Epidimiology & Social
Concurrent sexual partnerships and HIV prevalence in five urban communities of sub-Saharan Africa
Lagarde, Emmanuel; Auvert, Bertran; Caraël, Michela; Laourou, Martinb; Ferry, Benoîtc; Akam, Evinad; Sukwa, Tome; Morison, Lindaf; Maury, Bertrandg; Chege, Janeh; N'Doye, Ibrahimai; Buvé, Annej; Cities, the Study Group on Heterogeneity of HIV Epidemics in African*
From the Institut National de la Santé et de la Recherche Médicale, Unité 88, Saint-Maurice, France, aUNAIDS, Geneva, Switzerland, the bInstitut National de Statistiques et d'Analyses Economiques, Cotonou, Benin, the cCentre Français pour la population et le développement/Institut de Recherche pour le développement, Paris, France, the dInstitut de Formation et de Recherche Démographique, Yaoundé, Cameroon, the eTropical Diseases Research Centre, Ndola, Zambia, the fLondon School of Hygiene and Tropical Medicine, London, UK, the gLaboratoire d'Analyse Numérique, Université Paris VI, France, hThe Population Council, Nairobi, Kenya, the iComité National de Lutte contre le Sida, Dakar, Senegal, and the jDepartment of Microbiology, Institute of Tropical Medicine, Antwerp, Belgium. *See Appendix 2.
Requests for reprints to: E. Lagarde, INSERM U88, Hôpital National de Saint-Maurice, 14 rue du Val d'Osne, 94415 Saint-Maurice Cedex, France
Received: 2 June 2000;
revised: 21 December 2000; accepted: 1 February 2001.
Sponsorship: The study was supported by the following organizations: UNAIDS, Geneva, Switzerland; European Commission, Directorate General XII, Brussels, Belgium; Agence Nationale de Recherches sur le SIDA/Ministère français de la coopération, Paris, France; DFID, London, UK; The Rockefeller Foundation, New York, USA; SIDACTION, Paris, France; Fonds voor Wetenschappelijk Onderzoek, Brussels, Belgium; Glaxo Wellcome, London, UK; BADC, Belgium Development Cooperation, Nairobi, Kenya.
Objective: To estimate parameters of concurrent sexual partnerships in five urban populations in sub-Saharan Africa and to assess their association with levels of HIV infection and other sexually transmitted infections (STI).
Methods: Data were obtained from a multicentre study of factors which determine the differences in rate of spread of HIV in five African cities. Consenting participants were interviewed on sexual behaviour and at four of the five sites also provided a blood and a urine sample for testing for HIV and other STI. Data on sexual behaviour included the number of partnerships in the 12 months preceding the interview as well as the dates of the start and end of each partnership. Summary indices of concurrent sexual partnerships – some of which were taken from the literature, while others were newly developed – were computed for each city and compared to HIV and STI prevalence rates.
Results: A total of 1819 adults aged 15–49 years were interviewed in Dakar (Senegal), 2116 in Cotonou (Benin), 2089 in Yaoundé (Cameroon), 1889 in Kisumu (Kenya) and 1730 in Ndola (Zambia). Prevalence rates of HIV infection were 3.4% for Cotonou, 5.9% for Yaoundé, 25.9% for Kisumu and 28.4% for Ndola, and around 1% for Dakar. The estimated fraction of sexual partnerships that were concurrent at the time of interview (index k) was relatively high in Yaoundé (0.98), intermediate in Kisumu (0.44) and Cotonou (0.33) and low in Ndola (0.26) and in Dakar (0.18). An individual indicator of concurrency (iic) was developed which depends neither on the number of partners nor on the length of the partnerships and estimates the individual propensity to keep (positive values) or to dissolve (negative values) on-going partnership before engaging in another one. This measure iic did not discriminate between cities with high HIV infection levels and cities with low HIV infection levels. In addition, iic did not differ significantly between HIV-infected and uninfected people in the four cities where data on HIV status were collected.
Conclusion: We could not find evidence that concurrent sexual partnerships were a major determinant of the rate of spread of HIV in five cities in sub-Saharan Africa. HIV epidemics are the result of many factors, behavioural as well as biological, of which concurrent sexual partnerships are only one.
Over the past decades a lot of progress has been made in our understanding of the epidemiology of sexually transmitted infections (STI), in particular HIV infection. Empirical studies have identified a number of behavioural risk factors for STI at the individual level. The role of sexual networks in determining the rate of spread of STI in populations has been studied mainly with mathematical models [1–8].
In the last 10 years, concurrency of sexual partnerships has been proposed by several authors as playing a critical role in the dynamics of HIV epidemics [3–7]. For the same number of sexual partners, overlapping partnerships would be associated with a more rapid spread of HIV than serial partnerships. Kretzschmar and Morris proposed a measure of concurrency (called k) derived from the contact graph representing the whole network of sexual relationships [5,7]. This index k measures the fraction of sexual partnerships that are concurrent at any given point in time. It has the property of converging to a simple function of the mean (μ) and the variance (σ2) of the number of partners (k = σ2/ μ+μ−1). Results from modelling exercises suggest that sexual concurrency increases both the intensity and the variability of the intensity of an HIV epidemic and that the final size of the epidemic increases exponentially as k increases. These results are independent of changes in HIV transmissibility over time (due to changes in viral load).
So far very few empirical studies have been carried out to test the above hypothesis. Potterat et al. found that having concurrent sexual partnerships was a risk factor for transmitting Chlamydia trachomatis infection in Colorado Springs, USA . Rosenberg et al. identified concurrent sexual partnerships as a risk factor for the acquisition of an STI among adolescents attending a public sexually transmitted diseases clinic in the USA .
One of the hypotheses we tested in the Multicentre Study on Factors Determining the Differential Spread of HIV in four African Cities was that sexual concurrency was more frequent in the cities with high HIV prevalence than in the cities with relatively low and stable HIV prevalence . This paper presents the comparison of the extent of concurrent sexual partnerships in the four cities of the Multicentre Study and Dakar (Senegal), where a survey on sexual behaviour has been done using the same questionnaire as was used in the Multicentre Study. In addition, concurrency of sexual partnerships was explored as a risk factor for HIV infection in the four cities of the Multicentre Study.
The methods of the Multicentre Study on Factors Determining the Differential Spread of HIV in four African Cities are described in more detail elsewhere . Briefly, the study took place in two African cities with a high prevalence of HIV infection (Kisumu in Kenya and Ndola in Zambia) and two cities with a relatively low and stable HIV prevalence (Cotonou in Bénin and Yaoundé in Cameroon). In each city a cross-sectional survey was performed on a representative sample of about 1000 men and 1000 women aged 15–40 years, selected from the general population. Consenting study participants were interviewed about their socio-demographic characteristics and their sexual behaviour, and were tested for HIV and other STI (herpes simplex virus-2, syphilis, gonorrhoea, chlamydial infection, trichomoniasis). A standardized questionnaire was used in the four cities. This study took place in 1997/1998. In 1998 a similar survey, using the same questionnaire, but that did not include STI detection, was conducted in Dakar (Senegal). It is known from other studies that the prevalence of HIV infection among pregnant women has remained low and stable over the years in Dakar .
The questionnaire on sexual behaviour included questions about current spouses (up to four spouses) and on non-spousal partners of the last 12 months (up to eight partners). The dates of the start and the end of the partnerships were recorded, as well as the number of sex acts.
For each respondent we computed the number of partnerships in the last 12 months and its variance, at several time points. From the mean number of partners and the variance we derived for each city the kappa index, as described by Kretzschmar and Morris . In addition we computed for each individual the sum of the time periods of overlapping partnerships. The kappa index and the mean time period of overlap between concurrent sexual partnerships were compared between the five cities.
In order to explore concurrency as an individual risk factor for HIV infection we created an individual indicator of concurrency called iic. This indicator was designed in such a way that it fulfilled four criteria: (i) iic summarizes the individual propensity to keep or dissolve on-going partnership(s) before engaging in another one; (ii) iic does not depend on the number of partnerships; (iii) iic does not depend on the length of partnerships; (iv) iic covers the 12-month period preceding the interview. For each pair of partnerships declared by a respondent, we computed the actual duration of the overlap d and the expected duration ∊ given the length of both partnerships. Computation details of ∊ are provided in Appendix 1. The ratio d/∊ varies between 0 and infinity. We obtained a symmetric measure by computing r = (d/∊ − 1)/(1+ d/∊). This index r varies between −1 and 1, is zero when the duration of the overlap is that expected by chance, is positive when partnerships of the pair are more concurrent than expected, and is negative when partnerships of the pair are less concurrent than expected. The indices r were summed for each individual to give the individual indicator of concurrency iic. Mean values of iic were compared between the cities and between HIV-infected individuals and HIV-uninfected individuals in each city. Where the sample size allowed it, iic as risk factor for HIV infection was explored further in multivariate analysis.
All data were double-entered and validated in EPIINFO. Further data cleaning and analysis was done using SPSS 8.0 for Windows (SPSS Inc. 1997, Chicago, Illinois, USA).
Ethical approval for the study was obtained from the national ethical committee in each of the countries where the study took place, as well as from the ethical committees of the Institute of Tropical Medicine, Antwerp, the London School of Hygiene and Tropical Medicine and The Population Council.
The total number of individuals interviewed ranged from 1730 in Ndola to 2116 in Cotonou. In all five cities the response rates for women were higher than for men (Table 1). The response rates for men were especially low in Yaoundé, Kisumu and Ndola. The vast majority of eligible men who did not participate in the study were never found at home despite repeated visits by the study teams . The prevalence of HIV infection in men in Cotonou was 3.3%, in Yaoundé 4.1%, in Kisumu 19.8% and in Ndola 23.2%; the corresponding figures for women were 3.4%, 7.8%, 30.1% and 31.9% respectively . The prevalence of HIV infection among pregnant women in Dakar was less than 1% in 1998 . The prevalences of chlamydial infections and gonorrhoea are given in Table 1.
Table 1 gives the main demographic characteristics of the study populations. In Dakar, Cotonou and Yaoundé fewer men and women were married than in Kisumu and Ndola. The proportion of men declaring non-spousal partners in the last 12 months was highest in Yaoundé (74%) and ranged between 41% and 51% in the other cities. The proportion of women declaring non-spousal partners was 49% in Yaoundé and between 13% and 21% in the other cities.
The measures of concurrency are given in Table 2. The concurrency index k at the time of interview was high in Yaoundé (0.98), intermediate in Kisumu (0.44) and Cotonou (0.33) and low in Ndola (0.26) and Dakar (0.18). Fig. 1 gives the values of k at different points in time in the 12 months preceding the interview. The values decreased slightly over time but the comparison between the five cities remained the same all along: the k index was not higher in Kisumu and Ndola, the two high HIV prevalence cities, than in the other three cities where the prevalence of HIV has remained relatively low and stable. The same applies to the mean duration of the overlap between partnerships and to the individual index of concurrency iic (Table 2). The mean duration of overlap is higher for men than for women, in all cities, mainly because men reported more partnerships than women. The highest mean value was found in Yaoundé (271 days for men, 52 days for women), and the lowest in Dakar (38 days for men and 4 days for women) and in Ndola (73 days for men and 1 day for women). The mean value of iic ranged between −0.65 among men from Ndola and 0.14 among women from Kisumu (Table 2).
The mean iic was not higher in HIV-infected individuals than in HIV-uninfected individuals (Fig. 2). We further explored iic as a risk factor for HIV infection where the sample size allowed it, namely among men from Kisumu and men from Ndola. Propensity to concurrent sexual partnerships (individual iic > 0) was found to be slightly associated with HIV infection but the association was not statistically significant when adjusted for other risk factors (OR = 2.02; CI : [−0.84–4.88] in Kisumu; OR = 2.96; CI : [0.96–9.11] in Ndola). The complete risk factor analysis of the study is described elsewhere .
Sample sizes in Yaoundé allowed us to compare iic means by gonorrhoeal and chlamydial infection status. Mean (95% confidence interval) for iic was −0.19 (−0.85 to 0.47) for those infected with gonorrhoea versus −0.01 (−0.19 to 0.17) for those who were not infected. Mean (95% confidence interval) for iic was −0.15 (−0.59 to 0.29) for those infected with chlamydial infection versus −0.001 (−0.19 to 0.19) for those who were not infected.
The lack of association between levels of concurrent sexual partnerships and STI/HIV infections could be related to higher condom use among those with overlapping partnerships. We compared the mean sum of overlap between those who reported having used condoms always or most of the time with all non-spousal partnerships with those who did not. None of both measures were significantly associated with condom use except among women of Cotonou. For them, the mean sum of overlap was lower among those who reported condom use (27 days) than among those who did not (111 days; P = 0.01).
We compared different measures of sexual concurrency in five urban populations in sub-Saharan Africa with different levels of HIV infection. We have evidence that the differences in HIV prevalence are unlikely to be explained by differences in the time since introduction of the virus in the population . In our study concurrent sexual partnerships were not more frequent in the high HIV prevalence cities than in the relatively low HIV prevalence cities.
The fraction of sexual partnerships that were concurrent, as estimated by the k index proposed by Kretzschmar and Morris, was not bigger in Kisumu and Ndola than in the three ‘low’ HIV prevalence cities. Morris and Kretzschmar simulated the spread of HIV in virtual populations with an index k ranging from 0 to 0.67 . In the five populations we studied k ranged between 0.18 and 0.98. It would be interesting to see simulation outputs with k varying from 0.67 to 0.98. Morris and Kretzschmar also showed that the higher the value of k, the greater the uncertainty about the size of the epidemic at the end of the simulation. This may account for the relatively low HIV prevalence in Yaoundé despite a high value of k. But in Ndola k is low which should lead to a low HIV prevalence with low uncertainty, yet the prevalence was 28.4%. Both spousal and non-spousal partners were included when computing k and iic because such differences in marriage patterns (age at first marriage and frequency of polygamy) may partly explain the lack of correlation between the measures of concurrency and HIV prevalence in the population. However, we found very similar results when excluding spouses: the five populations ranked the same as when spousal partners were included (results not shown). Moreover, Cotonou was the city with the highest percentage of men and women in polygamous union (16% of men were in a polygamous marriage, compared with 6–12% in the other cities)  and relatively low values of k and iic.
Cities ranked slightly differently when k and iic were compared. Concurrency as measured by k was more frequent in Yaoundé than in Kisumu, but the mean iic was higher in Kisumu than in Yaoundé. This is explained by the fact that iic is an individual index that does not take into account partners of partners while k is intended to describe a feature of the whole network. The two indicators do not reflect the same aspect of concurrency.
It is clear that the frequency of concurrent sexual partnerships is only one parameter of sexual behaviour that may determine the rate of spread of HIV in a population. In addition, factors that influence the probability of transmission of HIV during sexual intercourse play an important role. In fact, in the Multicentre Study on Factors Determining the Differential Spread of HIV in four African Cities, differences in sexual behaviour could not explain the differences in HIV prevalence and we concluded that differences in factors that enhance the transmission of HIV outweighed differences in sexual behaviour .
There are several potential limitations to our study. Data were collected for limited number of partnerships (up to four spousal partnerships and up to eight non-spousal partnerships). However, the impact of this limitation was marginal because only 13 respondents reported more than four spouses and only 15 respondents reported more than eight non-spousal partnerships. Another concern is that our data are drawn from populations that have already reached relatively stable levels of HIV infection. Currently reported sexual behaviour patterns and concurrency of sexual partnerships may differ from those at the start of the HIV epidemics. Infected people may have changed their sexual behaviour as a result of their disease, thus partly explaining why we found no differences in iic between infected and non-infected people. However, STI are still very prevalent in these populations and HIV levels are high in young age groups, suggesting that the process that feeds the epidemic is continuing . It was not feasible to restrict the analysis to young age groups because of the small sample sizes. In addition, parameters of concurrency may also have been underestimated in cities with high levels of HIV infection as those who died from AIDS are likely to have had higher number of partnerships than average . Another consequence of the choice of cities is they are all in a mature phase of their HIV epidemic. Models predict that concurrency impacts on the establishment of STI epidemics but not on their maintaining  and this may explain why we found no impact of concurrency in our data. However, Yaoundé exhibited the highest level of concurrency and HIV prevalence has been relatively low for several years. According to models, Yaoundé fulfils the criteria for the establishment of a major epidemic, which has not occurred so far.
The analyses of the data on contacts with sex workers suggest that men in Cotonou and Ndola underreported their contacts with sex workers to a larger extent than men in Yaoundé and Kisumu . This could have led to underestimating average values of iic in Cotonou and Ndola. This is, however, unlikely to alter our conclusion as Cotonou is a city with a low level of HIV infection. Similarly, the validity study  suggested that women tended to report many fewer non-spousal partners than men (the ratio ranged from 2.0 to 3.3); again, we do not believe this could alter our conclusion as the discrepancy was not consistently higher in any of the cities with similar HIV levels.
Simulation exercises have suggested that concurrent sexual partnerships may play an important role in determining the rate of spread of HIV in a population. Several authors have used an individual indicator of concurrency to assess the risk of STI at the individual level. Potterat et al. found that concurrency is associated with being a transmitter of C. trachomatis. Rosenberg et al. found concurrent sexual partnerships to be an independent risk factor for STI among adolescents attending a public sexually transmitted diseases clinic in the USA . We did not find any association between concurrency and risk of HIV infection. There are several possible explanations for this lack of association. In our study concurrency was assessed over the 12 months preceding the interview. HIV infection could, however, have been acquired several years before the interview and in the meantime changes in behaviour could have taken place that have resulted in a change in ‘propensity’ for concurrent sexual partnerships (e.g. getting married and/or reducing non-spousal partnerships). Secondly, Rosenberg pointed out that concurrent sexual partnerships are a risk factor at the individual level in as far as the partners also have concurrent sexual partnerships. If the concurrent partners do not have other partners in their turn the risk of STI is not higher than for someone who has the same number of partners consecutively.
In conclusion, we were not able to find evidence that concurrent sexual partnerships were a major determinant of the rate of spread of HIV in five cities in sub-Saharan Africa. HIV epidemics are the result of many factors, behavioural as well as biological, of which concurrent sexual partnerships are only one.
1. Anderson R, May R, Boily M, Garnet G, Rowley J. The spread of HIV-1 in Africa: sexual contact patterns and the predicted demographic impact of AIDS. Nature 1991, 352: 581 –589.
2. Potterat JJ, Rothenberg RB, Muth SQ. Network structural dynamics and infectious disease propagation. Int J STD AIDS 1999, 10: 182 –185.
3. Watts CH, May RM. The influence of concurrent partnerships on the dynamics of HIV/AIDS. Math Biosci 1992, 108: 89 –104.
4. Hudson C. AIDS in rural Africa: a paradigm for HIV-1 prevention. Int J STD AIDS 1996, 7: 236 –243.
5. Kretzschmar M, Morris M. Measures of concurrency in networks and the spread of infectious disease. Math Biosci 1996, 133: 165 –195.
6. Garnett G, Johnson A. Coining a new term in epidemiology: concurrency and HIV. AIDS 1997, 11: 681 –683.
7. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997, 11: 641 –648.
8. Ghani A, Swinton J, Garnett G. The role of sexual partnership network in the epidemiology of gonorrhea. Sex Transm Dis 1997, 24: 45 –56.
9. Potterat JJ, Zimmerman-Rogers H, Muth SQ. et al
transmission: Concurrency, reproduction number, and the epidemic trajectory. Am J Epidemiol 1999, 150: 1331 –1339.
10. Rosenberg MD, Gurvey JE, Adler N, Dunlop MBV, Ellen JM. Concurrent sex partners and risk for sexually transmitted diseases among adolescents. Sex Transm Dis 1999, 26: 208 –212.
11. Buvé A, Caraël M, Hayes R, et al. Multicentre study on factors determining differences in rate of spread of HIV in sub-Saharan Africa: methods and general population prevalence of HIV infection. AIDS
2001, (suppl) (in press).
12. Meda N, Ndoye I, M'Boup S. et al
. Low and stable HIV infection rates in Senegal: natural course of the epidemic or evidence for success of prevention? AIDS 1999, 13: 1397 –1405.
13. Auvert B, et al. Risk factor analysis for HIV infection in four population with different rate of spread of HIV: male circumcision and HSV-2 infection play a major role in explaining differences in epidemic spread. AIDS
2001, (suppl) (in press).
14. Ferry B, Caraël M, Buvé A, et al. Comparison of key parameters of sexual behaviour in four African urban populations with different levels of HIV infection. AIDS
2001, (suppl) (in press).
15. The Study Group on Heterogeneity of HIV Epidemics in African Cities . The Multicentre Study on Factors Determining the Differential Spread of HIV in four African Towns: summary and conclusions. AIDS
2001, (suppl) (in press).
16. Ewald PW. The evolution of virulence. Sci Am April 1993, 268 (4): 86 –93.
17. Morison L, Weiss HA, Buvé A, et al. Commercial sex and the spread of HIV in four cities in sub-Saharan Africa. AIDS
2001, (suppl) (in press).
18. Buvé A, Lagarde E, Caraël M, et al. Validity issues in the multicentre study on factors determining the differential spread of HIV in four African towns. AIDS
2001, (suppl) (in press).
Computation details for overlap duration expectancy ∊
Let A and B be two partnerships with durations a and b, respectively. The A and B partnerships are represented by the time intervals Ix= [α − a/2; x + a/2] and Iy = [ y − b/2; y + b/2]. The observation period is [0, L], where L is 12 months. We assume in what follows that a ≥ b. Participants were asked to provide information on all partnerships in the last 12 months. We are therefore interested in situations for which Ix and Iy have a non-empty intersection with [0, L]. This requires that: EQUATION Let Ω i ω Ω be the subdomain in which xΩ Iy ≠ φ (when A and B overlap): EQUATION The probability that partnerships overlap is: EQUATION Let us now introduce χi the characteristic function of Ω i (χi = 1 inside Ω; χi = 0 outside Ω), and (x; y) the length of Ix ∩ Iy whenever it is not empty. The expected duration of overlap is then: EQUATION A straightforward calculation gives: EQUATION The expected duration of overlap in the case b ≥ a is obtained by interchanging a and b.
A particular case occurred when partnerships were still ongoing at the time of interview. For the computation of ∊, the duration of these partnerships was doubled. Similarly, when an observed overlap covered the time of interview, the observed overlap duration d was doubled. Cited Here...
Members of the Study Group on Heterogeneity of HIV Epidemics in African Cities
A. Buvé (coordinator), M. Laga, E. Van Dyck, W. Janssens, L. Heyndricks (Institute of Tropical Medicine, Belgium); S. Anagonou (Programme National de Lutte contre le SIDA, Benin); M. Laourou (Institut National de Statistiques et d'Analyses Economiques, Benin); L. Kanhonou (Centre de Recherche en Reproduction Humaine et en Démographie, Benin); E. Akam, M de Loenzien (Institut de Formation et de Recherche Démographiques, Cameroon); S.-C. Abega (Université Catholique d'Afrique Centrale, Cameroon); L. Zekeng (Programme de Lutte contre le Sida, Cameroon); J. Chege (The Population Council, Kenya), V. Kimani, J. Olenja (University of Nairobi, Kenya); M. Kahindo (National AIDS/STD control programme, Kenya); F. Kaona, R. Musonda, T. Sukwa (Tropical Diseases Research Centre, Zambia); N. Rutenberg (The Population Council, USA); B. Auvert, E. Lagarde (INSERM U88, France); B. Ferry, N. Lydié (Centre français sur la Population et le Développement/Institut de Recherche pour le Développement, France); R. Hayes, L. Morison, H. Weiss, J. Glynn (London School of Hygiene and Tropical Medicine, UK); N. J. Robinson (Glaxo Wellcome R&D, UK, formerly INSERM U88); M. Caraël (UNAIDS, Switzerland). Cited Here...
concurrency; HIV; networks; sexual behaviour; transmission; Africa
© 2001 Lippincott Williams & Wilkins, Inc.
What does "Remember me" mean?
By checking this box, you'll stay logged in until you logout. You'll get easier access to your articles, collections,
media, and all your other content, even if you close your browser or shut down your
To protect your most sensitive data and activities (like changing your password),
we'll ask you to re-enter your password when you access these services.
What if I'm on a computer that I share with others?
If you're using a public computer or you share this computer with others, we recommend
that you uncheck the "Remember me" box.
Highlight selected keywords in the article text.
Data is temporarily unavailable. Please try again soon.