Article type
Abstract
Objectives: With the increasing demand for high-quality evidence and the lack of professional knowledge for stakeholders, living system reviews and other tools to accelerate the translation of evidence into practice urgently require rapid evidence synthesis and grading. This study constructed a semi-automatic and user-friendly platform AutoMeta (http://www.autometa.org.cn/) to facilitate evidence translation and provide guidance for users to develop systematic reviews, enabling meta-analysis and grading certainty of evidence.
Methods: Establish a platform design expert group to provide guidance on platform design and construction, use the Shiny package in R software to build a web-based data interaction program, and use metafor, netmeta and mada package to conduct pairwise meta-analysis, inverse variance meta-analysis, network meta-analysis, and diagnostic test accuracy meta-analysis. On the basis of conducting research on the GRADE guidance and application of GRADE method in Cochrane Systematic Reviews, this study quantified the 5 downgrading factors of the GRADE through expert consensus, and automatically graded the confidence of evidence of each outcome by linking the meta-analysis results.
Results: Nowadays, pairwise meta-analysis and inverse variance meta-analysis and its confidence of evidence grading are available in the platform. After uploading data in the platform and selecting meta-analysis-related parameters in the evidence synthesis module, results including meta-analysis, subgroup analysis, publication bias, meta-regression and sensitivity analysis, as well as graphics such as forest plots and funnel plots, can be generated. By comparing the results analyzed by R and RevMan software, there was no difference in effect size estimates and heterogeneity results. The evidence grading module can generate risk of bias assessment results, evidence profile, and summary of findings, which can be downloaded and saved in Word format. Compared with the existing study, the results of risk of bias may be inconsistent due to differences assessment criteria compared to manual assessment results.
Conclusions: After repeated testing, compared to traditional meta-analysis software, the AutoMeta platform can significantly improve the efficiency of meta-analysis and certainty of evidence grading while ensuring accuracy. Other types of meta-analysis will be continuously updated in future versions of the platform.
Methods: Establish a platform design expert group to provide guidance on platform design and construction, use the Shiny package in R software to build a web-based data interaction program, and use metafor, netmeta and mada package to conduct pairwise meta-analysis, inverse variance meta-analysis, network meta-analysis, and diagnostic test accuracy meta-analysis. On the basis of conducting research on the GRADE guidance and application of GRADE method in Cochrane Systematic Reviews, this study quantified the 5 downgrading factors of the GRADE through expert consensus, and automatically graded the confidence of evidence of each outcome by linking the meta-analysis results.
Results: Nowadays, pairwise meta-analysis and inverse variance meta-analysis and its confidence of evidence grading are available in the platform. After uploading data in the platform and selecting meta-analysis-related parameters in the evidence synthesis module, results including meta-analysis, subgroup analysis, publication bias, meta-regression and sensitivity analysis, as well as graphics such as forest plots and funnel plots, can be generated. By comparing the results analyzed by R and RevMan software, there was no difference in effect size estimates and heterogeneity results. The evidence grading module can generate risk of bias assessment results, evidence profile, and summary of findings, which can be downloaded and saved in Word format. Compared with the existing study, the results of risk of bias may be inconsistent due to differences assessment criteria compared to manual assessment results.
Conclusions: After repeated testing, compared to traditional meta-analysis software, the AutoMeta platform can significantly improve the efficiency of meta-analysis and certainty of evidence grading while ensuring accuracy. Other types of meta-analysis will be continuously updated in future versions of the platform.