Saturday, April 27, 2013

A bizarre study on the safety of water fluoridation


Triggered by an exchange with Mel Rader, I studied a New Zealand research paper that was supposed to show that water fluoridation is safe.  I found the paper rather bizarre because not only does the paper use an incomplete analysis which allows safety problems to go undetected; it also cites a second paper by the same research group which explains precisely why the analysis in the first paper is misleading.

So we are in a situation where we are accumulating evidence associating fluoride in drinking water with decreased IQ scores in children, and we are lacking studies that establish a safe level of fluoride exposure with respect to the developing brain.

Motivated by the precautionary approach that scientists advocate for protecting our children's vulnerable brains, the only conclusion I can draw is that now is the time to stop further expansion of water fluoridation and demand the needed safety studies.


In March this year, I enjoyed having an exchange with Mel Rader, who is coordinating the campaign for water fluoridation in Portland, and Mel got me interested in a study from 1986 with the title "Exposure to fluoridated public water supplies and child health and behaviour" by Shannon, Fergusson and Horwood.  This study has been cited over the years in support for the safety of water fluoridation, and in our exchange, Mel pointed out:
One good study in epidemiology is worth more than a thousand bad studies, and this study is way better quality than all the Chinese studies that you are bringing up.
The Chinese studies are part of the accumulating evidence that I am concerned about because they show that fluoride in the drinking water is associated with disturbed brain development.  We don't know if the fluoride is the actual cause, and if it is, we don't know if there is a safe limit for fluoride exposure from drinking water for children.

So naturally, I decided to examine what I will call the "Shannon-Fergusson-Horwood safety study" more closely, to see if it held water and could help me to put my concerns to rest.

The study followed children from birth up to the age of seven years in different areas in the urban region of Christchurch, New Zealand.  Some of the areas had water fluoridation and others did not, and the study grouped the children according to how many years they had lived in one of the areas with water fluoridation.

The good news is that the Fergusson safety study does account for a number of possibly confounding variables that are important in IQ studies—such as socioeconomic status—and no significant difference was found in behavior measures.  The study concluded:
The results of this seven year study suggest that there is no evidence to indicate that that exposure to a fluoridated public water supply had any detectable effect on a large range of measures of childhood morbidity or problem behaviour, even when allowances were made for family social background.
Before we assess the safety study further, let us take a break and discuss Reference 6 of the study.

The Fergusson-Horwood dental health study

Reference 6 of the Shannon-Fergusson-Horwood safety study is another study called "Relationships between exposure to additional fluoride, social background and dental health in 7-year-old children", based on the same underlying study of the children in Christchurch, and is written by Fergusson and Horwood.  It illuminates the relationship between fluoride exposure and dental health in the Christchurch children.

Already in the abstract, the authors note the importance of taking the total fluoride exposure into account when analyzing the effects of fluoride:
This analysis showed that the level of exposure to additional fluoride was a complex variable influenced by at least three factors: a) the use of fluoride toothpaste; b) the child's length of residence in a fluoridated area; c) the length of time for which the child had been provided fluoride tablets.
The analysis showed that the important variables with associations to dental health scores are the number of years the children were exposed to fluoridated water,  the number of years the children were given fluoride tablets, and social background.  The longer the children were exposed to fluoride (through water or tablets), the better the dental health scores were.  Additionally, better social background was associated with better teeth.

In addition, the paper found relationships between the use of fluoride tablets and exposure to fluoridated water.  Unsurprisingly, longer time in a water-fluoridated area was associated with shorter time of using fluoride tablets.  I presume that parents in non-water fluoridated areas were recommended to compensate the lack of fluoride in the water by giving their kids fluoride tablets.  The paper expresses this as:
This correlation reflects the fact that as duration of residence in a fluoridated region increased, use of fluoride tablets decreased.
The negative correlation between fluoride exposure from the water and from tablets can be discerned in the paper's Table 1:
In particular, no parents in the study whose children lived in a water-fluoridated area for the whole period gave their children any fluoride tablets.

Another relationship that the paper found was between social background and the usage of fluoride tablets: higher social background was correlated to more years of tablet use.  Perhaps parents with higher social background were more inclined to follow public health recommendations and get fluoride tablets.

Table 1 also gives an indication of the importance of trying to take the total fluoride exposure into account when assessing health effects.  "dmf" score (number of decayed, missing or filled teeth) association was much more pronounced when taking both water and tablet exposure into account, compared to when one was looking only at water fluoridation, for example.  This can be seen by looking at the "Total" row in which tablet exposure is ignored: the dmf scores do not drop as clearly with increased exposure to fluoridated water.

So, rightly, the authors note in the discussion section of the paper (my emphasis):
This 7-yr study serves to illustrate some of the complexities which arise in assessing the exposure of children to additional fluoride.  In particular, for this cohort, additional fluoride was provided by fluoride toothpaste, fluoride tablets and fluoridated water.  While fluoride toothpaste was almost universally used, children had varying exposure times to both fluoridated water and tablets.  A major methodological implication of this result is that exposure to fluoride is a complex variable which requires measurement over a number of sources and time.  Failure to recognise this fact may lead to specious conclusions being draw in cross-sectional studies which examine only a single source of fluoride at a given time.  Typically such analyses are likely to have reduced sensitivity as a result of inaccuracies in the measurement of the sources of fluoride and varying exposure times to these sources.
In short: if you look at only a single source of fluoride, your analysis may fail to show an effect, which may lead to misleading conclusions. It was particularly important to take fluoride exposure through both water and tablets into account, because of the inverse association between the two.  For example, if one only looked at water fluoridation exposure, it could be the case that the total exposure to fluoride was roughly the same regardless of the number of years that the children lived in a water-fluoridated area, because parents compensated with fluoride tablets.

Now, let's turn back to the research group's safety study.

The Shannon-Fergusson-Horwood safety study

The introduction of the safety study gives a hint of a possible reason for its publication:
There is continuing debate about the advantages and liabilities of the fluoridation of public water supplies.  The advantages in terms of reduction of dental caries have been well documented but, on the other hand, the local anti-fluoridation lobby has always raised fears that fluoride may have unforeseen health consequences.
Perhaps the paper was meant to assure the public that there were no reasons to be concerned about adverse effects of water fluoridation.  And we are reassured already in the abstract, which noted:
This study showed no association between exposure to fluoridated water and a large range of measures of child health and behaviour taken during the period from birth to seven years, even when the possible effects of family social background were taken into account statistically.
The details were illustrated in the paper's Table 1, which showed that various health effects were not associated in a statistically significant manner (NS = not significant) with the number of years that the children had been exposed to fluoridated water:
The table shows that none of the adverse health effects that were studied had any association with the exposure time to fluoridated drinking water.

But wait a second—what on earth happened with exposure to additional fluoride?  Where are the fluoride tablets taken into account?

The bizarre truth is that the total exposure is not mentioned at all in the Shannon-Fergusson-Horwood safety study.  Not a single word about tablets.  I found it particularly confusing that only exposure to fluoride through the drinking water was included since two of the authors the same year published a paper where they pointed out that "[f]ailure to recognise this fact may lead to specious conclusions being draw in cross-sectional studies which examine only a single source of fluoride at a given time."

I contacted co-author John Horwood and he was kind enough to explain that the intent was not to estimate total fluoride exposure, but rather the consequences of exposure from the source where there was no individual choice, namely the drinking water.

I found this explanation even more confusing, since I wouldn't expect the human body to be able to distinguish between mandatory and non-mandatory exposure to fluoride when it comes to potential health effects.

So why did the safety study not find any difference in adverse health effects when looking at the number of years the children were exposed to fluoride from the water?  Was it because there are no such effects, or was it because there actually were effects, but the compensation with fluoride tablets made the total exposure to fluoride so similar that those effects could not be detected?

It would be interesting to revisit the original data and see what would come out of it if one attempted to account for total fluoride exposure.

Perhaps the analysis would still show that there were no negative health effects.  Such a result would at least rest on a somewhat stronger foundation than what was presented in the original paper.

But perhaps the analysis would show some less convincing and less desirable results that could have fueled the "anti-fluoridation lobby" in New Zealand back in the 80s.  And perhaps the safety paper would not have been used so much as an argument (for example by the Australian Dental Association in their FAQ) that water fluoridation is safe and "worth more than a thousand bad studies".

In search of a good safety study, showing the safe limit

With the accumulating evidence showing an association between fluoride in the drinking water and decreased IQ test scores, we need to see a good safety study that can show us that it must have been something else behind what looks like disturbed brain development and not fluoride in the drinking water.

Such a study should take the actual fluoride exposure into account, and it should look at the exposure during the first critical months of brain development before the blood-brain barrier is fully developed at six months' age.  Before the blood-brain barrier is in place, fluoride may enter the brain where it potentially could disturb the rapid brain development that is taking place.  The pregnant mother's exposure to fluoride must be estimated, and one must note to what extent the infants were breast fed or formula fed (breast feeding would have eliminated almost all fluoride exposure compared to formula, since the milk-production mechanism actively reduces fluoride transfer from plasma to milk).

One should also determine what safety factor would be acceptable.  Due to individual variation in exposure to fluoride, it is not enough to learn that "optimal" exposure through food and fluoridated water does not cause detectable brain development disturbance.  Since the brain can only be developed once and any disturbance is irreversible, we need a margin of safety and conduct a study that shows that a significantly higher fluoride exposure than "optimal" is safe.  Such higher fluoride exposure may not be present in the U.S., so we may have to conduct the study abroad.  We need to learn: where is the safe limit for fluoride exposure to our children's developing brains?

Until we have conducted such a study, the prudent thing to do is to heed the precautionary approach that Grandjean and Landrigan motivated in their 2006 Lancet review (my emphasis):
Prevention of neurodevelopmental disorders of chemical origin will need new approaches to control chemical exposures. The vulnerability of the human nervous system and its special susceptibility during early development suggest that protection of the developing brain should be a paramount goal of public health protection. The high level of proof needed for chemical control legislation has resulted in a slow pace of interventions to prevent exposures to lead and other recognised hazards. Instead, exposure limits for chemicals should be set at values that recognise the unique sensitivity of pregnant women and young children, and they should aim at protecting brain development. This precautionary approach, which is now beginning to be used in the EU, would mean that early indications of a potential for a serious toxic effect, such as developmental neurotoxicity, should lead to strict regulation, which could later be relaxed, should subsequent documentation show less harm than anticipated.  As physicians, we should use prudence when counselling our patients, especially pregnant mothers, about avoidance of exposures to chemicals of unknown and untested neurotoxic potential.
Our current situation is that we are accumulating indications that fluoride may disturb brain development, and we have no studies showing where the safe limit is under which there is no harm.  Therefore, according to the precautionary approach, now is the time to stop further expansion of water fluoridation and demand the studies that we need.


  1. This is an excellent analysis, Magnus -- the kind of critical analysis that is so woefully lacking in the pro-F camp. Any study on F/IQ in western populations that fails to carefully control for individual exposures, will be hard-pressed to find an effect as a function of water fluoride level. The ideal study, in my view, would not only focus on individual metrics of fluoride exposure (e.g., urinary F level), but would account for various nutrient statuses that can exacerbate fluoride's neurological effects (e.g., suboptimal iodine intake). And, unlike most studies on fluoridation's safety to date, the study should focus on socially disadvantaged populations that will generally have the health conditions that render one most vulnerable to fluoride toxicity. One study, by the way, that you may want to check out (if you haven't already done so), is the ongoing, NIH-funded, multimillion dollar Iowa Fluoride Study. I've summarized the study's key findings here: While the findings are damning in and of themselves, it's important to keep in mind that the study population is 97% white, and of relatively high socioeconomic status -- and thus a population that fails to reflect the full range of individual susceptibility.

  2. Thanks for your comment and link, Michael. It was interesting to hear about the Iowa Fluoride Study - hopefully we will gain more insight into fluoride's different effects from this longitudinal study as it progresses.

  3. Magnus, Your careful research and analysis have uncovered one serious weakness with the New Zealand paper. There are several other rather obvious weaknesses with this paper in regard to answering the question of whether fluoride has "... any detectable effect on a large range of measures of childhood morbidity or problem behaviour".

    Key qualifier words the New Zealand authors use are "detectable effect". If you use an insensitive test, you won't be able to detect effects. Most of the many likely or possible adverse effects of fluoride are relatively subtle, may be hard to distinguish from other common illnesses, or may only occur rarely in particularly sensitive or highly exposed individuals.

    How many suspected adverse effects of fluoride did the New Zealand study even try to detect?

    How sensitive was it?

    The study made no attempt to measure IQ, so it is hardly relevant to the mounting scientific evidence that fluoride lowers IQ. "Problem behavior" is sometimes related to IQ, but often not. The measures of problem behavior used in this study don't seem very sensitive. They are simply parent or teacher reports that the child did or did not exhibit "problem behavior" at certain ages. That seems pretty vague to me.

    What are some of the other amongst the "large range" of effects studied? Respiratory infections and asthma, neither of which have ever been identified as linked to ingested fluoride. Hmmm, are they padding their list of health effects with those unlikely to be associated with fluoride intentionally, or out of ignorance?

    The next category of effects can be combined as allergic or hypersensitivity effects: GI illness, eczema and other allergy. These have been associated with fluoride. But of course, now we have to remember what Magnus has uncovered. This study did not look at total fluoride exposure, but only water fluoridation exposure, so it's own authors would consider it incapable of detecting an effect.

    There are two other health effect categories they studied: number of consultations with a doctor, and number of hospital admissions. These are about as unspecific a health effect as can be imagined. They lump every possible health reason for visiting a doctor or hospital into a single measure, thereby overwhelming any detectable effect from specific health problems possibly caused by fluoride.

    So, other than the examination of allergic and hypersensitivity symptoms, this study seems to have been designed to be so insensitive that it is virtually guaranteed that no effect of fluoride will be found.

    Mel Rader, I don't know who fed you this low quality science on fluoride, but when you spout it back when defending fluoridation it reveals you are being duped by your advisors ... or your advisors don't know how to interpret scientific evidence.

    Magnus, keep up the good work.