Primarily because of the anonymity and accessibility of online spaces (websites, mobile apps), GBMSM have been early adopters of technologies designed for sexual partnering [15]. However, as promising as technology may be for HIV prevention and because Hispanic/Latino and black populations are more likely to own mobile phones and use mobile apps compared to white populations [16,17], HIV prevention technology also introduces often-overlooked methodological pitfalls such as low engagement of racial/ethnic minorities and recruitment bias. Historically, online HIV prevention work has had low representation of minority GBMSM [18-20]. There are two likely explanations for this disparity. First, black men constitute 13% of the US male population [21], which may account for the smaller proportion who complete online surveys. Second, online recruitment bias may occur, such that researchers may not be using targeted language or graphics, or recruiting from sites that cater to minority GBMSM [22].

Electronic health (eHealth) interventions have the potential to reach and engage GBMSM living with HIV from diverse racial/ethnic groups and socioeconomic statuses. These interventions are critical for risk-reduction efforts [23,24], can reach geographically dispersed men [12,25], and can adapt offline interventions [26]. However, few interventions have demonstrated efficacy in reducing HIV transmission from GBMSM who are HIV-positive. Recent eHealth and mobile health (mHealth) interventions for HIV-positive populations have focused on ART adherence [27,28], rather than sexual risk [29], and have been delivered through text messaging and on computers [30-33]. eHealth assessments of GBMSM living with HIV have been used to tailor provider-delivered interventions to decrease CAS with serodiscordant partners. Nevertheless, a recent provider-delivered eHealth intervention had low participation and retention, due mainly to factors such as time commitment and clinical setting [34]. Other eHealth interventions address HIV disclosure and condom use among GBMSM who are HIV-positive but are still in early stages of implementation (eg, Miranda et al [35]).

There is a great need for cost-effective HIV prevention strategies for HIV-positive populations, particularly GBMSM who constitute the majority of those living with HIV [36,37]. Even modest intervention effects can have a significant public health impact because the two most important factors that determine cost-effectiveness are (1) the HIV prevalence of the target population (preventing transmission from men who are HIV-positive rather than preventing acquisition by men who are HIV-negative) and (2) the cost per person reached [38]. Effective HIV prevention interventions that use digital media are also likely to be highly cost-effective because they can be easily replicated after development, require minimal staffing, and have unlimited geographic reach [31,38-40].

The aim of this paper is to describe the study protocol of a randomized controlled trial (RCT) evaluating the effectiveness of a video-based messaging intervention delivered online, by comparing intervention and attention control groups on reductions in serodiscordant CAS and increases in HIV disclosure to sex partners. Through our prior work [18,19,41], we have identified a risk-reduction intervention approach for GBMSM living with HIV. The Sex Positive!^[+] study is conducted over a 1-year period among 1500 US white, black, and Hispanic/Latino GBMSM who were virally unsuppressed at some point during the past year, or who report suboptimal adherence to ART, and report recent CAS with serodiscordant male partners. Sex Positive!^[+] encompasses many characteristics found to reduce risk among HIV-positive populations in that it is theory-driven, has intervention content focused on HIV transmission behaviors, uses videos that demonstrate risk reduction and health behaviors through modeling, is delivered in an intensive manner, and is delivered over a 1-year period [42].

Sex Positive!^[+] is a two-arm randomized controlled phase III clinical trial with a 1:1 allocation ratio.

The Institutional Review Board at Public Health Solutions in New York, NY, approved all study procedures. A waiver of documentation of written consent was obtained given the Internet-based research approach. Although Federal regulation requires that researchers obtain written informed consent for research on human subjects, under 45 CFR § 46.117(c) [43], written consent can be waived for research that involves minimal risk to participants and involves no procedures for which written consent is normally required outside of the research context. This research meets that criterion and we use an alternative approach where participants click a button signifying that they have read the informed consent page and agree to participate in the study. An advantage of online studies is that the consent form is available for the participant to review and/or print at any time. This strategy complies with the requirement of 45 CFR § 46.117(c) that participants are given a written statement describing the research and risks.

A Data and Safety Monitoring Board has been established to conduct semiannual reviews of study activities and to ensure participant safety, validity, and integrity of the data. The Board is comprised of experts, independent of the trial or funding agency, in RCTs, Internet research, Web design, and HIV-positive populations. Furthermore, a Certificate of Confidentiality has been obtained from the National Institute of Mental Health to help protect the privacy of HIV-positive participants enrolled in this health-related study.

The target sample is 1500 high-risk, virally unsuppressed or less than 90% ART-adherent, US white, black, and Hispanic/Latino GBMSM living with HIV. Individuals participating in any aspect of the study must (1) be biologically male, (2) be age 18 or older, (3) be able to read and respond in English, (4) reside within the United States or a US territory, (5) report CAS with any HIV-negative or unknown status (serodiscordant) male partners in the past 6 months, (6) identify as HIV-positive, (7) report a detectable viral load, not being on ART and not knowing their viral load in the past year, or an undetectable viral load but less than 90% ART-adherent in the past 30 days, (8) identify as white, black, or Hispanic/Latino, (9) be willing to participate in an online intervention study for 12 months, and (10) have a working email address and mobile phone for intervention follow-up. We use quota sampling and targeted recruitment to ensure balanced representation of white (n=500), black (n=500), and Hispanic/Latino (n=500) men. Further, we include the following black racial/ethnic categories: black, African American, Caribbean, African, or multiethnic black [44]. In addition, we use targeted recruitment to ensure that 20% of the sample are men between the ages of 18 and 29 years. This group is overrepresented in the current HIV epidemic, particularly young men of color, and is less likely to be adherent to HIV medications or in care [4]. Men who meet study criteria, consent, and register to participate are automatically randomized into one of the two study arms.

Men are identified for the study through social networking websites and gay-oriented sexual networking websites, dating websites, global positioning system (GPS)-based mobile phone apps that utilize targeted recruitment by city, race, and ethnicity, and online bulletin boards. By recruiting from different types of websites and mobile phone apps, we increase our chances of reaching a broader, more diverse pool of men with HIV. The goals of recruitment are to identify eligible participants online who are willing to participate in an online, longitudinal HIV risk-reduction intervention. Based on previous research findings [22], study banner advertisements mirror the racial/ethnic composition of each study subgroup. One of our sources of recruitment is POZ Personals, the dating site for POZ Magazine, which distributes internal system messages to a defined subset of US male members who are HIV-positive, at least 18 years of age, and self-identify as gay or bisexual.

Both HIV status disclosure to sex partners and serodiscordant CAS are assessed at each of five survey time points (baseline, 3-, 6-, 9-, and 12-month follow-up). The recall period for the primary outcome measures is the past 3 months for each of the five online assessments (Table 1).

Secondary outcomes assessed at each time point include self-reported adherence to HIV medications, viral load, and CD4 count. Syndemic factors, or co-occurring epidemics, thought to be related to HIV transmission risk will also be assessed at various survey time points, including drug and alcohol use [45,46], depressive and anxiety symptoms [47-49], condom use and HIV disclosure self-efficacy [50], sexual compulsivity [51], HIV stigma, and interpersonal violence [52-54]. Process measures are used to track participants’ interest in, acceptability of, and verification of video viewing by having men complete brief postvideo online surveys. These surveys are designed to assess participants’ likes and dislikes as well as to elicit critical thinking about the video as it pertains to study outcomes [19,55].

Based on the prevalence of behaviors in our previous studies, we calculated true proportions and sample sizes using chi-square tests for this two-arm design. We estimated that by enrolling approximately 750 men per group and retaining 75% at 12-month follow-up, we would have 80% power at a 5% alpha level to detect a minimum reduction of 8% in the number of serodiscordant CAS partners between the intervention and attention control arms.

Video messages are an effective way to deliver HIV prevention to GBMSM [19,56-58]. The first of our three theoretically grounded HIV prevention videos (from the HIV Big Deal project) tackling issues of CAS, HIV disclosure, and testing was rigorously evaluated among HIV-negative, HIV-positive, and untested GBMSM recruited online [19]. In our single-session video pilot for 971 GBMSM, we found significant reductions in CAS in the most recent encounter (9% decrease) and significant increases in HIV disclosure at 3-month follow-up (13% increase) compared to baseline [19]. In our subsequent online, single-session RCT for 3092 GBMSM that used videos from the HIV Big Deal project, we found significant reductions in CAS among men in the video study arm at 60-day follow-up (8% decrease) compared to baseline; HIV-positive men in the video study arm reduced their CAS (14% decrease), including with HIV-negative or status unknown partners, at 60-day follow-up (13% decrease) compared to baseline. Men living with HIV were also significantly more likely than men who were HIV-negative or untested to complete follow-up (57% vs 51%, P=.002) [18].

An advantage of online research is data validity for sensitive information. Indeed, a growing number of studies indicate higher reporting of sexual risk and substance-using behaviors with computer-based surveys compared to mail, phone, and in-person surveys [82-84]. However, compared to the gold standard of in-person interviewing, a limitation of online research—as with mail and phone surveys—is the challenge of verifying a participant’s identity [85]. Based on recommendations made during an open session of our Data and Safety Monitoring Board, as well as a meeting of Internet experts about the issue of online fraud in research studies [86], the Sex Positive!^[+] study implements several protections aimed at reducing the likelihood of fraud. Specifically, (1) contact information is verified during registration using multiple methods; (2) duplicate detection (of Internet Protocol [IP] address and mailing address) software is used to detect instances of participants attempting to create multiple study profiles; (3) mobile phone numbers are investigated to determine whether they are voice over IP (VoIP) numbers—an individual can obtain multiple VoIP numbers on the Internet, typically for no charge, that can be routed to the individual’s mobile device (eg, this helps a potential participant to sign up numerous times with unique phone numbers); (4) proxy IP addresses and invalid mailing addresses are flagged for further scrutiny by the research team; (5) trap questions are used in the baseline survey to flag cases with inconsistent data—“I am HIV-negative”—or careless responses (eg, straightlining); (6) study staff conduct weekly analyses of new screener data to identify individuals (by IP address) who make multiple attempts to join the study; (7) compensation is kept sufficiently low to reduce the chances of participating solely to gain incentive payments; and (8) to ensure participant authenticity, the final study incentive is mailed to a verified physical address.

Primary analyses for intervention efficacy will examine whether participants in the intervention arm report fewer CAS acts with serodiscordant partners, a higher percentage of anal sex acts with condoms, fewer sexual partners, and more disclosure of HIV status with partners compared to participants in the attention control arm. Dose-response analyses will allow us to examine whether a certain number of intervention videos—“doses”—are necessary to effect study goals.

Even with a robust retention plan, incentives, and survey programming that requires responses, missing data are inevitable because participants can “refuse” survey items or drop out of the study. For this study, data from the online screener will verify that blocked randomization produced equivalent groups and will also be used to assess possible sample attrition bias. Although there are many ways to handle missing data, our experience suggests that maximum likelihood estimation is the best approach, using the appropriate algorithm for estimation purposes.

We will also conduct analyses to assess the savings in averted HIV-related lifetime treatment costs, the total number of quality-adjusted life years (QALYs) saved by preventing a single HIV infection, and the cost of developing and implementing the Sex Positive!^[+]study. In general terms, after the 12-month follow-up we will analyze up to three partner-by-partner sexual encounters, for biological male partners only, as well as global number of condomless anal or vaginal sex partners (ie, biological male and female partners, transidentified females, and transidentified males) at each of the four follow-up assessment time points. This will provide an estimate of the number of secondary infections expected among the participant’s serodiscordant sex partners. The main analysis will assume that study participants with an undetectable viral load at a particular time point are noninfectious. Although a separate estimate will be calculated for each participant at each of the four follow-up time points, the sum of these time point findings will estimate the total number of secondary infections expected among the participant’s sex partners during the entire study period. We will then calculate the total number of estimated infections prevented by the intervention arm versus the control by measuring the difference in the mean number of expected secondary infections for all men in the intervention and control arms over the 12-month period. Based on the total number of prevented infections, we will calculate the corresponding savings in averted HIV-related lifetime treatment costs as well as the total number of QALYs saved by preventing a single HIV infection. Finally, we will calculate the net cost of the intervention, the cost per infection prevented, and the net cost per QALY saved (ie, the cost-utility ratio) [87]. The intervention can be considered “cost saving” if the net cost is negative and “cost-effective” if the cost-utility ratio is less than US $100,000 per QALY saved [88].

This study protocol has several limitations that deserve mention. All men were recruited online, through social networking and gay-oriented sexual networking websites and mobile phone apps. As such, the findings may not be generalizable to HIV-positive GBMSM who do not own a mobile phone or have Internet access, access these types of websites or mobile phone apps, to men who do not identify as gay, to individuals exposed to a study banner or email but choose not to click on it, or to men who do not identify as black, white, or Hispanic/Latino. Study content is only available in English, which limits its reach to participants. There is a need to translate content into Spanish because it is the second most-spoken language in the United States and represents a subpopulation of GBMSM with high rates of HIV [4,89]. Lastly, a potential limitation is that participants self-report their health outcomes, specifically viral load. However, a recent validation study of 639 individuals with HIV from an ongoing prospective study in New York found that participant recall of viral load agreed with the Department of Health’s registry data 85% of the time [90].

In conclusion, the Sex Positive!^[+] study addresses the lack of interventions designed for GBMSM living with HIV. This protocol describes the underlying theoretical framework and measures, study design, recruitment challenges, and antifraud measures. If efficacious, it will have a significant impact on reducing HIV transmission risk in a disproportionately affected population. Although this eHealth intervention is being implemented with virally unsuppressed men or men who struggle with ART adherence, it can be adapted for delivery in other settings and with other populations.