In urban areas of Rwanda and Zambia, most heterosexually transmitted HIV infections occur within marital or cohabiting relationships, according to a probability model that combined clinical data on couples' HIV status with population-based data on sexual behavior.1 The estimated proportion is particularly high in Rwanda, where more than 90% of such infections, among both women and men, may occur between cohabiting or married partners. The researchers estimate that voluntary testing and counseling interventions, which have been shown to reduce the rate of HIV transmission within couples, would prevent more than one-third, and perhaps as many as four-fifths, of heterosexually transmitted infections.

The new analyses incorporate data from several sources. Information on the types and durations of heterosexual relationships and on condom use at last sex were obtained from Demographic and Health Surveys conducted in Zambia in 2001–2002 and in Rwanda in 2005. The probability that one partner in a relationship, but not both, was infected with HIV (i.e., that the relationship was serodiscordant) was estimated using data collected in 2003–2005 by voluntary counseling and testing services in Kigali, Rwanda, and Lusaka, Zambia. The annual risk of transmission among serodiscordant couples was assumed to be 20%, reflecting older, preintervention data from Zambia and Rwanda. For each urban survey respondent who reported having had at least one sexual partner in the past 12 months, the researchers estimated the probability that he or she had acquired HIV during that time from a married or cohabiting partner and from a noncohabiting partner. From these values, the researchers created probability models to estimate the proportion of heterosexually transmitted infections that occur within marriage or cohabitation. In addition, because prior research has shown that joint voluntary testing and counseling interventions reduce the annual rate of HIV transmission within couples to just 3–7%, they estimated the proportion of heterosexually transmitted infections that could be prevented by such programs.

Analyses were based on data from 1,739 Zambian women, 540 Zambian men, 1,176 Rwandan women and 606 Rwandan men. In the year prior to being surveyed, 64–85% of respondents had had sex with a married or cohabiting partner, 16–51% had had sex with a noncohabiting partner and 1–15% had had sex with both types of partners. Levels of condom use at last sex were greater with noncohabiting partners (35–64%) than with cohabiting ones or spouses (5–11%).

In an estimated 9% of married or cohabiting couples and 16% of noncohabiting ones in Zambia, the female partner was HIV positive, but the male partner was not; in Rwanda, the corresponding values were 7% and 8%, respectively. Conversely, Zambian males were HIV positive, but their female partners were not, in an estimated 8% of married or cohabiting couples and 11% of noncohabiting ones; the corresponding values were 6% and 4%, respectively, in Rwanda.

An initial model that did not consider condom use predicted that more than half of heterosexually transmitted HIV infections occur in serodiscordant couples who are married or living together. The proportions varied by country and sex, ranging from 55% among Zambian males to 93% of both males and females in Rwanda. In analyses that took condom use (including the higher level of use by noncohabiting partners) into account and that assumed that condom use reduced the risk of HIV transmission by 50%, the proportion of heterosexually transmitted infections that occurred within marriage or cohabitation was slightly higher (60–94%).

Voluntary testing and counseling interventions that reduce the annual rate of HIV transmission within serodiscordant couples to 7% would prevent 36–50% of heterosexually transmitted HIV infections in Zambia and 60% of those in Rwanda. An intervention that reduced the annual transmission rate to 3% would prevent 51–80% of heterosexually transmitted infections.

In sensitivity analyses that tested a range of alternate assumptions about the annual rate of HIV transmission within married or cohabiting couples (10–30%), the annual probability of HIV transmission within noncohabiting couples (10–40%), the reduction in transmission risk resulting from condom use (25–85%) and the prevalence of serodiscordance (50–200% of the observed value), findings were largely similar to those described above. The results were also generally similar in analyses that varied the level of serodiscordance among couples and the number of partners a person had, with one exception: The proportion of new heterosexually transmitted infections occurring in marriage or cohabitation fell to 25–29% among Zambian men when extremely high or low levels of serodiscordance or numbers of noncohabiting partners were assumed.

The study's findings suggest that marital and cohabiting relationships pose a risk of HIV infection for men and women alike, the researchers note. They advocate collection of further data for couples of all types, and replication of the study in other populations to better define the proportion of new infections that occur within various relationship types. Although interventions that provide joint counseling to couples can reduce the risk of transmission, such programs are not the norm, the researchers observe. Moreover, services that focus on abstinence and fidelity without providing HIV testing will not be adequate for reducing transmission of the virus. Given the large potential reduction in new heterosexual infections that could be achieved by offering appropriate services, the researchers "call for increased promotion of voluntary counselling and testing for couples and for development and assessment of other interventions for couples that are both culturally and gender sensitive."—S. London


1. Dunkle KL et al., New heterosexually transmitted HIV infections in married or cohabiting couples in urban Zambia and Rwanda: an analysis of survey and clinical data, Lancet, 2008, 371(9631):2183–2191.