Problems in exposure assessment of fluoride in drinking water

We developed a source documentation approach that identified fluoride content of drinking water at the state or local level to estimate fluoride concentrations from public water systems. We then compared estimates from this approach with estimates obtained from a single source, the 1992 Centers for Disease Control and Prevention (CDC) Fluoridation Census.

We used residential histories from a case-control study. For each residence we attempted to determine fluoride concentrations using the 1992 CDC Fluoridation Census. For the source documentation method we utilized multiple sources from state and local contacts to verify and collect additional data. We compared the fluoride estimates obtained by the two methods.

When fluoride values were found using both methods, there was good correlation (Kendall's tau = 0.85; 95% confidence interval = 0.79, 0.90) and concordance was 96 percent. We obtained over 99 percent of the fluoride values needed using source documentation as compared to 49 percent of the values needed when we used a single publication. When fluoride values were missing using the 1992 CDC Fluoridation Census, 21 percent had source documentation estimates of at least 0.7 ppm.

Researchers need to consider limitations of using a secondary data source to estimate fluoride in drinking water, particularly in studies where exposure to fluoride is the primary exposure of interest.