Candidate Genes from an FDA-Approved Algorithm Fail to Predict Opioid Use Disorder Risk in Over 450,000 Veterans

Area of Science:

• Pharmacogenomics
• Computational Biology
• Clinical Genetics

Background:

• The Food and Drug Administration (FDA) approved an algorithm predicting opioid use disorder (OUD) risk based on genetic factors.
• Independent validation of the clinical utility of the candidate genes within this algorithm is crucial.

Approach:

• This case-control study analyzed 15 genetic variants in opioid-exposed individuals from the Million Veteran Program (MVP).
• Electronic health records (EHR) data spanning from 1992 to 2022 were used to assess the association between genetic variants and OUD diagnosis.
• Logistic regression and machine learning models evaluated the predictive performance of these variants.

Key Points:

• The 15 candidate genetic variants collectively explained only 0.4% of the variation in OUD risk.
• An ensemble machine learning model utilizing these genes achieved 52.8% accuracy in predicting OUD.
• The study included over 450,000 opioid-exposed participants, with 33,669 identified as OUD cases.

Conclusions:

• The candidate genes in the current FDA-approved algorithm lack sufficient efficacy for predicting OUD risk.
• The algorithm's limited predictive accuracy may result in significant false positive and negative outcomes in clinical settings.
• Development of more effective predictive models is necessary for identifying individuals at risk of OUD.