Tag Archives: Bashash

A conference paper on the maternal neonatal urinary fluoride/child IQ study has problems

Image credit: Do new mothers doing a Ph.D. get enough support?

The anti-fluoride movement has certainly mobilised over the neonatal maternal urinary fluoride study which reported an association with child IQ. They see it as the best thing since sliced bread and believe it should lead to the end of fluoridation worldwide.

They also seem to be putting all their eggs in this one basket and have started a campaign aimed at stopping pregnant women from drinking fluoridated water (See Warning to Pregnant Women: Do Not Drink Fluoridated Water).

So I was not surprised to see a newsletter this morning from the Fluoride Action Network reporting another output from this study – a conference paper (most likely a poster) presented at the  3rd Early Career Researchers Conference on Environmental Epidemiology. The meeting was in Freising, Germany, on 19-20 March 2018.

I had been aware of the poster for the last week so had expected FAN to gleefully jump on it and start promoting it in their campaigns.

Here is a link to the abstract:

Thomas, D., Sanchez, B., Peterson, K., Basu, N., Angeles Martinez-Mier, E., Mercado-Garcia, A., … Tellez-Rojo, M. M. (2018). Prenatal fluoride exposure and neurobehavior among children 1-3 years of age in Mexico. Occupational and Environmental Medicine, 75(Suppl 1), A10–A10.

It’s only an abstract and it may be some time before a formal paper is published, if at all. Posters do not get much in the way of peer review and often not followed by formal papers.  So I can’t say much about the poster at this stage as I never like to make an assessment of studies on the basis of abstracts alone.

But, in this case, I have Deena Thomas’s Ph.D. thesis which was the first place the work was reported. If you are interested you can access it from this link:

Thomas, D. B. (2014). Fluoride exposure during pregnancy and its effects on childhood neurobehavior: a study among mother-child pairs from Mexico City, Mexico. University of Michigan.

I will wait for a formal paper before properly critiquing the poster, but at the moment I find a big discrepancy between the Thesis conclusions and the conclusions presented in the poster abstract.

Thesis conclusions

In her work, Deena Thomas used the Mental Development Index (MDI) which is an appropriate way of determining neurobehavioral effects in young children.

She concluded in her thesis (page 37):

“Neither maternal urinary or plasma fluoride was associated with offspring MDI scores”

And (page 38):

“This analysis suggests that maternal intake of fluoride during pregnancy does not have a strong impact on offspring cognitive development in the first three years of life.”

And further (page 48):

“Maternal intake of fluoride during pregnancy does not have any measurable effects on cognition in early life.”

So – no association found of child MDI score with maternal neonatal urinary F concentrations.

Poster conclusions

But the poster tells a different story.

The abstract concluded:

“Our findings add to our team’s recently published report on prenatal fluoride and cognition at ages 4 and 6–12 years by suggesting that higher in utero exposure to F has an adverse impact on offspring cognitive development that can be detected earlier, in the first three years of life.”

So her conclusions reported in her thesis are exactly the opposite of the conclusions reported in her conference poster!

What the hell is going on?

The data

Obviously, I do not have access to the data and she does not provide it in her thesis. But from her descriptions of the data in her thesis and her poster perhaps we can draw some tentative conclusions.

The table below displays the data description, and a description of the best-fit line determined by statistical analysis, in her thesis and her poster.

Information on data Thomas Ph.D. Thesis Conference abstract
Number of mother/child pairs 431 401
Maternal Urinary F range (mg/L) 0.110 – 3.439 0.195 – 3.673
Mean maternal urinary F (mg/L) 0.896 0.835
Model β* -0.631 -2.40
Model p-value 0.391 – Not significant
95% CI for β -4.38 to -0.40

*β is the coefficient, or slope, of the best-fit line

Conclusions

Apparently at least 30 data pairs have been removed from her thesis data to produce the dataset used for her poster. Perhaps even some data pairs were added (the maximum urinary F value is higher in the smaller data set used for the poster).

This sort of change in the data selected for the statistical analysis could easily swing the conclusion from no effect to a statistically significant effect. So the reasons for the changes to the dataset are of special interest.

Paul Connett claims this poster “strengthens” the findings reported in the Bashash paper.  He adds:

“This finding adds strength to the rapidly accumulating evidence that a pregnant woman’s intake of fluoride similar to that from artificially fluoridated water can cause a large loss of IQ in the offspring.”

But this comes only by apparently removing the conflicting conclusions presented in Deela Thomas’s Ph.D. thesis. We are still left with the need to explain this conflict and why a significant section of the data was removed.

To be clear – I am not accusing Thomas et al. (2018) of fiddling the data to get the result they did. Just that, given the different conclusions in her thesis and the poster,  there is a responsibility to explain the changes made to the dataset.

From the limited information presented in the poster abstract, I would think the scatter in the data could be like that seen in the Bashash et al. (2017) paper. The coefficient of the best fit line (β) is relatively small and while the 95% CI indicates the fit is statistically significant its closeness to zero suggest that it is a close thing.

However, let’s look forward to getting better information on this particular study either through correspondence or formal publication of a research paper.

Other articles on the Mexican study

Fluoride, pregnancy and the IQ of offspring,
Maternal urinary fluoride/IQ study – an update,
Anti-fluoridation campaigners often use statistical significance to confirm bias,
Paul Connett “updates” NZ MPs about fluoride?
Paul Connett’s misrepresentation of maternal F exposure study debunked,
Mary Byrne’s criticism is misplaced and avoids the real issues

Similar articles

Advertisements

Mary Byrne’s criticism is misplaced and avoids the real issues

Image credit: BuildGreatMinds.Com

First, thanks to Mary Byrne and FFNZ for this response (see Anti-fluoride group coordinator responds to my article). Hopefully, this will help encourage some good faith scientific discussion of the issues involved in my original article (Paul Connett’s misrepresentation of maternal F exposure study debunked). I am pleased to promote such scientific exchange.

I will deal with the issues Mary raised point by point. But first, let’s correct some misunderstandings. Mary claimed I am a “fluoride promoter” and had “sought to discredit the study via his blog posts and tweets.”

  1. I do not “promote fluoride.” My purpose on this issue has always been to expose the misinformation and distortion of the science surrounding community water fluoridation (CWF). I leave promotion of health policies to the health experts and authorities.
  2. I have not “sought to discredit the study.” The article Mary responded to was a critique of the misrepresentation of that study by Paul Connett – not an attack on the study itself. This might become clear in my discussion below of the study and how it was misrepresented.

The study

The paper we are discussing is:

Bashash, M., Thomas, D., Hu, H., Martinez-mier, E. A., Sanchez, B. N., Basu, N., … Hernández-avila, M. (2016). Prenatal Fluoride Exposure and Cognitive Outcomes in Children at 4 and 6 – 12 Years of Age in Mexico.Environmental Health Perspectives, 1, 1–12.

Anti-fluoride activists have leaped on it to promote their cause – Paul Connett, for example, claimed it should lead to the end of community water fluoridation throughout the world! But this is not the way most researchers, including the paper’s authors, see the study. For example, Dr. Angeles Martinez-Mier, co-author and one of the leading researchers,  wrote this:

1. “As an individual, I am happy to go on the record to say that I continue to support water fluoridation”
2. “If I were pregnant today I would consume fluoridated water, and that if I lived in Mexico I would limit my salt intake.”
3.  “I am involved in this research because I am committed to contribute to the science to ensure fluoridation is safe for all.”

Was the reported association statistically significant?

Mary asserts:

“Perrott claims that the results were not statistically significant but his analysis is incorrect.”

That is just not true. I have never claimed their reported association was not statistically significant.

I extracted the data they presented in their Figures 2 and 3A and performed my own regression analysis on the data. This confirmed that the associations were statistically significant (something I never questioned). The figures below illustrating my analysis were presented in a previous article (Maternal urinary fluoride/IQ study – an update). These results were close to those reported by Bashash et al., (2017).

For Fig. 2:

My comment was – “Yes, a “statistically significant” relationship (p = 0.002) but it explains only 3.3% of the variation in GCI (R-squared = 0.033).”

For Fig 3A:

My comment was – “Again, “statistically significant” (p = 0.006) but explaining only 3.6% of the variation in IQ (R-squared = 0.0357).”

So I in no way disagreed with the study’s conclusions quoted by Mary that:

” higher prenatal fluoride exposure, in the general range of exposures reported for other general population samples of pregnant women and nonpregnant adults, was associated with lower scores on tests of cognitive function in the offspring at age 4 and 6–12 y.”

I agree completely with that conclusion as it is expressed. But what Mary, Paul Connett and all other anti-fluoride activists using this study ignore is the real relevance of this reported association. The fact that it explains only about 3% of the IQ variance. I discussed this in the section The small amount of variance explained in my article.

This is a key issue which should have been clear to any reader or objective attendee of Paul Connett’s meeting where the following slide was presented:

Just look at that scatter. It is clear that the best-fit line explains very little of it.  And the 95% confidence interval for that line (the shaded area) does not represent the data as a whole. The comments on the statistical significance and confidence intervals regarding to the best-fit line do not apply to the data as a whole.

Finally, yes I did write (as Mary quotes) in my introductory summary that “the study has a high degree of uncertainty.” Perhaps I should have been more careful – but my article certainly makes clear that I am referring to the data as a whole – not to the best fit line that Connett and Mary concentrate on. The regression analyses indicate the uncertainty in that data by the low amount of IQ variance explained (the R squared values) and the standard error of the estimate (about 12.9 and 9.9 IQ points for Fig 2 and  Fig 3A respectively).

The elephant in the room – unexplained variance

Despite being glaringly obvious in the scatter, this is completely ignored by Mary, Paul Connett and other anti-fluoride activists using this study. Yet it is important for two reasons:

  • It brings into question the validity of the reported statistically significant association
  • It should not be ignored when attempting to apply these findings to other situations like CWF in New Zealand and the USA.

Paul Connett actually acknowledged (in a comment on his slides) I was correct about the association explaining such small amount of the variance but argued:

  • Other factors will be “essentially random with respect to F exposure,” and
  • The observed relationship will not be changed by the inclusion of these other factors.

I explained in my article Paul Connett’s misrepresentation of maternal F exposure study debunked how both these assumptions were wrong. In particular, using as one example the ADHD-fluoridation study I have discussed elsewhere (see Perrott, 2017). I hope Mary will refer to my article and discussion in her response to this post.

While ignoring the elephant in the room – the high degree of scattering, Mary and others have limited their consideration to the statistical significance and confidence intervals of the reported association – the association which, despite being statistically significant, explains only 3% of the variation (obvious from the slide above.

For example, Mary quotes from the abstract of the Bashash et al., (2017) paper:

“In multivariate models we found that an increase in maternal urine fluoride of 0.5mg/L (approximately the IQR) predicted 3.15 (95% CI: −5.42, −0.87) and 2.50 (95% CI −4.12, −0.59) lower offspring GCI and IQ scores, respectively.”

I certainly agree with this statement – but please note it refers only to the model they derived, not the data as a whole. Specifically, it applies to the best-fit lines shown in Fig 2 and Fig 3A as illustrated above. The figures in this quote relate to the coefficient, or slope, of the best fit line.

Recalculating from 0.5 mg/L to 1 mg/L this simply says the 95% of the coefficient values, or slopes, of the best fit lines resulting from different resampling should be in the range  -10.84 to -1.74 CGI (Fig 2) and -8.24 to 1.18 IQ (Fig 3A).

[Note – these are close to the CIs produced in my regression analyses described above – an exact correspondence was not expected because digital extraction of data from an image is never perfect and a simple univariate model was used]

The cited CI figures relate only to the coefficient – not the data as a whole. And, yes, the low p-value indicates the chance of the coefficient, or slope, of the best-fit line being zero is extremely remote. The best fit line is highly significant, statistically. But it is wrong to say the same thing about its representation of the data as a whole.

This best-fit line explains only 3% of the variance in IQ – and a simple glance at the figures shows the cited confidence intervals for that line simply do not apply to the data as a whole.

The misrepresentation

That brings us back to the problem of misrepresentation. We should draw any conclusions about the relevance of the data in the Bashash et al., (2017) study from the data as a whole – not just from the small fraction with an IQ variance explained by the fitted line.

Paul Connett claimed:

“The effect size is very large (decrease by 5-6 IQ points per 1 mg/L increase in urine F) and is highly statistically significant.”

But this would only be true if the model used (the best-fit line) truly represented all the data. A simple glance at Fig 2 in the slide above shows that any prediction from that data with such a large scatter is not going to be “highly statistically significant.” Instead of relying on the CIs for the coefficient or slope of the line, Connett should have paid attention to the standard error for estimates from the data as a whole given in the Regression statistics of the Summary output. – For Fig. 2, this is 12.9 IQ points. This would have produced an estimate of “5-6 ± 36 IQ points which is not statistically significantly different to zero IQ points,”  as I described in my article

Confusion over confidence intervals

Statistical analyses can be very confusing, even (or especially) to the partially initiated. We should be aware of the specific data referred to when we cite confidence intervals (CIs).

For example, Mary refers to the CI values for the coefficients, or slopes, of the best fit lines.

Figs 2 and 3A in the Bashash et al., (2017) paper include confidence intervals (shaded areas) for the best fit lines (these take into account the CIs of the constants as well as the CIs of the coefficients). That confidence interval describes the region of 95% probability for where the best-fit line will be.

Neither of those confidence intervals applies to the data as a whole as a simple glance at Figs 2 and 3A will show. In contrast, the “prediction interval” I referred to in my article, does. This is based on the standard error of the estimate listed in the Regression statistics. Dr. Gerard Verschuuren demonstrated this in this figure from his video presentation.

Mary is perfectly correct to claim “it is the average effect on the population that is of interest” – but that is only half the story as we are also interested in the likely accuracy of that prediction. The degree of scatter in the data is also relevant because it indicates how useful this average is to any prediction we make.

Given the model described by Bashash et al., (2017) explained only 3% of the IQ variance, while the standard error of the estimate was relatively large, it is misleading to suggest any “effect size” predicted by that model would be “highly significant” as this ignores the true variability in the reported data. When this is considered the effect size (and 95% CIs) is actually “5-6 ± 36 IQ points which is not statistically significantly different to zero IQ points,”

Remaining issues

I will leave these for now as they belong more to a critique of the paper itself (all published papers can be critiqued) rather than the misrepresentation of the paper by Mary Byrne and Paul Connett. Mary can always raise them again if she wishes.

So, to conclude, Mary Byrne is correct to say that the model derived by Bashash et al., (2017) predicts that an increase of “fluoride level in urine of 1 mg/L could result in a loss of 5-6 IQ points” – on average. But she is wrong to say this prediction is relevant to New Zealand, or anywhere else, because when we consider the data as a whole that loss is “5-6 ± 36 IQ points.”

I look forward to Mary’s response.

Similar articles

Paul Connett’s misrepresentation of maternal F exposure study debunked

Title slide for Paul Connett’s presentation to parliament

Anti-fluoride campaigners are misrepresenting a recent Mexican study claiming its findings should cause governments around the world to abandon community water fluoridation (CWF). Their claims are unwarranted because the study has a high degree of uncertainty. Activists are misrepresenting the accuracy of the studies findings. Because Mexico has areas of endemic fluorosis the study itself is not relevant to CWF.

Misrepresentation of the Mexican study was a central argument used by US anti-fluoride activist Paul Connett in his recent New Zealand speaking tour. This is shown in the Powerpoint presentation he prepared for his meeting at parliament buildings last month (see Anti-fluoride activist commits “Death by PowerPoint”).

It may have not been used in the end as only 3 MPs turned up. But, given his status in the anti-fluoride movement, this presentation will present the current strongest arguments against CWF. It is therefore worth critiquing his presentation whether it was given or not.

In this article, I will concentrate on Paul’s presentation of the Mexican study and may deal with other arguments used in the presentation in later articles. The paper reporting the study is:

Bashash, M., Thomas, D., Hu, H., Martinez-mier, E. A., Sanchez, B. N., Basu, N., … Hernández-avila, M. (2016). Prenatal Fluoride Exposure and Cognitive Outcomes in Children at 4 and 6 – 12 Years of Age in Mexico.Environmental Health Perspectives, 1, 1–12.

In Connett’s mind, the study’s results are so overwhelming we should immediately stop fluoridation throughout the world! This was the first and main argument he presented. His title slide and slide no. 10 introducing the study demonstrates the importance to him.

Slide No. 10 introducing Connett’s presentation of the Bashash et al (2017) study.

I have critiqued this study in previous articles – readers can find them at:

Fluoride, pregnancy and the IQ of offspring,
Maternal urinary fluoride/IQ study – an update,
Anti-fluoridation campaigners often use statistical significance to confirm bias and
Paul Connett “updates” NZ MPs about fluoride?

Paul is clearly aware of these articles because he included a note in his presentation about them. I am honoured (it is the only comment in the presentation) and pleased he has made an effort to engage with my critique.

This is what he says:

“Ken Perrott and those who follow him will claim that the wide degree of scatter in the data means the findings of this study are unreliable.  That is an incorrect interpretation of this graph and the study.  The effect size is very large (decrease by 5-6 IQ points per 1 mg/L increase in urine F) and is highly statistically significant.  The fact that urine F can only explain a small amount of the variation of IQ does not invalidate the finding.  Rather, it is a reflection that there are many other factors that affect IQ, most of which are essentially random with respect to F exposure.  For example, individual genetics plays a huge role in IQ (it explains 80% or more of variation in IQ), therefore it would not be possible for F to explain more than the small remaining portion of variation in IQ.  Most studies of other developmental neurotoxins like Pb and Hg find very similar low correlation coefficients, yet there is no debate that their findings are valid.”

This comment provides me with a basis for a more detailed discussion of his use of the study.

The small amount of variance explained

Connett acknowledges my point that the observed relationship with urinary fluoride can explain only a very small amount of the variation in IQ – only 3%. A bit hard to deny considering the high degree of scatter in the data which is obvious even in the slides Connett uses:

Slide 20 where Connett reproduces Fig 2 from the Bashash et al. paper.

But he claims that this:

“does not invalidate the finding. Rather, it is a reflection that there are many other factors that affect IQ, most of which are essentially random with respect to F exposure.”

Here he is, of course, referring to his own “finding” or conclusion – not the authors.

Notice his assumptions:

  • Other factors will be “essentially random with respect to F exposure,” and
  • The observed relationship will not be changed by the inclusion of these other factors.

Those are huge assumptions. And they are wrong.

Here is a relevant example illustrating the danger of such assumptions – the association between ADHD prevalence and extent of fluoridation observed by Malin & Till (2015). Their association was able to explain between 22% and 31% of the variance in ADHD, depending on the specific data used. Far more than the 3% for the Bashash et al., (2017) study.

Yet, when other risk-modifying factors were included, in this case, mainly altitude, the significant association with fluoridation disappeared. A model including altitude, but not fluoridation, explained 46% of the variability in ADHD (see Perrott 2017 and a number of articles in this blog).

In this case, the incidence of fluoridation was correlated with altitude – fluoridation was simply acting as a proxy for altitude in the Malin & Till (2015) association. So much for Connett’s assurance that other factors “are essentially random with respect to F exposure.”

Other studies have found an association between symptoms of fluorosis and cognitive deficiencies. Choi et al., (2015), for example, reported an association of child cognitive deficits with severe dental fluorosis, but not with water F concentration. But there is a relationship between fluoride exposure and fluorosis prevalence – ie. fluorosis is not random with respect to F exposure. If the health effects resulting from fluorosis are the prime cause of the cognitive deficiency, the inclusion of fluorosis incidence in the multiple regression could produce a model where there is a statistically significant association with fluorosis but not with fluoride expose. That is, the urinary fluoride values could be simply acting as a proxy of fluorosis incidence.

A similar non-random association of premature births and low birth weight could occur because these problems do occur in areas of endemic fluorosis. These could be two of the health issues related to fluorosis but fluoride intake may not be the prime cause (see Premature births a factor in cognitive deficits observed in areas of endemic fluorosis?).

Connett is completely wrong to assume that other risk-modifying factors not considered in the Bashash study would necessarily be random with respect to fluoride exposure. And he is wrong to assume that inclusion of these factors would not change the association of child IQ with mothers’ urinary fluoride reported in the paper.

Notably, the Bashash et al (2017)study did not include any measure of fluorosis as a risk-modifying factor – despite the fact that Mexico has areas of endemic fluorosis. I believe its consideration of gestation period <39 weeks or >39 weeks was inadequate (the normal average period is 40 weeks). The cutoff point for birth weight (3.5 kg) was also high.

The size of the IQ effect

We only have the data in the Bashash et al., (2017) study to go with here and the associations they report are valid for that data. But what about the calculations Connett makes from the reported association.

For example, Connett declares:

” The effect size is very large (decrease by 5-6 IQ points per 1 mg/L increase in urine F) and is highly statistically significant.”

Let’s test this claim – using the association represented in Fig 2 from Bashash, which is the figure Connett and other anti-fluoride activists are using (his slide 20 above).

Firstly, we need to calculate prediction intervals from the data (see Confidence and prediction intervals for forecasted values). The shaded region in the figure used by Connett (Fig 2 in Bashash et al., 2017) represents the confidence interval – the region where there is a 95% probability that a best-fit line for the data lies. The region for the prediction intervals is much larger and Connett may be confused because he has interpreted the confidence interval wrongly. Yet, the prediction intervals are the important measure when considering the effect size.

Here are my graphs for the confidence interval and the prediction interval using data I digitally extracted from the paper (see Maternal urinary fluoride/IQ study – an update).

Let’s consider the predicted values of “child IQ” for urinary F concentrations of 0.5 and 1.5 mg/L.

Urine F (mg/L) Predicted value Lower Higher
0.5 99.8 74.4 125.2
1.5 93.0 67.5 118.4

The prediction intervals are very large. This means the real value for “child IQ” at a urine F value of 0.5 mg/L has a 95% probability of being in the range 74.4 – 125.2. The corresponding range for a urine F concentration of 1.5 mg/L is 67.5 – 118.4. When Connett claims that an increase of 1 mg/mL in mother’s urinary F produces a drop of 5 – 6 IQ points he actually means a drop of 5 – 6 ± 26 IQ points which is not statistically significantly different to zero.

The best-fit line for the data may be statistically significant – but Connett is wrong to say this about his predicted effect of urinary F on child IQ. In fact, over the whole range of urinary F measured there is a 95% probability that IQ remains at 100.

Connett’s claim of a “highly statistically significant” effect size is completely false. If he had simply and objectively looked at the scatter in the data points he would not have made that mistake.

Comparing maternal urinary F levels to other countries

Connett makes an issue of the similarity of maternal urinary F levels found in this Mexcian study to levels found elsewhere. One is tempted to say – so what? After all, I showed above that his claim of a “highly statistically significant” drop in child IQ with increases in maternal urinary F is completely wrong.

He does compare the urine F levels reported by Bashash et al., (2017) with some New Zealand data (Brough et al., 2015) and finds them to be very similar. Interestingly, Brough et al., (2015) reported their urinary F values as indicating fluoride intakes were inadequate for the women concerned. They certainly did not indicate toxicity.

The comparison does highlight for me one of the inadequacies in the Bashash (2017) paper – the inadequate measurements of urinary F. Whereas Borough et al., (2015) used the recommended 24-hr urine collection technique, the data used by Bashash et al (2017) relied on spot rather than 24 hr measurements. These spot measurements were only made once or twice during the pregnancy of these women.

Yes, these were the only F exposure measurements Bashash et al., (2017) had to work with but they are far from adequate.

Conclusions

Paul Connett, as a leader of the anti-fluoridation movement, is completely wrong about the Bashash et al., (2017) study. It will not lead to the end of community water fluoridation throughout the world – nor should it.

He has attempted to ignore, or downplay, the high scatter in the data and the low explanatory power of the relationship between children’s IQ and maternal F exposure found in the study (only 3%). His denial that this relationship may disappear when other more important risk-modifying factors are included is also wrong – as other examples clearly show.

Connett’s presentation of a size effect (5-6 IQ points with a 1 mg/L increase in F exposure) as “highly statistically significant” is also completely wrong. In fact, this size effect is more like 5 – 6 ± 26 IQ points which is not significantly different to zero.

The misrepresentation of this study by Paul Connett and other anti-fluoridation activists demonstrates, once again, that their claims should never be accepted uncritically. This is just one more example of the way their ideological and commercial interests drive them to misrepresent scientific finding.

Similar articles