Clinical Performance of a Gene-Based Machine Learning Classifier in Assessing Risk of Developing OUD in Subjects...

Area of Science:

• Pharmacogenomics
• Machine Learning in Medicine
• Addiction Science

Background:

• Guidelines recommend assessing Opioid Use Disorder (OUD) risk before prescribing oral opioids.
• Current subjective assessments lack clinical validation for objective risk stratification.
• Genetics-based objective risk assessment could enhance shared decision-making for opioid prescriptions.

Purpose of the Study:

• To evaluate the performance of a machine learning (ML) classifier for predicting Opioid Use Disorder (OUD) risk.
• To determine the clinical utility of an objective, genetics-based OUD risk assessment tool.

Main Methods:

• An observational cohort study of 385 adults prescribed oral opioids for 4-30 days was conducted.
• A machine learning (ML) classifier was developed and validated using genotyping data from SNP microarrays.
• The analysis utilized a representative sample reflecting U.S. demographics for accurate performance estimation.

Main Results:

• The ML classifier achieved 82.5% sensitivity and 79.9% specificity in predicting OUD risk.
• Performance metrics showed no significant differences across demographic factors including gender, age, race, or ethnicity.
• The study confirmed the classifier's robust performance in a diverse population.

Conclusions:

• A machine learning (ML) classifier based on genetic data can offer objective insights into a patient's risk of developing Opioid Use Disorder (OUD).
• This objective information can empower patients and healthcare providers in making more informed decisions regarding oral opioid prescriptions.
• The findings support the potential integration of genetic-based risk assessment into clinical practice for opioid prescribing.