2014 NZ Fluoride IQ study that Govt relies on is seriously flawed

The office of the Prime Minister’s Chief Science Advisor and the Director General of Health, Dr Diana Sarfarti, have dismissed the growing evidence from around the world that fluoride is a neurotoxin, preferring instead to rely on the findings from a very low quality New Zealand study.

The Broadbent study was published in the American Journal of Public Health in 2015. This research aimed to clarify the relationship between community water fluoridation and IQ using a prospective study design with data from the Dunedin Multidisciplinary Health and Development Study. The authors concluded that there were no differences in the IQ of children as a result of consuming fluoridated water after adjustment for potential confounding variables. However, there have been many criticisms raised about the validity of their findings. 

Criticisms of the study include:

1) Size difference of Groups

The non-fluoridated water group (n=99) was very small relative to the fluoridated water group (n=891). This disparity of numbers raises the question of whether the study had sufficient statistical power to detect differences in IQ between the two exposure groups. Questions of potential bias have also been raised, particularly as it wasn’t clear how many of the children living in the non-fluoridated areas were taking fluoride tablets and using fluoride toothpastes.

2) Other sources not controlled for

Data about other sources of fluoride exposure, e.g. from tablets and toothpastes, were collected but these were not controlled for in the models looking at the effect of community water fluoridation on IQ.

3) Fluoridation status of children who had taken fluoride tablets unknown

Although it was stated some children took fluoride tablets it was not made clear which ones. The children living in non-fluoridated areas would have been more likely to have been given fluoride tablets at the time. With each tablet providing 0.5 mg fluoride, intakes may not have been greatly different between the groups, making the likelihood of a null finding more likely.

4) Water consumption unknown

The authors did not consider how much water the children were drinking, nor if families were drinking all or some of their water from other sources (with or without fluoride), so there was no indication of the fluoride dose children were getting. Water consumption in childhood can vary considerably depending on age, sex, weight, whether a child is being breast fed and for how long, diet, activity levels, air temperature and humidity. In general, children drink more fluid relative to their body weight, so would be consuming more fluoride per kg than an adult. Results would have been more meaningful if total fluoride exposure had been calculated and an association between this and IQ investigated. Other researchers have estimated the likely total fluoride intakes between the two exposure groups in the study using publicly available data and found very little difference. They concluded that Broadbent et al (2015) could not have found a difference in IQ even if total fluoride had been measured!

5) Uses data from a study not designed to look at fluoride exposure

The Broadbent study is using data collected by the Dunedin Multidisciplinary Health and Development Study which had not been specifically designed to look at the issue of fluoride exposure when originally conceived. Had this been the case, more useful questions about fluoride exposure may have been asked. For instance, families were asked if children had “ever” or “never” taken fluoride supplements. Without further questions, this meant that a child who had only ever taken one fluoride tablet would be in the same category as those who consumed them every day. Designing questions without knowing how the data will be used in the future can be a limitation in longitudinal studies. Findings therefore need careful interpretation so as not to mislead.

6) Authors aggregated data for no apparent reason

The authors aggregated IQ scores at 7, 9, 11 and 13 years for each individual. It is not clear why this was done and whether the authors also looked at the IQ data that was collected at 5 years.

7) Unfluoridated area was known for having corrosive water

Concerns have been raised about the potential for confounding as the majority of the group not exposed to fluoridated water were living in an area known to have the most corrosive water in New Zealand, dissolving high levels of copper and possibly lead, well known as a neurotoxin, from the plumbing. The same water source was found to have high levels of natural manganese, another known neurotoxin.

8) Previous study had found loss of IQ was not considered

High blood lead levels, sufficient to cause a loss of four IQ points, were found in an earlier study of the same children but this was not considered by Broadbent and his colleagues.

9) Protective effect of breast feeding not discussed

A measure of breast feeding (+/- 4 weeks) was included in the analysis and found to be associated with IQ. Breast fed infants are protected from exposure to fluoride as breast milk does not contain fluoride regardless of maternal exposure. The authors did consider an interaction term for breast feeding but this was removed as it didn’t improve the model. No further information was given about this but their data was very limited. Breast feeding is likely to be advantageous for neurodevelopment in many respects, but could also be considered a proxy for fluoride exposure in early infancy. On this basis, it is a pity that more information about breast feeding was not available.

In Summary

This study was considered low-quality by the U.S. Government’s National Toxicology Program because of high risk of bias. It is obvious to anyone that since some of the most basic scenarios, such as which children were taking fluoride tablets, were not considered the study is unsuitable for the Government to rely on.