GOSH - a graphical display of study heterogeneity

Area of Science:

• Biostatistics
• Medical Informatics
• Epidemiology

Background:

• Statistical heterogeneity is common in meta-analyses, where individual study estimates disagree.
• Understanding the sources of heterogeneity is vital for advancing scientific knowledge and forming new hypotheses.

Purpose of the Study:

• To introduce a new graphical method for visualizing between-study heterogeneity in meta-analysis.
• To facilitate the exploration of heterogeneity, identification of influential studies, and subgroup analysis.

Main Methods:

• The proposed method employs combinatorial meta-analysis, performing meta-analyses on all possible subsets of studies.
• Summary effect sizes and statistics from these all-subsets analyses are used to generate informative graphs.

Main Results:

• The generated graphs provide a visual tool to investigate heterogeneity.
• The approach aids in identifying influential studies and exploring potential subgroup effects within a meta-analysis.
• Biomedical examples demonstrate the practical application and interpretation of the graphical display.

Conclusions:

• The combinatorial meta-analysis graphical approach offers a complementary method for exploring data in systematic reviews.
• This technique can enhance the understanding of between-study heterogeneity and support exploratory data analysis.