LAUSR

Dear editor, We thank Gu et al. [1] for their comments and interest in our recent publication [2] and for providing us with the opportunity to extend the discussion regarding our findings. In this reply, we would like to answer to the points raised in their letter. Regarding the first point, we agree that pre-publishing the protocol is an essential step in publishing a paper, though we did not clarify the registration information in the article, we developed a complete protocol in place before the study began and strictly implemented the plan. We have submitted the protocol at the end of the supplementary materials. We also appreciate the comments regarding the search strategy. We considered more databases when developing our search strategy, but ultimately we decided to run our search in the most commonly used medical databases related to our research topic. We performed a similar search in Scopus (n = 2220 articles), and there were no new studies which meet the inclusion criteria. For other databases mentioned by the author, the PsycINFO database is primarily focused on the field of psychology, NLM Gateway covers many resources that are not relevant to our research e.g. DIRLINE, AIDS Meetings and Health Services Research mtgs. Based on the topic of our paper, we don’t think it’s necessary to retrieve these database. The optimal retrieval strategy does not mean to include more databases and obtain more retrieval results, which will reduce the specificity of retrieval. In addition, we refer to the published systematic reviews [3, 4] for supplementary retrieval to avoid the omission of qualified literature. Therefore, we believe that the current retrieval strategy can ensure the comprehensiveness and accuracy of results. We will continue to update our systematic reviews as more high-quality evidence becomes available. Finally, we are interested in the inverse variance heterogeneity(IVhet) model [5] mentioned by Gu et al., although this model is not yet widely used, its advantages over traditional random-effects models are worth exploring. We learned and applied the IVhet model in our research, there was no significant difference from our results (e.g. the pooled OR was 2.35 [95% CI, 1.55–3.56] vs. 2.45 [95% CI, 1.76–3.41]). To sum up, addressing the points raised did not significantly change our results indicating that our conclusions are correct and reasonable. In the end, we would like to thank Gu et al. for their comments that provide valuable advice and new thinking.