Tag Archives: statistical analysis

Yet another fluoride-IQ study

As with most of these fluoride-IQ studies this one is only relevant to areas of endemic fluorosis (This is from a UNESCO paper and has been corrected for New Zealand. Identification of fluorosis in a country does not imply the whole country is high fluoride).

Yes, it’s a bit like groundhog day. Another fluoride-IQ study – and we expect this to be followed by another round of claims by anti-fluoride propagandists that this is the death knell to community water fluoridation. That this study provides the “irrefutable proof” that fluoride is a “neurotoxin.”

But that interpretation is completely wrong. This new study does nothing of the sort – in fact, quite the opposite.

The new study is:

Duan, Q., Jiao, J., Chen, X., & Wang, X. (2017). Association between water fluoride and the level of children’s intelligence: a dose-response meta-analysis https://doi.org/10.1016/j.puhe.2017.08.013

Now, why is this study absolutely useless for those opposing community water fluoridation?

It is not relevant to community water fluoridation

Because it is about a problem in areas of endemic fluorosis – where fluoride dietary intakes are much higher than where community water fluoridation exists.

From its first sentence it concentrates on fluorosis:

“Fluorosis is a progressive degenerative disease that causes skeletal fluorosis and dental fluorosis.”


“Currently, about 500 million people are exposed to environments high in fluoride content, while the incidence of fluorosis has already reached 200 million people worldwide.”

It’s not new research – it’s a meta-analysis of existing studies. Only studies dealing with areas of endemic fluorosis are considered in the meta-analysis. For example, the New Zealand (Broadbent et al., 2014) and Canadian (Barberio et a., 2017) papers which actually studied effects on IQ of community water fluoridation are not included. Nor is the Swedish study (Aggeborn & Öhman, 2016) which considered drinking water fluoride concentrations similar to that used in community water fluoridation.  So far these are the only reliable studies which considered low fluoride concentrations and they all show no effect of fluoride on IQ.

It is concerned with health effects in areas of endemic fluorosis

The meta-analysis includes 26 published studies in the meta-analysis. Most of the papers refer to “high fluoride water,” “fluorosis areas,” “endemic fluorosis” or similar terms in their titles. Low fluoride areas were only considered in the studies as “controls” and studies from areas of community water fluoridation were excluded.

Most of the considered studies simply compared IQ levels in “low fluoride” areas and “high fluoride” areas.  The mean drinking water fluoride concentration in the low fluoride levels of these studies was 0.6 mg/L (0.25 – 1.03 mg/L) and in the high fluoride areas, the mean drinking water concentration was  3.7 mg/L (0.8 – 11 mg/L).

As you can see the control or low fluoride areas, where the studies assumed there were no effects on IQ, have drinking water concentrations similar to that used in community water fluoridation (usually about 0.7 or 0.8 mg/L).

Yes, these studies did show statistically significant differences in IQ levels between the low and high fluoride areas. This is something for health authorities in areas of endemic fluorosis to be concerned about. And this, together with a range of other known health effects of excessive dietary fluoride intake, is the reason why attempts are made to reduce the fluoride levels in drinking water supplies in those areas.

People in high fluoride areas where fluorosis is endemic suffer a range of health problems. Credit: Xiang (2014)

Duan et al (2018) were able to present an overall estimate of the IQ difference between high and low fluoride areas – see figure. This is expressed as a standardised mean difference (SMD) – a necessary measure for a meta-analysis of a range of studies with different variability. The SMD = (difference in mean outcome between groups/standard deviation of outcome among participants) (see Cochrane Handbook).

All of the studies show a lower IQ in high fluoride areas than in low fluoride areas with the overall SMD being 0.52 (-0.62, -0.42 95% confidence interval).

To be clear – this is not 0.52 IQ points but can be interpreted as 0.52 x the standard deviation of IQ  in a population. Unfortunately, the authors do nothing to explain this, leaving readers to make the same mistake many did with a previous IQ meta-study (see Did the Royal Society get it wrong about fluoridation?).

Attempt to derive a dose-response relationship

The authors went on to attempt to derive an overall response curve relating SMD to drinking water fluoride concentration. Unfortunately, their results as presented in their  Fig 4 are confusing and the figure is not properly explained. Also, the modeling methods used to derive the response curve is not well explained.

However, the linear relationship they derived was not statistically significant. (They were able to derive a significant non-linear relationship, but again their methods and reason for doing this were not explained.)

I got the relationship shown in the figure below using the data provided in the paper without further modeling. This relationship is also not statistically significant (p=0.77).

The authors do suggest the possibility that lower intelligence may be associated with medium fluoride concentrations and “that very high fluoride concentration in water was associated with higher intelligence level than
medium fluoride.” However, although the figure above implies that IQ increases at higher fluoride concentrations, I do not think such conclusions are warranted with this data and its variability.

What causes the cognitive deficits?

Authors of these studies often seem to assume a direct chemical fluoride toxicity cause for the cognitive deficit. That also appears to be an assumption behind the desire to produce a dose-response relationship. Of course, anti-fluoride propagandists also prefer this mechanism because it enables them to argue that the effects also occur at low concentrations – they just haven’t been measured yet.

Although a dose-response relationship would be expected for a chemical toxicity mechanism this study did not produce a reasonable dose-response relationship. Some individual studies have claimed such a relationship but these claims are often not supported or the reported relationship is of only minor significance (see my discussion of Xiang et al., 2003 in Perrott, 2018).

The poor or non-existent relationship of cognitive deficits to drinking water fluoride concentration makes me suspect that there is not a direct effect. Rather the real causes of the cognitive deficits observed are dental or skeletal fluorosis or other health effects common in areas of endemic fluorosis. I suggested this in comments on Choi et al.,(2015) who observed a relationship with severe dental fluorosis but not water concentration (see Perrott 2015 – Severe dental fluorosis and cognitive deficits).

There I suggested consideration of the effects of severe dental fluorosis on quality of life and learning difficulties on cognitive deficits.  Another factor could be premature births and low birth weights which are known to influence cognitive development (see Premature births a factor in cognitive deficits observed in areas of endemic fluorosis?)

Duan et al., (2018) in their paper also allude to such possible mechanisms:

“Skeletal fluorosis is another very common and very serious side-effect of high fluoride intake, characterized by changes in the bone density, skeletal deformation, rickets, paralysis, disability, and even death. Patients with skeletal fluorosis have been reported to show neuronal nuclear vacuoles formations, cell loss in the spinal cord, and loss or solidification of Nissl bodies. Moreover, patients experience fatigue, sleepiness, headache, dizziness, and other symptoms related to the nervous system.”


The meta-analysis does confirm that there may be a problem with reduced of intelligence in children in areas of endemic fluorosis. This difference in IQ levels between high and low water fluoride levels is statistically significant.

However, this finding is of absolutely no relevance to community water fluoridation where the drinking water levels are similar to that in the low fluoride areas in the studies used for the meta-analysis.

The summarised data does not appear to be of sufficient quality to determine a reliable dose-response relationship. At least, the derived relationships are not statistically significant. An alternative explanation is that the observed reduced intelligence may not be directly related to drinking water concentration and instead related to dental or skeletal fluorosis, or other health effects common in areas of endemic fluorosis.

Similar articles