Article Text
PDF
Abstract
Background and aims Sexual violence (SV) is common during conflict. Despite reports of rape-related incidents of HIV infection, ecological analyses have found no association between SV and HIV at a population level. This has led to debate in the humanitarian, security and public health arenas about whether SV is an important HIV risk factor in conflict-affected settings. This paper uses published evidence on sexual violence in Africa and modelling to explore when SV may increase individual HIV risk and community HIV incidence.
Methods Publications on sexual violence in conflict settings were reviewed and a mathematical model describing the probability of HIV acquisition was adapted to include the potential effect of genital injury and used to estimate the relative risk of HIV acquisition in ‘conflict’ versus ‘non-conflict’ situations. An analytical equation was developed to estimate the impact of SV on HIV incidence.
Results A rape survivor's individual HIV risk is determined by potentially compounding effects of genital injury, penetration by multiple perpetrators and the increased likelihood that SV perpetrators are HIV infected. Modelling analysis suggests risk ratios of between 2.4 and 27.1 for the scenarios considered. SV could increase HIV incidence by 10% if rape is widespread (>40%); genital injury increases HIV transmission (threefold or more); at least 10% of perpetrators are HIV infected and underlying HIV incidence is low (<0.5%).
Conclusion The analysis illustrates that SV is likely to be an important HIV risk factor in some conflict-affected settings. More generally, it indicates the limitations of using broad aggregate analysis to derive epidemiological conclusions. Conflict-related initiatives offer important opportunities to assist survivors and prevent future abuses through collaborative programming on reconstruction, HIV and sexual violence.
- HIV
- rape
- sexual violence
- coerced sex
- transactional sex
- conflict
- health
- sexual assault
Statistics from Altmetric.com
Introduction
Sexual violence (SV), including gang rape, forced marriage and sexual slavery, is frequently part of the cruel reality of war.1 Although media reports often focus on mass rape,2 in practice, SV can take many forms. Women may be threatened or coerced into long-term sexual relationships by soldiers. Women and girls may also form sexual relationships with soldiers, border guards, peace keepers or even aid workers in exchange for food, shelter, protection or safe passage across borders. Although less reported, men are also raped during conflict.3
Until recently, there was significant concern that HIV would thrive during conflict periods in countries such as Liberia, Rwanda, Democratic Republic of Congo and Mozambique, and that widespread rape and sexual coercion would play a central part in the transmission of the virus.4 5 In Rwanda, for example, HIV infection levels as high as 70% were reported among rape survivors after the genocide.6 Epidemiological research on HIV from several non-conflict African settings also found evidence of increased HIV risk among women who had experienced SV,7–9 with a growing number of studies identifying a clustering of STI risk behaviours among men who are violent, suggesting that this increased risk is due to the use of force and the associated disruption to the genital epithelium and also to the increased likelihood that the SV perpetrator is HIV infected.10–12 Similarly, other factors commonly associated with being HIV infected have also been found to be connected to the perpetration of SV, including alcohol use; multiple sexual partners; unprotected casual sex; the likelihood of anal sex; STIs; age difference between sexual partners; access to and use of condoms (male and female) and exchanging money or goods for sex.13–16
Despite this identification of links between SV and HIV, a review by Spiegel et al (2004) concluded that there was limited evidence that conflict increases HIV risk, or that rape was associated with increases in HIV at a population level.17 Later, Spiegel et al (2007) argued that their data showed that conflict in general had no epidemiologically measurable impact and was not associated with population HIV prevalence.17 Subsequently, Anema et al (2008) used mathematical modelling in different conflict-affected sub-Saharan Africa settings to show that widespread rape only raises the overall level of HIV prevalence in a country by 0.023%.18
This evidence on conflict, SV and HIV has fuelled a heated debate within the humanitarian, security and public health arenas about whether SV is an important risk factor for HIV. Doubts about potential impact have led some groups to interpret the lack of association at a population level to indicate that rape has no association with HIV and to question the value of dealing with SV as part of HIV programming.
However, the studies that fail to find an association between SV and HIV have compared aggregated population-level data of HIV prevalence and experiences of violence. Such ecological analyses use several separate data sources and are inherently exploratory. More broadly, the use of aggregate, population data may mask the effect of competing forces particular to conflict settings, which may lead to reductions or increases in HIV infection. For example, although conflict often fosters SV, which may lead to increases in HIV risk, conflict-related factors, such as restrictions on population movements, population curfews and the separation of families, may also reduce opportunities for sex and widespread HIV transmission.19 20
Clearer insights into the ways SV affects HIV risk in different African settings where HIV is prevalent could be obtained by comparing the prevalence of HIV infection among women or men who have been assaulted with those who have not. To advance the debate, this paper presents current evidence on the extent of SV against women and men in different conflict-affected settings in sub-Saharan Africa, and then uses mathematical HIV modelling to explore the extent to which SV may increase individual HIV risk and investigate situations where SV may lead to substantial increases in community- or population-level HIV incidence.
Methods
We searched published and grey literature to compile existing evidence on the extent of SV against women and men in six sub-Saharan African countries that had recently experienced a period of armed conflict.
An established mathematical model of the probability of HIV acquisition21–23 was adapted to incorporate the potential for an increased risk of HIV transmission if genital injury occurs during sex and used to derive an analytical equation to describe the increased relative risk (RR) of HIV acquisition in a conflict situation versus a ‘comparable’ non-conflict scenario (online supplementary appendix). A number of plausible scenarios were developed to reflect situations that might arise in high- and low-conflict situations. Scenarios included gang rape, anal rape and transactional sex. Owing to lack of evidence about condom use in conflict situations, for the purposes of the modelling, we generally assumed that condom use was negligible.
Similarly, an analytical equation was developed (supplementary appendix) to consider the extent to which rape may increase HIV incidence at a community level. Different assumptions were made about the underlying HIV incidence, HIV and STI prevalence, the prevalence of rape and the effect of genital injury on transmission risk. In contrast to previous published analyses,18 we assessed the extent to which SV may contribute to HIV incidence rather than population prevalence. Given the different underlying HIV incidence that may occur between settings, in the analysis we consider the annual population HIV incidence in the range 0.5–2.5%.24
Mathematical modelling of the effect of genital injury requires an input to describe how the use of force may affect the risk of HIV acquisition for each sex act. There is evidence that during forced sex, genital injury and vaginal bleeding are much more common than in consensual sex, with the literature distinguishing between assaults with single versus multiple sites of trauma.25 Genital injury disrupts the multilayered stratified epithelium that lines a woman's reproductive tract and acts as a natural barrier to infection. Bleeding during sexual assault may also be associated with increased risk. For example, in a study of seroconversion rates in HIV discordant couples, those reporting non-menstrual bleeding during sexual intercourse were 4.9 times more likely to have seroconverted.26 In the absence of data on how this may affect the probability of HIV transmission for each sex act, in the modelling analysis we assume that genital trauma (and/or anal trauma for anal rape) increases the per sex act risk of HIV acquisition 1.5-, 3- or 6-fold, depending on the extent of trauma. These values were extrapolated from the epidemiological literature on the extent to which concurrent STI infection increases the risk of HIV acquisition for each sex act, with the cofactor of 3 comparable to the values commonly cited for a range of different STIs.27 28 These values were considered conservative, especially in situations where severe force has been used. In the anal rape scenario, penile-anal sex compared with penile-vaginal sex was assumed to entail a fourfold increase in risk of HIV transmission.29–31
An important input was the relative prevalence of HIV and STI infection among men who perpetrate SV compared with other men in that setting. The parameterisation of this input is difficult, however, as it depends on how long the perpetrator has been sexually active; their previous patterns of sexual behaviour; likelihood of exposure to HIV; and the extent to which they have previously used condoms or received STI treatment (which will reduce the likelihood that they are HIV infected). For the purposes of this model, we drew on evidence from non-conflict settings about the increased risk of HIV among the perpetrators of SV and assume that HIV and STI prevalence among men who perpetrate rape is twice as high as among less violent men.7 12 32
Other epidemiological parameter values used in the modelling of community HIV incidence were taken from the academic literature, and were set to the values commonly used in the modelling of HIV in sub-Saharan Africa. These include the probability of HIV transmission for each sex act of 0.002 for penile-vaginal sex;29 5% of those HIV-infected having high HIV viral load33–36; and high viral load increases probability of HIV transmission 10-fold.37 The parameters associated with high HIV viral load were assumed to be the same among perpetrators of SV as among less violent men.
Sensitivity analyses, where different assumptions were essentially turned off or varied, were used to explore the robustness of the findings to these different assumptions.
Results
Published evidence on the prevalence of SV in conflict is presented in table 1. Among settings with recently ended conflicts, lifetime levels of ‘sexual violence’ and ‘rape or sexual abuse’ experienced by women ranged from 42.3% in Liberia to 14%42 in Cote d'Ivoire.39 SV reported during conflict ranged from 35.3% in Liberia42 to 0.2% in Cote d'Ivoire.39 While in a setting with a prolonged conflict, service records from Democratic Republic of Congo report over 20 000 SV cases between 2003 and 2008.40 Owing to methodological and contextual differences it is difficult to compare these figures across settings as each study used different study populations; definitions of sexual abuse and time periods. Research on SV against men, though sparse, generally reported lower levels than SV against women.39 42 45
Sexual violence in selected sub-Saharan African countries with armed conflicts
Projections of the impact of sexual violence on individual HIV risk
Modelling projections can be used to get a sense of the extent to which an individual's risk of HIV may be influenced by SV, coercion and transactional sex. In practice, however, the absolute levels of risk are likely to be highly context specific. For this reason we focus here on estimating the risk ratios, comparing conflict-related scenarios with ‘comparable’ non-conflict scenarios. We consider risk ratios because these are less context specific. For the scenarios presented they are not dependent on underlying population prevalence of HIV, since this term cancels in the calculation of risk ratios (supplementary appendix). Hence, these findings apply for any given underlying prevalence of HIV.
The scenarios presented suggest that there may be large increases in individual HIV risk associated with SV in conflict-affected settings. The findings also illustrate how the risk ratios of HIV infection are largely determined by a combination of relative measures, which include the extent to which a perpetrator of SV may be infected with HIV or other STIs compared with other sexual partners; the number of forced sex acts compared with consensual sex acts; whether the sex acts are anal or vaginal and the number of assailants compared with other partners. The probability of HIV transmission during a violent sex act, in turn, also depends upon the extent of genital trauma, and the degree to which this increases the probability of HIV transmission.
To illustrate how the results are affected by the different factors, the right hand columns of table 2 show how the risk ratios vary according to the different assumptions. All scenarios show that individual risk of HIV more than doubles, even if there is no genital trauma or if genital injury does not increase the risk of HIV acquisition for each sex act (assuming HIV and STI prevalence are twice as high among violent men as among comparison men). Alternatively, individual risk of HIV also more than doubles if we instead assume that there is an increase in the risk of HIV acquisition for each sex act from genital trauma that occurs during forced sex and HIV prevalence is twice as high among violent men as among comparison men (and this holds even if STI prevalence is no higher among violent men).
Comparative model estimates of the relative increase in individual HIV risk for different illustrative conflict scenarios
Projections of the impact of sexual violence on community HIV incidence
A question raised in recent debates in the conflict field is whether SV could lead to increases in HIV at a population-level.17 Table 3 shows projections of how HIV incidence may be affected by SV.
Modelled projections of the potential increase in community HIV incidence, for different assumptions about underlying HIV incidence, and the prevalence of sexual violence, HIV, STIs and the extent of genital injury
In general, the proportion of HIV incidence that is attributable to SV increases proportionally with the prevalence of forced sex, the likelihood that the perpetrator is HIV infected and the cofactor effect of genital trauma. The contribution of SV to community HIV incidence is lower in high HIV incidence settings, because there is a greater underlying risk of HIV acquisition.
There are a range of situations where SV might potentially increase community HIV incidence by more than 10% (emboldened in table 3). For example, where HIV incidence is ≤0.5% in the absence of SV, community incidence would be increased by more than a tenth if more than 20% of women had experienced SV, genital injury increased HIV transmission sixfold or more and at least 10% of perpetrators were HIV infected.
If the STI prevalence is higher, an even larger increase in community HIV incidence may occur (data not shown). However, if genital injury increases HIV transmission just 1.5-fold, a greater than 10% increase in the 0.5% community HIV incidence would only occur if more than 40% of women were raped and at least 20% of perpetrators were HIV infected (not shown in table 3).
Discussion
SV is an extreme violation of human rights that results in profound and often enduring health problems. Although significant attention has been given recently to rape in war, there is still confusion about whether it may be an important risk factor for HIV in some conflict settings. SV is seldom included in epidemiological projections of HIV transmission.
Our review suggests that high levels of SV have occurred in several African conflict settings. Although the modelling presented here is limited by the data available and the challenge of developing comparable conflict versus non-conflict scenarios, where the SV and HIV epidemics overlap it is plausible that SV leads to substantial increases in individual HIV risk. This increased risk is likely to be due to both the biological effect of genital injury, and the potential clustering of behavioural risk factors among men who perpetrate rape, which make them more likely to be HIV- and STI-infected than non-perpetrators.
Our modelling also shows that SV may contribute to HIV incidence, particularly when the underlying incidence of HIV infection is low and/or the extent of sexual violence and genital trauma is high. These findings suggest that SV may lead to increases in HIV incidence at a community level—for example, in villages where systematic or mass rape has occurred. The main limitation of the modelling is that it is a static model, meaning that it cannot consider the transmission dynamics of HIV in terms of secondary and tertiary HIV infections, some of which may be indirectly attributable to SV. However, over a short timeframe in which the underlying sexual behaviour and epidemic remain fairly constant, other analyses suggest that findings from dynamic modelling are often still driven by the factors contained in the static risk equation.46
Epidemiologically, this analysis illustrates the dangers of confusing population-level and individual-level effects, and focusing on broad aggregate analysis to derive lessons about the drivers of HIV infection. If rebel forces have systematically raped certain populations or villages, for example, these individuals and groups may be at high HIV risk, yet this effect may not be detected at a population level if they are not considered separately. By aggregating data from dynamic and heterogeneous contexts such as conflict-affected settings, where there are competing influences on transmission, there is a danger that important risk factors will be overlooked.
The international community, through its various agencies, such as the Security Council, Inter-Agency Standing Committee and UN Action (a recent initiative that aims to prevent wartime rape becoming a peacetime norm) have provided a clear mandate for a strong response to sexual violence.47 For too long, we have had a narrow perspective of the epidemic and its drivers. We need a more nuanced understanding of the epidemiology of the virus and the contexts that create HIV vulnerability. Conflict-related initiatives offer important opportunities to assist survivors and prevent future abuses through collaborative programming that links reconstruction, HIV, sexual violence and women's empowerment.
Key messages
Sexual violence (SV) is common during conflict. Despite reports of rape-related HIV infection, population ecological analyses have found no association between violence and HIV.
For a range of SV scenarios, modelling suggests that a rape survivor's individual risk may increase by a factor of 2.4 to 27.1.
The contribution of SV to HIV incidence is determined by rape prevalence, genital injury, perpetrator HIV prevalence and underlying incidence. Situations exist where the incidence increases by 10%.
Conflict-related initiatives offer important opportunities to assist survivors and prevent future abuses through collaborative programming on reconstruction, HIV and sexual violence.
Acknowledgments
Support for this research was provided by the AIDS, Security and Conflict Initiative, convened by the Netherlands Institute of International Relations ‘Clingendael’ and the Social Science Research Council. Partial funding for this analysis also came from the Sigrid Rausing Trust. AMF and CHW are also members of the DFID-funded Research Programme Consortium for Research and Capacity Building in Sexual and Reproductive Health and HIV in Developing Countries of the LSHTM.
References
Supplementary materials
Web Only Data sti.2010.044610
Files in this Data Supplement:
Footnotes
Funding AIDS, Security and Conflict Initiative, convened by the Netherlands Institute of International Relations ‘Clingendael’ and the Social Science Research Council. Partial funding for this analysis also came from the Sigrid Rausing Trust.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.