Increased understanding about the relative contribution of genito-anal injury to HIV transmission may improve epidemic model estimates of the distribution of risk among and across different subpopulations. Better understanding about the distribution of HIV among subpopulations that are at highest risk of sexual violence can also improve the design and prioritization of combination prevention interventions that are most likely to reduce the risk of sexual violence and its potential contribution to HIV transmission. The effective incorporation of physiological and social variables into epidemic modelling will likely require new research approaches that can help communicate the level of risk associated with different types of ‘heterosexual’ transmission. Reference models that reflect the potential impact of sexual violence and genital injury can help direct attention toward key variables and uncertainties. For further research that clarifies these relationships will require multidisciplinary collaboration among groups with expertize in epidemiology, social science, public health, and clinical and basic science.
This paper expands on the central concepts introduced in the ‘Greentree White Paper on Sexual Violence, Genito-anal Injury and HIV Transmission’ and identifies related opportunities for improving model estimates of transmission risk, particularly among young adolescent women. The Greentree White Paper identifies important physiological cofactors of risk that have likely been underestimated in HIV epidemic modelling, basic science research and clinical investigations. Drawing from a wide range of scientific studies representing basic science, social science, public health, epidemiology, mathematical modelling and clinical research, the Greentree White Paper advances two hypotheses: first that genital injury plays a key role in HIV transmission, acquisition and pathogenesis; and second that age-related physiological and immunological factors are key mediators of risk.
This paper proposes several exploratory directions for research to increase understanding about the relative contribution of sexual violence and genito-anal injury to important predictive variables used in epidemic modelling. The long-term goal of this effort is, firstly, to improve understanding of causes of the distribution of HIV among various subpopulations and, secondly, to inform the design and prioritization of combination prevention interventions that are most likely to reduce the risk of sexual violence and its potential contribution to HIV transmission.
The role of sexual violence in HIV transmission
Sexual violence is a poorly studied yet potentially important risk factor in HIV transmission that may be significant in the overall expansion of the AIDS epidemic and its disproportionate geographic and gender distribution. This is especially so in sub-Saharan Africa, where 76% of all HIV-positive women are located and where prevalence among young women aged 15–24 is two-to-three times higher than among young men in the same age group. Similarly, disproportionate rates of infection are also reported in a number of fragile and conflict-affected countries (Table 1). According to the 2010 Report of the Secretary-General on Implementation of the Declaration of Commitment on HIV/AIDS and the Political Declaration on HIV/AIDS, this imbalance reflects, ‘not only the heightened physiological vulnerability of girls and young women, but also a high prevalence of intergenerational sexual partnerships, the lack of woman-initiated prevention methods and broader social and legal inequality that impedes the ability of young women to reduce their sexual risk’. According to UNAIDS, the disparities are even more extreme in many conflict-affected regions where ‘gender inequalities and all forms of violence against women and girls increase their vulnerability to HIV’.
Table 1. HIV Prevalence among Young Men and Women, Ages 15–24
Prevalence young male (15–24) (%)
Prevalence young female (15–24) (%)
Estimate [low-high estimate]
Estimate [low-high estimate]
Source: UNAIDS. Report on the Global AIDS Epidemic (2010).
Papua New Guinea
There is, however, little theoretical, legal and policy agreement about what constitutes sexual violence across socio-cultural settings. And despite its pervasiveness in many hyperendemic and conflict-affected areas, little is also known about the social and physiological co-factors of sexual violence and the conditions under which it can increase the probability of HIV transmission and acquisition among women and girls.1 In some conflict and post-conflict situations, for example, conflict-related rape can increase transmission risk at both individual and population levels.[5, 6] Research supported by the AIDS, Security and Conflict Initiative (ASCI) called attention to the potential linkages between HIV risk and the psycho-social effects of conflict, displacement, torture and violence on interpersonal and family relations and socialization processes. The breakdown of protective family and social structures, increased poverty, conflict-related rape, the mixing of high- and low-prevalence populations through displacement and migration, and the deployment of military and civilian personnel can increase the vulnerability of women to sex trafficking, forced and early marriage, transactional and survival sex; all of which have been associated with an increased risk of HIV. During conflicts and post-conflict transitions, the availability of and access to HIV prevention and services may also be limited or uneven, depending on the availability of humanitarian assistance and the functioning of national health infrastructure. HIV risk will be heightened among those population strata that do not have access to condoms, emergency and reproductive health care, treatment for sexually transmitted infections (STIs) and tuberculosis (TB), safe blood and other pharmacologic prevention.
A number of physiological factors can also influence the contribution of sexual violence to HIV transmission, acquisition and pathogenesis. These are elaborated in the Greentree White Paper and summarized in Fig. 1. The factors that are most likely to increase transmission efficacy are related to: (i) the perpetrator's HIV status, circumcision and other co-factors such as STIs and TB; (ii) the location, severity and frequency of genito-anal injuries; and (iii) access to pharmacologic prevention, both before and after sexual assault. Anatomic, biological and physiological factors that can increase susceptibility among women include (i) age-related mediators of risk that regulate immune protection, such as the role of sex hormones, the reproductive cycle and the maturation of the cervical and vaginal epithelium; (ii) the use of exogenous hormones; and (iii) opportunistic and co-infections.
Reassessing the explanatory value of ‘heterosexual transmission’ as a risk behaviour and a mode of transmission
According to UNAIDS and the World Bank, the majority of new infections in sub-Saharan Africa are expected to occur in the general population through heterosexual transmission: more than 90% in Lesotho, Uganda and Zambia; more than 80% in Swaziland, an estimated 60% in Kenya and 70% in Burkina Faso. Although model estimates of HIV incidence in generalized epidemics are improving rapidly, little progress has been made in their ability to estimate the distribution of risk patterns associated with heterosexual transmission or to explain the disproportionate rates of infection among young women as compared to young men, particularly in sub-Saharan Africa and in areas with endemic sexual violence.[9-13]
The term ‘heterosexual’ transmission was introduced in the late 1980s to distinguish the pandemic's trajectory in Africa from epidemic patterns seen among ‘homosexual’ populations and other key groups among whom incidence is high, such as sex workers and people who use intravenous drugs. The key variable used to distinguish risk among heterosexual populations is the number of overlapping sexual partners.[14, 15] In HIV estimation models and prevention interventions, the term ‘low-risk heterosexual transmission’ is often used to refer to individuals with one monogamous ‘heterosexual’ partner in the last year, while individuals with multiple heterosexual partners in the last year are considered to be ‘high-risk heterosexual’. The value that is most commonly used to estimate the per-act probability of heterosexual transmission (0.001) is, however, too low to explain the high prevalence of heterosexual transmission in epidemics where estimates are reported to be as high as 0.038 (female to male) and 0.30 (male to female).[16-19]
Many explanations have been offered to explain the dissonance between probability estimates and the findings of observational studies, including bias in self-reported behaviours,[18, 20, 21] underestimation of heterosexual anal transmission, infected instruments [23, 24] and the interaction between early high infectiousness and long-term concurrent sexual partners. But estimation methods are still limited by the quality of serosurveillance, and results are not very different across models that have begun to reflect the added impact of ulcerative STIs, co-infections, antiretroviral therapy, male circumcision and other facilitating factors, such as disease staging.
The limitations of the ‘Modes of Transmission Model’, used widely by decision-makers to inform research, policy, resource and prevention priorities, are well elaborated with respect to model structure and input parameters, but a number of limitations remain unexplored. The most central limitation among these is the yet unproblematized conceptualization of heterosexual transmission as a sexual orientation. This is in contrast to the concept of ‘homosexual transmission’, which, in the early nineties, was reconceptualized in the behavioural, social science and epidemiological literature as a range of sexual behaviours among ‘men who have sex with men’ (MSM) rather than as a sexual orientation (i.e. homosexual, bisexual, heterosexual or gay). In the case of heterosexual transmission, however, epidemic modelling continues to use a single value to estimate per-act heterosexual transmission probability despite the highly differentiated variation of risk associated with different male–female sexual practices. These risks are likely to vary according to (i) the type of sexual intercourse (anal or vaginal); (ii) the use of force (e.g. gang rape, incest, domestic violence); (iii) sex- and age-related anatomic and physiological characteristics and (iv) relationship typologies (e.g. polygany, levirate marriage, transactional sex).
A second limitation is that epidemiological models of risk transmission have not incorporated findings from behavioural and social science research that demonstrates various ways in which sexual violence perpetrators and survivors are a ‘risk group’ and that sexual violence is itself a ‘risk behaviour’. For example, research shows that men who have experienced sexual violence are more likely to rape and that men who perpetrate sexual violence are more likely to be HIV infected.[27-34] Research has also associated sexual violence with increased female risk-taking behaviour.[35-37] Female survivors of sexual violence and physical intimate partner violence often exhibit increased risk-taking behaviours in relation to partner selection, drug use and alcohol use, which in turn can influence sex practices (e.g. unprotected vaginal sex, anal sex, etc.).[32, 38-42] In other words, a woman (or man) in a stable, monogamous partnership who experienced rape or sexual assault at any point in their lives or on multiple occasions (by the same or different perpetrators) will still be assigned the same ‘low-risk’ probability as someone who has not, unless the perpetrator is reported as a ‘sexual partner’ in the past year.2
Third, estimation models have not taken into account a cluster of potentially significant sex- and age-related anatomic, biological and physiological factors that increase the susceptibility of young adolescent women. Summarized in Fig. 1, these factors are introduced in the Greentree White Paper and elaborated in Ghosh et al. Details regarding the potential dependencies between sex and age may be crucial to effective modelling. Fig. 3 offers a conceptual illustration of these potential dependencies, showing variation among key physiological risk factors with age. Among young adolescents, for example, physiological susceptibility may be amplified by the increased likelihood of cervical ectopy, STIs and the immaturity of the cervical and vaginal epithelium. Sex hormones play an important role in regulating immune protection against HIV acquisition, and increased susceptibility is distinct during adolescence, various stages of the menstrual cycle, pregnancy and menopause. Age is also a factor associated with the use of exogenous hormones, during peak reproductive years and among post-menopausal women. Additional research is needed to determine how the coexistence of multiple co-factors can increase susceptibility and the added contribution of genito-anal injury to HIV transmission, acquisition and pathogenesis when multiple co-factors are present.
Fourth, estimation models have not yet taken into account the role of genito-anal injury and intimate partner violence in HIV transmission, acquisition and pathogenesis. Depression and post-traumatic stress disorder – co-morbidities of intimate partner violence – can also contribute to increase physiological susceptibility and disease progression.[32, 37, 41, 44] And, as elaborated in the Greentree White Paper, the inflammation, abrasions and injury that result from consensual and forced sex can increase susceptibility to HIV infection by recruiting target cells to the site of injury and by disrupting the integrity of the mucosal barrier. The severity and type of injuries will be influenced by a number of factors including the sex- and age-related anatomic and physiological factors identified in Fig. 2.
Three important mediators of the impact of sexual violence and genito-anal injury on HIV transmission, acquisition and pathogenesis are identified in Fig. 3. They include (i) the location and type of injuries (anal or vaginal); (ii) the severity of injuries, which may be influenced by the degree of force, the number of perpetrators, sex- and age-related anatomic and physiological characteristics of the victim, and the age difference between the victim and the perpetrator; and (iii) the frequency of injuries (i.e. the cumulative incidence over time and the number of different perpetrators).
New directions for research
The Greentree White Paper identifies priorities for scientific research, policy and practice to improve understanding about the contribution of sexual violence and genito-anal injury to HIV transmission, acquisition and pathogenesis. It affirms the important contribution that epidemiologists and modellers can make in bridging these emergent lines of biomedical research with an already robust body of social and behavioural science on the gender dimensions of HIV risk and sexual violence.
New models are needed to probe these relationships and, as the science evolves, to quantify the relative contribution of sex- and age-related physiological factors to individual susceptibility. This would provide modellers and others with the current best estimates of parameter values and uncertainty ranges that can be associated with different forms of sexual violence and genito-anal injury. While these models and parameters would not themselves become a new standard, they could be used in preliminary modelling analyses to assess how various forms of sexual violence and genito-anal injury might influence HIV transmission across different age groups and over variable periods of time. This could possibly help explain heterogeneity in HIV prevalence across different subpopulations and identify the key parameters (concerning sexual violence and genito-anal injury) that most influence model estimates of new infections.
In particular, such plausible ranges of parameter values would allow researchers to assess (i) the likely relative contribution of different genito-anal injuries to the force of infection (i.e. the rate at which susceptible individuals become infected per unit time) and (ii) the relative contribution of age-related anatomic and physiological factors that place young girls and adolescents at increased risk. Identifying the relative importance of different parameters can help communicate the level of risk associated with different types of ‘heterosexual’ transmission – an overly aggregated category with little explanatory power. The results will also help identify the most critical areas of data needed to specify the social and physiological risk factors that are likely to have a significant impact on transmission and begin to signal directions for prioritizing prevention interventions.
In a significant departure from concurrency models that rely on quantitative indicators and use the individual as a unit of analysis, new approaches are encouraged that draw upon expert knowledge and empirical social and behavioural science to construct sociologically plausible reference models and analytic equations that can be used to characterize different typologies of sexual violence and estimate their potential contribution to HIV transmission, acquisition and pathogenesis. Examples of typologies include intimate partner violence, conflict-related sexual violence, forced and early marriage, child sexual assault and transactional/survival sex. This approach is similar to Kretzschmar's and Caraël's proposed ‘partnership typologies’, which assess the qualitatively different risks associated with the nature of partnership (e.g. polygamous families, transactional sex or rape), without assuming equal rates of dissolution, sexual activity and concurrency. Approaches like these can build on existing social science and behavioural studies that have demonstrated an association between sexual violence and HIV risk. They can also further develop the exploratory modelling that has already found significant correlations between conflict-related sexual violence and HIV incidence in some settings.
More accurate assessments will require understanding of the ‘risk environment’ and better understanding about how contextual factors – including sociocultural norms and institutions – influence individual behaviour as relationships change within a partnership, the family and broader community. Creating new reference models that incorporate both physiological parameters and social cofactors may be helpful in establishing correlates between different typologies of sexual violence and the probability of HIV transmission, and in specifying the characteristics of subpopulations that are more physiologically susceptible to HIV and socially vulnerable to sexual violence. Models can also help establish correlates between the age of victimization, the typology and duration of sexual violence, and transmission probability, as well as identify predisposing characteristics of sexual violence perpetrators and the reinforcing or enabling factors that increase their likelihood of acquiring and transmitting HIV.
Summary and conclusions
Despite significant improvements made in estimation and projection methodologies, infectivity estimates have not considered the potential contribution of sexual violence and genito-anal injury to per-act transmission probabilities in the context of heterosexual transmission. Important predictive variables that are used in epidemic modelling – such as ‘risk behaviour’, ‘modes of exposure’ and ‘high- and low-risk’ heterosexual transmission – associate overly aggregated and often culturally inaccurate categories of sexual behaviour with value estimates that often contradict the findings of observational studies. To the extent that different typologies of sexual violence can be constructed and associated with corresponding levels of transmission risk, sexual violence may constitute a category of exposure with more explanatory power than ‘heterosexual transmission’.
Increased understanding about the relative contribution of genito-anal injury to HIV transmission may improve epidemic model estimates of the distribution of risk among and across different subpopulations. Better understanding about the distribution of HIV among subpopulations that are at the highest risk of sexual violence can also improve the design and prioritization of combination prevention interventions that are most likely to reduce the risk of sexual violence and its potential contribution to HIV transmission. The effective incorporation of physiological and social variables into epidemic modelling will likely require new research approaches that can help communicate the level of risk associated with different types of ‘heterosexual’ transmission. Reference models that reflect the potential impact of sexual violence and genital injury can help direct attention toward key variables and uncertainties for further research that clarify these relationships. Pursuing this multidisciplinary agenda will require collaboration among groups with expertize in epidemiology, gender, social science, public health, and clinical and basic science.
We gratefully acknowledge the helpful comments received on an earlier version of this paper from Zahra Asgari, Miranda Berry, Marie-Claude Boily, Michel Caraël, Joseph O. Deasy, Catherine Hankins, Susan Newcomer, April Pei, Carel Pretorius, Fulvia Veronese and Charlotte Watts.
Sexual violence affects everyone: men and boys, women and girls. This paper, however, focuses on the disproportionate burden of HIV among young women in hyper-endemic settings who acquire it through heterosexual intercourse and for whom sexual violence, in all of its forms, is likely more prevalent.
Perpetrators of rape and other forms of sexual violence and exploitation are not likely to be identified as a ‘partner’ and HIV surveillance rarely includes questions about the experience of sexual violence over time.