Overlapping relationships: how important is long-term concurrency?

But HIV prevalence in populations is not related in any simple way to the average number of sexual partners people have. Differences in HIV prevalence between countries or ethnic groups in the same country cannot be explained this way.

One survey, for instance, found that men in Thailand and in Rio de Janeiro actually had a higher average number of lifetime sexual partners than men in four African countries (Tanzania, Kenya, Lesotho and Zambia), and women had low numbers of partners in any country, with very few having more than five a year.¹

Similarly, Daniel Halperin^2,3 compared three mid-1990s surveys of men in Thailand, the US and Uganda. He found that the proportion of men who said they had had more than ten lifetime sex partners was over 60% in Thailand, over 40% in the USA, and just over 20% in Uganda; yet HIV prevalence among men in those countries was respectively 2.2, 1 and 18% at the time the surveys were conducted.

Why, then, have the countries of Africa and especially southern Africa, an area with 2% of the global population but a third of the world’s HIV-positive people, ended up with HIV prevalence an order of magnitude greater than any other countries in the world?

Attempts to explain southern Africa’s HIV prevalence struggle to find anything unique about the area. Condoms in Africa have historically been used in a minority of sexual encounters⁴ but, except as a result of campaigns targeted at sex workers and their clients, this is also the case throughout much of the world.⁵ Surveys have found very high levels of sexually transmitted infections (STIs) amongst African men and women, but untreated and especially asymptomatic STIs are almost as common in other countries.⁶

Differences in the proportion of men who are circumcised may explain some of the differences between, for instance, southern and western Africa or between ethnic groups in single countries.⁷ It does not explain why countries outside Africa where circumcision is uncommon, such as Thailand or southern Europe, have not seen generalised epidemics along African lines, given that some risk behaviours (for instance, the proportion of men who pay for sex) are actually higher in those countries.

There is one pattern of sexual behaviour, however, that is much more common in Africa - and especially in southern Africa - than anywhere else in the world. Martina Morris of the University of Washington compared relationship patterns in countries ranging from South Africa to Uganda to patterns in Thailand and the US and found that having more than one regular partner at the same time was common in African countries.⁸

For instance, she found that Ugandan men reported fewer lifetime sexual partners than Thai men, but while the Thais usually combined one primary relationship with one-off encounters with sex workers, Ugandan men often had two or sometimes three long-term sexual relationships. In some groups, this has been codified into overt polygamy, but the pattern was more often that a man would have a wife but would also have a long-standing girlfriend, separated either geographically (for instance a wife in the home village and a girlfriend in the city, or vice versa) and temporally (for instance, a girlfriend during work hours and a wife at weekends).

In a number of cultures it has been common for wealthier men to keep both a wife and a mistress, but this has only ever applied to a minority of the population. Men who cannot afford two women might visit sex workers, but one-off or occasional casual sex encounters do not have the same potential for spreading HIV. The length of concurrency in Africa (where concurrent relationships overlapping for more than a year are common) is one difference with other cultures. The other, crucial difference is that, while in other cultures only men would tend to have long-term, concurrent relationships, research suggests that in certain African countries a significant minority of women do so too.

Carael, bringing together a number of studies¹ found that 13% of men in Kenya, 22% in Lusaka, Zambia, 36% in Ivory Coast and 55% of men in Lesotho had long-term, concurrent sexual relationships with at least two women. The highest proportion of men operating the same lifestyle anywhere else surveyed was in Rio de Janeiro, where 7% had concurrent relationships, but in four Asian locations– Thailand, Singapore, Sri Lanka and Manila in the Philippines – no more than 4% of men had concurrent relationships.

Women had a fairly accurate idea of whether their men were ‘seeing someone else’ in some places. When asked, the proportion of women in Thailand, Manila and Lusaka who were asked if their man had another long-term relationship agreed with the actual total to within a few per cent. Women in Ivory Coast, Tanzania and Kenya thought their men were less faithful than they actually were - while women in Lesotho underestimated the scale of unfaithfulness going on. But in no case did the women over- or underestimate ‘concurrency’ in their men by more than 25%.

However, in order for concurrent relationships to amplify HIV transmission in heterosexual epidemics, women have to be involved in them too; otherwise a chain of infection, from man to woman to man and so on, cannot be sustained. The proportion of women who had concurrent, long-term partnerships with more than one man was lower than the proportion of men who had them. But it was still about an order of magnitude higher in African countries than in other countries. Eleven per cent of women in Lusaka, Zambia, 9% in Tanzania, and no less than 39% in Lesotho had long-term relationships with two or more men, compared with 3% in Manila, 1% in Sri Lanka and less than 1% in Rio and the other Asian countries surveyed.

The degree to which women had concurrent relationships was greatly underestimated by their male partners; in all but one of the countries surveyed, the proportion of women involved in long-term concurrent relationships was more than double the proportion of men who thought their woman was involved in one.

One of the original messages of safer sex awareness was: “When you sleep with one person, you are sleeping with all their partners”. In cultures where serial monogamy is the norm (as in most of the developed world), both men and women may have a higher lifetime number of sexual partners than in other cultures, but one relationship is usually finished before another is started. Even if HIV is transmitted within a relationship, it essentially reaches a cul-de-sac as long as that relationship lasts. It is possible that HIV transmission via heterosexual sex is quite a rare event except in facilitating circumstances such as an exceptionally high viral load in the transmitting partner and, by the time the relationship has ended, the newly-infected partner will generally have seroconverted and will have passed the point at which they are maximally infectious. 

The same applies in cultures where men do have extramarital sex, but either as one-off assignations with sex workers or with mistresses or girlfriends who are themselves monogamous. Clearly in these circumstances men may acquire HIV during extramarital sex and pass it on to their wife, but if she is monogamous, HIV again reaches a cul-de-sac. Serial monogamy traps the virus within a single relationship for months or years.

In contrast, in situations where a significant proportion of both men and women have concurrent relationships, even if they only have two partners each, as soon as one person in the network of concurrent relationship contracts HIV, then all the others are at risk, both because more people are more often exposed to the virus and because recently infected individuals have many-fold higher viral loads.

In Lesotho, for instance, according to a national reproductive health survey conducted in 2002,³ 20% of men and nearly 10% of women reported having two or more sexual partners during the past four weeks, and in a 2005 survey in South Africa among young people aged 15 to 24, 40% of men and almost 25% of women reported having more than one current sexual partner.

Mathematical modelling and real-life observations have shown that small increases in the average number of concurrent relationships in a culture can have a huge effect on the size of sexual networks in that culture. As the average number of concurrent partners increases, islands of inter-related people link together until huge sexual networks arise, sometimes involving the majority of sexually active people in a locality, and it does not take a large increase in the number of concurrent sexual partners to link a large number of people together.

Martina Morris,⁹ for instance, has calculated by mathematical modelling that, where the mean number of concurrent partners in a population is 1.68, the largest single sexually-interlinked network comprised 2% of the local population. When the mean number of concurrent partners increased to 1.86, no less than 64% of the sexually-active population became linked in a single network.

Morris’s mathematical model was confirmed by Stephan Helleringer and Hans-Peter Kohler of the University of Pennsylvania,¹⁰ who have been documenting the existence of a large sexual network among the population of Likoma Island in Lake Malawi. Having surveyed everyone in the seven villages on the island who was in a sexual partnership, they found that because of concurrent relationships, two-thirds of the 1070 people surveyed were all connected to each other in a sexual network.

On the other hand, in many countries in western Africa, the pattern of extramarital sexual relationships is more like that seen in Asia. Surveys from Ghana, for instance, have found that extramarital relationships tend to be conducted with commercial sex workers, who form the only distinct group in the country with southern-African levels of HIV prevalence. HIV prevalence among sex workers in the capital, Accra, is in the order of 60 to 80%, whereas among the general population it is just over 2%, and among pregnant women 4%.² Contrast this with Mashonaland in Zimbabwe where the HIV prevalence in sex workers is in the region of 60% – but the prevalence among pregnant women is 35%. Not all African cultures are the same, and some may have had pre-existing sexual cultures that made them particularly vulnerable to HIV.

However, condom use tends to be less frequent in long-standing relationships. For example, in a 2003 Zambian survey¹¹, male truck drivers and uniformed personnel used condoms in over 80% of encounters with sex workers, 50% with regular partners, and virtually never with their wives. In a high-prevalence country like Zambia, it is therefore easy for a married woman who has only ever had sex with her husband to acquire HIV.

At the 2006 International AIDS Conference in Toronto, Suzanne Leclerc-Madlala of the University of KwaZulu Natal commented that the tradition of multiple sexual partners may have arisen in the area because it had cultural and economic value.¹² Concurrent relationships were originally based in patterns that assured the continuity of productivity in agricultural cultures, such as wife inheritance.

Daniel Halperin has commented that: “There are numerous social, cultural and economic reasons why multiple concurrent partnerships exist. In many societies, having multiple partners is a powerful signifier of masculinity, and a relatively wealthy man may even be expected to have more than one wife or girlfriend as long as he can afford to do so.”

The crucial difference in parts of Africa, as we have said, is that a lot of women also have concurrent partnerships.

Attitudes towards gender equality had an effect on the frequency of concurrent relationships in one South African study. It found that men who believed in greater gender equality were more likely to be monogamous while  women of the same belief were more likely to have multiple concurrent partners. The results suggest that beliefs about gender relations play a strong role in determining multiple concurrent partnering and HIV risk.¹³

Halperin² quotes comments from qualitative research interviews in South Africa as to people’s opinions of why concurrent relationships were so common. Men and women alike blamed men’s ‘natural’ promiscuity:

“Sometimes I think that men were born to be like that as well. Because it is just impossible to have one girlfriend. It simply does not happen. It is something that is impossible.” (Male, 26)

And women attributed their own unfaithfulness to revenge on promiscuous boyfriends:

“I cheated on him because I wanted revenge for what he had done to me. My friends and I went to a party in East London and there we decided to have new boyfriends. Although it was supposed to be a temporary arrangement, mine lasted longer until my boyfriend here found out that I was doing something like that.” (Female, 28)

But they acknowledged that, for women and even men, economic security and advancement were also factors:

“She has about seven different cars visiting her. And she says she does it for transport, so if she wants to go some place and one car is busy she has access to another. As well as money, so she can get money. She has men who get paid on the 15th, 20th, 30^th.” (Male, 20)

“Yes, for both men and women this is always a factor, that’s why you get young boys going after older women.” (Female, 28)

Research into concurrent relationships in Africa has until recently concentrated on heterosexual relationships. However, a degree of bisexual behaviour in men could clearly be a major contributor to higher rates of HIV incidence in this setting.

Investigators in Malawi, Namibia, and Botswana interviewed 537 men who had ever had sex with men about the gender of their sexual partners, partnerships and sexual risk behaviours.¹⁴

Just over a third of the men reported that they were married or had a stable female partner, and 54% said that they had had both male and female sex partners in the previous six months. Bisexual concurrency (having concurrent relationships with men and women) was common and was reported by 17% of men.

Analysis showed that having a concurrent relationship with both men and women was associated with 70% higher levels of reported condom use and a 60% lower likelihood of being 'out' to family. Bisexual concurrency was also twice as common in men who had paid men for sex.

The authors were “encouraged” that men in concurrent relationships with both men and women reported less sexual risk and higher levels of condom use than men who reported exclusively homosexual behaviour.

The investigators suggested that their findings should occasion a rethinking of the factors driving the HIV epidemic in the region.

Age-mixing intensifies the risk of concurrent relationships. Economic disparity and the hope of economic gain in relationships are often associated with age disparity between partners. It is easy to see that if, as in some cultures, the majority of people only have relationships with people of approximately their own age, and if, as it does in cultures as varied as gay America and heterosexual Uganda, HIV prevalence rises gradually with age and reaches a peak in the mid-30s, young people will not often come into contact with groups with higher HIV prevalence.

In cultures that age-mix, on the other hand, younger people (usually women, in heterosexual cultures) are much more likely to encounter HIV-positive partners. This can then dramatically reduce the average age at which people acquire HIV, and indeed we see that the peak years for HIV acquisition for women in southern Africa are in their late teens, some ten years younger than the peak years in men. For more on this topic, and an illustration of the impact of age-mixing in Swaziland, see Abstinence.   

Evidence that age-mixing makes a big difference to HIV acquisition comes from South Africa¹⁵ where a national survey found that women who reported that their sex partner(s) were generally less than five years older than them had an HIV prevalence rate of 15%, whereas women who had older partners had a prevalence of 30%.

Evidence that supports the importance of age-mixing also comes from a culture half a world away, in gay America. One theory as to why black gay men in America are so much more vulnerable to HIV than white gay men (despite having, on average, a slightly lower number of sexual partners and less use of recreational drugs) is that black gay men tend to age-mix more. Researchers in San Francisco¹⁶ found that black gay men were 45% more likely to have partners who were ten or more years older or younger than themselves than other ethnic groups. They were also 2.5 times more likely to have sex partners of the same ethnicity, which meant that upon sexual initiation, joining an already-connected network of sexual partners with high HIV prevalence was more likely.

If multiple partnerships and particularly concurrent ones have this degree of importance, some HIV prevention messages may need to be revised. HIV risk may be high in a network even when the average number of partners each person has is not very high. Daniel Halperin comments that: “Because most Africans do not have exorbitant numbers of partners, they may not realise how dangerous, especially in regions of high HIV prevalence, such behaviours actually are. In southern Africa, even people with only two lifetime partners – hardly high-risk behaviour by western standards – need to appreciate how risky that one extra partner can be if the relationships are long-term and concurrent.”

There is one paradox that one needs to be aware of in encouraging monogamous behaviour. In certain HIV-prevalence situations, it can make no difference to HIV incidence at all. For instance, in Zimbabwe and South Africa, steep declines in the incidence of bacterial STIs like syphilis and gonorrhoea have not been accompanied by declines in HIV. In these countries prevalence is so high that transmission within marriage or a monogamous relationship is extremely common. A monogamous marriage only reduces HIV transmission risk if both partners have the same HIV status, and in countries where HIV testing is the exception rather than the rule, promoting monogamy may in certain circumstances have the effect of spreading HIV from a core group of sexually active men and their female partners into the female population at large.

The importance of concurrent relationships as a primary factor in generalised HIV epidemics is not universally accepted. In two separate papers, Larry Sawers and Eileen Stillwaggon of the American University of Washington DC,¹⁷ and Mark Lurie and Samtha Rosenthal of Brown University, Rhode Island,¹⁸ have argued that:

• The most persuasive demonstrations of the impact of concurrency have come from mathematical models rather than empirical studies.

• Some of the assumptions used in mathematical models are questionable. For example, Sawers and Stillwaggon argue that both transmission rates and frequencies of sexual contact are overestimated, women are assumed to have the same sexual behaviours as men, and one-off encounters and casual sex are not taken into account.

• Concurrency is often vaguely and inconsistently defined in population studies, meaning that comparisons between studies in different places are problematic.

• There is no substantial evidence that levels of concurrency are significantly higher in Africa than elsewhere.

• In the few studies that have compared individuals’ participation in concurrent relationships and their HIV status, a consistent relationship has not been found.

This critique has stimulated a fierce debate. Proponents of the concurrency thesis argue that the critics’ analysis of the data is selective and that evidence from a wide range of sources supports the concurrency thesis. Moreover methodological limitations in the evidence should not justify a ‘do-nothing’ policy - it would be irresponsible for preventing programmes in Africa to ignore the issue.¹⁹

More specifically, Martina Morris²⁰ says that the critics have been misled by the complexity of the evidence needed to demonstrate how risky concurrency is. Studies that succeed or fail in showing a link between how many partners a person has and their likelihood of having HIV are no help in proving that concurrency spreads HIV. This is because concurrency is not a property of an individual, but of a network, and is therefore missed by studies that only look at individual risk factors:

“Concurrency is not a risk for the person who has concurrent partners, it is a risk for that person’s partners. Put another way, concurrency increases your risk of transmitting infection, not acquiring it. So the empirical signature is not the correlation between Person A’s behaviour and status, but between Person A’s behavior and their partner’s status.”