Patterns of Sexual Behaviors in Young Men Who Have Sex With Men in Mexico.

Recent research has documented the importance of understanding the multidimensional nature of sexual risk behavior. However, little is known about patterns of sexual behavior among men who have sex with men (MSM) in Mexico, men who are at greatest risk for HIV and sexually transmitted infections compared to other subpopulations in the country. This study applied latent class analysis to data from a large, HIV-negative sample of 18- to 25-year-old Mexican MSM recruited from a social and sexual networking website (N = 3,722) to uncover multidimensional patterns of sexual behaviors, partner factors, and protective behaviors, and examine how these were associated with health and well-being correlates. We selected a model with seven classes. The most common class included those who reported both insertive and receptive behaviors with more than one partner, but smaller groups of individuals were in classes marked by only insertive or receptive anal sex, romantic relationships, or sexual inactivity. Class membership differed by sexual orientation, age, depressive symptoms, alcohol problems, and self-acceptance, with individuals in a class marked by same-sex relationships generally reporting more positive outcomes. Findings suggest heterogeneity of behaviors among Mexican MSM and the possible efficacy of prevention messages tailored to individuals' specific patterns of sexual behavior.

Researchers have increasingly emphasized the multidimensionality of sexual behavior, which can include a variety of behaviors (e.g., vaginal, anal, and oral sex), and can occur in different types of relationships and with different protective strategies (Diamond & Savin-Williams, [17]; Lefkowitz & Vasilenko, [39]; Welsh, Rostosky, & Kawaguchi, [60]). Person-oriented methods, which allow researchers to examine profiles marked by different patterns of behaviors, have been increasingly used to study sexual behaviors in heterosexual adolescents (Beadnell et al., [7]; Haydon, Herring, Prinstein, & Halpern, [28]; Vasilenko, Kugler, Butera, & Lanza, [56]), as well as in sexual minorities, including men who have sex with men (MSM) (Calabrese, Rosenberger, Schick, & Novak, [11]; Masters, Beadnell, Morrison, Hoppe, & Wells, [42]; Rice, Norris Turner, & Lanza, [50]). This approach can help researchers to better understand the heterogeneity of a population and in turn can create more effective and potentially tailored prevention programs. However, most of this research has been conducted in the United States, and little is known about patterns of sexual behavior in other countries, including in Latin American countries like Mexico. In this study we used data from a large sample of young, Mexican MSM recruited from a sexual networking website to examine their patterns of sexual behaviors, and how these behavioral profiles were associated with health and well-being.

Mexico is generally viewed as having a traditional sexual culture, with sexual values shaped by the Catholic Church (Carrillo, [13]). In addition, gendered cultural values such as machismo, which emphasize male sexual dominance, control, and serving as a breadwinner to the family (Carrillo, [13]; Lozano-Verduzco, [41]), may influence sexual behaviors in Mexican MSM. For example, men with same-sex attractions may have female partners due to the need to conform to heteronormative ideas of masculinity (Lozano-Verduzco, [41]; Solórzano & Mendoza, [53]). Prior research has demonstrated that men in Latin America may see acting as the insertive partner in same-sex encounters to be more masculine, whereas being the receptive partner is akin to taking a female role (Almaguer, [1]; Carrier, [12]; Jeffries, [33]). In addition, some have argued that because of these sexual positioning dynamics, Latin American men are less likely to take both the insertive and receptive roles in sexual encounters (Jeffries, [33]).

In recent years, Mexico has experienced increased acceptance of same-sex relationships, culminating in the Supreme Court legalizing same-sex marriage in 2016, although this has not yet been implemented in all Mexican states. However, despite increased acceptability, same-sex marriage remains a contentious issue (Lozano-Verduzco, [41]). Additionally, an active LGBT (lesbian, gay, bisexual, transgender) community exists in urban areas like Mexico City, with many young gay men from rural areas moving to cities to be more free to explore their sexuality and meet same-sex partners (Langarita Adiego & Salguero Velázquez, [36]). As in the United States and other countries, young Mexican MSM are at higher risk for HIV and other sexually transmitted infections (STIs) than are their heterosexual peers. For example, a recent, national study estimated HIV prevalence among MSM to be around 17% (Bautista-Arredondo, Colchero, Romero, Conde-Glez, & Sosa-Rubí, [6]). In addition, about 12% of Mexican MSM reported an STI diagnosis in the past year (Biello et al., [8]), and rates of syphilis are rising among young men (Herrera-Ortiz et al., [29]). Thus, a greater understanding of young MSM's sexual behavior is important for both understanding the normative sexual behaviors during this period of cultural transition and for preventing HIV and STIs.

Researchers have increasingly emphasized the importance of examining sexuality as a multidimensional construct (Diamond & Savin-Williams, [17]; Lefkowitz & Vasilenko, [39]; Welsh et al., [60]). Sexual behavior can be conceptualized as involving different types of behavior (e.g., oral, vaginal, or anal sex), protective measures (e.g., condom use), and relationship factors (e.g., number and types of partners). Sexual risk may not be fully conceptualized by examining these behaviors in isolation, but an examination of the intersection of these different aspects of behavior may be important in understanding sexuality and STI or HIV risk (Beadnell et al., [7]; Pflieger, Cook, Niccolai, & Connell, [46]; Vasilenko et al., [56]; Vasilenko, Rice, & Rosenberger, [57]). For example, research often examines condom nonuse as an important predictor of STI or HIV; however, condom nonuse may carry a very different risk in the context of a single, monogamous relationship, a casual partner, or multiple, concurrent partners. Thus, person-centered methods can identify patterns with the greatest risk of negative health outcomes and assess their prevalence in the population. In addition, by uncovering the heterogeneity of a population, latent class analysis (LCA) can provide information about this population to researchers, allowing for the creation of more efficacious interventions that may be targeted to particular high-risk groups or tailored to address the circumstances of individuals with distinct profiles of risk (Lanza & Rhoades, [38]).

Research has increasingly used person-oriented methods such as LCA to examine patterns of sexual risk behavior in heterosexual adolescents (Beadnell et al., [7]; Haydon et al., [28]; Hipwell, Stepp, Keenan, Chung, & Loeber, [30]; Lanza & Collins, [37]; Vasilenko et al., [56]) and sexual minority individuals (Chan et al., [14]; Masters et al., [42]; Rice et al., [50]; Vasilenko et al., [57]) in the United States. For example, one recent study examined a large sample of young MSM recruited from a social and sexual networking website for MSM, and identified eight classes of sexual behavior in the past year (Vasilenko et al., [57]). The most prevalent class, which comprised about half of the sample, included individuals who engaged in both insertive and receptive anal sex with male partners, had multiple partners, were not in a romantic relationship, and used condoms inconsistently. Smaller classes included individuals who only engaged in insertive or receptive anal sex, were in a (not necessarily monogamous) romantic relationship, engaged in only nonpenetrative behaviors, or were not currently sexually active. The smallest two classes included men who had sex with both men and women or only women, and these together accounted for only about 4% of the sample. This study suggests both a normative pattern of behaviors and considerable heterogeneity in sexual behavior among young MSM in the U.S. (Vasilenko et al., [57]).

Research has suggested that Mexican MSM engage in high rates of risk behaviors; for example, more than half of men reported inconsistent condom use in the past 3 months, which is linked to HIV and STIs (Biello et al., [8]). To our knowledge no study has looked at patterns of sexual behavior among young MSM in Mexico. One recent study examined patterns of sexual behavior among heterosexual adolescents in Puebla, Mexico, and found classes similar to those found in research in the United States (Espinosa-Hernández & Vasilenko, [20]). However, Mexican adolescents were more likely to be in classes marked by committed relationships but with no sexual behavior compared to U.S. adolescents, who were more likely to be in a class marked by sexual behavior within relationships. This suggests the possibility of lower rates of sexual behavior and a stronger emphasis on relationships among young people in Mexico; however, it is not clear whether this extends to young MSM.

In summary, this study used LCA (Collins & Lanza, [15]) to examine multidimensional patterns of sexual behaviors among young Mexican MSM. We focused on young MSM (ages 18–25 years), to examine a more homogeneous cohort of similar-aged men who may be at unique risk for sexual risk behaviors. This age period involves exploration in many areas of life (Arnett, [3]), including in individuals in Mexico, particularly those who are more educated or living in urban areas (Fierro Arias & Moreno Hernandez, [24]; Galambos & Martınez, [26]). This may be especially true for sexual minority youth in Mexico, who may have limited opportunities for dating and sexual exploration in adolescence, but may engage in more sexual behaviors after moving away from parents to larger urban areas where they can meet potential partners (Langarita Adiego & Salguero Velázquez, [36]). Thus, by focusing on individuals ages 18–25 years we were able to capture the early years of this transition out of adolescence. Because of recent legal changes regarding same-sex marriage in Mexico and around the world, this cohort, compared to older ones, may also experience a Mexican society that is more accepting of homosexuality; thus, we wanted to restrict our study to a smaller, more homogenous age rage. In addition to documenting patterns of behavior among young Mexican MSM, we examined how these patterns are associated with risk and protective factors. In terms of risk factors, we were guided by prior theory and research on syndemic production in MSM (Halkitis et al., [27]; Mustanski, Garofalo, & Donenberg, [44]; Parsons, Grov, & Golub, [45]; Stall et al., [54]). Syndemic refers to connected or interacting epidemics or risk factors, often occurring within a marginalized population (Parsons et al., [45]; Singer et al., [52]). Studies of MSM in the United States have identified five syndemic factors that are associated with sexual risk: depression, substance use, child sexual abuse, intimate partner violence, and sexual compulsivity (Parsons et al., [45]). These factors have also been found to be prevalent and associated with anal sex without a condom in Latin American MSM (Mimiaga et al., [43]). In addition, we considered two protective factors, self-acceptance and parental support, which may be protective against negative health outcomes in sexual minorities (Ueno, [55]; Williams, Connolly, Pepler, & Craig, [61]).

We had three primary research aims:

• To uncover latent classes of young Mexican MSM based on their past three month sexual experiences.
• To examine how class membership differs by sexual identity and age.
• To examine how risk and protective factors (depressive symptoms, alcohol problems, child sexual abuse, intimate partner violence, sexual compulsivity, self-acceptance, and social support) are associated with latent class membership.

Data were collected through a partnership with the operators of an Internet website for Spanish and Portuguese-speaking men who seek social or sexual connections with other men (Biello et al., [8]). Participants who had logged on in the past 90 days and lived in Spain, Portugal, or Spanish/Portuguese speaking countries in Latin America and the Caribbean were sent an email link to the survey through their internal mailbox on the site. The survey was first translated into Spanish and Portuguese and then back-translated into English. The surveys were then pilot tested with multiple native speakers to ensure clarity. Nearly 246,620 recruitment emails were opened, with 56,584 individuals clicking the provided link to the survey, and 36,063 initiating the survey. Surveys took about 30 minutes to complete, and no compensation was offered. This study was approved by the Institutional Review Board at The Fenway Institute at Fenway Health in Boston, MA. Nearly 27% of the sample reported living in Mexico (n = 9,731). We focused on 18–25-year-old men living in Mexico (n = 3,722). We chose this subsample for a number of reasons. First, we focused on young MSM, given they are in a period potentially marked by higher exploration and risk behavior (Arnett, [3]; Fergus, Zimmerman, & Caldwell, [22]). Second, because of the rapid changes in acceptability of homosexuality in Mexico, we wanted to include a sample that was relatively homogenous in age in order to reduce cohort effects. Finally, including 18–25 year-olds allows us to make comparisons to similar research on MSM conducted in the United States (e.g., Vasilenko et al., [57]).

We included seven indicators of sexual behavior, which previous researchers found to be associated with HIV/STIs in MSM (Friedman et al., [25]; Koblin et al., [35]; Rice et al., [49]) and that are similar to those that have been successfully used to identify classes of sexual behavior among MSM in the United States (Vasilenko et al., [57]; see Table 1 for frequencies). Note that we retained individuals who did not identify as gay or bisexual and included indicators of sexual behaviors with both men and women, as prior research has suggested that cultural values may lead Mexican men with same-sex attractions to engage in sexual behavior with women (Lozano-Verduzco, [41]).

Table 1. Demographics and Frequency of Sexual Behaviors in Young MSM in Mexico

Six indicators were based on a series of questions asking participants whether they had engaged in a particular behavior in the past 3 months. Participants were asked about their number of male anal sex partners and female vaginal or anal sex partners, and then what behaviors they engaged in with these male and female partners. Number of male partners and number of female partners were recoded to indicate whether the participant had 0, 1, or 2+ partners of each gender in the past 3 months. Insertive anal intercourse and receptive anal intercourse responses indicated whether or not they had engaged in each of these behaviors with a male partner, and were defined as "your penis was in his/their rectum" and "his/their penis was in your rectum," respectively. Penetrative sex with a female partner indicated whether they had engaged in any vaginal or anal sex with a female partner in the past 3 months.

Condom use was recoded from questions asking how many times they engaged in the sexual behaviors listed earlier, and how many of those times they used a condom in the past 3 months. This variable was recoded into a three-level variable of no sex without a condom, some sex without a condom, and all occasions of sex without a condom.

Participants' relationship status indicated whether they were in a single relationship lasting more than 3 months. It was recoded from an item asking their current relationship status, with seven response options: not currently dating, dating more than one person, and a relationship shorter than 3 months, coded as no relationship, and a relationship of 3–6 months, 6 months–1 year, 1–5 years, and more than 5 years coded as in a relationship.

We examined group differences in class membership by demographic, health, and well-being indicators.

Sexual identity was based on a self-report question with five response options (homosexual/gay, bisexual, heterosexual/straight, unsure/questioning, and other). Because of the small number of individuals reporting heterosexual, unsure/questioning, and other (n = 186; 5% total), we combined them into a single "other" sexual identity category, and used a three-level indicator of sexual identity (homosexual, bisexual, and other). Age was an indicator of self-reported age (in years; M = 22.15).

Syndemic risk factors were based upon constructs identified in prior research (Parsons et al., [45]). Depressive symptoms were calculated from 10 items from the Center for Epidemiologic Studies Depression Scale (Radloff, [48]), rated on a 4-point scale from rarely or none of the time to all of the time; items were summed to create a continuous depressive symptoms score (M = 7.8; SD = 5.3; α =.82, range = 0–30). Alcohol problems were assessed with four items from the CAGE checklist (sample item: "Have you ever felt you needed to cut down on your drinking?"), which has been validated in prior studies (Ewing, [21]; Fiellin, Reid, O'Connor, & Review, [23]) and used in populations of Latin American MSM (Mimiaga et al., [43]). Participants were coded as having an alcohol problem if they reported at least two of these items (17.9%). Child sexual abuse was recoded from three items asking whether participants had been touched by, been forced to touch, or been forced to have vaginal or anal intercourse with a person at least 5 years older; it was recoded to indicate whether an individual had experienced any sexual abuse (33.8%). Intimate partner violence (IPV) was measured by three items indicating if an individual had experienced emotional, physical, or sexual violence from a male partner in the past five years; this was recoded to indicate any type of IPV (40.6%). Sexual compulsivity was measured using the Sexual Compulsivity Scale (Kalichman & Rompa, [34]), which has been validated in MSM (Dodge et al., [18]) and Spanish-speaking populations (Ballester-Arnal, Gómez-Martínez, Llario, & Salmerón-Sánchez, [4]). The scale contained ten items scored on a 4-point scale (α =.90, range = 0–30). Consistent with prior research (Mimiaga et al., [43]), we dichotomized the measure to indicate greater sexual compulsivity (24 or greater; 15.2%).

In addition to these individual items, we also created a risk score of the five syndemic items, indicating how many of the following issues an individual experienced: alcohol problems, child sexual abuse, intimate partner violence, high sexual compulsivity, and a clinical level of depressive symptoms (10 or more; Andresen, Malmgren, Cartner, & Patrick, [2]). On average, participants reported 1.3 of the syndemic factors (SD = 1.7).

In addition, we included two protective factors. Self-acceptance was measured with a single item defining "Self-acceptance [as] how much a person is comfortable with being who he/she is," and then asking participants to rate "how much would you say that you accept yourself?" on a scale of 1–10 (M = 8.3; SD = 1.6). Social support was measured by a single item asking "In general, how satisfied are you with the overall social support you get from your friends and family members?" rated on a scale of 1–4 (M = 2.9; SD = 1.1).

We modeled classes of sexual behavior using Latent Gold (Vermunt & Magidson, [59]). We ran models with 1–10 classes, and selected the model with the optimal number of latent classes based on information criteria (Akaike information criterion [AIC] and Bayesian information criterion [BIC]) and interpretability. For our chosen model, we examined the patterns of item response probabilities, which show how likely individuals in a class were to report a given response on an item. We used these to create class names that reflect characteristics that are most critical in differentiating classes from each other. Next, we examined correlates of class membership using the three-step command that implements the BCH (Bolck, Croon, and Hagenaars) approach. This method weights analyses by each individual's probability of being in a given latent class, which provides less biased results than classify–analyze approaches (Bolck, Croon, & Hagenaars, [9]; Vermunt, [58]). We used this approach to examine the prevalence of membership in each latent class by sexual identity, and to examine how class membership differed by age, depressive symptoms, alcohol problems, child maltreatment, IPV, sexual compulsivity, the sum syndemic scale, self-acceptance, and social support. Analyses of the effects of these factors on class membership use a latent class probability-weighted multinomial regression model, which examines how each predictor is associated with membership in a particular class compared to a reference group.

Descriptive statistics are presented in Table 1. Most men in the sample lived in an urban area, self-reported a middle-class or higher income, reported at least some university education, and identified as homosexual/gay. We fitted models with one through 10 classes (Table 2). The BIC indicated a six-class model, and the AIC indicated an eight-class model; thus, we considered models with six, seven, and eight latent classes. Based on interpretability and class separation, we selected the seven-class model (Table 3). We present the classes in order from most to least common. The most common class, termed Multiple Same-Sex Behaviors, contained 31% of participants. This class had a high probabilities of insertive and receptive anal intercourse, having multiple male partners, not being in a single, primary relationship, and using condoms some of the time. The next most common class, termed Not Currently Active (28%), was marked by low probabilities of engaging in any type of sexual behavior in the past 3 months. Receptive Anal (16%) was characterized by high probability of receptive, but not insertive, anal sex; individuals in this class were also unlikely to have had sex with female partners, were likely to have used condoms all times they had sex, and were slightly more likely to have only one partner than to have had two or more partners. Same Sex Relationship (15%) was the only class marked by a high probability of being in a primary relationship with a single male partner; individuals in this class also generally engaged in both insertive and receptive anal sex, did not use condoms consistently, and were equally likely to have one or two or more male partners. Insertive Anal Intercourse (6%) included individuals who had a high probability of insertive, but not receptive, anal intercourse with male partners; in addition, they were likely to not be in a relationship and to use condoms consistently. Finally, two classes were marked by having a high probability of vaginal intercourse: The Same and Other Sex class (3%) had high probabilities of both penetrative sex with women and receptive and insertive anal sex with men; they were also unlikely to be in a relationship. The Female Only class (1%) had high probability of penetrative sex with female partners, but no anal intercourse with men; individuals in this class were split between not being in a relationship and being in a relationship with a woman.

Table 2. Fit Statistics for LCA Models of Past-3-Month Sexual Behavior with One Through 10 Latent Classes

1 Note. AIC = Akaike information criterion; BIC = Bayesian information criterion.

Table 3. Prevalence of Latent Class Membership and Item-Response Probabilities for Seven-Class Model of Sexual Behavior in Mexican MSM

2 Note. Boldfaced text indicated item-response probabilities greater than.5.

Table 4 shows the prevalence of membership in each latent class by sexual orientation. For gay participants, the most common profile was Multiple Same Sex Behaviors (35%), whereas bisexual men and men reporting some other sexual identity were most likely to be in the Not Currently Active class (28% and 36%, respectively). Bisexual men were more likely than the other sexual identities to be in the Same and Opposite Sex Class (12% compared to 1% for gay and 6% for other sexual orientation), whereas men reporting another sexual identity were overrepresented in the Receptive and Insertive Anal Intercourse classes.

Table 4. Class Membership by Sexual Identity

Finally, we examined predictors of class membership, with the Multiple Same Sex Behaviors class as the reference group (Table 5). Thus, these analyses tested whether a particular predictor was associated with greater or lower odds of being in a particular class compared to the Multidimensional Same Sex Behaviors class. All predictors except social support were significantly associated with class membership in overall tests. Each additional year of older age was associated with 14% greater odds of being in the Same-Sex Relationship class compared to the Multiple Same Sex Behaviors class. In general, syndemic factors were more strongly associated with membership in the Multiple Same Sex Behaviors class compared to other classes. Specifically, having higher depressive symptoms was associated with lower odds of being in the Same-Sex Relationship (odds ratio [OR] =.95) and Insertive Anal (OR =.97) classes compared to the Multiple Same Sex Behaviors class. Alcohol problems were associated with significantly lower odds of membership in the Not Currently Active (OR =.55), Receptive Anal (OR =.34), and Same-Sex Relationship (OR =.65) classes compared to the Multiple Same Sex Behaviors class. Child sexual abuse was associated with lower odds of being in all classes compared to the Multiple Same Sex Behaviors class (ORs ranging from.54 to.86). Similarly, intimate partner violence was associated with lower odds of being in all classes compared to the Multiple Same Sex Behaviors class, with the exception of the Relationship class, which was not significantly different. Sexual compulsivity was associated with lower odds of being in all classes compared to Multiple Same Sex Behaviors, except for the Other Sex Only class. Sexual Compulsivity was associated with lower odds of being in all classes compared to Multiple Same Sex Behaviors (ORs ranging from.64 to.86).

Table 5. Correlates of Latent Class Membership

• 3 Note. LL = Log likelihood.
• 4 *p <.05, **p <.01, ***p <.001.

Greater self-acceptance was associated with 1.26 times greater odds of being in the Same-Sex Relationship class (OR = 1.26). Although the overall log-likelihood test was not significant, bivariate comparisons indicated that individuals with greater self-acceptance were more likely to be in the Same and Opposite Sex class compared to the Multiple Same Sex Behaviors class (OR = 1.25).

This study examined patterns of sexual behaviors in a sample of young MSM in Mexico. We found classes marked by a number of distinct patterns of behaviors, with the most common including both insertive and receptive, often condomless, anal sex with more than one partner in the past 3 months, and a nearly as large class marked by no anal or vaginal intercourse in the past 3 months. Smaller classes included individuals who engaged in only either receptive or insertive anal sex, or who were in a same-sex primary relationship. In addition, two very small classes included men who had sex with both men and women or only women. In general, these patterns are very similar to research on patterns of behavior in a comparable sample of young men in the United States (Vasilenko et al., [57]). However, there were a few differences in terms of the proportion of individuals in each latent class. For example, among young men in the United States, the insertive and receptive anal sex classes were roughly equal in size, whereas in the Mexican sample, the receptive class was considerably larger. This may suggest age-varying positioning behaviors among Mexican MSM, in which younger individuals are more likely to take a receptive role (Jeffries, [33]). In addition, the same-sex relationship class was somewhat larger than was observed in the United States. This is consistent with research on patterns of sexual behaviors in heterosexual Mexican adolescents, in which classes marked by committed relationships were more common compared to similar studies in the United States (Espinosa-Hernández & Vasilenko, [20]), and may reflect cultural values emphasizing the importance of relationships. There was little difference in the prevalence of engaging in sexual behaviors with men and women and only women (Vasilenko et al., [57]), suggesting that men with same-sex attractions are generally not engaging in heterosexual behaviors to demonstrate their masculinity to a greater extent than in the United States. Overall, Mexican MSM in this primarily urban sample engaged in patterns of behaviors largely similar to those in the United States.

Membership in these classes differed by individuals' sexual identity. Gay men were most likely to be in the Multiple Same Sex Behaviors class, whereas men with some other sexual orientation (e.g., straight, questioning) were most likely to be in the Not Currently Active class; bisexual men were relatively evenly split between these two groups. This may suggest that men who have already established a gay identity may be more likely to engage in same-sex behaviors, whereas men who may still be questioning their sexuality or same-sex attractions are more likely to abstain from sexual behavior, and may be beginning to explore their sexuality through online interactions prior to engaging in same-sex behaviors. Similarly, gay men were also more likely to be in the class marked by being in a same-sex relationship. Not surprisingly, bisexual men and men with other sexual orientations were more likely to be in the two classes involving sex with women than gay men. Men with some other sexual orientation were also the most likely to be in the insertive only class. This is consistent with cultural standards suggesting that being the insertive partner is viewed as not being "gay," that is, more active and masculine, and not necessarily at odds with heterosexuality (Jeffries, [33]).

Results also showed differences in health and well-being correlates by class membership. For example, individuals in the Multiple Same Sex Behaviors class had significantly highest odds of most syndemic factors, including IPV, child sexual abuse, and a higher overall syndemic score, compared to those in most other classes. Because this class included both insertive and receptive sex with a high likelihood of condom nonuse, individuals in this class may be at greater risk for HIV, STIs, and other adverse health outcomes. These risk behaviors may be driven, in part, by their higher level of experience with sexual abuse and intimate partner violence, which are associated with greater sexual risk behavior (Buller, Devries, Howard, & Bacchus, [10]; Lloyd & Operario, [40]). Individuals in this class may also be more integrated into an urban gay culture in which substance use is common. Individuals in the insertive class had lower depressive symptoms compared to those in the Multiple Same Sex Behaviors class, which could be related to the greater societal valuing of the insertive compared to receptive role (Jeffries, [33]), leading men who engage only in insertive behaviors to feel less depressed. In addition, men in the Same and Other Sex class reported being more satisfied with their social support than men in the Multiple Same Sex Behaviors class. It is possible that men who are attracted to men but also engage in sexual relationships with women may present themselves as more traditionally masculine and seemingly heterosexual, and thus may feel less in conflict with their families than men who are exclusively gay.

The class most consistently associated with health and well-being outcomes was the Same-Sex Relationship class. In addition to being lower than the Multiple Same Sex Behaviors class on child maltreatment, sexual compulsivity, and alcohol use, individuals in the Same Sex Relationship class had the lowest depressive symptoms and highest level of self-acceptance. This is consistent with research suggesting that being in a same-sex relationship is associated with higher self-acceptance and lower homophobia among young MSM in the United States (Bauermeister et al., [5]; Isay, [32]). Romantic relationships may be uniquely protective for sexual minority youth. For instance, romantic partners may be navigating similar experiences of stigma and could provide each other with emotional support (Russell & Consolacion, [51]). Our findings highlight the role of romantic relationships in well-being among young Mexican MSM, which underlies the importance of formalizing relationships through legal marriage.

This study has important public health and prevention implications. First, results suggest that young MSM who use sexual networking websites or apps may be an important population for prevention efforts, as they often reported risk behaviors for HIV and STIs, such as condomless sex with multiple partners in the past 3 months. However, this study also suggests heterogeneity in patterns of behavior, which may have differential health risks and may require different prevention messaging. For example, we identified one class in which men reported a male primary partner, but individuals in this class were also likely to have had other partners. Research in the United States has suggested that men in relationships may not find messages aimed at MSM in general to be relevant (Hoff, Chakravarty, Beougher, Neilands, & Darbes, [31]); thus, it is possible that men who are in relationships but have sex outside of their relationship may need different prevention messages. Similarly, messages aimed at MSM in general may not be relevant to men who identify as bisexual (Dodge et al., [19]); thus, different messages may be needed for individuals who engage in sex with both men and women or who have a bisexual or heterosexual identity. Approaches such as adaptive interventions (Collins, Murphy, & Bierman, [16]) may be a way to tailor programs to the needs of individuals with particular patterns of behavior. For example, a program implemented through smartphones or the Internet may collect information about individuals' prior patterns of sexual behavior, then provide intervention content that is tailored to, and changes with, what is most relevant to the individual at a given time.

There are a number of limitations of this study that provide areas for future research. First, because this sample was collected through social and sexual networking websites for MSM, it may not be generalizable to all MSM in Mexico. The majority of young adults in Mexico do have smartphones or Internet access (Poushter, [47]); however, the men in this sample primarily lived in urban areas and were relatively well educated. Thus, we were unable to examine differences by education and social class. In addition, because these sites cater to men looking to meet new partners, we may have underestimated the prevalence of men in relationships, making it difficult to examine the heterogeneity of individuals who are in relationships, which is an important area for future research. Despite these limitations, this data collection approach enables researchers to obtain large samples of MSM who are searching for new partners and may be engaging in sexual risk behavior, making such websites potentially important for research and prevention efforts (Biello et al., [8]).

We were not able to examine differences among some sexual identities, which were less prevalent, such as separating out individuals who were heterosexual, questioning, or other identity. Future research should better document patterns of sexual behavior within these groups. In addition, the study did not have measures of constructs related to degree of outness or other aspects of the coming out process, which may be associated with sexual behaviors. Our study was cross-sectional, and thus we know little about how individuals may move in and out of these classes over time, or the directionality of the associations between the health and well-being correlates and patterns of sexual behavior. In addition, analyses were somewhat exploratory due to both the analytic method and the lack of theoretical models addressing the multidimensionality of sexual behavior, and future work could attempt to replicate/refine this study, as well as to develop more specific theoretical or conceptual models on this topic. Data were also collected a few years before legalization of gay marriage in Mexico, and future research could better examine whether this event has led to any behavioral changes. In this analysis we focused only on young MSM who were HIV-negative, and future research should also examine patterns of risk and protective behavior, including factors like antiretroviral use and sexual behaviors, among men living with HIV infection.

Despite these limitations, this study advances our understanding of sexual behavior among young MSM in Mexico in a number of ways. First, it provides important descriptive information on this population by examining a range of sexual behaviors using a large sample. Second, it applies an innovative approach to understanding the heterogeneity of sexual behavior among young MSM and documents a number of distinct patterns of behavior. These include patterns of sexual behavior marked by engaging in a range of same-sex behaviors with more than one partner, as well as individuals who were sexually inactive, had sex with women, and were in same-sex relationships. Findings demonstrate the importance of promoting safe sex practices among this population, and providing information that can be used to target individuals at highest risk and tailoring interventions and prevention messages to be more relevant to MSM with different behavior profiles. In addition, it demonstrates many health and well-being correlates of being in romantic relationships for young MSM, demonstrating potential benefits that may come from legalization of same-sex marriage in Mexico.

No potential conflict of interest was reported by the authors.