Extracting standard error from the 95% confidence interval of a survival

Area of Science:

• Biostatistics
• Survival Analysis
• Meta-Analysis Methodology

Background:

• Accurate extraction of standard errors for survival data from 95% confidence intervals (95%CI) is essential for meta-analyses.
• Current methods often require assumptions about the transformation used to calculate 95%CIs, lacking assessment of incorrect assumption impacts.
• A need exists for a reliable method to identify the correct transformation and improve standard error extraction.

Purpose of the Study:

• To propose and assess the Logarithm of the Relative Asymmetry (LRA) approach for identifying the transformation used in survival 95%CIs.
• To evaluate the extraction error and the performance of the LRA method.
• To determine the impact of extraction errors and the LRA approach on meta-analysis outcomes.

Main Methods:

• The study introduces the Logarithm of the Relative Asymmetry (LRA) approach, which utilizes the asymmetry of reported 95%CIs to infer the transformation.
• Extraction error (difference between extracted and true standard errors) was assessed.
• The performance of the LRA approach and its impact on meta-analyses were evaluated.

Main Results:

• Incorrect assumptions about the transformation typically lead to overestimation of the standard error, particularly when survival approaches 0 or 1 or with smaller sample sizes.
• The LRA approach successfully identifies the transformation used for 95%CI calculation.
• LRA performance can be affected by the precision of reported survival data and CIs, or when CI bounds are exactly 0 or 1.

Conclusions:

• The LRA approach effectively identifies the transformation for survival 95%CIs, avoiding significant extraction errors.
• This method helps mitigate the propagation of errors into meta-analysis results.
• The LRA approach is recommended for improving the accuracy of standard error extraction in survival meta-analyses.