Does Advanced Paternal Age Cause Worse Neurodevelopmental Outcomes in Offspring?

1. Precise question and why it matters

Does a father's age at conception causally increase the risk of neurodevelopmental disorders — autism spectrum disorder (ASD), schizophrenia, ADHD — or reduce cognitive ability in his children? Or does the observed association mostly reflect confounding by genetics, socioeconomic status, family environment, and selection effects?

The question matters for two reasons. First, paternal age at first birth has risen steadily across high-income countries. If the association is causal, delayed fatherhood carries a real biological cost that public health messaging should address. If it is largely confounded, the same messaging would be misleading. Second, the answer informs how to interpret a family's risk: a 35-year-old father producing a child with autism has a different causal story than a 25-year-old father producing the same outcome.

2. Best current answer

Paternal age is a real but modest risk factor for ASD and schizophrenia. The evidence does not support a large causal effect on general cognitive ability or educational attainment. The dose-response shape differs across outcomes: a roughly monotonic increase for ASD and schizophrenia, a U-shape for ADHD with the lowest risk at 31–35 years, and a near-null or slightly negative effect on test scores once genetic confounding is controlled.

Absolute risk context: for a father conceiving at 40 versus 25, the absolute increase in ASD risk is roughly 1–2 percentage points above a baseline of approximately 2–3%. The increase in schizophrenia risk is roughly 0.3–0.5 percentage points above a baseline of approximately 1%. These are small relative to the overall variance in outcomes.

3. What the strongest studies show

3.1 Meta-analysis of 41 studies: parental age and autism

ObservationDehesh and colleagues (2024) pooled 41 observational studies. Advanced paternal age was associated with a 51% increased odds of ASD (adjusted OR 1.51, 95% CI 1.40–1.62). Advanced maternal age showed a similar magnitude (adjusted OR 1.47, 95% CI 1.33–1.62). Low parental age was not significantly associated with lower risk.

3.2 De novo mutation mediation: 28% of schizophrenia

Causal estimateFeng et al. (2025) performed whole-genome sequencing on multiplex schizophrenia families. Paternal age predicted +1.50 DNMs per year (95% CI: 0.81–2.19). Each additional DNM was associated with a 0.16-year earlier age of schizophrenia onset. The proportion of the paternal-age–onset association mediated through DNMs was 28% (95% CI: 18–38%). This is the first direct evidence that paternal-age-related mutations causally contribute to psychiatric risk, though it leaves approximately 70% unexplained by DNMs.

3.3 Mendelian randomization: no positive cognitive effect

Causal estimateGrätz et al. (2025) used Mendelian randomization with polygenic indices in the Norwegian MoBa cohort (n ≈ 15,600). The OLS association was positive: each year of delayed maternal age predicted +3.7% SD in test scores; paternal age +2.3% SD. Once the model controlled for the child's own PGI for age at first birth and parental PGI for educational attainment, the IV estimates turned significantly negative (maternal: −8.8% SD; paternal: −19.3% SD). The authors' conservative conclusion: "Our results do not provide evidence in favour of sociological theories that predict positive causal effects of parental age on children's educational attainment."

3.4 The DNM model underestimates the epidemiological association

ObservationTaylor et al. (2019) estimated the risk attributable to paternal-age-related de novo SNVs using trio exome-sequencing and Danish registries. For a father conceiving at 45 versus 25, dnSNV-attributable IRRs were: ASD 1.09 (~9%), schizophrenia 1.09 (~9%), intellectual disability 1.20 (~20%). The epidemiologic HRs for ASD and schizophrenia significantly exceeded what dnSNVs could explain. For ASD, the population-level association was approximately 9.3 times larger than the dnSNV-attributable risk.

3.5 ADHD: a U-shaped relationship

ObservationMin et al. (2021) meta-analyzed 11 studies (4.4 million participants). The lowest parental age (<20) carried the highest ADHD risk (adjusted OR 1.75 for fathers, 1.49 for mothers). The lowest risk was at paternal age 31–35. Paternal age over 45 showed a slight non-significant increase. This U-shape means very young fatherhood is a bigger ADHD risk factor than older fatherhood.

3.6 Mouse model: neuroinflammation and epigenetic inheritance

ObservationMao et al. (2024) found that advanced paternal age in mice produced offspring with cognitive impairment and autism-like behavior, mediated by neuroinflammation and microglial overactivation. MeRIP-seq revealed hypermethylation of the nuclear receptor Nr4a2 and elevated Ythdc1 in aging sperm and offspring hippocampus. Microglia ablation rescued behavioral deficits. Extrapolation: whether this rodent mechanism translates to humans is speculative.

4. Biological mechanisms

4.1 De novo mutation accumulation

The male germline accumulates point mutations at approximately 1.5 per year of paternal age, compared to approximately 0.4 per year for maternal age. This difference reflects continuous mitotic division of spermatogonial stem cells from puberty onward, versus the largely post-mitotic female germline. Each newborn inherits approximately 50–100 de novo point mutations, with roughly 80% of paternally-transmitted variants.

4.2 Selfish spermatogonial selection

Goriely and Wilkie (2012) proposed that mutations activating the RAS signaling pathway confer a selective growth advantage on spermatogonial stem cells, leading to clonal expansion within the testis. This mechanism preferentially enriches mutations in genes involved in cell-cycle control and early development — the same pathways disrupted in ASD and schizophrenia. A 2026 preprint combining whole-genome sequencing of 168 trios with duplex sequencing of paternal sperm confirmed that transmissible de novo disease risk is governed primarily by universal germline processes, while early developmental mosaicism produces uncommon but clinically meaningful outliers.

4.3 Epigenetic and neuroinflammatory pathways

Mao et al. (2024) found that advanced paternal age in mice produced offspring with neuroinflammation mediated by microglial overactivation, Nr4a2 hypermethylation, and elevated Ythdc1. Microglia ablation rescued behavioral deficits. Whether this rodent mechanism translates to humans is speculative.

4.4 Brain structure

Xia et al. (2021) followed 51 children from kindergarten through third grade and found that greater paternal age was associated with worse reading ability, mediated by volumetric maturation of the left posterior thalamus. This is preliminary (n = 43 in the behavioral analysis) and requires replication.

5. Why confounding is the central problem

5.1 The positive education association is confounded by parental genotype

The Grätz et al. (2025) Mendelian randomization study is the key evidence. Observational studies consistently find that children of older parents have higher test scores. But once the model controlled for the child's own polygenic index for age at first birth and parental polygenic index for educational attainment, the positive association disappeared and possibly reversed. The implication: much of what looks like a "paternal age effect" is actually the genetic footprint of parents who delay reproduction — more educated, higher-IQ parents tend to have children later and also transmit genes associated with higher cognitive ability.

5.2 Sibling designs partially support this

A 2024 Taiwan birth cohort (2.45 million children) used sibling-comparison analyses. In population analyses, older paternal age was associated with adverse perinatal outcomes. But within families, younger siblings with older paternal age had lower risks of under-five mortality, low birth weight, and congenital defects — suggesting that family-level resources and environment overwhelm the biological risk within sibling pairs.

5.3 The inherited-genetic-confounding account

Men who have children later may carry different genetic profiles than men who have children earlier. Janecka et al. (2017) estimated that approximately 50% of the APA effect may be explained by inherited factors, 10–20% by de novo mutations, and the remainder by non-genetic factors. The Norwegian MR study directly confirmed this for cognitive outcomes.

6. Counterevidence and unresolved disputes

The strongest challenge to the "paternal age causes harm" narrative comes from the Mendelian randomization evidence. Grätz et al. found no positive causal effect on children's test scores and a possible negative effect after controlling for parental educational attainment PGI. This does not disprove biological effects on ASD or schizophrenia risk — those are different outcomes with different genetic architectures — but it undermines the claim that older fatherhood carries a broad cognitive cost.

The strongest support for a real biological effect comes from the combination of consistent mutation-rate estimates across populations, direct mediation evidence (28% of schizophrenia onset–age association through DNMs), mouse models with plausible neurobiological mechanisms, and registry studies showing monotonic dose-response for ASD and schizophrenia even after adjustment for measured confounders.

The unresolved dispute is the proportion of the epidemiological association attributable to biology versus confounding. The DNM-attributable fraction is small (~9–20% depending on disorder). The remainder is a mixture of inherited genetics, selfish selection, epigenetic effects, and environmental confounding. Current evidence cannot cleanly separate these.

7. Decision implication

Fathers in their early-to-mid thirties sit in or near the lowest-risk window for ADHD (31–35) and carry only modestly elevated risk for ASD and schizophrenia relative to the population baseline.

The practical takeaway: for fathers in that range there is no evidence-based reason for parental anxiety about neurodevelopmental outcomes attributable to paternal age at conception. The absolute risk increases are small. The positive educational associations seen in observational studies are largely confounded by parental genotype, not caused by age. And the biological mechanisms, while real, produce small effect sizes against low base rates.

For public communication, the key message is: paternal age is a real but weak biological signal, overwhelmed by environmental, genetic, and socioeconomic factors that are more actionable. The popular narrative that "older fathers cause autism" overstates the evidence; the equally popular narrative that "older fathers are just wiser and richer, so it all works out" ignores the genuine but small biological risk.

8. Calibrated confidence

9. What would reverse this view

I would raise confidence in a causal paternal-age effect on cognition if a large sibling-comparison study with genetic data showed a consistent negative gradient. I would lower confidence in the biological risk if a Mendelian randomization study specifically targeting ASD or schizophrenia found no effect after controlling for pleiotropic PGI. I would revise upward if a human neuroimaging study replicated the thalamic finding in a large cohort with measured de novo mutation burden.

References

1. Dehesh, T., Mosleh-Shirazi, M. A., Jafari, S., Abolhadi, E., & Dehesh, P. (2024). A assessment of the effects of parental age on the development of autism in children: a systematic review and a meta-analysis. Behavioral Sciences, 14(11), 1008. doi:10.3390/bs14111008.
2. Feng, Y. C. A., Chen, W. J., Lin, M. C., et al. (2025). Paternal age, de novo mutation, and age at onset among co-affected schizophrenia sib-pairs: whole-genome sequencing in multiplex families. Molecular Psychiatry, 30, 3560–3567. doi:10.1038/s41380-025-02942-0.
3. Goriely, A., & Wilkie, A. O. M. (2012). Selfish spermatogonial selection: a novel mechanism for the association between advanced paternal age and neurodevelopmental disorders. American Journal of Psychiatry, 170(6), 599–608. doi:10.1176/appi.ajp.2012.12020269.
4. Grätz, M., Tropf, F. C., Torvik, F. A., Andreassen, O. A., & Lyngstad, T. H. (2025). No evidence of positive causal effects of maternal and paternal age at first birth on children's test scores at age 10 years. Nature Human Behaviour, 9, 731–736. doi:10.1038/s41562-025-02108-6.
5. Janecka, M., Mill, J., Basson, M. A., Goriely, A., Spiers, H., Reichenberg, A., ... & Choudhury, S. (2017). Advanced paternal age effects in neurodevelopmental disorders — review of potential underlying mechanisms. Translational Psychiatry, 7, e1019. doi:10.1038/tp.2016.294.
6. Kong, A., Thorleifsson, G., Frigge, M. L., et al. (2012). Rate of de novo mutations and the importance of father's age to disease risk. Nature, 488, 471–475. doi:10.1038/nature11396.
7. Mao, Y., Meng, Y., Zou, K., et al. (2024). Advanced paternal age exacerbates neuroinflammation in offspring via m6A modification-mediated intergenerational inheritance. Journal of Neuroinflammation, 21, 248. doi:10.1186/s12974-024-03248-8.
8. Min, X., Li, C., & Yan, Y. (2021). Parental age and the risk of ADHD in offspring: a systematic review and meta-analysis. International Journal of Environmental Research and Public Health, 18(9), 4939. doi:10.3390/ijerph18094939.
9. Taylor, J. L., Debost, J. C. P. G., Morton, S. U., et al. (2019). Paternal-age-related de novo mutations and risk for five disorders. Nature Communications, 10, 3043. doi:10.1038/s41467-019-11039-6.
10. Wan, W., Zhu, Y., Tian, J., et al. (2024). Associations of parental age at pregnancy with adolescent cognitive development and emotional and behavioural problems: a birth cohort in rural Western China. BMC Public Health, 24, 775. doi:10.1186/s12889-024-18309-z.
11. Xia, Z., Wang, C., Hancock, R., Vandermosten, M., & Hoeft, F. (2021). Development of thalamus mediates paternal age effect on offspring reading: a preliminary investigation. Human Brain Mapping, 42(16), 5196–5210. doi:10.1002/hbm.25567.
12. Taiwan birth cohort study on paternal age and offspring outcomes. (2024). BMJ Public Health, 2(2), e001113. doi:10.1136/bmjph-2024-001113.