Nucleus of the fluoride cluster. Fluoride is unusual among the wiki’s exposures: it is
deliberately added to drinking water and toothpaste for a real benefit (caries prevention) and
is a suspected developmental neurotoxicant at higher doses. The whole decision is a dose-stratified
benefit-versus-harm balance. As of 2026-08-14 all arms are landed — the caries-benefit pole
(2024 Cochrane) closed the last gap, joining the child-IQ and skeletal harm arms — so the nucleus now
carries the induced dose-stratified benefit/harm synthesis (below), not a list of gaps.
The one landed finding: higher fluoride, lower child IQ — above 1.5 mg/L
The US National Toxicology Program’s 2024 OHAT systematic review concluded, at moderate confidence, that higher fluoride exposure is associated with lower IQ in children — but the exposure bound is the whole decision and is routinely stripped off in secondary reporting. Quoted intact:
- «This review finds, with moderate confidence, that higher estimated fluoride exposures (e.g., as in approximations of exposure such as drinking water fluoride concentrations that exceed the World Health Organization Guidelines for Drinking-water Quality of 1.5 mg/L of fluoride) are consistently associated with lower IQ in children.» (National Toxicology Program, 2024)
- Below that level the review does not find a settled effect: «Associations between lower total fluoride exposure [… concentrations … lower than the WHO Guidelines … of 1.5 mg/L …] and children’s IQ remain unclear.» (National Toxicology Program, 2024)
A bare «fluoride lowers IQ» with no exposure bound overstates this source — the below-1.5 stratum (which includes US community-water fluoridation at 0.7 mg/L) is an insufficient-evidence state, not a demonstrated null and not a demonstrated harm. The dose-threshold decision, the study table, and the transportability appraisal live on Fluoride and Child IQ.
The exposure map — where the decision-relevant doses sit
| Exposure | Fluoride (drinking water) | What it is |
|---|---|---|
| US community water fluoridation | 0.7 mg/L | PHS-recommended added dose (historically 0.8-1.2) |
| WHO water guideline | 1.5 mg/L | NTP’s «higher»-exposure referent; anti-fluorosis, not anti-IQ |
| EPA secondary (non-enforceable) | 2.0 mg/L | severe dental-fluorosis protection |
| EPA MCL (enforceable) | 4.0 mg/L | skeletal-fluorosis protection |
Total fluoride exposure is not the water number: even in optimally fluoridated areas it «can vary widely … based on personal habits in the use of dental products and consumption of beverages such as black tea» (National Toxicology Program, 2024). ~1.9 million US residents (0.59%) are on community water naturally ≥1.5 mg/L — a small stratum, but the one where the moderate-confidence harm sits.
The benefit/harm menu — one axis, most arms still open
The telos’s one-axis (health) rule governs here: this is a health-outcome menu, not a fluoridation-policy verdict. Name each arm and its status; do not net them into a scalar.
- Child IQ / neurodevelopment (HARM): moderate confidence above 1.5 mg/L, unclear below. Landed.
- Caries prevention (BENEFIT): the reason fluoride is added at all — NTP explicitly excluded it
(«Benefits of fluoride with respect to oral health are not addressed in this monograph»
[EXTRACTED (NTP - Fluoride Neurodevelopment Monograph 2024) chunk 01]). Landed (2024 Cochrane, low certainty): contemporary (after-1975) CWF gives a dmf difference of «0.24 (95% CI -0.03 to 0.52)» — ~a quarter-tooth, CI crossing no-benefit — «with smaller effect sizes than pre-1975 studies», the large familiar benefit being a pre-toothpaste-era number (Iheozor-Ejiofor et al., 2024). At the same 0.7 mg/L dose, mild dental fluorosis is established (aesthetic concern ~12%, any level ~40%, low certainty). Full arm, era-attenuation and substitution mechanism on Fluoride and Caries Prevention. - Dental fluorosis (mild HARM at the fluoridation dose): the one harm established at 0.7 mg/L rather than above 1.5 — «12% (95% CI 8% to 17%)» aesthetic concern, ~40% any level, low certainty; a cosmetic/enamel outcome, below the IQ/fracture patient-important tier (Iheozor-Ejiofor et al., 2024). Detail on Fluoride and Caries Prevention.
- Skeletal fluorosis / bone fracture (HARM at high dose): the basis for the EPA 4.0 mg/L MCL. Landed (low confidence) — a dose-response MA finds fracture risk rising above ~1.5 mg/L water fluoride (overall RR 1.06/1.19/1.35 at 2/3/4 mg/L, all CIs crossing 1.0), significant only in postmenopausal females (RR 1.26, 95% CI 1.10-1.46 at 1.0 mg/L); BMD and fracture dissociate (a BMD rise is not a bone benefit). Full arm, U/J adjudication and BMD-surrogate trap on Fluoride and Bone Health (Mazzoli et al., 2025). The ~1.5 mg/L skeletal threshold coincides with the NTP IQ threshold — the same exposure line turns two harm arms upward.
- Adult cognition: low confidence, «a lack of evidence of an effect» (two studies). (National Toxicology Program, 2024)
- Other child neurodevelopmental (ADHD, infant neurobehavioral): low confidence. (National Toxicology Program, 2024)
The cluster-closing synthesis: the balance is DOSE-STRATIFIED, not a single benefit-vs-harm clash
With all arms landed, the benefit<->harm relationship is finally addressable — and the honest
structure is a dose-stratified distinction, not a [[tension]]. Running the not-joined checks
(Cooper, ii): the caries benefit and the serious harms hold at different exposures, so they are
consistent once matched by dose — filing a blanket «benefit vs harm» tension would compare
non-joined positions (a fake tension). What the fabric holds instead is a single balance whose sign
changes along the fluoride axis:
| Exposure | Benefit (caries) | Harm (IQ · skeletal · fluorosis) | Net decision structure |
|---|---|---|---|
| ~0.7 mg/L (US CWF) | modest, low/very-low certainty (dmf ~0.24, CI crosses 0), era-attenuated | IQ + fracture insufficient-evidence (not null, not established); only mild dental fluorosis established (~12%/40%, low) | decision under deep uncertainty on the serious-harm side — neither pole dominates; no joined clash of two established effects |
| >1.5 mg/L | little extra benefit (largely captured by ~1 mg/L + other sources) | child-IQ harm enters at moderate confidence; overall fracture risk begins rising | balance tips toward mitigation — harm turns on while marginal benefit is ~flat |
| >3-4 mg/L | none additional | skeletal fracture harm strengthens (EPA MCL 4.0) | mitigation (defluoridation / alternative source) |
The load-bearing insight (the hidden_insight a tension would have carried): at the fluoridation
dose there is no joined clash — the serious harms are insufficient-evidence rather than demonstrated,
so the 0.7 mg/L decision is under uncertainty, not between two knowns. The genuine trade-off only
appears above ~1.5 mg/L, where the moderate-confidence IQ harm and the fracture harm turn on
against a diminishing marginal caries benefit — and there the same ~1.5 mg/L line that turns up two
harm arms (see Fluoride and Child IQ, Fluoride and Bone Health) is where the benefit has
already flattened. Below it, the decision is dominated by uncertainty and by the person’s weighting of
a small caries gain against a mild cosmetic fluorosis risk (layer 3), not by any established serious
harm.
This is one axis (health): it is a benefit/harm map, not a fluoridation-policy verdict. Cost, environmental load and equity are separate axes the caries source touches (named on Fluoride and Caries Prevention) and are not priced in here; the health-outcome weighting stays the person’s.
Independence: NTP and Taylor 2025 are ONE backing (VERIFIED), not two
NTP’s dose-response rests on its own companion meta-analysis («DTT Meta-analysis, Taylor et al.
2024») — «The NTP authors of this monograph conducted a companion systematic review and meta-analysis»
(National Toxicology Program, 2024). That companion is now held as
the peer-reviewed primary Taylor 2025 (JAMA Pediatrics 179(3):282-292), and the one-backing framing
is CONFIRMED, not assumed: its ref 12 IS this NTP monograph, its authors are the NIEHS Division of
Translational Toxicology (Taylor, Rooney, Bucher — shared with the monograph), and «NIEHS did have a
role in the review approval of the manuscript»
(Taylor et al., 2025). So Taylor 2025 does not raise
confidence by type-E convergence — it is the same stream’s evidence in peer-reviewed form (a type-F
refinement of the harm arm, landed on Fluoride and Child IQ), counted as one line with NTP.
Confidence stays low on the below-1.5 stratum accordingly.
The third line (Veneri 2023): qualified-E, not independent backing — and a friction
A different stream did arrive — Veneri, Vinceti, Filippini et al. (CREAGEN / University of Modena), an
SR + cubic-spline dose-response (Veneri et al., 2023). It clears the necessary conditions for a type-E independent line: no author shared with
NTP/Taylor, independent funding (no NIEHS role), and different analytic instruments (ROBINS-E
RoB tool; one-stage cubic-spline). But independence of team is not independence of evidence — Veneri
pools a near-fully-shared primary-study base (its 33 studies sit almost entirely inside NTP’s 72 /
Taylor’s 59; the fluoride-IQ census is near-closed, sharing every marquee cohort — ELEMENT/Bashash,
MIREC/Till+Farmus, Dunedin/Broadbent). Two MAs pooling the same primary cohorts agree on the pooled
direction by construction — so Veneri’s directional concordance (water DR −3.05 IQ pts/mg/L above 1
mg/L) is shared-data robustness, NOT the independent second backing that would raise confidence by
convergence. This is a qualified / partial E (robust to team + analytic choices), and it is
NOT marked [E-independent] — doing so would be laundered-E. Confidence stays low.
Two forces cut against even a confidence bump. First, Veneri’s low-RoB and cohort-only strata null the association (low-RoB MD +1.11 NS; cohort-only −0.74 NS), the opposite of NTP/Taylor’s «holds in low-RoB studies» reading — a genuine type-D friction that traces to the RoB instrument (ROBINS-E’s aggressive confounding downgrade rates Bashash HIGH; OHAT rates it low), filed on Fluoride and Child IQ. Second, the same shared-data property means the two families cannot independently confirm each other on the low-dose stratum at all — that stays a genuine gap. A truly independent line would need a new primary cohort at fluoridation-relevant exposure, ideally a US one (none exists yet).
Note also the halo-across-a-lab caution: the Modena/Vinceti group also supplies the now-landed
skeletal-harm arm (Mazzoli 2025, on Fluoride and Bone Health) — different outcome, so no
laundering, but the cluster now rests on two arms (IQ, skeletal) from the same lab, and that is
not independent evidential breadth. Independent replication of either arm by a different group is a
named gap; the skeletal arm’s low confidence is held there for exactly this reason.
Why this is a decision-change, not a restatement
Public discourse collapses to fluoride is/ isn’t safe. The fabric’s move is the stratum split: a moderate-confidence harm signal that is real above ~1.5 mg/L water fluoride and unresolved at the 0.7 mg/L US fluoridation dose — so someone on a private well testing >1.5 mg/L faces a different decision (the harm arm is live; mitigation is on the table) than someone on optimally fluoridated municipal water, where at 0.7 mg/L the serious harm arms are insufficient-evidence, only mild dental fluorosis is established, and the caries benefit is genuine but small and low-certainty in the contemporary era (Fluoride and Caries Prevention). The exposure number, not the word «fluoride», carries the decision — and at the fluoridation dose the whole menu is small effects held at low certainty, on both sides.