Key points are not available for this paper at this time.
Summary and conclusions In all regions of sub-Saharan Africa, the predominant mode of transmission of HIV is sexual intercourse between men and women, but there are large variations in the rate and extent of the spread of HIV in different populations. We set out to try and identify factors that could explain why HIV has spread more rapidly in some parts of Africa than in others. To approach this question, we used a study design that combined ecological comparisons across populations with individual-level analyses within populations. The ecological analysis allowed us to explore the population-level effects of risk factor profiles that may be difficult to identify from individual-level analysis. Our study was carried out in four cities in sub-Saharan Africa. To increase the chances of detecting significant factors that could explain the differences in HIV prevalence between the cities, we selected four cities with widely contrasting levels of HIV infection: Cotonou (Benin), Yaoundé (Cameroon), Kisumu (Kenya), and Ndola (Zambia). It is very unlikely that the differences in HIV prevalence between these four cities are due to differences in time since the start of the HIV epidemics, and we believe that they are the result of differences in the rate of spread of HIV 1. We tested the hypothesis that high-risk sexual behaviour and/or factors that enhance the transmission of HIV during sexual intercourse are more common in Kisumu and Ndola than in Cotonou and Yaoundé. In each city, we interviewed and tested a representative sample of 600-900 men and 900-1000 women from the general population, as well as a representative sample of 300 sex workers 1. The questionnaires were pre-tested and standardized, and contained questions on socio-demographic characteristics and sexual behaviour, including questions about characteristics of spouses and non-spousal partners. Consenting study participants were tested for HIV and for other sexually transmitted infections (STIs) that have been proven or hypothesized to enhance the transmission of HIV, including syphilis, HSV-2 infection, chlamydial infection, gonorrhoea and trichomoniasis. Table 1 summarizes the ecological comparisons between the four cities. The outcome variable was the prevalence of HIV infection (high versus low), and the exposure variables included prevalence of sexual behavioural risk factors, prevalence of other STIs and prevalence of male circumcision. The sexual behaviour patterns that were explored have been shown in other studies, either empirically (high rates of partner change, contacts with sex workers) or by mathematical modelling (concurrent partnerships, mixing between different age and sexual activity groups), to be associated with an increased risk of HIV infection at the individual level or an increased risk of explosive HIV epidemics at the population level. In this study, high rates of partner change and being married were found to be risk factors for HIV infection in men in at least one city, high rate of partner change was found to be a risk factor for HIV infection in women in all four cities, and sex in exchange for money was found to be a risk factor in women in Yaoundé 2. However, high rates of partner change, contacts with sex workers, concurrent partnerships and large age differences between partners were not more common in the two high HIV prevalence cities than in the two low HIV prevalence cities 3-5. In fact, with the possible exception of contacts with sex workers, these risk behaviours were more common in Yaoundé, one of the low HIV prevalence cities, than in the high HIV prevalence cities. Levels of condom use reported by men were similar in the four cities, but women in the low HIV prevalence cities reported less frequent condom use than women in the high HIV prevalence cities 6. Condom use reported by sex workers did not show a difference between the low and the high HIV prevalence cities 3. The only parameters of sexual behaviour that distinguished Kisumu and Ndola from Cotonou and Yaoundé were age at sexual debut of women and age at first marriage of men and women. Compared with the low HIV prevalence cities, women in the high HIV prevalence cities began sexual activity at a younger age, and men and women got married earlier 5. But the differences in age at first sexual intercourse were not large enough to explain by themselves the differences in HIV prevalence in the age group younger than 20 years or in the whole population. Moreover, we did not find young age at first sexual intercourse to be an independent risk factor for HIV infection in young women, although it is biologically plausible that sexual intercourse at an early age increases the transmission of HIV 2. Early marriage would tend to shift exposure to an HIV-infected spouse and infection from this spouse to younger age groups, but the differences in age at first marriage are unlikely to explain a large part of the differences in HIV prevalence between the cities.Table 1: Summary of the distribution of risk factors for HIV infection across the four populationsThere are several possible explanations for the lack of association between prevalence of HIV infection and prevalence of (most) sexual behavioural risk factors at the ecological level, although such associations exist at the individual level. This ecological bias could be due to temporal ambiguity 7 (i.e., changes in sexual behaviour due to the impact of AIDS morbidity and mortality and prevention programmes). However, we compared different age groups in the four cities and did not find evidence for changes towards safer sexual behaviour in the high HIV prevalence cities 8. Also, differences in reporting of sexual behaviour between the high HIV prevalence cities and the low HIV prevalence cities are unlikely to fully explain the lack of association between HIV prevalence and sexual behavioural risk factors at the population level 8. The most plausible explanation is that the ecological association between HIV prevalence and sexual behavioural factors is confounded by factors that influence the transmission of HIV during sexual intercourse and that are responsible for differences in the effect of sexual behavioural parameters on the risk of HIV infection 9. In the low HIV prevalence cities nearly all men were circumcised, whereas in the high HIV prevalence cities the majority of men were not circumcised 10. Analysis of data from Kisumu confirms findings from other studies that male circumcision is associated with a decreased risk of acquisition of HIV infection by men 11. In addition, ulcerative sexually transmitted infections and trichomoniasis were more prevalent in Kisumu and Ndola than in Cotonou and Yaoundé (Table 2)12,13.Table 2: Prevalence of sexually transmitted infections in the general populationThe role of trichomoniasis in the transmission of HIV is not clear. A few studies only have provided evidence for an enhancing effect of Trichomonas vaginalis infection in the sexual transmission of HIV 14,15, and more research is needed on the interaction between trichomoniasis and HIV infection. With regards to the interaction between ulcerative STIs and HIV infection, there is convincing evidence that ulcerative STIs substantially enhance the vulnerability of non-HIV-infected individuals and the infectiousness of HIV-infected individuals 16. In this study, there was a strong and consistent association between HIV infection and herpes simplex virus type 2 (HSV-2) infection, even after adjusting for the confounding effect of sexual behaviour 12. The prevalence of HSV-2 infection was higher in Kisumu and Ndola than in Cotonou and Yaoundé, but the difference between Ndola and Yaoundé was rather small (55 and 51% in women, respectively). However, in the age groups younger than 25 years, the differences in prevalence rates were larger (40 and 26% in women in Ndola and Yaoundé, respectively). These differences cannot be explained by differences in sexual behaviour because reported rates of partner change were lower in Ndola than in Yaoundé. There are several other possible explanations. First, men and women in Ndola get married at an earlier age than men and women in Yaoundé, and being married was found to be a risk factor for HSV-2 infection 12. Secondly, the proportion of dually (HIV and HSV-2) infected individuals in the population (and among partners of young people) is higher in Ndola than in Yaoundé, and they are likely to transmit HSV-2 more easily than are HIV-negative HSV-2-infected individuals 17. The prevalence of syphilis was over 10% in Ndola, around 6% in Yaoundé, and less than 5% in Kisumu and Cotonou. Combined prevalences of HSV-2 or syphilis seropositivity in men were 13% in Cotonou, 29% in Yaoundé, 37% in Kisumu and 40% in Ndola. The corresponding figures for women were 31, 52, 69 and 58%. HIV-1 subtype A was the most prevalent subtype in one high HIV prevalence city (Kisumu), as well as in the two low HIV prevalence cities. Subtype C was the predominant subtype in the other high HIV prevalence city (Ndola). This does not imply that the high HIV prevalence in Ndola is the result of a higher transmissibility of subtype C, as this subtype has been found in the past in Cameroon. This suggests that differences in circulating subtypes of HIV-1 are not a major factor in determining the rate of spread of HIV in sub-Saharan Africa 18. We conclude that, in the four African populations we studied, differences in risky sexual behaviour are out-weighed by differences in factors influencing HIV transmission probability (i.e., lack of male circumcision and ulcerative STIs, in particular HSV-2 infection and syphilis). We hypothesize the following mechanism for the explosive HIV epidemics in Kisumu and Ndola. In the early stages of the epidemic, the prevalence of HIV infection increases gradually, due to risky sexual behaviour patterns in combination with factors that enhance the transmission of HIV (i.e., lack of male circumcision and other sexually transmitted infections). As the proportion of men and women in the population that are dually (HIV and HSV-2) infected increases, the average per-sex-act transmission of HIV and HSV-2 is enhanced and both the HIV and the HSV-2 epidemics reinforce each other. We cannot pretend to provide the final answer to the question why the HIV epidemics in sub-Saharan Africa are so heterogeneous. First of all, the limitations of ecological analyses are well known 7. Second, for practical reasons, we were only able to conduct this cross-sectional study in four cities and we cannot exclude the possibility that our results are chance findings due to the small sample size of our ecological study (n = 4). It would be premature to extrapolate our findings to other populations in sub-Saharan Africa without further research. Several other population-based studies have been conducted or are under way using the same protocol and standardized questionnaire as was used in the present multicentre study. A study among young people in Carletonville, South Africa seems to confirm the important role played by factors that enhance the transmission of HIV in the general population 19. Third, we did not study all possible factors that have been found or hypothesized to be associated with enhanced transmission of HIV and we cannot exclude that other factors also play an important role. For instance, we did not study genetic factors determining resistance to HIV infection. However, resistance to HIV infection associated with genetically determined immune responses has been found in few individuals only 20. Deletion of the CCR5 receptor gene, which has been found to be associated with decreased risk of HIV infection in homozygous individuals, has been found in only 0.1% of South Africans of African descent and 1.4% of HIV-negative sex workers in West Africa 21-23. Research into HIV receptors and specific immune responses is very active, but so far no genetic factor has been identified that confers protection against HIV infection and is prevalent enough in certain populations to be a major determinant of the rate of spread of HIV. Our study, however, has once more highlighted the importance of factors other than sexual behaviour in shaping HIV epidemics in sub-Saharan Africa. Work with mathematical models is ongoing to confirm the results of our study. Implications for prevention interventions The implications of these findings for the design of interventions are complex. There is now convincing evidence for a protective effect of male circumcision against the acquisition of HIV infection, and promotion of male circumcision should be considered as a strategy to reduce the spread of HIV. In Western Kenya, research is underway to assess the impact of a campaign of promotion of circumcision on the spread of HIV in the population, after it was established that male circumcision was acceptable for the traditionally non-circumcising population 24. More research of this kind is needed. Risks and benefits of the promotion of male circumcision will have to be weighed against each other, the major risk being that safe sex practices decrease if circumcision is perceived to provide full protection against HIV infection. Any programme aimed at promoting circumcision will have to be carefully evaluated. With regards to HSV-2 infection, more research is needed on the interaction between HIV and HSV-2. At present, there are no specific, feasible interventions to reduce the incidence and prevalence of HSV-2 in developing countries, apart from condom use and changes in sexual behaviour. Widespread suppressive treatment with acyclovir may not be a feasible option for sub-Saharan Africa because of the high HSV-2 prevalence rates and the need to take acyclovir for an extended period. A vaccine is not yet available, but research is ongoing 25. A vaccine against HSV-2 may be available sooner than a vaccine against HIV, and its impact on the spread of HIV could be large. When conducting phase III trials with an HSV-2 vaccine, it may be worthwhile including HIV infection as an end point, as well as HSV-2-related endpoints 26. Effective management of other treatable STIs, including syphilis, remains an important priority. Our findings, however, do not imply that changes in sexual behaviour do not play an important role in curbing the HIV epidemics. We found high rates of partner change to be associated with an increased risk of HIV infection at the individual level. In Uganda, which borders Western Kenya, the HIV prevalence among young pregnant women has decreased in recent years. This decrease is attributed to a delay in onset of sexual activity of young people, an increase in condom use and a decrease in the number of non-regular sexual partnerships reported by young men 27. We also believe that the slow spread of HIV from sex workers to the general population of Cotonou is at least in part due to interventions among sex workers. The high levels of HIV infection among young people, especially among female adolescents in Kisumu and Ndola, highlights the importance of interventions targeted at young people and their partners. We have evidence that young women in all four cities have under-reported their sex partners. More qualitative research will be conducted on the sexual behaviour of youth in the four cities. This research will be carried out in close collaboration with organizations that work with youth and it aims to identify further areas for intervention.
Buvé et al. (Wed,) studied this question.