Associations between persistent organic pollutants and endometriosis: A multipollutant assessment using machine...

Area of Science:

• Environmental Epidemiology
• Toxicology
• Gynaecology

Background:

• Endometriosis is a common gynaecological disease affecting women of reproductive age.
• Epidemiological studies suggest a link between endometriosis risk and exposure to organochlorine persistent organic pollutants (POPs).
• Simultaneous exposure to complex mixtures of POPs presents analytical challenges for traditional statistical models.

Purpose of the Study:

• To compare the performance of various machine learning techniques in identifying POPs associated with deep endometriosis.
• To explore associations between mixtures of POPs and deep endometriosis using advanced computational methods.
• To assess the utility of machine learning in handling highly correlated exposure biomarkers in epidemiological studies.

Main Methods:

• A case-control study dataset from France was analyzed.
• Five machine learning models were employed: regularised logistic regression, artificial neural network, support vector machine, adaptive boosting, and partial least-squares discriminant analysis.
• Models were used to identify POP biomarkers in adipose tissue associated with endometriosis and compare classification performance.

Main Results:

• Consistent identification of key POPs associated with deep endometriosis across all five models, including specific compounds like octachlorodibenzofuran and cis-heptachlor epoxide.
• High classification performance demonstrated by all machine learning models.
• Regularised logistic regression offered a balance between statistical interpretability and machine learning predictive power.

Conclusions:

• Machine learning models are effective tools for analyzing complex mixtures of environmental exposures and their association with endometriosis.
• These advanced techniques can overcome limitations of traditional statistical methods in epidemiological research.
• A battery of complementary algorithms is recommended for deciphering intricate exposome-health relationships.