Using Machine Learning to Identify Adverse Drug Effects Posing Increased Risk to Women

Area of Science:

• Pharmacovigilance
• Machine Learning
• Drug Safety

Background:

• Adverse drug reactions (ADRs) are a significant cause of mortality.
• Women experience twice the risk of ADRs compared to men, yet these differences are poorly understood.
• Real-world clinical data offer insights but are prone to confounding biases.

Purpose of the Study:

• To develop and validate a machine learning algorithm (AwareDX) to predict sex-specific drug risks.
• To quantify differential risks of adverse events between men and women.
• To create a resource identifying drugs with sex-specific risks.

Main Methods:

• Utilized a pharmacovigilance algorithm (AwareDX) incorporating machine learning.
• Mitigated confounding biases and correlated covariates in real-world clinical data.
• Validated the algorithm against clinical literature and pharmacogenetic mechanisms.

Main Results:

• Identified 20,817 adverse drug effects with sex-specific risks.
• Demonstrated high precision in predicting differential drug risks between sexes.
• Quantified the varying likelihood of adverse events based on sex.

Conclusions:

• AwareDX effectively predicts sex-specific drug risks, addressing a critical gap in pharmacovigilance.
• The developed resource aids in understanding and mitigating ADRs in understudied populations.
• Personalizing drug prescription and dosage based on sex can minimize adverse events.