A multiple imputation method for sensitivity analyses of time-to-event data with possibly informative censoring

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Time-to-event data analysis is crucial in many fields.
• Informative censoring can bias survival analysis results.
• Sensitivity analyses are needed to assess the impact of potential biases.

Purpose of the Study:

• To develop a multiple imputation method for sensitivity analyses of time-to-event data.
• To specifically address the challenge of possibly informative censoring.
• To provide a flexible framework for incorporating assumptions about post-discontinuation event tendencies.

Main Methods:

• A novel multiple imputation technique is proposed.
• Imputed times for censored values are drawn from a conditional failure time distribution.
• A hazard ratio parameter allows for modeling the tendency of events after follow-up discontinuation.

Main Results:

• The method allows for incorporating various assumptions about censoring mechanisms.
• Multiple-imputed datasets are analyzed using standard methods.
• Results are combined using Rubin's rules for valid inference.

Conclusions:

• The proposed multiple imputation method offers a robust approach for sensitivity analyses in time-to-event data.
• It provides a valuable tool for researchers dealing with potentially informative censoring.
• The method enhances the credibility of survival analysis findings by accounting for censoring uncertainty.