The Development And Validation Of Reporting Checklist For Network Meta Analysis

Background and objective:Network Meta analysis aimed to compare and rank all available interventions according to their effectiveness and safety, and choose the best intervention based on ranking results. Studies showed that the reporting quality of network Meta analysis was low, which affected the use of evidence. Available reporting checklists for network Meta analysis were not comprehensive; different reporting checklists have different requirements for network Meta analysis and could not be used to guide the reporting of network Meta analysis. So our study tried to develop the reporting checklist for network Meta analysis and validate using scientific methods.Methods:We included all available English network Meta analysis using systematic and comprehensive search to evaluate the application status of network Meta analysis based on the methods of social network analysis and bibliometrics methods; we compared the reporting checklists for network Meta analysis to analyze their differences and their completeness; we included those studies or reporting checklists which involved the reporting of network Meta analysis, abstracted all reporting items and integrated those items with PRISMA statements.25% network Meta analysis was randomly selected and assessed using the integrated reporting items. Afterwards, Exploratory Factor Analysis was used to develop the reporting checklist for network Meta analysis. All available network Meta analysis in surgery field was included and assessed using the reporting checklist in order to assess its reliability. Data analysis was conducted by SPSS 21.0; the reliability was tested by Cronbach a, Spearman-Brown coefficient, and Guttman coefficient. If Cronbach a is more than 0.8, it means that this checklist achieves high agreement among items, and has good reliability; Spearman-Brown coefficient or Guttman coefficient is more than 0.7, it also means this checklist has good reliability.Results:We retrieved 522 network Meta analysis. Our study showed the number of network Meta analysis increased by year and most of them focused on the drug treatment for chronic diseases. Six reporting checklists for network Meta analysis were identified. All of them were not comprehensive, and have different requirements for reporting. Most of them required to report the method and result sections, but did not require to report the background and discussion section.47 reporting items from 15 studies were collected and were integrated with reporting items from PRISM A checklist to form the embryo of reporting checklist for network Meta analysis which consisted of 55 reporting items. We used the integrated reporting items to evaluate the reporting of 25% network Meta analysis (n=113), and found the reporting of network Meta analysis was not adequate and comprehensive, especially in reporting of methods:45 (38.92%) reported the process of selecting studies,68 (60.18%) reported the process of data abstracting,61 (53.98%) reported the process of assessing the quality of included studies,10 (8.85%) reported the process of assessing clinical and methodological heterogeneity,54 (47.8%) reported the process of assessing statistical heterogeneity,7 (6.19%) reported the process of assessing the similarity,18 (15.93%) reported the process of assessing inconsistency. Based on the results of Exploratory Factor Analysis, we identified a 34-item reporting checklist for network Meta analysis. The reliability of this reporting checklist was validated by 73 network Meta analysis (involving six kinds of interventions and 11 kinds of diseases) in surgery field and the results showed that Cronbach a was 0.84, which means this reporting checklist could represent 84% of the real reporting quality of network Meta analysis. Spearman-Brown coefficient was 0.82, Guttman coefficient was 0.82.Conclusion:The reporting of network Meta analysis was poor, especially the key methodological parts. The reporting checklist based on systematic review and Exploratory Factor Analysis consisted of 34 reporting items, and has high reliability, it could be used to guide the reporting of network Meta analysis. However, studies were needed to validate the reliability and validity of reporting checklist using network Meta analysis in other field.