Aids Research Paper Introduction Paul


It is clear that many sources of evidence have contributed to our grasp of what does and does not work in HIV/AIDS education. Despite this, there has recently been a distinct move to narrow the evidence of success in this field to experimental and comparative work, with randomized controlled trials positioned as the `gold standard'. Here we take up the question of what constitutes evidence in HIV/AIDS education. We explore the social and historical factors which `privilege' certain kinds of evidence above others and question whether there exists but one way of understanding what works best in HIV/AIDS education. We draw expressly upon earlier insights and experience in educational evaluation per se and put a case that evidence gleaned through a range of research methods is more useful than exclusive reliance on experimental and comparative work.


In 15 years of responding to the HIV/AIDS epidemic, much has been learned about what does and does not work in HIV/AIDS education. Our knowledge of the characteristics that demonstrate effective and non-effective programming in this area comes from diverse sources. Some of the techniques that have been employed to determine whether a particular form of education works are less rigorous, though not necessarily less appropriate to a given set of circumstances, than others (Aggleton, 1997a). At times when it has been necessary to act with urgency—not uncommon throughout the epidemic—HIV educators have had to draw on personal experience, intuition and educated guesses to get education programmes under way. Over time, HIV educators have accumulated knowledge of what works in HIV/AIDS education through observation, especially observations that have been systematically applied. In other circumstances, assessments against the specific objectives of HIV/AIDS education programmes and demonstration projects have provided informative data on the outcomes of interventions. When there have been acceptable sites, participants and ethical considerations, as well as the necessary time and financial resources, randomized controlled trials (RCTs) and other experimental studies have contributed to our understanding of success in HIV/AIDS education.

Many sources of evidence, each with its corresponding strengths and limitations, have informed HIV/AIDS education and related public health policy. No single source has been sufficient to determine the appropriateness and effectiveness of different strategies. Successful approaches to HIV/AIDS education have been determined by bringing together evidence collected in different ways (Auerbach et al., 1994; Coates et al., 1996; Aggleton, 1997a). This is clearly evident in comprehensive lists of the characteristics of effective programming in HIV/AIDS education such as the compendium produced, for example, by the Washington State Department of Health (Washington State Department of Health, 1994). This lists some attributes of effective programming that are axiomatic (e.g. target the highest risk populations rather than the easiest to reach or teach, or the politically `fashionable'). Other features are based on the experiences and observations of educators (use the familiar language or vernacular of the targeted population, take into account cultural and social context, and ensure that programmes resonate with people's personal experiences). Some characteristics are designed to facilitate outcomes evaluation (e.g. have clearly defined objectives; specify what changes in behaviours are to be achieved and by whom). At times, the recommended strategies are supported by experimental results [e.g. use `natural leaders' among peers—see the work involving key opinion leaders (Kelly et al., 1991, 1992)].

Despite the fact that many sources of evidence have contributed to success in HIV/AIDS education (as measured through such outcomes as behaviour change or declining HIV infections), there has recently been a distinct move to narrow the evidence of success in this field. One way this has occurred is through effectiveness reviews that have privileged certain research methods rather than assessing the methods in terms of their appropriateness to the particular programme or intervention. There has been in the past few years a growing tendency to embrace experimental and comparative work—particularly RCTs as some sort of `gold standard'—as the only appropriate way to evaluate HIV/AIDS interventions. Correspondingly, other sources of evidence have been devalued. This push has come from several quarters, mainly in the UK (Oakley et al., 1995; Oakley and Fullerton, 1996) and in the US (Aral and Peterman, 1996; O'Leary et al., 1997).1

It is appropriate and timely to consider the question of what constitutes evidence in HIV/AIDS education, to consider the social and historical factors which `privilege' certain kinds of evidence above others, and to question whether there exists one way of understanding what works best in HIV/AIDS education. By doing so, we hope to explore some important issues concerning the research methods that are of greatest relevance to HIV/AIDS education, and how questions of evidence and research methods relate to the policy and practice of HIV/AIDS education. In this paper we examine these issues and draw expressly upon earlier insights and experience in educational evaluation per se to argue that evidence gleaned through a number of research methods is more useful than an exclusive reliance on experimental work. Our paper is a challenge to a certain prevailing orthodoxy, not a comprehensive review of successful HIV/AIDS education.

Evidence and health promotion

Without doubt, evidence is a basic requirement for rational decisions in all walks of life, in business, politics, law and medicine (O'Rourke, 1997). Nonetheless, there are many different and legitimate ways of understanding some things to be true, and of accumulating evidence of what works. Some people bring to their various professions useful characteristics such as business acumen, political nous, wisdom and even the healing touch. These traits may be enhanced through experience, observation and practice, not only through published data and accounts of controlled studies. So too in the HIV/AIDS arena. Educators and health policy analysts should be encouraged to draw upon the evidence which comes from many sources, some of it intuitive or experiential, some gleaned from systematic observation and some based on experimentation. Evidence and information derived from various means should be valued, carefully examined and filtered to see how applicable they are to a given set of circumstances. Likewise, it is important to acknowledge researcher expertise as well as practitioner expertise. Effective evaluation needs researchers skilled at understanding research methods in order to evaluate `success' appropriately.

The debate surrounding the movement towards evidence-based medicine points to the pitfalls of accepting the results of clinical research trials as the only `true' kind of medical knowledge (O'Rourke, 1997; Sweet, 1998). Not only does such a view fail to recognize the situated, historically and socially constructed nature of human understandings, it also omits the possibility that such understandings can and do change over time. The published evidence is highly selected, incomplete and often dated. The patients in RCTs are often not representative of the general patient population and trials reported in the literature rarely contain more than 50% of patients initially screened. RCTs usually test only one therapeutic agent, even though many patients with common conditions require multiple drugs. Many journals are still reluctant to publish negative or inconclusive outcomes so there is a publication bias for positive results. Major controlled trials take time to set up, conduct, analyse and report, so that by the time benefit can be ascribed to a particular treatment it may have been superseded by another with apparently greater efficacy and fewer side effects (O'Rourke, 1997). In one recent study of surgical operations, for example, only 39% of treatment evaluation questions could have been answered by an RCT in an ideal clinical research setting (Solomon and McLeod, 1995). Likewise, in the field of nursing, White (White, 1997) has argued that not all nursing problems are capable of being reduced to a clear issue that can be solved by controlled experimental work and many require artistry to find a solution. White has also pointed to the possibility that researchers applying for funding for non-experimental research may be disadvantaged, and the temptation to use experimental findings to justify restricting choice of interventions for both practitioner and patient. Moreover, there is the problem of the gap between research-based knowledge and its implementation in health-related practice (Berrow et al., 1997).

Each of these problems associated with clinical research trials in medicine has its parallel in, and is equally applicable to, experimental studies in the field of HIV/AIDS education. Apart from difficulties in controlling in a multi-faceted, socially-based intervention, another major problem remains. If knowledge of what works in HIV/AIDS education were to be limited to experimental evidence, one would surely want to ensure that the evidence was not skewed in favour of particular interventions and populations. This is certainly not a tenable proposition. Some interventions lend themselves to experimentation (e.g. teacher-led sex education), others do not (e.g. community-based HIV/AIDS education). Likewise, some populations are relatively amenable to experimental work (e.g. children who attend the same school every day), others not at all (e.g. non-gay identifying men who have sex with other men who visit cruising spots on an irregular basis). It is a perverse notion indeed to admit as evidence only that which can be tailored to the experimental suit. We must be willing to be much more realistic and flexible in our approach to evaluation if we wish to capture that which is most relevant to the frequently complex world in which infections occur and may sometimes be avoided.

Gaining evidence

As indicated above, the position taken here is that evidence in HIV/AIDS education comes, by necessity and by design, from diverse sources. This position is supported by researchers with long-standing experience in HIV/AIDS. As Coates et al. [(Coates et al., 1996), p. 6] have recently remarked, `a variety of research methods (including observational studies, quasi-experimental designs, demonstration projects, and evaluations of interventions and societal changes) are needed to illuminate effective HIV prevention strategies'. Such a position finds tentative support among researchers from other social disciplines: `The realities of social work...render the use of RCTs difficult to realise routinely. Their cost, and the fairly loose relationship that often exists between our understanding of social problems and the responses we make to them...can mean it is not sensible to place all our evaluative eggs in the RCT basket' [(Macdonald, 1996), p. 43]. Indeed, one of the strongest present day advocates for RCTs in an earlier time recommended, `This is not, of course, to say that the procedure of randomized controlled evaluation is the only means to reliable knowledge, is sufficient in itself, or is always the right approach' [(Oakley, 1990), p. 193, emphases in original], a view which we fully endorse.

The problems of affording RCTs the same status in social science as in biomedicine have been outlined numerous times, such as in a recent editorial for Health Education Research (Tones, 1997): random allocation of participants, except on a relatively small scale, is expensive and problematic, and so artificial that generalization to `normal' populations is rarely possible. It is almost impossible to stop `leakage' from experimental to control groups or to avoid competing inputs from other sources. Experimental designs may work when comparing a narrowly focussed intervention with no intervention, but educational interventions typically involve complex multifactorial inputs.

Despite the limitations of comparative and experimental designs in health promotion generally, and in HIV/AIDS education specifically, they continue to be enthusiastically embraced in some quarters. The structure of academia and the politics of research funding are contributing agents (Chapman, 1993; Aggleton, 1997b). Academic advancement is largely dependent on research output. Research units rely on grants for their continued existence and individual researchers on funds to generate output. To persons untrained in the social and educational sciences, evidence accumulated through experimental trials has greater face validity. Wherever such persons control the funding of research and they indicate, whether by commission or by omission, that a particular research method is to be employed or is to be given preference in the distribution of research grants, researchers are likely to oblige. In a regimen where, for example, only controlled experimental studies are funded, researchers are faced with the stark choice between abandoning their research endeavours or ensuring that their research proposals include both experimental and control groups.2

The latter option may be pursued irrespective of the merits of the research method or the overall importance of the research question to be explored (Speller et al., 1997).

Community-based HIV/AIDS education campaigns have been highly effective in bringing about rapid, large-scale risk behaviour change, at least initially.3

Bye (Bye, 1990), for example, has described the significant nature of behavioural change among gay men in San Francisco in the early 1980s where community-wide attitudinal and behavioural change was accomplished through planned change in gay community norms, values and institutions. Multiple strategies focussed not on individuals, mass media, and personal testing and counselling, but on full-scale community mobilization, the use of local gay media, interpersonal communication within informal peer networks, grass-roots outreach, building community, and participation and involvement through volunteer work. The effectiveness of the community mobilization strategies was documented through various assessments of the acceptance of the education messages, and high awareness and recall of the messages. A substantial decline in unsafe sexual behaviours among gay men in San Francisco and the decline in the number of new HIV infections among this group provided sound evidence of success. The efficacy of multi-level, community-based interventions in restricting the development of the epidemic among gay men has also been documented in London (King, 1993) and in Sydney (Kippax et al., 1993).

As Bye (Bye, 1990) correctly points out, research methods which might be used in the evaluation of individual-level behavioural change interventions may not be appropriate for the evaluation of community-level strategies that emphasize sociocultural change accomplished through a diffusion process. Carefully designed experiments are rarely possible and certainly not appropriate when the emphasis is on the mobilization of communities, largely informal social networks and peer-to-peer communication. Evidence of success here is best furnished through data on the acceptance and awareness of the educational messages, community change (e.g. the uptake of safer sexual practices), decreased morbidity (e.g. fewer new infections) and project replication.

For determining whether or not a particular HIV/AIDS education programme has been successful, the standard outcome measures have often been changes in individual's knowledge, attitudes, beliefs, intentions, behavioural skills (such as condom use skills, refusal skills and negotiation skills) and behaviours (such as needle and syringe sharing, unprotected anal intercourse, and condom use). Self-report remains the most feasible procedure for obtaining information about AIDS risk behaviours in diverse populations, but it is useful to corroborate these data with reliable seroconversion data where possible (Auerbach et al., 1994) or valid population-level HIV incidence rates (Aral and Peterman, 1996). Low incidence and latency periods preclude the use of seroconversion rates as the outcome measure for experimental interventions which usually have few participants and typically are of short duration. However, in situations where HIV testing is normative and frequent (e.g. among gay men in Australia), seroconversion and HIV incidence rates may provide a useful outcome measure of the impact of community and population level HIV/AIDS education.

An alternative approach to understanding the effects of HIV/AIDS education might be to use mathematical modelling to provide evidence for how prevention activity may reduce HIV transmission. Multi-level models incorporating relevant demographic, epidemiological, risk behaviour and prevention parameters can be generated to show how a given educational programme may reduce infection (Auerbach et al., 1994). Furthermore, in a climate of fiscal restraint, cost–benefit analyses setting the cost of the intervention against the medical care and other savings that would be achieved through averting HIV infections could be invaluable for garnering support for a particular intervention. Both of these approaches offer ideal ways of engaging politicians and other decision makers in HIV work. Alongside RCTs, they have a distinctly political tone, for as has recently been suggested: `Perhaps the only real justification for the experimental design is `political', and derives from the need to provide `hard' evidence of effectiveness and efficiency in order to gain the funding needed to stay in business!' [(Tones, 1997), p. ii].

A feeling of déjà vu

Now we want to look back briefly to a preAIDS era, and a time when there was a marked shift away from quantitative and experimental evaluation towards alternative, non-experimental evaluation strategies among those working in another area of social intervention. A landmark in the development of educational evaluation was US domestic reaction to the launch of the Soviet Union's Sputnik I in 1957 and the subsequent launching of a person into space in 1961 (Haney, 1981; Glaser and Silver, 1994; McBride and Schostak, 1995). Questions were raised about the quality and content of American education, prompting federal support for school curriculum improvement in a number of subjects, notably mathematics and science. Large-scale federal and state-funded evaluations of teaching projects were implemented. The majority of these evaluations were quantitative and experimental in nature, and, in retrospect, yielded little by way of helpful information (McBride and Schostak, 1995). Following a great deal of debate in the educational literature throughout the 1960s and early 1970s, a new breed of evaluators and several different prototypes of evaluation gradually emerged. Experimental studies declined from a position as the pre-eminent research method to become by the mid-1970s just one among an array of techniques for gauging success (Jenkins, 1976).

This decline of experimental work in educational evaluation came about for many reasons. There is scope here [drawing on (Jenkins, 1976)] to mention only a few of the central issues of what was then a complex debate: the tradition of individual measurement associated with experimental studies attracted increasing criticism as being too narrowly focused. Evaluators were encouraged to take a broader view of what is eligible for collection as evaluation data. Various audiences have different information needs not all of which can be met through traditional experiments. `Formative' evaluations rather than `summative' ones were required to provide the data to improve ongoing courses. Education came to be seen as a dynamic process, subject to adaptation and change, with constant interaction between the end result and the means adopted to achieve it. It was considered bad evaluation practice to undervalue the experience of practitioners, in this case teachers in classrooms. It was recognized that all research is initially based on value judgements of some kind and so no research is totally `objective' or free from bias.

Other key players in the debate argued along similar lines. For example, Cronbach (Cronbach, 1963) called for evaluation that is formative and ongoing, deploying a wide battery of techniques necessary to describe and assess the outcomes of educational programmes. In his view, the greatest service evaluation can perform is to identify aspects of a course where revision is desirable. White (White, 1971) argued that there is no single process of evaluation: there are many different forms of evaluation, all of which should play a part in assessing whether or not a new curriculum is any good. Parlett (Parlett, 1975) criticized the experimental or `agricultural–botany' paradigm of evaluation with its reliance on testing, quantitative data and notions of objectivity, reliability and neutrality. In contrast, a more social-anthropological paradigm with a wider information base, and emphasis on interpreting, explaining and discovering patterns of coherence and interconnectedness was asked for.

What then were some of the alternative paradigms to emerge in educational evaluation in the 1970s and how might HIV/AIDS education benefit from these earlier understandings of the evaluation process? Jenkins (Jenkins, 1976) summarized six prototypes of evaluation, each serving different purposes and employing different technologies and raising different problems. Here we briefly outline the various prototypes, adapting Jenkins' (Jenkins, 1976) curriculum evaluation typology to the perspective of the HIV/AIDS educator. Table I sets out the main characteristics, strengths and limitations of each prototype.

Admittedly, Table I oversimplifies the situation particularly in respect of important ways in which the alternative models supplement each other. The key message is that there is a variety of methods for evaluating programmes, serving a wide range of purposes and audiences. The conventional paradigm of educational research (i.e. experimentation) is complemented by other approaches. The objective, goal-based model determines the degree to which desired changes in knowledge, attitudes and behaviour are actually taking place. Goal-free evaluation attempts to escape the `tunnel vision' of goal-based evaluation by assessing the worth of programmes against a set of a priori checkpoints. Self-study or self-deliberation is an ongoing evaluative activity focussing on the process of educational change to gain knowledge critical in strengthening our efforts and those of our clients. Action research, conducted by practitioners and involving the clients or community in the change process, fits within the self-study paradigm. Illuminative evaluation [embraced recently in the broader health education literature (Tones, 1997)] is concerned to portray day-to-day aspects of the setting: to isolate its significant features; delineate cycles of cause and effect; and comprehend relationships between beliefs and practices, and between organizational patterns and the responses of individuals (Parlett and Hamilton, 1972). Decision-making models serve planning decisions at the various stages of programme development: at the outset when setting objectives to suit a particular context; when determining the procedure to follow; when evaluating how well the programme is proceeding; and at the conclusion when deciding whether to reject or recycle the project.

Here we encounter no less than six different prototypes and six different ways of looking at what works in HIV/AIDS education. As has been our thesis throughout, no single model supplants all others. Rather, there are useful ways in which the techniques supplement and complement each other. In so doing, and by accumulating evidence from a variety of sources, data triangulation may be achieved. If the findings from different sources point in the same direction it is good evidence that a programme really works. So much the better if the data also contain rich information on the context and process so that the intervention may be replicated and improved.

Policy and practice

Given the urgency of the epidemic, and the need for effective prevention and care, it is imperative to examine the important relationship between knowledge and methods, on the one hand, and HIV/AIDS education policy and practice, on the other. The uncompromising position that the only useful data are those derived from RCTs fails to understand the rather loose association between evidence, public health policy and the idiosyncratic, at times contradictory, choices individuals make about their own health. A prime example is that of injecting drug use (IDU) and HIV transmission. There is clear evidence—some from RCTs, though most from comparative and case studies—that multiple prevention measures of education, needle/syringe exchange programmes, methadone maintenance and accepting injecting drug users as legitimate citizens are effective (Wodak, 1996). Some of the strongest evidence comes from data which show declining IDU HIV transmission in countries which have adopted such multiple prevention measures (e.g. Netherlands, Switzerland) compared with those which have not (e.g. US), but see also Stimson et al. (Stimson et al., 1998) concerning recent research in the developing world. Despite this evidence, there is still a reluctance on the part of politicians and policy makers in many countries to rethink their belief in the effectiveness of drug law enforcement and to implement proven measures such as needle/syringe exchanges. The key challenge for HIV prevention in this area is how to establish greater openness among political leaders and decision makers.

Experimental evaluations with their emphasis on behavioural change at the individual level have limited explanatory power, and tend to obscure key social and relational factors involved in behaviour, such as peer pressure, norms, obligations, emotions, cultural beliefs and organizational structures of communities (Auerbach et al., 1994; Aggleton, 1996b). For this reason, research involving ethnographies, social network analyses and community outreach evaluations is needed to illuminate the important linkages between social structure and individual behaviours. For example, it is imperative that we understand the social conditions which may contribute to HIV transmission in developing countries. Lack of access to universal health care, lack of funding for HIV/AIDS education, and inequalities between women and men are some of the many conditions which may fuel an epidemic. Likewise, an understanding of the social, political and cultural dimensions of IDU is needed if we are to deal with vulnerability to HIV among injecting drug users. Broadly, racial and ethnic discrimination and chronic unemployment are two factors that are integral to high rates of IDU and the subsequent transmission of HIV among certain segments of the population in Western industrialized countries. At the group level, initiation practices and notions of trust and friendship have to be understood and taken into account when developing HIV/AIDS education for injecting drug users (Auerbach et al., 1994; Loxley and Ovenden, 1995; Crofts et al., 1996).

One final point about research methods and HIV/AIDS educational practice is that successful interventions are of their essence likely to be multifaceted. In developed world gay communities, for example, where men have adopted safe sex practices and reduced HIV transmission, success has come about not through a single intervention that can be studied experimentally. On the contrary, change has been achieved through multiple approaches which operate synergistically—including community development, skills building, peer education, media campaigns, destigmatization, and counselling and testing (Bye, 1990; King, 1993; Kippax et al., 1993; Auerbach et al., 1994). Similarly, most interventions to reduce the transmission of HIV among injecting drug users contain many potentially active ingredients (Des Jarlais and Friedman, 1996). To gain real insight into successes in gay communities or among injecting drug users, illuminative rather than experimental evaluations are required. An RCT of this element or that may tell us something about an individual educational factor, but it reveals little about its relative contribution as part of a multifaceted programme. Of even greater concern, it may suggest to those who fund programmes that less expensive, singular interventions are sufficient to achieve programme success when clearly they are not.


We have argued against the hardline position that the only useful evidence in HIV/AIDS education comes from experimentation, including RCTs, and have called for a more encompassing view of what constitutes evidence in this field. The uncompromising proposition that we can only truly know which forms of education are effective after they have been judged against RCT criteria unnecessarily restricts our knowledge base, devalues useful alternative research methods, and compromises policy and practice. Experimental researchers are not the sole repository of knowledge, nor indeed are comparative evaluations the only way in which we have come to know about the world of HIV prevention and care. Much of what we know in HIV/AIDS education depends on insights from practitioners, programme managers, community members, theorists and scholars from non-experimental traditions. Some evidence has been gleaned from true experimental designs, other evidence from qualitative, illuminative and community-based work. Given the limited role that RCTs have in this area, putting a false premium on experimental and controlled studies will weaken rather than strengthen future efforts to minimize HIV transmission.

Looking back to a parallel debate about appropriate research methods for use in evaluating educational programmes more generally, we find that after some early cold war flirtation with experimental comparative designs, educational evaluators came later to adopt a more inclusive stance. Leading educational evaluators from the 1970s onwards have embraced multiple and diverse evaluation strategies. It is recognized that different technologies—goal-based evaluation, staff self-study, experimental control, illuminative evaluation and such-like—contribute worthwhile information according to the various purposes and audiences of the assessment. No single evaluation prototype can answer all relevant questions and none is without its limitations. The strongest evidence is that which is consistent and comes from a variety of investigators, employing dissimilar methods and taking different theoretical perspectives.

The recent focus on experimental studies and the call for more RCTs has one other hidden trap. It diverts our attention from important emerging issues in HIV/AIDS education. At the HIV Prevention Works Satellite Symposium of the XI International Conference on AIDS in Vancouver it was suggested that experimental procedures could contribute little to current priorities in HIV/AIDS education. Other techniques were needed to address priority areas for HIV prevention research, e.g. the effects of multiple interventions; how best to characterize and conceptualize interventions such as peer education, community mobilization and outreach work; how best to involve people living with HIV and AIDS in the design, implementation and evaluation of programmes; and how best to achieve commensurability and synergy between public sector and NGO/community-based responses (Aggleton, 1996b). If the current fashion for more RCTs and other experimental studies persists, there is a real danger that these and other important new directions for HIV/AIDS education will go quite unresearched.

Table I.

Prototypes of evaluation

Prototype Key emphasis Purpose Key activities Key viewpoint used to delimit study Outside experts needed HIV/AIDS educator involvement Risks Pay-off 
Objectivesmodel Objectives of the HIV/AIDS programme To measure target group progress towards objectives Specify the objectives; measure progress Programme manager;HIV/AIDSeducator Measurement specialists Conceptualize objectives; collect and interpret data Oversimplify goals; ignore processes Ascertain target group progress towards specific objectives 
Self-study Staff self-deliberation To review content and procedures of the HIV/AIDS programme DiscussHIV/AIDS programme; make professional judgements Programme manager;HIV/AIDSeducator None, but possible outside authenticationor technicalhelp Committee discussionsand decisionmaking Exhaust staff; ignore valuesof outsiders Increase staff responsibility 
Illumination Descriptionand judgementdata To report the ways different people see the HIV/AIDS programme Discover what stakeholders want to know; observe; gather opinions Audience of final report `Social anthropologists' Keep logs;give opinions Stir up value conflicts;ignore causes Broad picture of the natureof theHIV/AIDSprogramme 
Decisionmaking Decision making and accountability To facilitate rational and continual decision making Identify upcoming alternatives; study implications; set up quality control Administrator; programme manager Operations analysts Anticipate decisions, contingencies Overvalue efficiency; undervalue goals Programmes sensitive to feedback 
Experimentation Cause and effect relationships To seek simple but enduring explanation of what works in HIV/AIDS education Exercise experimental control and systematic variation Theorist;HIV/AIDSresearcher HIV/AIDS research; statistical analysts Tolerate experimental constraints Artificiality; lack of generalizability Rules for developing newHIV/AIDSprogrammes 
Goal-free evaluation Goal-free checklist To assess the HIV/AIDS programme's effects (irrespectiveof its goals) Ignore proponent claims; follow checklist Evaluator Independent analyst Make theHIV/AIDSprogramme accessible Overvalue documents and record keeping Data on effects with little risk that the evaluator will beco-opted byproponents 
Adapted from Jenkins (1976). 
Prototype Key emphasis Purpose Key activities Key viewpoint used to delimit study Outside experts needed HIV/AIDS educator involvement Risks Pay-off 
Objectivesmodel Objectives of the HIV/AIDS programme To measure target group progress towards objectives Specify the objectives; measure progress Programme manager;HIV/AIDSeducator Measurement specialists Conceptualize objectives; collect and interpret data Oversimplify goals; ignore processes Ascertain target group progress towards specific objectives 
Self-study Staff self-deliberation To review content and procedures of the HIV/AIDS programme DiscussHIV/AIDS programme; make professional judgements Programme manager;HIV/AIDSeducator None, but possible outside authenticationor technicalhelp Committee discussionsand decisionmaking Exhaust staff; ignore valuesof outsiders Increase staff responsibility 
Illumination Descriptionand judgementdata To report the ways different people see the HIV/AIDS programme Discover what stakeholders want to know; observe; gather opinions Audience of final report `Social anthropologists' Keep logs;give opinions Stir up value conflicts;ignore causes Broad picture of the natureof theHIV/AIDSprogramme 
Decisionmaking Decision making and accountability To facilitate rational and continual decision making Identify upcoming alternatives; study implications; set up quality control Administrator; programme manager Operations analysts Anticipate decisions, contingencies Overvalue efficiency; undervalue goals Programmes sensitive to feedback 
Experimentation Cause and effect relationships To seek simple but enduring explanation of what works in HIV/AIDS education Exercise experimental control and systematic variation Theorist;HIV/AIDSresearcher HIV/AIDS research; statistical analysts Tolerate experimental constraints Artificiality; lack of generalizability Rules for developing newHIV/AIDSprogrammes 
Goal-free evaluation Goal-free checklist To assess the HIV/AIDS programme's effects (irrespectiveof its goals) Ignore proponent claims; follow checklist Evaluator Independent analyst Make theHIV/AIDSprogramme accessible Overvalue documents and record keeping Data on effects with little risk that the evaluator will beco-opted byproponents 
Adapted from Jenkins (1976). 

View Large

This paper was prepared when P. Van de Ven was on a Public Health Travelling Fellowship awarded by the Australian National Health and Medical Research Council. He had discussions with various researchers, policy analysts and practitioners in the Netherlands and the UK but the views herein are the authors' own.


Aggleton, P. (


) Global priorities for HIV/AIDS intervention research.

International Journal of STD and AIDS


7 (Suppl. 2)




Google Scholar

Aggleton, P. (1996b) Research for prevention education: future directions. Keynote address at the HIV Prevention Works Satellite Symposium, XI International Conference on AIDS, Vancouver, Canada.

Google Scholar

Aggleton, P. (1997a) Success in HIV Prevention: Some Strategies and Approaches. AIDS Education and Research Trust, Horsham, West Sussex.

Google Scholar

Aggleton, P. (


) Developing and sustaining interventions for prevention and care.

National AIDS Bulletin






Google Scholar

Aral, S. O. and Peterman, T. A. (


). Measuring outcomes of behavioural interventions for STD/HIV prevention.

International Journal of STD and AIDS


7 (Suppl. 2)




Google Scholar

Auerbach, J. D., Wypijewska, C. and Brodie, H. K. H. (1994) AIDS and Behavior: An Integrated Approach. National Academy Press, Washington, DC.

Google Scholar

Berrow, D., Humphreys, C. and Hayward, J. (


) Understanding the relationship between research and clinical policy: a study of clinicians' views.

Quality in Health Care






Google Scholar

Bye, L. L. (1990) Moving beyond counselling and knowledge-enhancing interventions: a plea for community-level AIDS prevention strategies. In Ostrow D. G. (ed.), Behavioral Aspects of AIDS. Plenum, New York, pp. 157–167.

Google Scholar

Chapman, S. (


) Unravelling gossamer with boxing gloves: problems in explaining the decline in smoking.

British Medical Journal






Google Scholar

Coates, T. J., Chesney, M., Folkman, S., Hulley, S. B., Haynes-Sanstad, K., Lurie, P., VanOss Marin, B., Roos, L., Bunnett, V. and Du Wors, R. (


) Designing behavioural and social science to impact practice and policy in HIV prevention and care.

International Journal of STD and AIDS


7 (Suppl. 2)




Google Scholar

Crofts, N., Louie, R., Rosenthal, D. and Jolley, D. (


) The first hit: circumstances surrounding initiation into injecting.







Google Scholar


That said, we recognize that health education and health promotion, especially in the UK, encompasses a variety of positions and approaches. We also recognize that the views of individual researchers and the official stances of the organizations for which they work change over time.


One health promotion manager in England described an interesting approach received from a social research unit (personal communication with first author, 18 August 1997). A group of researchers had secured funding for an RCT of an HIV/AIDS intervention. The problem was that the research group did not have a specific intervention to evaluate and were thus turning to the health promotion manager to provide one. This anecdote draws attention to the absurdity of the methodological tail wagging the health promotion dog.


Whereas community-based approaches may have been successful initially, behaviour change in relation to gay men and HIV/AIDS may not have been sustained over time in all countries. This does not mean that the methods used to assess and validate the impact and effectiveness of early prevention measures were inappropriate, merely that there is an important difference between engendering an initial response and sustaining it over time. We recognize also that community-based strategies are usually most relevant to a particular time and place, and to a particular set of historical, social and medical circumstances.


Acquired immunodeficiency syndrome (AIDS) of humans is caused by two lentiviruses, human immunodeficiency viruses types 1 and 2 (HIV-1 and HIV-2). Here, we describe the origins and evolution of these viruses, and the circumstances that led to the AIDS pandemic. Both HIVs are the result of multiple cross-species transmissions of simian immunodeficiency viruses (SIVs) naturally infecting African primates. Most of these transfers resulted in viruses that spread in humans to only a limited extent. However, one transmission event, involving SIVcpz from chimpanzees in southeastern Cameroon, gave rise to HIV-1 group M—the principal cause of the AIDS pandemic. We discuss how host restriction factors have shaped the emergence of new SIV zoonoses by imposing adaptive hurdles to cross-species transmission and/or secondary spread. We also show that AIDS has likely afflicted chimpanzees long before the emergence of HIV. Tracing the genetic changes that occurred as SIVs crossed from monkeys to apes and from apes to humans provides a new framework to examine the requirements of successful host switches and to gauge future zoonotic risk.

Acquired Immune Deficiency Syndrome (AIDS) was first recognized as a new disease in 1981 when increasing numbers of young homosexual men succumbed to unusual opportunistic infections and rare malignancies (CDC 1981; Greene 2007). A retrovirus, now termed human immunodeficiency virus type 1 (HIV-1), was subsequently identified as the causative agent of what has since become one of the most devastating infectious diseases to have emerged in recent history (Barre-Sinoussi et al. 1983; Gallo et al. 1984; Popovic et al. 1984). HIV-1 spreads by sexual, percutaneous, and perinatal routes (Hladik and McElrath 2008; Cohen et al. 2011); however, 80% of adults acquire HIV-1 following exposure at mucosal surfaces, and AIDS is thus primarily a sexually transmitted disease (Hladik and McElrath 2008; Cohen et al. 2011). Since its first identification almost three decades ago, the pandemic form of HIV-1, also called the main (M) group, has infected at least 60 million people and caused more than 25 million deaths (Merson et al. 2008). Developing countries have experienced the greatest HIV/AIDS morbidity and mortality, with the highest prevalence rates recorded in young adults in sub-Saharan Africa ( Although antiretroviral treatment has reduced the toll of AIDS- related deaths, access to therapy is not universal, and the prospects of curative treatments and an effective vaccine are uncertain (Barouch 2008; Richman et al. 2009). Thus, AIDS will continue to pose a significant public health threat for decades to come.

Ever since HIV-1 was first discovered, the reasons for its sudden emergence, epidemic spread, and unique pathogenicity have been a subject of intense study. A first clue came in 1986 when a morphologically similar but antigenically distinct virus was found to cause AIDS in patients in western Africa (Clavel et al. 1986). Curiously, this new virus, termed human immunodeficiency virus type 2 (HIV-2), was only distantly related to HIV-1, but was closely related to a simian virus that caused immunodeficiency in captive macaques (Chakrabarti et al. 1987; Guyader et al. 1987). Soon thereafter, additional viruses, collectively termed simian immunodeficiency viruses (SIVs) with a suffix to denote their species of origin, were found in various different primates from sub-Saharan Africa, including African green monkeys, sooty mangabeys, mandrills, chimpanzees, and others (Fig. 1). Surprisingly, these viruses appeared to be largely nonpathogenic in their natural hosts, despite clustering together with the human and simian AIDS viruses in a single phylogenetic lineage within the radiation of lentiviruses (Fig. 2). Interestingly, close simian relatives of HIV-1 and HIV-2 were found in chimpanzees (Huet et al. 1990) and sooty mangabeys (Hirsch et al. 1989), respectively. These relationships provided the first evidence that AIDS had emerged in both humans and macaques as a consequence of cross-species infections with lentiviruses from different primate species (Sharp et al. 1994). Indeed, subsequent studies confirmed that SIVmac was not a natural pathogen of macaques (which are Asian primates), but had been generated inadvertently in US primate centers by inoculating various species of macaques with blood and/or tissues from naturally infected sooty mangabeys (Apetrei et al. 2005, 2006). Similarly, it became clear that HIV-1 and HIV-2 were the result of zoonotic transfers of viruses infecting primates in Africa (Hahn et al. 2000). In this article, we summarize what is known about the simian precursors of HIV-1 and HIV-2, and retrace the steps that led to the AIDS pandemic.

Figure 1.

Origins of human AIDS viruses. Old World monkeys are naturally infected with more than 40 different lentiviruses, termed simian immunodeficiency viruses (SIVs) with a suffix to denote their primate species of origin (e.g., SIVsmm from sooty mangabeys)....

Figure 2.

Phylogeny of lentiviruses. The evolutionary relationships among Pol sequences (∼ 770 amino acids) derived from various mammalian lentiviruses; host species are indicated at the right. Exogenous viruses are depicted in black, with HIV-1, HIV-2,...


Lentiviruses cause chronic persistent infections in various mammalian species, including bovines, horses, sheep, felines, and primates. The great majority of lentiviruses are exogenous, meaning that they are transmitted horizontally between individuals. However, it has recently become clear that, on several occasions in the past, lentiviruses have infiltrated their hosts’ germlines and become endogenous, vertically transmissible, genomic loci (Fig. 2). Examples include the rabbit endogenous lentivirus type K (RELIK), which became germ-line embedded approximately 12 million years ago (Katzourakis et al. 2007; van der Loo et al. 2009), and two prosimian endogenous lentiviruses, which independently invaded the germ-lines of both the grey mouse lemur (pSIVgml) and the fat-tailed dwarf lemur (pSIVfdl) about 4 million years ago (Gifford et al. 2008; Gilbert et al. 2009). These “viral fossils” are of particular interest because they provide direct evidence of the timescale of lentivirus evolution. Molecular clocks derived from extant SIV sequences suggested that ancestral SIVs existed only a few hundreds of years ago (Wertheim and Worobey 2009), but it has long been suspected that such analyses may grossly underestimate deeper evolutionary timescales (Sharp et al. 2000; Holmes 2003). Recent studies of SIV-infected monkeys on Bioko Island, Equatorial Guinea, partly substantiated this conclusion, showing that geographically isolated subspecies have been infected with the same type of SIV for at least 30,000 years and probably much longer (Worobey et al. 2010). The endogenous viruses in lemurs reveal that the span of evolutionary history of primate lentiviruses as a whole is at least two orders of magnitude greater still. Thus, it is possible that at least some SIVs, such as those infecting four closely related species of African green monkeys (Chlorocebus species), have coevolved with their respective hosts for an extended period of time, perhaps even before these hosts diverged from their common ancestor (Jin et al. 1994a). So far, SIV infections have only been found in African monkeys and apes, and so it seems likely that primate lentiviruses emerged in Africa sometime after the splits between lineages of African and Asian Old World monkeys, which are believed to have occurred around 6–10 million years ago (Fabre et al. 2009). However, because neither Asian nor New World primates have been sampled exhaustively, the conclusion that SIVs are restricted to African primates must remain tentative, especially because none of these primate species has been examined for endogenous forms of SIV (Ylinen et al. 2010). Thus, our understanding of the evolutionary history of primate lentiviruses is still incomplete.

To date, serological evidence of SIV infection has been reported for over 40 primate species, and molecular data have been obtained for most of these (also see Klatt et al. 2011). The latter studies have shown that the great majority of primate species harbor a single “type” or “strain” of SIV. That is, viral sequences from members of the same species form a monophyletic clade in evolutionary trees. This host-specific clustering indicates that the great majority of transmissions occur among members of the same species; however, there are also numerous documented instances when SIVs have crossed between species. Examples range from incidental “dead-end” infections (e.g., SIVver infections of baboons) (Jin et al. 1994b; van Rensburg et al. 1998) to the generation of new SIV lineages with substantial secondary spread (e.g., SIVgor infection of gorillas) (Van Heuverswyn et al. 2006). In addition, cross-species transmissions have generated mosaic SIV lineages through superinfection and recombination in species that already harbored an SIV (e.g., SIVsab infection of sabaeus monkeys) (Jin et al. 1994a). In both mandrills (Mandrillus sphinx) and moustached monkeys (Cercopithecus cephus), such recombination events have led to the emergence of a second SIV strain that cocirculates with the original virus (Souquiere et al. 2001; Aghokeng et al. 2007). Thus, it is clear that in addition to more long-standing virus/host relationships, a number of naturally occurring SIVs have emerged more recently as a result of cross-species transmission and recombination. What remains unknown is when and how often these cross-species transfers have occurred, what impact they had on virus and host biology, and whether AIDS is a frequent consequence of SIV host switching. The prevalence of naturally occurring SIV infections varies widely, ranging from 1% in some species to over 50% in others (Aghokeng et al. 2010), and it is tempting to speculate that less ubiquitous SIVs were acquired more recently and/or may be more pathogenic.


Of the many primate lentiviruses that have been identified, SIVcpz has been of particular interest because of its close genetic relationship to HIV-1 (Fig. 2). However, studies of this virus have proven to be challenging because of the endangered status of chimpanzees. The first isolates of SIVcpz were all derived from animals housed in primate centers or sanctuaries, although infection was rare in these populations. Collective analyses of nearly 2,000 wild-caught or captive-born apes identified fewer than a dozen SIVcpz positive individuals (Sharp et al. 2005). Because other primate species, such as sooty mangabeys and African green monkeys, are much more commonly infected, both in captivity and in the wild (Fultz et al. 1990; Phillips-Conroy et al. 1994; Santiago et al. 2005), this finding raised doubts about whether chimpanzees represented a true SIV reservoir. To resolve this conundrum, our laboratory developed noninvasive diagnostic methods that detect SIVcpz specific antibodies and nucleic acids in chimpanzee fecal and urine samples with high sensitivity and specificity (Santiago et al. 2003; Keele et al. 2006). These technical innovations, combined with genotyping methods for species and subspecies confirmation as well as individual identification, permitted a comprehensive analysis of wild-living chimpanzee populations throughout central Africa.

Chimpanzees are classified into two species, the common chimpanzee (Pan troglodytes) and the bonobo (Pan paniscus). Common chimpanzees have traditionally been further subdivided into a number of geographically differentiated subspecies (Groves 2001). Four subspecies were defined on the basis of mitochondrial DNA sequences (Gagneux et al. 1999), namely western (P. t. verus), Nigeria-Cameroonian (P. t. ellioti, formerly termed P. t. vellerosus), central (P. t. troglodytes), and eastern (P. t. schweinfurthii) chimpanzees. To determine the distribution of SIVcpz among these populations, fecal (and in some cases urine) samples were collected at different field sites and tested for the presence of virus specific antibodies. Antibody positive fecal specimens were then subjected to RNA extraction and reverse transcriptase polymerase chain reaction (RT-PCR) amplification to molecularly characterize the infecting virus strain. At select field sites, mitochondrial and microsatellite analyses of host DNA were also used to confirm sample integrity and to determine the number of tested individuals. Figure 3A summarizes current molecular epidemiological data derived from the analysis of over 7,000 chimpanzee fecal samples collected at nearly 90 field sites (Santiago et al. 2002, 2003; Worobey et al. 2004; Keele et al. 2006; Van Heuverswyn et al. 2007; Li et al. 2010; Rudicell et al. 2010). These studies have identified common chimpanzees as a natural SIVcpz reservoir, but also revealed important differences between the epidemiology of SIVcpz and that of other primate lentiviruses. First, only two of the four chimpanzee subspecies were found to harbor these viruses. SIVcpz was detected at multiple sites throughout the ranges of both central and eastern chimpanzees in an area ranging from Cameroon to Tanzania, but there was no evidence of infection in western and Nigeria-Cameroonian chimpanzees, nor in bonobos, despite testing of multiple communities. In addition, SIVcpz prevalence rates among central and eastern chimpanzees varied widely, ranging from 30% to 50% in some communities to rare or absent infection in others. In contrast, other SIVs, such as those of sooty mangabeys and African green monkeys, are much more widely and evenly distributed and infect their hosts at generally higher prevalence rates (Phillips-Conroy et al. 1994; Santiago et al. 2005). Nonetheless, the puzzle of why SIVcpz was so scarce among captive chimpanzees was finally resolved: As it turned out, most of these apes were imported from West Africa and thus were members of the P. t. verus subspecies, which does not harbor SIVcpz (Prince et al. 2002; Switzer et al. 2005).

Figure 3.

Geographic distribution of SIVcpz and SIVgor infections in sub-Saharan Africa. Field sites where wild-living (A) chimpanzees and bonobos, and (B) gorillas have been sampled are shown (each site is identified by a two-letter code; because of space limitations,...

The absence of SIVcpz from two of the four subspecies suggested that chimpanzees had acquired this virus more recently, after their divergence into different subspecies. Indeed, phylogenetic analyses of full-length proviral sequences revealed that SIVcpz represents a complex mosaic, generated by recombination of two lineages of SIVs that infect monkeys (Bailes et al. 2003). In the 5′ half of the genome, as well as the nef gene and 3′ LTR, SIVcpz is most closely related to SIVrcm from red-capped mangabeys (Cercocebus torquatus); however, in the vpu, tat, rev, and env genes, SIVcpz is most closely related to a clade of SIVs infecting several Cercopithecus species, including greater spot-nosed (C. nictitans), mustached (C. cephus), and mona (C. mona) monkeys (Bailes et al. 2003). Chimpanzees are known to hunt and kill other mammals, including monkeys (Goodall 1986), suggesting that they acquired SIV in the context of predation. The current range of the central chimpanzee overlaps those of red-capped mangabeys and the various Cercopithecus species, and so it is likely that the cross-species transmission events that led to the emergence of SIVcpz occurred in that area, and that SIVcpz later spread to eastern chimpanzees, although it is unclear whether this occurred during or subsequent to their divergence from the central subspecies. Importantly, all of more than 30 sequenced SIVcpz strains show an identical mosaic genome structure. Moreover, there is no evidence that chimpanzees harbor any other SIV, although they, as well as bonobos, are routinely exposed to SIVs through their hunting behavior (Mitani and Watts 1999; Surbeck and Hohmann 2008; Leendertz et al. 2011).


Initially, SIVcpz was thought to be harmless for its natural host. This was because none of the few captive apes that were naturally SIVcpz infected suffered from overt immunodeficiency, although in retrospect this conclusion was based on the immunological and virological analyses of only a single naturally infected chimpanzee (Heeney et al. 2006). In addition, SIV-infected sooty mangabeys and African green monkeys showed no sign of disease despite high viral loads in blood and lymphatic tissues (Paiardini et al. 2009), leading to the belief that all naturally occurring SIV infections are nonpathogenic. However, the sporadic prevalence of SIVcpz, along with its more recent monkey origin, suggested that its natural history might differ from that of other primate lentiviruses. To address this, a prospective study was initiated in Gombe National Park, Tanzania, the only field site where SIVcpz infected chimpanzees are habituated and so can be observed in their natural habitat.

Gombe is located in northwestern Tanzania on the shores of Lake Tanganyika. The park is home to three communities, termed Kasekela, Mitumba, and Kalande, which have been studied by Goodall and colleagues since the 1960s, 1980s, and 1990s, respectively (Pusey et al. 2007). Prospective studies of SIVcpz in Gombe began in 2000 (Santiago et al. 2002). By 2009, infections were documented in all three communities, with mean biannual prevalence rates of 13%, 12%, and 46% in Mitumba, Kasekela, and Kalande, respectively (Rudicell et al. 2010). Analysis of epidemiologically linked infections revealed that SIVcpz spreads primarily through sexual routes, with an estimated transmission probability per coital act (0.0008–0.0015) that is similar to that of HIV-1 among heterosexual humans (0.0011) (Gray et al. 2001; Rudicell et al. 2010). SIVcpz also appears to be transmitted from infected mothers to their infants, and in rare cases, possibly by aggression (Keele et al. 2009). Migration of infected females constitutes a major route of virus transmission between communities (Rudicell et al. 2010).

Behavioral and virological studies also provided insight into the pathogenicity of SIVcpz. Age-corrected mortality analyses revealed that infected chimpanzees had a 10- to 16-fold increased risk of death compared to uninfected chimpanzees (Keele et al. 2009). SIVcpz-infected females were less likely to give birth and had a much higher infant mortality rate than uninfected females. Postmortem analyses revealed significant CD4+ T-cell depletion in three infected individuals, but not in either of two uninfected individuals. One infected female, who died within 3 years of acquiring the virus, had histopathological findings consistent with end-stage AIDS. Taken together, these findings provided compelling evidence that SIVcpz was pathogenic in its natural host. Subsequent studies of both wild and captive chimpanzees confirmed these findings. By the end of 2010, the Kasekela and Mitumba communities had experienced three additional deaths, all SIVcpz related. One case concerned an infant born to an infected mother, whereas the other two were adult females, one of whom died with severe CD4+ T cell depletion within 5 years of acquiring SIVcpz (KA Terio et al., submitted). Moreover, demographic studies revealed that the Kalande community, which showed the highest SIVcpz prevalence rates (40%–50%), had suffered a catastrophic population decline, whereas the sizes of the Mitumba and Kasekela communities, which were infected at a much lower level (12%–13%), remained stable (Rudicell et al. 2010). It has been suggested that only members of the P. t. schweinfurthii subspecies, or more particularly the chimpanzees of Gombe, are susceptible to SIVcpz-associated pathogenicity (Weiss and Heeney 2009; Soto et al. 2010). However, a prospective study of orphaned chimpanzees in Cameroon identified an SIVcpz infected P. t. troglodytes ape that suffered from progressive CD4+ T cell loss, severe thrombocytopenia, and clinical AIDS (Etienne et al. 2011). Thus, it seems likely that SIVcpz has a substantial negative impact on the health, reproduction, and lifespan of all chimpanzees that harbor SIVcpz in the wild.


Noninvasive testing also led to the unexpected finding of a new SIV lineage in wild-living gorillas (Fig. 4). Analysis of ∼ 200 fecal samples from southern Cameroon identified several HIV/SIV antibody positive gorillas, and amplification of viral sequences revealed the existence of a new SIV lineage, termed SIVgor (Van Heuverswyn et al. 2006). This lineage fell within the radiation of SIVcpz, clustering with strains from P. t. troglodytes apes, suggesting that gorillas had acquired SIVgor by cross-species infection from sympatric chimpanzees (Fig. 4). Phylogenetic analyses of full-length SIVgor sequences confirmed this conclusion, indicating that SIVgor resulted from a single chimpanzee-to-gorilla transmission event estimated to have occurred at least 100 to 200 years ago (Takehisa et al. 2009). Subsequent screening of over 2500 fecal samples from 30 field sites across central Africa uncovered additional SIVgor infections, but only in western lowland gorillas (Gorilla gorilla gorilla) and not in eastern gorillas (Gorilla beringei). Although virus-positive apes were present at field sites more than 400 km apart, only four such sites were identified, and prevalence rates in these communities did not exceed 5% (Fig. 3B). Thus, SIVgor appears to be much less common in gorillas than SIVcpz is in chimpanzees (Neel et al. 2010). Whether SIVgor is more prevalent in communities in parts of west central Africa that have not yet been tested is not known. It is also unclear whether SIVgor is pathogenic for its host, because there has been no opportunity to study the natural history of this infection in either captive or wild-living gorillas. Finally, it remains a mystery how gorillas acquired SIVgor, because they are herbivores and do not hunt or kill other mammals. Nonetheless, gorillas and chimpanzees feed in the same forest areas, which must have led to at least one encounter that allowed transmission.

Figure 4.

HIV-1 origins. The phylogenetic relationships of representative SIVcpz, HIV-1, and SIVgor strains are shown for a region of the viral pol gene (HIV-1/HXB2 coordinates 3887–4778). SIVcpz and SIVgor sequences are shown in black and green, respectively....


HIV-1 has long been suspected to be of chimpanzee origin (Gao et al. 1999); however, until recently, the perceived lack of a chimpanzee reservoir left the source of HIV-1 open to question. These uncertainties have since been resolved by noninvasive testing of wild-living ape populations. It is now well established that all naturally occurring SIVcpz strains fall into two subspecies-specific lineages, termed SIVcpzPtt and SIVcpzPts, respectively, that are restricted to the home ranges of their respective hosts (Figs. 3 and ​4). Viruses from these two lineages are quite divergent, differing at about 30%–50% of sites in their Gag, Pol, and Env protein sequences (Vanden Haesevelde et al. 1996). Interestingly, population genetic studies have shown that central and eastern chimpanzees are barely differentiated, calling into question their status as separate subspecies (Fischer et al. 2006; Gonder et al. 2011). However, the fact that they harbor distinct SIVcpz lineages suggests that central and eastern chimpanzees have been effectively isolated for some time. In addition, molecular epidemiological studies in southern Cameroon have shown that SIVcpzPtt strains show phylogeographic clustering, with viruses from particular areas forming monophyletic lineages, and the discovery of SIVgor has identified a second ape species as a potential reservoir for human infection (Van Heuverswyn et al. 2006). Collectively, these findings have allowed the origins of HIV-1 to be unraveled (Keele et al. 2006; Van Heuverswyn et al. 2007).

HIV-1 is not just one virus, but comprises four distinct lineages, termed groups M, N, O, and P, each of which resulted from an independent cross-species transmission event. Group M was the first to be discovered and represents the pandemic form of HIV-1; it has infected millions of people worldwide and has been found in virtually every country on the globe. Group O was discovered in 1990 and is much less prevalent than group M (De Leys et al. 1990; Gurtler et al. 1994). It represents less than 1% of global HIV-1 infections, and is largely restricted to Cameroon, Gabon, and neighboring countries (Mauclere et al. 1997; Peeters et al. 1997). Group N was identified in 1998 (Simon et al. 1998), and is even less prevalent than group O; so far, only 13 cases of group N infection have been documented, all in individuals from Cameroon (Vallari et al. 2010). Finally, group P was discovered in 2009 in a Cameroonian woman living in France (Plantier et al. 2009). Despite extensive screening, group P has thus far only been identified in one other person, also from Cameroon (Vallari et al. 2011). Although members of all of these groups are capable of causing CD4+ T-cell depletion and AIDS, they obviously differ vastly in their distribution within the human population.

Figure 4 depicts a phylogenetic tree of representative HIV-1, SIVcpz, and SIVgor strains. It shows that all four HIV-1 groups, as well as SIVgor, cluster with SIVcpzPtt from central chimpanzees, identifying this subspecies as the original reservoir of both human and gorilla infections. HIV-1 groups N and M are very closely related to SIVcpzPtt strains from southern Cameroon, indicating that they are of chimpanzee origin. It has even been possible to trace their ape precursors to particular P. t. troglodytes communities. HIV-1 group N appears to have emerged in the vicinity of the Dja Forest in south-central Cameroon, whereas the pandemic form, group M, likely originated in an area flanked by the Boumba, Ngoko, and Sangha rivers in the southeastern corner of Cameroon (Keele et al. 2006; Van Heuverswyn et al. 2007). Existing phylogenetic data support a gorilla origin of HIV-1 group P, but too few SIVgor strains have been characterized to identify the region where this transmission might have occurred. In contrast, the immediate source of HIV-1 group O remains unknown, because there are no ape viruses that are particularly closely related to this group (Fig. 4). Thus, HIV-1 group O could either be of chimpanzee or gorilla origin. Nonetheless, the fact that group O and P viruses are more closely related to SIVcpzPtt than to SIVcpzPts suggests that both groups originated in west central Africa, which is consistent with their current distributions.

How humans acquired the ape precursors of HIV-1 groups M, N, O, and P is not known; however, based on the biology of these viruses, transmission must have occurred through cutaneous or mucous membrane exposure to infected ape blood and/or body fluids. Such exposures occur most commonly in the context of bushmeat hunting (Peeters et al. 2002). Whatever the circumstances, it seems clear that human–ape encounters in west central Africa have resulted in four independent cross-species transmission events. Molecular clock analyses have dated the onset of the group M and O epidemics to the beginning of the twentieth century (Korber et al. 2000; Lemey et al. 2004; Worobey et al. 2008). In contrast, groups N and P appear to have emerged more recently, although the sequence data for these rare groups are still too limited to draw definitive conclusions.

Eastern chimpanzees are endemically infected with SIVcpzPts throughout central Africa (Fig. 3A). Although prevalence rates have not been determined for all field sites, the P. t. schweinfurthii communities that have been studied show infection rates that are very similar to those found in P. t. troglodytes (Keele et al. 2006, 2009; Rudicell et al. 2010). Given that SIVcpzPtt strains have been transmitted to gorillas and humans on at least five occasions, it is striking that evidence of similar transmissions from eastern chimpanzees is lacking. There are a number of possible explanations. First, the risk of human exposure to SIVcpzPts may be lower, perhaps because of differences in the frequencies or types of human–ape interactions in central and east Africa. Second, SIVcpzPts infections of humans may have occurred, but gone unrecognized, because of limited human sampling and a lack of lineage-specific serological tests. Finally, as discussed below, SIVcpzPtt has evolved to overcome human restriction factors, such as tetherin, which may pose a barrier to cross-species transmission; because SIVcpzPts is highly divergent from SIVcpzPtt, viruses from this lineage may not have been able to adapt in the same way. Although SIVcpzPtt and SIVcpzPts strains replicate with similar kinetics in human CD4+ T cells in vitro (Takehisa et al. 2007), such cultures are unlikely to accurately recapitulate the conditions of viral replication and transmission in vivo.


Since its first discovery, HIV-2 has remained largely restricted to West Africa, with its highest prevalence rates recorded in Guinea-Bissau and Senegal (de Silva et al. 2008). However, overall prevalence rates are declining, and in most West African countries HIV-2 is increasingly being replaced by HIV-1 (van der Loeff et al. 2006; Hamel et al. 2007). Viral loads tend to be lower in HIV-2 than HIV-1 infected individuals, which may explain the lower transmission rates of HIV-2 and the near complete absence of mother-to-infant transmissions (Popper et al. 2000; Berry et al. 2002). In fact, most individuals infected with HIV-2 do not progress to AIDS, although those who do, show clinical symptoms indistinguishable from HIV-1 (Rowland-Jones and Whittle 2007). Thus, it is clear that the natural history of HIV-2 infection differs considerably from that of HIV-1, which is not surprising given that HIV-2 is derived from a very different primate lentivirus.

A sooty mangabey origin of HIV-2 was first proposed in 1989 (Hirsch et al. 1989) and subsequently confirmed by demonstrating that humans in West Africa harbored HIV-2 strains that resembled locally circulating SIVsmm infections (Gao et al. 1992; Chen et al. 1996). SIVsmm was found to be highly prevalent, both in captivity and in the wild, and to be nonpathogenic in its natural host (Silvestri 2005). In a wild-living sooty mangabey community, SIVsmm was primarily found in higher-ranking females, suggesting that virus infection had no appreciable negative effect on reproductive behavior or success (Santiago et al. 2005). Finally, the fact that sooty mangabeys are frequently hunted as agricultural pests in many areas of West Africa provided plausible routes of transmission.

Since its first isolation, at least eight distinct lineages of HIV-2 have been identified, each of which appears to represent an independent host transfer (Fig. 5). By analogy with HIV-1, these lineages have been termed groups A–H, although only groups A and B have spread within humans to an appreciable degree. Group A has been found throughout western Africa (Damond et al. 2001; Peeters et al. 2003), whereas group B predominates in Cote d’Ivoire (Pieniazek et al. 1999; Ishikawa et al. 2001). All other HIV-2 “groups” were initially identified only in single individuals, suggesting that they represent incidental infection with very limited or no secondary spread. Of these, groups C, G, and H have been linked to SIVsmm strains from Cote d’Ivoire, group D is most closely related to an SIVsmm strain from Liberia, and groups E and F resemble SIVsmm strains from Sierra Leone (Gao et al. 1992; Chen et al. 1996, 1997; Santiago et al. 2005). Because of their sporadic nature, groups C–H have been assumed to represent “dead-end” transmissions. However, a second divergent HIV-2 strain has recently been placed in group F (Fig. 5). This virus was identified in an immigrant in New Jersey, who came from the same geographic area in Sierra Leone where this lineage was first discovered (Smith et al. 2008). Unlike the original index case, the newly identified group F infection was associated with reduced CD4 T cell counts and high viral loads (Smith et al. 2008). It is presently unknown whether group F has been spreading cryptically in humans, or whether the two group F viruses represents independent transmissions from sooty mangabeys.

Figure 5.

HIV-2 origins. The phylogenetic relationships of representative SIVsmm and HIV-2 strains are shown for a region of the viral gag gene (SIVmac239 coordinates 1191–1921). SIVsmm and SIVmac are shown in black; the eight groups of HIV-2, each of which...


HIV and SIV must interact with a large number of host proteins to replicate in infected cells (Fu et al. 2009; Ortiz et al. 2009). Because the common ancestor of Old World monkeys and apes existed around 25 million years ago, the divergence of these host proteins may pose an obstacle to cross-species infection. In addition, primates (including humans) encode a number of host restriction factors, which have evolved as part of their innate immune response to protect against infection with a wide variety of viral pathogens (Malim and Emerman 2008; Neil and Bieniasz 2009; Kajaste-Rudnitski et al. 2010). Although viruses have, in turn, found ways to antagonize these restriction factors, these countermeasures are frequently species-specific. Thus, a number of adaptive hurdles have to be overcome before primate lentiviruses can productively infect a new species.

The first evidence of host-specific adaptation of HIV-1 came from an analysis of sites in the viral proteome that were highly conserved in the ape precursors of HIV-1, but changed—in the same way—each time these viruses crossed the species barrier to humans (Wain et al. 2007). This analysis identified one site in the viral matrix protein (Gag-30) that encoded a Met in all known strains of SIVcpzPtt and SIVgor but switched to an Arg in the inferred ancestors of HIV-1 groups M, N, and O, and has subsequently been conserved as a basic amino acid (Arg or Lys) in most strains of HIV-1. The fact that the same nonconservative amino acid substitution occurred on each of the three branches involving cross-species transmission to humans suggested that this matrix residue was under strong host-specific selection pressure. This conclusion was subsequently confirmed by two additional observations. First, it was found that a reciprocal transmission, in which a chimpanzee was experimentally infected with HIV-1, led to the reversion of this host-specific signature; that is, a basic residue at Gag-30 in HIV-1 changed back to a Met on in vivo propagation in a chimpanzee (Mwaengo and Novembre 1998). Second, it was found that in chimpanzee CD4+ T lymphocytes, a virus with a Met at position 30 replicated more efficiently that an otherwise isogenic virus with a Lys at the same position, whereas the opposite was true in human cells (Wain et al. 2007). Interestingly, only one of the two recently discovered HIV-1 group P strains has switched from a Met to a Lys at Gag-30 (Vallari et al. 2011). Although the structure of the HIV-1 matrix protein has been determined (Hill et al. 1996), the function of the amino acid at position 30 is not known, and it remains to be determined why this site is under such strong selection pressure.

The potential of an SIV to infect a new primate species is also influenced by its ability to counteract different host restriction factors. Three classes of restriction factors have been shown to constitute barriers to SIV cross-species transmission. These include (1) APOBEC3G (apolipoprotein B mRNA editing enzyme catalytic polypeptide-like 3G), which interferes with reverse transcription (Sheehy et al. 2002); (2) TRIM5α (tripartite motif 5α protein), which interferes with viral uncoating (Stremlau et al. 2004); and (3) tetherin (also termed BST-2 and CD317) which inhibits the budding and release of virions from infected cells (Neil et al. 2008). Of these, tetherin appears to have had the greatest impact on the precursors of HIV-1 and HIV-2 (Fig. 6). Tetherin is comprised of a cytoplasmic amino-terminal region, a trans-membrane domain, a coiled-coiled extracellular domain, and a carboxy-terminal glycosylphosphatidylinositol (GPI) anchor (Fig. 6B). Recent studies have shown that most SIVs use their Nef protein to remove tetherin from the cell surface by targeting its cytoplasmic domain (Jia et al. 2009; Zhang et al. 2009). In contrast, HIV-1 (Neil et al. 2008; Van Damme et al. 2008) as well as SIVs from greater spot-nosed, mona, moustached, and Dent’s monkeys (Sauter et al. 2009; Schmokel et al. 2011) use their Vpu protein to degrade tetherin by binding to its membrane-spanning domain (Iwabu et al. 2009; Rong et al. 2009). Still other viruses use their envelope glycoprotein to interfere with tetherin, by interacting with either its extracellular or its cytoplasmic domain (Bour et al. 1996; Gupta et al. 2009; Le Tortorec and Neil 2009; Serra-Moreno et al. 2011). These various antitetherin responses appear to have emerged as a direct result of host-specific selection pressures following cross-species transmission (Sauter et al. 2009; Evans et al. 2010; Lim et al. 2010).

Figure 6.

Tetherin function and virus specific antagonism in different primate hosts. (A) Mechanism of restricted virion release by tetherin; two alternative models are shown (for details, see Evans et al. 2010); (B) Viral antagonists of tetherin and their sites...

Because of the constant onslaught of viral pathogens, host restriction factors evolve rapidly (Sawyer et al. 2004, 2005; McNatt et al. 2009; Lim et al. 2010). Most notably, the human tetherin gene differs from that of other apes by a five-codon deletion in the region encoding the cytoplasmic domain (Sauter et al. 2009). Because Nef interacts with the cytoplasmic domain of tetherin, this deletion rendered the SIVcpz Nef protein inactive on transmission to humans. Gorilla tetherin does not have this deletion, and thus Nef continued to function as a tetherin antagonist on transmission of SIVcpz to gorillas (Fig. 6C). Thus, to facilitate replication in humans, SIVcpz and SIVgor had to find alternative routes to overcome tetherin. One option was to switch back to using Vpu, as in their monkey ancestors (Fig. 6C). However, this required that the SIVcpz and SIVgor Vpu proteins regained this function because neither has antitetherin activity in chimpanzee or humans cells. Not surprisingly, this was not successful in all instances (Fig. 6C). When representatives of each of the HIV-1 groups were analyzed, only the Vpu proteins of group M viruses showed potent antitetherin activity (Sauter et al. 2009). Group O and P Vpu proteins were completely inactive, whereas the group N Vpu showed only marginal activity (Sauter et al. 2009; Kirchhoff 2010). Moreover, even the latter adaptation came at a cost, because the group N Vpu lost its ability to down-modulate CD4. Thus, of the four transmitted ape viruses, only the precursor of HIV-1 group M succeeded in mounting a full antitetherin defense in human cells. It may thus not be a coincidence that only HIV-1 group M resulted in a global epidemic (Gupta and Towers 2009; Sauter et al. 2009).

Like many other SIVs, SIVsmm does not encode a vpu gene and uses its Nef protein to combat tetherin. Thus, on transmission to humans, SIVsmm had to overcome the same hurdle of a cytoplasmic tail deleted human tetherin. In this case, the envelope glycoprotein (gp41) was recruited as an alternative tetherin antagonist (Le Tortorec and Neil 2009). However, this escape mechanism has thus far only been observed for epidemic HIV-2 group A, and it will be interesting to determine which, if any, of the other HIV-2 groups have acquired a similar antitetherin activity. It seems likely that the cytoplasmic domain deletion in human tetherin has represented a significant barrier to SIV zoonoses, including those viruses that have succeeded in infecting humans.


HIV-1 evolves around one million times faster than mammalian DNA (Li et al. 1988; Lemey et al. 2006), because the HIV-1 reverse transcriptase is error prone and the viral generation time is short (Ho et al. 1995; Wei et al. 1995). This propensity for rapid genetic change has provided a unique opportunity to gain insight into when and where the AIDS pandemic had its origin. Phylogenetic and statistical analyses have dated the last common ancestor of HIV-1 group M to around 1910 to 1930, with narrow confidence intervals (Korber et al. 2000; Worobey et al. 2008). This indicates that after pandemic HIV-1 first emerged in colonial west central Africa, it spread for some 50 to 70 years before it was recognized. The probable location of the early epidemic has also been identified. Molecular epidemiological studies have indicated that most, if not all, of the early diversification of HIV-1 group M likely occurred in the area around Kinshasa, then called Leopoldville. All of the known HIV-1 group M subtypes were identified there, as well as additional lineages that have remained restricted to this area (Vidal et al. 2000). Leopoldville was also the place where the earliest strains of HIV-1 group M were discovered (Zhu et al. 1998; Worobey et al. 2008). Genetic analysis of infected blood and tissue samples collected from residents of Kinshasa in 1959 and 1960, respectively, revealed that HIV-1 had already diversified into different subtypes by that time (Worobey et al. 2008). Finally, demographic data indicate that pandemic HIV-1 emerged at a time when urban populations in west central Africa were expanding (Worobey et al. 2008). Leopoldville was the largest city in the region at that time and thus a likely destination for a newly emerging infection. Moreover, rivers, which served as major routes of travel and commerce at the time, would have provided a link between the chimpanzee reservoir of HIV-1 group M in southeastern Cameroon and Leopoldville on the banks of the Congo (Sharp and Hahn 2008). Thus, all current evidence points to Leopoldville/Kinshasa as the cradle of the AIDS pandemic.

As HIV-1 group M spread globally, its dissemination involved a number of population bottlenecks—founder events, which led to the predominance of different group M lineages, now called subtypes, in different geographic areas. HIV-1 group M is currently classified into nine subtypes (A–D, F–H, J, K), as well as more than 40 different circulating recombinant forms (CRFs), which were generated when multiple subtypes infected the same population (Taylor et al. 2008). It has been possible to trace the migration pathways of some of these subtypes and CRFs. For example, subtypes A and D originated in central Africa, but ultimately established epidemics in eastern Africa, whereas subtype C was introduced to, and predominates in, southern Africa from where it spread to India and other Asian countries. Subtype B, which accounts for the great majority of HIV-1 infections in Europe and the Americas, arose from a single African strain that appears to have first spread to Haiti in the 1960s and then onward to the US and other western countries (Gilbert et al. 2007). The recombination event that created CRF01 probably occurred in Central Africa, but this viral lineage was first noted in the late 1980s causing a heterosexual epidemic in Thailand, contemporary with subtype B viruses spreading among intravenous drug users (Taylor et al. 2008). CRF01 has gone on to dominate the AIDS epidemic in southeast Asia. Although the initial distribution of these subtypes and CRFs may have been largely caused by chance events, recent studies have suggested that viruses of different subtypes vary in their biological properties, which may influence their epidemiology (Taylor et al. 2008). For example, subtype D has been associated with greater pathogenicity (Kiwanuka et al. 2010) and an increased incidence of cognitive impairment and AIDS dementia (Sacktor et al. 2009). It thus appears that not only the genetic but also the biological diversity of HIV-1 group M subtypes and CRF is increasing.


Although primate lentiviruses were first identified in the late 1980s, it is only very recently that the complexities of their evolutionary origins, geographic distribution, prevalence, natural history, and pathogenesis in natural and nonnatural hosts have been appreciated. The preceding sections summarize what is known about the origins and evolution of the simian relatives of HIV-1 and HIV-2, their propensity to cross species barriers, and the host factors that govern such transmissions. Given that there are numerous additional primate species that harbor SIV, the question arises what to expect from future zoonoses? Host-specific restriction factors play a major role in preventing cross-species transmission, and they may well be responsible for the fact that only two types of SIV have thus far succeeded in colonizing humans. However, as exemplified by the various HIV-1 and HIV-2 outbreaks, they are certainly not insurmountable. Determining the entire spectrum of host restriction factors and their mechanisms of action will be required to gauge the likelihood of future zoonoses. In this regard, the role of tetherin should be examined further. Because this protein “tethers” virions to the cell surface, a lack of effective antitetherin measures may result in reduced titers of infectious virus in genital secretions. This may explain why the precursors of the rare groups of HIV-1 and HIV-2 were able to infect humans but unable to establish epidemic infections.

From the above, it is also clear that any newly introduced SIV must replicate to some extent to accumulate the necessary mutations that are required to adapt to divergent host proteins and restriction factors. Circumstances that enhance human-to-human passage would thus be expected to increase the chance of such adaptation. It has been suggested that large-scale injection campaigns conducted in west central Africa at the beginning of the twentieth century (Pepin et al. 2006, 2010; Pepin and Labbe 2008), together with the destabilization of social structures (Chitnis et al. 2000), the rapid growth of cities (Worobey et al. 2008), and an increased prevalence in sexually transmitted diseases, including genital ulcers (de Sousa et al. 2010), may have facilitated the early dissemination and adaptation of both HIV-1 and HIV-2. The fact that HIV-1 groups M and O as well as HIV-2 group A all emerged around the same time is consistent with this hypothesis (Korber et al. 2000; Lemey et al. 2003, 2004; Worobey et al. 2008; de Sousa et al. 2010). However, whether these medical interventions and/or social factors really played a role in the emergence of HIV-1 and HIV-2, and more importantly, whether such “jump-starts” were required to spawn the AIDS pandemic, will remain unknown.

Finally, it is important to view the pathogenesis of simian and human AIDS viruses in the context of their evolution (Kirchhoff 2009). One feature of pathogenic HIV-1 and HIV-2 infection that distinguishes them from nonpathogenic SIV infections is a high level of chronic immune activation, which is a strong predictor of disease progression. In HIV-1 infection, this immune activation is fueled, at least in part, by the inability of the Nef protein to down-modulate TCR-CD3, a function that is conserved in most nonpathogenic SIV infections (Schindler et al. 2006; Arhel and Kirchhoff 2009). Lack of this Nef function is associated with increased T-cell activation and apoptosis in vitro and loss of CD4+ T cells in natural SIV infection in vivo (Schindler et al. 2008). A higher state of T- cell activation is associated with enhanced levels of proviral transcription and viral replication, but also with increased expression of interferon-induced restriction factors, such as tetherin. In HIV-1, Vpu compensates for this by providing potent antitetherin activity. It has thus been proposed that to overcome the barriers of cross-species transmission, primate lentiviruses must induce an inflammatory milieu to increase their ability to replicate and accumulate mutations necessary for more adaptation (Kirchhoff 2009). If this were indeed the case, AIDS would be an inevitable consequence of SIV cross-species transmission.


We thank Lilian Pintea for maps of current ape ranges, Frank Kirchhoff for unpublished results, Gerald Learn for phylogenetic tree construction, and Jamie White for artwork and manuscript preparation. This work was supported by grants from the National Institutes of Health (R01 AI50529, R01 AI58715, P30 AI 27767), and the Bristol Myers Freedom to Discover Program.


*Reference is also in this collection.

Aghokeng AF, Bailes E, Loul S, Courgnaud V, Mpoudi-Ngolle E, Sharp PM, Delaporte E, Peeters M 2007. Full-length sequence analysis of SIVmus in wild populations of mustached monkeys (Cercopithecus cephus) from Cameroon provides evidence for two co-circulating SIVmus lineages. Virology360: 407–418 [PMC free article][PubMed]

Aghokeng AF, Ayouba A, Mpoudi-Ngole E, Loul S, Liegeois F, Delaporte E, Peeters M 2010. Extensive survey on the prevalence and genetic diversity of SIVs in primate bushmeat provides insights into risks for potential new cross-species transmissions. Infect Genet Evol10: 386–396 [PMC free article][PubMed]

Apetrei C, Kaur A, Lerche NW, Metzger M, Pandrea I, Hardcastle J, Falkenstein S, Bohm R, Koehler J, Traina-Dorge V, et al. 2005. Molecular epidemiology of simian immunodeficiency virus SIVsm in U.S. primate centers unravels the origin of SIVmac and SIVstm. J Virol79: 8991–9005 [PMC free article][PubMed]

Apetrei C, Lerche NW, Pandrea I, Gormus B, Silvestri G, Kaur A, Robertson DL, Hardcastle J, Lackner AA, Marx PA 2006. Kuru experiments triggered the emergence of pathogenic SIVmac. AIDS20: 317–321 [PubMed]

Arhel NJ, Kirchhoff F 2009. Implications of Nef: Host cell interactions in viral persistence and progression to AIDS. Curr Top Microbiol Immunol339: 147–175 [PubMed]

Bailes E, Gao F, Bibollet-Ruche F, Courgnaud V, Peeters M, Marx PA, Hahn BH, Sharp PM 2003. Hybrid origin of SIV in chimpanzees. Science300: 1713. [PubMed]

Barouch DH 2008. Challenges in the development of an HIV-1 vaccine. Nature455: 613–619 [PMC free article][PubMed]

Barre-Sinoussi F, Chermann JC, Rey F, Nugeyre MT, Chamaret S, Gruest J, Dauguet C, Axler-Blin C, Vezinet-Brun F, Rouzioux C, et al. 1983. Isolation of a T-lymphotropic retrovirus from a patient at risk for acquired immune deficiency syndrome (AIDS). Science220: 868–871 [PubMed]

Berry N, Jaffar S, Schim van der Loeff M, Ariyoshi K, Harding E, N’Gom PT, Dias F, Wilkins A, Ricard D, Aaby P, et al. 2002. Low level viremia and high CD4% predict normal survival in a cohort of HIV type-2-infected villagers. AIDS Res Hum Retroviruses18: 1167–1173 [PubMed]

Bour S, Schubert U, Peden K, Strebel K 1996.


Leave a Reply

Your email address will not be published. Required fields are marked *