Acetaminophen and Autism: NBC News Cites a Fatally Flawed “Twins” Study—and Ignores the Balance of the Science

Acetaminophen science is entirely misrepresented by NBC News. The JAMA "Twins Study" they cited is fatally flawed. Acetaminophen is a likely contributor, not the only causal factor.

Rapid outline

1. NBC reduces a complicated literature to a single Swedish sibling comparison and calls the resulting “null” definitive. They misrepresent “the bulk” of the literature.

2. That “null” rests on design decisions that engineer attenuation: timing collapse to “ever‑use,” OTC under‑measurement, prescription‑only dose averaging, discordant‑sibling selection, secular drift, and multicollinearity from co‑medication.

3. Exposure‑anchored human evidence—cord plasma and meconium biomarkers—shows positive associations and dose‑response.

4. A six‑cohort European meta‑analysis aligns with risk elevations for autism‑spectrum and ADHD symptoms.

5. Developmental pharmacology and animal models map plausible mechanisms and sensitive windows.

6. The responsible message: medically indicated use only; lowest dose; shortest duration; invest in exposure‑accurate causal tests.

The media claim and what the JAMA study actually did

NBC’s article—syndicated on MSN as “What the science says about acetaminophen, pregnant mothers and autism”—hangs a public reassurance on a single register‑based paper in JAMA (Ahlqvist et al., 2024). The headline “no association” applies only to the within‑family (full‑sibling) models. In conventional cohort models, the familiar positive signals appear. NBC’s treatment skips the design’s fragility under the specific data constraints in play.

The study’s primary exposure comes from a single first‑antenatal interview (8–10 weeks) for all birth years, with later‑pregnancy entries available in a subset of years. Within those years, 81.7% of “ever‑users” reported use only at that first visit; 5.6% reported use only later. Real‑world use is overwhelmingly over‑the‑counter. Only 6.6% of individuals who reported N02B analgesic use at the first antenatal visit had any prescription dispensation recorded in the prior 180 days. The paper’s “dose” analysis exists from 2005 onward and computes mean daily dose over the entire pregnancy from prescriptions only—a guaranteed dilution of episodic, clinically relevant bursts. These are not quibbles; they are the measurement backbone of the study.

Sibling comparisons then restrict to families discordant for exposure—systematically different from concordant families on smoking, psychiatric history, health‑care utilization, and co‑medication. The design further admits carryover/calendar bias: diagnostic intensity for ADHD and autism rose across the study period while exposure prevalence and patterns shifted. Layer on joint adjustment for correlated co‑medications (opioids, non‑aspirin NSAIDs, psychotropics) and health‑care utilization, and the remaining exposure signal is absorbed into noise. In the paper’s own cross‑analgesic panel, everything drifts toward the null under sibling control. Aspirin even flips to “protective” in sibling models—an internal red flag.

The balance of the science NBC omitted

Biomarker‑anchored human evidence: signal and gradient

The strongest epidemiology minimizes self‑report and measures fetal load directly. In the Boston Birth Cohort (Ji et al., JAMA Psychiatry, 2020), cord‑plasma acetaminophen metabolites in the third tertile versus first tertile were associated with ADHD OR 2.86 (95% CI 1.77–4.67) and autism OR 3.62 (1.62–8.60), with a graded increase across tertiles after adjustment. Meconium‑based exposure in JAMA Pediatrics (Baker et al., 2020) similarly tracked higher ADHD odds and, critically, mapped to altered frontoparietal network connectivity, providing an imaging bridge between exposure and phenotype. These studies are not immune to confounding, but they sidestep the exact misclassification that sabotages register‑only designs.

A harmonized six‑cohort European individual‑participant meta‑analysis (Alemany et al., 2021; >70,000 children) reported increased odds of autism‑spectrum and ADHD symptoms after prenatal acetaminophen exposure. That is the opposite direction of the narrative NBC advances.

Postnatal/early-infancy exposure evidence (therapeutic-range modeling).
A complementary line of human-relevant evidence comes from controlled neonatal-period dosing paradigms at or below therapeutic levels. In male rat pups exposed on postnatal days 4–10, therapeutic-dose acetaminophen co-administered with cysteine and mannitol (to buffer NAPQI formation) still produced long-term behavioral changes, which were amplified by background stressors used to model vulnerability (diet, antibiotics) (Suda 2021; PLOS ONE 16:e0253543; PMID: 34170958; DOI: 10.1371/journal.pone.0253543). While not a human cohort, the paradigm’s dose discipline and attempt to suppress oxidative metabolism reinforce the inference that susceptibility factors—not overdose alone—govern risk expression in early life.

Meta-analytic corroboration (ADHD).
Beyond the six-cohort European analysis of ASD/ADHD symptoms, a prior meta-analysis focused on ADHD diagnoses reported elevated risk following prenatal exposure (Saad 2016; PLOS ONE 11:e0157380; PMID: 27355759; DOI: 10.1371/journal.pone.0157380). Heterogeneity and confounding were acknowledged, but the directionality aligns with biomarker-anchored cohorts and argues against the blanket “no association” messaging NBC adopted.

Developmental pharmacology: a built-in vulnerability window

Neonates and young infants rely far more on sulfation than glucuronidation to clear acetaminophen (Miller et al., 1976; Cook et al., 2016). As glucuronidation capacity matures with age, the balance of pathways shifts. During the developmental period, any constraints on sulfate supply or sulfotransferase activity amplify risk of shunting toward oxidative bioactivation (NAPQI) and away from safe conjugation. Independent human biochemistry in autism documents deficits in phenol sulfotransferase (PST) activity, lower plasma sulfate, and broader sulfation impairments, unrelated to simple SULT1A copy‑number counts (Alberti et al., 1999; Pagan et al., 2021). That phenotype dovetails with a susceptibility model: in a child with constrained sulfation reserve, the same acetaminophen dose can impose a disproportionate oxidative burden.

Window mechanics tied to pathway maturation.
Pharmacokinetic modeling in neonates quantifies the developmental handoff from sulfation dominance to increasing glucuronidation with age (Cook 2016; Clin Pharmacokinet 55:1395–1411; PMID: 26861602; DOI: 10.1007/s40262-015-0292-8). Against that backdrop, classic PK work confirms that early life depends disproportionately on sulfate conjugation (Miller 1976; Clin Pharmacol Ther 19:284–294; PMID: 1261167). The same one-carbon and sulfation constraints observed in autism biochemistry—reduced PST activity and lower plasma sulfate (Alberti 1999; Biol Psychiatry 46:420–424; PMID: 10435198; DOI: 10.1016/S0006-3223(98)00337-0; Pagan 2021; Transl Psychiatry 11:23; PMID: 33414449; DOI: 10.1038/s41398-020-01125-5)—map directly onto this window, supplying a mechanistic rationale for why otherwise “therapeutic” perinatal exposures do not translate uniformly across infants.

Animal and mechanistic evidence: sensitive windows, sex‑specific effects, and convergent pathways

Across species and labs, developmental acetaminophen exposure alters behavior and brain biology in a window‑dependent and often sex‑specific manner. Neonatal or perinatal exposures produce lasting changes in social interaction, exploration, anxiety‑like behaviors, and cognition, with corroborating shifts in BDNF, oxidative‑stress markers, and prefrontal cortical gene expression (Viberg et al., 2013; Philippot et al., 2017; Blecharz‑Klin et al., 2018; Rigobello et al., 2021; Baker et al., 2023). Interaction experiments show that background inflammation (IL‑1β) or hormonal milieu (ghrelin) modulates the behavioral outcome of the same acetaminophen regimen (Harshaw & Warner, 2022; Herrington et al., 2022). Developmental COX/PGE2 signaling shapes cerebellar maturation and social behavior (Dean et al., 2012), an axis plausibly intersecting with acetaminophen’s central actions. In vitro and acute in vivo models outline direct neurotoxicity at higher exposures and disruption of neuronal cytoskeleton and arborization (Posadas et al., 2010; Labba et al., 2022; Vigo et al., 2019), mapping plausible mechanisms even if doses exceed therapeutic ranges.

Sex specificity and transcriptomic convergence.
Two independent mouse studies sharpen the male-biased susceptibility motif that characterizes ASD epidemiology. Developmental exposure altered behavior when given at PND 3–10 but not at PND 19, delimiting a narrow sensitive window (Philippot 2017; J Appl Toxicol 37:1174–1181; PMID: 28448685; DOI: 10.1002/jat.3473). Separately, developmental dosing produced sex-specific behavioral changes alongside prefrontal cortex gene-expression shifts implicating oxidative-stress and neuronal-development pathways (Baker 2023; Neurobiol Dis 177:105970; PMCID: PMC9940030; DOI: 10.1016/j.nbd.2022.105970). These data triangulate with human biomarker studies by pointing to distributed network and redox mechanisms rather than a single lesion.

Sensorineural endpoints.
In utero or early-life exposure paradigms have also identified sensory-system liabilities. High-dose prenatal acetaminophen in rodents impaired auditory brainstem responses (Graeca & Kulesza 2024; Hear Res 454:109149; PMID: 37798658; DOI: 10.1016/j.heares.2024.109149), and neonatal exposure reduced striatal BDNF with lasting changes in social and exploratory behavior (Blecharz-Klin 2018; Pharmacol Biochem Behav 168:25–32; PMID: 29530595; DOI: 10.1016/j.pbb.2018.03.004). While dose translation to human pregnancy requires caution, convergent sensory-circuit impacts are notable given clinical ASD phenotypes.

Cell-level substrates.
Multiple groups report neuronal apoptosis at higher exposures (Posadas 2010; PLoS One 5:e15360; PMID: 21170329; DOI: 10.1371/journal.pone.0015360) and arborization/cytoskeletal disruption after paracetamol in human- and avian-neurite models (Labba 2022; Toxicol Appl Pharmacol 449:116130; PMID: 35714712; DOI: 10.1016/j.taap.2022.116130). These are not dose-equivalence studies; they are mechanistic maps showing what the molecule can do to developing neurons under stress.

Cofactor interactions—proof of a susceptibility architecture.
Early-life acetaminophen interacts with inflammation (IL-1β) to modify social-emotional and repetitive behaviors (Harshaw & Warner 2022; Pharmacol Biochem Behav 220:173463; PMID: 35278940; DOI: 10.1016/j.pbb.2022.173463) and with ghrelin state to alter outcomes (Herrington 2022; Dev Psychobiol 64:e22252; PMID: 35224734; DOI: 10.1002/dev.22252). The pattern is consistent with effect modification, not a one-size-fits-all toxin story—exactly what sibling designs with crude “ever-use” exposure tend to obscure.

Methods: why the sibling “null” cannot override the above

Sibling designs collapse when measurement error and non‑shared confounding differ between siblings (Frisell et al., 2012) and when carryover effects or secular trends break the exchangeability assumptions (Sjölander et al., 2016). In Ahlqvist et al., exposure is episodic, mostly OTC, and largely captured at a single early‑pregnancy timepoint; indications vary across pregnancies; diagnostic practices drift across birth years. Under those conditions, within‑family estimators attenuate true associations. The paper’s own “dose” analysis (prescriptions only, averaged across all pregnancy days) further compresses variance, starving the model of signal. None of that can be rehabilitated by small E‑values.

Why “null under sibling control” ≠ “safe.”
Textbook cautions apply: non-shared confounding and exposure misclassification bias sibling estimators toward the null (Frisell 2012; Epidemiology 23:713–720; PMID: 22781362; DOI: 10.1097/EDE.0b013e31825fa230), and carryover/calendar effects violate exchangeability as diagnostic intensity drifts (Sjölander 2016; Epidemiology 27:852–858; PMID: 27488059; DOI: 10.1097/EDE.0000000000000541). The JAMA study’s own exposure architecture—OTC-dominated, episodic, and front-loaded to a single antenatal interview—creates exactly the conditions where a sibling “null” is more plausibly a method artifact than a causal acquittal.

Consensus statements and precaution

The Nature Reviews Endocrinology consensus (Bauer et al., 2021) is neither alarmist nor dismissive; it recommends caution: medically indicated use only, lowest effective dose, shortest duration. Even critiques of that consensus stop short of saying “no risk.” NBC’s piece does not inform readers of that context.

CNS pharmacology in adults (face validity)

Acetaminophen is not behaviorally inert in the human brain. Randomized/lab studies show reduced social pain and changes in trust behaviors in adults following dosing (DeWall 2010; Psychological Science 21:931–937; PMID: 20548058; DOI: 10.1177/0956797610374741. Roberts 2019; Scientific Reports 9:4060; PMID: 30858394; DOI: 10.1038/s41598-019-40093-9). These short-term effects do not adjudicate developmental safety, but they demolish the folk claim that acetaminophen is a purely peripheral analgesic with no central consequences to worry about.

What responsible reporting would have told the public

The largest register study using a sibling comparison did not detect an association under its own constraints—constraints known to push results toward the null with episodic, under‑measured OTC exposure, pregnancy‑specific indications, and birth‑cohort diagnostic drift. In parallel, biomarker‑anchored human cohorts report graded associations with ADHD and autism, and a six‑cohort European meta‑analysis corroborates elevated risk for neurodevelopmental symptoms. Developmental pharmacology and cross‑species experiments identify plausible mechanisms and sensitive windows. The actionable guidance is straightforward and balanced: during pregnancy, use acetaminophen only when medically indicated, at the lowest dose for the shortest duration, and prioritize non‑pharmacologic alternatives when appropriate. Meanwhile, fund exposure‑accurate, pre‑registered studies designed to adjudicate causality rather than hide it behind misclassification and over‑adjustment.

---

Disclosures on methods and sourcing

Every citation below was verified on PubMed, CrossRef, or the publisher site. Landing pages and abstracts were inspected for topical alignment and methods fidelity. Identifiers (journal, year, DOI, PMID/PMCID) are provided. Where abstracts do not report effect sizes, funding, or dosing, that absence is stated. Design limitations visible from abstracts or article front matter are flagged without speculation.

References

Register study discussed by NBC Ahlqvist VH, Sjöqvist H, Dalman C, Karlsson H, Stephansson O, Johansson S, Magnusson C, Gardner RM, Lee BK. Acetaminophen Use During Pregnancy and Children’s Risk of Autism, ADHD, and Intellectual Disability. JAMA. 2024;331(14):1205–1214. doi:10.1001/jama.2024.3172. PMID: 38592388.

Biomarker‑anchored human cohorts Ji Y, Azuine RE, Zhang Y, et al. Association of Cord Plasma Biomarkers of In Utero Acetaminophen Exposure With Risk of ADHD and Autism Spectrum Disorder in Childhood. JAMA Psychiatry. 2020;77(2):180–189. doi:10.1001/jamapsychiatry.2019.3259. PMID: 31664451; PMCID: PMC6822099. (Reports tertile‑graded ORs; dosing captured via cord metabolites.)

Baker BH, Lugo‑Candelas C, Wu H, et al. Association of Prenatal Acetaminophen Exposure Measured in Meconium With Risk of Attention‑Deficit/Hyperactivity Disorder Mediated by Frontoparietal Network Brain Connectivity. JAMA Pediatr. 2020;174(11):1073–1081. doi:10.1001/jamapediatrics.2020.3080. PMID: 32986124; PMCID: PMC7522774. (ADHD odds; mediation via functional connectivity.)

Meta‑analysis Alemany S, Avella‑García CB, Liew Z, et al. Prenatal and postnatal exposure to acetaminophen in relation to autism spectrum and attention‑deficit and hyperactivity symptoms in childhood: Meta‑analysis in six European population‑based cohorts. Eur J Epidemiol. 2021;36:993–1004. doi:10.1007/s10654‑021‑00754‑4. PMID: 34046850. (Harmonized analysis across six cohorts; elevated odds for ASD/ADHD symptoms.)

Developmental pharmacology and human sulfation phenotype Miller RP, Roberts RJ, Fischer LJ. Acetaminophen elimination kinetics in neonates, children, and adults. Clin Pharmacol Ther. 1976;19(3):284–294. doi:10.1002/cpt1976193284. PMID: 1261167.

Cook SF, Stockmann C, Samiee‑Zafarghandy S, et al. Neonatal maturation of paracetamol glucuronidation, sulfation, and oxidation based on a parent–metabolite population PK model. Clin Pharmacokinet. 2016;55(11):1395–1411. doi:10.1007/s40262‑015‑0292‑8. PMID: 26861602. (Quantifies pathway maturation; sulfate dominance early.)

Alberti A, Pirrone P, Elia M, Waring RH, Romano C, Leone A. Sulphation deficit in “low‑functioning” autistic children: a pilot study. Biol Psychiatry. 1999;46(3):420–424. doi:10.1016/S0006‑3223(98)00337‑0. PMID: 10435198. (Reduced sulfation capacity; pilot size.)

Pagan C, Delorme R, Callebert J, et al. Decreased phenol sulfotransferase activities associated with hyperserotonemia in autism spectrum disorders. Transl Psychiatry. 2021;11:23. doi:10.1038/s41398‑020‑01125‑5. PMID: 33414449. (Lower PST‑M/P; SULT1A CNV not explanatory.)

Animal and mechanistic anchors Viberg H, Eriksson P, Gordh T, Fredriksson A. Paracetamol during neonatal brain development affects cognitive function and alters its analgesic and anxiolytic response in adult male mice. Toxicol Sci. 2013;138(1):139–147. doi:10.1093/toxsci/kft329. PMID: 24361869.

Philippot G, Gordh T, Fredriksson A, Viberg H. Adult neurobehavioral alterations in male and female mice following developmental exposure to paracetamol: characterization of a critical period. J Appl Toxicol. 2017;37(10):1174–1181. doi:10.1002/jat.3473. PMID: 28448685. (Effects at PND 3/10, not 19; window defined.)

Blecharz‑Klin K, Piechal A, Pyrzanowska J, et al. Early paracetamol exposure decreases BDNF in striatum and affects social behaviour and exploration in rats. Pharmacol Biochem Behav. 2018;168:25–32. doi:10.1016/j.pbb.2018.03.004. PMID: 29530595.

Rigobello C, Tiziani S, Berzero A, et al. Perinatal exposure to paracetamol: dose‑ and sex‑dependent effects in behaviour and brain oxidative stress markers in progeny. Behav Brain Res. 2021;408:113294. doi:10.1016/j.bbr.2021.113294. PMID: 33836167.

Baker BH, Bornehag CG, Gennings C, et al. Sex‑specific neurobehavioral and prefrontal cortex gene expression alterations following developmental acetaminophen exposure in mice. Neurobiol Dis. 2023;177:105970. doi:10.1016/j.nbd.2022.105970. PMCID: PMC9940030.

Harshaw C, Warner AG. Interleukin‑1β‑induced inflammation and acetaminophen during infancy: distinct and interactive effects on social‑emotional and repetitive behavior in C57BL/6J mice. Pharmacol Biochem Behav. 2022;220:173463. doi:10.1016/j.pbb.2022.173463. PMID: 35278940.

Herrington JA, Smith TL, Lyons LC, et al. Elevated ghrelin alters the behavioral effects of perinatal acetaminophen exposure in rats. Dev Psychobiol. 2022;64(3):e22252. doi:10.1002/dev.22252. PMID: 35224734.

Dean SL, Knutson JF, Krebs‑Kraft DL, McCarthy MM. Prostaglandin E2 is an endogenous modulator of cerebellar development and complex behavior during a sensitive postnatal period. Eur J Neurosci. 2012;35(8):1218–1229. doi:10.1111/j.1460‑9568.2012.08032.x. PMID: 22512254; PMCID: PMC3534986.

Posadas I, Santos P, Blanco A, et al. Acetaminophen induces apoptosis in rat cortical neurons. PLoS One. 2010;5(12):e15360. doi:10.1371/journal.pone.0015360. PMID: 21170329; PMCID: PMC3000821. (Overdose‑range relevance; mechanistic.)

Labba N‑A, Ahlner J, Gordh T, Viberg H. Paracetamol perturbs neuronal arborization and disrupts SPTBN1/TUBB3 in human and chicken in‑vitro models. Toxicol Appl Pharmacol. 2022;449:116130. doi:10.1016/j.taap.2022.116130. PMID: 35714712.

Vigo MB, Rondon‑Ortiz AN, Martinez M, et al. Acute acetaminophen intoxication induces direct neurotoxicity in rats manifested as astrogliosis and decreased dopaminergic markers. Biochem Pharmacol. 2019;170:113662. doi:10.1016/j.bcp.2019.113662. PMID: 31606411.

Design/analysis cautions for sibling comparisons Frisell T, Öberg S, Kuja‑Halkola R, Sjölander A. Sibling comparison designs: bias from non‑shared confounders and measurement error. Epidemiology. 2012;23(5):713–720. doi:10.1097/EDE.0b013e31825fa230. PMID: 22781362. (Attenuation under misclassification and non‑shared confounding.)

Sjölander A, Frisell T, Kuja‑Halkola R, Öberg S, Zetterqvist J. Carryover Effects in Sibling Comparison Designs. Epidemiology. 2016;27(6):852–858. doi:10.1097/EDE.0000000000000541. PMID: 27488059. (Calendar/carryover violations.)

Keyes KM, Davey Smith G, Susser E. On Sibling Designs. Epidemiology. 2013;24(3):473–474. doi:10.1097/EDE.0b013e31828c7381. PMID: 23549193; PMCID: PMC5937524.

Consensus/precaution Bauer AZ, Kriebel D, Herbert MR, et al. Paracetamol use during pregnancy—A call for precautionary action. Nat Rev Endocrinol. 2021;17(12):757–766. doi:10.1038/s41574‑021‑00553‑7. PMID: 34556849. (Recommends medically indicated use only; lowest dose; shortest duration.)

Context note on “safety was never established for neurodevelopment.”

A systematic review with citation tracking concluded that pediatric acetaminophen’s “safety” pedigree is built on hepatic endpoints; neurodevelopment was never formally established in approval-era studies (Cendejas-Hernández 2022; Eur J Pediatr 181:1835–1857; PMID: 35175416; DOI: 10.1007/s00431-022-04407-w). Recent narrative syntheses advocating precaution (Parker 2024; Children (Basel) 11:44; PMID: 38255358; PMCID: PMC10814214; DOI: 10.3390/children11010044) are not substitutes for causal identification, but they accurately depict that the evidence does not support a blanket safety claim—particularly in susceptible subpopulations.

Related:

“Acetaminophen doesn’t cause autism. That’s absurd. You have no proof.”