Simple imputation method for meta-analysis of survival rates when precision information is missing

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Medical Informatics

Background:

• Meta-analyses of survival rates often lack precision information (standard errors or confidence intervals) in clinical studies.
• Exclusion of studies with missing precision leads to biased pooled estimators and reduced precision.
• Existing methods for handling missing precision data are insufficient.

Purpose of the Study:

• To develop a simple and effective method for imputing missing precision information in survival meta-analyses.
• To improve the accuracy and precision of pooled estimators by utilizing commonly available study statistics.
• To provide a practical solution for incorporating incomplete data into synthesis analyses.

Main Methods:

• Developed a novel imputation method using individual study statistics: sample size, number of events, and risk set size.
• Validated the method through extensive simulation studies comparing it to naive deletion and ideal scenarios.
• Applied the method to a systematic review of radiotherapy data to demonstrate its robustness.

Main Results:

• The proposed imputation method significantly improved the accuracy and precision of pooled estimators compared to naive deletion.
• The method's performance was comparable to analyses with complete precision information.
• No underestimation bias of standard errors was observed, though overestimation risk exists if risk set size is unavailable.

Conclusions:

• The developed imputation method effectively addresses missing precision data in survival meta-analyses, mitigating bias and enhancing precision.
• This approach offers a valuable tool for researchers to include more studies in meta-analyses, leading to more reliable results.
• An R package is available to facilitate the implementation of this procedure in practice.