By age two, the average child in high-income countries has received roughly three courses of antibiotics — most often a beta-lactam such as penicillin, amoxicillin, or cephalosporin — for ear infections, respiratory illnesses, or post-surgical prophylaxis. The clinical benefits of these prescriptions are well established. What is less widely appreciated is that the same medication crosses a developmental window during which the gut microbiome is actively shaping the brain, immune system, and neuroendocrine axis.
The mouse work of Volkova and colleagues (2021) showed that low-dose penicillin given during this window produced lasting changes in microbial composition and altered transcription in the frontal cortex and amygdala — two regions central to executive control and emotional regulation. A decade of mechanistic and epidemiological work now situates that finding inside a coherent framework: a critical microbial-developmental window, biologically plausible signaling pathways, modest but reproducible cohort-level associations with neurodevelopmental disorders, and an active debate about how much of the association is causal.
The critical microbial-developmental window
The first three years of life are the formative period for both the human gut microbiome and the human brain. Microbial colonization begins at birth, accelerates with the introduction of solid foods, and reaches a relatively adult-like configuration by approximately 36 months. Over the same interval, the brain forms most of its synaptic connections — approximately 100 trillion synapses across roughly 86 billion neurons — and the immune system establishes durable patterns of tolerance and reactivity.
The mechanistic case for a “critical window” comes most directly from the Cox et al. (2014) study published in Cell, which used low-dose penicillin in mice and demonstrated that microbial perturbation limited to early life was sufficient to produce lasting metabolic effects. Mice that received penicillin only during weaning developed durable changes in body composition and hepatic gene expression, even after the microbiome had largely recovered to normal composition by adulthood. The mechanism was not “antibiotic in the system causes the phenotype.” It was “antibiotic-induced microbial change during a developmental window leaves a permanent signature.”
Volkova et al. (2021) extended that logic to brain gene expression. Mice exposed to low-dose penicillin from late pregnancy through weaning showed altered microbiome composition together with transcriptomic changes in the frontal cortex and amygdala. The differentially expressed genes mapped to pathways previously implicated in neurodevelopmental and neuropsychiatric disorders, and informatic analyses linked specific bacterial taxa — notably Lactobacillus, Allobaculum, and members of the Rikenellaceae family — to gene-expression patterns in those regions.
Signaling pathways from gut to brain
The gut–brain axis is not a single pathway but a parallel set of communication routes, and antibiotics can disrupt several at once.
Short-chain fatty acids. Acetate, propionate, and butyrate are produced when gut bacteria ferment dietary fiber. They modulate blood–brain barrier permeability, influence microglial maturation, and serve as signaling molecules to enteric nerves. Beta-lactam antibiotics reduce the abundance of fiber-fermenting taxa (notably Bacteroidetes and butyrate producers), which lowers SCFA output during the developmental window when those signals are most influential.
Vagal afferents. The vagus nerve carries information from the gut to the brainstem and onward to limbic and cortical regions. Bacterial metabolites — particularly butyrate and certain bile-acid derivatives — directly stimulate vagal afferent terminals. Animal work has shown that microbial perturbation alters vagally-mediated behaviors, and severing the vagus nerve attenuates many of the behavioral effects of microbial transfer.
Tryptophan metabolism and serotonin. Approximately 90% of the body’s serotonin is synthesized in the gut, and gut microbes regulate the availability of tryptophan and the activity of tryptophan hydroxylase 1 (TPH1). When microbial composition shifts, the balance between serotonin synthesis and the kynurenine pathway shifts with it, with downstream consequences for mood, sleep, and HPA-axis reactivity.
Immune signaling. Gut microbes prime the developing innate and adaptive immune systems. Disruption during early life produces a low-grade inflammatory state and altered cytokine profiles that the brain reads through circulating signals and through microglia at the blood–brain barrier. Several human and rodent studies have shown elevated frontal-cortex cytokine expression after early-life penicillin.
None of these pathways acts alone, and antibiotic-induced disruption typically perturbs all of them in parallel — which is why mechanistic dissection has been slow and why effect sizes in single-pathway models tend to be modest.
Evidence from human cohort studies
The mouse evidence is mechanistically clean but ecologically narrow. Human evidence comes from large epidemiological cohorts that ask whether antibiotic exposure in pregnancy or early childhood is associated with downstream neurodevelopmental outcomes.
The largest single estimate comes from a Swedish nationwide cohort of 483,459 first-born singletons by Njotto and colleagues (2023). After adjustment for indication, infection burden, parental psychiatric history, and socioeconomic confounders, the study reported:
- Maternal antibiotic exposure during pregnancy: odds ratio 1.16 for autism spectrum disorder (ASD) and 1.29 for attention-deficit/hyperactivity disorder (ADHD).
- Early-life antibiotic exposure: odds ratio 1.46 for ASD and 1.90 for ADHD.
The directional pattern — early-life exposure showing larger associations than prenatal exposure — is consistent across multiple cohorts and aligns with the critical-window biology.
A 2024 Taiwanese population-based study by Yang et al. analyzed roughly one million children and found a more modest signal: an adjusted hazard ratio of 1.06 for combined ASD/ADHD in the singleton cohort and 1.03 in a sibling cohort that controls for shared family environment. Crucially, when the authors restricted analysis to exposure-discordant sibling pairs — the strongest available control for unmeasured family-level confounding — the hazard ratio dropped to 0.92, with the confidence interval crossing one.
The contrast between cohort and sibling-discordant estimates is the central methodological lesson of the literature. Some of the apparent association between antibiotic exposure and later neurodevelopmental diagnosis reflects the underlying infections (or the genetic and environmental correlates of frequent infections) rather than the antibiotic itself. The honest summary is: there is a small, plausible, residual association after accounting for the obvious confounders, but the magnitude is well below the level that would justify clinical alarm.
Confounding by indication and the causality question
Antibiotics are not prescribed at random. Children who receive more antibiotics tend to have more infections, which tend to track with parental atopy, lower socioeconomic status, daycare attendance, prematurity, and a constellation of other factors that themselves correlate with neurodevelopmental risk. Disentangling antibiotic-as-cause from antibiotic-as-marker is the central methodological challenge.
Three design strategies have been deployed:
- Within-sibling comparisons. Siblings share roughly half their genome and most of their early environment. When one sibling is exposed and another is not, the comparison removes much of the shared family-level confounding. The 2024 Taiwan analysis used this design and saw the association attenuate substantially (Yang et al., 2024).
- Negative-control exposures. Some studies compare antibiotics with non-systemic medications used for similar indications (for example, topical preparations) to isolate the systemic-microbiome effect from the indication. These comparisons typically show smaller associations for non-systemic exposures.
- Mendelian randomization. Genetic variants that predispose to microbial composition or to inflammation can serve as instrumental variables. The few applied analyses of this kind suggest a small but non-zero microbiome-mediated component.
The synthesis from these triangulating designs is that early-life antibiotic exposure plausibly carries a small causal contribution to neurodevelopmental risk, but most of the raw cohort association reflects confounding by indication and shared family environment.
Implications for prescribing
The clinical implication is not “avoid antibiotics in young children.” Untreated bacterial infections in infancy produce serious immediate harm and their own downstream developmental consequences, including hearing loss from untreated otitis, scarring from untreated streptococcal disease, and severe sequelae from sepsis. The relevant question is “use antibiotics when they are clinically indicated, and avoid prescribing when they are not.”
That distinction is more than rhetorical. Approximately 30% of antibiotic prescriptions written for children in primary-care settings in high-income countries are for indications where antibiotics provide little or no benefit — most commonly viral upper respiratory infections, viral bronchiolitis, and self-limited diarrheal illness. Reducing these unnecessary prescriptions is a low-cost intervention that addresses both antimicrobial resistance and the developmental concerns reviewed here.
For prescribed courses, narrow-spectrum agents disturb the microbiome less than broad-spectrum alternatives, and the duration of treatment matters: shorter courses (when clinically appropriate) produce smaller and more transient microbial perturbations than longer courses.
Probiotics and microbial restoration
The natural follow-up question is whether the microbiome can be “restored” after disruption. The honest answer is partial. Specific probiotic strains — particularly Lacticaseibacillus rhamnosus, Lactobacillus acidophilus, and Lactiplantibacillus plantarum — produce short-chain fatty acids and modulate serotonin metabolism in animal models, and a small number of pediatric trials have shown modest benefits on infant cognition and behavior after antibiotic courses. However, the field has not converged on a defined regimen with reproducible effects, and probiotic supplementation does not reconstitute the full ecological diversity that antibiotics displace.
Better-supported microbial-restoration strategies are environmental: breastfeeding (where feasible), exposure to diverse environments, and a fiber-rich diet that supports the recovery of butyrate-producing taxa. These interventions act on the developmental window itself rather than attempting to replace specific lost organisms.
Conclusion
Early-life antibiotic exposure perturbs the gut microbiome during a critical developmental window, alters gene expression in brain regions central to neurodevelopment, and shows a small but reproducible association with later neurodevelopmental diagnoses across multiple large cohorts. The animal mechanisms — short-chain fatty acid signaling, vagal communication, tryptophan metabolism, and immune priming — are well characterized. The human evidence is consistent with a modest causal contribution after accounting for confounding by indication, but the residual effect is not large enough to override clear clinical indications. The actionable conclusion for clinicians and parents is the same one that follows from antimicrobial stewardship more broadly: prescribe antibiotics when they are clinically indicated, avoid them when they are not, and prefer narrow-spectrum, short-course regimens when alternatives exist.
References
- Volkova, A., Ruggles, K., Schulfer, A., Gao, Z., Ginsberg, S. D., & Blaser, M. J. (2021). Effects of early-life penicillin exposure on the gut microbiome and frontal cortex and amygdala gene expression. iScience, 24(7), 102797. https://doi.org/10.1016/j.isci.2021.102797
- Cox, L. M., Yamanishi, S., Sohn, J., Alekseyenko, A. V., Leung, J. M., Cho, I., Kim, S. G., Li, H., Gao, Z., Mahana, D., Zárate Rodriguez, J. G., Rogers, A. B., Robine, N., Loke, P., & Blaser, M. J. (2014). Altering the intestinal microbiota during a critical developmental window has lasting metabolic consequences. Cell, 158(4), 705–721. https://doi.org/10.1016/j.cell.2014.05.052
- Njotto, L. L., Simin, J., Fornes, R., Odsbu, I., Mussche, I., Callens, S., Engstrand, L., Bruyndonckx, R., & Brusselaers, N. (2023). Maternal and early-life exposure to antibiotics and the risk of autism and attention-deficit hyperactivity disorder in childhood: a Swedish population-based cohort study. Drug Safety, 46(5), 467–478. https://doi.org/10.1007/s40264-023-01297-1
- Yang, K.-L., Yen, T.-A., Lin, F.-J., Hsu, C.-N., & Wang, C.-C. (2024). Antibiotic use and risk of autism spectrum disorder and attention-deficit/hyperactivity disorder: a population-based cohort study. Child and Adolescent Psychiatry and Mental Health, 18(1), 82. https://doi.org/10.1186/s13034-024-00774-4
- Borrego-Ruiz, A., & Borrego, J. J. (2024). Neurodevelopmental disorders associated with gut microbiome dysbiosis in children. Children, 11(7), 796. https://doi.org/10.3390/children11070796