Refining adverse drug reaction signals by incorporating interaction variables identified using emergent pattern...

Area of Science:

• Pharmacovigilance and Pharmacoepidemiology
• Computational Health Sciences
• Biostatistics

Background:

• Longitudinal observational data is crucial for identifying adverse drug reactions (ADRs).
• Confounding interaction terms can obscure true ADR signals in healthcare data.
• Accurate signal refinement is essential for drug safety monitoring.

Purpose of the Study:

• To develop a framework for identifying and incorporating candidate confounding interaction terms.
• To refine adverse drug reaction signals using regularized Cox regression analysis.
• To improve the analysis of longitudinal observational data for ADR detection.

Main Methods:

• Utilized emergent pattern mining to identify potential confounding interaction terms.
• Applied a cohort study design with regularized Cox regression (elastic net).
• Incorporated identified confounding terms into the regression model to account for their effects.

Main Results:

• The methodology successfully accounted for confounding, leading to accurate ADR signal identification.
• Regularized Cox regression with confounding terms correctly ranked known ADR drug families higher.
• Without accounting for confounding, non-ADR-related factors were incorrectly ranked highest.

Conclusions:

• The proposed framework efficiently identifies high-order confounding interactions without expert input.
• This method can be applied to any outcome of interest for rapid signal refinement.
• The approach shows significant potential to reduce false positive rates in longitudinal data analysis for ADRs.