Examining Reproducibility of Literature Search in Meta-Analysis

Meta-analysis, a systematic retrieval from literature databases is an essential and prevailing method for combining data from multiple studies. Unfortunately, few studies have examined its rigor, which affects its reproducibility of results. We identified 22 meta-analyses on cervical cancer in PubMed for examining the parameters defined by PRISMA, relating to the rigor of literature retrieval. We found that 16 literature databases were used, and EMBASE was a leading resource, accounting for the highest frequency (81.82%). About half (45.45%) of the meta-analyses presented a complete, reproducible search strategy for at least one database. The ratio of included to retrieved articles after redundancy removal was only 6.58%, indicating low precision due to unclear or unreported processes. Our work serves as an initial step to examine the planning and execution of meta-analysis. Future efforts need to enhance reliability on literature retrieval in meta-analysis and compliance to PRISMA.