By visiting our site, you agree to our privacy policy regarding cookies, tracking statistics, etc.

WA Dental Health Report

This poor-quality dental report provides little evidence of fluoride’s effectiveness

In 2016, a dental health study or report (commissioned by WA Health) was published, comparing tooth decay rates between children residing in fluoridated Perth and children residing in unfluoridated Bunbury and the South-West. It is obvious this study was designed to achieve a desired outcome, but a detailed analysis reveals it is very poor-quality, and therefore should not be relied on as evidence to support the fluoridation of Bunbury’s drinking water.
 
Facebook
Twitter
LinkedIn
WhatsApp
Email

BUNBURY FLUORIDATION MENU

Definitions

Covariate: A covariate is a patient or study participant variable (e.g. age, sex, race), that may or may not be related to outcome being studied. If the covariate is related to the outcome and the exposure/risk factor, it becomes a confounder.
Confounder: A confounder is a covariate that is related to both the outcome and the exposure/risk factor. A confounder may either increase of decrease the likelihood of the study outcome. In the case of dental caries for children, the major confounders are sugar consumption (including and especially sugar-sweetened beverages), dental hygiene (including visits to the dentist), overall dietary profile and socioeconomic status (SES), however there are many other confounders.
Dental caries: Tooth decay.
Dmft: decayed, missing and filled primary teeth.
DMFT: Decayed, missing and filled permanent teeth.
SIC or SiC: Significant Caries Index. The SiC Index is calculated by taking the mean DMFT of the one-third of the individuals having the highest of DMFT values in a given population.
SIC10 or SiC10: Significant Caries Index (10%). The SiC10 Index is calculated by taking the mean DMFT of the 10 per cent of the individuals having the highest of DMFT values in a given population.
Socioeconomic status (SES): Socioeconomic status is the social standing or class of an individual or group. It is often measured as a combination of education, income, occupation and area or region of residence.

An analysis of the study commissioned and published by the WA Department of Health titled: “Dental Health Outcomes of Children Residing in Fluoridated and Non-Fluoridated Areas of Western Australia”, Crouchley, K. & Trevithick, R. (2016).
Overview

Before delving into this study’s design detail and outcomes, it is worth examining the first paragraph, which sets the tone for this biased study, stating:

“Water fluoridation has been introduced in a number of countries across Europe, South America, North America and Asia (Cheng et al, 2007) yet remains a contentious issue for some individuals.”

The authors of this study are obviously not aware that less than five percent of the world’s population receives artificially fluoridated water and most European governments (representing 98 percent of Western Europe’s population) have ceased, rejected or prohibited water fluoridation based on a lack of safety, lack of effectiveness in reducing tooth decay, and lack of ethics. This clearly shows artificial fluoridation is not just a contentious issue for “some individuals”, but rather for most of the world’s population and their decision makers.

Also reflecting this lack of popularity, there is an obvious trend worldwide for de-fluoridation. Since 2010, more than 230 communities have rejected the practice of adding “fluoride” to drinking water, as numerous substantial research studies continue to confirm the adverse health effects of low-level, long-term ingestion of artificial fluoridation chemicals – the most common of which is hexafluorosilicic acid, a highly-toxic industrial waste chemical created as a by-product in the processing of phosphate fertiliser.

Study design

This is a fundamentally very weak study design.

The study included a total of 10,825 children aged 5 to 12 years, 9,972 of these were from fluoridated metropolitan Perth and 853 were from non-fluoridated regions in the South-West of WA, from Australind, north of Bunbury, down to Margaret River and Bridgetown in the south. In other words, about 92 percent of the cohort were from a fluoridated region. This obvious bias means this study is already statistically weak.

Outcomes and covariates were known at individual level and yet the study presented exposure of dental caries at the group level, therefore it is a semi-ecological study.

This study is cross-sectional, which means it compared children over a single time period rather than followed children longitudinally, over an extended period of time.  Cross-sectional studies are weaker than longitudinal studies. 

Starting with this fundamentally weak design, the study quality is crippled even more because it only compared two areas, one area with fluoridation and one area without.  These weak cross-sectional ecological studies can be greatly strengthened by examining larger numbers of areas.  A comparison of just two areas is by far the weakest design (even though it is extremely common amongst fluoridation effectiveness studies) because it is impossible to adjust for any confounding group-level variable that differs between the two areas. 

For example, if there was no individual-level SES information, but the group-level SES was lower in the non-fluoridated area or areas, it is impossible to control for this difference because there is no way to know how large an effect on caries SES has when fluoridation status is held constant.

On top of these built-in weaknesses there is added the severe limitation that no socio-economic, diet, oral hygiene, or several other important risk factors for caries were measured. The incidence of dental caries involves many risk and protective factors, so failure to control for any of them means the study has a high risk of bias. The report acknowledges this weakness, but only lists a few of the risk factors that were not measured:

“The study controlled for age, sex, Aboriginal status and having an initial examination at a DTC but was unable to control for the effects of other individual level factors which may contribute to the risk of caries, including socio-economic status, diet, and dental and oral hygiene. “ [p. 23]

Socio-economic status (SES), however, is a very well-known potential confounder, and it should have been possible to at least obtain some group-level information on SES, such as average per capita income for a local area. The fact that this study did not try to obtain any SES data suggests the authors were not very serious about conducting a rigorous valid study of caries risk.

It is quite likely that at least some of the non-fluoridated areas had lower SES compared to Perth, were more rural and they definitely had different types of water source containing varying amounts of natural fluoride.

In a 2017 Communications Plan, obtained via FOI, Aqwest stated the following:

“Aqwest has 6 points of input of water to its system with 6 associated Water Treatment Plants. The raw water at these plants has natural fluoride levels ranging from 0.00 mg/litre to 0.80 mg/litre with an overall average of 0.26 mg/litre. The desired level stipulated by the Department of Health is 0.80 mg/litre.”

With such a range of naturally occurring fluoride in the unfluoridated children, it was therefore surprising no attempt was made to measure existing levels of fluoride or control for this potentially significant confounder.

Most importantly, the differences between the fluoridated and non-fluoridated area’s outcome were so small that even a slight amount of confounding could account for the differences, rather than the difference in fluoridation status.

By far the most important analysis in this type of study is the multivariable regression model. This is where potential confounding is dealt with. This study has a glaring omission with respect to multivariable regression models:  It only reports the results of such an analysis on a single outcome. It only looked at the outcome “percent of children caries free”. This is the crudest outcome that the study investigated and is not very informative about the full distribution and differences in disease.  All the other outcomes are more informative.  They are:

  • dmft, and the individual components d, m, f
  • DMFT, and the individual components D, M, F
  • SIC
  • SIC10

No multivariable regression analyses were conducted for any of these more informative outcomes and no explanation was offered for this omission.

Also, there is another glaring omission in this study: levels of dental fluorosis – from both natural fluoride contained in water in the south-west and from artificial fluoridation in Perth – should have been recorded at the same time the children were assessed. But of course the authors and WA Health did not want to draw any possible attention to the damage water fluoridation, especially in the fluoridated Perth region, may be causing.

Study findings

This study found:

For deciduous teeth:

  • NO significant difference in filled teeth at ages 5, 6, 8, years
  • NO significant difference dmft except age 5
  • NO significant difference in decayed teeth at ages 6, 7, 8 or 9
  • NO significant difference in missing teeth at ages 6, or 8

For permanent teeth:

  • NO significant difference in filled teeth at ages 6, 7, 8 or 9
  • NO significant difference in decayed teeth or missing teeth or filled teeth at age 11-12
  • Aboriginal children had much worse odds of one or more dmft 4.44 (3.56-5.28) and 2.41 (1.77-3.29) DMFT

The differences between the fluoridated and non-fluoridated were very small.  For example, the mean SIC10 scores (the DMFT for the 10% of children with the highest DMFT) were 2.5 and 2.8 in the fluoridated and non-fluoridated groups aged 9 respectively. That difference of 0.3 DMFT is extremely small, especially considering this is for the children with the most decay, which are arguably those in most need of help from any preventative measures.

The study does not do a good job of indicating the variance and statistical significance, but it appears most of the differences do not reach statistical significance. The report does not say whether the difference in the SIC10 for 9 year olds reached statistical significance, which almost certainly means it did not, since in every other case where a result was significant, the report mentions this.

The very small apparent benefits also mean this study is especially prone to bias from even small amounts of confounding.  For example, if SES had been measured, and it was found that SES has even a relatively small effect on DMFT, and SES was lower in the non-fluoridated area, then even the small degree of confounding by SES could be enough to completely explain the lower DMFT in the non-fluoridated area.

Finally, it is completely speculative for the report to suggest that since other studies that controlled for potential confounders like SES still found a benefit from fluoridation, that it wasn’t as much of a weakness for this study to have not controlled for SES and other important risk factors. Our epidemiologists can’t remember ever seeing such a speculative claim in an epidemiological report, and it suggests these authors were desperately trying to find excuses for their weak study design.

Some interesting excerpts:
The authors of this study are obviously not aware that less than five percent of the world’s population receives artificially fluoridated water and most European governments (representing 98 percent of Western Europe's population) have ceased, rejected or prohibited water fluoridation based on a lack of safety, lack of effectiveness in reducing tooth decay, and lack of ethics.
In other words, about 92 percent of the cohort were from a fluoridated region. This obvious bias means this study is already statistically weak.
On top of these built-in weaknesses there is added the severe limitation that no socio-economic, diet, oral hygiene, or several other important risk factors for caries were measured. The incidence of dental caries involves many risk and protective factors, so failure to control for any of them means the study has a high risk of bias. 
The fact that this study did not try to obtain any SES data suggests the authors were not very serious about conducting a rigorous valid study of caries risk.
With such a range of naturally occurring fluoride in the unfluoridated children, it was therefore surprising no attempt was made to measure existing levels of fluoride or control for this potentially significant confounder.
Most importantly, the differences between the fluoridated and non-fluoridated area’s outcome were so small that even a slight amount of confounding could account for the differences, rather than the difference in fluoridation status.
The report does not say whether the difference in the SIC10 for 9 year olds reached statistical significance, which almost certainly means it did not, since in every other case where a result was significant, the report mentions this.
Also, there is another glaring omission in this study: levels of dental fluorosis – from both natural fluoride contained in water in the south-west and from artificial fluoridation in Perth – should have been recorded at the same time the children were assessed. But of course the authors and WA Health did not want to draw any possible attention to the damage water fluoridation in the fluoridated Perth region may be causing.
Finally, it is completely speculative for the report to suggest that since other studies that controlled for potential confounders like SES still found a benefit from fluoridation, that it wasn’t as much of a weakness for this study to have not controlled for SES and other important risk factors. Our epidemiologists can’t remember ever seeing such a speculative claim in an epidemiological report, and it suggests these authors were desperately trying to find excuses for their weak study design.