Online interventions to address HIV and other sexually transmitted and blood‐borne infections among young gay, bisexual and other men who have sex with men: a systematic review

Abstract Introduction Globally, young gay, bisexual and other men who have sex with men (gbMSM) continue to experience disproportionately high rates of HIV and other sexually transmitted and blood‐borne infections (STBBIs). As such, there are strong public health imperatives to evaluate innovative prevention, treatment and care interventions, including online interventions. This study reviewed and assessed the status of published research (e.g. effectiveness; acceptability; differential effects across subgroups) involving online interventions that address HIV/STBBIs among young gbMSM. Methods We searched Medline, Embase, PsycINFO, CINAHL, and Google Scholar to identify relevant English‐language publications from inception to November 2016. Studies that assessed an online intervention regarding the prevention, care, or treatment of HIV/STBBIs were included. Studies with <50% gbMSM or with a mean age ≥30 years were excluded. Results Of the 3465 articles screened, 17 studies met inclusion criteria. Sixteen studies assessed interventions at the “proof‐of‐concept” phase, while one study assessed an intervention in the dissemination phase. All of the studies focused on behavioural or knowledge outcomes at the individual level (e.g. condom use, testing behaviour), and all but one reported a statistically significant effect on ≥1 primary outcomes. Twelve studies described theory‐based interventions. Twelve were conducted in the United States, with study samples focusing mainly on White, African‐American and/or Latino populations; the remaining were conducted in Hong Kong, Peru, China, and Thailand. Thirteen studies included gay and bisexual men; four studies did not assess sexual identity. Two studies reported including both HIV+ and HIV− participants, and all but one study included one or more measure of socio‐economic status. Few studies reported on the differential intervention effects by socio‐economic status, sexual identity, race or serostatus. Conclusion While online interventions show promise at addressing HIV/STBBI among young gbMSM, to date, little emphasis has been placed on assessing: (i) potential differential effects of interventions across subgroups of young gbMSM; (ii) effectiveness studies of interventions in the dissemination phase; and (iii) on some “key” populations of young gbMSM (e.g. those who are: transgender, from low‐income settings and/or HIV positive). Future research that unpacks the potentially distinctive experiences of particular subgroups with “real world” interventions is needed.

subgroups of young gbMSM (e.g. those who are: economically deprived; ethno-racial minorities; living in regions with discriminatory policies and/or political and cultural influences) [6]. At the same time, the increasing incidence among young gbMSM of viral (e.g. syphilis, Hepatitis C) and bacterial infections (e.g. gonorrhoea) [7,8] further signals the importance of developing new and innovative intervention approaches to meet the needs of today's generation of young gbMSM. As such, there are strong public health imperatives to identify effective prevention, treatment and care interventions that address HIV and other STBBIs among young gbMSM [2].
The Internet provides a medium to address the prevention, care, and/or treatment of HIV/STBBIs [9][10][11], particularly among youth 1 less than 30 years of agea highly "connected" generation [12], including among young gbMSM [13,14]. As such, many new online health promotion interventions have emerged during key transitional periods in the life course of the current generation of young gbMSM (e.g. as they move from childhood into adolescence or early adulthood). Previous research has illustrated how online interventions can change both mediators of safer sex (e.g. knowledge about sexual health, self-efficacy), in addition to behavioural (e.g. condom use, testing) and biomedical outcomes (e.g. incident infections) [15,16]. Online approaches to intervention are also considered scalable and cost-effective and may provide opportunities to overcome challenges with delivering HIV/STBBI interventions to "hidden" or "hard-to-reach" populations who may not otherwise access in-person programmes [9], including young gbMSM [17]. Moreover, global access to the Internet via a variety of devices (e.g. mobile phones, smartphones, notebooks, desktop computers, and tablets) is widespread, particularly among youth <30 years, including within many low-, middle-and high-income settings [18][19][20]. Policy makers and intervention strategists are also increasingly aware that the Internet provides opportunities to meet young people "where they are at, " including via social and sexual networking applications ("apps") which are often widely used by young gbMSM. For instance, mobile apps like Grindr, Scruff, and Tinder have millions of gbMSM users active across most areas of the globe [21], with a recent systematic review identifying that the majority of gbMSM using geosocial networking apps are ≤30 years of age [13]. As such, while web-based technologies may facilitate sexual risk behaviour among young gbMSM (e.g. "low-threshold" access to multiple and concurrent partners), they also provide innovative and promising opportunities to provide the right intervention to the right groups of gbMSM at the right time [22].
While online approaches have shown promise in providing sexual health promotion and care to young people, less is known about how online interventions can address the prevention, care, and/or treatment of HIV/STBBIs among young gbMSM, and reviews of online approaches to address HIV/ STBBIs have been notably absent [23]. This article provides a comprehensive review of the literature of online interventions that aim to address HIV and other STBBIs among young gbMSM by answering two primary research questions: (i) What is the status of research (e.g. effectiveness; acceptability) involving online interventions to address HIV/STBBIs among young gbMSM?; and (ii) What are the differential intervention effects according to intervention type (e.g. behavioural, biomedical, structural), social positioning (e.g. by SES; sexual identity) and research design? By answering these research questions through a systematic review of the peer-reviewed literature, our aim is to identify effective intervention strategies and to inform a renewed research agenda regarding the development of evidence-based online interventions for young gbMSM.

| METHODS
The research questions, outcome measures, search strategy, study selection process, and data analysis plan were based on an internal unpublished protocol developed prior to the initiation of the activities involved in this review process.

| Search strategy
Following the Systematic Reviews and Meta-Analyses (PRISMA) checklist [24] (see Appendix S1), we searched for studies related to online STBBI/HIV prevention and care among young gbMSM that were in English and published in a peer-reviewed journal in the following databases from inception through 15 November 2016: Medline, PsycINFO, CINAHL, EMBASE, and Google Scholar (the first 300 hits) [25]. Search terms were combined using appropriate Boolean operators and included subject heading terms or key words for four key themes and were tailored to fit each database requirements: men who have sex with men (e.g. homosexuality OR bisexuality OR men who have sex with men OR gay men OR MSM) AND HIV/STI (e.g. HIV OR AIDS OR STI/STD OR gonorrhea OR syphilis OR chlamydia OR herpes OR hepatitis) AND intervention (e.g. prevention OR intervention OR programme OR implementation OR evaluation) AND online (e.g. Internetbased OR web-based OR online OR e-health). Hand searches of the bibliographies of relevant published works and previous reviews were also performed. Our full electronic search strategy is included as a supplemental file.

| Eligibility criteria
The population, interventions, comparisons, outcomes and study designs considered for review are listed in Table 1. Studies were only included if they had provided post-intervention results.

| Data extraction, analysis, and quality assessment
Titles and abstracts of retrieved articles were screened to identify studies that potentially met our inclusion criteria. Full texts of all potentially eligible articles were retrieved by coauthor MK and independently assessed for full inclusion criteria by two review authors (MK and RK). Disagreement or uncertainty between the review authors was resolved through further discussion at weekly team meetings. Each study included was coded by two reviewers for study characteristics (e.g. study date and location), participant characteristics (e.g. target population, age, ethnicity), intervention characteristics (e.g. components, delivery method, duration, setting, theoretical framework), and outcomes (e.g. outcomes measured, main findings). Extracted data were summarized across included studies with respect to: participants and characteristics of studies; interventions and effects; and differential effects in outcomes across participant subgroups.
Risk of bias was assessed using the Cochrane risk of bias instrument for randomized controlled trials (RCTs) [26] and the modified Newcastle Ottawa scale for non-randomized studies [27]. For RCTs, studies were examined for selection bias, performance bias, detection bias, attrition bias, reporting bias, and other potential sources of bias. RCTs were considered at high risk of bias when at least one item was assessed as high risk of bias. For non-randomized studies, evaluations were made for selection bias, comparability, and outcome assessment.

| Interventions and effects
All but one study [30] reported a statistically significant effect on one or more outcomes. Of the 17 articles, all focused on behavioural and/or knowledge outcomes at the individual level in order to address HIV/STBBIs in the following intervention categories: [1] reduction of risky sexual behaviours (e.g. condomless sex) via knowledge acquisition and/or attitude change; and [2] testing promotion interventions. One trial [33] assessed an existing "live" intervention in dissemination phase (a website called healthMpowerment.org); the remaining 16 were at the "proof-of-concept" stage (i.e. at a stage seeking to determine whether an intervention is sufficiently promising to develop and scale). Two reported using tailored interventions (e.g. interventions with the capacity to refine to the level of the individual user) [33,38], while the remaining used targeted approaches (i.e. focused at the group level, such as at "MSM" Interventions included Internet-enabled apps, webpages and/or social media. This also included interventions that users could use on Internet-enabled devices such as mobile smartphones, handheld tablet computers (e.g. iPads), laptops and/or desktop computers. We did not include mHealth (i.e. mobile-based) interventions that did not feature an Internet-based component for the end-user (e.g. SMS text messaging interventions).
or "gay men"). Further details on each intervention, including study limitation and main findings, are reported in Table 5 Seven RCTs aimed to change knowledge, attitudes, or behaviours, and included an assessment of behaviour change as a primary or secondary outcomes. Bowen et al. 2008 [29] conducted an RCT in which they offered six modules of various scenario content (including on topics of HIV prevention, "contexts" of risk and experiences with new and casual partners). They reported a statistically significant change in knowledge, self-efficacy, and motivation to engage in risk-reduction practices, in addition to reduced anal sex and significant increases in condom use. Lau et al. 2008 [30] provided an experimental group with bi-weekly "visually appealing and professionally designed, educational, email graphical messages" on the topics of HIV/STD prevention, including HIV transmission, correct condom use, HIV testing, relationship and love, and the relationship between drugs and sex. The authors did not find a significant change in risk behaviour and perceptions following intervention and warned that the effectiveness of online interventions should not be taken for granted. Carpenter et al. [32] provided an experimental group with a 90-minute motivational, informational and skills training modules, including interactive materials, multimedia presentation and didactic text, followed by an opportunity to test their knowledge about HIV risk and learn up-to-date information. Findings indicated reductions among the experimental group with risky sexual practices with those with the "riskiest" sexual partners (those who reported having partners who were either seropositive or of an "unknown" status), not including    Table 3. Quality assessment of non-randomized studies using the modified Newcastle Ottawa Scale Author ( Studies were considered high quality if they scored above median (i.e., four points). Low rates of participation by minorities and those of lower socio-economic status; The brevity and short-term nature of the intervention; Intervention was delivered remotely (i.e. there is no reliable information about "dosage" and no assurance that the intervention was completed as intended) Table 5.

(Continued)
Author (   condomless receptive anal intercourse. Christenson et al. [34] provided a simulation using an avatar that is designed to reduce shame associated with sexual stigma among MSM by allowing participants to view their sexual desires as being "normal. " The study reported reductions in self-reported feelings of shame among the experimental group, though there was not a direct effect to reductions in risky sexual behaviour at follow-up. Mustanski et al. [35] offered a total of seven online learning modules to participants on a variety of topics that were designed for young MSM upon receiving an HIVnegative test. Compared to the control group, participants in the experimental arm had a lower rate of unprotected anal sex acts at follow-up. Lau et al. [43] assessed two fear appeal approaches to video-based interventions (five to ten minutes) one related to enhancing fear of health implications of contracting an STI and the other related to enhancing fear of the social losses associated with contracting an STI. The study did not find a statistically significant difference in unprotected anal intercourse among the intervention and that of a "factual" text-based website control. Hightow-Weidman et al. [33] conducted an RCT that provided the experimental group with access to an existing online informational website called Mpowerment; key features of the site included tailored live chats with an HIV expert, interactive quizzes, "hook-up/sex" journals, and HIV/STBBI risk assessment tools. The study reported changes in intention to use condoms and engage in preparatory condom use behaviours, though this was not a condition of the intervention (i.e. both control and experimental groups experienced this); no change in risk behaviour was reported.
The remaining two RCTs aimed to change knowledge and/ or attitudes but did not include an assessment of behaviour change following the intervention. Bowen et al. [28] conducted an RCT that reported a statistically significant improvement in HIV/AIDS-related knowledge and safer sex attitudes after delivering two online scenario-based modules. The intervention consisted of approximately 20-minute module regarding various scenarios about risk behaviour and an "inexperienced" man's experiences with a risky sexual encounter and the potential of having become infected. Mustanski et al. [37] conducted an RCT to measure the effects of HIV prevention messaging videos about multiple biomedical and behavioural HIV prevention methods (including nPEP, PrEP, rectal microbicides, and condoms) and MSM's intentions to use these strategies. The study found that the number of prevention messages did not produce differential attitudes and intentions regarding condoms; however, receiving multiple messages at once was associated with greater intentions to use PrEP and nPEP, but not rectal microbicides.

| Testing promotion interventions
Three RCTs assessed testing behaviour change following an online intervention. Blas et al. [31] randomized MSM to receive either a traditional public health text-based intervention (control group) or a five-minute video-based HIV testing promotion video (experimental group) targeted to either (i) gay, or (ii) non-gay MSM. They reported a statistically significant increase in intention to get tested among non-gay identified MSM following intervention, as well as following through to do so. Young et al. [41] conducted an RCT with peer-leaders creating a Facebook group and inviting participants to join and encouraging them to test throughout the duration of the 12-week study. Those in the intervention were more likely to test than those in the control. Bauermeister et al. [38] tailored the content of an online intervention based on the experimental group's socio-demographic data, including age, race/ethnicity, sexual identity, relationship status, testing history, sexual behaviour and structural barriers (e.g. Black MSM saw pictures of Black men). While testing practices were higher among the intervention group, this was not statistically significant; however, the difference was clinically meaningful with Cohen's d = 0.34, leading the authors to suggest preliminary efficacy.
Two non-randomized studies assessed testing behaviour following the intervention. Huang et al. [42] recruited participants from Grindr to receive a self-test kit (either via a pharmacy rebate code, via the mail or through a vending machine at a local LGBTQ centre). The study found that social network advertising that links users to a self-test was successful. Solorio et al. [44] conducted a multi-media campaign that included social media outreach and web-based informational pages to encourage testing among Latino MSM. The study found a significant impact on testing behaviour.
The majority of studies did not report on or examine differential effects by sexual identity. However, two studies comprised a research design that was specifically designed to assess intervention effect by sexual identity. An RCT based in Peru by Blas et al. [31] split their study sample between MSM that were "gay-identified" (including those who identify as "gay" or "caleta"i.e. those who identified as "closeted" or "semi-closeted") and "non-gay-identified" (including those who identify as heterosexual, bisexual or "flete"i.e. young male prostitutes). Each cluster was split and therefore received both the control and experimental conditions; the intervention a five-minute HIV testing health promotion videowas targeted towards either gay-or non-gay-identified MSM. The authors found a significant difference only among non-gay identified MSM, thereby theorising that the non-gay-identified men were more receptive to interventions that promote HIV testing.
A non-randomized study by Solorio et al. [44] hypothesized that non-gay-identified Latino men in Seattle, Washington, would be less likely to be responsive to campaigns that were not targeted specifically towards the Spanish-speaking Latino community. As such, a multi-media campaign was targeted to Spanish-speaking MSM who do not identify as gay and also for those who identify as gay. The study reported that MSM who did not identify as gay were just as likely to seek HIV testing following the intervention as those who identified as gay. The authors considered these results successful as nongay-identified Latino MSM represented a "difficult-to-reach" population.

| Gender identity
Blas et al. [31] reported including both transgender-and cisgender-identified MSM. This study included an intervention arm that was specifically designed for transgender MSM; however, they lacked sufficient power to evaluate the transgender arm (n = 21) and the transgender group was therefore excluded from the analysis. Mustanski et al. [40] reported that they included four transgender men; no sub-analysis was conducted by gender identity. Bauermeister et al. [38] reported that transgender men were excluded from their study. The remaining 14 studies did not report the gender identities of participants.
Two studies reported including HIV-positive participants. An RCT by Hightow-Weidman et al. [33] reported they randomized more HIV-positive participants into their control group. Unfortunately, due to a low sample size, they were unable to account for serostatus in their analysis. As such, the authors suggest that they may have been unable to detect behaviour change resulting from the intervention because the HIV-positive participants were likely engaging in condomless sex with sero-concordant partners. These authors suggest future studies should account for differential effects of HIV status both through statistical controls and a stratified randomized design in order to ensure serostatus differences are sufficiently powered to assess both study conditions (i.e. intervention and control). Bauermeister et al. [38] also assessed and reported on HIV serostatus; their study sample included four (3.0% of the sample) who were HIV positive in their HIV/STBBI testing promotion intervention. While they did not assess the differential effects of the intervention by serostatus, they did indicate that all four HIV-positive participants reported seeing a HIV/STI provider in the past 30 days.

| Ethno-racial characteristics
Among the studies from Asia, one from China [43] and one from Hong Kong reported sample [30] compositions that were entirely Chinese and one study from Thailand reported an entirely Thai sample [36]. One study from Peru did not report ethno-racial identity [31] while another reported a mixed sample composition (mixed, White and Black) [41]. One study from the US specifically focused on recruiting Black MSM [33], another Latino MSM [44] and a third both Black and Latino MSM [42], as these populations were specifically identified as being at an elevated risk for HIV/STBBIs. One study [28] from the US reported on ethno-racial identity as being either "White" or "non-White. " The remaining eight studies [29,32,34,35,[37][38][39][40] from the US included at least four or more measures for ethno-racial identities, including (in order of most frequently used to least) White, Latino or Hispanic, Black or African American, "other, " Middle Eastern, Native American, Asian Pacific and Hawaiian Pacific Islander. None of the studies reported on differential effects of interventions by ethno-racial identity.

| DISCUSSION
Our systematic review of online approaches to address the prevention, care, and/or treatment of HIV/STBBIs among young gbMSM included 12 RCTs and five non-randomized studies. Sixteen of the studies in our review were "proof-ofconcept" efficacy trials of interventions not specifically designed for further dissemination; accordingly, the sample sizes were generally small (median: 130 participants). One study assessed a "live, " "real-world" intervention. All of the studies focused on behavioural or knowledge outcomes at the individual level (e.g. condom use, testing behaviour, and knowledge and/or attitudes about HIV/STBBI risk), and all but one reported a statistically significant effect on one or more primary outcomes. Twelve studies described theory-based interventions. Twelve were conducted in the United States, with study samples focusing mainly on White, African-American and/or Latino populations; the remaining were conducted in Hong Kong, Peru, China, and Thailand. Thirteen studies included gay and bisexual men; four studies did not assess sexual identity. Two studies reported including both HIV-positive and HIV-negative participants, and all but one study included one or more measure of SES (e.g. income, educational attainment). While most (n = 13) of the interventions included and reported upon measures of intervention acceptability, five of these did not provide details on how this was assessed; the remaining four did not report on intervention acceptability.
The statistically significant changes in one or more primary outcome in all but one of our included studies underscores the promise that online approaches have for addressing HIV/ STBBIs among young gbMSM. As such, our review supports previous research [9] suggesting that efforts to change behaviour at the population level may benefit from evidenceinformed online approaches. Nevertheless, as no trial had a low risk of bias for all quality criteria, the promising results need to be interpreted with caution and confirmed in further high-quality trials. Moreover, there were several limitations associated with the measurements used across the studies included in our review. First, all of the studies focused on behavioural interventions, with no studies assessing the efficacy or effectiveness of other kinds intervention types (e.g. biomedical, structural). Second, although behavioural outcomes (e.g. condom use, HIV testing) resonate with the National HIV/AIDS Strategy for the United States (NHAS) 2015 indicators for young MSM [72], study outcomes were measured and reported in highly variable and inconsistent ways across studies, making it difficult to compare findings across studies and precluding our ability to pool the results. Future studies should seek to use standardized measures whenever possible to assess the effects of online interventions on different outcomes.
The majority of research to date in this area is largely focused on "proof-of-concept" and/or "one-off" interventions that are not sustained beyond the completion of the study. Out of the 17 studies included in our review, only one [33] sought to assess an intervention that was within the dissemination phase (i.e. it was "live" and available to the public while the study was taking place). Moving beyond the "proof-of-concept" research phase into the dissemination of interventions in "real-world" conditions will benefit from including additional implementation measures during all phases of intervention research [73]. For example, online HIV/STBBI intervention research should be designed with a variety of implementationoriented considerations in mind to systematically identify the factors that will influence intervention scalability (e.g. equitable reach; rate of uptake) among key groups of young gbMSM within and across a variety of settings [16].
We also suggest that future intervention research in this area will benefit from enhanced efforts to assess the effects of various "real-world" and "live" interventions, (e.g. rather than focusing on "proof-of-concept" trials), including riskreduction interventions that have been developed and implemented from outside of the health or community-based sectors. For example, within the private technology sector, Grindr recently provided users with the option to disclose their serostatus, viral load (e.g. "undetectable") and/or use of PrEP on users' profile pages. Identifying the effects that these kinds of "real-world" interventions may have on the social and sexual health outcomes of young gbMSM is a critical "next-step" for intervention research in this area.
Most of the studies included one or more socio-demographic measures to describe the sample composition (e.g. by SES, sexual identity, serostatus). However, with a few notable exceptions, few reported on the differential intervention effects by SES, sexual or gender identity, ethno-racial characteristics or HIV serostatus, often due to small sample sizes and sample compositions that were too homogenous. We suggest that future research regarding online sexual health interventions undertake differential analyses, particularly in light of growing evidence that suggests individually-oriented interventions tend to (re)produce inequalities in health [74,75]. We agree with Hightow-Leidman et al. [33] suggestion that future research designs shouldwhenever possibleseek to account for differential effects through the use of statistical controls and/or stratified randomized designs, or stepped-wedge designs for those interventions in the dissemination phase. This may also require additional measures to effectively ensure key differences (e.g. by serostatus, sexual identity, SES) are sufficiently powered to assess differential effects (e.g. within and across study conditions).
We were surprised that few studies focused on evaluating interventions among several "key" populations of young gbMSM, including those living with HIV, transgender gbMSM and those living in low-income settings. For example, given the ongoing shifts in the field of HIV emphasizing the treatment and prevention needs for those living with HIV, research assessing online HIV/STBBI interventions for HIV-positive populations seems notably absent [16]. Furthermore, out of the 4669 participants included in our review, only 25 participants were reported as identifying as transgender. Finally, to date, no studies assessed online approaches to addressing HIV/STBBI among young gbMSM in low-income settings. Future research in lowand middle-income settings is needed. Critically, it is important to emphasize that if key groups of MSM are not included in online intervention research, these groups are likely to be excluded from population-specific interventions to reduce HIV/ STBBI risk as they are scaled up to the population level [76].
There was also a tendency for the interventions in our review to be based on the premise that a "one-size-fits-all" approach to intervention delivery can work for all populations of gbMSM, regardless of serostatus, SES, and other circumstances. For example, we were surprised that most interventions employed "targeted" approaches (i.e. at a population's group characteristics), particularly given the public health science in this area that indicates that tailored (i.e. at an individual's characteristics) web-delivered behaviour change intervention are significantly more effective than non-tailored websites in achieving behavioural outcomes [77]. Indeed, only Hightow-Leidman et al. [33] and Bauermeister et al. [38] reported tailoring content and user experience based on userspecific data profiles (e.g. based on age, ethnicity, sexual identity). Given that interventions that tailor approaches to an individual's specific "profile" (e.g. based on various features of their social positioning) tend to better capture a user's attention, contain less redundant information and overall be more acceptable among users [77], future online intervention development in this area may benefit from developing sophisticated approaches to tailored web-based service delivery systems.

| Limitations
Our reviewthe first review focusing on online interventions regarding the prevention, care, and/or treatment of HIV/STBBIs among young gbMSMhas several strengths and limitations. First, a limitation of our review is that because of the high-level of heterogeneity between the different risk-reduction and testing promotion interventions and measured outcomes, it is not feasible to calculate the pooled effects of the interventions included in our review. Second, while our approach to searching the literature was comprehensive and employed a robust set of search strategies, including the use of multiple databases, potentially relevant studies that are reported in other domains (e.g. technical reports in the grey literature; non-English peer-reviewed publications) are not accounted for in our review. Third, because our inclusion criteria required a sample of at least 50% gbMSM under 30 years of age, studies that may have reported relevant findings about online intervention with gbMSM are excluded ; future analyses of these studies will provide important details about the experiences of older generations of gbMSM with online interventions. Nevertheless, these findings provide a key "first step" in informing both potentially effective strategies and a renewed research agenda regarding the development of evidence-based online interventions to address HIV/STBBIs among young gbMSM.

| CONCLUSIONS
On the basis of our findings, we support a call for more rigor and attention within the creation of study designs that have the capacity to report differential effects within and across population sub-groups in intervention research [74] in order to unpack the potentially distinctive experiences of particular subgroups of young gbMSM (e.g. lower vs. higher income gbMSM). We also urge researchers in this area to identify the effects of "realworld, " "live" interventions, including online sexual health programmes that provide a service to the public (e.g. online testing platforms) and/or interventions that are put forth from the private technology sector (e.g. risk-reduction strategies that are programmed within social and sexual networking apps). Finally, future research must also assess intervention effects among young gbMSM who are transgender, living in low-income settings and/or who are living with HIV.

A U T H O R S ' C O N T R I B U T I O N S
All co-authors contributed to the research questions. MK offered technical expertise in developing our search strategy. RK and MK extracted data and conducted the data analysis. RK drafted the first version of the manuscript and received extensive feedback from all co-authors.

A C K N O W L E D G E M E N T S
We are thankful to Kyle Sutherland and Michelle Pang for assisting with manuscript preparation activities. We are also thankful to all of the participants from a Summit we hosted with our partners YouthCO HIV/Hep C Society and the British Columbia Centre for Disease Control in September 2016 in which we discussed analytic questions together for this systematic review. Note 1 We define youth as being <30 years, in accordance with emerging trends in the operationalization of young adult age groups (e.g. European Union; Millennial Generation), accounting for a set of secular trends, including delayed transitions into "adulthood" (e.g. leaving parental home; achieving financial independence).