Most HIV infections in Zambia and Rwanda happen in marriage: prevention programmes for couples recommended

Mathematical modelling of Zambian and Rwandan populations, based on the rates of extramarital sex and the proportion of couples who are HIV-serodiscordant, shows that the proportion of HIV infections acquired in any one year that are acquired within marriage or in stable cohabiting relationships ranges from 55% to 93%.

The model, published in The Lancet therefore suggests that, at this point in time of the African epidemic, most infections are occurring within marriage. The authors therefore urge that prevention programmes concentrate on counselling and testing couples rather than individuals.

For the frequency of marital and extramarital relationships within the general urban population, and also condom use statistics, the authors used data from Demographic and Health Surveys (DHS) dating from 2001-02 (Zambia) and 2005 (Rwanda).

For the prevalence of serodiscordance within relationships, the model used hard data from voluntary counselling and testing services targeted specifically at couples in the capital cities of Lusaka and Kigali. In these programmes, partners are present at counselling sessions both before and after HIV tests, with simultaneous disclosure of results, result-specific counselling, training in condom use, and appropriate referrals to care.

Excluding from the DHS people from rural areas, those who had never had sex, and those who had been abstinent for over a year, the authors ended up analysing the likely serodiscordance and HIV acquisition risk for 4061 individuals; 1739 women and 540 men in Zambia and 1176 women and 606 men in Rwanda.

They estimated the probability that each person had acquired HIV during the previous year from any sexual partner and divided these into infections acquired prom the primary partner and extra-marital partners.

As in all mathematical models, the results obtained depend on the assumptions fed in. In this model, the authors did not attempt to reflect real-world scenarios exactly. They asked themselves: if you take a sample of serodiscordant couples, both cohabiting and non-cohabiting, casual and regular, what are the chances that over the course of the year, the positive person passes on their HIV to the negative one?

Importantly, they assumed for the sake of simplicity that couples where both partners were HIV-negative stayed that way during the year, and so excluded the effect of new incident infections in one partner. They also assumed that the likelihood of transmission between partners did not vary throughout the year. In real-life-scenarios, of course, this would not happen; some partners would acquire HIV from outside the relationship and HIV risk would vary according to factors such as pregnancy, changes in frequency of sex, sexually transmitted infections and the viral load of the positive partner. And finally, they assumed that the positive partner was undiagnosed.

They used data from Zambia and Rwanda that indicate a transmission rate of 20-25% a year in cohabiting serodiscordant couples who are untested for HIV, and set their HIV transmission probability at 20%. They also tested the model to see if rates varying from 10% to 40% made a difference.

Other variables within the model included condom efficacy (the degree to which, given documented condom use levels, they actually prevented infection, with rates varying from 25% to 85%); the degree to which HIV serodiscordance was greater in cohabiting than non-cohabiting couples (ranging from half as common to twice as common); and the average number of non-marital sexual partners per year, ranging from 0.5 to five.

The DHS statistics showed that more people were in their 20s than any other age group, except for Rwandan men, where the largest group were in their 30s. The Zambian group were on average younger, with 14% of women and 10% of men under 20 compared with 4% of Rwandan women and 2% of men.

Just over half of Rwandan men and women were married compared with just under two-thirds of Zambian men and three-quarters of Zambian women, but far more also said they were ‘living with someone’. These differences may reflect different cultural patterns or just the way researchers asked the question. The vast majority had only one partner during the year, whether cohabiting or non-cohabiting, and the proportion who reported one cohabiting partner closely tracked the proportion of people who reported that they were married or living together. The way the surveys were conducted could not rule out that a proportion of these people had one martial partner and one regular non-marital one.

However for all but Zambian men sexual activity with more than one non-cohabiting partner was rare; only 1% of Rwandan women and 2% of Zambian women and Rwandan men reported this. Zambian men bucked this trend, with over 10% reporting two non-cohabiting partners and 5% three or more.

Condom use was quite uncommon with spouses or cohabiting partners, with about one in ten Zambians and one in 20 Rwandans using condoms with their spouses. In contrast condom use was high with non-cohabiting partners; around 60% in the men, 50% on Zambian women and 35% in Rwandan women (who were the least likely to have non-marital sex). The DHS surveys only asked about condom use at last sex, so these figures may overestimate consistent condom use.

The VCT centre surveys showed that in Zambia, the woman was positive and the man negative in 9% of cohabiting relationships and 16% of non-cohabiting relationships. The man was positive in 8% of cohabiting relationships and 11% of non-cohabiting relationships. In Zimbabwe, the woman was positive and the man negative in 7% of cohabiting relationships and 11% of non-cohabiting relationships. And the man was positive and the woman negative in 6% of cohabiting and 4% of non-cohabiting relationships.

Therefore, in somewhat more relationships it was the woman who was positive, which reflects demographics, and serodiscordance was somewhat more common in non-marital or quasi-marital relationships.

The likelihood of HIV infection within all sexual relationships, marriage/cohabiting relationships and non-cohabiting relationships was then derived from these figures by multiplying the frequency of serodiscordancy within each group and type of relationship by the relative proportion of relationships that were cohabiting and non-cohabiting.

Given these data, what proportion of HIV infections would be transmitted within marriage and living-together partnerships? The answer is, in most cases, the vast majority. In Rwanda nearly 93% of HIV infections transmitted by either men or women occurred within marriage. For Zambian women the figure was 77%. In Zambian men, both because they have more extramarital relationships and because they are more likely to meet a serodiscordant partner within those relationships, the answer comes out as just over half – 55%.

If a programme of HIV prevention for couples were to reduce the annual transmission likelihood from 20% to 7% it would produce significant reductions in the proportion of HIV infections acquired or transmitted in marriage and therefore their absolute number. This proportion would decrease in Rwanda to 65% in women and 60% in men, and in Zambia from 50% in women and 36% in men. Current levels of condom use make very little difference, because they’re not often used in marriage; the figures were almost identical whether researchers factored in a reduction in transmission due to documented condom use or not.

Variables that did make a difference to the model were if there was a higher or lower risk of HIV transmission through casual sex contact than through marital sex, and if people had more non-marital/cohabiting sex partners. If the likelihood of a Zambian man getting/passing on HIV through casual sex was twice what it was in marital sex, for instance, the proportion with HIV who got/received HIV from/to their wives fell to 42%. And if he had five casual partners as well as a cohabiting one during the year, then within-marriage HIV transmission fell to a quarter of the total.

As the authors point out themselves, this model gives higher figures for within-marriage HIV than retrospective epidemiological surveys have done because it concentrates solely on relationships that are already serodiscordant. If, for instance, as some surveys have done, one were to establish the source of HIV infection for each individual in a sample of couples, it would lead inevitably to the conclusion that more that 50% of HIV is caught outside marriage – because one person in each couple had to catch it from someone else first.

In a mature epidemic that has passed its early rapid growth phase, however, the majority of HIV may well be being passed on in marriage – hence the call for better prevention interventions for couples.

An accompanying editorial criticises Dunkle and colleagues’ model for two reasons. Firstly, the data on serodiscordance is derived from voluntary HIV counselling and testing centres, and people presenting themselves for testing might have better reason than most to suspect they have HIV. Secondly, it relies on population data from several years ago and doesn’t factor in the possible impact of antiretroviral treatment or male circumcision. However antiretrovirals and circumcision would only serve to increase the effectiveness of a comprehensive package of HIV prevention measures for couples.

Above all, the editorial concludes, Dunkle and colleagues’ paper is an argument for higher rates of HIV testing in Africa. In a continent where 80% of people with HIV do not know their status and 90% do not know their partner’s status, prevention for couples can’t start to work until there is a huge scale-up of testing.