Investigation 006
The Fluoride Decision
A quarter of a tooth, and everything it was traded for
Environment & exposure
Fluoride & cognition
Seventy-four studies, a federal court, and one number that keeps moving: where does the dose-response actually start?
What we are reading
Gathered for the investigation and not yet written up. Each note says what the paper does and does not show.
- Taylor KW, Eftim SE, Sibrizzi CA, Blain RB, Magnuson K, Hartman PA, Rooney AA, Bucher JR. Fluoride Exposure and Children's IQ Scores: A Systematic Review and Meta-Analysis.10.1001/jamapediatrics.2024.5542 · PMID 39761023
The National Toxicology Program's own meta-analysis, and the single most important number in the fluoride debate. 74 studies, 10 of them cohorts. Pooled SMD -0.45 (95% CI -0.57 to -0.33). Individual-level data: IQ falls 1.63 points per 1 mg/L of URINARY fluoride (1.14 points among low-risk-of-bias studies only). THE CRITICAL NUANCE, which both sides misquote: for DRINKING WATER measured alone the inverse association held below 4 and below 2 mg/L but was NULL below 1.5 mg/L. For URINARY fluoride it stayed inverse below 1.5. The authors' own conclusion names 'limited data and uncertainty in the dose-response association ... when fluoride exposure was estimated by drinking water alone at concentrations less than 1.5 mg/L'. US fluoridation is 0.7 mg/L. So the honest sentence is: the effect is real and dose-dependent above 1.5, and at 0.7 the drinking-water evidence does not resolve it while the urinary evidence suggests it does not stop.
- Taylor KW, Eftim SE, Sibrizzi CA, Blain RB, Magnuson K, Hartman PA, Bucher JR, Rooney AA. Addressing Critiques of the Evidence Linking Fluoride and Children's IQ.10.5334/aogh.4853 · PMID 41393310
The NTP authors answering their critics in December 2025. Their claim here is stronger than in the JAMA paper: 'among the high-quality evidence, inverse associations were still observed at fluoride exposure levels below 1.5 mg/L, based on both urinary AND drinking-water measurements'. Quote it as their position, not as settled fact, and put it next to the JAMA abstract, which says the water-only association was null below 1.5. That gap between a team's meta-analysis and the same team's defence of it is itself worth airtime.
- Veneri F, Vinceti M, Generali L, Giannone ME, Mazzoleni E, Birnbaum LS, Consolo U, Filippini T. Fluoride exposure and cognitive neurodevelopment: Systematic review and dose-response meta-analysis.10.1016/j.envres.2023.115239 · PMID 36639015
The best dose-response modelling anyone has done: a cubic-spline one-stage meta-analysis of 30 studies. Highest vs lowest exposure differs by -4.68 IQ points (95% CI -6.45 to -2.92); for drinking water specifically, -5.60. The curve is substantially LINEAR above 1 mg/L — about -3.05 IQ points per 1 mg/L up to 2 mg/L, and steeper above that. Urinary fluoride shows a weaker slope starting around 0.28 mg/L, which they note corresponds to roughly 0.7 mg/L in water. AND THE LINE THAT MUST BE READ ON AIR: 'the inverse association ... was particularly strong in the studies at high risk of bias, while NO ADVERSE EFFECT emerged in the ONLY study judged at low risk of bias.' They could not rule out residual confounding.
- Choi AL, Sun G, Zhang Y, Grandjean P. Developmental fluoride neurotoxicity: a systematic review and meta-analysis.10.1289/ehp.1104912 · PMID 22820538
The Harvard paper that started the modern argument. 27 studies, standardised mean difference -0.45 — the identical figure the NTP would land on thirteen years later with 74 studies. Its own limitation is the one everyone forgot to mention when it went viral: these were high-fluoride endemic areas in China, comparing badly-affected villages with less-affected villages, not fluoridated cities with unfluoridated ones.
- Grandjean P. Developmental fluoride neurotoxicity: an updated review.10.1186/s12940-019-0551-x · PMID 31856837
Grandjean's update, and the paper that moved the argument from ecological studies to individual exposure: three prospective cohorts from Mexico and Canada with measured maternal urinary fluoride, all negative for cognition. His benchmark-dose calculation concludes safe exposures are 'likely to be below currently accepted or recommended fluoride concentrations in drinking water'. Also makes the argument the dental literature has never answered well: topical fluoride works in the mouth, so why swallow it?
- Kumar JV, Levy SM, Warren JJ. An update on community water fluoridation, part 2: Fluoride exposure and children's intelligence (IQ) scores.10.1016/j.adaj.2026.02.010 · PMID 41941356
The 2026 counter-case, in the Journal of the American Dental Association. Conclusion: 'no association between F concentrations relevant to CWF and IQ', the endemic-area studies are methodologically weak, and 'there is no established biological mechanism to explain F's effect on IQ'. Two things to say about it on air, both true: the methodological critique of the Chinese cross-sectional literature is largely correct, and this is the dental profession reviewing the safety of the dental profession's flagship public health measure. Give it its argument and give it its address.
- Kumar JV, Moss ME, Liu H, Fisher-Owens S. Association between low fluoride exposure and children's intelligence: a meta-analysis relevant to community water fluoridation.10.1016/j.puhe.2023.03.011 · PMID 37120936
The other side's primary evidence, and it is a real analysis, not a press release. Restricted to eight studies from NON-endemic areas: SMD 0.07 (95% CI -0.02 to 0.17), I-squared 0%, no significant fluctuation across the fluoride range on spline modelling. Their honest closing sentence matters: 'the reported association observed at higher fluoride levels in endemic areas requires further investigation.' Eight studies is a small evidence base to clear a public health measure with, and they say so.
- Gunasekaran S, Sakthivel S, Rajan RE, Balasubramanian M, Latkar YS, Mathew BM. Fluoride-induced effects on cognitive development in Indian children: A systematic review and meta-analysis.10.4103/jisppd.jisppd_259_25 · PMID 41026552
Eleven studies, 6,000-plus children, September 2025. Above 2.0 ppm the pooled effect is -6.5 IQ points (95% CI -7.3 to -5.7), worst in 6-to-10-year-olds. Useful because it is recent and specific about the exposure level; limited by the same problems as the rest — mostly cross-sectional, substantial heterogeneity, and the authors report some publication bias themselves.
- Tang QQ, Du J, Ma HH, Jiang SJ, Zhou XJ. Fluoride and children's intelligence: a meta-analysis.10.1007/s12011-008-8204-x · PMID 18695947
The 2008 Chinese meta-analysis, sixteen case-control studies, weighted mean difference about -5 IQ points. Mostly of historical interest now, but it establishes that the signal was in the literature a full four years before Harvard found it and sixteen years before a US court did.
Connected on the map
Also in Environment & exposure