Quantitative bias analysis for time to event data needs good validation data

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Quantitative bias analysis (QBA) methods primarily address binary outcomes, leaving time-to-event outcomes underdeveloped.
• Misclassification in time-to-event analyses is complex, involving errors in both events and person-time, necessitating specific bias parameters.

Purpose of the Study:

• To highlight the underdeveloped nature of QBA for time-to-event outcomes.
• To discuss current limitations and propose future directions for improving QBA in this area.

Main Methods:

• Review of existing QBA methodologies and their limitations for time-to-event data.
• Discussion of recent work utilizing expert-informed ranges and simulation for bias parameter estimation.
• Emphasis on the need for validation studies and methodological innovation.

Main Results:

• Current QBA methods for time-to-event outcomes are limited by a lack of validation data, especially for person-time measurement error.
• Existing approaches, while better than qualitative assessments, require further development for broader application.

Conclusions:

• Improving QBA for time-to-event outcomes requires focused efforts on designing and publishing validation studies.
• Methodological innovation, accessible software, and extensions of regression calibration and risk-based adjustment are crucial.
• Expanding QBA methods will enhance the validity, transparency, and interpretability of epidemiologic research.