SGEM#485: I Want a New Drug – One Not Associated with Neurodevelopmental Disorders

Date: Sept 16, 2025 Reference: Prada et al. Evaluation of the evidence on acetaminophen use and neurodevelopmental disorders using the Navigation Guide methodology. Environ Health. August 2025 Guest Skeptic: Dr. Andrew Martin is an emergency physician practicing in Jacksonville, Florida.  Case: A 27-year-old at 24 weeks’ gestation presents to the emergency department (ED) with fever (38.6 °C), myalgias, and sore throat. She took 650 mg of acetaminophen (Tylenol) six hours ago with partial relief. She hesitated to repeat the dose after reading online posts about “Tylenol and autism.” She has no abdominal pain, no vaginal bleeding, and normal fetal movement. Vitals otherwise stable; pharynx erythematous, no exudate. She asks, “Is it safe to take another dose, or could this hurt my baby’s brain later?” Background: Acetaminophen (paracetamol) is the most used analgesic–antipyretic in pregnancy. A recent prospective cohort study suggests ~40 to 65% of pregnant people report using it. They are typically using acetaminophen for headache, myalgias, or fever, with most use being short and intermittent. Alternatives, particularly non-steroidal anti-inflammatory drugs (NSAIDs), carry well-described fetal risks in late gestation. This is one of the reasons why acetaminophen remains the default first-line choice [1,2]. Biologically, acetaminophen crosses the placenta and achieves fetal levels like maternal levels, making the developing brain theoretically exposed during critical windows [3]. This has motivated a large observational literature examining whether prenatal exposure is linked to later neurodevelopmental outcomes such as ADHD and autism. Meta-analyses generally report small associations (summary effects around 1.2 to 1.3) and signal stronger effects with longer duration of use, though heterogeneity in exposure measurement and outcome ascertainment is substantial.  Professional bodies, including the American College of Obstetricians and Gynecologists (ACOG) and the Society of Obstetricians and Gynecologists of Canada [SOGC], continue to recommend acetaminophen for appropriate indications at the lowest effective dose and shortest duration. At the same time, they do acknowledge ongoing research and the limitations of observational data (including confounding by indication). For emergency clinicians, the practical tension is familiar. The dilemma is that untreated maternal fever and significant pain can themselves harm pregnancy, yet patients are increasingly asking about possible long-term neurodevelopmental potential harms of using acetaminophen.  Clinical Question: Is acetaminophen exposure during pregnancy associated with ADHD, ASD, or other neurodevelopmental disorders (NDDs) in children? Reference: Prada et al. Evaluation of the evidence on acetaminophen use and neurodevelopmental disorders using the Navigation Guide methodology. Environ Health. August 2025 Population: Observational studies assessing children of pregnant individuals for neurodevelopmental outcomes. Excluded: Postnatal exposures, non-human studies for the primary analysis, non-original publications, and duplicate reports from the same cohort. Exposure: Prenatal acetaminophen (maternal self-report, biomarkers such as meconium/cord blood, or medical records/prescription registries). Comparison: Children who were not exposed prenatally to acetaminophen, or those exposed to alternative analgesics. Outcome: Primary Outcome: NDDs (particularly ADHD and ASD) and related symptomatology measured by clinical diagnoses, medication use, or validated behavioural scales. Secondary Outcomes: Timing and dose–response patterns, broader cognitive/behavioural domains (language & executive function), and triangulation across design types. Type of Study: Systematic review using the Navigation Guide methodology with a qualitative synthesis (no meta-analysis) due to substantial heterogeneity. Authors’ Conclusions: “Our analyses using the Navigation Guide thus support evidence consistent with an association between acetaminophen exposure during pregnancy and increased incidence of NDDs. Appropriate and immediate steps should be taken to advise pregnant women to limit acetaminophen consumption to protect their offspring’s neurodevelopment.” Quality Checklist for Systematic Review: The main question being addressed should be clearly stated. Yes The search for studies was detailed and exhaustive. Unsure Were the criteria used to select articles for inclusion appropriate? Yes Were the included studies sufficiently valid for the type of question asked? Unsure Were the results similar from study to study? No Were there any financial conflicts of Interest? Yes Result: They searched PubMed (primary) through Feb 25, 2025, with confirmatory checks in Web of Science and Google Scholar. A total of 46 studies were included.  The studies consisted of a mix of prospective cohorts, retrospective/historical cohorts, sibling-controlled cohorts, and two case-control studies.  Sample sizes ranged from ~100 to a national registry scale (Sweden’s registry cohort included 2.5 million births). Key Result: The preponderance of studies reported positive associations between prenatal acetaminophen exposure and ADHD/ASD or related symptoms, with dose–response suggested in several biomarker and prospective cohorts. However, some sibling-controlled analyses attenuated associations toward null, and overall heterogeneity was high. Of 46 studies, 27 reported positive associations, 9 null, and 4 inverse (protective) associations. Several higher-quality or biomarker-based studies suggested dose-response relationships. No pooling into a meta-analysis due to heterogeneity 1) Exposure Misclassification: Most contributing cohorts ascertained prenatal acetaminophen exposure via maternal self-report during pregnancy or post‑partum recall, or from prescription/registry data that do not fully capture over-the-counter use. The authors’ own risk‑of‑bias summaries flag the exposure domain as a frequent concern, and they rate retrospective self-report, especially when collected after child diagnosis, as high risk for recall bias. In contrast, biomarker studies (meconium, cord blood, maternal plasma/urine) mitigate recall bias and, in several instances, suggest dose–response, but they remain snapshots that imperfectly reflect timing and cumulative dose. The review also highlights how low-sensitivity exposure measurement can deflate associations. An example is the large Swedish sibling‑analysis cohort that reported 7.5% use based on midwife interviews, despite contemporaneous sources indicating ~50 to 60% use. This implies substantial underascertainment (misclassification bias or ascertainment bias) that likely biases the estimate toward the null and compounds loss of power in within-family models. Together, these exposure‑measurement limitations should make us more cautious in both the direction and magnitude of observed associations in the qualitative synthesis. 2) Residual Confounding: Although many included studies adjusted for a broad set of factors (maternal age, socioeconomic status, smoking, alcohol, illness, fever, and infection) and some used negative‑control exposure periods or propensity approaches, the review acknowledges that unmeasured and residual confounding remain possible. Importantly, the authors did not implement a quantitative bias analysis (E‑values, probabilistic bias analysis) to bound the strength of confounding necessary to explain the findings. They explicitly list this as a limitation of the review. Given that indications for acetaminophen (fever, pain, intercurrent infection) may themselves relate to neurodevelopmental outcomes and can be difficult to measure with sufficient granularity, any qualitative conclusion about “persistence after adjustment” should be interpreted skeptically. 3) Outcome Heterogeneity & Variable Ascertainment:  The review pools evidence across disparate outcome definitions: registry-based clinical diagnoses (ICD‑coded ADHD/ASD), ADHD medication use, and multiple validated behavioural scales (CBCL, SDQ), often at different child ages and with parent vs teacher report. This heterogeneity in outcome measurement (plus differences in timing of assessment) creates non-comparability that complicates causal interpretation and precludes valid pooling. Appropriately, the authors do not conduct a meta-analysis, citing the substantial heterogeneity in exposure assessment, outcome measures, and confounder adjustment. Risk‑of‑bias summaries also show concerns in the outcome domain for some outcome types. Mixing clinical diagnoses, proxy markers (medication), and symptom scales across varied ages increases measurement error and between-study variability, limiting the precision and generalizability of the qualitative synthesis. 4) Search Scope & Evidence‑Grading: The search strategy relied on PubMed as the primary database with confirmatory checks in Web of Science and Google Scholar. No additional eligible studies were identified beyond PubMed, and no discussion of searching the grey literature.  Heavy reliance on a single primary database risks missing studies indexed elsewhere or gray literature, potentially introducing retrieval bias.[4] In addition, while the Navigation Guide offers a structured approach, the authors note that its numeric domain‑averaging can imply unwarranted precision and assign equal weight to domains even when certain biases (confounding, exposure error) likely dominate. They attempted sensitivity analyses (excluding lowest‑scoring papers and up-weighting confounding) but acknowledge the framework’s default “moderate” rating for observational evidence could skew certainty assessments. These methodological choices reasonably reflect transparency but also represent limitations of the review’s grading and synthesis.