The Correlation between Rates of Cancer and Autism: An Exploratory Ecological Investigation

Hung‐Teh Kao, Stephen L. Buka, Karl T. Kelsey, David F. Gruber, Barbara PortonPLoS 2010 · DOI

Abstract

This exploratory ecological investigation examines the potential association between autism prevalence and cancer incidence across the United States. By analyzing state-level data, the study identifies a highly significant correlation between autism and specific female-specific cancers, particularly in situ breast cancer and uterine cancer. These findings suggest that shared genomic instability or hormonal signaling pathways, such as estrogen exposure, may serve as common risk factors for both conditions.

Introduction

The etiology of autism spectrum disorder (ASD) remains a complex puzzle in modern medicine. While significant genomic aberrations—including chromosomal rearrangements and de novo copy-number variations (CNVs)—are known to play a role, the specific environmental or genetic triggers remain elusive. Interestingly, these patterns of genomic instability are also characteristic hallmarks of cancer. This study addresses whether autism and cancer share population-level risk factors, using ecological data to identify potential biological links.

Methodology

The researchers utilized Spearman rank correlations to compare state-wide autism prevalence with cancer incidence rates. To ensure statistical rigor and mitigate the risk of Type I errors (false positives) inherent in large datasets, several advanced techniques were employed:

  • Multiple Comparison Adjustments: The Bonferroni correction was applied to maintain conservative significance thresholds.
  • P-value Integration: Three distinct methods—Fisher’s, Brown’s (adjusted for dependent data), and Simes’—were used to evaluate the overall significance of observed correlations.
  • Diagnostic Stratification: States were categorized into four groups based on their autism diagnostic criteria (e.g., strict federal guidelines vs. expanded DSM-IV criteria) to account for variability in reporting.

Data Sources

  1. Autism Prevalence: Data for children aged 3–21 (2000–2007) was sourced from the Individuals with Disabilities Education Act (IDEA) database.
  2. Cancer Incidence: Age-adjusted rates for 24 female and 22 male cancers (1999–2005) were obtained from the Centers for Disease Control and Prevention (CDC).
  3. Demographics: U.S. Census Bureau data provided the population denominators for prevalence calculations.

Results and Key Insights

The investigation revealed a striking gender-specific pattern in the correlation between autism and cancer.

The Estrogen Hypothesis and Gender Asymmetry

The most robust statistical links were found exclusively with female-specific cancers. Specifically, in situ breast cancer and uterine cancer showed highly significant correlations with autism rates. In states utilizing the strictest diagnostic criteria, the correlation with in situ breast cancer reached a significance level of $p < 10^{-10}$ (Brown’s method).

Conversely, male cancers showed virtually no significant correlation (combined $p = 1.0$). This asymmetry suggests that the shared risk factors are likely not universal genomic stressors but may instead involve gender-specific biological mechanisms, such as cumulative estrogen exposure or specific germline mutations.

Impact of Diagnostic Rigor

The strongest signals emerged when the analysis was restricted to states with the most rigorous diagnostic criteria. This suggests that diagnostic "noise" in broader datasets can obscure underlying biological signals, emphasizing the importance of data quality in ecological studies.

Discussion and Limitations

While these findings point toward shared genomic roots, the authors emphasize the ecological nature of the study. A correlation at the state level does not imply individual-level causation (the "ecological fallacy").

Limitations include: * Diagnostic Substitution: Historical shifts in labeling (e.g., from "learning disabilities" to "autism") may affect early data points. * Prevalence Underestimation: The IDEA database tracks children in special education, potentially missing high-functioning individuals.

Conclusion

This study reinforces the hypothesis that autism may be part of a broader spectrum of conditions driven by genomic fragility. The specific association with estrogen-sensitive female cancers warrants further molecular investigation into the roles of germline mutations and hormonal signaling in neurodevelopment. Identifying these shared factors could eventually lead to improved screening and preventive strategies for both autism and cancer.

Summary updated

Abstract

These findings suggest that there may be an association between autism and specific forms of cancer.
BibTeX
@inproceedings{kao2010correlation,
  author = {Hung‐Teh Kao and Stephen L. Buka and Karl T. Kelsey and David F. Gruber and Barbara Porton},
  title = {{The Correlation between Rates of Cancer and Autism: An Exploratory Ecological Investigation}},
  booktitle = {PLoS},
  year = {2010},
  doi = {10.1371/journal.pone.0009372},
}
Figure