Quantitative bias analyses to address measurement error in time-to-event endpoints

Area of Science:

• Health Research Methods
• Biostatistics
• Real-World Evidence

Background:

• Single-arm trials using external real-world data (RWD) face outcome measurement error due to differing patient assessment schedules.
• This error can bias time-to-event endpoint comparisons and impact findings.
• Existing bias mitigation methods often require validation samples, which are not always feasible.

Purpose of the Study:

• To demonstrate novel statistical methods for quantitative bias analysis (QBA) to address outcome measurement error in RWD.
• To provide a framework for contextualizing RWD findings when validation samples are infeasible.
• To guide researchers in applying QBA when outcome measurement error is a concern.

Main Methods:

• Leveraging novel statistical methods as Quantitative Bias Analyses (QBA).
• Implementing QBA with Cumulative Incidence Curve Correction and Survival Regression Calibration.
• Generating plausible parameter values through simulation for QBA.

Main Results:

• QBA enables the estimation of plausible error ranges when direct measurement is not possible.
• Demonstrated practical application of QBA using a Newly Diagnosed Multiple Myeloma cohort.
• Provided guidance on conducting QBA for outcome measurement error and interpreting results.

Conclusions:

• QBA offers a robust approach to manage outcome measurement error in RWD for comparative effectiveness.
• These methods are valuable when validation samples are infeasible, enhancing the reliability of RWD.
• The study provides practical tools and guidance for applying QBA in real-world research settings.