MetaLLMReporter: An R Shiny App Integrating Meta-Analysis Execution with LLM-Assisted Reporting

Background Statistical tests are numerous in conducting a meta-analysis goes beyond the simple calculation of a group estimate. Combining these diverse results (heterogeneity, bias, sensitivity, subgroups, etc.) into a coherent whole suitable for various groups of people (e.g., academic papers, plain language summaries) is a major challenge. Methods We created MetaLLMReporter, an interactive web tool with R (v4.4.0) and Shiny framework (v1.8.1.1) with a bs4Dash (v2.3.2) interface. It accepts user-supplied CSV data for continuous measurements (mean, sd, n) carries out a series of standard meta-analysis procedures using functions from meta, metafor, and dmetar packages. Above all, it integrates Google’s Gemini large language model (LLM) using API calls (httr, jsonlite) to generate automatically systematic written text reports consolidating the analyses in various formats (Cochrane, NEJM, Lancet, Plain Language). Results/Functionality The MetaLLMReporter carries out a standard meta-analysis (meta::metacont) and performs additional analyses, including heterogeneity assessment, leave-one-out sensitivity analysis, publication bias tests (meta::metabias), meta-regression (metafor::rma), subgroup analysis (meta::metacont), cumulative meta-analysis (meta::metacum), Bayesian meta-analysis (bayesmeta::bayesmeta), trim-and-fill (meta::trimfill), outlier detection (dmetar::find.outliers), and p-curve analysis (dmetar::pcurve). Text summaries for each analysis is displayed. Users can then trigger the LLM to generate detailed reports formatted in specific journal formats or plain language. Conclusions MetaLLMReporter makes it easier to generate textual symbolises various aspects of meta-analysis and uses LLM technology to help write reports for various audiences. It is intended to assist researchers in interpreting and reporting complicated meta-analysis results more effectively.