Fluoride in drinking water and osteosarcoma incidence rates in the continental United States among children and adolescents.  Levy Ledrec

Cancer Epidemiol. 2012 Apr;36(2):e83-8. doi: 10.1016/j.canep.2011.11.008. Epub 2011 Dec 19.

This paper itself acknowledges its ecological (group level) design means it can not control for confounders on an individual level. The group level for this study is states in the US, which means they are very broad groups, rather than small groups which would be closer to an individual level and more reliable.

The only potential confounders they controlled for were gender and age. But for age, they only used 5 year age brackets in their main analysis of CDC WONDER data. Their analysis of SEER data did use 1 year age brackets, but they did not look at fluoridation in their SEER analysis! They do not give any good explanation for why they failed to examine the question of whether fluoridation was a risk factor for osteosarcoma in the SEER data. The SEER data is available at a much finer grained level than CDC WONDER, so it would have been reasonable to do analyses with the SEER data on the effect of fluoridation. The SEER data is available by single years of age and by country, rather than just by state. In fact, the first ecological study to find a relationship between fluoridation and osteosarcoma (by Hoover 1991) used exactly this SEER data, at a county level, and found a positive association. It almost seems as if Levy has intentionally omitted looking at the SEER data in a way that might confirm Hoover’s findings.

Even more concerning, is that Levy did not use all of the SEER data. Instead, without any good explanation, he chose to omit the data from Hawaii and Utah. His reasoning is that these two states had very low fluoridation rates. That should have meant they would be especially important to include in any analysis of the relationship between fluoridation and osteosarcoma. If the only states that are included cover only a narrow range off fluoridation rates, one will not expect to find strong evidence that fluoridation increases risk for osteosarcoma, even if it is a strong biological effect. Therefore, including the broadest range of fluoridation rates is essential. Furthermore, to omit 2 of 9 states means you omitting a fairly large portion of your data … potentially enough to bias your results.

As it turns out, both Hawaii and Utah have relatively low osteosarcoma rates. If they had been included in the analysis, an association between fluoridation may well have turned up.

Finally, the authors reveal they have a fundamental misunderstanding of Bassin’s study and results. Here is what they say about Bassin’s study: “Similarly, our analysis from SEER 9 data shows similar incidence rates between males and females between 5 and 14 years of age. We observed no evidence of ‘‘peaking’’ in male incidence rates or risk ratio between ages 5 and 8 as was reported in Bassin’s 2006 study [17].”

Bassin did not find an incidence rate peak or risk ratio (high F to low F) at ages 5-8. Instead, she found that exposure during this age range to fluoride caused an increased odds ratio (or risk ratio) for osteosarcoma diagnosed almost 10 years later around age 15. Levy is completely misusing the SEER data to claim it does not support the age-incidence pattern of osteosarcoma found by Bassin. In fact, Bassin’s age-incidence pattern closely follows SEER data, and it was Bassin’s key finding that exposure at ages 5-8 lead to increased risk of osteosarcoma being diagnosed later in life.

Basically, Levy and Leclerc used a very weak methodology, weakened it further by failing to even look at the fluoridation in relationship to their SEER data set, and chose to omit the two SEER states that might have helped show there was a positive relationship between fluoridation and osteosarcoma. They completely misunderstood the main finding of Bassin, and perhaps that was why they erroneously chose to omit Hawaii and Utah. They thought Bassin found an unusual age-incidence curve in those exposed to fluoridation (nothing could be further from the truth), so they thought they needed to only look at states where there was fairly widespread fluoridation and omit those that had little fluoridation. It has very little scientific weight compared to Bassin’s study.

Chris Neurath

Research Director AEHSP – American Environmental Health Studies Project