Fifty-five percent of women who participated in a study on gender-based violence and HIV/AIDS in Soweto, South Africa,1 reported that they had experienced physical or sexual abuse by a male partner. Those who had experienced both types of abuse or frequent violence were significantly more likely to test positive for HIV than those who reported little or no abuse, even when sexual risk behaviors common among women who have a history of gender-based violence were taken into account (odds ratio, 1.5). In addition, women who reported a large power imbalance in their current relationship had higher odds of testing positive than those who said the partnership was equal (1.5).

The findings are based on data from 1,366 women aged 16-44 who were recruited while seeking antenatal care at clinics in 2001-2002; those who elected to have a routine HIV test were included in the study. In structured interviews conducted before the women learned their HIV status, same-sex interviewers asked participants about their demographic characteristics; their experience of physical and sexual violence perpetrated by male partners, childhood sexual abuse and coercion at first intercourse; the level of male control in their current (or most recent) relationship; and a range of sexual risk behaviors. The researchers conducted multivariate logistic regression analyses to examine independent associations between these factors and HIV status.

Fifty percent of participants were married or living with a partner, 78% were in a relationship of at least a year's duration, 41% had had 12 or more years of schooling and 37% said that there was not enough to eat in their household. Thirty-four percent of the women tested positive for HIV.

Overall, 55% of participants reported that they had ever experienced physical or sexual violence perpetrated by a male partner; very few women reported sexual violence without also indicating some form of physical abuse. The majority of abused women indicated that the violence had occurred "a few times" or "many times" (which the researchers categorized as medium or high frequency). The women were roughly evenly distributed across a scale measuring low, moderate or high levels of power imbalance between women and their current partners. Sexual abuse in childhood, at first intercourse and by males other than partners were each reported by fewer than 10% of participants.

The prevalence of sexual risk factors varied: Roughly half of participants reported having had five or more male partners, never having used a condom or having had at least one casual partner; about one in five had ever received material compensation for sex. Approximately one-third of the women had ever drunk alcohol, but only a small proportion had problems with drug or alcohol use.

In preliminary analyses, several measures of intimate partner violence appeared to predict HIV infection: having experienced physical violence only, both physical and sexual violence, a medium or high frequency of either type of violence, or "broad violence" (a summary measure defined as both physical and sexual violence or one type at a high frequency). Reporting a large power disparity in the current sexual relationship and having participated in any of the sexual risk behaviors examined, except never having used a condom, were also associated with an increased likelihood of HIV infection. All of the risk behaviors—including never having used a condom—were linked to various forms of intimate partner violence, child abuse and power disparities in women's current relationships.

When demographic characteristics and sexual risk factors were taken into account, intimate partner violence and large power differentials continued to be strong predictors of HIV infection: Women who had experienced broad violence were significantly more likely than those who had experienced limited or no violence to have positive test results (odds ratio, 1.5), and participants who reported a large imbalance of power in their current partnership had elevated odds of HIV diagnosis (1.5).

The researchers acknowledge the limitations of cross-sectional and self-reported data, and note the imprecision of a number of measures. They suggest that because abused and disempowered women have an increased risk of HIV infection even after their own risk behavior is considered, abusive and controlling men may be more likely than other men to be infected with HIV or with other STIs that can increase women's susceptibility to HIV. The researchers conclude that "addressing problems of gender-based violence and HIV will require broad community and societal level transformations that challenge entrenched cultures of violence and male-dominated norms of gender relations."

The authors of an accompanying commentary point out that many major initiatives to stem the spread of HIV/AIDS—including the U.S. President's Emergency Plan for AIDS Relief—acknowledge the importance of addressing gender inequities, intimate partner violence and resultant challenges to HIV prevention.2 However, they note that many implementation strategies, such as those that focus on abstinence among young people, do not account for the circumstances and effects of gender-based violence and may therefore be of limited use to women who have been abused or disempowered. The commentators stress the need to "turn the global rhetoric into effective action" in order to make "a real difference in the lives of women and girls worldwide."—R. MacLean


1. Dunkle KL et al., Gender-based violence, relationship power and risk of HIV infection in women attending antenatal clinics in South Africa, Lancet, 2004, 363(9419):1415-1421.

2. Martin SL and Curtis S, Gender-based violence and HIV/AIDS: recognising links and acting on evidence, commentary, Lancet, 2004, 363(9419):1410-1411.