Sexually transmitted infections amongst men who have sex with men (MSM) in South Africa

There is limited data about bacterial STIs in MSM populations in sub-Saharan Africa. Our retrospective analysis used data from the HVTN 702 HIV vaccine clinical trial (October 2016 to July 2021). We evaluated multiple variables. Polymerase chain reaction testing was conducted on urine and rectal samples to detect Neisseria gonorrhoea (NG) and Chlamydia trachomatis (CT) every 6 months. Syphilis serology was conducted at month 0 and thereafter every 12 months. We calculated STI prevalence and the associated 95% confidence intervals until 24 months of follow-up. The trial enrolled 183 participants who identified as male or transgender female, and of homosexual or bisexual orientation. Of these, 173 had STI testing done at month 0, median age was 23 (IQR 20–25) years, with median 20.5 (IQR 17.5–24.8) months follow-up (FU). The clinical trial also enrolled and performed month 0 STI testing on 3389 female participants, median age 23 (IQR 21–27) years, median 24.8 (IQR 18.8–24.8) months FU and 1080 non-MSM males with a median age of 27 (IQR 24–31) years, median 24.8 (IQR 23–24.8) months FU. At month 0, CT prevalence was similar in MSM and females (26.0% vs 23.0%, p = 0.492) but was more prevalent in MSM compared to non-MSM males (26.0% vs 14.3%, p = 0.001). CT was the most prevalent STI among MSM at months 0 and 6 but declined from month 0 to month 6 (26.0% vs 17.1%, p = 0.023). In contrast, NG did not decline in MSM between months 0 and 6 (8.1% vs 7.1%, p = 0.680) nor did syphilis prevalence between months 0 and 12 (5.2% vs 3.8%, p = 0.588). Bacterial STI burden is higher in MSM compared to non-MSM males, and CT is the most prevalent bacterial STI amongst MSM. Preventive STI vaccines, especially against CT, may be helpful to develop.

The authors seem to have data on behaviors during follow-up (ln 123-129). If so, it is unclear why they did not also evaluate the changes in behavior over time to corroborate their findings of changes in STI prevalence. These need to be added to the analysis, if available.
Response: Risk behavior data was collected throughout the trial but not at the same time points as the STI data collection. For example, testing for Neisseria gonorrhoea (NG) and Chlamydia trachomatis was performed at month 6 but risk behavior was collected at month 6.5. As the manuscript focuses on the assessment of differences in STI prevalence at the first two time points, we evaluated whether demographic and month 0 risk behavior variables differed between MSMs with and without STIs at months 0 and 6. These results are reported in Table 4 and we have added a description of this analysis to the Statistical Analysis section.

The analysis on risk factors for STI is ambiguous (Table 5) and the footnotes do not adequately describe what was being modeled as an outcome. Any STI within 24
Sexually transmitted infections amongst men who have sex with men (MSM) in South Africa PLOS Global Public Health months? And risk factors evaluated at which time points (as the authors state "between months 0-24", ln 148)? If the covariates were time-updated, they could simply model the outcome STI at each time point along with corresponding covariates measured at the same time points. Techniques to account for repeated measured (i.e., GEE or mixedeffects) could be applied accordingly. Regardless, this analysis needs to be described in more detail in the statistical analysis section.
Response: Thank you for pointing out that this analysis could be better clarified. We have added a description to the Statistical Analysis section and revised Table 4 (formerly Table 5) to report and compare the rates of demographic and risk behaviour characteristics at baseline among MSMs with at least one STI at months 0 and/or 6 vs. those without an STI at months 0 and 6. Information about the MSM groups now appears in the column headers rather than in footnotes to the table and we restrict to STIs at the first two time points which are of greatest interest and contain the most complete data.

The reviewers asked for more clarity in the statistical analysis section and the authors did not accommodate these requests accordingly (even though their answerers were sufficient). Please state the endpoint being modelled. This is a logistic regression model with general estimating equations; please state as such (if this model was indeed entertained).
Response: Additional detail has been added to the Statistical Analysis section to clarify that STI prevalence was modelled as a binary endpoint using general estimating equations with a logit link and independent correlation structure.
One final general comment: please adhere to the STROBE reporting guidelines. There are several items that appear to be missing in this manuscript.
Response: Thank you for the suggestion to align to STROBE, which we have done as far as possible. We have improved clarity in the manuscript that we performed a retrospective analysis on clinical trial data, i.e. HVTN 702 was not an observational study. We changed the Abstract to read: The clinical trial also enrolled and performed month 0 STI testing on 3389 female participants, median age 23 (IQR 21-27) years, median 24.8 (IQR 18.8, 24.8) months FU and 1080 non-MSM males with a median age of 27 (IQR 24-31) years, median 24.8 (IQR 23, 24.8) months FU. The primary follow-up period was 24 months.
We have added comments about differential LTFU. Line 268 -273 now reads: Following non-efficacy declaration, there was a reduction in study follow up visits. This was the largest contributor to the reduction in number of participants and STI data over time. HIV infection, study dropouts and missed visits also contributed to the loss of study participants and data. There is potential for differential loss to follow up (LTFU) bias as participants who either tested positive for HIV or an STI dropped out of the study. This would result in an underestimation of the prevalence of STIs at later time points. ln 52. What type of "health-seeking behavior" is being referenced here? Also, there are no citations to support this claim.
Response: Thank you for this question. Citation has been added. Line 52 now reads: There is poor health-seeking behavior (accessing HIV/STI screening and testing services, and disclosure of sexual practices to healthcare workers) amongst MSM within South Africa and other Sub-Saharan countries [6] ln 58. In what ways are the "profiles" being described? "changes in STI prevalence" would be more appropriate.
Response: We agree and have replaced "profiles" with "prevalence", avoiding the mention of changes in prevalence as these are only considered for the first two time points whereas prevalence is described for the entire follow-up period. The sentence has been changed to "In our analysis, we describe the STI prevalence amongst MSM enrolled in a preventive HIV vaccine trial which had routinely tested for sexually transmitted infections as part of study procedures" Sexually transmitted infections amongst men who have sex with men (MSM) in South Africa PLOS Global Public Health Table 1. Why include a column on all MSM? Since the main outcome is STI prevalence, would it be more obvious to exclude those without an STI measurement? Or did those without STI measured at month 0 have an STI measured at a later time point?
Response: STI testing was added under version 2 of the protocol with STI testing initiated on 08 May 2017. Participants enrolled under version 1 of the protocol did not have STI testing performed. This resulted in 10 MSM participants not being tested for STI at month 0. However, these participants did receive STI testing at later time points. Table 1 includes a column for all MSM to indicate that demographics are similar among all MSM and MSM with STI testing at month 0.
The inclusion and exclusion criteria could be more thoroughly described.
Response: Thank you for this suggestion. Line 75-77 now reads: To be enrolled into HVTN 702, participants had been assessed as being healthy, 18-35-year-old HIV-uninfected adults at risk of HIV acquisition who were willing to receive HIV counselling and testing. The final decision to either include or exclude a volunteer from the study was based on laboratory test results, medical history, physical examination findings and answers to questionnaires. Possible barriers to retention e.g. employment were also considered. The detailed inclusion/exclusion criteria is provided in the supplementary section Table 3 is excessive. The authors could very well add a sentence, such as: "XX CT infectious were observed in NN individuals." to begin the description of prevalence over time.
Response: Thank you for your comment about Table 3. Table 3 has been deleted. Line 149 -151 now reads, Overall, there were 41 unique participants that accounted for the 50 participants/STI combinations that had multiple positives across the time-points Months 0, 6, 12, 18, 24.
ln 252-256. The authors seem to be missing the point of the reviewer. Normally, if you test in more places, you are more likely to find an STI. Is there a differential diagnostic bias between MSM, females at birth and non-MSM? For example, if the non-MSM never engaged in insertive anal sex with a male partner, their risk of having an anal STI is essentially nil, hence no reason to test for rectal STIs. The potential for differential diagnostic bias would likely depend on whether MSM was correctly classified in this instance. Please modify the discussion accordingly.
Response: Thank you for your comment. The Discussion now reads this: For our analysis, CT and NG results among MSM were each considered positive if either the urine or rectal sample testing was positive. Urine testing for non-MSMs was regarded as sufficient for detection of STIs in this population. Rectal testing was excluded in non-MSMs,