnant women with a small dip in their children’s IQ.

American Dental Hygienists’ Association Statement Regarding a Study Linking Maternal Fluoride Exposure during Pregnancy and the IQ Scores of Their Children: 
In reviewing this study, it is important to consider the following limitations identified by the authors:
• The timing of the urine sampling may not have precisely represented fetal exposure throughout the pregnancy.
• Maternal IQ was not taken into consideration.
• The reported fluoride intake did not measure the actual concentration in the tap water in the participant’s home. Fluoride concentration was estimated based on the postal code; some locations could be receiving water from multiple treatment plants.
• The estimated fluoride intake only considered beverages; dietary sources and oral hygiene products such as toothpaste were not considered. 
• Information on the consumption of tap water and other water-based beverages was obtained via a self-report, non-validated questionnaire that may be subject to recall bias. 
• The method for estimating maternal fluoride intake has not been validated.  
• Post-natal fluoride exposure or consumption was not assessed.”

Podcast by Dr. Christopher Labos (cardiologist with a degree in epidemiology) and Johathan Jarry (a science communicator) includes “a comprehensive look at the [Green, et al.] fluoride-and-IQ study that came out” that begins 30 minutes into the discussion:
Some of the things that we have discovered over time have been done with observational research. So the fact that it’s observational research in and of itself doesn’t make it wrong, but you have to realize the limitations. The link between smoking and lung cancer came out of observational research.   But it came out of observational  research  that was replicated many times over in many different populations and eventually became incontrovertible.    And you can also get similar evidence using animal models – convergence of evidence, it all moves in the same direction.  Because with an observational study you see ‘A’ is associated with ’B’:  ‘A’ could be causing ‘B’;  ‘B’ could be causing A; it could be a random association; or there could be some ‘C’ factor that is causing both ‘A’ and ‘B’.”
Listing of study limitations: a relatively small sample size, urinary fluoride level issues, potential recall bias on fluoride/water consumption reporting; a  statistical drop in IQ was found only in boys (but girls’ IQ increased slightly); the small decrease in boys’ IQ was only found in performance IQ, not verbal IQ (which went up); the scatter plots of IQ/fluoride data look almost random and showed no obvious relationship; parent’s IQ was not tested; fluoride consumption of children after birth was not measured; no other IQ/fluoride studies have reported an IQ difference between boys and girl; many other possible confounding factors that might affect young children’s IQ were not measured or included in the analysis.
As we go through all of these issues, you will notice none of them are terrible, but when you add all of them up, we could really be off by a lot.  It’s a question of certainty – how certain are you this is true.   The minute you do something you know is going to have this huge media response, do you have a responsibility to be more certain.  You have to understand your research has consequences.  If you are going to do a study about how fluoride is affecting brain development of children, maybe you could be a bit more certain before you publish it because you know you are going to induce fear in the population.

Earlier article related to potential dangers of subgroup analysis:
Analysis of subgroup results in a clinical trial is surprisingly unreliable, even in a large trial. This is the result of a combination of reduced statistical power, increased variance and the play of chance. Reliance on such analyses is likely to be more erroneous, and hence harmful, than application of the overall proportional (or relative) result in the whole trial to the estimate of absolute risk in that subgroup. Plausible explanations can usually be found for effects that are, in reality, simply due to the play of chance. When clinicians believe such subgroup analyses, there is a real danger of harm to the individual patient.  - Debate: Subgroup analyses in clinical trials: fun to look at - but don't believe them!

Bruce Y Lee, MD, Senior Contributor@Forbes:
If you think these findings prove that fluoride in drinking water leads to lower intelligence, remember that associations do not mean that one thing causes another. Otherwise, we'd be urging Nicholas Cage to stop starring in movies to prevent drownings in swimming pools. Observational studies, like this JAMA Pediatrics one, cannot, cannot, cannot, cannot prove cause and effect. People who live in areas where tap water is fluoridated could also be exposed to other things or have behaviors that may affect their children's IQ scores. For example, could they or their children be eating more processed foods with artificial ingredients? Could they be exposed to more chemicals in the environment? We do not know enough about the details of the study participants' lives to know what really is happening.”

NHS UK - No proof that a mother's intake of fluoride in pregnancy affects their child's IQ:
"The findings need to be interpreted carefully. First, this is an observational study that cannot prove that the mother's fluoride exposure in pregnancy is directly responsible for the child's later IQ. Many hereditary, environmental and lifestyle factors could influence the child's IQ. Although the researchers have tried to adjust for potential confounders, it's very difficult to account for all the things that may be having an influence."  

Dental Health Services Victoria strongly supports health benefits of water fluoridation:
While the researchers did account for some confounding factors which could impact IQ scores, such as maternal education, employment and certain health conditions, many other factors known to impact brain development were not accounted for by the study, such as maternal alcohol intake, breastfeeding practices and nutritional intake in the early years of life. The accuracy of the methods used by the researchers to estimate fluoride intake are also untested, as no external validation of these methods has been carried out.  It is important to consider this new study in line with the overwhelming weight of existing scientific evidence that shows fluoride is safe and beneficial in reducing tooth decay.  Extensive evidence shows there is no connection between levels of fluoride in drinking water and IQ in children. Two recent studies in New Zealand and Sweden looked at the same question as the Canadian researchers and found no association between water fluoridation and IQ scores.”

The Right Chemistry: Fluoride-IQ link doesn't appear significant
When it comes to risk factors that can affect a fetus, I would rank tobacco smoke, alcohol consumption, air pollution and exposure to lead, mercury, cadmium and various potentially endocrine disrupting chemicals such as phthalates well ahead of worries about fluoride.” (Joe Schwarcz is director of McGill University’s Office for Science & Society)

Science-Based Medicine: Maternal Fluoride and IQ – The Scientific Community Pushes Back – Discussion of the Green, et al. study concerns and the request for independent analysis of the study data. (Dr. Grant Ritchey) "The NIEHS provided funding for the Green et al. study, and we felt it was imperative to express our concerns to this body, not because the conclusions did not meet with our approval, but because there was a lack of transparency in the data they used as well as fundamental statistical errors in analyzing the data. Our letter to the NIEHS: …[requested] that NIEHS formally ask the Green authors to release the HIPAA-compliant, Research Identifiable File (RIF) data sets from their study, as well as a complete explanation of their methods and the computer program/codes used in their data management and analysis. As of this post no release of data has been forthcoming..."

Science-Based Medicine: Maternal Fluoride and IQ: “Further, when you have to do subgroup analysis in order to find significant results that is a red flag for noisy data. There are other red flags as well, like the huge variance in results, and the disconnect between performance and verbal IQ. So we have noisy correlational data at odds with prior research. Perhaps even more damning is that the negative results were driven largely by a small number of boys who had extreme levels of exposure. Take out the outliers, and the overall effect becomes non-significant.” (Steven Novella)

There are many extremely important, potential factors that can significantly contribute to a child’s IQ that apparently were not collected, analyzed and included in the analysis?  Without considering any of the possible factors that can have an important impact on a child's IQ, why would any extremely low (boys-only) potential association of IQ with low levels of exposure to fluoride be of any value?

Some of these important, ignored variables include:

  1. Parental IQ:IQ and education for both parents were statistically correlated to child IQ. However, paternal IQ and education were not significant after accounting for maternal IQ effects.” (Meador, et al., Epilepsy Behav. 2011) “IQ correlations – father-child, 0.51; mother-child, 0.55. So intelligence clearly has a powerful genetic component.”(Intelligence and IQ, Dr. C. George Boeree)
  2.  Diet and nutrition of mothers during pregnancy & Diet and nutrition of children after birth:After adjusting for maternal education, socioeconomic status, the Home Observation for Measurement of the Environment (HOME) score, and total caloric intake, the good prenatal and childhood nutrition indices predicted more favorable neurodevelopment, while both poor nutrition indices predicted poorer neurodevelopment. These associations were stronger in prenatal than childhood models.” (Malin, et al., Nutrients 2018)
  3.  Birth weight and early-life weight:Within this cohort of typically developing children, early-life weight status was inversely associated with children's perceptual reasoning and working memory scores and possibly with full-scale intelligent quotient scores.” (Li, et al., Obesity, 2018)
  4.  Length of time breast-feeding:Breastfeeding duration of 1 month or shorter compared with longer periods was associated with approximately three points lower IQ… In conclusion, in this large sample with high quality assessment of child IQ, we found support of a beneficial association for breast feeding with child IQ at 5 years of age, while adjusting for maternal IQ and parental education, which only few previous studies have been able to do.” (Strøm, et al., BMJ Open 2019)
  5.  Amount of time parents spent with child: Using a large longitudinal British dataset, I show that paternal involvement in childhood has positive associations with offspring IQ at age 11, and offspring social mobility by age 42, though not with numbers of grandchildren. For IQ, there is an interaction between father's socioeconomic status socioeconomic status (SES) and his level of involvement, with high-SES fathers making more difference to the child's IQ by their investment than low-SES fathers do.” (Evolution and Human Behaviour, 2008)
  6.  “Maternal education, household income, parents’ skin color, duration of breastfeeding, head circumference and number of siblings were the most powerful predictors of low IQ at age six. Of a broad set of potential social and biological predictors explored those essentially social were the most impactful ones, which could mean that a high proportion of these children may require intervention.”  "Another major finding of this study is the effect of growth, nutrition, and breastfeeding during the first year of life on cognitive ability. Children who were breastfed for a longer period were less likely to have low IQ than those who were not breastfed." "Other predictors of low IQ were smoking during pregnancy and maternal perception of the child’s health." (Camargo-Figuera, et al., BMC Pediatr.)
  7.  Amount of time each child spent in front of TV or playing electronic games:In multivariate models, being a third-born or more child and watching television ≥1 h/day at 24 months were negatively associated with all IQ scores…” (Plancoulaine, et al., Sleep Med. 2017)
  8.  Many other factors were ignored that influence intelligence and cognitive development in infants and young children: that influence intelligence and cognitive development in infants and young children, for example: Amount of time each child spent outside playing; Amount of time each child spent socializing with others; any environmental stress, genetics, etc.
    “The IQ of an individual is multifactorial and is determined by a multitude of factors. Nature and nuture work together in determining human intelligence. Even though the genetic susceptibility plays a crucial role on the IQ of the individual, various modifiable environmental factors like education, premature birth, nutrition, pollution, drug and alcohol abuse, mental illnesses, and diseases can have an influence on an individual’s IQ. These modifiable factors can reinforce or weaken genetic susceptibility.” (Arun Oommen, Journal of Neurology & Stroke, 2014)
Almost none of the potential factors described above (and many others) that can significantly affect IQ were considered in the Green, et al. study.  If the Green, et al. study had actually been designed to find a legitimate, potential association between prenatal and early childhood exposure to low levels of fluoride, and IQ, then data on these (and other) extremely critical factors which have known negative effects on IQ and cognitive development should have been collected and analyzed in addition to the several factors mentioned in the report.

As noted above, there are a number of criticisms of the statistical methods used in they Green, et al. study which cast serious doubt on any conclusions that there was an actual, relevant association (or correlation) between fluoride exposure (at levels expected from community water fluoridation) and IQ. 

Bottom Line:  Trust the Majority of Relevant Science Experts instead of accepting the interpretation of vocal outliers who, after over 75 years of desperately trying, are still unable to provide ANY relevant, quality studies sufficient to change the scientific consensus that community water fluoridation is a safe and effective public health measure for reducing the risk of dental decay and related health issues. 

Below are graphs from the Green, et al. study that compared various measurements of prenatal maternal fluoride exposure with children's IQ at 3 - 4 years old (regression lines removed).  Along the x-axis are the measured fluoride exposure levels, and along the y-axis are the measured IQ scores.  Each dot represents the IQ of a child plotted against the prenatal fluoride exposure of its mother. 

The highlighted dot in graph A below shows a female child with a measured FSIQ of about 142 and the mother's average urinary fluoride concentration (from three samples during pregnancy) was near 0.8 mg/L.  If you follow the 142 IQ line to the left, you will see several children with similar IQs with mothers who had lower average urinary fluoride concentrations, and if you follow the 142 IQ line to the right, you will see several girls and a boy with IQs above 125 who had average urinary fluoride concentrations up to about 2.0 mg/L.

If you follow the 0.8 mg/L down, you will find a large scatter of boys and girls with IQs between about 85 and 125.  At the very bottom of the 0.8 mg/L line are the records of two boys with IQs between about 50 and  60 whose mothers' average urinary fluoride concentrations were between 0.6 and 0.8 mg/L.  Is it reasonable to assume this level of fluoride exposure would cause such a drop in IQ?

Carefully examine the scatter plot showing a huge variation in how the IQ of children is actually associated (correlated) with an average maternal fluoride level from just three urine samples collected during about 270 days of pregnancy.  Then consider what dozens of other graphs might look like that showed possible relationships between a child's IQ and the many other factors described above that have been shown to be associated with IQ development.  An example is shown below.  The second graph (B) also shows a huge variation between fluoride exposure and IQ. 

The wide scatter alone is sufficient to cast serious doubt on the validity and/or relevance of any conclusions without even considering the other study limitations described above by the experts.  When a  one sees scatter graphs like those in the Green, et al. study where it looks as though the data points are randomly scattered, one should consider any alleged association between the two variables being studied with extreme skepticism.

Scatter Plot Green Iq, Fluoride Study

Monkey/Shotgun Scatter

For some examples of how a graph can be constructed to "prove" almost any association or correlation I went to a site that lists average IQ by country along with other factors that can be demonstrated to be associated/correlated with IQ.  The Green, et al. study claimed that a 1-mg/L increase in Maternal Urinary Fluoride level was associated with a 4.49-point lower IQ score in boys.  One can easily see from the graphs below that a $20,000 increase in average income is associated with a 10 point increase in IQ.  Extremely concerning, however, is the graph that clearly shows an 8 degree increase in average maximum temperature is associated with a 10 point drop in IQ.

The main point of this demonstration is to show that finding an association/correlation between two factors - even a statistically significant association - does not prove there is any actual cause and effect relationship between the two factors.  The income/IQ association might demonstrate a legitimate cause and effect association, but it could be that a higher IQ causes a higher average income.  The temperature/IQ association  is probably due to a number of other causes.  For additional examples of how relationships between different factors can be presented search on "correlation does not prove causation" or go to:

IQ by Country