Globally, HIV has become a major public health issue among men who have sex with men (MSM) [1]. It has been reported that, compared to the general male population, MSM are 28 times more likely to be affected by HIV globally [2], and this difference is about 100 times in China [3]. The overall prevalence of HIV among Chinese MSM during 2001–2018 has been estimated to be 5.7% [4]. Besides, some other sexually transmitted infections (STIs), such as the syphilis epidemic [5], also have a higher prevalence in Chinese MSM. The higher risk for STIs among MSM is mainly due to sexual risk behaviors, such as condomless sex [6]. A meta-analysis has suggested that the rate of condomless sex in Chinese MSM was 55.2% in the past 6 months [7].

Internalized homophobia (IH) is the assemblage of negative attitudes toward homosexuality in general or homosexual features and behavior among sexual minorities [8,9]. According to Meyer’s conceptualization of the Minority Stress Theory, there are distal and proximal stressors unique and chronic for minority populations. IH is a form of stigma-related proximal stressor, attributed to distal stressors (e.g., acts of antigay discrimination and victimization) [10]. Moreover, IH is in the most proximal position along the continuum from environment to self [10]. In countries with heterosexism-dominated culture, discrimination and prejudice for MSM may be pervasive [11]. Even if one’s homosexuality is concealed, MSM may be harmed by social values toward self, especially in countries where collectivist culture is highly accepted [12]. From a different perspective, symbolic interaction theorists also contend that interactions with others are important for the development of a sense of self, and negative regard from others (e.g., stigma) leads to negative self-regard (e.g., IH) [13]. Persistent IH negatively impacts health, including low self-esteem, defenses that can lead to isolation [14], anxiety, depression, and suicidal ideation [15]. Importantly, IH has further effects on risky behavior such as condomless sex [16,17] due to a low level of self-validation, coping with psychological distress resulting from the perceived stigma, and temporarily escaping from shame and depression [18,19]. Besides, a higher level of IH might be related to more substance and alcohol use that may impair decision-making processes [20]. However, the results regarding the association between IH and condomless sex remain inconsistent [21], reflecting a need for more nuanced studies, such as examining the mediator or moderator [22].

In the past 20 years, there has been a pronounced increase in studies testing potential roles of social capital in health [23]. Social capital (SC), which was described as a source of support and benefits embodied in interpersonal and social networks [24], may play an essential mediation role between mental health and risky behaviors [25–27]. The social capital theory contends that the accumulation of human capital is determined by resources of social relationships [28]. However, whether individuals can access group-level resources (e.g., medical and health services) depends on many factors, including personality and mental health [10]. For those with high IH, some group-level resources (e.g., support from family, safer sex information from MSM community) may be difficult to access. For example, due to a high level of IH, some old MSM have been cut off from their family of origin [29]. Besides, MSM indicating greater IH are more isolated from the MSM community and are less likely to discuss HIV prevention with sexual partners [30]. Therefore, IH might show a direct effect on SC, and further affect health through SC. In addition, SC has been suggested to mitigate poor mental status (e.g., stigma, and depression) and condomless sex. For example, a study conducted among 98 male sex workers in the northeast of the US has found that SC acts as a buffer in the relationship between stigma and sexual risk behaviors [16]. This has suggested that when estimating mediation effects of SC, the potential interaction effect between SC and exposure should also be considered. A qualitative study has suggested that reducing IH and fostering SC are vital elements to improve the effectiveness of HIV risk prevention in this population [31]. Although this finding is encouraging, it is unclear whether and how SC may mediate the effect of IH on condomless sex.

Moreover, individuals of stigmatized groups generally evaluate themselves by comparing with others who look similar to them [10], and MSM who are non-homosexual men have a risk of double stigma cast by gays besides heterosexuals [32]. The previous study has also suggested that MSM who are non-homosexual subgroups may be exposed to more stressors and have more mental health problems than homosexual subgroups [33]. Therefore, the association of IH, SC and condomless sex should be considered separately in these two subgroups. Besides, different SC dimensions are worth concerning due to their different roles on health outcomes [34,35].

Prior work on examining mediation effects may be limited in considering the interaction between the mediator and exposure variables. This study aimed to examine the potential mediation effect of SC on the association between IH and condomless sex in Chinese MSM, with consideration of the interaction between SC and IH. We therefore incorporated a novel causal mediation analysis with a four-way decomposition approach, which is suitable in the presence of a mediator with which the exposure may interact [36] and can help understand how much of a total effect may be attributed to mediation, interaction, both mediation and interaction, or neither. These analyses could help design targeted intervention strategies to reduce risky sexual behaviors among MSM.

This study on Chinese MSM suggested that IH and SC were closely associated with condomless sex. Using a novel four-way composition method, we found that for those with a high level of SC, IH is still a risk factor for condomless sex. Improving SC could reduce and/or mitigate the negative effect of IH on condomless sex, with heterogeneity observed in different domains of SC. Such a mechanism was only found in the homosexual subgroup.

Social connections may have “mixed” effects and paradoxically exacerbate the effect of IH, especially if such connections entail role strain associated with obligations to provide social support to others (e.g., a man was supported and expected to carry on the family line) [52]. The heterogeneity of results in the IF-based SC and CS-based SC might be attributed to the “mixed” effects. Specifically, for CS-based SC, our study reinforced the evidence that reduction of IH could decrease condomless sex by encouraging community participation and improving government support [26,31]. In contrast to the CS-based SC, however, we observed an adverse trend for IF-based SC, i.e., the risk of condomless sex due to the direct effect of IH increased when IF-based SC was fixed at a higher level. The higher IF-based SC may be explained by the support from family members and MSM peers, which might result in more social interaction among MSM. However, for those with a high level of IH, family obligations based on strong IF-based SC become a possible stressor [52], and access to social support from other sexual minorities might be restricted [53]. They also suffer more adverse mental status (e.g., anxiety, low condom use self-efficacy) [15,54] and behavior (e.g., substance use) [20], increasing their condomless sex behavior. Besides, subgroup analysis suggested that the risk of condomless sex due to IH and the mediation and interaction effect of SC were only significant in the homosexual subgroups. However, similar to a recent study in Kazakhstan [55], our findings suggested that the non-homosexual population reported a higher level of IH, more condomless sex, and a lower level of SC than their counterpart.

We provided new findings that deserve discussion using the causal mediation analysis with the four-way method. First, with total SC fixed at a high level, mediation effects of total SC contributed significantly to the total effect, highlighting the dominant role of SC as a mediator between IH and condomless sex in MSM. Second, we found that even attained a high level of SC, MSM with a high level of IH still had a higher risk of condomless sex than those with a low level of IH. For those with a high level of SC, structural stigma may still shape the fear of being stigmatized and crisis sense of support loss from social contacts (e.g., family members, colleagues) [56,57]. External prejudice events (i.e., distal stressors) and IH might further increase the risk of condomless sex. This finding suggested that joint intervention of reducing IH and fostering SC should be designed to improve the effectiveness of HIV risk prevention [31]. Third, we found that a higher IF-based SC may not mediate or mitigate the effect of IH on condomless sex; however, risk attributable to interactions gradually increased with the improvement of CS-based SC. Those with enough community support can obtain more available health services, and self-efficacy in safe sex also mitigates the effect of IH [58].

The findings on the relationship between IH and risky sexual behavior seem mixed since the first definition of IH was proposed [59]. Williamson’s review revealed some inconsistent evidence about IH and risky sexual behavior [17]. Our study on Chinese MSM suggested a significant association of IH with condomless sex, which is in line with the findings in the US [20,30], Israel [60], and Ugandan [61]. However, inconsistencies also exist [21,62], which might be explained by one or more potential moderators [21]. Since the previous study proposed the need for more nuanced studies, such as examining the mediator or moderator [22], we, therefore, extended our previous study, which focused on the relationship between IH and sexual risk behavior [63], considering the potential role of SC in the association between IH and condomless sex. More evidence is still needed to better understand the effect of IH on condomless sex through other factors.

Our findings have strong policy implications. Although homosexuality and same-sex sexual behavior were no longer considered a disease since 2001, homosexuality is still not widely accepted [9], with structural stigma and prejudice leading to a high level of IH among Chinese MSM [64]. A previous report showed that nearly 85% of Chinese MSM kept their homosexuality secret [65]. In recent years, the Chinese government has issued some policies to enhance health service delivery for MSM [66], while mental health services for this population are still lacking [15]. As our findings suggested, those with both high SC and IH still had a high risk of condomless sex. These findings provided information on the professions or service providers of the indispensable reduction of psychological problems and minority stress like IH and to improve the family, community, and social support towards MSM to mitigate the impact of IH on condomless sex [15,26,31,67]. Further efforts are needed at civil and political levels to address structural discrimination, establish a supportive environment and strengthen peer support [68,69].

Several limitations should be mentioned in our study. First, we used a health-related SC scale for the general population and could not differentiate the SC from MSM and non-MSM peers. The estimates due to mediations and moderation of peer-related SC may differ, which requires future research. Second, the reporting bias might exist, especially for some information such as sexual orientation. However, we used anonymous questionnaires to reduce this bias as much as possible. Third, findings from the participants in Sichuan may not be generalized without caution to the overall MSM in China. Fourth, we adopted a non-probability sampling and cannot guarantee the representativeness of the total sample. Fifth, some potential covariates (e.g., pre-exposure prophylaxis use) were not included in our study, which may lead to an overestimation of the contribution of IH to the risk for condomless sex.