CONTEXT: Since 1990, HIV infection in Brazil has spread among the heterosexual population, particularly in the north. Containment of the epidemic can be informed by a better understanding of men's sexual risk behavior.

METHODS: Logistic, Poisson and multilevel logit models were applied to data on married and cohabiting men who had participated in the 1996 Brazilian Demographic and Health Survey.

RESULTS: Twelve percent of married or cohabiting men reported having had at least one extramarital partner in the previous 12 months; half of these had had two or more. The majority (77%) of partners were described as friends or lovers; 4% had been prostitutes and 15% strangers. Among men who had had sex with an extramarital partner in the last year, 40% reported having used condoms during last extramarital sex. Compared with members of evangelical religions, other men were significantly more likely to report having had an extramarital partner (odds ratios, 3.0-4.7) and unprotected extramarital sex in the last 12 months (3.4-7.9). Region of residence was also strongly correlated with extramarital sex: Compared with men in southern or central Brazil, those in the north had more than three times the odds of having had extramarital sex and unprotected extramarital sex in the last year (3.1-3.8).

CONCLUSION: In Brazil, religious affiliation and region of residence exert a major influence on risk behavior.

International Family Planning Perspectives, 2004,30(1):20-26

Although the spread of HIV has been slower in Latin America than in Africa, the pandemic is now a well-established and growing problem there. Brazil has the highest number of AIDS cases in Latin America, and HIV prevalence among adults is 0.7%, which is close to the average for Latin America. Initially, AIDS in Brazil was concentrated among welleducated homosexuals living in the large cities in the south, but since 1990, rates have been increasing among heterosexuals, the less educated and northerners. By 1998, 61% of sexual transmission of HIV occurred via heterosexual contact; the male-to-female ratio of those who were infected was 2:1, compared with 16:1 in 1986; and 66% of men who had contracted the infection through heterosexual contact had had multiple partners, compared with 27% of women.1

In contrast to the plethora of surveys on sexual behavior conducted in Africa, Europe and North America in response to the HIV pandemic, only a few surveys have been conducted in Latin America, particularly on men. One such survey is the 1996 Brazil Demographic and Health Survey (DHS), in which a nationally representative sample of men were interviewed on a range of topics, including extramarital sex.

Extramarital sex by men is likely to have different origins and consequences than risky sexual behavior by unmarried men, and thus merits separate analysis. The reasons for examining extramarital sexual behavior in the context of the spread of HIV and AIDS are obvious; however, as a four-site study2 sponsored by UNAIDS confirms, the actual risk of infection depends on the overall prevalence of sexually transmitted infections (STIs) and HlV in particular communities and networks, and on the prevalence of cofactors, such as ulcerating infections. The epidemiological data from Brazil show that HIV and AIDS are increasing in prevalence among heterosexuals and within marital relationships, thus making extramarital sex an increasingly important topic for investigation.3 Understanding such risk behavior is essential for sound policy and program development, allows interventions to be better designed, targeted and evaluated, and may aid in predicting the future spread of the disease.

This article uses 1996 DHS data to analyze the correlates of extramarital sexual partnerships among Brazilian males, focusing on the potential behavioral impact of the increasing popularity of evangelical religions. The majority of evangelical groups in Brazil are Pentecostal, and although Pentecostals are a diverse group, they share a belief that life should be centered around religion. Adherents tend to be devout, reject "sinful" behavior and are encouraged to have strict moral codes that challenge the conduct of the larger society.4 Pentecostalism is a relatively new religion in Brazil and, unlike Catholicism, is more a religion of conversion than of birth. There are no recent official statistics on the number of Pentecostals in Brazil, but it is widely acknowledged that the growth is explosive.5 Globally, the number of Pentecostals in the world is increasing by almost 2% a year, and it is estimated that there will be over 800 million Pentecostals by 2025.6

METHODS

Data

In the 1996 Brazil DHS, 3,986 men aged 15-59 were selected for interview and 74% were successfully interviewed. Among those not interviewed, 76% were absent, 12% refused and 12% were not interviewed for other reasons.7 The subsampIe used in our analysis consists of men who were married or living with a partner at the time of the survey. We refer to these men as married-that is, we make no distinction between common law and registered marriages, and thus refer to any sexual intercourse with partners other than those with whom the men lived as extramarital sex. The weighted number of such men was 1,667 (unweighted 1,640).

The male questionnaire collected data on men's number of sexual partners in the last 12 months, the identity of their last three partners, condom use during their most recent sexual act with their last three partners, and their knowledge and perceptions of HIV and AIDS. As the spouse was one of the last three partners for all but three of the men surveyed, information on partner characteristics and condom use is available for a maximum of two extramarital partners. For the analysis, we selected three main outcomes: whether the respondent had an extramarital sexual partner in the last year, the number of such partners and condom use with each partner.

Human sexuality and sexual conduct have been analyzed from many different disciplinary perspectives: biology, economics, psychoanalysis, gender studies, social anthropology and so on. Because sexual expression is such as a lundamental aspect of the human experience, it is unlikely that any single analytical framework will gain wide acceptance. Certainly, there is no agreed on or well-founded theory of sexual networking in married men, nor was the 1996 Brazil DHS designed to test specific causal hypotheses. Nevertheless, the survey collected information on a wide range of factors that permit an unusually thorough examination of the correlates of behavior.

The predictor variables selected for the analysis include demographic factors (current age), prior sexual and marital history (age at first intercourse, number of marriages and nature of current marriage), ecological factors (urban-rural residence and region of residence), socioeconomic characteristics (education, occupation, value of household assets [as an indicator of wealth], household headship) and religious affiliation. The rationale for this selection is that research in other countries has revealed them to be potentially powerful influences on sexual behavior. For instance, age at first intercourse and nature of the main partnership have emerged as predictors of extramarital networking in industrialized and developing countries;8 socioeconomic status is often positively associated with the rate of acquisition of extramarital partners, though the health risks may be offset by greater condom use;9 and religion has been found to be a critically important factor in some societies.10

The questionnaire also collected data on men's knowledge of HIV and AIDS, and their perceived risk of or vulnerability to HIV infection. Causal interpretation of any statisiical associations between these cognitive and attitudinal factors, however, is problematic. For instance, perceived risk of HlV infection may act as an influence on behavior, or it may be a consequence of recent behavior or may reflect both of these elements. Hence, these factors have been excluded from the main analyses.

Statistical Analysis

We used stepwise logistic regression to assess the correlates of having an extramarital partner in the last 12 months; Poisson regression was used to model the predictors of the number of such partners. We applied standardized sample weights (i.e., the inverse of the probabilities of selection corrected for nonresponse) and adjusted standard errors for both clustering and stratification using the survey command in Stata 7. For the analysis of condom use with the last two partners, a multilevel logit model was used to control for between-respondent variability, as well as to estimate any unobserved heterogeneity in the probability of condom use with these partners. We systematically examined correlates in the final models for significant firstorder interactions, but found none.