Is it worth publishing data and recommendations that could be misconstrued and may not make much of a programmatic difference in the field? Yes.

Data, if collected and analysed correctly and interpreted carefully, help to improve our understanding of complicated and nuanced situations. Even if programmes in the field do not significantly change, our understanding of what the outcomes of such interventions can achieve will be more realistic. It also helps decision-makers prioritise their funding and interventions.

Sub-Saharan Africa (SSA) is overwhelmingly affected by HIV/AIDS 1. It is also the region with the highest number of armed conflicts worldwide. In 2005, armed conflicts were active in 15 of SSA's 47 countries 2, and resulted in the displacement of approximately 12.5 million people 3. Armed conflicts have traditionally been considered catalysts for HIV transmission due to associated sexual violence against women, lack of access to preventive and curative heath services, and increased vulnerability and risks incurred by population displacement and food insecurity 45. The use of sexual violence as a weapon of war has been particularly associated with concerns about heightened HIV incidence among women and girls 67. Widespread rape by armed combatants has been documented in Burundi 8, Sierra Leone 9, Rwanda 10, the Democratic Republic of Congo (DRC) 11, Liberia 12, Sudan 13 and Uganda 14. For the last decade, international organisations have perceived rape as a major cause of HIV transmission in conflict settings, and prioritised public health interventions to mitigate its effects 15161718.

However, empirical research suggests that HIV incidence in armed conflicts may in fact diminish, remain constant or not increase to the same magnitude compared with those countries not in conflict. This may be due to reductions in population mobility, access, and urbanisation 1920. A recent systematic review of the seven conflict affected countries found that existing data do not support a link between widespread rape and increased HIV prevalence at the population level 21. This finding has been met with some criticism from non-governmental organisations and advocacy groups 22, partially because of the perceived HIV-risk associated with rape in these contexts.

This paper builds on Spiegel et al.'s systematic review of HIV in conflict affected countries. To further examine the assertion that widespread rape may not increase HIV prevalence at the population level, we analyse the impact of varying scenarios of widespread rape on HIV prevalence in 7 SSA countries that have undergone conflict.

We built a model to determine the potential impact of varying scenarios of widespread rape on HIV prevalence at the population level in the seven African countries listed above (Table 1). Specifically, we investigated how the total population HIV prevalence would be affected if 1%, 5%, 10% and 15% of the female population aged 5–49 years were raped in each country. We looked at extreme scenarios, by assuming low, intermediate and high estimates of HIV prevalence and transmission rates among male assailants. We assumed that assailants of widespread rape were armed combatants. Given the absence of seroprevalence data for armed combatants in the countries under study, we assumed the HIV prevalence among assailants was commensurate to that of national militaries in SSA. To take into account potential higher prevalence in selected groups (e.g. HIV used as a weapon), we multiplied the baseline population prevalence by 2, 4 and 8. Primary outcomes of our analysis included absolute and relative increases in population level HIV prevalence, per different rape prevalence estimates, for each country, and for age groups 5–49 years. To be conservative in our estimates, we decided to use the 5–49 year age group compared with the 15–49 year group because rape in conflict has been reported among girls less than 15 years of age.

We obtained annual population estimates for the countries under study from United States Census Bureau's International Data Base 23. These population estimates are based on all residents of nations within SSA regardless of legal status or citizenship, and represent a compendium of data on population, fertility, mortality, contraceptive use, and related demographic topics provided by the US Census Bureau 24. High, low and mid-range estimates of national HIV prevalence estimates for women aged 5–49 years were obtained from UNAIDS/WHO's country-specific end-2006 Epidemiological Fact Sheets for HIV/AIDS and Sexually Transmitted Infections 25.

We determined the number of women at risk of becoming newly infected with HIV from rape by subtracting the estimated number of HIV prevalent cases from the total adult female population. The lower population prevalence was used as a conservative estimate. To account for varying heterosexual transmission rates according to disease stage, multiple rapes, presence of sexually transmitted infections, and other confounding factors, we assumed low, intermediate and high rates of HIV transmission. On the low end, we assumed an HIV transmission rate of 0.0028, which is the acknowledged rate for HIV heterosexual HIV transmission among asymptomatic and symptomatic individuals (i.e. >500 to >200 CD4+ counts). On the upper end, we assumed a rate of 0.008, the rate of heterosexual transmission among individuals with acute and recent primary HIV infection 26. This upper rate was then multiplied two and four times to take into account potential factors associated with higher rates of transmission, such as genital and/or rectal trauma associated with rape by multiple assailants 27. The number of women at risk of HIV transmission is the product of the total population of women in this age-group by the probability of rape, taking into account those who are already HIV positive. The number of newly infected women is, therefore, the product of the number of women at risk by the probability of the assailant being positive and the probability of transmission. The absolute increase in prevalence is the number of newly infected women divided by the total population. The relative increase in the total population is therefore this rate divided by the country specific rate (Table 1).

Our findings show that even in the most extreme situation, where 15% of the female adult population level was raped, where HIV prevalence among assailants was 8 times the country population prevalence, and where the HIV transmission rate was highest at 4 times the average high rate, widespread rape increased the absolute HIV prevalence of these countries by only 0.023% (range: 0.007–0.046%). These projections support the finding that widespread rape in conflict affected countries in SSA has not incurred a direct major population-level change in HIV prevalence 21.

Our lowest (0%) or intermediary (0.004%) estimates may, in fact, be most representative of the actual situation. Little information exists about HIV prevalence among armed combatants in conflict affected SSA countries. UNAIDS has previously assumed that rates of HIV among armed combatants is 2 to 5 times higher than among civilian populations 28. However, a recent study now suggests that only long-serving soldiers and soldiers belonging to less disciplined forces may be at increased risk of infection 29. HIV prevalence in militaries may be lower than previously assumed due to the common military practice of recruiting young men from low-prevalence rural areas, to screening out HIV-positive individuals through routine testing, and to stationing soldiers in remote areas with limited mobility and remuneration 30.

Some persons will justifiably be concerned that publishing such an article will do harm to all of the important efforts that have occurred to ensure that sexual exploitation and violence are recognised as the most basic of human rights violations and essential interventions must be provided to all survivors of such heinous acts in all contexts. Since widespread rape in conflict situations does not appear to directly increase HIV prevalence at the population level, should donors and other decision makers decide to put their limited funds, personnel and interventions towards other groups and programmes in different contexts that may have a larger public health affect? They may and are free to do so. However, we would strongly recommend against it due to the reasons stated above. Despite the uncomfortable findings of this article and the possibility that people may misinterpret or correctly interpret the findings of this article and decide to prioritise programmes other than sexual violence, we still believe it is important to publish such a paper. We dread the possibility that some journalist may try to grab the headlines by writing "Rape does not increase HIV". However, that concern does not justify not having an open, honest, intelligent and nuanced discussion about rape and its affect on HIV transmission at the individual and population levels.

The authors declare that they have no competing interests.

AA undertook the modeling and drafted the manuscript. MRJ underook the modeling and commented on the manuscript. EM undertook the background research, helped to draft the manuscript, and participated in the revisions of the manuscript. PS conceptualised the design of manuscript, helped to draft the manuscript, and participated in the revisions of the manuscript. All authors read and approved the final manuscript.