Referee 1's review:

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.
**********

Overall this is an important manuscript that adds to the body of knowledge on the recruitments and retention of participants in HIV prevention research.

1. Title ~ given the nature of the study population, it would be helpful to add "female sex workers" and "urban South Africa" to the title.

2. Abstract ~ the conclusions/significance section simply summarizes the results, and says that there are important lessons to be learned. Instead of this, it would be preferable to state what the implications of the findings are for HIV prevention research, and what the exactly lessons are, in brief.

3. Introduction ~ first para ~ "research of the HIV epidemic" is awkward ~ "research into"?

4. Introduction ~ second para ~ the phrase "growing understanding that the acute and early phrases of HIV infection...." Needs a citation

5. Introduction ~ third para, final sentence ~ the phrase "within a very challenging environment" is vague, could be more specific about what makes this environment challenging?

6. Introduction ~ general comment ~ the experiences reported here focus on the recruitment and retention of high-risk women, predominantly female sex workers, from urban areas, and this needs to be reflected in the introduction (ie, mentioned in the fourth paragraph)

7. Introduction ~ general comment ~ the introduction needs to recognize that there has been a large amount of work conducted already on experiences / lessons learned in HIV prevention, and summarise the key findings of this (which will allow the authors to highlight what new insights their data provide in the results/discussion)

8. Methods ~ first para under cohort development ~ "women within the already high risk category of urban South Africa" ~ what does this mean?

9. Methods ~ what was the distance, on average, from where participants lived to the study clinic(s)? Was transport readily accessible (and if so, at what cost), or provided for participants? This information is helpful to understand the retention issues.

10. Methods ~ participant protection ~ what steps were taken to ensure that participants who seroconverted during the study had access to HIV clinical care specifically, including ART?

11. Methods ~ given the focus of the study on seroconversion, more detail is required on the behavioral prevention efforts that were included at enrollment and follow-up. What form did the counseling take (eg, theory based?)? Administered by whom? What duration, and frequency over time?

12. Tables 2-3 are very short and could easily be merged into Table 1, making the results section much easier to follow (they do not need to be separate tables).

13. The content of Table 4, and the corresponding text, does not seem relevant to the topic of the paper and is distracting (unless the authors can explain why BP, BMI, HB and other general clinical descriptors are meaningful for HIV prevention research, and participant retention specifically?), and can be deleted.

14. Causes of death ~ 'natural causes' should be grouped with 'unknown'; 'thrombocytopenia' is not a cause of death per se, but a clinical/laboratory finding (was this autoimmune? Drug-induced? idiopathic, etc)

15. The incidences of drop-out (loss to follow-up), HIV, STI and pregnancy are all interesting/important, and I was surprised to see the analysis stop at only a basic description (an in fact, no incidence of STI are reported). With the space saved by merging tables 2 and 3 into table 1, and deleting table 4 (and the corresponding text), it would be important to present an analysis of the risk factors for at least (a) LTFU and (b) pregnancy (two major sources of censoring in HIV prevention research), both in crude and adjusted (multivariate) analyses. This would strengthen the manuscript substantially (as few 'lessons learned' papers present data on predictors of LTFU and pregnancy, the latter being an issue recognized only relatively recently) and is directly relevant to the focus of the paper. (The authors may also wish to include risk factors for HIV and other STI infection, but this may also fall beyond the scope of this manuscript).

16. Given the pregnancy rate observed, what were the contraceptive use patterns in the cohort? Patterns of barrier method use during follow-up?

17. Were there any patterns in the results according to whether women were female sex workers or not? With about 25% of participants not sex workers, this is worth examining and reporting, if only briefly.

18. Results, final para ~ the mention of "enrichment of the negative cohort with women less susceptible to HIV infection due to hose genetic resistance factors" is curious, and not substantiated by anything else in the manuscript. This either needs to be explained, carefully (was there recruitment from different ethnic groups, or any other aspect that would suggest patterns in host genetics, that is not mentioned in the methods?), or deleted. Also, the grammar of the last sentence needs attention.

19. Same para ~ the last two sentence of this paragraph are highly interpretive, and belong in the results (and again, the "host genetic factors" discussion needs considerable explanation / support).

20. Figure 3 does not convey the pattern in HIV seroconversion rate with time particularly well. It should be deleted or replaced with a clearer graphic presentation, eg, instantaneous hazard of HIV infection over time (from product-limit analyses)

21. Discussion, first para, first sentence ~ this needs to refer to "high-risk South African female sex workers", since that group forms the bulk of the study population.

22. Discussion / results ~ given the important points about transactional sex, it makes sense to present data on transactional sex practices versus fully-fledged female sex work ~ in the results

23. Discussion ~ general comment ~ the discussion contains some important points that will be of interest to readers working in HIV prevention research, but it can be hard to follow the flow of the discussion due to a lack of structure. Subheadings to help organize the discussion would help greatly in this regard.

24. The discussion lacks any attention to how these results and related experiences build on what is known already about recruitment and retention in HIV prevention research, particulary from sub-Saharan Africa. There is a sizable body of literature on these issues (including a host of lessons learned) from HIV vaccine trial work (including preparatory trial cohorts in South Africa analogous to this one) and also from microbicide trials (including a publication by Ramjee in AIDS 2000, JAIDS 2007 from what seems to be a similar study population). The work needs to be contextualized previous experiences/lessons learned (and mention of existing publications on lessons learned from HIV prevention research should be added to the introduction) ~ which of these lessons simply reinforce what has been presented in the literature already, and what experiences presented in this manuscript are novel?

25. Discussion ~ general ~ this study population is comprised predominantly of female sex workers. First, this needs to be acknowledged as a qualification on the generalisability of these results (versus general population samples of women); second, the authors may wish to draw out any differences in the results between FSW and women who were not sex workers; and third, when the authors add discussion of existing experiential knowledge on the conduct of HIV prevention research in resource limited settings (per point #24) to the intro and discussion sections, it will be useful to draw out distinctions between general population and female sex worker samples.

26. References: The abbreviation of Int J Epidemiol is incorrect in ref #1; Dunkle and McIntyre are misspelled in reference #9 (and not clear if this journal name should be abbreviated?); The abbreviation for JID is incorrect in reference #4; the editor for reference #22 is clearly misspelled; AIDS should be capitalized in references #19 and #24.