A web of risk: multilevel factors and feedback loops (re)produce HIV ‘risk’ among gay, bisexual and other men who have sex with men – a global systematic review

Abstract

Background

HIV literature shows that gay, bisexual and men who have sex with other men (GBMSM) experience inequities across social and contextual factors. Given growing inequities, this study used complex systems theory, a scientific approach to understanding the interconnected parts, to identify and visualise the system of factors that shape the emergence or (re)production of HIV risk among GBMSM.

Methods

A meta-synthesis of systematic reviews and meta-analyses was conducted to examine risk factors for HIV in alignment with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses criteria and quality assessments using A Measurement Tool to Assess Systematic Reviews 2. After screening 255 studies, data were synthesised and visualised from 29 articles with moderate-quality or high-quality assessments. Study characteristics and risk factors for HIV were extracted, and data were thematically analysed into higher-order themes and respective subthemes aligned with Bronfenbrenner’s socio-ecological model. Kumu.io, a system mapping software, was used to visualise the system of factors.

Results

Our thematic analysis and visualisation portray a dynamic and complex web of HIV risk that GBMSM experience implicated across all levels of the socio-ecological model: individual, interpersonal, community, institutional/organisational and structural/policy levels. These risk factors, in tandem, interact with one another to create pathways and patterns that generate feedback loops, such that the systems of factors create the emergence of GBMSM’s HIV risk beyond that accounted for at the individual level.

Conclusion

GBMSM’s HIV risk is socially patterned by a diversity of multilevel and interacting risk factors, which creates a dynamic and reinforcing system of HIV risk that requires attention in its totality to fully address HIV risk.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• HIV risk arises from a diverse set of factors, including sexual behaviour, access to biomedical HIV prevention, substance use and numerous other socioeconomic and political factors, but little is understood about their interconnections.

WHAT THIS STUDY ADDS

• This study found a dynamic and diverse set of factors that interact to (re)produce the HIV risk that gay, bisexual and men who have sex with men experience globally.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• HIV risk emerges from a web of risk factors that operate across all levels of the socio-ecological model, thus requiring a focus on creating a risk-reducing environment; however, research on structural and social drivers is critically under-represented in the HIV scientific literature.

Introduction

The median global HIV prevalence rate is 0.8%; however, it is 7.7% among gay, bisexual and men who have sex with men (GBMSM).¹ GBMSM are considered a ‘key population’ in HIV programming, policy and practice, given their high HIV risk.^{2 3} Risk, defined here as likelihood of infection, can arise from numerous factors. Consistent and correct use of condoms can reduce risk by over 90%, proper use of pre-exposure prophylaxis (PrEP) indicates nearly 100% protection, and undetectable=untransmittable (U=U) shows that persons with suppressed HIV do not transmit during condomless anal sex.4,7 However, the science of HIV does not serve all communities equitably.

Greater attention to how the structural and social determinants of health shape GBMSM’s HIV ‘risk’ is critical. The socio-ecological model (SEM) posits that HIV risk is influenced by factors across structural (eg, policy), institutional (eg, norms and beliefs), community (eg, shared identity and values)¹, interpersonal (eg, sexual relationships and friendships) and individual levels (eg, risk perception and behaviour).⁸ For this paper, community is defined in a broad sense focused on a shared identity, behaviours or values of GBMSM globally, nationally and locally. In Europe, GBMSM living in more stigmatising country contexts had elevated probabilities of HIV diagnosis than GBMSM in countries with less stigmatisation, regardless of their behaviours.⁹ Another review indicated the negative impacts of structural stigma on HIV prevention.¹⁰ A global systematic review and meta-analysis indicated multilevel factors influence GBMSM’s PrEP use, which in turn would reduce risk.¹¹ Although the SEM recognises multilevel factors, the lack of specificity of relationships between risk factors is a challenge.¹²

The syndemic theory is also important when considering HIV. First described by Singer,¹³ it examines the interconnected and multiplicative effect of co-occurring social and health conditions among key populations that drive their HIV risk.¹³ Specifically, it acknowledges that co-occurring health conditions are exacerbated by harmful social conditions, including stigmatisation, marginalisation and violence that increases disease burden and disparities. Further syndemic studies have found that multiple health conditions cluster or co-occur with one another and that often these relationships have a multiplicative effect that increases or can reduce HIV risk if adequately addressed.14,16 However, a similar criticism is that it does not showcase the specific mechanisms or explain the interactions between the health conditions and their causes. In fact, Singer described in a 2003 article that “…the mechanisms of (unidirectional, bidirectional or dialectical) disease interplay, and how deleterious disease enhancement occurs—including assessment of the effects of malnutrition, stress, stigma, toxic exposure or other challenges to health on disease interactions—are important lines of investigation”.¹⁷

Complex systems theory can be practical in specifying relationships. In a complex system, HIV risk is considered an emergent property, defined as a collective property that arises from interactions, interdependencies, adaptations and feedback loops.¹⁸ Complex systems are interconnected items that (re)produce patterns of behaviour over time. It has three key factors: (1) elements (ie, GBMSM and HIV), (2) interconnections (ie, risk factors across GBMSM’s lived experiences) and (3) a function or purpose of the ‘system’ (ie, HIV’s goal is to replicate).¹⁹ Owing to the dynamic nature of complex systems, they offer a holistic understanding of the physiological, biological, biomedical and socio-ecological factors that (re)produce GBMSM’s HIV risk. Combining complex systems with evidence synthesis, the scientific study of interpreting studies in rigorous and transparent ways to support the systematic presentation of global knowledge, can offer insights into how HIV risk arises from complex interactions.

This systematic review of systematic reviews aims to offer a novel visualisation of HIV research that maps the convoluted system of factors that shape HIV risk among GBMSM. We aimed to (1) synthesise elements related to HIV risk, (2) thematically categorise the elements across levels of the SEM and (3) visualise the interactions among various risk factors, which shape the emergence of HIV ‘risk’ among GBMSM.

Materials and methods

We conducted a meta-synthesis of systematic reviews aligned with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria.²⁰ The following terms were used: HIV/AIDS, HIV, AIDS, gay, bisexual, men who have sex with men, male-to-male sexual intercourse, systematic review, scoping review and meta-analysis. The full MeSH search is provided in online supplemental table 1. This review relied on search terms and data from a previous review on HIV infection.²¹

All studies from the previous review (n=63) and those coded as having no HIV outcome (n=123) were migrated. We conducted an updated search for publications after April 2022. The final searches were completed in January 2023. We searched CINAHL (n=7), Global Health (n=14), PsycINFO (n=10), PubMed (n=44), Scopus (n=65), and Web of Science (n=16) databases (online supplemental table 1). We used Covidence, an evidence synthesis software, which systematised the review process²² and Zotero.²³ After manually removing duplicates, 342 studies were imported into Covidence, where another 87 duplicates were removed, leaving 255 for the title and abstract screening.

The inclusion criteria were as follows: (1) examined factors that influence HIV risks, (2) GBMSM in the population of study, (3) peer-reviewed systematic review or meta-analyses, (4) written in English, Spanish or Serbo-Croatian (language skills of the team) and (5) non-intervention studies (given the aetiologic purpose of this review). There were no limits on the timeframe or geography.

Three reviewers (KG, NP and KS) independently conducted the title and abstract screening in duplicate and the full-text review. At both stages, conflicts were resolved in meetings between reviewers. Fifty studies did not meet the inclusion criteria at the title and abstract screening, leaving 205 for full-text reviews. Reasons for exclusion were not tracked at the screening stage in line with PRISMA.²⁴ During the full-text review, 150 papers were excluded, of which a majority did not measure risk factors (n=75) and were intervention studies (n=38). Other reasons can be found in figure 1, which captures the review process.

Figure 1. PRISMA flow diagram of systematic review of reviews on HIV risk among gay, bisexual and other men who have sex with men globally. GBMSM, gay, bisexual and other men who have sex with men.

Analysis and conceptualising HIV risk

Three reviewers (KO, EK and NP) conducted data extraction, and KO, EK and KS conducted qualitative thematic analysis of the literature sample (n=55). The following information was extracted and organised into tables: (1) authors, (2) year of publication, (3) geographic area of the study, (4) populations under study and the sample size and (5) risk factors for HIV and their associated risks. Quality assessments of each review were conducted using the A Measurement Tool to Assess Systematic Reviews 2 guidelines for systematic reviews.²⁵ Our analysis only included studies that were of high and moderate quality (n=29) to ensure the quality of the results. Online supplemental table 3 presents each systematic review’s quality assessment and the extracted data and results. After the primary results were extracted, KS, KO and EK categorised them into higher-order themes and respective subthemes and identified the level of Bronfenbrenner’s SEM (tables 1–3, online supplemental table 4).²⁶

Table 1. Associations and level of SEM of factors associated with adherence, awareness, use and willingness to use medications that influence HIV risk.

Variable	Factors negatively associated with the variable	Citations	Factors positively associated with the variable	Citations
PrEP adherence
Individual	Substance use Fear of PrEP side effects Poor health status Inconvenience of PrEP	Ching et al29 Ching et al29 Ching et al29 Ching et al29
Interpersonal	Lack of support from partner	Ching et al29
Institutional	High cost of PrEP Inadequate healthcare provider training	Ching et al29
		Ching et al29
Structural	Social stigma	Ching et al29
PrEP use
Individual	Low perceived HIV risk	Russ et al30, Yuan et al31	Condom use	Yuan et al31
	Low perceived need	Russ et al30
	Medical mistrust	Russ et al30
	PrEP risk compensation	Russ et al30
	PrEP-related stigma	Russ et al30
	Low perceived efficacy of PrEP	Yuan et al31
	Lack of knowledge of PrEP	Yuan et al31
	Inconvenience of PrEP	Yuan et al31
	Concerns of side effects	Yuan et al31
Interpersonal	Interpersonal stigma	Russ et al30	Social support	Yuan et al31
	Poor familial relationships	Yuan et al31
Community	Lack of community support	Russ et al30
	AIDS discrimination	Yuan et al31
Institutional	High cost of PrEP	Yuan et al31
Structural	Social stigma	Yuan et al31
Willingness to use PrEP
Individual	Low perceived efficacy of PrEP	Sun et al32
	Concerns of side effects	Sun et al32
	Higher education	Sun et al32
	Higher number of sexual partners	Sun et al32
	Non-LGBT orientation	Sun et al32
	Stigma towards PrEP	Sun et al32
	Stigma towards PrEP users	Sun et al32
PEP awareness
Individual	Casual partners Closeted LGBT status Low education Unemployment	Jin et al33 Jin et al33 Jin et al33 Jin et al33	Chemsex Higher number of partners Knowledge of HIV status LGBT identity Meeting partners via Internet Previous STI Unprotected sex	Jin et al33 Jin et al33 Jin et al33 Jin et al33 Jin et al33 Jin et al33 Jin et al33
Community			Interaction with gay culture Lower levels of HIV stigma	Jin et al33 Jin et al33
Institutional			Access to healthcare	Jin et al33
PEP use
Individual	Concerns of side effects Denied housing Experience with abusive language Lack of knowledge of PEP	Jin et al33 Jin et al33 Jin et al33 Jin et al33	Circumcised Engagement in sexual risk behaviours In a relationship Substance use Chemsex	Jin et al33 Jin et al33 Jin et al33 Jin et al33 Maxwell et al34
Institutional	High cost of PEP	Jin et al33
nPEP use
Individual	Lack of knowledge of nPEP Low perceived HIV risk	Wang et al35 Wang et al35	Higher number of sex partners Inconsistent condom use Substance use	Wang et al35 Wang et al35 Wang et al35
Institutional	Low access to healthcare	Wang et al35
ART adherence
Individual	Poor body image Endorsing masculinity	Nowicki et al36 Zeglin et al37

ART, antiretroviral therapy; LGBT, lesbian, gay, bisexual, transgender+community; nPEP, non-occupational post-exposure prophylaxis; PEP, post-exposure prophylaxis; PrEP, pre-exposure prophylaxis; SEM, socio-ecological model; STI, sexually transmitted infection.

Given the large body of HIV research, we analysed reviews because they interrogate discrepant information in our scientific literature. We analysed the overlap across reviews to ensure that outcomes were not overstated (online supplemental table 2). Out of 1411 unique studies analysed in our sample, 75 appeared across multiple reviews. Six (out of 29) reviews had no overlap, whereas the remaining 23 reviews had at least one overlapping article.

We used Kumu.io, a systems mapping software, to visually represent the HIV risk environment for GBMSM.²⁷ Green arrows indicate a positive association, red represents a negative association and black means both positive and negative. The lines represent the predictor to outcome associations that were examined in the reviews. The complex systems model is dynamic despite its static representation within the text. The dynamic model, which respects the principles of complex systems, is available in the following hyperlink: https://embed.kumu.io/2a5f78e7531da8b57709d028ae96dce7#hiv

Results

Review characteristics

Our sample included 29 reviews (23 moderate and 6 high quality), encompassing 1411 unique studies (online supplemental table 3). All reviews, except one,²⁸ revealed individual-level risk factors. Six reviews examined interpersonal factors, six examined community factors, eight examined institutional factors and five examined the structural level.

Synthesised themes

After ‘connecting the dots’, the associations between HIV risk and the risk factors portray a dynamic nature of HIV risk (figure 2). HIV risk operates as an emergent property arising from complex interactions that involve all levels of the SEM, creating feedback loops such that the complex system (re)produces the emergence of HIV risk beyond just the individual level behaviours. The meta-synthesis identified four main themes spanning multiple levels of the SEM: (1) non-adherence, awareness, use and willingness to use medications (table 1); (2) sexual and sex-seeking behaviours and related factors (table 2); (3) non-usage of HIV prevention and testing services (table 3); and (4) factors contributing to additional HIV vulnerability (online supplemental table 4).

Figure 2. Complex systems visualisation of HIV risk among GBMSM globally. Square, HIV risk factor; circle, risk of risk factor; green arrow, positive association; red arrow, negative association; black arrow, both positive and negative association; yellow, individual-level factor; blue, interpersonal-level factor; orange, community-level factor; light green, institutional-level factor; purple, structural-level factor.

Table 2. Associations and level of SEM of factors associated with sexual and sex-seeking behaviours and related factors that influence HIV risk.

Risk behaviour	Factors negatively associated with behaviour	Citations	Factors positively associated with behaviour	Citations
Unprotected sex
Individual	Knowledge of partner’s HIV status Depression Self-efficacy Knowledge of HIV status Use of ART Poor body image	Crepaz et al45 Huebner et al46 Lacefield et al40 Malekinejad et al44 Malekinejad et al44 Nowicki et al36	Low perceived HIV risk High perceived efficacy of HAART Depression Alcohol use Substance use Negative attitudes towards condoms LGBT identity Sexual compulsivity Transactional sex Public sex/gay venues Higher number of sex partners Intentional non-use of condoms Meeting partners via Internet Sex with HIV+partner Chemsex	Crepaz et al38, Zou & Fan39, Lacefield et al40 Crepaz et al38 Heubner et al46 Lacefield et al40, Lewis et al41 Lacefield et al40, Lewis et al41 Lacefield et al40 Lacefield et al40 Lacefield et al40 Lacefield et al40 Lacefield et al40 Lacefield et al40 Lacefield et al40 Liau et al42, Yang et al43 Malekinejad et al44 Maxwell et al34
Interpersonal	Social support	Lacefield et al40
Condom use
Individual	Depression Higher education Higher income Low perceived risk of partner Older age PrEP use PrEP use not associated Public sex/gay venues Sex with HIV- partner Substance use Substance use not associated Younger age Lack of HIV status transparency	Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Freeborn et al47 Stephens et al48	PrEP use Knowledge of partner’s HIV status	Freeborn et al47 Malekinejad et al49
Sexual compulsivity
Individual			Depression Anxiety Alcohol use Substance use Intimate partner violence Childhood sexual abuse Unprotected sex Casual partners	Rooney et al50 Rooney et al50 Rooney et al50 Rooney et al50 Rooney et al50 Rooney et al50 Rooney et al50 Rooney et al50
Higher number of partners
Individual			PrEP use Substance use	Freeborn et al47 Moradi et al51
Casual partners
Individual			Chemsex	Maxwell et al34
Group sex
Individual			Transactional sex	Chow et al52
Chemsex
Individual			Transactional sex Presence of other STIs LGBT identity Lack of knowledge of services	Chow et al52 Maxwell et al34 Maxwell et al34 Maxwell et al34
Transactional sex
Individual	Lack of knowledge of PrEP	Russ et al30
Sexual risk behaviours (general)
Individual			PrEP use	Traeger et al53
Meeting partners via internet
Individual	Lower education Lower income	Zou and Fan39 Zou and Fan39

ART, antiretroviral therapy; HAART, highly active antiretroviral therapy; HIV+, human immunodeficiency virus (positive); HIV-, human immunodeficiency virus (negative); LGBT, lesbian, gay, bisexual, transgender+community; PrEP, pre-exposure prophylaxis; SEM, socio-ecological model; STI, sexually transmitted infection.

Table 3. Associations and level of SEM of factors associated with non-usage of HIV prevention and testing services.

Variable	Factors negatively associated with the variable	Citations	Factors positively associated with the variable	Citations
HIV testing/services
Individual	Previous negative test experience Low perceived HIV risk Low perceived test accuracy Fear of being outed Internalised stigma Fear of positive diagnosis Younger age Lack of knowledge of PrEP Lack of knowledge of services Endorsing masculinity	Kularadhan et al54 Kularadhan et al54, Magno et al55 Kularadhan et al54 Kularadhan et al54, Stephens et al48 Kularadhan et al54 Magno et al55, Stephens et al48 Magno et al55 Russ et al30 Stephens et al48 Zeglin et al37	Convenience of service Obligation to test LGBT identity	Kularadhan et al54 Kularadhan et al54 Maxwell et al34
Interpersonal	Stigma from healthcare providers	Stephens et al48
Community	Community stigma	Magno et al55
	Gender norms	Stephens et al48
Institutional	High cost of test	Kularadhan et al54	Peer-test facilitators	Kularadhan et al54
	Low access to services	Magno et al55
	LGBT stigma from religious organisations	Stephens et al48
Structural	Societal stigma	Kularadhan et al54, Stephens et al48
	Criminalisation of LGBT	Stephens et al48
	Societal-level LGBT stigma	Wang et al35
Access to healthcare
Institutional	Institutional stigma	Wilson et al28

LGBT, lesbian, gay, bisexual, transgender+community; PrEP, pre-exposure prophylaxis; SEM, socio-ecological model.

Non-adherence, awareness, use and willingness to use medications

Factors related to PrEP use, substance use, fear of PrEP side effects, poor patient health status, inconvenience of taking PrEP, lack of partner support, high cost, inadequate healthcare provider training and MSM social stigma were negatively associated with use²⁹ (table 1). In two reviews, low perceived HIV risk was negatively associated with PrEP use^{30 31} (table 1). Low perceived need of PrEP, medical mistrust, PrEP stigma and engaging in sexual risk behaviours while on PrEP were negatively associated factors in a review conducted in the USA.³⁰ Low perceived efficacy and knowledge of PrEP, inconvenience of taking PrEP and concerns about side effects were individual-level factors negatively associated with PrEP use in China,³¹ whereas condom use was positively associated.³¹ Interpersonal stigma and lack of community support were also negatively associated with PrEP use in the USA.³⁰ Poor familial relationships, AIDS discrimination, high PrEP cost and social stigma were negatively associated with PrEP use in China, whereas social support was positively associated with it.³¹

Willingness to use PrEP was examined in one global review investigating only individual-level factors. Some examples were low perceived PrEP efficacy, side effect concerns, higher patient education and higher number of sexual partners, which were negatively associated with willingness³² (table 1).

One global review examined PEP awareness and found casual partners, closeted LGBT status, low patient achieved education and unemployment were negatively associated with PEP awareness.³³ Chemsex, unprotected sex, higher number of partners, knowledge of HIV status, interaction with gay culture, lower levels of community-level HIV stigma and access to healthcare were positively associated with PEP awareness³³ (table 1).

PEP use was examined by two global reviews^{33 34} (table 1). One review found that concerns of side effects, being denied housing, experience with abusive language, lack of PEP knowledge and high cost of PEP were negatively associated with PEP use.³³ This review also found that being circumcised, engagement in sexual risk behaviours, substance use and being in a relationship were positively associated with PEP use. The second global review found chemsex to be positively associated with PEP use.³⁴

A global review observed factors related to a lack of nPEP knowledge and found that low perceived HIV risk was negatively associated with nPEP use, and a higher number of sex partners, inconsistent condom use and substance use were positively associated with nPEP use.³⁵ At the institutional level, low access to healthcare was negatively associated with it (table 1)³⁵.

Adherence to anti-retrovirals (ARVs) was examined in two studies. Poor body image showed a negative association with ART adherence in a global review,³⁶ and endorsing masculinity through sexual means also demonstrated a negative association in another review³⁷ (table 1).

Sexual and sex-seeking behaviours and related factors

Nine reviews reported factors positively associated with condomless sex, including, but not limited to, low perceived HIV risk,^{38 39} alcohol and substance use,^{40 41} meeting partners via the internet,^{42 43} sex with HIV-positive partners⁴⁴ and chemsex³⁴ (table 2). Knowledge of partner’s HIV status,⁴⁵ self-efficacy,⁴⁰ knowledge of HIV status and use of ART⁴⁴ and poor body image³⁶ were negatively associated with condomless sex. The effect of depression on condomless sex showed both positive and negative association in one review.⁴⁶ Social support was the only interpersonal-level factor found to be negatively associated with condomless sex⁴⁰ (table 2).

A global review found that depression, higher education and income, low perceived partner HIV risk, older age and other individual-level factors were negatively associated with condom use⁴⁷ (table 2). Another review in Asia observed that a lack of HIV status transparency was negatively associated with condom use.⁴⁸ PrEP use⁴⁷ and knowledge of partner’s HIV status⁴⁹ were positively associated with condom use.

One meta-analysis revealed that depression and anxiety, alcohol and substance use, intimate partner violence, childhood sexual abuse, condomless sex and casual partners were individual-level factors positively associated with sexual compulsivity.⁵⁰ PrEP use⁴⁷ and substance use⁵¹ were positively associated with having a higher number of partners. In a global review, chemsex was positively associated with having casual partners.³⁴ Engaging in transactional sex was positively associated with engaging in group sex and chemsex in China.⁵² The presence of other STIs, LGBT identity and the lack of knowledge of protective services were also positively associated with chemsex in a global review.³⁴

In the USA, a lack of knowledge of PrEP was negatively associated with engaging in transactional sex (table 2).³⁰ In a global review, PrEP use was positively associated with engaging in unspecified sexual risk behaviours.⁵³ Lastly, in one review, lower achieved education and income were negatively associated with meeting partners via the internet.³⁹

Non-usage of HIV prevention and testing services

Another body of research clustered around the use of HIV and testing services. For example, low perceived HIV risk,^{54 55} fear of being outed as MSM,^{48 54} fear of a positive diagnosis^{48 55} and a lack of knowledge of services⁴⁸ (table 3) were negatively associated with use of services. Convenience of the service,⁵⁴ obligation to test⁵⁴ and identifying as LGBT³⁴ were positively associated with HIV service use.

Interpersonal stigma from healthcare providers towards MSM,⁴⁸ community stigma⁵⁵ and community expressed gender norms⁴⁸ were negatively associated with HIV service use. Cost of testing,⁵⁴ low access to services locally⁵⁵ and religious stigma towards MSM⁴⁸ were institutional-level factors that impeded HIV service use. Institutional-level stigma such as ethnic stereotyping was found to negatively affect access to services²⁸ (table 3). Participants highlighted peer-led facilities as a service quality that motivated HIV service use.⁵⁴

Societal stigma,^{48 54} specifically criminalisation of LGBT individuals,⁴⁸ and stigma towards the LGBT community³⁵ were structural factors negatively associated with HIV service use.

Factors contributing to additional HIV vulnerability

Individual-level health was examined in four reviews (online supplemental table 4). Transactional sex,⁵² sex in public or at gay venues⁵⁶ and chemsex³⁴ were positively associated with the presence of other STIs. Low income and homelessness were positively associated with substance use.³⁴ Non-adherence to ART was positively associated with viral load.³⁷

Community-level racial stigma was the only factor negatively associated with employment,²⁸ which in turn reduced access to healthcare and services (online supplemental table 4). Community-level HIV and LGBT stigma were negatively associated with patients disclosing their HIV status to healthcare providers and partners.²⁸ Stigma from religious organisations was also negatively associated with HIV status transparency^{28 48} (online supplemental table 4).

A lack of community support was positively associated with internalised stigma within MSM²⁸ (online supplemental table 4). Drug use, penis size and sexual positioning were positively associated with endorsing masculinity through sexual means³⁷ (online supplemental table 4).

Discussion

This systematic review of reviews provides specificity on the relationships between risk factors that work in relation and create feedback loops that (re)produce GBMSM’s HIV risk globally, according to the literature. Our analyses and visualisation, although not exhaustive, indicate that centring the entire system is needed to address the interactional, adaptive and aggregative nature of GBMSM’s HIV risk.

New HIV infections should only be theoretically possible and not a lived reality, given advances in science. However, as our review shows, HIV infections are inequitably borne by GBMSM with lived experiences that interact in untoward ways that amplify HIV risk. Research in Vietnam attributes rising HIV prevalence among GBMSM to insufficient protective resources and social norms calling for alternative HIV testing strategies.⁵⁷ A 2013 study in the USA examined the efficacy of evidence-based interventions targeted at MSM and HIV-risk behaviours but found limited interventions specific to MSM of colour in southern regions, which is where the US HIV epidemic is predominately concentrated.⁵⁸ Another review examined the complexity of factors that shape PrEP use, which included groupings such as substance use, sociodemographic, sexual factors, networks and disease histories.¹¹ Linking those findings to our review and visualisation provides even greater specificity to the relationships that need to be areas of focus beyond the individual level.

As with all studies, limitations exist, which should be considered when interpreting the findings. Although we systematically reviewed existing reviews, the processes outlined in figure 2 are only a start towards offering the research, policy and practice community a more nuanced way to think about HIV ‘risk’. The figure builds on previous work that examined factors directly associated with HIV infection and not just HIV ‘risk’, reifying the complex system.²¹ Although the SEM helps identify levels of influence, it often oversimplifies the dynamic, non-linear processes at play, resulting in the application of complex systems in this study. Moreover, SEMs can often (over) emphasise the ‘surround’ and minimise individual and community agencies in shaping the environment. Another limitation of the visualisation and analysis is that many pathways were identified as associations, given the large cross-sectional nature of the research literature, which limits understanding of causal influences. Missing reviews also limit the interpretation of our results, as many focused on individual-level factors. For example, community viral load is important because if GBMSM living with HIV in a particular community are overwhelmingly undetectable, the risk of transmission would not exist. Sexual mixing patterns are also important. Studies have found that racial preferences are at play in sexual relationships among GBMSM, such that black GBMSM tend to have sex within their racial group and white men within theirs.59,62 Given the high prevalence and incidence of HIV among black GBMSM and less PrEP access, each sexual event carries more HIV ‘risk’. However, advances in PrEP, including injectables, and in combination with U=U, with its global diffusion, fully recognise that HIV cannot be transmitted if enough of a concentration within the network uses the prevention tools. Community organising and mobilising by gay men in organisations such as ActUP and Stop AIDS San Francisco also sped up US and global responses to HIV that are directly related to recent advances.⁶³ Lastly, the visualisation of the HIV risk environment may be viewed as too broad, creating a ‘one-size-fits-all’ model, which is not how it should be interpreted. Rather, the visualisation speaks to the complexity of HIV risks, knowing that some will vary across time, geography and the prevalence and incidence of HIV within any given population. Future reviews that focus on causal study designs are needed to map the system of HIV risk among GBMSM and other priority populations. Future studies could leverage the advancement of artificial intelligence to conduct a more robust and expansive systematic review of the global literature, particularly for non-individual-level and associational research. Interventionists should also consider implementing strategies that can act in a domino effect, creating larger, more preventative feedback loops. This is particularly important as other studies have identified even more complex systems of HIV, including psychosocial, biological, syndemics and health systems.64,67

Conclusion

This visualisation of the scientific literature indicates that GBMSM’s HIV risk, globally, is structured by adaptive feedback loops and dynamic interactions that induce a high HIV risk environment. Interventions that create an HIV risk-reducing environment are urgently needed to reach the global goal of ending the HIV epidemic.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.