Model of a fetus in the womb. Photo credit: CP PHOTO/ Alliance Atlantis/ HO)
The maternal urinary fluoride/IQ study (Bashash et al., 2017) continues to get attention – but mainly from anti-fluoride organisations. The scientific community will evaluate the published report after giving it due consideration and there have already been criticisms. But anti-fluoride campaigners consider it the best thing since sliced bread. The Fluoride Action Network (FAN) describes it as “a cannon shot across the bow of the 80 year old practice of artificial fluoridation” and Fluoride Free NZ insist that it “must spell an end to fluoridation in New Zealand.”
We expect confirmation bias from the anti-fluoride organisations. But the misrepresentations in the propaganda from these organisations are of more concern because they are blatantly meant to scaremonger.
Misrepresentation by anti-fluoride organisations
These people have worked hard to stress the respectability of the authors of the Bashash et al., (2017) paper and claim the study is impeccable. They are not interested in a critical analysis of the data and the conclusions. And they are completely silent about the evidence from the study showing no association of children’s urinary fluoride levels and IQ – normally they are quick to criticise authors reporting such a lack of association.
But this time as well as their normal misrepresentations they have actually manipulated a figure from the paper. I wonder what copyright law would say about this.
I provided the relevant figures from the paper my earlier article (see Fluoride, pregnancy and the IQ of offspring) and commented on the large amount of scatter in the data. This scatter should be a warning to any sensible reader – so FAN simply overcomes that problem by deleting the data points in their presentation of the figure.
Here is the original Figure 2 and the FAN misrepresentation of it:
Notice 2 things:
- The original figure showed the data for GCI – general cognitive index. It is not IQ and not presented as IQ in the original paper. But it is a measure of “verbal, perceptual performance, quantitative, memory, and motor abilities of preschool-aged children.” Perhaps a fine point and FAN may be excused for inserting the more popularly understood term IQ. Or perhaps they decided not to use the real figure for IQ (Figure 3A) because it implied no effect at normal urinary fluoride levels (see figure 3A in Fluoride, pregnancy and the IQ of offspring);
- FAN removed all the data points in their presentation of the figure. I am sure FAN would argue this was to “simplify” the figure. But in doing so they have removed what is the most important information in Figure 2 – the wide scatter of the data points. That scatter suggests that even though the reported association is “statistically significant” it explains very little of the observed variation and is therefore not important (and may not even be real).
Association of maternal urinary F with child IQ poor and probably misleading
In Fluoride, pregnancy and the IQ of offspring I estimated that “the reported relationships with maternal urinary fluoride could explain no more than a few percent of the variation in the data.” Purely an estimation because I did not have the data to analyse myself and the authors did not give the relevant statistical information.
I have since used a plot digitiser programme to extract the data for these figures and performed my own statistical analysis.
These are the results:
For Figure 2:
Yes, a “statistically significant” relationship (p = 0.002) but it explains only 3.3% of the variation in GCI (R-squared = 0.033)
For Figure 3A:
Again, “statistically significant” (p = 0.006) but explaining only 3.6% of the variation in IQ (R-squared = 0.0357).
So my estimate was pretty good. And my evaluation is valid:
“In this case, I would expect that other risk-modifying factors that explain the variation more completely could be found. And if these were included in the multiple regressions there may not be any observable relationship with urinary fluoride.”
Considering that this work was unable to explain about 97% of the variation in CGI and IQ I really question its publication. Certainly, scientific evaluations will conclude that this paper should not have any influence on policymakers.
It’s a pity that with all the data the authors had they did not seek out, or properly evaluate, other possible risk-modifying factors.
Other work by group showing no association ignored
Strangely, the Bashash et al., (2017) paper did not include relevant IQ information from the PhD thesis of one of their team Deena B. Thomas. This is her thesis citation:
Thomas, D. B. (2014). Fluoride exposure during pregnancy and its effects on childhood neurobehavior: a study among mother-child pairs from Mexico City, Mexico. PhD thesis, University of Michigan.
It can be downloaded from the full-text link.
The data in chapter 2 of this thesis – Urinary and Plasma Fluoride Levels During Pregnancy and Determinants of Exposure Among Pregnant Women from Mexico City, Mexico – was published. The citation is:
Thomas, D. B., Basu, N., Martinez-Mier, E. A., Sánchez, B. N., Zhang, Z., Liu, Y., … Téllez-Rojo, M. M. (2016). Urinary and plasma fluoride levels in pregnant women from Mexico City. Environmental Research, 150, 489–495.
Bashash et al., (2017) did reference this paper – after all, it dealt with the data they used for estimating fluoride exposure. But they did not reference the thesis itself – and two other chapters in that thesis are directly relevant to the relationship of fluoride exposure to child IQ.
Chapter 3 – Prenatal fluoride exposure and neurobehavior: a prospective study – is directly relevant except that where Bashash et al., (2017) reported data for the children when 4 years old and 6-12 years old Thomas reported data for child neurobehavioral outcomes at ages 1, 2 and 3.
She concluded:
“that maternal intake of fluoride during pregnancy does not have a strong impact on offspring cognitive development in the first three years of life.”
OK – perhaps the difference is purely due to age. But surely it is part of the picture and should at least been mention in the Bashash et al., (2017) discussion.
Chapter 4 – Concurrent Fluoride and Total WASI in 6-15 year old children from Mexico City, Mexico – is also directly relevant because Bashash et al., (2017) did include that data in their paper. They concluded that:
“there was not a clear, statistically significant association between contemporaneous children’s urinary fluoride (CUFsg) and IQ either unadjusted
or adjusting for MUFcr.”
This differs a little from the findings in Thomas’s thesis:
“In the overall population, urinary fluoride appears to have no significant impact on total WASI scores (β =1.32, p=0.33), but this association changes once the models are separated by male and female children. Male children showed a significantly positive trend (β=3.81, p=0.05), and females showing a negative trend that was not significant (β= -1.57, p=0.39).” [WASI score is a measure of IQ]
And she wrote:
“analysis suggests concurrent urinary fluoride exposure has a strong positive impact on cognitive development among males aged 6-15 years.”
She concludes:
“These results were surprising in that they show opposite trends to what has been reported in the literature so, more studies with similar reliable methodology, which account for plasma fluoride, diurnal variations in urinary fluoride and children’s SES, are needed. If these results are substantiated, different fluoride interventions may be needed for male children versus female
children.”
I would have thought these findings and conclusions were worthy of discussion by Bashash et al., (2017). It’s not as if the authors were unaware of their colleague’s findings.
Maybe internal politics are involved. but that does not justify the omission.
Conclusion
The anti-fluoride people, and particularly FAN, are misrepresenting the study and have manipulated a figure to hide information in an unethical way. The data presented in the Bashash et al., (2017) study shows maternal urinary fluoride can only explain 3 – 4 % of the variation in General Cognitive Index and IQ of the children. The inclusion of a more viable risk-modifying factor would probably remove even that small amount explanation. Bashash et al., (2017) also neglected to discuss relevant information from a colleague which contradicted their conclusions.
Open Parachute 