Methods for time-varying exposure related problems in pharmacoepidemiology: An overview

Area of Science:

• Pharmacoepidemiology
• Observational Studies
• Biostatistics

Background:

• Time-varying exposures can introduce significant bias in treatment effect estimations.
• Accurate control for time-varying factors is essential in pharmacoepidemiology and other observational research.

Purpose of the Study:

• To provide an overview and guidance on methodologies for addressing time-varying exposures and confounding.
• To explore conceptual approaches for managing complex time-dependent variables in research.

Main Methods:

• Literature review focused on time-varying exposure concepts.
• Exploration of methods addressing four key pharmacoepidemiological challenges: treatment episodes, time-varying confounders, cumulative exposure/latency, and treatment switching.

Main Results:

• Correct construction of treatment episodes is vital for unbiased treatment effect estimates.
• Advanced methods like marginal structural models are underutilized due to complexity and lack of user-friendly software.
• Simpler methods, including those without adjustment or with time-varying covariates, are more common, but exposure magnitude requires careful modeling.

Conclusions:

• Further research is needed to implement and apply complex methodologies for time-varying exposures.
• Comparing results from multiple methods is recommended due to potential differences in treatment effect estimates.
• Accurate treatment episode estimation and exposure quantification are critical for reliable effect estimations.