PEER REVIEW HISTORY BMJ Open publishes all reviews undertaken for accepted manuscripts. Reviewers are asked to complete a checklist review form (http://bmjopen.bmj.com/site/about/resources/checklist.pdf) and are provided with free text boxes to elaborate on their assessment. These free text comments are reproduced below. ARTICLE DETAILS TITLE (PROVISIONAL) AUTHORS HIV testing among men who have sex with men in Tijuana, Mexico: a cross-sectional study Pines, Heather; Goodman-Meza, David; Pitpitan, Eileen; Torres, Karla; Semple, Shirley; Patterson, Thomas VERSION 1 - REVIEW REVIEWER REVIEW RETURNED Carl Kendall Center for Global Health Equity Department of Global Community Health and Behavioral Sciences Tulane University School of Public Health and Tropical Medicine 1440 Canal Street, Suite 2350 Mail Stop #8319 New Orleans, LA 70112 USA 10-Nov-2015 GENERAL COMMENTS This is a well-written and paper from an excellent research team using appropriate methodology to explore an important issue concerning HIV testing among MSM. The paper could be made much better by: 1. Adhering to the new STROBE guidelines for RDS studies especially describing characteristics of the RDS data collection, such as waves, duration of the study, and other characteristics. Among the reasons to adhere to these guidelines: (2-7) 2. The researchers include both traditional MSM (homosexual, gay, bisexual men) and transwomen in their sample. This is especially relevant for RDS, since network characteristics of traditional MSM and TW are likely to be different in different parts of the world. For example, in Brazil, networks of MSM persons identifying as homosexual, gay, bisexual and hetero rarely overlap with travesti, and travesti have a very different risk profile. I wonder if inclusion on the analysis or exclusion of the small number of TW subjects would generate a difference? Using netdraw to map the sample might be helpful to see how the populations are connected. By the way, I did not see if any of the seeds were TW? The authors discuss this issue, concerning the small TW sample size, which would justify separate RDS for each population. Thye cite several studies that separate the two populations. 3. The researchers do not describe how they came up with their sample size or if they considered RDS design effects. These issues are likely to affect their results. 4. The researchers need more detail on how they identified and assigned weights for multivariate analysis. (Also in needs to be inserted in l.21, p.9). Field sample weights depend on the partition analyzed. (RDS-A provides a better solution.) 5. The researchers do not discuss when RDS estimates differ substantially from unadjusted values, or the extremely wide CI.
6. Not sure if Table 2 reports adjusted values. 7. In discussion, the researchers might chose to comment on testing guidelines, since there is substantial ambiguity about recommendations for MSM in established guidelines (p. 17, l. 16). 8. The discussion of the apparent contradiction of the role of age (p. 17, l. 34 and following) could be clarified. Yes, older participants could have already tested and know of their positive status. This brings up the issue of reporting who did and didn t test in the study, which is not reported. 9. The researchers bring up the issue of origin of their sample. The paper would be greatly enhanced if we had information on time of residence in Tijuana and national origin, since sex work populations are often very mobile. This would also influence the characterization of the RDS (are the coupons distributed casually, or to someone the recruiter really knows?). Researchers often ask: reside in X or reside in X for the past Y months as part of the inclusion cascade. Of course, (and while not a scientific consideration) if they had information about participant visits to the US or number of US clients for sex workers they would raise the profile of this paper. 10. Discussion (BTW, a Conclusions section would help the speed reader) would be enhanced mentioning PEP and PrEP programs in Tijuana. The comments on social media interventions appear out of the blue. Perhaps the authors mean to say that the successful conclusion of the study demonstrates that the population is networked, and intervention might take advantage of that. Peer Driven Interventions (Robert Broadhead) were part of the origin of RDS. The reviewer uploaded file for STROBE guidelines for RDS. Please contact the publisher to obtain this file. REVIEWER REVIEW RETURNED M. Eugenia Socías BC Centre for Excellence in HIV/AIDS, Vancouver, Canada 15-Nov-2015 GENERAL COMMENTS This is a well-conducted study of HIV testing behaviours among MSM and transwomen two populations with high burden of HIV infection in Tijuana, Mexico. Overall, it provides important information on prevalence and correlates of lifetime and recent HIV testing that could help inform interventions to increase HIV testing among these populations. Given that the authors collected information on specific barriers to HIV testing, instead of hypothesizing the causes of lower uptake of HIV testing among certain subgroups (e.g., migrants, less educated), it would have been interesting to know what the actual main barriers to HIV testing among these groups were in order to help inform the design of interventions targeted to these particular subpopulations of MSM and TW. Below I provide other minor suggestions that I hope would contribute to strengthen the manuscript Methods: 1. Please specify which statistical criteria were use to include covariates in the multivariable model (eg., all significant variables in the bivariate analysis? at which p-level?), as well as which were forced into the model based on literature review.
Results 2. Page 9, line 32. Wasn t male sex an inclusion criterion? If the authors refer to the self-identified gender of participants they should consider changing male, which refers to biological sex assigned to birth to men throughout the text and in the tables. 3. Page 9, line 47: given that recent HIV test, refers to the last 12 months a time from last HIV test in months may be more appropriate and easier to interpret. 4. Table 1: 3500 pesos, why was that threshold chosen? VERSION 1 AUTHOR RESPONSE Reviewer: 1 1. Adhering to the new STROBE guidelines for RDS studies (attached), especially describing characteristics of the RDS data collection, such as waves, duration of the study, and other characteristics. Among the reasons to adhere to these guidelines: (2-7) Response: We thank the reviewer for bringing the new STROBE guidelines for RDS studies to our attention. Our revised manuscript now adheres to these guidelines. In the attached STROBE-RDS checklist, we indicate the page or table number where information addressing each item on the checklist can be found. 2. The researchers include both traditional MSM (homosexual, gay, bisexual men) and transwomen in their sample. This is especially relevant for RDS, since network characteristics of traditional MSM and TW are likely to be different in different parts of the world. For example, in Brazil, networks of MSM persons identifying as homosexual, gay, bisexual and hetero rarely overlap with travesti, and travesti have a very different risk profile. I wonder if inclusion on the analysis or exclusion of the small number of TW subjects would generate a difference? Using netdraw to map the sample might be helpful to see how the populations are connected. By the way, I did not see if any of the seeds were TW? The authors discuss this issue, concerning the small TW sample size, which would justify separate RDS for each population. Thye cite several studies that separate the two populations. Response: We believe that the social and sexual networks of MSM and TW do overlap in Tijuana, Mexico. However, the small number of TW (n=14) included in our sample is likely due to the fact that (1) none of the seeds were TW and (2) participants were instructed to invite male peers who have sex with men within their social network to participate in the study. Thus, it is more accurate to describe RDS implementation in this study as a strategy to recruit MSM (not TW). We now clarify these instructions in the study population and sampling methods section and refer to our study population throughout the manuscript as MSM only. In a sensitivity analysis, we excluded participants who identified as TW, but the results were qualitatively similar (this information has been added to the results section). Thus, we present findings from the full sample of eligible peer-recruits who represent biological males who have sex with men without a previous HIV diagnosis (n=189), including those who identified as TW. 3. The researchers do not describe how they came up with their sample size or if they considered RDS design effects. These issues are likely to affect their results. Response: The primary purpose of this study was to determine the prevalence of HIV among MSM in Tijuana, Mexico (Pitpitan et al., J Int AIDS Soc, 2015). Recruitment continued until a sample size of 200 was achieved. A sample size of 200 was calculated as the minimum sample size needed to
detect an HIV prevalence of 20% with at least 80% power assuming a design effect of 2 and an alpha-level of 0.05. This information has been added to the study population and sampling methods section. 4. The researchers need more detail on how they identified and assigned weights for multivariate analysis. (Also in needs to be inserted in l.21, p.9). Field sample weights depend on the partition analyzed. (RDS-A provides a better solution.) Response: Individual RDS weights were generated following an RDSAT partition analysis of prior HIV testing (outcome of interest) and were applied to the analysis sample for use in RDS-weighted logistic regression analyses. This information has been added to the statistical analysis section. The word in has also been inserted as suggested. 5. The researchers do not discuss when RDS estimates differ substantially from unadjusted values, or the extremely wide CI. Response: Differences in RDS-unadjusted and RDS-adjusted estimates may be explained by preferential recruitment of peers with similar characteristics and differential recruitment by personal social network size (now summarized in Supplemental Table 2), which are accounted for in RDSadjusted estimates. This information has been added to the results section. RDS weighting procedures can produce imprecise estimates (Winship et al., Sociol Methods Res, 1994; Heckathorn, Social Methodol, 2007), and may explain the wide confidence intervals obtained for some odds ratios in our RDS-weighted logistic regression analyses. We now include this as a limitation in the discussion section. This fact, combined with the fact that our RDS-weighted multivariate logistic regression analysis of recent ( 12 months) HIV testing consisted of only 116 participants, likely explains the extremely wide confidence intervals presented in Table 4. Rather than report such unstable estimates that contribute little information to the literature on correlates of recent HIV testing among MSM, we decided not to report our findings from that analysis and have deleted Table 4. 6. Not sure if Table 2 reports adjusted values. Response: Table 2 reports RDS-unadjusted sample estimates. To clarify this, we have added a footnote (a) to the Table. 7. In discussion, the researchers might chose to comment on testing guidelines, since there is substantial ambiguity about recommendations for MSM in established guidelines (p. 17, l. 16). Response: While the recommendation that MSM be tested at least annually is clear, we agree the recommendation for more frequent testing (i.e., every 3 to 6 months) among MSM at substantial risk is much less clear, particularly with respect to how substantial risk should be defined. However, given that this manuscript focuses on identifying the prevalence and correlates prior HIV testing to inform the development of testing strategies to promote testing uptake among MSM in general, and not only those at substantial risk, we feel that commenting on this ambiguity is beyond the scope of the manuscript. 8. The discussion of the apparent contradiction of the role of age (p. 17, l. 34 and following) could be clarified. Yes, older participants could have already tested and know of their positive status. This brings up the issue of reporting who did and didn t test in the study, which is not reported. Response: Because our analysis excluded previously diagnosed HIV-positive peer-recruits (n=2), our
contradictory findings with respect to age cannot be due to the fact that older participants already knew their HIV-positive status and thus were less likely to have tested for HIV in the past 12 months. Rather we believe that this difference is due to older participants greater cumulative access to HIV testing services throughout their lifetime than younger participants. While additional research is needed to understand why younger age was associated with recent HIV testing, younger participants may have sought HIV testing in the past 12 months because they perceived themselves to be at greater risk of HIV infection than older participants. However, since we have decided not present the findings for our RDS-weighted logistic regression analysis of recent ( 12 months) HIV testing for the reasons provided in our response to item 5 above, we no longer present or discuss our contradictory findings with respect to the role of age. All participants underwent rapid and confirmatory (if necessary) HIV testing as part of the study. This has been clarified in the study procedures section. 9. The researchers bring up the issue of origin of their sample. The paper would be greatly enhanced if we had information on time of residence in Tijuana and national origin, since sex work populations are often very mobile. This would also influence the characterization of the RDS (are the coupons distributed casually, or to someone the recruiter really knows?). Researchers often ask: reside in X or reside in X for the past Y months as part of the inclusion cascade. Of course, (and while not a scientific consideration) if they had information about participant visits to the US or number of US clients for sex workers they would raise the profile of this paper. Response: Information on time of residence in Tijuana and number of US clients among sex workers is not available. RDS-unadjusted and RDS-adjusted estimates for birthplace (Tijuana, Mexico; Outside Tijuana in Mexico; or US) have been added to Table 1. RDS-unadjusted and RDS-adjusted estimates for travel to the US and deportation from the US have also been added to Table 1. However, due to a large number of missing values for travel to the US (n=34) and deportation from the US (n=39) these covariates were not examined in regression analyses as explained in the statistical analysis section. Participants were recruited by friends (78%), sex partners (15%), acquaintances (5%), and relatives (2%). This information has been added to the beginning of the results section. 10. Discussion (BTW, a Conclusions section would help the speed reader) would be enhanced mentioning PEP and PrEP programs in Tijuana. The comments on social media interventions appear out of the blue. Perhaps the authors mean to say that the successful conclusion of the study demonstrates that the population is networked, and intervention might take advantage of that. Peer Driven Interventions (Robert Broadhead) were part of the origin of RDS. Response: Post-exposure prophylaxis (PEP) is not widely used among key populations and antiretrovirals are not yet approved for use as pre-exposure prophylaxis (PrEP) in Mexico. However, provider training on culturally competent care for sexual minorities may support the delivery antiretrovirals for use as both PrEP once approved in Mexico and PEP to MSM in Tijuana. As such, in addition to promoting open discussions about sexual practices, facilitating risk reduction counseling, increasing HIV testing, and ensuring timely linkage to HIV care for MSM in Tijuana, we now refer to the potentially positive impact provider training on culturally competent care for sexual minorities could have on the development and implementation PrEP and PEP delivery programs in Tijuana. We agree with the reviewer, and now state that our ability to recruit MSM via RDS networks suggests interventions delivered via social networks that rely on the diffusion of information through networks may increase HIV-related knowledge and support regular HIV testing uptake among MSM in Tijuana.
Finally, we have added a conclusions section to the end of the manuscript as suggested. Reviewer: 2 This is a well-conducted study of HIV testing behaviours among MSM and transwomen two populations with high burden of HIV infection in Tijuana, Mexico. Overall, it provides important information on prevalence and correlates of lifetime and recent HIV testing that could help inform interventions to increase HIV testing among these populations. Given that the authors collected information on specific barriers to HIV testing, instead of hypothesizing the causes of lower uptake of HIV testing among certain subgroups (e.g., migrants, less educated), it would have been interesting to know what the actual main barriers to HIV testing among these groups were in order to help inform the design of interventions targeted to these particular subpopulations of MSM and TW. Response: Barriers to HIV testing were not measured via open-ended questions in this study. However, we agree with the reviewer, and now suggest the need for future qualitative research among MSM subgroups in Tijuana with a low prevalence of prior HIV testing to better understand their barriers to testing and inform targeted HIV testing interventions that support testing uptake among these vulnerable MSM. This suggestion can be found in the final paragraph of the discussion section. Methods: 1. Please specify which statistical criteria were use to include covariates in the multivariable model (eg., all significant variables in the bivariate analysis? at which p-level?), as well as which were forced into the model based on literature review. Response: Covariates were selected for inclusion in multivariate regression analyses based on previous research indicating their association with HIV testing among MSM (i.e., age, sexual orientation, education, exchange of money for sex in the past 6 months, number of condomless anal intercourse acts [past 2 months], history of abuse, outness about having sex with men, internalized homophobia, HIV knowledge, self-reported STI diagnosis [past 2 months]) and univariate results (i.e., p-value 0.05; birthplace, venues visited to meet male sexual partners, substance use). This information has been added to the statistical analysis section. Results: 2. Page 9, line 32. Wasn t male sex an inclusion criterion? If the authors refer to the self-identified gender of participants they should consider changing male, which refers to biological sex assigned to birth to men throughout the text and in the tables. Response: While biological male sex was an inclusion criteria, participants could still identify their gender as male or transgender female. We have clarified on page 9 (line 32) and in Table 1 that 93% of participants identified as male. We have also clarified that we examined the association between male gender identity and prior HIV testing in Table 3. 3. Page 9, line 47: given that recent HIV test, refers to the last 12 months a time from last HIV test in months may be more appropriate and easier to interpret. Response: Median time since last HIV test is now reported in months (8.9 months [IQR: 4.1-24.9] in the results section.
4. Table 1: 3500 pesos, why was that threshold chosen? Response: The categories used to collect information on monthly income (no income, <$1000, $1000- $1499, $1500-$1999, $2000-$2499, $2500-$2999, $3000-$3500, >$3500) have been added to the measures section. These categories have been used in previous research with key populations in the Mexico-US border region (Strathdee et al., JAIDS, 2008; Strathdee et al., PLoS ONE, 2008), which selected a threshold of 3000 pesos when examining factors associated with HIV infection. However, for consistency with the primary outcome paper from this study (Pitpitan et al., J Int AIDS Soc, 2015), we selected a threshold of 3500 pesos. VERSION 2 REVIEW REVIEWER REVIEW RETURNED Eugenia Socias BC Centre for Excellence in HIV/AIDS, Vancouver, BC, Canada 13-Jan-2016 GENERAL COMMENTS The authors have carefully addressed reviewers' concerns. I am satisfied with their responses.