Don’t take our word for it. Here’s every fact on this site, with its evidence grade, its caveats, and a link to the source, so you can check us.
We start you on the strong ones: findings that have been replicated or come from one large, well-controlled study. Moderate and weak facts are here too, labeled for what they are. Use “Link to this fact” to share a single card.
Health effects / Other evidence
Parabens showed little link to ADHD in this study: only propyl paraben differed between teens with and without ADHD, and it was detected in only about 9% of samples.
Weak evidence: Not significant in adjusted models; HPLC-UV method with many non-detects; single spot urine of short-half-life compounds; parabens are preservatives, not fragrance ingredients.
Source: Terzioğlu 2026, Psychiatry Investigation Link to this fact
Health effects / Review of other studies
A Chinese case-control study cited in the review found higher early-pregnancy urinary monoethyl phthalate (MEP, the breakdown product of the fragrance solvent diethyl phthalate) in women with subclinical hypothyroidism (adjusted OR 1.42), alongside DEHP metabolites (ORs 1.81 to 1.92).
Weak evidence: Single study, association only, review does not report CIs; MEP may mark the whole fragrance mixture rather than DEP itself; DEHP is not a fragrance ingredient.
Health effects / Other evidence
In one experiment, office typing, adding and proof-reading all went slightly faster and more accurately with more ventilation, about 1.7% better per doubling of outdoor air on average, but only typing reached statistical significance.
Weak evidence: Trend fitted to three condition means; ANOVA not significant; 30 young women, 4.6 h; single research group; treat as a lead, not a productivity figure.
Health effects / Other evidence
In one lab experiment with 144 healthy adults, people tested in a room with a lavender oil diffuser scored lower on a working-memory measure (152 vs 173 with no scent) and reacted more slowly than people tested with no scent.
Weak evidence.
Caveats
Single session of about 30 minutes, 48 per group, young adults, no correction for many outcomes, no odor detection or hedonic rating, oil composition and air level not measured, author affiliated with the battery's commercial maker and oils from Tisserand (no funding stated); a lead, not a consumer claim.
Source: Moss 2003, International Journal of Neuroscience Link to this fact
Health effects / Other evidence
In the same experiment, rosemary oil raised a secondary-memory score (200 vs 177 with no scent) but slowed memory reaction times; its overall memory-quality score did not differ from the no-scent group.
Weak evidence: Same small single-session design and caveats; the memory gain came from one subfactor among six tested; rosemary vs control on quality of memory was not significant.
Source: Moss 2003, International Journal of Neuroscience Link to this fact
Health effects / Other evidence
Neither lavender nor rosemary made people feel calmer in this study; rosemary raised self-rated alertness and both oils reduced the fall in contentedness seen in the no-scent group.
Weak evidence: Visual analog scales completed by participants who were deceived about the aim; change scores; single session; not a test of calming claims in anxious people.
Source: Moss 2003, International Journal of Neuroscience Link to this fact
Health effects / Other evidence
In the Atlanta survey, 30.2% of the people who called themselves chemically sensitive also reported asthma, against 12.1% of the whole sample.
Weak evidence: Descriptive cross-tabulation only: the brief prints no asthma rate for the non-sensitive group and tests no association; self-reported asthma.
Source: Caress & Steinemann 2004 (AJPH), American Journal of Public Health Link to this fact
Health effects / Other evidence
Among Atlanta adults who reported being chemically sensitive or unsure, about 3 in 10 (29.9%) said it made shopping in stores difficult and 13.5% said they had lost a job because of it.
Weak evidence: Self-report; asked of 'yes' plus 'not sure' respondents, not only the 199 'yes'; causation of job loss not verified.
Source: Caress & Steinemann 2004 (AJPH), American Journal of Public Health Link to this fact
Health effects / Other evidence
Atlanta (12.6%) and California (15.9%) used the same chemical-sensitivity question; the authors call the two 'statistically similar' without printing a test, though a simple two-proportion check (wiki arithmetic, unweighted) puts the 3.3-point gap at about z = 3.3.
Weak evidence: Unweighted check; the California survey was weighted and the Atlanta brief reports no weighting or design effect, so the real gap test is uncertain.
Source: Caress & Steinemann 2004 (AJPH), American Journal of Public Health Link to this fact
Health effects / Other evidence
In a small German clinic study of 60 children and adolescents, unmedicated children with ADHD detected a rose-like odorant (2-phenylethanol) at lower concentrations than matched healthy children (a large difference, Cohen's d 1.27), while telling odors apart and naming them was normal.
Weak evidence.
Caveats
One single-center study with 20 per group, no replication in our evidence base; it measures lab odor detection of one non-irritant odorant, not reactions to fragrance or chemical intolerance.
Source: Romanos 2008, Biological Psychiatry Link to this fact
Health effects / Other evidence
Children with ADHD who were taking methylphenidate did not differ significantly from healthy controls on odor detection, but this compares separate groups of 20, not the same children before and after treatment.
Weak evidence: Cross-sectional between-group comparison; medication was not randomized; dopamine mechanism is a hypothesis from animal work.
Source: Romanos 2008, Biological Psychiatry Link to this fact
Health effects / Other evidence
Nearly half (48.9%) of US adults with asthma in this survey said their asthma began before age 11, and 32.6% said it began after age 20.
Weak evidence: Retrospective recall; percentages computed on 135 not 137; small sample.
Source: Caress & Steinemann 2009 (Toxicol Ind Health), Toxicology and Industrial Health Link to this fact
Health effects / Other evidence
Among adults who knew or suspected a cause of their asthma (59 people), 13.6% (8) named toxic exposure; counted against all 137 asthmatics that is about 6%, and most could not name a cause.
Weak evidence: Self-reported attribution, tiny numerator (8); the abstract words the 13.6% as a share of asthmatics.
Source: Caress & Steinemann 2009 (Toxicol Ind Health), Toxicology and Industrial Health Link to this fact
Health effects / Other evidence
In the review's cited rat studies, the sunscreen UV filters 4-MBC and octyl-methoxycinnamate lowered circulating thyroid hormone and raised TSH, and benzophenone-2 strongly inhibited human recombinant thyroid peroxidase in vitro.
Weak evidence.
Caveats
Animal and in vitro results only, secondhand; the review states no human or wildlife thyroid studies of UV filters existed. Benzophenone-3 is named as suspected with no BP3-specific result.
Source: Boas 2012, Molecular and Cellular Endocrinology Link to this fact
Health effects / Other evidence
Risk genes at the 12 loci include FOXP2, SORCS3, DUSP6 and SEMA6D, which fit brain development and dopamine regulation, but the authors call gene assignment at most loci uncertain and found little support for previously proposed candidate genes.
Weak evidence: Gene assignment from credible sets and eQTL/Hi-C annotation, not functional experiments; several loci are cryptic.
Health effects / Other evidence
Because ADHD is strongly familial (siblings of affected children have about nine times the risk), studies linking parents' exposures or traits to child ADHD can partly reflect shared genes unless parental genetics are controlled.
Weak evidence: This is our inference from the review, not a finding of the review; whether specific chemical studies are affected has to be checked paper by paper.
Source: Faraone 2019, Molecular Psychiatry Link to this fact
Health effects / Other evidence
Within 32 Taiwanese girls with ADHD, urinary MEP correlated with serum LH (r = 0.646), testosterone (r = 0.780) and free testosterone (r = 0.407), but there was no ADHD-control hormone difference and the correlations are age-adjusted only.
Weak evidence.
Caveats
Within-ADHD correlations in 32 girls, ADHD girls younger and shorter than controls, no pubertal staging, no multiplicity correction; MEP marks diethyl phthalate from fragrance and other products, so no fragrance effect on hormones follows.
Source: Tsai 2020, Toxics Link to this fact
Health effects / Other evidence
The authors conclude that EDCs adversely affect the endocrine system and link to ADHD, but no hormone differed by ADHD status, mediation was not tested, and the authors state all findings would be non-significant after multiplicity correction.
Weak evidence: Within-sample correlations only; cross-sectional.
Source: Tsai 2020, Toxics Link to this fact
Health effects / Other evidence
In zebrafish larvae, butyl-, propyl-, ethyl- and methylparaben lowered T3 and T4 at micromolar concentrations, with toxicity rising with alkyl chain length (BuP > PP > EP > MP).
Weak evidence: Single aquatic in vivo study relayed by a review; survival also fell, so effects may reflect general toxicity; not a human finding.
Source: Azeredo 2023, International Journal of Molecular Sciences Link to this fact
Health effects / Other evidence
The review relays a Quebec meconium study as showing a decrease in free T4 with methylparaben; the primary study's continuous free T4 result is null (1.00, 0.95-1.05) and only a binary low-free-T4 outcome (OR 2.50) points that way.
Weak evidence: Cross-check by this evidence base against the primary; the 16% lower total T3 (relative concentration 0.84) matches Baker 2020.
Source: Azeredo 2023, International Journal of Molecular Sciences Link to this fact
Health effects / Other evidence
The review's title says parabens have 'hazardous effects' on the thyroid axis and it urges WHO and regulators to limit exposure, but its own conclusions say human data are conflicting and the mechanisms are not known.
Weak evidence: Title and recommendation outrun the body of a narrative review; this is a judgment about framing, not about the cited primaries.
Source: Azeredo 2023, International Journal of Molecular Sciences Link to this fact
Health effects / Other evidence
Studies of urinary parabens and attention in children and teens disagree in direction: a Dutch adolescent cohort found better ratings with methylparaben, while a six-country child cohort found worse attention with methylparaben.
Weak evidence: Different ages, tests and designs, all cross-sectional single spot urines with no or limited multiplicity control. Not a fragrance result.
Health effects / Other evidence
In a Turkish two-clinic review of 214 headache patients, about two-thirds of migraine patients (87 of 129) and about a quarter to a third of tension-type headache patients (22 of 85) reported osmophobia, a clear difference (p<0.001) but far from zero in tension-type headache.
Weak evidence.
Caveats
Cureus; retrospective chart review at two clinics; group n and TTH percentage are internally inconsistent (129/85 vs 128/86; 25.9% vs 31.3%); self-reported, no olfactory testing or controls.
Health effects / Other evidence
In the same Turkish study, perfume was the smell that bothered migraine patients most (about 31%, 40 of 128), while in tension-type headache patients food smells came first (10.5%) and perfume third (5.8%).
Weak evidence: Small retrospective chart-based sample; one odor per patient apparently recorded; 'food' not broken down; says nothing about causing attacks.
Health effects / Other evidence
Osmophobia in this study did not differ by sex, age, disease duration or disability score within migraine or tension-type headache groups.
Weak evidence.
Caveats
Underpowered (only 21 migraine men; 85-129 per group); a null result in a small retrospective sample is not evidence of no association; contrasts with a longer-duration link in Gossrau 2022.
Health effects / Other evidence
Higher infant urinary MEP went with slightly lower Mullen early-learning scores across ages 6 to 36 months (composite beta -0.16, 95% CI -0.31 to -0.02), but the fine-motor and visual-reception intervals touch or include zero and nothing was corrected for multiple testing.
Weak evidence: Association only; one small enriched-risk cohort; nominal p-values; MEP may reflect a fragrance or personal-care mixture, not DEP itself.
Source: Sotelo-Orozco 2024, International Journal of Hygiene and Environmental Health Link to this fact
Health effects / Other evidence
The replacement plasticizer DEHTP was associated with lower infant composite and visual-reception scores (composite beta -0.11, 95% CI -0.21 to -0.01) but with a lower, not higher, risk of an ASD classification (RRR 0.50, 0.30-0.83) in the same children.
Weak evidence: Opposite directions in one dataset suggest chance or genetic-risk confounding in an enriched-risk cohort; DEHTP is not a fragrance ingredient.
Source: Sotelo-Orozco 2024, International Journal of Hygiene and Environmental Health Link to this fact
Health effects / Other evidence
In girls in that cohort, higher third-trimester methylparaben went with higher age-3 Mullen scores while propylparaben went with lower scores in both trimesters (about -2.1 to -2.6 per interquartile contrast, 95% credible intervals excluding zero), so parabens did not move in one direction.
Weak evidence: Methylparaben and propylparaben are highly correlated, so opposed weights may be collinearity; units are not plain Mullen points; single small cohort.
Source: Oskar 2024 (MARBLES), Science of The Total Environment Link to this fact
Health effects / Other evidence
In a small Iranian lab study, 20 adults did a digit task-switching test faster (mixed block about 965 to about 845 ms) and more accurately (about 83% to about 89%) on a second run after smelling lavender oil, with more frontal alpha and theta brain-wave power.
Weak evidence.
Caveats
Single group, fixed order (no-odor run first, lavender run second), no control odor, no blinding, one practice block; practice, fatigue and expectation cannot be separated from lavender. Values read from Fig. 4 are approximate.
Source: Afghan 2024, Journal of Medical Signals & Sensors Link to this fact
Health effects / Other evidence
The study's title says lavender 'improves' attentional shifting, but the design cannot show that: nobody did the test with a different smell or with no smell on the second run, so a plain repeat of the test could produce the same gains.
Weak evidence: Authors' own wording is 'may facilitate' in the abstract conclusion but 'improves' in the title and 'effectively induces' in the discussion.
Source: Afghan 2024, Journal of Medical Signals & Sensors Link to this fact
Health effects / Other evidence
Both easy (non-switch) and hard (switch) trials got faster and more accurate after the lavender breaths, which looks like a general speed-up rather than an improvement specific to mental flexibility.
Weak evidence: The switch cost (switch minus nonswitch) was not compared before vs after; inference from the reported bars.
Source: Afghan 2024, Journal of Medical Signals & Sensors Link to this fact
Health effects / Other evidence
Students in the chamber rated the air as worse at 28 C than at 23 C whatever the VOC level, and reports of eye irritation, fatigue, dizziness, headache and stuffy nose rose during the session in every condition, including the low-VOC one.
Weak evidence: Self-report, unblinded to temperature, small groups, symptom rise appears in low-VOC rooms too so cause is unclear (heat, test fatigue or chamber).
Source: Zhao 2025, Science of The Total Environment Link to this fact
Health effects / Other evidence
In an online survey of 461 women in the UAE, 37.5% said air fresheners or deodorizers cause them health problems, most often breathing problems (16.3%), headaches or migraine (15.8%), skin problems (11.9%) and watery eyes or congestion (11.5%).
Weak evidence: Open-link snowball sample of 461 women, mostly young and university-educated; self-report, no exposure measurement; not weighted, so not a UAE prevalence.
Source: AlBlooshi 2026, Frontiers in Toxicology Link to this fact
Health effects / Other evidence
About 3 in 10 women in a UAE online survey (31.7%) reported symptoms from fragranced products in general, and 24.1% from being near someone wearing fragrance, a pattern close to the Western weighted surveys.
Weak evidence.
Caveats
Open-link snowball sample of 461 women, mostly young and university-educated; self-report, no exposure measurement; not weighted, so not a UAE prevalence. The similarity to Steinemann's 32.2% is not independent confirmation because the sample is self-selected.
Source: AlBlooshi 2026, Frontiers in Toxicology Link to this fact
Health effects / Other evidence
In a rat study, 30 days of breathing the vapor of one undiluted commercial perfume (5 mL on cotton in a closed box, 1 h/day) raised markers of oxidative stress (serum MDA 48.7 vs 32.3 nmol/L, lung MDA about 1.8-fold) and lowered antioxidant defenses in serum and lung versus untreated and olive-oil rats.
Weak evidence.
Caveats
Animal only; five male rats per group; one perfume; airborne concentration, box volume and ethanol content never measured; no empty-box or ethanol-vapor control, so confinement stress and solvent are not excluded; no exact p-values; heart and brain oxidative markers did not change.
Source: Al-Sowayan 2026, Frontiers in Pharmacology Link to this fact
Health effects / Other evidence
In the same rat study, perfume-exposed animals had higher blood pressure (130 vs 112 mmHg), heart rate (376 vs 352 bpm), cardiac troponin T (3.41 vs 2.06 ng/mL) and creatinine (0.99 vs 0.60 mg/dL) and lung and heart tissue changes on slides.
Weak evidence: Animal only, n = 5; single troponin and creatinine values with large SE; no empty-box or ethanol control; histology unscored; dose unknown relative to human use.
Source: Al-Sowayan 2026, Frontiers in Pharmacology Link to this fact
Health effects / Other evidence
The perfume rat study found no change in serum brain-injury markers (GFAP, UCH-L1) and no significant change in brain KLF4 expression; the brain claim rests on cortical gliosis on slides alone.
Weak evidence: The abstract says Klf4 was downregulated, but the results call it a non-significant trend; five rats per group; histology qualitative.
Source: Al-Sowayan 2026, Frontiers in Pharmacology Link to this fact
Health effects / Other evidence
In this five-country survey 34.6% of adults scored as highly chemically intolerant (25% in Mexico to 51% in India), far above clinician-diagnosed rates of about 0.5 to 6.5% that the authors cite.
Weak evidence: Opt-in online panel with 4 to 21% of records removed for quality; the questionnaire cutoff measures reported symptoms, not diagnosed illness. Do not treat as population prevalence.
Source: Palmer & Kattari 2026, J Xenobiotics, Journal of Xenobiotics Link to this fact
Health effects / Other evidence
In a 20-person Iranian EEG study with no placebo or no-odor group, brain-wave responses (P300) during a number-switching task were slightly larger at one electrode (+0.29 microvolts, P = 0.029) and slightly earlier at another (about 3 to 8 milliseconds) after participants sniffed lavender essential oil for a few breaths; the authors themselves say the design limits causal conclusions.
Weak evidence.
Caveats
Single-group pre-post, n = 20, no blinding or control odor, oil not characterized, four uncorrected tests, a switch-latency statistic that is internally inconsistent, and behavioral results asserted but not shown; practice, fatigue and expectation are not separable from the odor.
Source: Afghan 2026, Journal of Medical Signals & Sensors Link to this fact
Health effects / Review of other studies
In six small case-control studies reviewed in 2026 (30 to 80 per group), children with ADHD had different gut bacteria on average (lower alpha diversity, fewer Lactobacillus and Bifidobacterium, lower fecal short-chain fatty acids), but no study measured chemical exposure, the microbiome and ADHD in the same children.
Weak evidence: Cross-sectional, heterogeneous methods; one tabulated study is itself a meta-analysis; effect sizes not reported.
Source: Wu 2026, Biological Psychiatry Global Open Science Link to this fact
Health effects / Review of other studies
The proposed route from chemical exposure through gut dysbiosis (barrier leak and inflammation, fewer short-chain fatty acids, stress-hormone dysregulation) to ADHD is a hypothesis: the supporting barrier, inflammation and HPA data are from rodents or from other conditions, and the review's only EED-linked human microbiome evidence is two cross-sectional urine studies.
Weak evidence: Animal doses described as often supraphysiological; mediation was never tested; the review's own text says direct evidence is lacking.
Source: Wu 2026, Biological Psychiatry Global Open Science Link to this fact
Health effects / Other evidence
Modeled ADHD prevalence varies about sixfold across world regions, highest in Australasia (about 5.4%) and lowest in South Asia and sub-Saharan Africa (about 0.8-0.9%), and rises with national income-education development (SDI).
Weak evidence.
Caveats
The authors attribute the pattern largely to differences in perceived impairment, access to child psychiatric care and diagnostic practice, not demonstrated biological or environmental differences; many regional intervals are wide.
Source: Cortese 2026 (Mol Psychiatry), Molecular Psychiatry Link to this fact
Health effects / Other evidence
A rise in ADHD diagnoses in clinics and surveys is not by itself evidence that ADHD itself, or any environmental exposure effect, has increased: the GBD model of standard-criteria prevalence did not rise, and its authors point to awareness and recognition.
Weak evidence: GBD input data end about 2017, so it cannot test post-2017 or pandemic-era changes; the contrast with survey and registry trends is not resolved by this paper alone.
Source: Cortese 2026 (Mol Psychiatry), Molecular Psychiatry Link to this fact
Health effects / Other evidence
Four small differences in DNA methylation at single sites in newborns' cord blood tracked their mothers' early-pregnancy butylparaben, isobutylparaben, isopropylparaben and summed paraben levels, but these are molecular markers, not health outcomes.
Weak evidence.
Caveats
Single cohort with no replication; hits did not survive restriction to the 450k array and one moved above the threshold with a different imputation; isopropylparaben hit rests on 27 exposed mothers and an inflated QQ plot (lambda 1.62); no health outcome analyzed.
Source: Mäkinen 2026, Environment International Link to this fact
Health effects / Other evidence
The Finnish study could not reproduce the cord-blood methylation sites reported in the one earlier paraben study, a California farmworker cohort; its one nominal match went the opposite direction.
Weak evidence: Different populations, exposure imputation and sex-stratified design; the authors offer explanations but no test of them.
Source: Mäkinen 2026, Environment International Link to this fact
Health effects / Other evidence
Cities where outdoor ozone was most strongly linked to daily deaths (top 10 of 95 US cities, average 1.1% per 10 ppb) tended to be older, slower-growing and less air-conditioned than the 10 cities with the weakest links (average 0.1%), which the author reads as a possible sign of indoor ozone exposure.
Weak evidence.
Caveats
A descriptive city-level pattern with no statistics or adjustment; climate, age structure and co-pollutants are alternative explanations, and Dallas (91% AC, 1.1%), Denver and San Diego go against it. The author says the studies were not designed to test the hypothesis.
Source: Weschler 2006, Environmental Health Perspectives Link to this fact
Health effects / Other evidence
After adjusting for study methods, ADHD prevalence did not differ between North America and Europe (p=0.40), arguing against ADHD being a purely North American cultural construct; only Africa and the Middle East (four studies each) came out lower.
Weak evidence: Few studies for Africa and the Middle East (authors urge caution); null on region is not proof of cultural equivalence; the paper does not claim culture has no influence.
Source: Polanczyk 2007, American Journal of Psychiatry Industry-funded Link to this fact
Health effects / Other evidence
ADHD prevalence in the 2007 review was higher in boys than girls (about 10% vs about 4%) and in children than adolescents (about 6.5% vs about 2.7%).
Weak evidence: Values read from a chart, not printed; sex and age strata came from fewer than half the studies and were not in the adjusted model.
Source: Polanczyk 2007, American Journal of Psychiatry Industry-funded Link to this fact
Health effects / Other evidence
In a 2011 German study of 29 adults with ADHD and 29 matched controls, smell threshold, odor discrimination and odor identification did not differ between groups (threshold 10.0 vs 10.0, p = 0.91), so the childhood finding of lower odor thresholds in ADHD was not replicated in adults.
Weak evidence: Single small center; patients had more smokers (17 vs 7) and comorbidities; one lab, same group as the 2008 child study.
Source: Schecklmann 2011, Neuropsychobiology Link to this fact
Health effects / Other evidence
In the same adult ADHD study, brain oxygenation measured by near-infrared spectroscopy during smelling rose-like 2-phenylethanol and bergamot-like linalool was significant in more regions in controls than in ADHD patients, but only 3 of 40 direct group comparisons reached p < 0.05, and no correction for multiple comparisons was applied.
Weak evidence: Exploratory; the authors acknowledge high variance, low power and no multiple-comparison correction; 'significant versus not significant' is not a group difference.
Source: Schecklmann 2011, Neuropsychobiology Link to this fact
Health effects / Other evidence
Among adults with ADHD, stronger brain responses to high-dose linalool went with more childhood ADHD symptoms (r about 0.41 to 0.58) and higher trait impulsivity (r about 0.44 to 0.50), mostly in the combined subtype.
Weak evidence: Small subgroup, one odorant condition, no correction; correlation only; one lab.
Source: Schecklmann 2011, Neuropsychobiology Link to this fact
Health effects / Review of other studies
A 2025 review's 'Schecklmann 2011' result of higher odor discrimination after stimulant washout comes from a different paper (children on and off methylphenidate), not this adult study, where discrimination did not differ (13.55 vs 13.21 of 16, p = 0.43).
Weak evidence: The J Psychiatr Res paper itself has not been read in this evidence base; attribution rests on its PubMed abstract.
Source: Schecklmann 2011, Neuropsychobiology Link to this fact
Health effects / Other evidence
In the same review, ADHD subtype did not predict sensory processing problems, but co-occurring oppositional defiant disorder and anxiety predicted more severe problems (ODD with tactile and auditory under-responsiveness, anxiety with sound hypersensitivity).
Weak evidence: Several of the supporting studies are by the review's author, single clinic, parent-report; associations only.
Source: Ghanizadeh 2011, Psychiatry Investigation Link to this fact
Health effects / Other evidence
Olfactory findings in the 2011 review were mixed: one study found a lower odor threshold in ADHD (improved by stimulants) with normal identification and discrimination, another found poorer odor identification; the review did not reconcile them.
Weak evidence: Small samples; medication status and task differ between the studies; no odor-exposure or fragrance measure.
Source: Ghanizadeh 2011, Psychiatry Investigation Link to this fact
Health effects / Other evidence
Pooling five double-blind crossover trials, restriction (elimination) diets reduced ADHD symptoms by about 0.29 standard deviations, roughly a third of a stimulant effect, with a conservative estimate that about 33% of children with ADHD respond.
Weak evidence.
Caveats
Only a handful of old, small trials; the effect is not specific to food colors; the abstract and text give different confidence intervals (0.07-0.53 vs 0.16-0.52); one outlying study was excluded to resolve heterogeneity.
Source: Nigg 2012 Industry-funded Link to this fact
Health effects / Other evidence
Laboratory attention tests showed a reliable but small-sample effect of food colors (6 studies, 154 children, 0.27, 95% CI 0.07 to 0.47); the authors judge it too small and too old to support clinical or policy recommendations.
Weak evidence: Only 68 children in FDA-color-only studies; the colors-only result is vulnerable to single-study removal.
Source: Nigg 2012 Industry-funded Link to this fact
Health effects / Other evidence
The paper's estimate that about 8% of children with ADHD may have symptoms related to food colors multiplies a 24% color-reaction rate by a 33% diet-responder rate, and the authors say its confidence interval would be large and could not be computed.
Weak evidence: Rate comes from children preselected as diet responders; not a population prevalence.
Source: Nigg 2012 Industry-funded Link to this fact
Health effects / Other evidence
Studies measuring lead in hair gave much larger ADHD-symptom correlations (inattention r .38, 4 studies) than other methods (r .14, 23 studies); the authors call hair a less accurate measure, so the pooled value excluding hair is the safer figure.
Weak evidence: Only 4 to 5 hair studies, three from one research team; hyperactivity/impulsivity difference not significant (p = .12).
Source: Goodlad 2013, Clinical Psychology Review Link to this fact
Health effects / Other evidence
The 2013 lead and ADHD meta-analysis found no larger effect in studies with higher average blood lead (beta = .03 and .11, both nonsignificant), including samples averaging about 1 ug/dL, which the authors read as consistent with no clear threshold.
Weak evidence: Between-study comparison, not a within-person dose-response; authors say it suggests but does not prove a linear association.
Source: Goodlad 2013, Clinical Psychology Review Link to this fact
Health effects / Other evidence
The 2013 lead and ADHD meta-analysis tested publication bias only with a fail-safe N (2,006 null studies for inattention; 937 for hyperactivity/impulsivity), with no funnel plot or Egger test, and excluded non-English papers and dissertations.
Weak evidence: Fail-safe N is a weak bias check; up to 4 non-English papers may have been eligible.
Source: Goodlad 2013, Clinical Psychology Review Link to this fact
Information, not medical advice. See also: myths we won’t tell you.