PrEP use and PrEP use disclosure are associated with condom use during sex: A multilevel analysis of Latino MSM egocentric sexual networks

Abstract

Pre-exposure prophylaxis (PrEP) is integral to HIV prevention; however, the influence of PrEP use and PrEP use disclosure on condom use is unclear among Latinx men who have sex with men (LMSM). This study explored associations of LMSM PrEP use and use disclosure on consistent dyadic condom use in the past 6 months. Participants were 130 HIV-negative PrEP and non-PrEP using LMSM ages 20–39 years. Two-level logistic regression models assessing individual- and dyadic-level predictors on condom use were fitted using R. Participants reported a mean of four sexual partners (n=507 dyads). Participants who reported using PrEP or having more sexual partners were more likely to use condoms; however, participants who reported disclosing PrEP use were less likely to use condoms. Future longitudinal studies should characterize approaches to increase informed personal health choices and conversations about PrEP, condom use, and other HIV risk-reduction strategies using network methodologies.

INTRODUCTION

Although the national rate of new HIV diagnoses has plateaued in the general population from 2014–2018, HIV incidence has increased among young Latinx men who have sex with men (LMSM) ages 13–24 years (1–3). If current incidence rates persist, approximately one in five LMSM will be diagnosed with HIV in their lifetime (4). Miami-Dade County, Florida, is one of the counties with the highest number of new HIV cases in the United States (U.S.), and experiences an HIV incidence four times that of the national rate (1, 2). In Miami-Dade County, more than one out of every two new HIV diagnoses occur among LMSM, signifying a pressing need for additional biomedical and behavioral (together termed “biobehavioral”) prevention efforts utilizing condoms and pre-exposure prophylaxis (PrEP) within this population facing substantial HIV disparities (5).

Previously established predictors of condomless sex has been associated with partner closeness (6), serodifference (6–9), use of alcohol/drugs (7, 9–11), HIV status disclosure (8), exposure to discrimination due to gay-identification (12), and gay affirmation (a self-affirmation of gay identity in which feeling that being gay is an important and positive part of one’s identity and being gay is normal and fulfilling) (13, 14). Findings from previous studies suggest that gay affirmation is correlated with increases in condom use self-efficacy (15–17).

Additionally, endorsement of the traditional Latinx cultural values of simpatía (promotion of harmony in relationships) and machismo (expectation of men to detach themselves from any feminine attributes and that men have uncontrollable sexual urges) has been negatively associated with condom use during anal sex for LMSM (9). The Latinx cultural value familismo (familism) is associated with increases in condom use (18). Familism emphasizes the importance of individuals accepting responsibility of their families and close friends (19). Based on this interpretation of familism, we hypothesize that LMSM with higher beliefs of familism may be more likely to use PrEP or condoms. The association between Latinx-specific cultural values and uptake of HIV preventive behaviors necessitates the importance of examining the role that cultural factors play on PrEP use, PrEP use disclosure, and condom use among LMSM.

In accordance with the risk compensation framework, in which risk behaviors increase following use of other risk reduction methods (20–22), findings from previous studies suggest that condom use intention may decrease after PrEP uptake (23, 24). Reasons for decreased condom use intention include barriers to arousal and an increased temptation for condomless sex, which is perceived as more pleasurable (25). However, other studies have found that PrEP uptake does not influence condom use (26–29). Marcus et al. (2019) suggested that there is a population-level decline in condom use among MSM that is not due to the effects of risk compensation. Instead, MSM may be experiencing an increasing need for alternatives to condom dissemination programs, such as PrEP programs (30). These mixed findings suggest that additional factors may influence condom use (20–22, 26–29).

Geospatial networking apps used to find sexual partners allow users to display and/or disclose PrEP use status (24); however, few studies have investigated the effects of PrEP use disclosure on condom use. Due to the perception that PrEP users engage in condomless sex, MSM who disclose PrEP use may find it difficult to negotiate condom use (31). Hooks and Gross (2020) found that MSM are more likely to perceive PrEP users as being more warm and/or trustworthy relative to individuals who do not use PrEP (32). Additionally, the authors found that White PrEP users are perceived to be more likely to engage in condomless sex and group sex compared to non-White groups, suggesting the need to study the role that race plays on the association between PrEP use disclosure and condom use (32).

The role of PrEP use and PrEP use disclosure on condom use is unclear. The Theory of Planned Behavior (TPB), which is often posited to explain condom use among MSM (33), can only predict limited variance in condom use intentions and condom use behavior (24% and 12.4% variance, respectively) (33). Thus, additional factors at other levels, such as social network factors may be needed to explain condom use. In the context of this study, the Social Action Theory can explain condom use based on both personal characteristics (e.g., PrEP use) and dyadic social interaction factors (e.g., PrEP disclosure) (34). Egocentric network studies have successfully used multilevel analyses to describe correlates of condom use within a prespecified time period (35). The present study aimed to identify correlates of condom use at the personal and dyadic social interaction levels using multilevel modeling. We hypothesized that, among LMSM, PrEP use and PrEP use disclosure during sex will be associated with consistent condom use with each sexual partner. Further, we hypothesized that there would be an interaction effect between PrEP use and racially identifying as White on consistent condom use.

METHODS

Study sample, eligibility criteria, and recruitment

We utilized data from a cross-sectional social network study of the sexual practices of LMSM living in Miami-Dade County. Participants were recruited from a South Florida community-based organization using respondent-driven sampling methods. Briefly stated, the parent study, PrEParados enrolled 130 participants grouped into 1 of 10 social networks. The project coordinator first recruited 12 seeds. Then, each of the 12 seeds recruited 3 friends who then, in turn, each recruited 3 friends (13 egos per network). This formed a total of 10 networks of 13 LMSM (n=130). Additional information about study recruitment can be found elsewhere.

To be included in the study, participants had to report: 1) cis-male identity, 2) HIV-negative status, 3) sex with a man in the past six months, 4) Hispanic, Latino, or Latinx identification, and 5) behaviors which would qualify them for PrEP prescription based on the CDC PrEP Clinical Practice Guidelines (36). Data were collected from October 2018 to August 2019.

Data collection

Trained bilingual interviewers employed with our community partner obtained written consent from participants. Participants completed the assessment at a community partner site. All study materials were developed in English and Spanish using the back-translation method supported by Marin & Marin (1991). Oral and written data we collected using tablets and a paper-based survey. Study data were collected and managed using REDCap software (Research electronic data capture) hosted at the University of Miami (37, 38).

Measures

Participant/Ego-level Characteristics-

Participants provided individual-level data, including race, PrEP use, number of sexual partners in the past 6 months, and information surrounding their endorsement of familism (39) and self-affirmation of gay identity (14). Familism was measured using the previously validated Mexican American Cultural Values Scale, which consists of 16 items and utilizes a 5-point Likert scale (“Strongly Disagree” to “Strongly Agree”) (39). Total possible mean scores could range from 1–5. Although originally developed for Mexican Americans, this scale is widely used for other foreign-born and US-born Hispanic/Latinx subgroups and nationalities (40). Based on previously established correlates of condom use, this study used a modified version of the full scale to examine the effects of three familism subscales: support (6 items), obligations (5 items), and referents (5 items). In this study, the Cronbach’s alpha for all 16 items in the three familism subscales was 0.98. Due to the high internal consistency among the subscales, a confirmatory factor analysis model was fitted and predicted values for the familism latent variable were extracted for each participant using the R “lavaan” package (41). This new construct was used in our models to account for the effect of familism on condom use.

Gay affirmation was measured using the previously validated Internalized Homonegativity Inventory, a reliable and validated scale used commonly among MSM to measure internalized homophobia and self-affirmation of gay identity (14). The original gay affirmation subscale consists of 7 items utilizing a 6-point Likert scale ranging from “Strongly agree” to “Strongly disagree” (14). However, 2 items were removed from the original subscale for this analysis based on feedback from our community partners during pre-testing. The internal consistency for the final 5 items in this subscale had a Cronbach’s alpha of 0.85.

Situational/Dyadic-level Characteristics -

Participants provided demographic data, and descriptive information of up to 12 sexual partners in the past six months, including anonymous sex partners, using measures developed for this survey. For each sexual partner, participants reported situational/dyadic-level characteristics such as condom use during insertive and receptive anal sex, alcohol use, HIV status disclosure, emotional closeness to each partner, and PrEP use disclosure. Only male alters were included in our analysis.

Emotional closeness was measured using the question, “On a scale of 1 to 5, how close do you feel to this person?” and values could range from “Not at all close” to “Extremely close” with total possible scores ranging from 1 to 5. Reported closeness of ego to alter was group mean centered by ego.

PrEP use disclosure by the alter was measured using the question, “Did this person tell you he was using PrEP?” and the response was binary (Yes/No). PrEP use disclosure by the ego was measured using the question, “Did you ever tell this person that you were using PrEP?” and the response was binary (Yes/No). It is important to note that although the ego may have disclosed PrEP use, this does not necessarily mean the ego was actually using PrEP at that time.

The outcome of interest, consistent condom use during anal sex, was measured by asking each participant, “Please indicate if you had: 1) Active anal sex without a condom refers to when you inserted your penis into another person’s anus without a condom (you were a top); 2) Passive anal sex without a condom refers to when another man penetrated you in the anus with his penis (you were bottom)” for each alter. The occurrence of consistent condom use during anal sex was coded as a binary variable, (1) for consistent condom use and (0) for inconsistent condom use (i.e., ever engaging in condomless sex with that alter).

Data analysis

All analyses were conducted using R (42). Descriptive statistics of participants (egos) and sexual partners (alters) are displayed in Tables I and II, respectively. For each model, odds ratios (ORs) and 95% confidence intervals were estimated. Our models were fitted using a Bernoulli sampling distribution. At level 1, the unit of analysis was the alter characteristic (n=507), and at level 2, the unit of analysis was the ego-level characteristic (n=130). As different random and fixed effects were added to the models discussed below, Likelihood-Ratio Tests (LRT) were used to compare if each subsequent step was a significantly better fit than the previous one. We specified the null model, which reported an intra-class correlation of 0.78, signifying that 78% of the variance in condom use is due to between participants (ego) effects. Our baseline model (Table III: Fixed Model) included the variables of interests for both the sexual partner attributes (level 1) and the participants’ attributes (level 2).

Table I:

Participants’ sociodemographic characteristics

	PrEP nonuser n (%)	PrEP user n (%)
Total	38 (30)	89 (70)
Mean age in years (SD)	28.54 (4.4)	27.71 (3.8)
Race
White	68 (78)	23 (66)
Multi-racial	14 (16)	10 (29)
Black/African American	3 (3)	2 (6)
American Indian or Alaskan	1 (1)	0 (0)
Asian or Pacific-Islander	1 (1)	0 (0)
Country of nativity
US-born	40 (51)	21 (55)
Foreign-born	39 (49)	17 (45)
Income
$24,999 or less	11 (12)	5 (13)
$25,000 - $34,999	34 (38)	20 (53)
$35,000 - $49,999	26 (29)	9 (24)
$50,000 or more	18 (20)	4 (11)
Marital status
Single or never married	83 (91)	29 (76)
Married or domestic partner	8 (9)	9 (24)
Mean sexual partners (SD)	3.7 (3.0)	4.4 (2.4)
Mean gay affinity score (SD) (alpha = 0.85)	4.8 (1.0)	5.2 (1.0)
Mean Latino cultural values score	3.0	3.4
Mean religiosity score (alpha = 0.97) (SD)	2.3 (1.3)	2.4 (N/A)
Mean family referent scores (alpha = 0.93) (SD)	3.2 (N/A)	3.4 (0.9)
Mean family support score (alpha = 0.93) (SD)	3.3 (N/A)	3.8 (1.0)
Mean family obligation score (alpha = 0.93) (SD)	3.3 (1.2)	3.8 (0.9)
Mean respect score (alpha = 0.97) (SD)	3.2 (1.2)	3.6 (0.9)

Table II:

Descriptive statistics of alters, as perceived by participants

	Alter n (%)
Condomless sex
Used condom during sex	144 (28)
Ever had condomless sex	363 (72)
Ethnicity
Hispanic/Latino	146 (29)
Non-Hispanic/Latino	361 (71)
Age
Under 18 years	1 (0)
18 – 29 years	299 (59)
30 – 39 years	141 (28)
40 – 49 years	23 (5)
50 years or older	8 (2)
No response	35 (7)
Partner PrEP use disclosure
Partner did not disclose PrEP use	406 (80)
Partner disclosed PrEP use	101 (20)
Participant PrEP use disclosure
Participant did not disclose PrEP use	388 (77)
Participant disclosed PrEP use	119 (23)

Table III:

Multi-level analyses of condom use during sex

	Fixed Model			Random Slope Model			Final Model with Interactions
Predictors	Odds Ratios	CI	p	Odds Ratios	CI	p	Odds Ratios	CI	p
(Intercept) Situational/Dyadic-level characteristics	0.00	0.00 – 0.01	<0.001	0.00	0.00 – 0.00	<0.001	0.00	0.00 – 0.01	<0.001
Closeness with alter	0.70	0.46 – 1.04	0.079	0.37	0.15 – 0.91	0.030	0.71	0.47 – 1.06	0.090
Alter is living with HIV and undetectable	1.30	0.64 – 2.63	0.465	1.35	0.58 – 3.13	0.483	1.28	0.64 – 2.59	0.487
Drank alcohol	0.69	0.39 – 1.22	0.207	0.71	0.38 – 1.32	0.279	0.68	0.39 – 1.19	0.179
Alter disclosed PrEP use	1.05	0.60 – 1.82	0.871	0.96	0.52 – 1.79	0.901	1.01	0.58 – 1.76	0.971
Ego disclosed PrEP use Participant/Ego-level characteristics	0.49	0.23 – 1.04	0.063	0.45	0.18 – 1.09	0.078	0.42	0.20 – 0.91	0.028
Race is White	0.35	0.08 – 1.45	0.147	0.35	0.07 – 1.71	0.196	0.28	0.07 – 1.17	0.081
Ego is using PrEP	2.40	1.05 – 5.45	0.037	2.54	1.01 – 6.38	0.046	9.48	2.54 – 35.47	0.001
Number of sex partners	1.41	1.12 – 1.79	0.004	1.61	1.19 – 2.17	0.002	1.44	1.14 – 1.81	0.002
Gay self-affirmation score	3.34	1.37 – 8.18	0.008	4.24	1.49 – 12.04	0.007	3.04	1.30 – 7.14	0.011
Familism – latent variable	3.31	1.19 – 9.17	0.021	3.11	1.00 – 9.70	0.051	4.83	1.64 – 14.24	0.004
White x Using PrEP							0.11	0.02 – 0.55	0.007
Random Effects
σ2	3.29			3.29			3.29
τ00	5.32 ego			12.75 ego			4.52 ego
τ11				2.63 ego-alter closeness
ρ01				0.94 ego
ICC	0.62			0.82			0.58
N	116 ego			116 ego			116 ego
Observations	428			428			428
Marginal R2 / Conditional R2	0.483 / 0.802			0.382 / 0.890			0.538 / 0.805

Our next step was to identify a model by specifying the random effect of participant’s closeness to the alter, as suggested by prior research (Table III: Random slope model) (6). This model was a better fit than the baseline (χ²(2)=6.68, p<.05), and the closeness effect was significant. The overall variance for this random slope was 2.63, which contributed to an increase in the Conditional R² (from 0.80 to 0.89) but decreased the Marginal R² (from 0.48 to 0.38). This meant that the fixed effects in the random slope model explained less of the overall variation in our data.

The following interaction effects were then explored in a subsequent model: a) the ego using PrEP and the alter disclosing their PrEP use, and b) both the ego and the alter disclosing PrEP use. This interaction effect did not improve the model. Based on findings from prior research (32, 43), we then included the interaction effect of White race and PrEP use. There was a significant interaction effect between participants identifying as White and being on PrEP. This is represented in Table III: Final Model with Interactions. The final model with interactions was a better fit overall than the baseline fixed model (χ²(1)=8.65, p<.01). The final model failed to converge when the random slope for ego-alter closeness was added; thus, this random effect was dropped. Finally, a simple slope analysis was conducted to further investigate the interaction between ego being on PrEP and identifying as White on the outcome variable of condom use (Figure 2 and Table IV).

Figure 2:

Predicted probability of condom use, based on interaction of White race and PrEP use

Table IV:

Simple slope analysis of condom use, based on interaction of white race and PrEP use

Slope of White LMSM when ego is using PrEP
Est.	S.E.	z-value	p-value
−3.44	1.17	−2.93	0.00

Slope of White LMSM when ego is not using PrEP
Est.	S.E.	z-value	p-value
0.89	0.99	0.90	0.37

Ethical approval and informed consent

Written consent in either English or Spanish (based on participant preference) was provided and obtained from all participants. The present study was approved by the Institutional Review Board of the University of Miami (IRB # 20180284). All study team members, including our community-based partners, received Human Subjects Research training prior to study commencement. Participants received a $50 gift card as compensation for their time.

RESULTS

Participant characteristics- Thirty percent of participants reported currently using PrEP. Almost all participants self-identified as gay (96%), with few identifying as bisexual (4%). Participants were predominantly White (72%), employed full-time (84%), single or never married (87%), and born in the U.S. (56%). Our sample was heterogenous for Latinx nativity: of those born outside of the U.S. (n=56), 39% reported Cuba nativity, 9% reported Nicaragua nativity, 9% reported Dominican Republic nativity, and the remainder from 5 other Latin American countries.

Alter Characteristics

Egos perceived the majority of their sexual partners to be Latinx (71%) and gay (88%). Except for one female alter, all sexual partners were male—the female alter was excluded from the analysis. As seen in Table II, the majority of sexual partners were reported to be of similar age to the participants (18 to 39 years old).

Hierarchical logistic regression model

The odds of condom use with a sexual partner were 9.48 (95% CI: 2.54, 35.47; p<0.01) times larger with ego’s PrEP use, 1.44 (95% CI: 1.14, 1.81; p<0.01) times larger for every additional sexual partner, 3.04 times larger (95% CI: 1.30, 7.14; p<0.05) for each increment in gay self-affirmation score, and 4.83 (95% CI: 1.64, 14.24; p<0.05) times larger for each point increment in familism score. The odds of condom use with a sexual partner were 0.42 (95% CI: 0.20, 0.91; p<0.05) for ego disclosing PrEP use compared to non-disclosure to this sexual partner during sex. Finally, although there was no effect of White race on condom use, White LMSM who used PrEP were less likely to use condoms compared to Non-White LMSM (OR=0.11; 95% CI: 0.02, 0.55; p<0.05). The interaction effect is further described in Table IV.

Figure 1 demonstrates the multilevel correlation between condom use and PrEP use, stratified by race (only significant effects are shown). The first two models were not significant for ego’s PrEP use disclosure to the alter on condom use; however, this association became significant when controlling for the interaction between White race and ego’s PrEP use. As seen in Figure 2 and Table IV, the interaction between PrEP use and White race is negative and significantly different than zero (β=−3.44; SE=1.17; z-value=−2.93; p<0.05), meaning that a non-White LMSM PrEP user is more likely to use condoms relative to a White LMSM PrEP user. Conversely, for PrEP non-users, racial identity did not yield a significant change in slope (both had low odds of consistent condom use, as expected for PrEP non-users).

Figure 1:

Plot of final multi-level model of correlates with condom use during sex

DISCUSSION

This study used multilevel modeling of egocentric sexual networks to identify correlates of consistent condom use with a sexual partner by LMSM in Miami-Dade County, Florida. We found that PrEP disclosure, PrEP use, number of sexual partners, and higher endorsement of the Latinx cultural belief familism and self-affirmation of gay identity were associated with condom use. The most relevant finding of this study is that PrEP use is positively correlated with consistent condom use; however, when the ego disclosed PrEP use (regardless of whether they were actually using PrEP), consistent condom use decreased. Previous studies have found contradictory conclusions on the risk compensation effects of PrEP use on condom use (20–24).

Our findings suggest that disclosing PrEP use is associated with an increase in condom use. This may be due to underlying sociocultural barriers. There may be a heightened stigma attached to PrEP use, in which individuals who use PrEP are perceived to engage in condomless sex (31, 44–46). In addition, PrEP users have reported being labelled as “promiscuous,” by family and friends for their PrEP use (31). Endorsement or internalization of PrEP-related stigma may cause individuals to promote condom use with partners who use PrEP, believing these individuals may be more likely to have a sexually transmitted infection (31). This finding supports the Social Action Theory which considers condom use in the context of both individual- and situational-level factors. To ensure the dual protection method of both PrEP and condoms, condom use negotiation scripts should also consider incorporating PrEP use disclosure with the ultimate goal of promoting condom use. These negotiation scripts can be developed for specific priority populations using community-based participatory research approaches to ensure that language is stigma-free and tailorable for diverse situations, conversation modes (e.g., online, text, face-to-face), and sexual partnerships (e.g., casual, open relationship). Notably, the alter’s disclosure of PrEP use did not affect condom use. Future studies can examine condom use through a more intensive data collection tool, such as a sex journal which can more accurately capture situational level information during sex including conversations surrounding condom and PrEP use.

Some findings from previous research have suggested that Black and Latinx MSM have higher rates of condomless sex than Whites (47). Another study found that MSM perceived White MSM to have higher rates of condomless sex relative to other racial/ethnic identities (32). Few studies, if any, have looked at condomless sex among Latino racial subgroups. Our study, which focuses on LMSM, found that White LMSM were less likely to use condoms relative to non-White LMSM. When we performed a stratified analysis using a subsample of LMSM who did not use PrEP, condom use did not change by racial identification. Future studies should explore whether due to the disproportionate burden of HIV among Black MSM, White LMSM may feel less at risk for HIV due to their White self-identification despite being an ethnic minority.

Previous studies have found that LMSM who have more sexual partners engage in higher rates of condomless sex (6). However, our study found that as the number of sexual partners increased, LMSM reported more consistent condom use with alters. In addition, of partnered men, MSM in a monogamous relationship are more likely to perceive being protected from HIV relative to men in an open relationship; however, men in an open relationship have higher odds of recent HIV testing than men in a monogamous relationship (48). Future studies should examine how the number of sexual partners and type of sexual partnership (e.g., monogamy, polyamory) influence HIV prevention decision making processes, especially in the context of PrEP use, PrEP use disclosure, and condom use.

We found that familism and gay self-affirmation were positively associated with condom use. LMSM with higher endorsement of familism may feel more responsibility towards the health of their sexual partners. LMSM who feel more positive about their gay identity may feel more comfortable utilizing services such as PrEP navigation and condom dissemination programs at gay-friendly HIV prevention organizations. Our findings on the association between positive endorsement of familism and gay self-affirmation with condom usage suggests that future HIV prevention programs for LMSM should consider the integration of these factors in prevention programming. For example, HIV prevention programs could integrate gay-affirming messaging to help LMSM internalize positive beliefs about being gay, which in turn could lead to increased condom use (49).

Limitations to the present study include recall and social desirability bias due to the self-reported nature of data, lack of generalizability to the broader U.S. Latinx population as Miami-Dade County has a heterogenous and diverse Latinx community, and the use of perceived sociodemographic information (e.g., “street race”) of sexual alters which may not accurately reflect alters’ characteristics. In order to minimize recall and social desirability bias, we used forced-choice items, allowed for guided self-administration of the questionnaire, and employed bilingual LMSM interviewers. Multilevel modeling has strength in elucidating the multiple layers of both ego and dyadic/situational characteristics on the outcome of condom use; however, the utility of multilevel models in the analysis of egocentric network data requires that three conditions be met: 1) the dependent variable must be measured at the alter level; 2) observations at the ego level must be independent of each other; and 3) egocentric sexual networks cannot overlap. As participant recruitment was based on respondent-driven sampling, there is a small chance that conditions 2 and 3 may have been violated. Finally, we did not capture previous PrEP use in our analyses. Participants may have been previously using PrEP but were not using PrEP at the time of the study.

Conclusion

This study examined the association between social network characteristics and condom use among LMSM. A novel finding is that PrEP use was positively associated with condom use; however, the ego’s disclosure of PrEP use decreased odds of condom usage with sexual partners. Notably, the alter’s disclosure of PrEP use did not affect consistent condom use. In addition, we found that increased numbers of sexual partners yielded greater odds of condom usage. Previous studies on PrEP have focused on White MSM so studies such as this, with a Latinx population diverse in race and country of origin, are important (50). Social network methodologies can be used to increase the efficacy of informed personal health choices about PrEP, condoms, and other risk reduction strategies (U=U, negotiated safety) by LMSM.