Are polypharmacy side effects predicted by public data still valid in real-world data?

Area of Science:

• Pharmacovigilance
• Computational drug discovery
• Clinical informatics

Background:

• Growing interest in predicting drug-drug interactions (DDIs).
• Need for real-world data validation of predicted polypharmacy side effects.
• Current predictions often lack empirical verification.

Purpose of the Study:

• To confirm if predicted polypharmacy side effects align with actual patient data.
• To validate a deep learning model for predicting drug interactions.
• To assess the real-world risk of lung edema from cefpodoxime-chlorpheniramine co-administration.

Main Methods:

• Utilized a deep learning-based polypharmacy side effect prediction model.
• Retrospective analysis of patients (≥18 years) from January 2000 to December 2020.
• Employed inverse probability of treatment weighting (IPTW) for group balancing and analyzed outcomes using Kaplan-Meier and Cox proportional hazards models.

Main Results:

• Identified cefpodoxime-chlorpheniramine-lung edema as a high-risk combination.
• Simultaneous use of cefpodoxime and chlorpheniramine significantly increased 1-year cumulative incidence of lung edema (p=0.001).
• Increased risk of lung edema observed compared to monotherapy (HR 2.10 for cefpodoxime, HR 1.64 for chlorpheniramine).

Conclusions:

• Real-world data validation of polypharmacy predictions aids clinical decision-making.
• Simultaneous cefpodoxime and chlorpheniramine use is associated with elevated long-term lung edema risk.
• Findings support enhanced monitoring for patients on this drug combination.