Municipal water fluoridation, adolescent IQ, and cognition across the life course: Evidence from the Wisconsin Longitudinal Study.

Table 1 describes a) outcome measures of IQ at age 16 and of cognition at several ages in later life and b) individual- and school-level measures that may confound associations between CWF and cognition. While cognitive outcomes are highest for Wisconsin Longitudinal Study (WLS) (5) participants who were first exposed to CWF in late adolescence, it is also true that socioeconomic advantage is highest among those sample members.

Fig. 1.

As shown in Fig. 1, participants exposed to CWF did not perform significantly worse (or better) than their peers who were never exposed; this is true regardless of the age at which cognition was assessed and regardless of whether we restrict the sample to participants who did not move since ages 11 or 1. Only two of the 45 key coefficients are statistically distinguishable from zero—about what we would expect by chance.

We replicate and extend recent evidence (2) that CWF is not associated with reduced adolescent cognition. Our analysis improves on previous work, however, by using a measure of adolescent IQ (instead of academic achievement) and by more carefully considering adolescents’ duration of time in communities (and thus consistency of exposure to CWF). Like previous work, our research is limited by our inability to directly quantify adolescents’ consumption of fluoride (e.g., via urine samples).

We find no evidence that CWF is associated with lower adolescent IQ or cognition later in life.

We utilize data from the WLS, which includes a 1/3 random sample (n = 10,317) of the Wisconsin high school graduating class of 1957. WLS data are available at https://www.ssc.wisc.edu/wlsresearch/data/, and were collected in accordance with protocols approved by the Institutional Review Board at the University of Wisconsin at Madison.

WLS data include the addresses of participants’ schools in 1957; because more than 90% of WLS records have been linked to the decennial US Censuses of 1940 and 1950—when sample members would have been ages 1 and 11, respectively—we are able to identify which sample members lived in their same communities over time.

Following the same methodology used by Warren et al. (2), we characterize sample members’ CWF exposure through age 14—prior to when IQ tests were typically administered—using information about CWF practices (6) and naturally occurring fluoride levels in untreated well water (7). Students were classified as exposed from birth if one or more untreated wells in their county had naturally sufficient fluoride levels. Otherwise, they were classified as being exposed to CWF beginning from the year in which their community began water fluoridation. See the SI Appendix available on the journal’s website for more details.

Adolescent IQ was measured using the standardized Henmon-Nelson test by schools in students’ freshman and junior years. Cognition at ages 53, 64, and 72 was measured using the similarities task of the Wechsler Adult Intelligence Scale (WAIS). Cognition at age 80 was measured using the Modified Telephone Interview for Cognitive Status (TICSm).

Control variables—introduced to account for selection into CWF—were collected in 1957 and include parental education, father’s occupation, family income, and size of community. We also include school means of family socioeconomic background measures. Unfortunately, the WLS did not collect information about dental care in adolescence or about non-water sources of fluoride exposure.

Missing values for student-level background measures were imputed using chained equations, with 20 imputed datasets. OLS regressions were estimated using “mi estimate” commands in Stata/SE 19.5.

We appreciate assistance from Carol Roan and Joseph Savard. Since the 1990s, the Wisconsin Longitudinal Study has been supported mainly by the National Institute on Aging (NIA) (R01AG060737 and R01AG009775), the Vilas Estate Trust, the NSF (SBR932066), and the Spencer Foundation. We also benefited from NIA center grant support (P30AG017266 and P30AG066613).