Multiple imputation methods for inference on cumulative incidence with missing cause of failure

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Estimating cumulative incidence, or absolute risk, is crucial in survival analysis.
• Missing cause-of-failure data complicates accurate cumulative incidence estimation.
• Existing methods may be unreliable when data is incomplete.

Purpose of the Study:

• To develop and validate statistical methods for estimating cumulative incidence with missing failure data.
• To provide a robust approach using multiple imputation and asymptotic theory.
• To assess the performance of these methods in simulated and real-world data.

Main Methods:

• Development of asymptotic theory for multiple imputation in the context of competing risks.
• Modeling cause-specific hazards using separate proportional hazards models.
• Application of the methods to survival data with missing cause-of-failure information.

Main Results:

• The proposed multiple imputation methods provide accurate estimates of cumulative incidence under missing at random assumptions.
• Simulation studies demonstrate that the procedures perform well in cohorts of 200 and 400 subjects.
• The methods effectively handle missing data, yielding reliable estimates for cumulative incidence and prediction error.

Conclusions:

• Multiple imputation based on asymptotic theory offers a reliable approach for estimating cumulative incidence with missing failure data.
• The developed methods are applicable to various settings, including cancer survival analysis.
• This work provides a valuable tool for researchers dealing with incomplete survival data.