Portal de Periódicos da CAPES

Epidemiological considerations in planning HIV preventive vaccine trials

2001; Lippincott Williams & Wilkins; Volume: 15; Linguagem: Inglês

10.1097/00002030-200100005-00008

1473-5571

José Esparza, Donald S. Burke,

Vaccine Coverage and Hesitancy

Introduction There is growing consensus that a safe, effective and accessible HIV vaccine offers the best long-term hope to control the AIDS epidemic, especially in developing countries, where more than 95% of all new HIV infections are occurring [1,2]. Much has been learned since 1987, when the first human trial of an HIV candidate vaccine was launched in the United States. The first generation of candidate vaccines was aimed mainly at inducing neutralizing antibodies using monomeric envelope glycoproteins derived from laboratory-adapted strains of HIV-1 (X4 strains, those using CXCR4 as second receptor). Today, a whole range of new candidate vaccines is being developed, with the goal of inducing broadly reactive neutralizing antibodies against clinical isolates of HIV-1 (R5 strains, those using CCR5 as second receptor), as well as cell-mediated immune responses [3]. Many of those candidate vaccines have been tested, or are undergoing testing, in primate models. Although animal experiments are instructive, human trials remain in the critical path to develop HIV vaccines, and the expectation is that several new candidate vaccines should soon move to trials in human volunteers. In the absence of definitive information regarding potential immune correlates of protection, several products, based on different vaccine concepts, would have to be tested simultaneously in phase I/II trials, to evaluate their safety and immunogenicity in human volunteers. The best candidate vaccines will then be selected for phase III efficacy evaluation in large-scale field trials. Several trials will be needed to assess the efficacy of different vaccine concepts, against different virus subtypes, and in different populations (which may differ in route of transmission of the virus, and genetic, nutritional or health characteristics). To address these multiple questions, a number of phase III trials will have to be conducted in both industrialized and developing countries, and this will require intense international cooperation and collaboration. Identification and strengthening of potential sites for trials The effort to identify and strengthen potential sites for HIV vaccine efficacy trials started in the United States and in other industrialized and developing countries during the early 1990s [4]. Before field trials of HIV vaccines could be conducted, several critical issues had to be addressed, such as strengthening the research infrastructure for clinical trials, monitoring HIV variability, and ethical and social-behavioural matters [5]. The conduct of large-scale trials requires strong political and community commitment and support. Scientists, community members, and national authorities from developing countries should be involved as full partners in the planning and conduct of vaccine trial preparations. This serves two functions: first, to avoid foreseeable logistics and technical problems; and second, to begin to build a relationship of trust that will be essential for the successful completion of difficulty and lengthy efficacy trials. Vaccine trial preparedness efforts implemented during the 1990s included the development of prospective cohorts of HIV-negative volunteers at high risk of HIV infection that could be enrolled into efficacy trials. Unfortunately, only one candidate vaccine entered phase III clinical trials in the United States and Thailand, in 1998 and 1999, respectively [6]. As a result, few of those well-characterized cohorts were ever used to conduct phase III trials, although they have provided important information that would facilitate the implementation of future trials. The establishment of prospective cohorts is an expensive and time-consuming exercise, and most of the recently published data, which will be reviewed in this paper, come from studies that were initiated several years ago. Transitioning from phase I to phase III trials The clinical and epidemiological requirements for phase I, II and III trials are different, but they should be considered as sequential steps of a continuous process. Phase I trials provide initial safety and immunogenicity data, and are usually conducted among small numbers of volunteers (30-50) at relatively low risk of HIV infection. Phase II trials provide additional safety and immunogenicity information in different populations, and are usually conducted among a larger number of volunteers (in the hundreds) including people at higher risk of HIV infection representing the population in which the phase III trial would be implemented. Phase III trials are designed as large-scale, double-blinded controlled trials, involving thousands of volunteers (depending on HIV incidence), and conducted to assess the efficacy of the candidate vaccine in preventing HIV infection or disease. Traditionally, phase I trials are conducted in the country of origin of the candidate vaccine, usually an industrialized country. This approach has been justified by several scientific, ethical and political reasons. From the scientific point of view, it is essential that the first introduction of a candidate vaccine in humans is properly monitored, especially in relation to potential side effects, and some developing countries may not have the necessary scientific infrastructure readily available. In addition, some developing countries may lack appropriate regulatory mechanisms to identify candidate vaccines that are ready for testing in human trials. Having said this, it is important to emphasize that there are no formal ethical impediments for the conduct of phase I trials in developing countries, provided such trials are carried out for valid scientific and public health reasons, and provided the host country can ensure sufficient scientific standards and ethical safeguards [7]. One satisfactory approach could be to begin trials simultaneously in both the country where the candidate vaccine was produced and in the developing country where phase III trials are planned. The conduct of phase I/II trials requires considerable clinical and laboratory expertise. On the laboratory side, it is critical to use state-of-the-art technology to assess humoral and/or cellular immune responses to the candidate vaccine. Likewise, it is essential that vaccine evaluation centres have access to the appropriate methodology to study virologic and immunological aspects related to intercurrent (breakthrough) HIV infections, including measurement of virus loads and the genetic, biological and immunological characterization of intercurrent strains. The complexity of these techniques, which are constantly evolving, requires a cooperative approach, in collaboration with selected reference laboratories. The conduct of phase I/II trials not only provides critical information needed to make decisions about moving to phase III trials, but also offers an opportunity for training and capacity building in preparation for the larger and more complex phase III trials. In addition, phase I/II trials offer an opportunity for communities to become acquainted with and to build consensus around HIV vaccine activities in the country, and to develop mechanisms for the scientific and ethical review and monitoring of vaccine-related research. The first HIV vaccine trial in a developing country was initiated in 1993 in China. Since then, 11 phase I/II trials have been conducted or initiated in developing countries, with seven trials conducted in Thailand, and one in each of Brazil, China, Cuba and Uganda [1]. Results from phase I/II trials in developing countries were recently reported from Thailand, Cuba and Uganda. Two different groups in Thailand [8,9] tested gp120 candidate vaccines, with safety and immunogenicity results comparable with those obtained from trials in the United States [10]. Although these initial Thai trials used candidate vaccines based on subtype B strains (MN and SF-2), they provided a basis for the development of similar antigens based on subtype E, which is the HIV genetic variant prevalent in Thailand [11-14]. Investigators from Cuba also reported initial results of a phase I trial using a multi-antigenic synthetic peptide bearing several V3 sequences (TAB9) [15,16]. Initial results from the first HIV vaccine trial in Africa were reported at the XIII International Conference on AIDS, Durban, South Africa [17], using a subtype B canarypox-HIV recombinant candidate vaccine (ALVAC vCP205). Because most infections in Uganda are caused by subtype A and D strains [18], the choice of a subtype B vaccine was based on an anticipated cross-recognition of conserved cytotoxic T lymphocyte epitopes between different HIV subtypes [19,20]. Although a relatively low frequency of HIV-specific cytotoxic T lymphocyte responses was documented in vaccinated volunteers, the trial successfully demonstrated the feasibility of conducting scientifically valid vaccine research in Uganda. Several new phase I/II trials are expected to begin in industrialized and developing countries in 2001. A multicentric phase II trial has been approved for implementation in several countries in Latin America and the Caribbean (Brazil, Haiti, and Trinidad & Tobago) to assess the safety and immunogenicity of a subtype B prime-boost regimen (canarypox-HIV followed by gp120), in anticipation of a phase III trial planned to be conducted in those countries and in the United States. The next phase I/II trial in Africa is planned in Kenya, using DNA immunization and a Modified Vaccinia Ankara vector expressing subtype A antigens [21]. The conduct of phase III trials presents some difficult challenges, recently reviewed by Excler and Beyer [22]. In addition to the necessary political and community support, the host country must satisfy a number of epidemiological requirements (to be reviewed in the following two sections). It should also have in place the necessary mechanisms to guarantee an appropriate review and approval process (regulatory, scientific and ethical). And it should have the clinical, laboratory and epidemiological infrastructures to handle thousands of volunteers participating in a trial. Identification of populations with high HIV incidence for phase III trials To obtain statistically significant information on potential protective efficacy of HIV candidate vaccines, phase III trials should be conducted in well-defined populations with relatively high HIV incidence. The calculation of sample size depends primarily on the incidence of the primary end-point that is used to measure vaccine efficacy, the duration of study follow-up, the rate of retention of trial participants, and the minimum level of efficacy that the trial is powered to detect. For example, assuming a 1-year recruitment period, 3 years of follow-up, 5% of volunteers lost to follow-up per year, and a 1% annual HIV incidence, 2500 participants would be required for a placebo-controlled, two-arm trial powered to detect a minimum 60% efficacy with a 30% lower 95% confidence bound [23]. Potential populations suitable for HIV vaccine efficacy trials can be identified from cohorts of HIV-negative volunteers at higher risk of HIV infection [24]. Accurate estimates of HIV incidence can then be prospectively estimated, taking into consideration the effect of nonvaccine interventions (education, counselling, condom promotion, and, possibly, treatment of sexually transmitted diseases). In addition, cohorts provide essential information on the ability to recruit and retain volunteers over the several years that a phase III trial will last. Cohort studies also provide important information on variation of the HIV incidence rates due to intense counselling and education. HIV incidence may vary over time in a closed cohort, due to a saturation effect, or by the simple fact of participating in the study ('cohort effect'). A new proposed approach to estimate HIV incidence is utilizing the combination of two serological tests with different sensitivities to discriminate recent seroconverters with lower levels of antibodies ('detuned' assays) [25]. Detuned assay strategies may represent a convenient surrogate method for rapid identification of populations with high incidences of HIV. Detuned assay strategies, however, still need to be validated and standardized, especially in populations with non-subtype B infections. Cross-sectional studies of HIV prevalence have also been conducted in an attempt to identify populations that could be enrolled in phase III trials of HIV vaccines, accepting that high HIV prevalence is driven by high HIV incidence. These two types of simplified studies (detuned test and cross-sectional prevalence) have, however, the major disadvantage of not providing other important information for phase III trials, such as social behavioural characteristics of the population, or the ability to recruit and retain volunteers. In 1992, the World Health Organization assisted several developing countries (including Brazil, Thailand and Uganda) in the development of National AIDS Vaccine Plans, which included the establishment of preparatory cohorts for HIV vaccine trials [5,26]. The National Institutes of Health (NIH) of the United States has also supported cohort development in several sites. Four cohorts of homosexual and bisexual men were established in Brazil (in Belo Horizonte, Rio de Janeiro and Sao Paulo). In addition to fostering community understanding and support, these cohorts have also provided essential epidemiological information for the future conduct of phase III trials in Brazil. The cohort in Belo Horizonte, which enrolled 407 volunteers from 1994 to 1999, exhibited an annual HIV incidence of 1.75% [27]. Another Brazilian cohort of 753 high-risk male homosexuals in Rio de Janeiro had an overall annual HIV incidence of 3.1% and a follow-up rate of 88% after 18 months [28-30]. Different populations have been explored in Thailand as potential populations for phase III trials. A prospective cohort of injecting drug users (IDU) established in Bangkok in 1995 exhibited an overall HIV incidence rate of 5.8% [31,32]. HIV incidence has remained high in this cohort, which is the one from which volunteers were recruited for the phase III trial that started in 1999 in that country [6]. Persons attending sexually transmitted diseases clinics in Thailand were also assessed as a potential population for HIV vaccine trials, although incidence was lower than expected (1.4%), indicating that vaccine trials in this population would have to be larger than previously thought [33]. In view of the declining HIV incidence in traditional high-risk populations, large community-based studies are now being considered in Thailand [34]. Indeed, the notion of conducting vaccine trials in identified populations 'at risk' is relatively unique to HIV. Most pivotal vaccine efficacy studies have been conducted in large community-based trials. Some examples include polio (n > 600 000), Japanese encephalitis (n > 70 000) and hepatitis A (n > 40 000). HIV prevalence and incidence data have been recently published from several African countries, and the results show the different dynamics of the epidemic, with decreasing incidences in countries with mature epidemics (such as Uganda and Kenya) and increasing rates in countries with more recent epidemics (such as South Africa). A prospective cohort of female sex workers in Mombasa, Kenya documented a dramatic decline in the risk of HIV infection during participation in the study, with most seroconversions occurring during the first year of the study. In this particular study, HIV incidence declined 10-fold during the 3 years of follow-up, from 17.4 to 1.7% [35]. The authors warned that failure to anticipate variations in HIV incidence within high-risk populations could result in vaccine trials with an insufficient number of seroconversions to demonstrate protection [36]. A recent study conducted among women attending an antenatal clinic in the Gulu District of Northern Uganda also showed that HIV prevalence has declined from 26% in 1993 to 16.1% in 1997 [37]. A cohort of 2733 police officers in Dar es Salaam, which had a 13.8% HIV prevalence at the time of recruitment, demonstrated a crude annual incidence of 1.99%, with the authors suggesting that this could be a potential population group for HIV vaccine evaluation [38]. Another proposed potential population for HIV vaccine trials is young women in the Hlabisa health district in rural South Africa [39,40]. In this population, the estimated annual HIV incidence increased from 4% in 1992/1993 to 10% in 1996/1997. HIV prevalence among commercial sex workers in Addis Ababa, Ethiopia was extremely high (73.7%) [41], although this may not be an appropriate population for HIV vaccine trials due to the inherent difficulties to ensure appropriate follow-up of commercial sex workers. The presented studies exemplify the difficulties of identifying apparently appropriate populations for HIV vaccine trials, which in fact may no longer be appropriate by the time the candidate vaccine is ready for administration, or when the trial is actually approved by the national authorities. Although the initial HIV vaccine trials in developing countries were approved only after prolonged periods of evaluation and discussion, with the experience gained over the past few years, the approval process has now been considerable streamlined. HIV genetic subtypes and planning of efficacy trials The potential relevance of the genetic variability of HIV in terms of vaccine-induced protection is not known [42,43]. Research has already shown that HIV-1 genetic subtypes (or clades) do not strictly correspond to immunotypes. More than one genetic subtype could share common protective epitopes, and it is also possible that more than one immunotype is contained within a single genetic subtype [1]. The identification of vaccine-relevant immunotypes may also depend on what type of immune response is responsible for protection. In general, neutralizing antibodies seem to be more strain specific, whereas cell-mediated immune responses are more cross-reactive [11,19,44-46]. Information regarding the geographic distribution of different HIV genetic subtypes, and their immunological relevance, is extremely important to guide the design of new candidate vaccines and to plan clinical trials, especially phase III trials. To optimize chances for success, it would be strategically important that initial phase III trials are conducted with candidate vaccines that closely match strains prevalent in the trial population. However, at some point in the future, it would be necessary to carry out additional efficacy trials to explore the possibility of achieving cross-protection between different subtypes [1]. Alternative approaches to determining the degree of cross-clade protection would be: (i) to compare two candidate vaccines produced to correspond to two different HIV genotypes, and test these in a three-limbed trial against placebo; or (ii) to conduct the trial in a population where two HIV genotypes are highly prevalent, and power the study to determine efficacy against both. A significant effort has been made by different national and international HIV vaccine programmes to obtain information on the geographic distribution of the different genetic subtypes of HIV-1 [47,48]. As a result, we have a reasonably good understanding of the geographic distribution and dynamics of HIV-1 subtypes around the world, including that of the circulating recombinant forms [49]. The prevalent virus in the Americas, Western Europe, Australia and Japan remains subtype B, which is precisely the subtype on which most existing candidate vaccines are based. This situation, however, may change in the future, with new subtypes being introduced in different regions of the world. For instance, in addition to subtype B, subtype F viruses are also present in several South American countries, and subtype F is already the most prevalent subtype in Uruguay [50]. In addition to subtypes B and F, a B/F intersubtype recombinant form is widely circulating in Argentina [51,52]. Preliminary epidemiological analysis of the Argentinean samples revealed that the frequency of infection with subtype F viruses was significantly higher among heterosexual patients (71%) compared with homosexual patients (11%), in which subtype B is prevalent [51]. A similarly complex epidemic is evolving in Brazil. In one study from three potential HIV vaccine evaluation sites in that country, the majority of strains were subtype B (89.9%), followed by subtypes F (14.3%) and C (2.9%) [53]. In this study, four variants of subtype B with different tetrapeptides at the tip of the V3 loop were found, with a predominance of GWGR, which is a rare sequence in viruses from the United States. Although the possible immunological significance of the V3 sequence variability is not known, it was suggested that different strains may differ in the rate of progression to AIDS, which should be taken into account in the design of candidate vaccines and vaccine trials in Brazil [54]. In addition, it was reported that 7.6% of studied infections in Brazil were caused by recombinant viruses showing five distinct B/F mosaic patterns [55]. The epidemic in Thailand has been well documented, with subtype E causing the widespread heterosexual epidemic and subtype B being originally associated with the epidemic among IDU, but being replaced by subtype E infections. Of potential relevance for the selection of 'matched' candidate vaccines for testing in Thailand is the observation that the genetic diversity of the envelope gene among subtype E strains is growing with time, from 4.4% in 1990-1994 to 9.5% in 1996, representing the 'maturing' of the epidemic [56]. The epidemic in Africa is associated with multiple subtypes, with subtype C being the most prevalent in Southern Africa [57,58] and Ethiopia [59,60], subtypes A and D in East Africa [61-64], and an AG recombinant (IbNG) being most prevalent in Western Africa [65-67]. Although subtypes A and D represent the prevalent strains in East Africa, the spread of subtype C seems to be expanding over time. A recent study from Uganda found almost equal proportions of subtypes A and D (49 and 48%, respectively), with subtype C identified in 2.5% of the samples [61]. It is important to indicate, however, that although both subtypes A and D viruses are present in all regions of Uganda, their distribution is unequal among different villages [62]. Two other East African countries, Kenya and Tanzania, also have epidemics caused by subtypes A and D, as well as by subtype C viruses [63,64]. Vaccine strategies for East Africa may have to be re-evaluated to address the seemingly increasing incidence of subtype C infections, in addition to the initially established subtype A and D epidemics [68]. An extensive survey of HIV strains from different countries in West and West-Central Africa revealed that 20% of the samples had discordant subtypes between gag and env, with subtypes A and G predominantly involved in the recombination events, exemplified by the prototype AG-IbNG strain, now known as circulating recombinant form CRF02-AG [65-67]. As we move to Central Africa, the picture becomes more complex. Phylogenetic analysis of viruses from Cameroon showed a majority to be IbNG-AG-like, followed by subtypes D, G, F and H [69,70]. A large survey conducted in the Democratic Republic of Congo (formerly Zaire) revealed that all known HIV subtypes cocirculate, and that 6% of the samples could not be subtyped [71]. Subtype A was the predominant one (between 50 and 60%), with subtypes C, D, G and H having prevalence ranging from 7 to 9%, and subtypes F, J, K and CRF01-AE strains representing 2-4% of the samples. Only one subtype B strain was identified. Eighteen (29%) of 62 samples had discordant subtype designations between env and gag. The presence of different genetic subtypes of HIV in different regions of the world, which will certainly increase in the future, presents both a challenge and an opportunity. The challenge is to identify sites where the candidate vaccine can be properly matched to the circulating virus. The opportunity is based on the need to think globally, and to accept the fact that a future HIV vaccine will have to protect against all genetic subtypes of HIV. With carefully designed efficacy trials, sites with multiple subtypes may present interesting opportunities to obtain information on the breadth of protection of different HIV candidate vaccines and vaccine concepts. Recruitment and retention of volunteers As already indicated, cohort and other preparatory studies are necessary to assess the feasibility of recruiting and retaining volunteers, as well as to explore their willingness to participate in vaccine trials. In addition, these studies offer the opportunity of obtaining information on the effect on HIV incidence of non-vaccine risk-reduction interventions that, for ethical reasons, should be made available to all volunteers. Considerable experience has been accumulated in the United States through the work of the NIH-sponsored HIVNET Vaccine Preparedness Study Protocol Team. For epidemiological reasons, homosexual and bisexual men have been identified as a potential population for HIV vaccine trials in the United States. However, to obtain information on the efficacy of candidate vaccines against heterosexual transmission, a major effort is also been made to recruit women into trials in the United States [72,73]. In one study, the majority of homosexual men expressed willingness to participate in HIV vaccine trials, although younger men were less likely to answer questions about vaccine concepts, and were more likely to be lost to follow-up [74]. Their most common reason to participate in HIV vaccine trials was the desire to contribute to stop the epidemic and to help others [75]. Willingness to participate was found to decrease with time, and this may require a continuous educational effort at community and individual levels [76]. This effort must take into account the beliefs of vaccine recipients, their motivations for trial participation, and perceptions of potential side effects [77,78]. Since phase III trials will require the enrolment of thousands of volunteers, novel approaches for collecting information on risk behaviours are being developed, including audio, computer-assisted self-interview strategies [79]. Behavioural and social issues have been investigated among participants in a phase I/II trial in Thailand, where one-fifth of the volunteers reported having experienced overtly negative reactions from family or friends, although no problems with discrimination in employment, health care, or insurance did occur [80]. In preparation for a phase III trial in Thailand, willingness to participate was assessed among 193 IDU in Bangkok, with more than 50% reporting that they would definitely participate in the trial, and only 3% not willing to do it [81]. Altruism, regular HIV tests, and family support were found to be important motivations for participation. Information has also been recently published from three of the Brazilian cohorts of HIV-negative homosexual and bisexual men. In the Belo Horizonte cohort, 50% of participants reported that they would definitively participate in HIV vaccine trials, with 30% indicating that they might participate [27]. Almost 70% of volunteers from the cohort of high-risk homosexual men in Rio de Janeiro reported their willingness to participate, with altruism being the main reason [28,82]. Fear of vaccine-induced infection and HIV-positive serological tests were the main reasons for not wanting to participate [82]. Of relevance to this discussion are the results obtained from another cohort of homosexual/bisexual men in Rio de Janeiro, indicating that knowledge and awareness of risk do not easily translate into behavioural change, since significant proportions of the cohort continued practising unprotected sex [83]. In Uganda, knowledge and attitudes regarding vaccines in general, and HIV vaccines in particular, was assessed among 10 848 people participating in the Community Epidemiological Research project in Rakai [84]. Although the community had good knowledge of vaccination in general, 67% believed that HIV vaccines were already available, 41% knew that HIV vaccines were being tested, and only 13% that a vaccine would control HIV/AIDS. Ethical considerations The perspective of conducting multiple HIV vaccine trials in industrialized and developing countries has triggered an intense international debate on the ethical conduct of those trials [85-90]. To address some of those issues, the Joint United Nations Programme on HIV/AIDS (UNAIDS) embarked on a process of international consultation to define important ethical considerations, and to formulate guidance that might facilitate the ethical design and conduct of those trials [91]. As a result, in May 2000, UNAIDS issued a Guidance Document on 'Ethical considerations in HIV preventive vaccine research' [7]. In 18 Guidance Points, the document highlights some of the critical elements that must be considered when planning and implementing HIV vaccine trials. General recommendations include recognition of the ethical responsibility of the international community to promote HIV vaccine development, based on collaborative partnerships between the different players. This leads to capacity building in the host countries, where many of the trials will be conducted, and where future vaccines will be needed as a matter of urgency. The UNAIDS document makes a strong recommendation to involve communities, to ensure the ethical and scientific quality of the proposed research, and its relevance to and acceptance by the affected community [7,92]. Special attention is recommended to cross-cultural aspects, such as in the implementation of informed consent [93], and to engage the media to facilitate an informed discussion about scientific and social issues related to HIV vaccine evaluation [94]. The UNAIDS document also recommends early planning to make a future vaccine available to the community where trials are conducted, and to other populations in need of the vaccine [7,95]. Conclusions After many years of preparation, the first phase III trial of an HIV vaccine was initiated in the United States in 1998 using a bivalent BB gp120 candidate vaccine [6]. Despite some initial concerns regarding the difficulty of recruiting sufficient number of volunteers for HIV vaccine trials [96], the trial successfully recruited the required sample size of 5415 volunteers in 61 sites in North America and Europe [97,98]. An equivalent bivalent gp120 candidate, based on B and E strains, also entered phase III trial in Bangkok, in a population of 2500 IDU [99,100]. Interim efficacy analysis in the North American and Europe trial will be conducted in late 2001, and final efficacy analysis in late 2002. In the Thai trial, first efficacy analysis will be conducted in late 2002, and final analysis in late 2003. In the meantime, it is essential to increase efforts to develop and evaluate new candidate vaccines [1,101,102], which will require the preparation of multiple vaccine evaluation sites in industrialized and developing countries. New initiatives are being promoted by the US NIH, the US Walter Reed Army Institute of Research, the US Centers for Disease Control and Prevention, the International AIDS Vaccine Initiative, the French National Agency for Research on AIDS, the European Community, and the WHO-UNAIDS HIV Vaccine Initiative, among others. Examples of these new initiatives are: (i) the US NIH recently announced the selection of international sites for their HIV Vaccine Trial Network, in Brazil, China, Haiti, India, Peru, South Africa, Thailand and Trinidad; (ii) the International AIDS Vaccine Initiative is developing a number of partnerships to accelerate the development of HIV vaccines appropriate for developing countries; and (iii) the WHO-UNAIDS HIV Vaccine Initiative is assisting in the development of a collaborative 'African AIDS Vaccine Programme'. In parallel to this vaccine preparedness effort, it is also important to expand the pipeline of candidate vaccines for testing in human volunteers, including products based on different HIV subtypes. Ideally, a large number of HIV candidate vaccines should be tested in phase I/II trials, to accelerate the identification of the best product for subsequent phase III evaluation. During the 1990s, a major strategic approach was to prepare cohorts for future HIV vaccine efficacy trials. As already discussed, few of those cohorts were actually enrolled in trials because of the lack of appropriate candidate vaccines for testing (or the lack of decision to test available products). We also learned that HIV incidence may tend to decrease with time in those cohorts, due to the intense educational effort that is provided during the studies, although we do not know for how long that lower HIV incidence can be sustained. Alternative approaches should be explored, including the rapid identification and recruitment of populations with emerging epidemics, in which HIV incidence tends to be higher than in populations with mature epidemics. These populations may be located in 'non-traditional' sites, and they could be identified with the help of new laboratory techniques such as the 'detuned' assay. Another approach that merits additional consideration is the conduct of 'community-based' trials, enrolling several thousands of volunteers with lower HIV incidence (< 1%). This community-based approach, however, would require a significantly higher level of funding than that which has been available until now, not only to establish and follow the appropriate study populations, but also to manufacture the required larger number of doses of the candidate vaccine. After vaccine efficacy is demonstrated in a phase III trial, it would still be unknown whether HIV vaccines will be equally protective against infections acquired by different routes. At some point in the development process, efficacy should be determined in different at-risk populations. Finally, appropriate consideration should also be given to the epidemiological needs for the conduct of future effectiveness trials [103] that may be required after initial efficacy of HIV candidate vaccines is demonstrated in phase III trials. Those effectiveness trials may need to enrol hundreds of thousands of participants, and they might be necessary to provide information needed to develop appropriate vaccination strategies for public health use of future safe, effective and affordable HIV vaccines. Acknowledgements The authors thank Saladin Osmanov and Giovanni Rezza for helpful comments.