The following is an excerpt from FAN’s comprehensive analysis of the pre-2001 research on fluoride and osteosarcoma.
The case-control study by Gelberg, published first as a PhD dissertation (Gelberg 1994) and then later in two peer-reviewed journals (Gelberg 1995, 1997), may represent the most substantive study on fluoride/osteosarcoma previous to Bassin’s 2001 analysis.
In assessing Gelberg’s data, we were at first struck by the existence of several notable errors in both the thesis and papers. In all three documents, Gelberg states that the majority (68%) of her cases were female. In her thesis, Gelberg states this repeatedly and cites 18 studies finding the opposite, that males have higher incidence than females (Gelberg 1994; p. 82, 85, 102). However, if one looks at the data presented in Gelberg’s tables, it is clear that 68% of the cases were in fact males, not females. Personal communication with Gelberg confirmed this error. We are not sure how Gelberg got her data confused, and how this error managed to slip by peer review twice.
Another error in Gelberg’s paper can be gleaned by comparing Tables 2 and 3 in her published paper with Table 1 (Gelberg 1995). Upon comparing these tables, it is evident that the cases and controls in the “Total Fluoride” and “Toothpaste” categories in Tables 2 and 3 were reversed. Again, we are surprised to see that these errors were not corrected in peer review. These errors do not affect her final results as they were apparently made separately from her statistical analysis for fluoride.
While these errors do raise questions about the study, our primary concern with Gelberg’s work relates to the methods she used to analyze her data.
Gelberg’s study population was New York State except New York City and all cases came from the NY Cancer Registry from the years 1978-1988. Unlike most other studies, Gelberg tried to include exposure to fluoride from sources other than from drinking water. From interviews she estimated how often subjects used fluoride supplements, toothpaste, or received fluoride dental treatments. However, she had to use very broad assumptions to assign how much fluoride was ingested from each of these sources. These broad assumptions combined with possible subject recall biases, bring into question the accuracy of these fluoride exposures. For example, between her thesis in 1994 and her published paper in 1995, she apparently revised her methods of assessing toothpaste fluoride exposure so that her assessments for lifetime cumulative exposure went down by a factor of three. No explanation is offered for this change in exposure assignment but it highlights the difficulties in determining fluoride exposures from toothpaste and other non-water sources. We therefore believe that Gelberg’s analysis of drinking water fluoride is likely to be the most reliable and will focus on that.
Gelberg, like Hoover 1991, never analyzes her data with subjects divided into a simple two-category model: exposed versus unexposed. Instead, she uses milligrams of total lifetime fluoride exposure as a continuous variable. But then she presents the data broken into “quartiles” of exposure level without ever presenting the outcome of the continuous variable model as an Odds Ratio (OR). So, the reader is left to estimate what such an OR might be from her “quartile” ORs. She then calculates the p-values for the trend of her analysis with a method that assumes linearity of response to the variable. Yet from looking at the ORs for “quartiles” it is apparent that many of the OR trends do not show a linear trend, thereby violating the assumption of her method.
It is still possible, however, to get some feel for the relative risks from drinking water exposure to fluoride. For males the lower “quartile” group shows a borderline statistically significant increased risk OR of 2.8 (95%CI 1.0-8.1). For females the OR is even higher and statistically significant at 10.5 (95%CI 1.2-91). For both males and females in the higher “quartiles” of exposure, the ORs are no longer significant, but the risk for osteosarcoma generally stays above 1.0. If, instead of breaking the data into “quartiles”, it had been broken into just “exposed” and “unexposed”, it is quite possible the exposed group would have a significantly elevated risk for osteosarcoma compared to the unexposed group.
The best way to resolve this question would be for Gelberg to share her data so that researchers could see whether other methods of analysis might reveal associations that her limited approach failed to uncover. Another benefit of sharing this data is because it is some of the only data for which detailed histories of fluoride exposure are available. This means her data could be reanalyzed looking at age-specific exposure effects as in Bassin’s study. Re-using Gelberg’s data as a check on Bassin’s findings would be much more efficient than the expensive and time-consuming process of collecting new data.
In looking for other possible risk factors for osteosarcoma, Gelberg (1994) found that a history of exposure to dental x-rays was significantly related to the development of osteosarcoma (OR 4.0; 95%CI 1.3-12) . Dental x-rays were, in fact, one of the few variables Gelberg examined that had an effect reaching statistical significance. While Gelberg does not provide data on the number of cases who had osteosarcoma of the jaw or skull (sites where osteosarcoma is known to develop in humans), it would be interesting to see if there was any correlation between the use of dental x-rays and osteosarcoma of the jaw/skull. If a correlation does in fact exist, it would be important to determine if this correlation impacted the fluoride analysis. In other words, if the fluoride analysis was limited to only those children with osteosarcoma at other bone sites, would the OR values shift? If so, in what direction?
Whether or not controlling for dental x-rays impacts the fluoride analysis, we believe the finding is of public health significance in and of itself, and are surprised to see that this is the only positive finding from Gelberg’s thesis that was not mentioned in her two published papers (Gelberg 1995, 1997).
Bassin E.B. (2001). Association Between Fluoride in Drinking Water During Growth and Development and the Incidence of Osteosarcoma for Children and Adolescents. Doctoral Thesis, Harvard School of Dental Medicine.
Gelberg K.H. (1994). Case-control study of osteosarcoma. Doctoral Thesis, Yale University.
Gelberg K.H., Fitzgerald E.F., Hwang S., Dubrow R. (1995). Fluoride exposure and childhood osteosarcoma: a case-control study. American Journal of Public Health 85:1678-83.
Gelberg KH, Fitzgerald EF, Hwang S, Dubrow R. (1997). Growth and development and other risk factors for osteosarcoma in children and young adults. International Journal of Epidemiology 26:272-8.
Hoover R.N., Devesa S.S., Cantor K.P., Lubin J.H., Fraumeni J.F. (1991). Time trends for bone and joint cancers and osteosarcomas in the Surveillance, Epidemiology and End Results (SEER) Program. National Cancer Institute. National Cancer Institute. In: Review of Fluoride: Benefits and Risks Report of the Ad Hoc Committee on Fluoride of the Committee to Coordinate Environmental Health and Related Programs US Public Health Service. Appendix E and Appendix F.