Article Text

Original article
Using conjoint analysis to determine the impact of product and user characteristics on acceptability of rectal microbicides for HIV prevention among Peruvian men who have sex with men
1. Eric C Tang1,
2. Jerome T Galea2,3,
3. Janni J Kinsler4,
4. Pedro Gonzales5,
5. Magdalena E Sobieszczyk1,6,
6. Jorge Sanchez5,7,
7. Javier R Lama5,7
1. 1Columbia University College of Physicians & Surgeons, New York, New York, USA
2. 2Epicentro Salud, Lima, Peru
3. 3Socios En Salud, Lima, Peru
4. 4Division of General Internal Medicine and Health Services Research, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, California, USA
5. 5Asociación Civil Impacta Salud y Educación, Lima, Peru
6. 6Department of Medicine, Division of Infectious Disease, Columbia University College of Physicians & Surgeons, New York, New York, USA
7. 7Department of Global Health, University of Washington, Seattle, Washington, USA
1. Correspondence to Dr Eric C Tang, Graduate Medical Office, 2425 Geary Blvd, M160, San Francisco, CA 94115, USA; ect2119{at}columbia.edu

## Abstract

Objectives Men who have sex with men (MSM) are in need of novel and acceptable HIV prevention interventions. In Peru, a Phase II clinical trial was recently completed evaluating rectally applied tenofovir gel among Peruvian MSM and transgender women. If deemed safe and acceptable, the product could move into efficacy testing, but acceptability data for similar products are needed now in order to prepare for future implementation. Peru is in need of expanded, national acceptability data among likely users.

Methods Using conjoint analysis of an online cross-sectional survey taken by 1008 Peruvian MSM and transgender women, we tested the acceptability of eight hypothetical rectal microbicide (RM) products comprising six, dual-value attributes. We also assessed the relationship of select product attributes with sample characteristics.

Results Highest acceptability was found for a RM that was 90% effective, used before and after sex, without side effects, costing approximately $0.30, had no prescription requirement and had a single-use applicator. Product effectiveness and presence of side effects were the factors most likely to drive RM acceptance and use. Education, sexual orientation, sexual role and concern for HIV infection were also related to aspects of RM acceptability. Conclusion RM acceptability was high, confirming the results of earlier, smaller studies and placing confidence in the acceptability of RMs. Analysis of the relationships with product attributes and sample characteristics underscore the need to consider the impact of factors such as sexual orientation, sexual role, level of education and concern for HIV acquisition on RM acceptability. • HIV • LATIN AMERICA • MICROBICIDES • PREVENTION • ATTITUDES ## Statistics from Altmetric.com ## Introduction Men who have sex with men (MSM) and transgender women are key groups vulnerable to HIV1 ,2 who are in great need of safe, effective—and acceptable—prevention strategies.3 Peru is emblematic of an escalating global disparity in HIV rates between MSM and the general population,4 where the HIV prevalence among Peruvian MSM is estimated at 12.4% compared with 0.4% reported for the general population.5 Considerable research worldwide has focused on biomedical HIV prevention interventions using antiretrovirals6 including treatment as prevention, oral pre-exposure prophylaxis (PrEP) and, most recently, rectal microbicides (RMs). RM clinical trials have investigated tenofovir-based gels for rectal use to prevent HIV infection in MSM and transgender women,7 including a Phase II clinical trial recently completed in Peru.8 RMs formulated as gels could be easily adopted, since using lubricants during anal sex is common among MSM and transgender women.9–11 Nonetheless, they will need to be acceptable by target populations and product characteristics like effectiveness, cost, side effects, frequency of use and delivery mode could affect uptake.12–15 Additionally, certain demographic, behavioural and sociocultural factors external to RMs may influence product uptake.16 ,17 Previous studies have demonstrated that MSM in multiple countries have been interested in RMs.10 ,13 ,17–19 In Peru, one large study found that 77% of MSM were willing to use a lubricant to prevent HIV transmission.10 A second RM acceptability study, including 63 Peruvian MSM and transgender women, used conjoint analysis (CJA) to understand the preferences that potential users had for specific RM characteristics,13 finding that most preferred an RM that was 80% (vs 40%) effective; had no (vs some) side effects; was used before sex (vs daily); formulated as a gel (vs a liquid); 15 mL (vs 35 mL) in volume and available by prescription (vs over-the-counter). The strength of the latter study is that CJA permits the individual measurement of the value that study participants placed on each of the RM characteristics thereby allowing the reliable prediction of the use of real products with the same characteristics.20 For example, CJA measuring patients’ HIV treatment medication preferences successfully predicted their actual medication choices.21 For this reason, CJA has been increasingly used in the HIV biomedical prevention acceptability field.22 ,23 In order for CJA to be able to accurately predict product use, the hypothetical products tested (called ‘scenarios’) must closely resemble the actual product of interest. The CJA study in Peru mentioned above, conducted in 2009–2010,13 used product characteristics based on the state of RM development at the time. Five years later, as RM development has evolved, some of the product attributes studied are no longer under consideration or have updated values, thereby weakening the predictive value of the study results. At the same time, we need more information on how RM acceptability is influenced by factors external to the product themselves—such as user demographic and behavioural characteristics—from larger and more diverse samples.16 The present cross-sectional study was conducted to provide the most robust RM acceptability data to date using CJA and enhanced socio-behavioural data among a national sample of Peruvian MSM and transgender women. ## Methods ### Study participants and procedures Between April and June 2014, we recruited participants throughout Peru using paid advertisements on Facebook.com and two Peruvian gay websites (peruesgay.com and gaysperuanos.com). On Facebook, we targeted ads to reach male-identified users aged ≥18 years in Peru who liked pages, had interests or listed activities related to MSM-related terms such as ‘gay’, ‘bisexual’, ‘transgender’ and ‘gay pornography’. The advertisements directed potential participants to an online consent form requiring confirmation that it had been read and accepted before proceeding to the study survey. The cross-sectional survey included study eligibility screening questions (≥18 years old, biologically male at birth, sex with a man or transgender woman in the past year). Survey responses were anonymous, IP addresses were not recorded and browser cookies limited one survey completion per computer. Financial compensation was not provided. ### Measures The survey included a maximum of 59 items and required approximately 10 minutes to complete. Basic demographic information was first collected, followed by an explanation of RMs: ‘Rectal microbicides are gels that contain anti-HIV compounds that could be utilized for the prevention of HIV during anal sex. Currently, there are no rectal microbicides available for use, but there are many on-going studies testing their safety and acceptability’. Next, two CJA exercises assessed participants’ preferences for eight different RM scenarios comprising six, dual-value attributes: effectiveness (90% vs 50%); frequency of use (before and after sex vs daily); side effects (none vs some); cost ($0.30 vs $4.50); prescription (none vs needed) and applicator type (single-use vs reusable); table 1 lists the attributes, their values and selection justification. Table 1 Rectal microbicide (RM) scenario attribute, preferred and non-preferred attribute values and justification for conjoint analysis inclusion A full-factorial design for six attributes, each with two values, would yield 64 (26=64) different RM scenarios. Since asking participants to rate all 64 scenarios would be difficult and burdensome, a fractional factorial orthogonal design is often used in CJA experiments to permit the efficient estimation of several main factors using a subset of all possible combinations.29 We reduced the number of scenarios from 64 to 8 while ensuring that all of the attribute/level combinations appeared the same number of times. The attributes were described in lay language with examples to aid comprehension. For example, the applicator was described as ‘a device that one would use to put the RM (gel, lubricant) in the rectum before anal sex’, and a single-use applicator was described as ‘prefilled, but can be used only once before throwing it away’, while a reusable applicator was described as ‘having to be washed and refilled yourself, but can be used again and again’. For the first exercise, participants marked their likelihood of using each RM on a scale from 1 (‘definitely would use’) to 5 (‘definitely would not use’), and multiple scenarios could share the same rating. In the second exercise, participants ranked the eight RM scenarios from 1 (‘most likely to use’) to 8 (‘least likely to use’), but none of the scenarios could share the same value. For both exercises, the scenarios were presented randomly to prevent order effect bias. Following the CJA exercises, participants were asked to explicitly indicate which of the two options they preferred for three of the RM attributes in table 1 (frequency of use, prescription requirement and applicator type) with the additional choice of no preference. The survey ended with questions about sexual health, HIV status and more detailed demographics. ### Statistical analysis Univariate analyses were used to describe the study population. For the first CJA exercise, the acceptability of each of the eight hypothetical RM scenarios was derived by converting the five-point Likert scale into a 0–100 scale, whereby ‘definitely would use’=100 and ‘definitely would not use’=0 and then summing and averaging the RM acceptability ratings across respondents. For the second CJA exercise, a one-way analysis of variance (ANOVA) model was applied to fit each respondent's acceptability ratings across the eight RM scenarios. The six RM attributes served as independent variables in the model. The effect of each RM attribute from the ANOVA model was the impact score of the attribute on RM acceptability for the individual respondent. Attribute values were sorted as ‘preferred’ and ‘non-preferred’ in order to arrive at the overall impact score for each attribute, which was derived by subtraction (impact score=(acceptability of preferred RM attribute)−(acceptability of non-preferred RM attribute)). Individual impact scores were averaged across respondents for each attribute to compute its influence on overall RM acceptability. A one-sample t-test determined the statistical significance of the impact of each attribute on RM acceptability. Finally, unadjusted and adjusted ORs (UOR/AOR) assessed the associations between sample characteristics and frequency of use, prescription needed and applicator type. Participants who reported being exclusively activo (only practicing insertive anal sex) and/or those who self-reported being HIV-positive were not excluded from study participation since use of a RM may involve partners who apply or provide the product for anal intercourse. However, since the interpretation of RM use may differ for exclusively activo or HIV-positive persons, we conducted three separate sensitivity analyses by removing participants from these groups from the overall analyses. The study was approved by the Institutional Review Boards at both the Columbia University Medical Center (IRB-AAAK9302) and the Asociación Civil Impacta Salud y Educación (N° 0281-2012-CE). ## Results ### Sample characteristics Of the 6385 individual views of the consent form, 4448 participants (69.7%) agreed to complete the survey. A total of 2997/4448 participants (67.4%) answered the first page of questions, which included basic demographic and screening questions. Of the 2997 participants, 2370 (79.1%) met the inclusion criteria (≥18 years old, born male and had sex with a man or transgender woman in the past year), and 1173/2370 (49.5%) completed the survey. An additional two participants were removed because they lived outside of Peru, leaving a total of 1171 target participants that completed the survey. The sample for this analysis was based on 1008/1171 participants (86.1%) who answered all items for both CJA exercises. Most (895/1008 or 89%) participants were recruited from Facebook while 113/1008 (11%) were recruited from the two Peruvian websites. Socio-demographics, sexual risk behaviours and HIV-related factors are presented in table 2. Table 2 Socio-demographics, sexual risk behaviours and HIV-related factors of total sample (n=1008) The majority of participants self-identified as gay (65%) or bisexual (32%). Approximately half (42%) of the participants self-identified as pasivo (receptive partner during anal intercourse), 59% reported unprotected anal sex with one or more partners in the previous 3 months and 8% of participants reported being infected with HIV. ### Conjoint analysis Table 3 shows the acceptability of all eight RM scenarios (exercise 1) for the total sample and by sexual orientation (gay, bisexual, heterosexual/other). Table 3 Comparison of hypothetical RM acceptability scenarios among the total sample (n=1008) and by self-identified sexual orientation: gay (n=654), bisexual (n=322) and heterosexual/other (n=29) Table 4 shows the impact of each RM attribute on acceptability (exercise 2) for the total sample and by sexual orientation. Table 4 Comparison of impact scores of RM attributes on hypothetical RM acceptability among the total sample (n=1008) and by self-identified sexual orientation: gay (n=654), bisexual (n=322), and heterosexual/other (n=29) Mean RM acceptability among the total sample was 61.6 on a 0–100 point scale (range: 51.9–85.1) and was highest among heterosexuals (66.3; range: 54.3–86.2), followed by bisexuals (63.3; range: 52.1–85.8) and gays (60.7; range: 50.5–85.0). The RM scenario with the highest acceptability for the total sample and each subgroup was scenario 8: 90% effective; use before and after sex; no side effects;$0.30 per dose; no prescription and single-use applicator.

In exercise 2, effectiveness had the single greatest impact on acceptability across the six RM attributes for the total sample and for each subgroup (table 4). Participants reported significantly higher RM acceptability for a product that was 90% effective (acceptability=70.0 on a 0–100 point scale) compared with 50% effective product (acceptability=53.2 on a 0–100 point scale), yielding a net impact score of 16.8 (p<0.001). Net impact scores for the effectiveness attribute for gay men, bisexuals and heterosexuals were 17.6 (p<0.001), 15.2 (p<0.001) and 14.9 (p<0.001), respectively.

Side effects had the second-greatest impact on RM acceptability for the total sample, gay men and bisexuals. Among the total sample, participants reported significantly higher RM acceptability for a product with no side effects (acceptability=66.9), compared with a product with some side effects (acceptability=56.3), yielding a net impact score of 10.6 (p<0.001). Net impact scores for gay men and bisexuals were 10.7 (p<0.001) and 10.9 (p<0.001), respectively.

Prescription requirement had the third-greatest impact on RM acceptability for the total sample, gay men and bisexuals. Among the total sample, participants reported significantly higher RM acceptability for a product that did not require a prescription (acceptability=65.0), compared with one requiring a doctor's prescription (acceptability=58.2), yielding a net impact score of 6.8 (p<0.001). Net impact scores for gay men and bisexuals were 7.2 (p<0.001) and 5.9 (p<0.001), respectively.

### Sensitivity analyses

When excluding HIV-positive individuals, exclusively activo participants, and those who reported either being HIV-positive or exclusively activo, overall RM acceptability was 61.2, 62.2, and 61.7, respectively. These scores were not significantly different from the overall acceptability of RM when all 1008 participants were included (61.6).

### Association of sample characteristics: frequency of use, prescription requirement and mode of RM delivery

#### Frequency of use

Participants who were concerned about contracting HIV in the future were more likely to prefer using a RM before and after sex compared with participants who were not (UOR=1.96; 95% CI 1.15 to 3.37; p<0.05; AOR=1.95; 95% CI 1.10 to 3.47; p<0.05).

#### Prescription requirement

Participants with greater than a high school education were less likely to prefer using a RM not requiring a prescription compared with participants with a high school degree or less (UOR=0.37; 95% CI 0.22 to 0.62; p<0.001; AOR=0.35; 95% CI 0.20 to 0.63; p<0.001), and activo participants were less likely to prefer using a RM not requiring a prescription compared with pasivo participants (UOR=0.48; 95% CI 0.30 to 0.78; p<0.01; AOR=0.46; 95% CI 0.28 to 0.78; p<0.01).

#### RM applicator

Activo participants were more likely to prefer a single-use applicator compared with pasivo participants (UOR=1.41; 95% CI 1.00 to 2.00; p<0.05; AOR=1.49; 95% CI 1.01 to 2.18; p<0.05).

## Discussion

This study found a high acceptability of RMs among a diverse population of Peruvian MSM and transgender women with an overall mean RM acceptability score of 85.1 out of 100 for the highest-rated CJA scenario (90% effective, use before and after sex, no side effects, \$0.30 cost, no prescription required, single-use applicator). We found that a higher effectiveness and the absence of side effects are factors most likely to drive the actual use of RMs similar to what has been reported in HIV vaccine acceptability (using full rank discrete choice modelling)30 and vaginal microbicide work (which used full-profile CJA).w1 These results bolster confidence in the uptake of the same or similar product evaluated in MTN-0178 among Peruvian MSM and transgender women, which had few side effects and was tested for pericoital use with a single-use applicator,8 just like the most favoured scenario in our study.

The results also confirm findings from the previous RM CJA study in South America,13 with one exception. In the present study, MSM preferred products available without a prescription while in the previous study, a prescription was preferred. This discrepancy could be related to the fact that the previous study was conducted in-person at study sites, potentially favouring persons comfortable with their sexuality and interacting with healthcare personnel, while the use of the internet in the present study may have favoured persons preferring anonymity. This discrepancy warrants further research since product access will be mediated by much larger sociocultural issues such as homoprejudice, stigma and discrimination that may prevent MSM and transgender women from purchasing a product for anal intercourse regardless of the prescription requirement.16

We analysed acceptability data by sexual orientation, demographic and sexual behaviours to better understand the characteristics of potential RM users. We found that the mean RM acceptability was highest among heterosexuals (66.3) and lowest among gay men (60.7), yet the overall mean impact scores for all six attributes were statistically significant for gay men and bisexuals while only the effectiveness impact score was statistically significant for heterosexuals. This may be due to the significantly smaller sample size for heterosexuals. In general, the attributes with the greatest impact for both gay and bisexual men were effectiveness, side effects and prescription needed while the attributes with the greatest impact for heterosexuals were effectiveness, cost and side effects.

Participants with more than a high school degree were much more likely to prefer a RM with a prescription requirement. These higher-educated participants may feel more wary of their widespread consumption and believe a healthcare professional should assess the appropriateness of their use, given their potential side effects.

Regarding sexual role, exclusively or mostly activo participants preferred RMs with a single-use applicator and one that required a prescription. Qualitative data could help interpret these findings, though one explanation for the preference of single-use applicators may be the desire to have the product available when needed for their receptive partners.

Participants who were concerned about contracting HIV in the future were more likely to prefer using a RM before and after sex versus daily. While reasons for being concerned about acquiring HIV are complex, one explanation may be that those most concerned about contracting HIV would want to use the microbicide around times of highest risk, which for MSM, is before and after anal sex.

Our study adds to a growing body of RM acceptability literature among MSM now encompassing Peru,10 ,11 ,13 ,16 w2 the US,14 ,19 Puerto Rico,w3 and Thailand.17 w4 An overarching message across these studies is that MSM (including those identifying as gay, heterosexual or bisexual), male sex-workers and transgender women generally find the concept of topically applied rectal product that prevents HIV acceptable. But the literature has also signalled potential barriers to uptake, such as less-than perfect products (lower effectiveness or with side effects) and a host of sociocultural issues including stigma and discrimination faced by MSM and transgender women.16 ,17 These issues are vital to understand as the first RM to come to market should be prioritised to those most vulnerable to HIV, like transgender women, who are also among the most stigmatised sexual minorities worldwide.

This study had limitations. Using non-randomised sampling, our data are not generalisable beyond the study sample. For instance, the use of the internet excluded those who did not visit the internet recruitment sites or those without internet access thereby introducing additional bias. The sample was better educated than other studies of Peruvian MSM and transgender women,13 w2 potentially due to the internet use. We chose CJA in order to build on our previous RM acceptability work using a similar approach; it is only as robust as the attributes/levels included in the model. Our attribute/levels choices were informed by current RM development, but other attributes not included could influence RM acceptability. Furthermore, we assume orthogonality (ie, the attributes vary independently) in our CJA model, but this may not be the case, and the external validity of CJA has been challenged when compared with related approaches like discreet choice experiments.w5

## Conclusion

Both product attributes and user characteristics affect RM acceptability. This study builds confidence in the acceptability of the RM that was recently evaluated in Peru or a product with similar characteristics if found efficacious in preventing incident HIV infection. MSM and transgender women worldwide are in urgent need of new and HIV prevention options like RMs, and continued research is needed to fully prepare for their arrival.

### Key messages

• If found acceptable, rectal microbicides may be a potential tool to curb the HIV endemic among Peruvian men who have sex with men and transgender women.

• Among this population, rectal microbicide acceptability is high, though product effectiveness and side effects appear to drive rectal microbicide acceptability and use.

• Prescription requirement, cost, frequency of use and applicator type may also be important considerations in the uptake of a future rectal microbicide.

• Rectal microbicide development should consider their acceptability among target populations and may be influenced by sexual orientation/role, education level and concern for acquiring HIV.

## Acknowledgments

The authors would like to acknowledge the study participants as well as the staff at IMPACTA with whom this study was made possible.

• ## Supplementary Data

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

• ## Abstract in Spanish

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

## Footnotes

• Handling editor Jackie A Cassell

• Contributors ECT, JTG, JRL, MES and JS designed the survey and ECT collected the data. ECT, JTG and JJK wrote the manuscript with statistical analyses done by JJK. PG, MES, JS and JRL participated in the results interpretation and revised the final manuscript.

• Funding Doris Duke Charitable Foundation to Columbia University College of Physician and Surgeons (Grant #2010062, ECT), Infectious Diseases Society of America (ECT) and IMPACTA.

• Disclaimer Funding sources played no role in the study design, data collection, data analysis, data interpretation, writing of the manuscript or the decision to submit the paper for publication.

• Competing interests None declared.

• Ethics approval The IRB of Columbia University Medical Center; The Institutional Committee of Bioethics (or ‘IRB’) of Asociación Civil Impacta Salud y Educación.

• Provenance and peer review Not commissioned; externally peer reviewed.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.