Many obstacles have contributed to the lack of ART programs in conflict or post-conflict affected areas. They include issues related to health need prioritisation, feasibility, effectiveness, safety and ethics. Our study provides important information that can be useful in addressing many of these concerns.
1. "HIV treatment is complex. It will not be possible to assure safe and effective treatment in conflict affected settings."
Experience from vertical HIV programmes in resource-limited settings has led to simplification of treatment with standardised treatment protocols, fixed-dose combination drug regimens, minimal monitoring, and intensive adherence support. This approach has proved safe and effective  and can be applied without major changes in conflict settings. In our programmes, fixed-dose combinations facilitated adherence, procurement, and stock management, and reduced costs. Relatively complex monitoring tools such as CD4 counts and liver function tests allowed a paradoxical simplification of management and increased the ease of decision making by less experienced clinical staff . HIV activities were introduced in a stepwise manner that avoided overwhelming teams and allowed staff time to gain experience in HIV care and systems to be put in place.
The evidence presented here and previously from our programmes and from others suggests that ART outcomes are equivalent to those in stable resource-limited settings [11–14]. In this study, the median 12-month survival of 0.89 (0.88-0.91) compares favourably with that in Malawi (0.81;0.79-0.83), Zambia (0.82)  and South Africa (0.93;0.92-0.94) . In addition, it is comparable with the ART-LINC study, to date the largest combined analysis of cohorts in stable resource-limited settings; mortality in our programmes (9% versus ART-LINC 6%) and lost-to follow-up rates (11% versus ART-LINC 15%) after 12 months of treatment, as well as median immunological gains after 6 months of treatment (129 cells/mm 3 versus ART-LINC 106 cells/mm 3) were similar. This analysis involves a very large dataset involving many programmes in conflict affected environments and thus provides important data supporting the effectiveness of providing ART in these environments.
2. "Adherence to treatment is likely to be poor due to forced displacement and population mobility."
There is no evidence that, providing the drug supply is well-managed, people will be any less adherent to ART in conflict settings. Our experience is that people adhere well , but treatment should be provided free of charge given the negative impact of user fees on access to services and adherence to ART . Acute programme disruptions were uncommon and rarer than expected; the only programme to face such disruption was Bukavu, DRC, for 2 weeks in 2005. Disruptions can occur even in 'stable' settings, either due to unexpected conflict such as in Kenya in 2008 or through drug ruptures secondary to mismanagement or financial limitations . Thus many of the practical measures used in these settings could be applied in all HIV programmes and also to particular populations at higher risk of interruption such as migrants and nomadic populations.
With the proliferation of access to treatment in stable settings, leakage of drugs into the informal sector is inevitable even in conflict settings, and the absence of treatment programmes forces people to look to these costly and sub-standard sources. In such contexts provision of treatment is also a form of harm reduction.
Populations in conflict and post-conflict settings can be mobile. In Liberia, our patients would often cross the borders into neighbouring Sierra Leone and Guinea seeking health care or better conditions. Programmes need to allow for population movement (Appendix 1). If population movements are planned in advance, management strategies can include the provision of 3-6 months of ARVs for patients stable on treatment, or if medically stable, delaying initiation of ART until the patient arrives in the new destination . Simple, cheap, and readily available regimens may be preferred to more complex and expensive ones that may not be available elsewhere.
3. "Treatment should be life-long. People may be started on treatment, only to stop after 6 months or a year."
In resource-limited settings significant health benefits are usually obtained within 6 months of starting ART; mortality can be reduced by up to 78% , rates of opportunistic infections reduced by 56%, and robust immunological gains obtained. We saw good survival and immunological outcomes at 6 months. People also became informed about their illness and the benefits of treatment. This improved health and knowledge may enable them to better manage their illness if treatment stops, reduce the risk of them transmitting the virus by adapting behaviour, and help them to potentially seek treatment elsewhere as it becomes increasingly available in resource-limited settings .
Predictions of how long people will be able to take treatment and what will happen in the future are rarely possible. Where there is uncertainty people should be given the chance of receiving treatment, but informed patient consent should be obtained regarding risks, benefits, and potential for interruption or cessation of ART. However it is important to determine the minimum time that treatment should be available to obtain benefit. While absolute rules on this are difficult, we and others feel that around 3-6 months on treatment should be seen as a minimum .
4. "Stopping treatment will lead to resistance."
It is clear that while regular treatment interruptions do promote resistance, the risk of developing resistance due to a single stop of treatment is low. The risk is further reduced if those taking an NNRTI-based regimen receive a 1-week continuation of dual nucleoside reverse transcriptase inhibitor (NRTI) therapy (ie AZT/3TC or D4T/3TC; a 'washout' course) to cover the drug's longer half-life , or, for those on a PI-based regimen, all drugs are stopped together .
5. "Resources should be directed to other more important acute health priorities."
In high HIV prevalence areas many competing health priorities such as malaria, diarrhoea, and tuberculosis occur more frequently and have a higher mortality due to the presence of underlying HIV. Addressing underlying HIV substantially contributes to addressing these needs and reduces their demands on medical services. In our experience, integration of HIV activities strengthened other health activities such as the diagnosis and treatment of tuberculosis and maternal and reproductive health by sharing resources and improving laboratory services, procurement, supply, and monitoring mechanisms. This contributes to improving the health of all victims of conflict but also increases the likelihood of sustainability for implemented activities [see Additional file 1, Table S1] .
Nevertheless, there are certainly low HIV-prevalence settings or especially difficult and unstable environments where the potential benefits of introducing ART may not justify the resources required. In some conflict-affected areas such as Somalia, Darfur, and Sri Lanka, MSF-OCA has not introduced HIV treatment activities (apart from post-exposure prophylaxis) because of a lack of HIV-related medical needs identified by field teams, often compounded by potentially serious negative consequences for those testing positive. A decision to provide treatment should, like all health-care decisions, be made on an informed and unprejudiced assessment of the needs and priorities of the population and the feasibility of an effective response.
6. "HIV programmes should be sustainable"
Continuation or handover of HIV programmes started in a conflict-affected area, usually without prior HIV activities, can pose significant challenges. Our experience was that an established and effectively running programme was a catalyst to engage other actors, especially National Aids Programmes and MoHs, to provide ART in the region . Despite the difficult conditions in which the programmes were instituted, a handover partner was found for all programmes that MSF-OCA closed [see Additional file 1, Table S1]. All programmes were handed over to MoH, sometimes with national or international non-governmental organisations providing technical support, training, support to procurement and supply channels, and funding of key staff. For example, in Mindouli, Republic of Congo, the National AIDS Programme funded key staff involved in HIV activities and accredited the hospital as an ART site . Early integration with existing MoH systems and structures and planning and discussion with potential partners facilitated the process. Identification and onsite training of key staff to remain in the programme was vital.
Further challenges in conflict environments
Prevention of Mother to Child Transmission (PMTCT)
Despite the recognised importance of the intervention and a strong desire to implement it by programme managers, the initial inclusion of PMTCT was surprising difficult in many settings, usually due to the resistance of health-care staff. Their concerns included perceived potential negative consequences for women diagnosed HIV-positive, the complexity of the intervention in environments where programme disruption was possible, a lack of understanding of its potential benefits, and limited staff experience in managing HIV. Nevertheless, HIV transmission rates can be reduced by partial interventions even if full ones are prevented by programme disruption . Our experience was that with simplified protocols and tools, quality education and counselling of women and staff, and the provision of extra resources, PMTCT activities were possible. Infant feeding was complicated by the potential for programme interruption to leave mothers who formula fed without infant feeding options, and therefore exclusive breastfeeding with early rapid weaning at 6-9 months was usually promoted, apart from Angola where almost all women were offered and adopted formula feeding.
Paediatric HIV care
Similarly to programmes in stable settings, the inclusion of infants and children was limited. This was influenced by difficulties diagnosing HIV in children, low clinician confidence in clinical HIV paediatric care, and lack of drugs in adapted formulations and fixed-dose combinations . Strategies targeting children, especially orphans, and improved and adapted diagnostic tools and medications are needed in these settings.