Discordant Reporting of Partner Labels, HIV Testing and Sexual Behavior Among a Sample of Partnered Men Who have Sex with Men in the US

Abstract

There is growing evidence that sexual behaviors among male couples are strongly shaped by emotional and quality characteristics of the relationship, and that the labels that men attach to their relationships may indicate how men perceive and engage in risk taking. There has been a lack of attention to how male couples label their relationships, and how discordant understandings of relationship labels may shape HIV risk behavior. Using data from a sample of 804 partnered men who have sex with men, this analysis examines associations between discordant relationship labeling and participation in HIV testing and sexual behavior. Men who labeled their relationship differently from their partners were less likely to have been recently tested for HIV and more likely to engage in sexual risk. The results underscore the need to develop interventions that provide spaces and skills for men to learn how to effectively navigate HIV risks in their relationships.

Introduction

There is now strong evidence for the role of male dyads in the US HIV epidemic, with primary partners identified as the source of approximately one-third [1] to two-thirds [2] of new HIV infections. This understanding of the role of primary partners in HIV transmission is leading to a slow, but significant, paradigm shift in HIV prevention, with an increased recognition of the need to develop dyadically focused interventions. HIV prevention programming has traditionally focused on individual men who have sex with men (MSM), in particular gay-identifying men, approaching prevention through the lens of the risk of HIV acquisition from casual sex, and has largely ignored the risk of HIV transmission that occurs within primary partnerships. Historically, HIV prevention efforts have focused on reducing the number of casual sex partners [3], indirectly messaging that primary partnerships pose reduced risk of HIV and conveying a misplaced sense of protection associated with primary partners [4, 5]. Understanding relationship characteristics and dynamics among partnered MSM is essential in identifying targetable modes of prevention that are specific to dyadic contexts. However, before interventions can be developed, is it imperative to understand how HIV risk behaviors exist within and across the spectrum of relationship typologies [6].

Data indicate that compared to their single peers, partnered MSM participate in less frequent testing for HIV [7, 8] and report more frequent condomless anal sex (CAS) [2]. In order to address prevention promotion among this population, it is important to understand the relationship contexts and characteristics that may be driving sexual risk behavior and lowered engagement in HIV prevention within dyads, as these factors likely represent the most potent intervention points (i.e. providing couples with behavioral skills to discuss HIV risk within their relationship). Much of the literature regarding sexual risks among male couples has focused on sexual agreements. A sexual agreement refers to an explicit mutual understanding made by a male couple regarding sexual behaviors that are permitted within and outside of the relationship [6, 9]. There is now a significant body of research illustrating that the majority of male couples form sexual agreements [6, 10–12]. Sexual agreements may take the form of monogamy, permission for sex with outside partners under conditions (i.e. no emotional attachments), or sex with outside partners with no conditions. Sexual agreements are formed for a number of reasons, including as a tactic for managing the risk of HIV, or to shape sexual pleasure, intimacy, communication, and trust within the relationship [6, 9, 13–15].

There is growing evidence that sexual behaviors—including the formation of and adherence to sexual agreements—among male couples are strongly shaped by emotional and quality characteristics of the relationship [16–22]. The greater prevalence of condomless anal sex (CAS) in primary partnerships has been postulated to be shaped by feelings of intimacy, trust, and love [17, 23, 24]. Forgoing condoms can be a tactical move among MSM to denote trust and foster intimacy with their partners [25–29]. However, some relationship dynamics have also been documented to be associated with preventive behaviors. Couples with more constructive communication styles and higher levels of relationship satisfaction report higher levels of investment in their sexual agreements [11, 14, 30], and these relationship traits may make it easier for couples to discuss and decide on sexual agreements [31].

While it is clear that male couples represent an important intervention target for reducing the transmission of HIV and other sexually transmitted infections (STIs), and that there are distinct relationship factors that shape sexual risk behaviors in dyads, research to date has largely focused on a comparison of single versus partnered men. Bavinton et al. [32] note that partnerships exist on a continuum, ranging from casual (often anonymous) sex partners through to long-term romantic, committed relationships. Researchers, however, has tended to categorize relationships as primary/committed versus casual [30, 33, 34]. There is also emerging evidence that the manner in which MSM perceive, or label, their relationships can shape sexual behavior with their partners. Sullivan et al. [35] analyzed qualitative data from a sample of 28 young (18–24 years) MSM and identified six fluid relationship types reported among participants: (1) hooking up that turned into dating, (2) date that turned into a hook-up, (3) hooking up that turned into friends-with-benefits, (4) friends-with-benefits that turned into semi-dating, (5) breaking up with strings, and (6) ambiguous exploration. Sexual and emotional expectations varied across these relationship typologies, and differences were in part shaped by perceived levels of commitment and desires for the future direction of the relationship. Using qualitative interviews with 13 gay and bisexual men, Goldenberg et al. [36] identified that how MSM labeled or perceived their partner shaped their sexual risk decisions with that partner. Men who perceived that their sex partners were not gay-identified, not out, or were behaviorally bisexual, considered these partners to be a lower risk for HIV transmission. The qualitative findings by Sullivan et al. [35] and Goldenberg et al. [36] illustrate the potential for sexual behavior to vary considerably among dyads dependent on how individuals perceive, or label, their relationship. These relationship labels (i.e. partner, friend-with-benefits) connote degrees of commitment and trust, and thus shape how men perceive the risk of HIV acquisition in the relationship. Missing from the literature, however, is an understanding of how differential labeling of relationships by partners is associated with participating in prevention and sexual risk behaviors. Reporting different relationship labels may be a sign of poor communication and discordant expectations around sex and intimacy. Discordant relationship labels may be a risk factor for reduced engagement in HIV prevention, working through a pathway of reduced or ineffective communication among couples. The current analysis uses dyadic data from a survey of partnered MSM to examine associations between discordant reporting of relationship type and HIV testing and recent sexual behavior. This new information has the potential to inform dyadic interventions that address the behavioral skills required by male couples to effectively engage in and manage HIV prevention.

Methods

Data for this study are from a randomized controlled trial (RCT) evaluating an innovative dyadic HIV testing and counseling intervention for male couples (ClinicalTrials.gov Identifier: NCT02335138). The online recruitment process and data collection methods have been described previously [37]. Participants were recruited from across the US via targeted advertising on social media platforms (e.g., Facebook, Instagram), and geospatial mobile apps for MSM (e.g., Grindr, Scruff, Bareback Realtime). Participants for each couple were eligible for participation in the RCT if they were both (1) cis-gender male, (1) 18 years of age or older, (2) had been in a sexual relationship with each other for more than 6 months, (3) had not had an HIV test in the last 3 months, (4) did not report severe intimate partner violence (IPV) (experience or perpetration) within the past 12 months, (5) were willing to receive rapid home HIV test kits at an address of their choice, (6) had internet access, and (7) self-reported being concordant HIV negative or HIV sero-discordant. Transgender participants were excluded as the sample size for the RCT was insufficient to permit an analysis stratified by gender identity, and it was felt inappropriate to combine transgender and MSM participants into one group for analysis.

Participants provided their contact information on the study’s website including their preferred name, email address, cell phone number, and mailing address. Joint emails were sent to the couples requesting each partner to complete separate baseline surveys. Upon survey completion, couples were randomized to receive either the control condition (rapid home HIV test kits followed by self-reporting of test results via the study’s website) or the intervention condition (rapid home HIV test kits followed by testing together under the facilitation of a remotely located counselor during a prescheduled video session). Participants completed separate follow-up surveys at 3 and 6 months. Study approval was obtained from the Institutional Review Board at the University of Michigan (HUM00102906).

The current analysis is restricted to data from the baseline survey. Between April 2016 and June 2017, 13,592 individuals accessed the study’s landing page that contained a short description of the proposed activities, 2926 individuals (21.5%) clicked through to the eligibility screener, of which 862 individuals (29.5%), i.e. 431 couples met the inclusion criteria. Of these, 868 individuals (424 couples) (98.4%) completed separate baseline surveys. Analysis is restricted to individuals whose partner also completed the survey and those without missing data, for an analysis sample of 804. Analysis of the variable measuring whether a participant had CAS with an outside partner when there was no agreement present is restricted to those with agreements, for a sample of 681. Demographic information obtained from participants included age, race and ethnicity, sexual identity, highest educational level, employment status, and legal marital status. The survey included a series of questions pertaining to their relationship including its duration (in years and months) and engaging in CAS in the past 3 months (with primary and outside sex partners). The survey described a sexual agreement (an agreement you have both talked about and decided regarding sex with others outside of your relationship) and where asked whether they had a sexual agreement with their primary partner. Participants who reported having a sexual agreement were asked if it was “open” (i.e. sex with outside partners was allowed without restrictions), “open with restrictions” (i.e. sex with outside partners was allowed with restrictions (e.g. type of sex) or “closed” (i.e. sex with outside partners was not allowed). Participants were also asked about their own and their partners’ past HIV testing behavior in the past 12 months: participants reported whether they had been tested in the past 12 months and the date of their most recent test.

Mutual trust was measured using the Dyadic Trust Scale [38], a 5-point Likert tool ranging from “strongly disagree” to “strongly agree” on eight items (one’s partner is more interested in his own welfare, there are times when one’s partner cannot be trusted, one’s partner is honest and truthful, one feels he can trust his partner completely, one’s partner is truly sincere in his promises, one feels his partner does not show him enough consideration, one’s partner treats him fairly and justly, one can count on his partner for help). This scale has been tested and validated among sexual minority populations, including men in same-sex relationships [39]. Data from the baseline survey showed excellent internal reliability for this scale (Cronbach’s alpha = 0.87).

Two sub-scales from the Communal Coping to Reduce HIV Threat Scale were used to measure the extent to which couples share preferences for general lifestyle issues and sexual health. Each of the sub-scales was measured using a 5-point Likert tool ranging from “not to any extent at all” to “to a great extent”. The Preferences for General Lifestyle Scale includes six items including diet, exercise, finances, general preventive health, social activities, and sexual relationship. The Preferences for Sexual Health Outcomes Scale relates to sexual health includes seven items including using condoms when having sex with each other, limiting the number of outside sex partners, having sex outside the relationship, using condoms when having sex outside the relationship, getting regularly tested for HIV and other STIs, being the insertive or receptive partner when having sex with each other, and being sexually faithful to each other. The Communal Coping to Reduce HIV Threat Scale was specifically developed and validated among MSM who had been partnered for at least 3 months [40]. Data from the baseline data showed excellent internal reliability for this scale (General Lifestyle Cronbach’s alpha = 0.92: Sexual Health Cronbach’s alpha = 0.89).

Communication Style was measured using the Communication Patterns Questionnaire—Short Form (CPA-SF), a 11 item scale that measures participants’ perceptions of their interactions with their partner [41]. Items include two stems: “when issues or problems arise, how likely is it that” (responses include both partners avoid discussing the problem) and “during a discussion of issues or problems, how likely is it that” (responses include both partners express feelings to each other). Data from the baseline data showed good internal reliability for this scale (Cronbach’s alpha = 0.78).

Binary variables were created to measure dyadic differences within partnerships, including whether partners reported differing races, sexual identities or sexual agreements (the binary variables denotes whether there was any differences in those characteristics between partners). Continuous variables were created for differences in age, trust, communication, desire for future of relationship, preferences for general lifestyle, and preferences for sexual health. The continuous difference variables were created by subtracting the response of partner A from partner B, to create variables representing the absolute difference in reporting of characteristics between partners.

The analysis considers four outcomes. The first outcome measures whether partners report different relationship labels. Each partner was asked “What term best describes your relationship with [name]?” Options included: Boyfriend, Lover, Husband, Spouse, Partner, Fuck buddy, Hook-up, Friend with benefits, and other (please describe). Less than 1% of participants selected other, and all wrote in descriptions. Relationship labels were categorized as formal (Boyfriend, Lover, Husband, Spouse, Partner) and informal (Fuck buddy, Hook-up, Friend with benefits). Written in descriptions for other were included in the most appropriate category (e.g. “boo” was considered formal). A binary variable was created to measure discordant partner labels, coded one if the participant reported a different relationship label than their partner. The second outcome measures recent HIV testing, and is coded 1 if the participant reports having tested for HIV in the 12 months prior to the survey. The third and fourth outcomes measure recent sexual behavior. A binary variable is coded one if the participant reports CAS in the past 3 months with both their primary partner and an outside sex partner and does not currently report taking PrEP. Another binary variable is coded one if the participant reports that they do not have an agreement that allows for CAS with an outside sex partner and they have had CAS with an outside sex partner in the past 3 months.

Separate logistic regression models were fit to identify independent associations of demographic, dyadic and dyadic difference characteristics with each of the four binary outcomes. Multilevel models were fit for individual participants nested within dyads, with a random intercept for the dyad. For each model, N equaled the total number of individuals and not the total number of couples, as the analysis seeks to understand how relationship factors shape individual behaviors. Results are presented as adjusted odds ratios (aORs) with 95% confidence intervals (CIs). Statistical analyses were conducted using STATA v15.

Results

This online recruited sample of partnered MSM was predominately white (75.3%), educated (52.6% had at least college level education) and gay identifying (90.2%). Respondents represented 43 US states. The sample was relatively young, with three-quarters under the age of 35 years (Table 1). In terms of relationship characteristics, 28.4% of men were legally married to their study partner, approximately one-third had been in their relationship for 1–3 years, and 84.7% reported having a sexual agreement. Monogamy was the most commonly reported sexual agreement (72.0%) (Tables 1 and 2). Approximately one-fifth of men reported a different race from their partner, and 16.8% reported a different sexual identity. More than half reported a different sexual agreement from their partner, and this did not vary by the type of sexual agreement they reported. Approximately one-third reported discordant relationship labels. Just under half reported having tested for HIV in the past 12 months. In terms of recent sexual behavior, 47.4% reported CAS with both a primary and outside sex partner in the past 3 months and 13.9% reported CAS with an outside sex partner in the absence of a sexual agreement that permitted it.

Table 1.

Participant and relationship characteristics of 804 partnered men who have sex with men, United States, April 2016–June 2017

	N (%)
Participant characteristics
Age
18–24	236 (29.4)
25–34	379 (47.4)
35–44	117 (14.4)
> 45	72 (8.8)
Race
White non-Hispanic	606 (75.3)
Black non-Hispanic	61 (7.7)
Other non-Hispanic	137 (17.0)
Ethnicity
Hispanic	160 (19.9)
Non-Hispanic	644 (80.1)
Education
High school/GED	118 (14.7)
Some college	263 (32.7)
College graduate and graduate level	423 (52.6)
Currently unemployed	129 (16.0)
Currently employed	675 (84.0)
Identifies at gay	725 (90.2)
Identifies as bisexual or other identity	79 (9.8)
Had a HIV test in the past 12 months	393 (48.9)
Reports recent CAS with both primary and outside sex partner in past 3 months	381 (47.4)
Reports recent (3 months) CAS with outside sex partner when sexual agreement does not permit it	112 (13.9)
Relationship characteristics
Relationship length
< 1 year	135 (16.8)
1–3 years	282 (35.1)
3–5 years	139 (17.3)
6–10 years	156 (19.3)
> 10 years	92 (11.5)
Currently married	229 (28.4)
Currently unmarried	575 (71.6)
Sexual agreement
No agreement	123 (15.3)
Monogamy	579 (72.0)
Sex with outside partners with conditions	91 (11.4)
Sex with outside partners with no conditions	11 (1.3)
Interracial relationship
Same race	630 (78.4)
Interracial	174 (21.6)
Difference in sexual identities
Same sexual identity	669 (83.2)
Different sexual identities	135 (16.8)
Difference in sexual agreements
Report same sexual agreement	332 (41.3)
Report different sexual agreement	472 (58.7)
Report different partner labels
Report same partner label	539 (67.0)
Report different partner labels	265 (33.0)

Table 2.

Distribution of relationship characteristics and dyadic differences in relationship characteristics among 804 partnered men who have sex with men, United States, April 2016–June 2017

	Mean	Standard deviation	Range
Trust in partner	39.08	13.63	8–93
Shared preferences for general lifestyle	28.23	7.99	0–40
Shared preferences for sexual health	32.11	7.08	0.40
Communication style with partner	36.93	15.17	11–93
Age difference with partner	0.27	8.75	−37–46
Difference in trust	0.43	10.49	−49–52
Difference in shared preferences for general lifestyle	−0.27	5.86	−26–24
Difference in shared preferences for sexual health	−0.38	6.64	−26–24
Difference in communication style with partner	0.50	10.85	−49–52

Table 3 shows the factors significantly associated with men reporting discordant relationship labels than those reported by their partners. The only individual factor that was significant was sexual identity, with gay identifying men less likely to report discordant relationship labels. Reporting of discordant relationship labels was significantly less likely at relationship lengths over 3 years, among men who were married and among those with any sexual agreement. While no individual reports of relationship characteristics were associated with reporting discordant relationship labels, dyadic differences in age, trust, and preferences for sexual health were all associated with increased odds of reporting discordant relationship labels. Men in interracial relationships, those who reported different sexual orientations from their partner and those who reported different sexual agreements from their partner were more likely to report discordant relationship labels.

Table 3.

Factors associated with discordant reporting of partnership labels 804 partnered men who have sex with men, United States, April 2016–June 2017

	Participant reports different relationship label than their partner
	AOR (95% CI)
Participant characteristics
Age (18–24)
25–34	1.20 (0.81, 1.79)
35–44	1.20 (0.68, 2.11)
> 45	1.46 (0.73, 2.86)
Race (White non-Hispanic)
Black non-Hispanic	1.40 (0.78, 2.51)
Other non-Hispanic	1.03 (0.65, 1.59)
Ethnicity (Non-Hispanic)
Non-Hispanic	1.05 (0.69, 1.59)
Education (high school/GED)
Some college	0.96 (0.59, 1.56)
College graduate and graduate level	0.79 (0.48, 1.27)
Currently employed (unemployed)	0.79 (0.52, 1.20)
Identifies at gay (bisexual or other identity)	0.53 (0.32, 0.98)
Relationship characteristics
Relationship length (< 1 year)
1–3 years	0.85 (0.55, 1.36)
3–5 years	0.74 (0.52, 0.97)
6–10 years	0.69 (0.38, 0.98)
> 10 years	0.58 (0.32, 0.89)
Currently married (unmarried)	0.45 (0.23, 0.75)
Sexual agreement (no agreement)
Monogamy	0.76 (0.52, 0.98)
Sex with outside partners with conditions	0.36 (0.12, 0.58)
Sex with outside partners with no conditions	0.56 (0.23, 0.84)
Trust in partner	0.98 (0.96, 1.01)
Shared preferences for general lifestyle	1.01 (0.98, 1.05)
Shared preferences for sexual health	0.97 (0.95, 1.08)
Communication style with partner	0.95 (0.97, 1.07)
Partner differences
Interracial relationship (same race)	1.23 (1.06, 1.42)
Age difference with partner	1.64 (1.45, 1.87)
Different sexual orientation to partner (same)	1.68 (1.45, 2.02)
Difference in trust	1.09 (1.04, 1.12)
Difference in shared preferences for general lifestyle	1.02 (0.98, 1.07)
Difference in shared preferences for sexual health	1.11 (1.05, 1.18)
Difference in communication style with partner	0.99 (0.84, 1.13)
Report different sexual agreements	1.45 (0.89, 1.67)

Results in italics are significant at p < 0.05

Factors associated with HIV testing, recent CAS with primary and outside partners, and CAS with outside partners when not permitted as per sexual agreements are presented in Table 4. Black men and those with higher levels of education were more likely to report recent HIV testing. No individual characteristics were significantly associated with sexual risk. The odds of recent HIV testing declined significantly with relationship length, whereas the odds of both sexual behavior outcomes increased significantly with relationship length. Married men were less likely to report a recent HIV test. Those with agreements other than monogamy were more likely to report a recent HIV test and also more likely to have had CAS with both primary and outside partners. Once again, while individual reports of relationship characteristics were not associated with HIV testing or sexual behavior, dyadic differences in these reports were. Those with partner age differences, differences in preferences for sexual health, and those who reported different sexual agreements were less likely to report recently testing for HIV. Those with differences in trust, preferences for sexual health and who reported different sexual agreements were significantly more likely to report recent CAS with a primary and outside sex partner, and recent CAS with an outside sex partner in the absence of a sexual agreement. Those who reported discordant relationship labels from their partner were significantly more likely to report recent CAS with a primary and outside sex partner, recent CAS with an outside sex partner in the absence of a sexual agreement and less likely to report recent HIV testing.

Table 4.

Factors associated with HIV testing (n = 804), recent CAS with primary and outside partners (n = 804) and CAS with outside partner when sexual agreement did not permit it (n = 681) among partnered men who have sex with men, United States, April 2016–June 2017

	Reports HIV testing in the past 12 months (n = 804)	Reports CAS with primary partner and an outside partner (n = 804)	Reports CAS with an outside partner when their agreement doesn’t allow for it (n = 681)
	AOR (95% CI)	AOR (95% CI)	AOR (95% CI)
Participant characteristics
Age (18–24)
25–34	1.67 (0.10, 2.36)	1.02 (0.58, 1.80)	0.99 (0.53, 1.84)
35–44	1.61 (0.94, 2.76)	2.07 (1.04, 3.85)	1.63 (0.74, 2.54)
> 45	0.98 (0.51, 1.91)	1.01 (0.39, 2.01)	0.50 (0.14, 1.78)
Race (white non-Hispanic)
Black non-Hispanic	1.94 (1.08, 3.48)	0.61 (0.22, 1.45)	0.69 (0.23, 2.17)
Other non-Hispanic	1.04 (0.69, 1.55)	1.01 (0.60, 1.91)	1.44 (0.80, 2.61)
Ethnicity (Non-Hispanic)
Non-Hispanic	0.79 (0.53, 1.18)	0.67 (0.38, 1.23)	0.82 (0.47, 1.42)
Education (high school/GED)
Some college	1.55 (1.14, 2.55)	1.73 (0.79, 2.54)	2.21 (0.92, 3.43)
College graduate and graduate level	1.43 (1.12, 2.03)	0.60 (0.74, 2.85)	1.96 (0.80, 4.01)
Currently employed (unemployed)	1.30 (0.86, 1.96)	1.42 (0.75, 2.81)	1.36 (0.65, 2.81)
Identifies at gay (bisexual or other identity)	0.91 (0.55, 1.49)	1.22 (0.55, 1.79)	1.06 (0.52, 2.19)
Relationship characteristics
Relationship length (< 1 year)
1–3 years	0.36 (0.23, 0.58)	3.21 (1.35, 4.85)	4.04 (1.51, 6.86)
3–5 years	0.24 (0.14, 0.42)	3.59 (1.40, 5.41)	3.31 (1.12, 4.12)
6–10 years	0.14 (0.08, 0.26)	3.13 (1.22, 5.46)	3.91 (1.34, 6.74)
> 10 years	0.30 (0.15, 0.60)	4.21 (1.44, 6.85)	5.53 (1.61, 8.51)
Currently married (unmarried)	0.74 (0.68, 0.93)	0.80 (0.47, 1.37)	0.74 (0.41, 1.35)
Sexual agreement (no agreement)
Monogamy	0.89 (0.79, 1.12)	0.35 (0.12, 0.47)	–
Sex with outside partners with conditions	1.42 (1.11, 1.65)	2.21 (1.89, 2.54)	–
Sex with outside partners with no conditions	1.42 (1.21, 1.74)	3.45 (2.54, 4.09)	–
Trust in partner	1.01 (0.98, 1.03)	1.02 (0.97, 1.05)	0.99 (0.96, 1.04)
Shared preferences for general lifestyle	1.02 (0.97, 1.03)	0.97 (0.93, 1.05)	1.01 (0.95, 1.04)
Shared preferences for sexual health	1.02 (0.99, 1.05)	1.02 (0.96, 1.05)	1.01 (0.96, 1.05)
Communication style with partner	0.99 (0.97, 1.02)	1.03 (0.99, 1.06)	1.03 (0.99, 1.06)
Partner differences
Interracial relationship (same race)	1.08 (0.75, 1.56)	0.97 (0.55, 1.77)	0.88 (0.49, 1.59)
Age difference with partner	0.85 (0.74, 0.98)	0.99, (0.97, 1.02)	0.99 (0.97, 1.02)
Different sexual orientation to partner (same)	0.81 (0.54, 1.22)	1.10 (0.60, 1.89)	1.14 (0.62, 2.12)
Difference in trust	0.99 (0.97, 1.01)	1.23 (1.05, 1.38)	1.15 (1.03, 1.26)
Difference in shared preferences for general lifestyle	1.03 (0.98, 1.06)	0.98 (0.93, 1.03)	0.99 (0.95, 1.08)
Difference in shared preferences for sexual health	0.75 (0.56, 0.94)	1.32 (1.11, 1.46)	1.36 (1.05, 1.89)
Difference in communication style with partner	0.99 (0.97, 1.02)	1.02 (0.97, 1.03)	0.89 (0.75, 1.23)
Report different sexual agreements	0.54 (0.23, 0.79)	1.78 (1.35, 2.23)	4.56 (2.19, 6.48)
Report discordant relationship labels	0.69 (0.54, 0.73)	1.51 (1.23, 1.79)	1.89 (1.45, 2.33)

Results in italics are significant at p < 0.05

Discussion

The results presented here provide new insight into associations between relationship perceptions and HIV-related risk behavior within relationships. Approximately one-third of men reported a relationship label that was very different from that reported by their partner. As hypothesized, discordance in relationship label reporting was significantly associated with reduced HIV testing and increased recent sexual risk taking. Recent qualitative studies [35, 36] have shown strong links between how men label their relationships and the behaviors that they are willing to undertake in their relationships. Relationship quality factors such as a desire to create intimacy or trust, or hopes for the future direction of the relationship act as decision-making tools for behavioral risk. For example, men may negate condom use as a mechanism for building trust or creating intimacy [17, 23, 24]. The labels men apply to these relationships thus become a marker for both their emotional desires for that relationship and the behavior they are willing to participate into achieve those desires. Labels applied to the relationship, by the individual, or through partner discussions, also set a script for the expected future direction of the relationship (i.e. degree of emotional commitment or fidelity) and the behavior that is expected in the relationship. In the current analysis, it was surprising to find that one-third of men labeled their relationship differently from their partner, an indicator that partners may have very different behavioral and emotional expectations for the relationship. This was apparent in the associations between relationship labeling discordance and HIV testing and sexual risk. Previous work has shown that men with high levels of trust and communication are less likely to break their sexual agreements with their partner Gomez et al. [31]. Similarly, men who have not talked to their partners about what to call their relationship, or who have talked but have differing perceptions, are likely also not talking about or agreeing on expectations around sexual behavior outside of the relationship or participating in regular HIV testing. Men who differed from their partner on relationship quality indicators—including trust and preferences for sexual health—were less likely to engage in HIV testing and more likely to engage in sexual risk behavior. Similarly, these results likely reflect a lack of coherence among partners on expected and desired relationship characteristics. These dyadic differences could be addressed through interventions that provide a couple with the space and skills to explore their individual and joint expectations for their relationship.

Men who had been in relationships for longer periods of time, those who were married and men with sexual agreements were all less likely to report discordant relationship labels. These results highlight the central role of communication: to create a sexual agreement and to agree to be married, both require a couple to communicate effectively and make joint decisions. With longer relationship lengths come more opportunities to develop communication styles that work for the couple. However, men at longer relationship lengths were also more likely to have sex outside of their relationship when their agreement did not allow for it. While men may be better at communicating and defining their relationship over time, it may not necessarily translate into risk reduction behavior. Interventions are needed that can address this paradox. Interventions that focus on teaching couples the behavioral skills for effective communication and to talk about and agree upon risk reduction behaviors have the potential to reduce discordance in relationships that is linked to increased HIV risk borne out of discordant relationship expectations.

Interestingly, men who were age, race, or sexual identity discordant were more likely to report different relationship labels. It is possible that demographic differences are associated with differences in cultural ideals, lived experiences and relationship histories and that each of these differences may create obstacles to couples talking effectively about their relationship desires. Dyadic differences in male couples have also shown to increase HIV risk behaviors [42], and so careful attention is required to tailor HIV prevention interventions for the unique contexts of demographically different couples. However, further work, ideally using qualitative methodologies, is needed to fully understand how demographically discordant male couples communicate and navigate relationship building.

The results also further support for the role of sexual agreements in shaping HIV risks for partnered MSM. Previous studies have shown that the establishment and adherence to a sexual agreement helps couples reduce their risk for HIV [30, 43]. Men who had a relationship other than monogamy were more likely to have had recent CAS with both their primary and outside sex partners but they were more likely to have tested recently for HIV. Hence, partnered men who have created sexual agreements may be enacting testing behaviors based on the behaviors that their agreements allow. Importantly, those who differed in the reporting of their sexual agreement were less likely to have tested recently for HIV and more likely to have engaged in each of the sexual risk indicators. This raises an important point about sexual agreements: they require men to talk through and agree upon terms with their partner, otherwise they become assumptions rather than agreements. Desires for and adherence to agreements are also fluid and change over time, requiring men to have the confidence and behavioral skills to routinely check-in with their partner on their sexual agreement.

There are several limitations to the results. The cross-sectional data from a convenience sample precludes making causal inferences and limits the generalizability of the results to all male couples in the US, who may or may not use social media platforms and geospatial mobile apps. To enroll in the RCT required men to provide personal identifying information, and thus social desirability may have shaped how participants reported their sexual or HIV testing behaviors. Although we requested participants to complete the survey separately from their partners, it is possible that couples answered questions together, potentially influencing each other’s responses. The categorization of relationship labels into the binary variable used in the current analysis has not previously being validated. Participants were only asked to choose a relationship label, and did not designate them as formal or informal. It is also possible that participants are choosing labels that reflect the behavior they want to, or expect to, engage in within their relationship. The sample is also predominantly non-Hispanic White, and future research attention is needed to explore these issues among a more racially and ethnically diverse sample of male couples. While the strengths and uniformity of the results suggests the variable is capturing relationship differences, further work is warranted to refine the measurement of relationship labels.

Collectively the results highlight the need for dyadic interventions that provide male couples with opportunities and skills to talk about and explore their relationships as a pathway to engagement in HIV prevention and risk reduction. Couples’ HIV testing and counseling (CHTC), originally developed for heterosexual couples in sub-Saharan Africa [44], has been adapted for same-sex male couples [45], and is a CDC Public Health Strategy [46]. CHTC provides an opportunity for partnered men to talk about their relationships and develop sexual agreements in the context of the HIV test results. There are several other examples of dyadic interventions that aim to address HIV risk among same-sex male couples. Connect with Pride was an intervention for methamphetamine-using, Black/African American male dyads that involved seven in-person sessions to address issues such as communication, joint problem-solving, and condom negotiation [47]. 2GETHER was an intervention for young male couples aged 18–29 years that involved four interactive weekly sessions focusing on enhancing communication skills, coping with relationship stress, applying problem-solving techniques to relationship issues, and formulating an agreement to reduce their risk for HIV [48]. Post-test decreases in sexual risk behaviors, increases in skills related to HIV prevention, and improvement in relationship investment were observed. We Prevent is a novel intervention for 15-19 year old male dyads that is currently being piloted in the US [49], involving telehealth-delivered sessions to increase relationship communication skills around HIV prevention. Each of these interventions focuses on dyadic skill building as a mechanism for reducing HIV risk and recognizes that by teaching partnered men to work together, and creating shared expectations for the relationship, male couples can understand, react to and plan for the HIV risks experienced while in their relationships. It is imperative that this work be extend to all types of male couples, particularly young male couples for whom there is the greatest risk of HIV acquisition at a time when they have not yet developed or practiced the behavioral skills necessary to communicate and plan with a partner.