Recently, I wrote an article about the published study that seemed to show that fluoride lowers IQ in children. I predicted that the anti-vaccine religion would jump all over it as an indication that public health authorities are useless – it took like an hour before the anti-vaxxers employed all of their useless pseudoscience and conspiracy theories to jump all over that.
I was somewhat ambivalent about the research, but I thought it might be an indicator that there could be an issue with water fluoridation. However, upon further review, I’ve come to the realization that this was pretty much worthless.
So do we know if fluoride lowers IQ? Based on research published in JAMA Pediatrics, no we don’t. In fact, I’m willing to state that we should be confident that fluoridated water is still safe while preventing cavities.
Let’s take a look at what I’ve noticed.
Fluoride lowers IQ – the paper
As I mentioned, I reviewed this paper just a few days ago, here’s a quick summary of the research.
The article, by Christine Till Ph.D. of the Department of Psychology, York University, Toronto, was published in JAMA Pediatrics in August 2019. This Canadian study examined whether higher fluoride intake during pregnancy was associated with IQ scores in children.
The researchers looked at 512 mother-child pairs and found that each 1 mg/L increase in fluoride concentration maternal urine seemed to be linked with a decrease of 4.49 Full-Scale IQ points in boys, aged 3-4 years. The results were adjusted for demographic, quality-of-home, and some, but not all, confounders.
Of course, as a result of this study, there has been a lot of attention-grabbing headlines about fluoride in our water supply and its link to the IQ of children. But I’m not seeing a lot of in-depth criticism of that study, so that’s what I’m going to try to do with a long list of issues make the conclusions presented by Dr. Till as problematic. In other words, I am unconvinced that fluoride reduces IQ of children.
Critique of fluoride lowers IQ paper
- Sample size. This is my first reaction to this study – if we’re going to make a broad condemnation of an important public health issue, then it better be with more than an n=512 sample size. That’s tiny in the world of epidemiology, where many studies include hundreds of thousands or even millions of data points. This is how we know that the HPV vaccine is safe – not from a 512 person study, but from huge studies that included +2 million patients. The problem with such a small sample size, in this case, is that “noise” in the data can be amplified by just a few data outliers.
- IQ measurement statistics. I mentioned that the authors seemed to show that the IQ dropped for boys but increased for girls. But the study wasn’t set up to determine a gender difference, and when you look at the total data, there was a non-significant drop in IQ when the data were pooled (which the authors conveniently ignored in the abstract). The authors performed a “subgroup analysis” where they tried to dig into the data to find something, anything, that would make this study interesting. This is an inappropriate methodology for epidemiological research for a variety of reasons, the most important of which, if you do enough digging, you will always find a subgroup that shows you something.
- Gender difference. If fluoride was really toxic, it’s difficult to determine a biologically plausible reason why there would be a gender difference in IQ. Moreover, there was not a verbal IQ decrease in boys or girls. This is beginning to look more and more like random differences that are not related to fluoride.
- More on gender differences. Oddly, the males included two boys with extremely low IQs – they were in the 50s. In education, IQs below 70 would indicate special needs. In such a small sample size, the dramatically low scores would have a substantial effect on the average IQ. In fact, if the authors had eliminated those two individuals from the data, then reanalyzed the data and still showed a difference, it would be more impressive. My guess is that they did, and they found that the results became random noise.
- Study design and gender. This is a troubling issue. Apparently (or at least it’s not very clear), the researchers did not pre-register participants for the study. Without this, it’s clear that they made a decision post hoc to check gender differences in IQ, which makes their study troublesome. If they had designed the study upfront to examine gender differences, it may have made us more comfortable with the gender differences.
- More IQ issues. A 4.49 point IQ drop is large enough that we would have seen it in other studies. As I mentioned in my previous review of this research, a study in New Zealand showed no difference in IQ.
- The randomness of water fluoridation. This is a critical point that I missed the first time around – high fluoride levels are not random, so there are a huge number of confounders that will influence the data. For example, are there certain other issues that may account for why a person would live in a high or low fluoride area (apart from knowing if there’s fluoride in the water, something few people check)? Not knowing these confounders immediately adds bias and noise to the data.
- They didn’t find the right confounders. Interestingly, in their data, it appears that adjusting for confounders didn’t really change the data much. That implies that the weren’t adjusting for all of the confounders, which makes their data even more suspect from a statistical standpoint.
- No information on the actual amount of fluoride consumed. The researchers measured urinary fluoride in pregnant women once each trimester as a proxy for total fluoride consumption. I cannot stress this enough – this is nearly useless data. The half-life of fluoride in the body is around 5 hours. So, it’s not a measurement of everything that the woman consumed over the past 3 months, but only within a few hours. She could have brushed her teeth, used mouthwash, or any number of activities that would raise or lower the urinary concentration of fluoride. It just doesn’t give us any information whatsoever.
- Urinary fluoride has not been validated as a proxy. Even the authors admit this – urinary fluoride, especially only taken once every three months, has not been established to be evidence of what is consumed. This is a major flaw of the article.
- The actual amount of fluoride consumed was small if we are to trust their urinary concentration as a proxy. In fact, most of the participants showed that they consumed much less than 1 mg/L of fluoride. In other words, the authors concluded that fluoride lowers IQ based on most individuals who did not even reach the threshold of 1 mg/L. Most of the very noisy data seem to arise from outliers, which makes it even more statistically troublesome.
- Post-natal fluoride. Although I am unconvinced that the authors showed that fluoride consumption during pregnancy had any effect on IQ, if it were true, it would also be observed in children who drank this water from birth to 4 years old. And if that were true, then the IQ differential would potentially be much larger. It isn’t. Not determining whether pre-natal or post-natal fluoride had an effect is a major concern with this data.
- Exposure to substances other than fluoride. Although the authors focused on fluoride and tried to control for substances like lead, they failed to examine all of the other stuff in the water – pesticides, other natural salts, organic material, or just about anything else. Again, without this data, we have no information as to whether it was fluoride or anything else on the planet.
- Observational study. The women were not separated by particular levels of fluoride. They just observed the IQ of the children and tried (poorly) to determine fluoride intake.
- Statistics. I am not a big fan of p-values, but it is where we are at with biomedical research. It can be abused, and there’s no black line after which the data is “perfect.” In this case, the p-value is 0.04 which is slightly below the 0.05 threshold where scientists will declare “significance.” Moreover, it appears (or we can’t tell) how many other hypotheses where they ran a p-value test. This gives the appearance that they were looking for results by running a bunch of statistical tests, rather than establish a hypothesis and determine whether the statistics supported or rejected that hypothesis. I know this sounds picky, but it’s a part of statistical analysis that ALL scientists use. But importantly, a p-value of 0.04 isn’t something to scream out “fluoride lowers IQ.”
- Average IQ in children in fluoridated and non-fluoridated groups. The results were 108.07 vs. 108.21, statistically exactly the same. Yes, the most basic analysis they could have done, separating the children into a simple binary set of groups – no fluoride vs fluoride – show no difference. If they had written the title of their paper as “No IQ differences in children from mothers who consumed fluoridated water versus those who did not. An observational study,” I would have read it, and just saved it for a day that I needed to show that fluoridated water was safe. Because that’s what their data say.
- Statistical errors. Although IQ tests can be accurate, there is still an error around the mean depending on the age of the participant, the tester, and the environment. There’s not a definitive blood test for IQ, it is a fairly subjective measurement. Then there’s the measurement of urine fluoride which can vary for any number of issues – more error. So what happens is that there is a fairly complicated statistic measurement that has error piled upon error – this makes a small study even more prone to improper conclusions.
Sorry vaccine deniers, this doesn’t say what you think it says
I know that the usual suspects in the anti-vaccine groups are using this study to “prove” that the CDC, WHO and anyone else who advocates for children’s health through vaccinations are misguided and wrong, just because they advocated for fluoride in water.
Well, they ought not to rely upon this study as a “definitive” proof that fluoride lowers IQ. Because this study does not actually serve as robust, repeated, and conclusive evidence of any issue with fluoridated water.
So, I was wrong by giving this study a slight pass and suggest that maybe, out of an abundance of caution, there should be a public health discussion about fluoride. Well, if there is a reason to do so, this study isn’t it. It’s almost useless, and it makes me wonder why JAMA Pediatrics published it. Well, good journals have published nonsense before (see Andrew Wakefield).
But like Wakefield and autism, this bad study will be used by the anti-health crackpots to block water fluoridation throughout the world. And they’ll use it to attack vaccines.
- Broadbent JM, Thomson WM, Ramrakha S, Moffitt TE, Zeng J, Foster Page LA, Poulton R. Community Water Fluoridation and Intelligence: Prospective Study in New Zealand. Am J Public Health. 2015 Jan;105(1):72-76. doi: 10.2105/AJPH.2013.301857. PubMed PMID: 24832151; PubMed Central PMCID: PMC4265943.
- Green R, Lanphear B, Hornung R, Flora D, Martinez-Mier EA, Neufeld R, Ayotte P, Muckle G, Till C. Association Between Maternal Fluoride Exposure During Pregnancy and IQ Scores in Offspring in Canada. JAMA Pediatr. 2019 Aug 19;. doi: 10.1001/jamapediatrics.2019.1729. [Epub ahead of print] PubMed PMID: 31424532.