Dear Editor Danrolf de Jesus and Dr Jesus M Garcia Calleja Re: MS#9952457645244487 We are most grateful for the helpful comments of the referees. The following are our response to each of their comments. We have revised our manuscript accordingly. We are attaching two copies of the revised manuscript for your consideration, in one we highlight the changes we have made. Yours truly, Yingjie Zheng Response to reviewers’ comments on MS#9952457645244487 1. English language writing We have our manuscript reviewed by Professor Monhun Ng from University of Hongkong. 2. The title is weak. I would resentence it, and would not use past tense in a title. The language has been corrected and the title has been changed to “Should chronic hepatitis B mothers breastfeed? A meta analysis”. 3. Most included studies (94%) are Chinese studies, almost all from Chinese journals, which is difficult for researchers or readers outside of China to get a hold of these articles. Looking from another perspective, this study provides us with information that would be difficult to reach and there would be a language barrier. Our meta analysis included 32 studies, and 30 of them are from China: (1) 2 are indexed in PUBMED: 1 study is from Hongkong [study 2: Lancet 1988, 2(8618):1032], and the other study from Shanghai [study 8: Int J Clin Pract 2003, 57(2):100-102]. (2) Carefully evaluating the quality of studies based on Newcastle-Ottawa Quality Assessment Scale (NOS), we can confirm another Chinese study with high quality [study 15]. Similar rating has been done for the rest Chinese studies. And all the studies meet our criterion of inclusion and exclusion. (3) It’s not surprising that Chinese scientists are much interested in breastfed transmitted hepatitis B infection since the disease is highly epidemic in this country. Thus, we believe that this meta analysis should lost much information if the Chinese studies have been excluded. Actually for any meta analysis, the completeness of the included studies should be ensured. Language barrier should be a common problem. To be a scientist, I think it’s my responsibility to introduce high-quality Chinese studies into the world and an international cooperation should effectively overcome this problem. 4. The authors present their results clearly, with figures and tables, but are lacking with making clearer sentences for the reader to understand the methodology. We correct it accordingly. 5. Abstract: Don’t use the past tense. We have corrected the abstract, and re-written the results and conclusion. 6. Introduction: Second paragraph, first sentence. Here you are making a statement I would advice to state a reference. We have re-written the paragraph and added one reference. 7. Introduction: Fourth paragraph: The last sentence in this paragraph, what do you mean with uncomfortable? We have re-written the whole paragraph. 8. I would already start defining high infectivity in this section. High HBV DNA levels? HBeAg-positiveness? We have re-written the whole paragraph. 9. Is my understanding correct in this? That your aim was to understand whether immunization helps to prevent CHB infection in infants born to CHB infected women, especially those with high infectivity. Yes. 10. Methods: First paragraph, data sources… Add the starting date of the search from year…to December 31, 2010. We have added this information. 11. Methods: Paragraph 4, The last sentence, general characteristics of the participants… are the participants the mothers? Or the infants? Its not clear in the methods. We change “participants” to “babies”. 12. Methods: 5th paragraph, Please explain fourth and 2nd outcome of NOS, its not clear to the reader. In NOS, the 4th item under selection is “Demonstration that outcome of interest was not present at start of study” and the 2nd item under outcome is “Was follow-up long enough for outcomes to occur”. We add this information in our manuscript. 13. Methods: Statistical analysis section, Because of the language some things are left unclear. What is Begger’s test? I tried to find reference 21, but havent’t come across, I did find Biometrics 2000, Jun 56 (2) 455-63. We add a reference to Begger’s test. Reference 21 is for the method of nonparametric trim and fill. 14. Results: What is meant by backward check? Do you mean manual reference checking? Backward search means that the reference lists of potentially relevant manuscripts were scanned backward to obtain extra suitable studies. We add this information in the Methods section. 15. In the 3rd paragraph you mention that 1 study did not mention if the feeding methods were decided by the mother. Is this important for your results? Analysis? If yes, explain the reason. If not, leave it out. Only one study randomized the feeding methods. I think it’s important and add this information. 16. 6th paragraph, I believe you mean Subgroup analysis showed. Yes. 17. Discussion We have re-written it. 18. This is how I understand the research question, but it is formulated differently in the abstract, introduction and conclusion. We have re-written it. 19. My main methodological question is why 'risk difference' was used instead of 'risk ratio'? The a priori chance of infection is low to start with, so one can never show a large difference between the groups. Using risk ratio would better highlight the differences between the two groups compared. We use RD instead of RR due to: (1) we are much concerned about breastfed transmitted HBV and want to quantify the risk in the breast-fed infants that is attributable to breastfeeding if it really exists. (2) The calculation of RDs is easy to implement in common software for the studies with zero outcome in both groups. During the analysis, those studies are often automatically excluded by the software if RR or OR is used as measurement of effect. (3) We also calculate RR, and the results are similar. We give one result below (see Figure 1). The RRs and its 95% confidential limits in those studies (De Martino M 1985; Gu XQ 2003; Zhang JH 2005; Chen H 2008; SUN LY 2010) with zero outcomes in both groups were firstly calculated with Monte-Carlo simulation and then integrated into the other 27 studies for further analysis. (4) The linear relationship between RDs and RRs was showed in Figure 2. RR Figure 1 Effect of breastfeeding on occurrence of chronic hepatitis B infection among the infants born of CHB mothers analyzed by random-effect model (RR: risk difference; CI: confidence interval) 3.0 RRs 2.5 2.0 1.5 1.0 0.5 0.0 -0.08 -0.06 -0.04 -0.02 0.00 RDs 0.02 0.04 0.06 0.08 Figure 2 the relationship between RRs and RDs of the 32 studies. 20. The funnel plots are nice to see, but do not add much information. This information can be given in words. We agree to the reviewer’s comments. But the funnel plots would help the readers to visually judge how the publication bias happens. 21. Other major modifications We re-write the sections of Result, Discussion and Conclusion; We reformat and replace Table 1 with a new one; We replace Figure 3 with a new one. Only “RD” in the vertical axis has been changed to “RDs”. Other revisions can be found in our manuscript with the tracked changes.