Causal diagrams for disease latency bias

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Disease latency, the period from disease onset to diagnosis, presents challenges in epidemiological research.
• Disease latency bias (DLB) can occur when disease inception precedes exposure assessment, potentially skewing study results.
• The specific mechanisms and structures of DLB in chronic disease studies remain underexplored.

Purpose of the Study:

• To elucidate the various ways DLB can introduce bias into epidemiological studies of latent outcomes.
• To provide a framework for understanding the structural nature of DLB.
• To discuss strategies for identifying and mitigating DLB in research.

Main Methods:

• Utilized directed acyclic graphs (DAGs) to illustrate four distinct scenarios of DLB.
• Employed causal diagrams to map the pathways through which DLB influences epidemiological findings.
• Reviewed potential methods for addressing DLB in study design and analysis.

Main Results:

• Demonstrated that DLB can manifest through unmeasured confounding.
• Identified reverse causality as a potential consequence of DLB.
• Showcased selection bias and bias through a mediator as additional mechanisms of DLB.
• Illustrated four specific bias structures arising from DLB using DAGs.

Conclusions:

• Disease latency bias is a significant factor affecting epidemiological studies with latent outcomes.
• Causal diagrams are valuable tools for researchers to identify and control DLB.
• Addressing DLB is essential for improving the validity of epidemiological research on chronic diseases.