OBJECTIVES: Meta-analysis plays an important role for evidence-based medicine to comprehensive synthesis of evidence from multiple clinical studies. In practices of meta-analysis, it is well known that some studies might have markedly different characteristics from the others, and they possibly yield biases and misleading results. The influences of the “outlying” studies should be adequately addressed in practices.

METHODS: Recently, we developed effective robust statistical inference methods based on machine learning theory. The robust inference methods are designed to adjust the influences of outliers even when there are multiple and serious influential outliers based on a robust criterion, density power divergence. In this study, we present a R package robustmeta that is an easy-to-handle computational package for the robust inference methods and its applications.

RESULTS: We will show the robust estimation can be implemented by a simple command. Also, we will provide several illustrative examples whether and how the new methods can treat the influences of outlying studies using real systematic review datasets published in recent medical journals. We can provide the summary estimate and its robust 95% confidence interval adjusting the outliers' effects. Also, we can assess the contribution rates of individual studies after the adjustment; usually, the contribution rates of potential outlying studies are adequately reduced by the robust methods.

CONCLUSIONS: We would present the main conclusions are possibly changed by applying the effective robust methods and these methods can provide new insights in practices of meta-analyses.