Role versatility among men who have sex with men in urban Peru.

Role versatility refers to the practice in which individual men who have sex with men (MSM) play both insertive and receptive sexual roles over time. Versatility has been thought to be relatively uncommon among Latin American MSM but possibly is rising. Versatility also has been shown to be a potentially large population-level risk factor for HIV infection. In this study we examine the correlates of versatile behavior and identity among 2,655 MSM in 6 Peruvian cities. Versatile behavior with recent male partners was found in 9% of men, and versatile (moderno) identity was reported by 16%. Significant predictors included high education, white-collar occupation, sex work, and residence in Lima. Age was not significant in any analysis. Since sex work is correlated negatively with other predictors, versatile men appear to comprise two distinct subpopulations. Insertive-only men appear to play a strong role in bridging the HIV epidemic between MSM and women.

Homosexual sex is largely distinguished from heterosexual in that either partner may play either role (insertive or receptive) during a single type of sexual act. This simple fact allows for two distinct patterns of sexual role to emerge among homosexuals: those of role segregation (individuals only engaging in one of the two roles over time) and role versatility (individuals engaging in both). Although this distinction can in fact be drawn for some sexual acts for any couple--whether same-sex or opposite--research largely has focused on anal sex between men, given its central role in the spread of HIV and other sexually transmitted infections (STIs). The prominence of role segregation among MSM in Latin America has been well documented, although role versatility also clearly has been present as long as investigators have been exploring the topic (Arboleda, 1995; Caceres & Jimenez, 1999; Carrier, 1976, 1989, 1995; Carrillo, 2002; Gutmann, 2003; Murray, 1995; Parker, 1999). Role-segregated behavior often is accompanied by role-segregated identities, marked by the terms activo (insertive, "top") and pasivo (receptive, "bottom"). Terms indicating a versatile role identity (e.g., moderno in Peru, internacional in Mexico) also have been present since at least the 1960s (Carrier, 1976). These terms suggest that the identity (and corresponding behavior) is viewed as less traditional, and perhaps is conceived of as a diffusion of gay-identified sexual norms from the United States and Western Europe, although the extent to which this reflects its true origin is debated (e.g., Carrillo, 2002; Gutmann, 2003; Murray & Arboleda, 1995; Parker, 1989). Anecdotal evidence from urban Peru points to a rise in moderno identity, which suggests that versatile behavior also may be increasing, although this is not automatically true--it may simply represent more behaviorally versatile men choosing to identify thus.

With the advent of HIV, the phenomena of role segregation and versatility have taken on additional public health significance. Mathematical modeling has demonstrated that patterns of role versatility and segregation can have a potentially large impact on HIV transmission within a population. Trichopoulos and colleagues (1988) first argued that the presence of versatile men in an MSM population should allow the virus to spread more easily through the whole population; given differences in transmissibility for insertive and receptive anal sex, such men are capable of both becoming infected efficiently while receptive and then transmitting efficiently while insertive. Role segregation, in contrast, should create bottlenecks to the efficient transmission of HIV through long chains of people, providing protection to the entire population. The population-level impact of versatility may be in operation even when versatility does not act as an individual risk factor--that is, versatile men are fueling a larger overall epidemic for everyone, while still remaining at lower individual risk than receptive-only men. Mathematical models exploring this phenomenon initially incorporated the assumption of no transmission from receptive to insertive partner (Van Druten, Van Griensven, & Hendriks, 1992; Wiley & Herschkorn, 1989). Subsequent evidence showed, however that unprotected insertive anal intercourse (UIAI) poses a non-negligible risk of infection for the insertive partner (Tovanabutra et al., 2002; Vittinghoff et al., 1999). More recent modeling work has shown that the effect of versatility remains large even when there is considerable risk from UIAI, as long as it is less than that for unprotected receptive anal intercourse (URAI; Goodreau, Goicochea, & Sanchez, 2005).

Role segregation and versatility represent a continually shifting framework around which male homosexual relations in Latin America are structured. Given the strong relevance of versatility to the epidemic vulnerability of MSM populations, we sought to identify the characteristics that distinguish versatile men from nonversatile men among one set of MSM communities in Latin America--those of urban Peru.

Method

Study Population

We collected data on sexual behavior from MSM in 6 larger cities in Peru (Arequipa, Iquitos, Lima, Piura, Pucallpa, and Sullana; Figure 1). These were selected based on higher HIV prevalence and incidence relative to other Peruvian cities, as found in the nationwide 2000 HIV Sentinel Surveillance. Males were eligible for the study if they were 18 + years old and reported sexual intercourse with at least one man during the previous year. Piura and Sullana are twin cities and were combined for the purposes of analysis.

[FIGURE 1 OMITTED]

Subjects were recruited by trained recruiters and peer educators using outreach work and snowball techniques over a 3-month period in each city. Recruitment occurred at venues known to be frequented by MSM (saunas, pornographic movie theaters and arcades, prostitution areas, discotheques, bars, beauty parlors, transvestite houses, and sport courts). Methods used to contact potential participants and to refer them to study clinics included posters, flyers, and informational meetings. The reliance on convenience sampling was unavoidable given the marginalized nature of homosexuality in Peru.

Sexually transmitted disease (STD) clinics from the Ministry of Health and from private nongovernmental organizations served as sentinel sites during the study period. At each site, one counselor explained the objectives of the study to potential participants and obtained informed consent for their participation in the questionnaire and blood testing. Those men who agreed and provided written informed consent underwent a structured interview using Computer-Assisted Self-Interviewing (CASI) to collect information about demographics, sexual risk behavior, sexual identity, and HIV testing and diagnosis. The characteristics of the three most recent sexual partners were described individually as were sexual practices with each partner. Counselors assisted participants in cases of illiteracy or lack of familiarity or comfort with computers. Physicians obtained a medical history and conducted a targeted physical examination for STIs. Peripheral venous blood samples were obtained for detection of HIV antibodies by ELISA with confirmation by Western Blot.

All participants received risk reduction counseling, condoms, and lubricants. Participants with diagnoses of HIV or STI received postdiagnosis counseling and education. Syndromic treatment was provided to those with curable STI, while HIV-infected men were referred for standard care, which in Peru at the time (2002) did not include antiretroviral therapy.

Definitions of Versatility

We use three definitions of role versatility. The first is self-identified moderno role. This stems from participants' response to the question, "What is your role in the bedroom?" (activo, pasivo, moderno): The question purposefully excluded a specific time frame to allow participants to report their overall self-perception. The second definition involves role reported during anal sex with men in the last 3 months. We refer to this as AAI role ("any anal intercourse"). The third definition is similar to the second, but is limited to acts not involving a condom--the UAI ("unprotected anal intercourse") definition. The UAI definition is the most epidemiologically relevant one, since UAI is the behavior in which HIV is most readily spread. The AAI definition also is useful for understanding versatility generally, however and for delimiting a set of men who may be likely to practice unprotected role versatility with unreported or future partners.

These definitions rely on reports of behavior with sexual partners in the last 3 months. For the sake of brevity and recall accuracy, however respondents were asked to provide detailed sexual behavior with only their last three partners in this time period. Those who had more than 3 partners in the last 3 months (48.6%) thus did not provide complete behavioral information for the last 3 months. In addition, some reported partners were female. Together this means that some men who were versatile among all recent male partners would appear to be role segregated. The analyses involving the AAI or UAI definitions of versatility were conducted separately for men reporting on 1, 2, and 3 male partners to avoid confounding the tendency toward versatility with the tendency for many or exclusively male partners.

Results

Descriptive Statistics 

The study enrolled 2,703 men. Of these, 2,655 are included for analysis; 43 (1.6%) did not complete the survey, and an additional 5 did not respond to questions on the sex of their partners. An unknown number of men were screened out at the many informal outreach venues.

Table 1 provides basic demographic and behavioral information on the sample. Although the majority of respondents declared their sexual identity as homosexual or bisexual, 10.3% declared themselves heterosexual. Self-described current sexual role was evenly divided between activo (41.8%) and pasivo (40.6%); only 16% declared themselves to be moderno (versatile). Moderno identity ranged from a high in Lima (56.5%) to a low in Piura/Sullana (7.2%). Behavioral bisexuality is common among these men; 35.0% reported a female partner among their most recent 3 partners in the last 3 months.

Predictors of Moderno Identity

In Table 2, we consider the predictors of moderno identity for the entire study population using multiple logistic regression. Thirteen additional men who did not answer all relevant questions are excluded from this and subsequent analyses. We see a significant and consistent increase in moderno identity with increasing education. An independent effect is seen for higher-status occupations (professional, employee [empleado], university student) versus lower-status ones (worker [obrero], taxi driver, hairdresser). Engaging in sex work also increases the odds of being moderno. Travel abroad and to other regions domestically were included in the analysis, given the belief that moderno identity in Latin America represents cultural diffusion from the north. Nevertheless, these predictors were not significant, perhaps partly because they correlate highly with education and occupation.

Surprisingly, age (reclassified into five categories) shows no significant effect on moderno status. We can examine the issue further with a simple graph of all three role identities by age (Figure 2). Although moderno status remains relatively constant across all ages, the two types of segregated role identities (activo and pasivo) reverse, with activo being more common among younger men and pasivo among older men. This suggests that the set of men who are moderno at each age may differ from one another; at younger ages, they are selecting moderno identity largely in contrast to activo, while in older ages it represents men who choose moderno identity in contrast to pasivo. Because of this pattern, we repeated the analysis of moderno predictors from above for each of the five age classes separately (not shown). The qualitative results do not change; the same predictors appear, although some lose significance among the smaller age categories.

[FIGURE 2 OMITTED]

Predictors of Versatile Behavior

We run the same model using AAI and UAI as outcomes instead of moderno identity. Because of the confounding with reported partner numbers discussed above, these analyses are stratified by number of male reported partners. The qualitative results are again the same as for moderno identity, so Table 3 lists only the significant predictors and the magnitude of the effect for these analyses. Overall, the predictors among these different subsets remain generally the same as those for moderno identity; higher education, higher-status occupation, and residence in Lima are positive predictors of role versatility. When the factors are limited to unprotected sexual behavior, differences in education remain significant, despite the fact that one might expect increased versatility to be offset by increased condom use among this population. Having a serious male partner is either significant or close to significant in most analyses, suggesting that men may be more inclined to be versatile with their serious partners than with others.

Sex Work, Transvestites, and Versatility

Some of the predictors of versatility were themselves negatively correlated: sex work and high status as marked through education and occupation. This suggests that the set of men who are versatile may be a heterogeneous group, with men following two different paths to versatility: one opportunistic (sex work) and one reflecting a cosmopolitan or international perspective of educated, affluent men. To explore this, we rerun the analyses separately for sex workers and nonsex workers. We add an additional predictor for transvestite status, as this divides the world of sex work into two distinct types, widely assumed to relate to the client's desired sexual role (specifically insertive with transvestite sex workers). For the 10% of respondents who consider themselves sex workers, only residence in Lima is significant (OR 4.17); transvestite status was in fact not. For nonsex workers, the effect of education becomes even stronger than in the initial analysis--for instance, college-educated men are now 6.3 times more likely to be moderno than low-educated men, as opposed to only 4.4.

Versatility, Female Partners, and HIV Bridging

In this population, men who are exclusively receptive during UAI with male partners have a higher prevalence of HIV infection (16.6%) than either versatile (12.9%) or insertive (6.5%) men. As demonstrated by Goodreau and colleagues (2005), this pattern, where versatility is not an individual risk factor, is to be expected under some circumstances even though versatility is placing the population as a whole at greater risk. We also find that the number of female partner(s) reported on is largest for insertive-only men, intermediate for versatile men, and smallest for receptive men. Together, these points mean that it is not immediately clear who presents the greatest risk of bridging the HIV epidemic to females: insertive men (who have many female partners but low HIV), receptive men (who have few female partners but high HIV), or versatile men (who are intermediate on both). Table 4 shows a simple analysis in which we consider the question: On average how many women might a typical man in each role category be exposing? We simply assume the two effects (female partnering and HIV status, given role) are independent and take their product. Note that by doing this we are implicitly assuming that the men obtained HIV from their male partners such that their female partners are in turn at risk from them, which we suspect is most often the case among behaviourally bisexual men in this population.

We find that an individual man who is always insertive during UAI with other men is potentially exposing more women to HIV than is a man who is either receptive only or versatile; the lower HIV prevalence among insertive-only men is more than counteracted by their higher rates of female partnering. Insertive-only men comprise roughly one third of our sample, and this is likely a considerable underestimate of their true fraction in the MSM population, since our sampling venues disproportionately exclude married, heterosexually identified men. It appears, then, that a considerable fraction of the bridging of HIV from MSM to females occurs through men who are strictly insertive, although putting an exact number on this attributable risk is difficult given our sampling scheme.

Discussion

This study considered the predictors of versatile behavior among MSM in urban Peruvians. The versatile population appears to consist largely of two different subpopulations: one is educated, white-collar men of all ages and in all cities, although highest in Lima; the other is sex workers across the board, but, again, especially in Lima. Education and occupation remained as predictors when limiting the question to unprotected anal sex only; the higher levels of condom use that might be expected with higher education did not offset the high level of versatility among these men.

Previous work has identified the disproportionate impact on the HIV epidemic that versatile men can have, even under the counterintuitive scenario where they have relatively low HIV prevalence themselves (Goodreau et al., 2005). Changing the behavior of such men should thus have relatively large indirect effects for reducing HIV transmission overall. We believe that efforts to reduce versatility itself would be inappropriate and ineffective, however; this is the case regardless of the source and strength of role preferences, but it seems especially unhelpful in cases where versatility emerges from increasingly egalitarian norms about same-sex relations. Furthermore, the risk created by versatility is less to oneself than to one's potential future partners, who may not even be known yet; changing one's own behavior from versatile to exclusively receptive may in fact increase one's own immediate risk of infection. Rather, we propose that further behavioral interventions promoting higher condom use, lower rates of partner exchange, or both be targeted to subpopulations in which versatility is high. This work has demonstrated the existence of two such communities. Although numerous HIV interventions are already targeted at sex workers in Lima, relatively little focus has been placed on the well-off and highly educated, under the assumption that they already generally have received the message and are relatively safe. The fact that these sociodemographic characteristics remain as strong predictors for unprotected versatile anal sex indicates that this assumption is not correct. Interventions that are targeted specifically to the bars and discos that cater to such men may be especially helpful in lowering HIV rates for everyone given the disproportionate role that versatile men may play in sustaining an HIV epidemic.

Age was found to play surprisingly little role in predicting versatility, suggesting that versatility may not be as much on the rise as is popularly believed. An alternate (but less parsimonious) explanation, however, is that there are two countervailing effects in play: a period effect (probability of versatility rises as one ages) and a cohort effect (each subsequent generation begins its sexually active life more versatile than the previous). If both of these phenomena are present in roughly equal strengths, one would see little variation in versatility by age at any single time point, yet versatility would be continuously on the rise through time. Each of these two pieces is plausible, but distinguishing this more complex scenario from the simple case of complete stasis is not possible from this simple cross-sectional study. Doing so requires multiple similar studies, a longitudinal study, or additional qualitative work. Goodreau and colleagues (2005) reported on the first wave of a longitudinal study examining sexual roles among MSM; as further waves become available, the answer to this question should become clearer.

Finally, we identified that men who are insertive only with their male partners play a major role in bridging the HIV epidemic to their female partners. The general perception among MSM in Peru is that being exclusively insertive is a way to avoid HIV infection (Caceres & Rosasco, 1997), which is strictly true relative to being receptive and when focusing only on one's own infection. At the population level, however, the greatest risk to women from MSM may in fact be from insertive-only men, although given our sampling approaches an exact number is difficult to estimate. Nevertheless, it is clear that stronger messages are called for on the high indirect risk to a man's wife, girlfriend, or other female partners when he engages in unprotected insertive anal sex with men.

References

Arboleda, G. M. (1995). Social attitudes and sexual variance in Lima. In S. O. Murray (Ed.), Latin American male homosexualities (1st ed., pp. xvi, 304). Albuquerque: University of New Mexico Press.

Caceres, C. F. & Jimenez, O. G. (1999). Fletes in Parque Kennedy: Sexual cultures among young men who sell sex to other men in Lima. In P. Aggleton (Ed.), Men who sell sex: International perspectives on male prostitution and HIV/AIDS (pp. xix, 282). Philadelphia: Temple University Press.

Caceres, C. F. & Rosasco, A. M. (1997). The correlates of safer behavior among homosexually active men in Lima. AIDS, 11 (Suppl. 1), S53-59.

Carrier, J. M. (1976). Cultural factors affecting urban Mexican male homosexuality. Archives of Sexual Behavior, 5, 103-124.

Carrier, J. M. (1989). Sexual behavior and spread of AIDS in Mexico. Med Anthropol, 10(2-3), 129-142.

Carrier, J. M. (1995). De los otros: Intimacy and homosexuality among Mexican men. New York: Columbia University Press.

Carrillo, H. (2002). The night is young: Sexuality in Mexico in the time of AIDS. Chicago: University of Chicago Press.

Goodreau, S. M., Goicochea, L. P., & Sanchez, J. (2005). Sexual role and transmission of HIV type 1 among men who have sex with men, in Peru. J Infect Dis, 191(Suppl. 1), S147-158.

Gutmann, M. C. (2003). Changing men and masculinities in Latin America. Durham: Duke University Press.

Murray, S. O. (1995). Modern male homosexuality in Mexico and Peru. In S. O. Murray (Ed.), Latin American male homosexualities (1st ed.). Albuquerque: University of New Mexico Press.

Murray, S. O. & Arboleda, G. M. (1995). Stigma transformation and relexification: Gay in Latin America. In S. O. Murray (Ed.), Latin American male homosexualities (1st ed., pp. xvi, 304). Albuquerque: University of New Mexico Press.

Parker, R. (1989). Youth, identity, and homosexuality--The changing shape of sexual life in contemporary Brazil. Journal of Homosexuality, 17(3-4), 269-289.

Parker, R. G. (1999). Beneath the equator: Cultures of desire, male homosexuality, and emerging gay communities in Brazil. New York: Routledge.

Tovanabutra, S., Robison, V., Wongtrakul, J., Sennum, S., Suriyanon, V., Kingkeow, D., et al. (2002). Male viral load and heterosexual transmission of HIV-1 subtype E in northern Thailand. J Acquir Immune Defic Syndr, 29(3), 275-283.

Trichopoulos, D., Sparos, L., & Petridou, E. (1988). Homosexual role separation and spread of AIDS. Lancet, 2(8617), 965-966.

Van Druten, H., Van Griensven, F., & Hendriks, J. (1992). Homosexual role separation--Implications for analyzing and modeling the spread of HIV. Journal of Sex Research, 29(4), 477-499.

Vittinghoff, E., Douglas, J., Judson, F., McKirnan, D., MacQueen, K., & Buchbinder, S. P. (1999). Per-contact risk of human immunodeficiency virus transmission between male sexual partners. Am J Epidemiol, 150(3), 306-311.

Wiley, J. A. & Herschkorn, S. J. (1989). Homosexual role separation and AIDS epidemics--Insights from elementary models. Journal of Sex Research, 26(4), 434-449.

Jesus Peinado

Asociacion Civil Impacta Salud y Educacion, Lima, Peru

Steven M. Goodreau

University of Washington

Pedro Goicochea

Asociacion Civil Impacta Salud y Educacion, Lima, Peru

Jorge Vergara

Asociacion Civil Impacta Salud y Educacion, Lima, Peru

Nora Ojeda

Asociacion Servicios Generales dela Salud y Educacion, Piura, Peru

Martin Casapia

Asociacion Civil Selva Amazonica, Iquitos, Peru

Abner Ortiz

Asociacio Civil Cayetano Meredia, Pucallpa, Peru

Victoria Zamalloa

Instituto Sur Peruano de Infectologia, Arequipa, Peru

Rosa Galvan

Asociacion Civil Impacta Salud y Educacion, Lima, Peru

Jorge R. Sanchez

Asociacion Civil Impacta Salud y Educacion, Lima, Peru

Correspondence should be addressed to Steven Goodreau, University of Washington, Seattle, Washington. E-mail: [e-mail]

Table 1. Demographics of Study Population

                                                   Number   Percent

Total                                                2655
Age-median (range 18-71)                               24
City
  Piura/Sullana                                       738      27.8
  Lima                                                908      34.2
  Iquitos                                             304      11.5
  Pucallpa                                            278      10.5
  Arequipa                                            427      16.1
Self-reported sexual orientation
  Homosexual                                         1438      54.2
  Bisexual                                            936      35.3
  Heterosexual                                        273      10.3
  NA                                                    8       0.3
Self-described current sexual role
  Activo (Insertive)                                 1127      42.4
  Pasivo (Receptive)                                 1095      41.2
  Moderno (Both insertive & receptive)                432      16.3
  NA                                                    1       0.0
Sexual self-perception
  Hombre (man)                                        966      36.4
  Gay                                                 824      31.0
  Transvestite                                        389      14.7
  Buse (closeted, predominantly pasivo or             216       8.1
    moderno man)
  Mostacero (closeted, predominantly activo man)      159       6.0
  Other (woman, fete, none)                            95       3.6
  NA                                                    6       0.2
Number of sex partners (male and
  female, last 90 days)
  0                                                   142       5.3
  1                                                   668      25.2
  2                                                   554      20.9
  3+                                                 1291      48.6
Gender of all sex partners *
  Only male                                          1583      59.6
  Only female                                         211       7.9
  Both male and female                                719      27.1
  None                                                142       5.3
Relationships to sexual partners *
  At least one male "primary" partner                1028      38.7
  At least one female "primary" partner               549      20.7
  At least one male "one-time" partner               1425      53.7
  At least one female "one-time" partner              424      16.0
Role during unprotected anal
  intercourse with males **
  Only receptive                                      697      26.3
  Only insertive                                      496      18.7
  Both insertive and receptive                        116       4.4
  No unprotected anal intercourse                     993      37.4
    with male partners
  No male partners                                    353      13.3
Role during anal intercourse with males **
  Only receptive                                     1109      41.8
  Only insertive                                      840      31.6
  Both insertive and receptive                        246       9.3
  No anal intercourse with male partners              107       4.0
  No male partners                                    353      13.3
Sexually transmitted infections
  Genital Herpes Reactivity (by IR > 3.5)            1207      45.5
  Syphilis Reactivity (MHA-TP + & RPR +)              350      13.2
  HIV (Western Blot)                                  329      12.4

* For up to the 3 most recent sexual partners involving contact within
the last 90 days.

** For up to the 3 most recent male sexual partners involving contact
within the last 90 days.

Table 2. Global Predictors of Moderno Role Identity,
(Total n = 2,642; # Moderno = 429)

Predictor                        p     Odds ratio   95% C.I.

Age class                      0.838
  21-25 *                      0.598      1.09      0.79-1.49
  26-30 *                      0.259      1.22      0.86-1.74
  31-40 *                      0.578      1.11      0.77-1.61
  41+ *                        0.954      1.02      0.60-1.71
Education                      0.000
  Secondary **                 0.070      1.55      0.97-2.50
  Trade school **              0.000      2.72      1.61-4.58
  University **                0.000      4.41      2.60-7.48
Occupation                     0.000
  High status ***              0.000      2.40      1.74-3.31
  None/other ***               0.050      1.40      1.00-1.97
  Sex worker                   0.022      1.49      1.06-2.11
Has long-term male partner     0.005      1.39      1.10-1.74
Has long-term female partner   0.056      0.64      0.40-1.01
Any sex with women             0.006      0.57      0.38-0.85
Lima                           0.000      2.57      2.04-3.24
Travel abroad                  0.724      1.02      0.91-1.15
Domestic travel                0.673      0.97      0.86-1.10
Constant                       0.000      0.04      0.02-0.07

* Reference category = 18-20.

** Reference category = Less than secondary.

** Reference category = Low status.

Table 3. Significant Predictors for Behavioral Versatility

Outcome           n     n versa.   Significant predictors

AAI w/1         1,068      71      Education (trade school) *
male partner                       Education (university) *
                                   Lima

AAI w/2          494       60      Education (university) *
male partners                      Occupation (high status) **
                                   Has long-term male partner
                                   Lima

AAI w/3          736      115      Occupation (high status) **
male partners                      Lima

UAI w/1         1,068      40      Education (University) *
male partner                       Lima

UAI w/2          494       24      Has long-term male partner
male partners

UAI w/3          736       52      Has long-term male partner
male partners                      Travel abroad
                                   Lima

Outcome         Odds ratio (95'% CI)

AAI w/1         12.90 (1.67-99.85)
male partner    10.69 (1.33-85.78)
                 2.35 (1.37-4.04)

AAI w/2          5.24 (1.27-21.71)
male partners    2.45 (1.01-5.94)
                 3.19 (1.64-6.22)
                 2.97 (1.51-5.84)

AAI w/3          2.35 (1.25-4.40)
male partners    3.54 (2.19-5.72)

UAI w/l          8.93 (1.07-74.25)
male partner     2.18 (1.08-4.38)

UAI w/2          3.33 (1.22-9.08)
male partners

UAI w/3          2.19 (1.09-4.39)
male partners    0.27 (0.11-0.64)
                 2.47 (1.27-4.78)

* Reference category = Less than secondary.

** Reference category = Low status.

Table 4. Potential Exposure of Female Partners to HIV

UAI role with all the partners--last            HIV infection
90 days                                    (Western blot positive) (a)

                                                 n         %

Only receptive during UAI with male            116      16.6
  partner(s) (n = 697)
Only insertive during UAI with male             32       6.5
  partner(s) (n = 496)
Both insertive and receptive during UAI         15      12.9
  with male partner(s) (n = 116)

                                             Avg. # of female
UAI role with all the partners--last            partners in
90 days                                         last 3 mos.

                                              mean (95% C.I.)

Only receptive during UAI with male          0.31 (0.15-0.47)
  partner(s) (n = 697)
Only insertive during UAI with male          1.81 (1.56-2.05)
  partner(s) (n = 496)
Both insertive and receptive during UAI      0.63 (0.28-0.98)
  with male partner(s) (n = 116)

                                             Mean no. of females a
UAI role with all the partners--last        man in this category is
90 days                                   potentially exposing to HIV

Only receptive during UAI with male                   0.05
  partner(s) (n = 697)
Only insertive during UAI with male                   0.12
  partner(s) (n = 496)
Both insertive and receptive during UAI               0.08
  with male partner(s) (n = 116)

(a) Chi-square test < 0.00.