The effectiveness of HIV prevention and the epidemiological context

Nicholas C. Grassly,¹ Geoff P. Garnett,² Bernhard Schwartländer,³ Simon Gregson,⁴ & Roy M. Anderson⁵

Abstract Planning an intervention to prevent infections with the human immunodeficiency virus (HIV) should be guided by local epidemiological and socioeconomic conditions. The socioeconomic setting and existing public service capacity determine whether an intervention can have a significant outcome in terms of a reduction in a defined risk. The epidemiological context determines whether such risk reduction translates into a measurable impact on HIV incidence. Measurement of variables describing the epidemiological context can be used to determine the local suitability of interventions, thereby guiding planners and policy-makers in their choice of intervention. Such measurements also permit the retrospective analysis of the impact of interventions where HIV incidence was not recorded. The epidemiological context is defined for four different categories of intervention, shown to be effective in lower-income countries by randomized controlled trials. Appropriate indicators for the epidemiological context and methodological guidelines for their measurement are proposed. Their use in the transfer of a successful intervention from one context to another and in scaling up the effort to control HIV infection is explored. These indicators should provide a useful resource for those involved in planning HIV prevention interventions.

Keywords HIV infections/prevention and control/transmission; Disease transmission/prevention and control; Communicable disease control/methods; Epidemiologic factors; Epidemiologic measurements; Randomized controlled trials; Logistic models; Developing countries (source: MeSH).

Mots clés HIV, Infection/prévention et contrôle/transmission; Transmission maladie/prévention et contrôle; Lutte contre maladie contagieuse/méthodes; Facteurs épidémiologiques; Mesures épidémiologiques; Essai clinique randomisé; Modèle logistique; Pays en développement (source: INSERM).

Palabras clave Infecciones por VIH/prevención y control/transmisión; Transmisión de enfermedad/prevención y control; Control de enfermedades transmisibles/métodos; Factores epidemiológicos; Mediciones epidemiológicas; Ensayos controlados aleatorios; Modelos logísticos; Países en desarrollo (fuente: BIREME).

Introduction

Public health interventions to prevent new human immunodeficiency virus (HIV) infections can be targeted at different risk behaviours (e.g. heterosexual sexual intercourse, injecting drug use), different aspects of a given risk behaviour (e.g. heterosexual sexual intercourse: age at sexual debut, levels of pre- and extramarital sexual intercourse, condom use, etc.), and different at-risk populations (e.g. prostitutes, adolescents and school youth, young women). Interventions can range from education to promote behaviour change, to treatment of bacterial sexually transmitted infections (STIs) that enhance HIV transmission, to social marketing of condoms. The impact of such interventions depends very much upon the local context, in terms of risk behaviour, attitudes to risk, prevalence of cofactor STIs, stage of the HIV epidemic, existing health services, etc. Intervention design must take the local context into account.

Implementation of interventions

The implementation of an intervention to prevent the spread of HIV can be considered to have an immediate 'outcome' in terms of a change in risk behaviour or the riskiness of that behaviour, and a consequent `impact' on HIV incidence (Fig. 1). The likelihood of a significant outcome in this sense is determined by the socioeconomic, cultural and legislative context, and the existing public health service infrastructure. For example, an intervention that promotes access to educational material about safer sex may have different outcomes, in terms of changes in sexual risk behaviour, in different localities depending on community attitudes to sex. Whether or not the outcome of an intervention translates into an impact on HIV incidence depends on the epidemiological context. Although a significant programme outcome implies a relative reduction in HIV incidence, the relations between these two parameters are complex and non-linear (1, 2). Even an intervention with a large outcome may have minimal impact on HIV incidence if implemented late in the HIV epidemic or targeted at the wrong people (3). Dramatic increases in condom distribution may have very little impact on HIV spread until use during sexual intercourse is close to 100% in high-risk partnerships (4). Researchers and policy advisers must therefore be very careful in advocating interventions that have been shown to be successful using only measures of programme outcome and not impact (1).

Epidemiological context

The epidemiological context will depend on the intervention of interest, but can be broadly defined as the current state and trends in the behavioural and biological factors that determine the transmission dynamics of a given disease and the impact of a specified intervention. The appropriateness of an intervention to the local context can be assessed by the use of intervention trials, most rigorously by randomized controlled trials (RCTs). However, for the majority of HIV/AIDS programme managers this is not an option, and intervention selection typically proceeds by a process of analogy: those interventions that have proved effective in similar socioeconomic and epidemiological contexts are taken as models for the local intervention. This process often results in the selection of those interventions shown to be effective by the few published and well-publicized RCTs. When the same intervention carried out in different contexts has different impacts, confusion can arise over its appropriateness. For example, the reduction in HIV incidence resulting from an intervention offering improved syndromic treatment of cofactor STIs in the United Republic of Tanzania was not replicated by a community-wide STI treatment programme in Uganda (5, 6). Although these interventions implemented STI treatment in different ways, a significant cause of their different impacts is likely to have been the different contexts (7). In Uganda, the prevalence of one cofactor STI (herpes simplex virus type 2) was higher and the HIV/AIDS epidemic was at a later stage — a point at which the proportion of new HIV infections attributable to STI presence tends to be lower than earlier on, when core groups with high STI prevalence may be responsible for the more generalized spread of HIV.

This article describes the epidemiological contexts for different types of intervention aimed at preventing the spread of HIV in lower-income countries. Appropriate indicators are proposed, together with guidelines for their measurement. Ways of using these indicators to guide the choice of intervention and their role in the process of scaling up interventions to the national level are also considered. The relation between indicators recording the epidemiological context and national monitoring and evaluation indicators (8) is explored, and the complementary roles of national programme monitoring and evaluation, and local intervention planning are highlighted. The article begins with a review of selected interventions that have been shown to be effective at reducing HIV spread in lower-income countries by means of rigorous RCTs.

Randomized controlled trials

The "gold standard" for dealing with the influence of context is an assessment of specific interventions by means of an RCT prior to their widespread implementation (9). Randomization across individuals or communities is aimed at eliminating the influence that variation in contextual factors may have on apparent programme success or failure. In practice, RCTs, especially those undertaken in the community, have to be carried out within reasonably narrow geographical and political bounds and may thus fail to control for broader contextual issues. This restricts the general applicability of results and can give conflicting outcomes for different RCTs (7). In addition, there may be design problems, such as contamination and insufficient documentation of implementation or difficulties in replication (9).

RCTs of interventions to prevent heterosexual and perinatal HIV transmission that have shown a statistically significant outcome or impact are illustrated in Table 1 (10–24). This updates the 1997 World Bank study of selected interventions to prevent HIV infection in developing countries (10), and focuses on studies published in the last three years. A recent review of interventions in developing countries that targeted heterosexual transmission and injecting drug users, which also includes prospective cohort studies and repeat cross-sectional surveys, with and without comparison groups, has been carried out by Merson et al. (11).

RCTs have identified a substantial number of interventions that have potential for wider implementation and also scaling-up. The interventions fall into four categories, based on their outcome: sexual behaviour change; improved STI service use and quality; increased number of mothers receiving antiretroviral therapy to prevent mother-to-child transmission; and increased use of safe alternatives to breastfeeding. A number of other trials are currently in progress, e.g. information, education and counselling (IEC) with and without STI treatment in rural Uganda (25), replacement feeding of infants and antiretroviral therapy of mothers to prevent mother-to-child transmission in Côte d'Ivoire (26), and peer education and STI management in rural Zimbabwe (S. Gregson, personal communication, 2001). A reduction in the relative incidence of HIV infections is the clear aim of interventions, and can be considered the ultimate measure of intervention success (see Fig. 1). However, measures of success other than intervention impact are typically recorded, such as a decline in average numbers of extramarital sexual partners, more consistent condom use, or an increase in the average age at sexual debut. Of the RCTs of behavioural interventions listed in Table 1, four recorded changes in sexual behaviour, three recorded knowledge and attitudes, and one recorded the impact on STI incidence — none recorded the impact on HIV incidence. This pattern reflects that seen by Merson et al. (11) and the World Bank (10).

The epidemiological context

The impact on HIV incidence of an intervention with a significant outcome in a particular location depends on the epidemiological context. It is therefore essential to assess the epidemiological context in order to predict the likely impact of an intervention and determine whether it is potentially transferable to other locations. Proposed indicators describing the epidemiological context for the four previously identified intervention types are listed in Table 2 (27–35). Reasons for their inclusion, and how they determine the impact of an intervention on HIV incidence are described in the Annex (available at www.who.int/bulletin). The guidelines for their measurement are given, and this information is provided as a hypertext document with the relevant links (available at www.epidem.org/indicators.htm). Comments on the usefulness of the indicators and suggestions for improvement are welcomed and should be sent to the corresponding author.

If these indicators are measured in the appropriate way, the local suitability of an intervention can be assessed and its impact on HIV incidence estimated prior to implementation. Similarly, if an intervention has reduced a risk behaviour by a certain amount, but HIV incidence was not recorded, the impact of this intervention on the incidence can be estimated by a consideration of the relevant contextual indicators. For example, a national HIV/AIDS programme manager may be interested in determining whether a government education and condom distribution programme targeted at commercial sex workers has been successful in decreasing HIV incidence in the local population. The intervention has been evaluated in terms of the change in rates of condom use with clients and the average number of clients per day. The relevant epidemiological contextual indicators would therefore be the phase of the HIV epidemic, cofactor STI prevalence in the target and general population, sexual behaviour of the general population, and patterns of mixing of sex workers with the general population. It may emerge that, although HIV infection is reasonably prevalent throughout the population, cofactor STIs are concentrated among sex workers and their male clients, while self-reported sexual behaviour in the general population indicates a low intrinsic risk. This would suggest that many new HIV infections arise because of the high rate of transmission from sex workers to their clients (enhanced by cofactor STIs), and subsequently to the spouses and other sexual partners of the clients. Thus the intervention targeting commercial sex workers is likely to have been effective in reducing overall HIV incidence, and its continued implementation is therefore of key importance. Further, it may become clear that clients of sex workers disproportionately represent a particular section of society, such as migrant labourers, and that an intervention targeting these individuals would also be effective.

If the programme manager was subsequently to consider implementing this intervention elsewhere, the same contextual indicators should be recorded in the new location. If the epidemiological context is favourable, implementation of the intervention can follow. For example, the epidemic may be at an early stage in the new location, and so most new HIV infections in the general population will be due to contact with sex workers who are likely to be HIV-positive. An intervention targeting sex workers should therefore be highly effective in reducing overall HIV incidence.

Measurement of the components of the epidemiological context allows the local suitability of an intervention with a predicted outcome in terms of risk reduction to be determined. Once we have identified a set of interventions with a suitable outcome, we need to assess whether and how they can be implemented to achieve this outcome. This is dependent on broader socioeconomic, cultural and legislative factors, such as determinants of an individual's ability to control his or her own HIV risk, existing maternal and child services, availability of voluntary counselling and testing facilities, vulnerability of those individuals revealing their HIV status, community attitudes to sexual risk behaviour, adherence to antiretroviral drug therapy, alcohol consumption, and logistic problems with programme implementation (31, 36). A discussion of these factors is beyond the scope of this paper (8, 36).

An unfavourable epidemiological context suggesting a low impact for a specific intervention does not necessarily rule out implementation of that intervention, particularly if the broader socioeconomic context is such that a large intervention outcome can be achieved at low cost. Rather, the unfavourable components of the epidemiological context can themselves be targets of another intervention. For example, high STI prevalence among the clients of commercial sex workers will make improved STI treatment of the sex workers alone ineffective in reducing overall HIV incidence. By targeting another STI treatment intervention at the clients, for example by offering workplace STI treatment and counselling, the epidemiological context for the original intervention may become favourable.

Broader determinants of socioeconomic, cultural and legislative contexts can also be the target of interventions. Such interventions are typically termed "structural" or "environmental", the former resulting in changes in laws, policies or administrative procedures, and the latter in living conditions, resources or opportunities (11, 37).

The use of epidemiological models

By assessing and recording indicators for the epidemiological context, programme managers and policy-makers will be able to identify those interventions likely to be effective in their local setting, subject to the constraints of implementation discussed above. The value of this process can be increased through the use of mathematical models that predict the impact of interventions on the number of new HIV infections, while including, explicitly and quantitatively, potentially important contextual variables. Models are particularly useful when determining the impact of an intervention that results not only from a decline in susceptibility of targeted individuals, but also from the decrease in transmission of HIV to other individuals.

Many mathematical models of interventions to prevent the spread of HIV and results obtained by applying them to particular scenarios have been reported in the literature (4, 38–42). Although mathematically complex, such models have recently been implemented as user-friendly software designed to assist policy-makers and national HIV/AIDS programme managers. Two such applications were demonstrated at the XIII International AIDS Conference in Durban, South Africa, 2000. The first, a suite of programs entitled HIVTools, has been designed to assess interventions targeted at different populations including sex workers, school youth, and injecting drug users (43). These programs include costing spreadsheets for the interventions modelled, and were recently used to assess a harm-reduction programme aimed at injecting drug users in Svetlogorsk, Belarus (44). The second application, AIDS Strategic Intervention Support Tool (ASIST), developed by the US National Institute of Mental Health, models interventions aimed at reducing heterosexual transmission of HIV. Interventions that can be modelled include condom distribution, education for sexual behaviour change, bacterial STI treatment, microbicidal gel distribution, and antiretroviral therapy.

Both of these programs are able to deal explicitly with the epidemiological context as described by the indicators listed in Table 2 through entry of the relevant parameters, and can predict the effectiveness of interventions under different scenarios. In ASIST, a simple "wizard" guides users through the process of setting parameters consistent with a broadly defined scenario. Two ways of using such programs are illustrated in Box 1.

Scaling up to the national level

To what extent can and should interventions that are successful at a local level be implemented at the national level? Can an intervention shown to be effective by an RCT in a different specific epidemiological context be taken as the model for a component of the national HIV/AIDS programme? These questions concern the process of scaling up.

The effectiveness at the national level of an intervention targeted at a specific risk group is determined in part by the epidemiological context and how it varies across the country, for example, the mixing of the target population of the intervention (e.g. miners or truck drivers) with at-risk groups throughout the country. However, there are issues that are specifically relevant to the national level, such as the heterogeneous distribution of at-risk populations across the districts of a country, in terms of both type of risk and concentration. The most obvious distinctions in this case are rural/urban differences in HIV prevalence and risk behaviours. Such heterogeneity has implications both for HIV/AIDS surveillance (29), and for the impact of different types of intervention. For example, an intervention targeted at commercial sex workers may be more effective in urban areas, where sex work is a more clearly defined activity, and numbers of sex workers and their accessibility are often higher. The epidemiological context may also vary from one urban setting to another or from one rural area to another within the same country. For example, the relative roles of injecting drug use and commercial sex in the spread of HIV in India, and the interaction of these at-risk populations vary considerably from state to state (National AIDS Control Organisation of India, unpublished data).

A national HIV/AIDS programme should therefore allow regional approaches to the HIV/AIDS epidemic, and this naturally implies a certain amount of decentralization of the organization and financing of prevention activities. The use of flexible financing structures is crucial to such decentralization (45), permitting funds earmarked for HIV/AIDS prevention to be used for interventions reflecting local culturally adapted best practice (46). Local relevance can be established by examining the epidemiological context as well as broader structural and environmental factors, and by carrying out costing studies. Capacity to carry out analyses of costs and effectiveness may be centrally located or based within external organizations and accessible through technical assistance programmes. Alternatively, financing and aid instruments may be designed in such a way as to promote local capacity in carrying out such analyses. Organizational and institutional capacity-building for HIV/AIDS prevention activities can be made the explicit goal of financing arrangements. Indicators for the dimensions of this capacity have been designed and their performance examined (47).

Scaling up an intervention may produce economies of scale, not only in terms of costs, but also in terms of effectiveness. For example, in Rakai, Uganda, one of the problems with mass STI treatment was the rapidity of reinfection due to migration and contact with untreated infected individuals (7). An intervention that aims at improved STI treatment and is implemented at the national level would reduce this rate of reinfection. In other words, the intervention has a "positive externality". Scaling up those interventions that have positive externalities may result in increasing returns in terms of effectiveness. The logical conclusion of such interventions could be disease eradication, which is the rationale behind many accelerated health programmes (such as immunization against poliomyelitis). However, HIV eradication is not realistic given the maintenance of infection in difficult-to-reach groups such as commercial sex workers and injecting drug users, its long duration of asymptomatic infection, and the lack of an effective vaccine.

As an intervention programme is scaled up to the regional and national level, questions naturally arise about the possibility of targeting an expanded range of risk groups. For example, an STI and HIV/AIDS education programme targeted at commercial sex workers may be expanded to include all adult attendees of STI clinics. This concerns intervention design rather than scaling up. Prior to implementation, the relevant epidemiological context and socioeconomic environment will need to be reassessed with respect to this new intervention with a different target population.

Although consideration of the epidemiological context may give some indication as to the likely effectiveness of a national intervention package, pilot programmes remain of fundamental importance. Pilot programmes differ from RCTs in that they tend not to include a matched control group. Although they are useful in clarifying the relative importance of different components of the epidemiological context, their key role is in identifying broader socioeconomic and financial constraints to effective programme implementation. The monitoring and evaluation of pilot programmes should therefore include indicators of programme outcome, costs and financial sustainability, in addition to those for the epidemiological context. For example, a list of indicators for the monitoring and evaluation of a pilot programme for the integrated prevention of mother-to-child transmission should include the epidemiological context, with indicators such as HIV prevalence in pregnant women, but also effectiveness indicators (e.g. receipt of antiretroviral drugs) and cost indicators (e.g. cost of drugs) (31).

Establishing the effectiveness of national AIDS control programmes

The importance of different levels of indicators, corresponding to programme input, output, outcome and impact, for the monitoring and evaluation of national HIV/AIDS programmes has recently been clarified (8, 48). This framework has been expanded to make clear the role of proximate determinants of HIV spread, which are the immediate biological and behavioural determinants of HIV spread, in other words the components of the basic reproductive rate, R₀: the efficiency of transmission, the rate of exposure of a susceptible to an infectious person, and the duration of infectiousness (48–50). It is through these proximate determinants that the impact of the broader socioeconomic context and the output of interventions on HIV incidence are seen.

Some of the indicators for these proximate determinants correspond to a subset of indicators for the epidemiological context for different interventions, as we have defined in this paper, while others correspond to those for the intervention outcome. For example, recommended indicators for monitoring the sexual behaviour of youth include sexual mixing by age, and the proportion of the population with multiple sex partners (8). The first of these can be compared with indicator 3 for the epidemiological context of interventions to promote sexual behaviour change and/or condom distribution (see Table 2). The second is a component of the variables recording the outcome of this intervention.

With the completion of recommendations for second-generation surveillance in 2000, some of these monitoring and evaluation indicators have been incorporated into the latest country-specific HIV fact sheets prepared by WHO/UNAIDS. They include indicators for proximate determinants, such as prevalence of curable STIs and self-reported sexual behaviour (including age at first sexual encounter and non-regular sexual partnerships), and context and outcome indicators, such as the rate of STI treatment and counselling, general health service indicators for accessibility and blood safety, condom availability, and knowledge of methods of HIV protection.

At the national level, indicators recording proximate determinants are necessarily restricted, and therefore lack the detail that would permit the local relevance of interventions to be assessed. Indeed, this is not their role (8). Monitoring and evaluation of national HIV/AIDS programmes serve a different function to that of planning future interventions. However, the two processes are complementary. National monitoring and evaluation, together with appropriate costing studies, can guide the choice of interventions that should be assessed with respect to the epidemiological context at the local level. If the epidemiological context is favourable, consideration of the broader socioeconomic context and the use of pilot studies can lead to the appropriate choice of intervention. Once implemented, this intervention becomes part of the national monitoring and evaluation effort.

Acknowledgements

We thank Hélène Carabin for detailed comments and criticism. Thanks are also due to UNAIDS, the Royal Society, London and the Wellcome Trust for their financial support to individual authors.

Conflicts of interest: none declared.

Résumé

Efficacité de la prévention du VIH et contexte épidémiologique

La planification d'une intervention destinée à la prévention des infections par le virus de l'immunodéficience humaine (VIH) doit être guidée par les conditions épidémiologiques et socio-économiques locales. Le contexte socio-économique et la capacité des services publics existants détermineront si une intervention peut avoir un impact significatif en termes de réduction d'un risque déterminé. Le contexte épidémioque déterminera si une telle réduction du risque est susceptible d'avoir un impact mesurable sur l'incidence du VIH. La mesure des paramètres décrivant le contexte épidémiologique peut être utilisée pour déterminer si une intervention peut convenir au niveau local, et guidera les planificateurs et les décideurs dans leur choix. Ces mesures permettront également de procéder à l'analyse rétrospective de l'impact des interventions là où l'incidence du VIH n'a pas été enregistrée. Le contexte épidémiologique a été défini pour quatre catégories d'interventions dont l'efficacité dans les pays à faible revenu a été démontrée lors d'essais contrôlés randomisés. Des indicateurs appropriés pour le contexte épidémiologique sont proposés avec des directives méthodologiques pour leur mesure. Leur application au transfert d'une intervention réussie d'un contexte à un autre et au passage à une plus grande échelle de lutte contre l'infection à VIH est explorée. Ces indicateurs devraient constituer un outil précieux pour les responsables des choix politiques.

Resumen

Influencia del contexto epidemiológico en la eficacia de la prevención de la infección por el VIH

La planificación de una intervención de prevención de las infecciones por el virus de la inmunodeficiencia humana (VIH) debería hacerse en función de las condiciones epidemiológicas y socioeconómicas locales. La situación socioeconómica y la capacidad de la administración pública determinan si una intervención puede tener un efecto importante de reducción de un riesgo definido, y el contexto epidemiológico determina si esa reducción del riesgo se traducirá en un impacto cuantificable en la incidencia de infección por el virus. Es posible medir las variables que caracterizan el contexto epidemiológico para establecer la idoneidad de las intervenciones a nivel local y orientar la elección de las intervenciones por parte de los planificadores y los formuladores de políticas. Esas mediciones también permiten analizar retrospectivamente el impacto de las intervenciones en los casos en que no se registró la incidencia de infección por el VIH. Se define el contexto epidemiológico para cuatro categorías de intervención cuya eficacia en los países de bajos ingresos se ha demostrado mediante ensayos controlados aleatorizados. Se proponen indicadores apropiados para el contexto epidemiológico, así como pautas metodológicas para su medición, y se analiza el empleo de esos instrumentos para transferir una intervención exitosa de un contexto a otro y ampliar las actividades de lucha contra la infección por el VIH. Estos indicadores deberían constituir un valioso recurso para los formuladores de políticas.

References

1. Fishbein M, Pequegnat W. Evaluating AIDS prevention interventions using behavioural and biological outcome measures. Sexually Transmitted Diseases, 2000, 27: 101–110.        

2. Schwartländer B et al. AIDS special report — AIDS in a new millennium. Science, 2000, 289: 64–67.        

3. Wasserheit JN, Aral SO. The dynamic topology of sexually transmitted disease epidemics: implications for prevention strategies. Journal of Infectious Diseases, 1996, 174 (suppl. 2): S201–S213.        

4. Garnett GP, Anderson RM. Strategies for limiting the spread of HIV in developing countries: conclusions based on studies of the transmission dynamics of the virus. Journal of AIDS, 1995, 9: 500–513.        

5. Grosskurth H et al. Impact of improved treatment of sexually transmitted diseases on HIV infection in rural Tanzania: randomised control trial. Lancet, 1995, 346: 530–536.        

6. Wawer MJ et al. Control of sexually transmitted diseases for AIDS prevention in Uganda: a randomised community trial. Lancet, 1999, 353: 525–535.        

7. Grosskurth H et al. Control of sexually transmitted disease for HIV- 1 prevention: understanding the implications of the Mwanza and Rakai trials. Lancet, 2000, 355: 1981–1987.        

8. National AIDS programmes: a guide to monitoring and evaluation. Geneva, Joint United Nations Programme on HIV/AIDS, 2000 (UNAIDS/00.17E, Internet communication, 24 November 2000 at http://www.unaids.org/publications/documents/epidemiology/surveillance/JC427-Mon&Ev-Full-E.pdf).        

9. Hayes R. Design of human immunodeficiency virus intervention trials in developing countries. Journal of the Royal Statistical Society, Series A, 1998, 161: 251–263.        

10. Confronting AIDS: public priorities in a global epidemic. Oxford, Oxford University Press, 1997 (World Bank Policy Research Report).         

11. Merson MH, Dayton JM, O'Reilly K. Effectiveness of HIV prevention interventions in developing countries. AIDS, 2000, 14 (suppl. 2): S68–S84.        

12. Stanton BF et al. Increased protected sex and abstinence among Namibian youth following a HIV risk-reduction intervention: a randomised, longitudinal study. AIDS, 1998, 12: 2473–2480.        

13. Fawole IO et al. A school-based AIDS education programme for secondary school students in Nigeria: a review of effectiveness. Health education research, 1999, 14: 675–683.        

14. Egger M et al. Promotion of condom use in a high-risk setting in Nicaragua: a randomised controlled trial. Lancet, 2000, 355: 2101–2105.        

15. Harvey B, Stuart J, Swan T. Evaluation of a drama-in-education programme to increase AIDS awareness in South African high schools: a randomised community intervention trial. International Journal of STD and AIDS, 2000, 11: 105–111.        

16. The Voluntary HIV-1 Counselling and Testing Efficacy Study Group. Efficacy of voluntary HIV-1 counselling and testing in individuals and couples in Kenya, Tanzania and Trinidad: a randomised trial. Lancet, 2000, 356: 103–112.        

17. Coplan P et al. A randomised controlled trial of an intervention to improve treatment of STDs among Nigerian youth. Paper presented at: XIII International AIDS Conference, Durban, South Africa, 9–14 July 2000. (Abstract WePeD4723)         

18. Harrison A et al. Syndrome packets and health worker training improve sexually transmitted disease case management in rural South Africa: randomized controlled trial. AIDS, 2000, 14: 2769–2779.        

19. Wiktor SZ et al. Short-course oral zidovudine for prevention of mother-to-child transmission of HIV-1 in Abidjan, Côte d'Ivoire: a randomized trial. Lancet, 1999, 353: 781–785.        

20. Dabis F et al. 6-month efficacy, tolerance, and acceptability of a short regimen of oral zidovudine to reduce vertical transmission of HIV in breastfed children in Côte d'Ivoire and Burkina Faso: a double-blind placebo-controlled multicentre trial. Lancet, 1999, 353: 786–792.        

21. DITRAME ANRS 049 Study Group. 15-month efficacy of maternal oral zidovudine to decrease vertical transmission of HIV- 1 in breastfed African children. Lancet, 1999, 354: 2050–2051.        

22. Saba J. Interim analysis of early efficacy of three short ZDV/3TC combination regimens to prevent mother-to-child transmission of HIV-1: the PETRA trial. Paper presented at: Sixth Conference on Retroviruses and Opportunistic Infections, Chicago, 31 January–4 February 1999 (Abstract S6).        

23. Guay LA et al. Intrapartum and neonatal single-dose nevirapine compared with zidovudine for prevention of mother-to-child transmission of HIV-1 in Kampala, Uganda: HIVNET 012 randomised trial. Lancet, 1999, 354: 795–802.        

24. Nduati R et al. Effect of breastfeeding and formula feeding on transmission of HIV-1: A randomised clinical trial. JAMA, 2000, 283: 1167–1174.        

25. Kamali A et al. Seven-year trends in HIV-1 infection rates, and changes in sexual behaviour, among adults in rural Uganda. AIDS, 2000, 14: 427–434.        

26. Chorba TL et al. Implementation and effectiveness of a short- course zidovudine and replacement feeding program to prevent mother-to-child transmission of HIV-1 in Abidjan, Côte d'Ivoire. Paper presented at: XIII International AIDS Conference, Durban, South Africa, 9–14 July 2000 (Abstract ThPpC1446)         

27. Family Health International. BSS: Guidelines for repeated behavioural surveys in populations at risk of HIV. Arlington, VA, Family Health International, 2000.        

28. Looking deeper into the HIV epidemic: a questionnaire for tracing sexual networks Geneva, Joint United Nations Programme on HIV/AIDS, 1998 (UNAIDS/98.27, Internet communication, 24 November 2000 at http://www.unaids.org/publications/documents/epidemiology/surveillance/una98e27.pdf).        

29. Second generation surveillance for HIV: the next decade. Geneva, World Health Organization/Joint United Nations Programme on HIV/AIDS, 2000 (UNAIDS/00.03E, Internet communication, 24 November 2000 at http://www.unaids.org/publications/documents/epidemiology/surveillance/2^nd_GENERATION_ FINAL.PDF).        

30. Guidelines for sexually transmitted infections surveillance. Geneva, World Health Organization/Joint United Nations Programme on HIV/AIDS Working Group on Global HIV/AIDS/STI Surveillance, 1999 (UNAIDS/99.33E, Internet communication, 24 November 2000 at http://www.unaids.org/publications/documents/impact/std/jc240%2Dsextransminfsurv%2De.pdf).        

31. Local monitoring and evaluation of the integrated prevention of mother to child HIV transmission in low- income countries. Geneva, UNICEF/UNAIDS/WHO, 2000 (draft 03/2000, Internet communication, 24 November 2000 at http://www.unaids.org/publications/documents/mtct/ME2000.doc).        

32. Vertical transmission of HIV – rapid assessment guide. Geneva, United Nations Children's Fund/Joint United Nations Programme on HIV/AIDS UNAIDS, 1998 (Internet communication, 24 November 2000 at http://www.unaids.org/publications/documents/mtct/sitanag.pdf).        

33. Sherman GG et al. CD4(+) T cell enumeration in HIV infection with limited resources. Journal of Immunological Methods, 1999, 222: 209–217.        

34. Semba RD et al. Human immunodeficiency virus load in breast milk, mastitis, and mother-to-child transmission of human immunodeficiency virus type 1. Journal of Infectious Diseases, 1999, 180: 93–98.        

35. Indicators for assessing vitamin A deficiency and their application in monitoring and evaluating intervention programmes. Geneva, World Health Organization, 1996 (unpublished document WHO/NUT/96.10).        

36. Sumartojo E. Structural factors in HIV prevention: concepts, examples, and implications for research. AIDS, 2000, 14 (suppl. 1): S3–S10.        

37. Sweat MD, Denison JA. Reducing HIV incidence in developing countries with structural and environmental interventions. AIDS, 1995, 9 (suppl. A): S251–S257.        

38. Korenromp EL et al. Model-based evaluation of single-round mass treatment of sexually transmitted diseases for HIV control in a rural African population. AIDS, 2000, 14: 573–593.        

39. Rehle TM et al. AVERT: a user-friendly model to estimate the impact of HIV/sexually transmitted disease prevention interventions on HIV transmission. AIDS, 1998, 12 (suppl. 2): S27–S35.        

40. Robinson NJ et al. Proportion of HIV infections attributable to other sexually transmitted diseases in a rural Ugandan population: simulation model estimates International Journal of Epidemiology, 1997, 26: 180–189.

41. Rowley JT, Anderson RM. Modeling the impact and cost- effectiveness of HIV prevention efforts. AIDS, 1994, 8: 539– 548.        

42. Wood E et al. Extent to which low-level use of antiretroviral treatment could curb the AIDS epidemic in sub-Saharan Africa. Lancet, 2000, 355: 2095–2100.        

43. HIVTools — a cost-effectiveness toolkit. Geneva, Joint United Nations Programme on HIV/AIDS/Health Economics and Financing Programme, London School of Hygiene and Tropical Medicine, 2000 (Internet communication, 24 November 2000 at http://www.unaids.org/publications/documents/economics/index.html#cost_model).        

44. Walker D et al. Responding to HIV/AIDS in Belarus: the importance of early intervention. Eurohealth, 2000, 6: 53–55.        

45. Walt G et al. Health sector development: from aid coordination to resource management. Health Policy and Planning, 1999, 14: 207–218.        

46. Binswanger HP. Scaling up HIV/AIDS programs to national coverage. Science, 2000, 288: 2173–2176.        

47, Kotellos KA et al. Field experiences: measuring capacity building efforts in HIV/AIDS prevention programmes. AIDS, 1998, 12 (suppl. 2): S109–S117.        

48. Boerma T et al. A framework for the evaluation of national AIDS programmes. Chapel Hill, NC, MEASURE, 1999 (Evaluation Working Papers WP-99-17; Internet communication, 1 Novem- ber 2000 at http://www.cpc.unc.edu/measure).        

49. Anderson RM, May RM. Infectious diseases of humans: dynamics and control. Oxford, Oxford University Press, 1991.        

50. Gregson S et al. HIV and fertility change in rural Zimbabwe. Health Transition Review, 1997, 7 (suppl. 2): 89–112.        

¹ Coordinator, UNAIDS Epidemiology Reference Group, Department of Infectious Disease Epidemiology, Imperial College School of Medicine, St Mary's Campus, Norfolk Place, London W2 1PG, England (email: n.grassly@ic.ac.uk). Correspondence should be addressed to this author.

² Reader, Department of Infectious Disease Epidemiology, Imperial College School of Medicine, London, England.

³ Chief, Strategic Information Division, Joint United Nations Programme on HIV/AIDS, Geneva, Switzerland.

⁴ Wellcome Trust Research Fellow, Department of Infectious Disease Epidemiology, Imperial College School of Medicine, London, England.

⁵ Head, Department of Infectious Disease Epidemiology, Imperial College School of Medicine, London, England.

Ref. No. 01-1172

Annex. Reasons for inclusion of the indicators of epidemiological context

Education to alter sexual behaviour/condom distribution

1. Phase of the HIV/AIDS epidemic

The phase of the HIV/AIDS epidemic, as recorded by age- and sex-stratified HIV prevalence for different risk groups will determine the possible impact of a targeted intervention. Early on in a heterosexual HIV/AIDS epidemic, HIV infections tend to be concentrated in individuals with high rates of partner change, before spreading from such core groups to the general population. In other words, a generalized epidemic in which HIV prevalence exceeds 1% in the general population is usually preceded by a concentrated epidemic with >5% of core groups but 1).

2. Cofactor STI prevalence

There is evidence that STIs such as syphilis, gonorrhoea, chlamydial infections, trichomoniasis, and genital herpes (HSV-2) act as cofactors, enhancing the probability of sexual transmission of HIV (2, 3). Likewise HIV-1 may act as a cofactor for the spread of HSV-2 by increasing viral shedding (4). In areas with high prevalence of untreated STIs, HIV prevalence is also often high. An education programme to alter sexual behaviour and promote condom use will result in a decline in both HIV and other STIs, each decline acting synergistically with the other. Thus in populations with high STI prevalence, the impact of such programmes on HIV incidence may be greater than in populations with low STI prevalence.

3. Mixing of target population with other at- risk populations or age-groups

If there is extensive mixing of the target population with other at-risk groups, the intervention can have a large impact on HIV incidence in the total population. This is the case if the target population corresponds to a spread network (a cause for a substantial number of new infections in other populations), with a large effective reproductive rate, R_e (2, 5). Conversely, if HIV incidence within the target population is mainly due to sexual contacts outside that risk group, and the intrinsic R_e is less than 1, an intervention targeting this group may have limited impact. Assessments of the mixing patterns of the target population through social and geographical mapping, and sexual network questionnaires, and of baseline sexual behaviour (which determines R_e) allow the likely impact of an intervention on overall HIV incidence to be determined.

4. Sexual behaviour of populations not targeted by the intervention

The risk behaviour of untargeted populations determines the relative contribution of the targeted and untargeted populations to incident HIV infections. For example, if the target population is acting as a spread network, then assessment of sexual behaviour in untargeted populations can give an indication as to the likely impact on overall HIV incidence of a decline in this source of infection. If men who visit prostitutes are not using condoms with their regular sex partners (e.g. spouses), an intervention targeting prostitutes will have a bigger impact on overall HIV incidence than if these men use condoms all the time.

It may also be important to measure sexual behaviour in populations not targeted by the intervention if it is thought that the intervention may significantly affect the balance of supply and demand for sexual services. Thus an intervention that enables brothel-based female prostitutes to enforce condom use may reduce demand for brothel-based commercial sex and result in increased demand for commercial sex from non-brothel-based prostitutes — a group for which HIV prevention activities may not be in place.

Improved STI treatment

1. Phase of particular STI epidemic

As with HIV infection, interventions aimed at other STIs will be more or less appropriate at different phases of the STI epidemic (1).

2. Phase of HIV/AIDS epidemic

As for behavioural interventions described above.

3. Prevalence of other cofactor STIs including genital herpes (HSV-2)

Interventions to reduce HIV incidence by treating bacterial STIs, such as gonorrhoea, may be less effective if the prevalence of other untreated STIs is high. For instance, a high prevalence of viral STIs, such as HSV-2, may mean that the probability of HIV transmission remains high despite treatment of bacterial STIs (6). The prevalence of other STIs should therefore be measured and, if appropriate, interventions to prevent and treat them should be implemented. The epidemiological context for one intervention may be the target of another. In this way interventions can act in synergy.

4. Mixing of target population with other at- risk populations and age-groups

As for behavioural interventions described above.

5. Prevalence of STIs in neighbouring populations and importance of reintroduction following treatment

If there is extensive mixing with other risk groups or sexual networks, as determined by indicator 4, then a high prevalence of STIs in these risk groups may result in rapid re-introduction of the STIs to the targeted population following treatment. This may make periodic mass treatment less effective than continuous symptomatic treatment in reducing STI prevalence and therefore HIV incidence.

Antiretroviral drugs to prevent mother-to-child transmission

1. Levels of breastfeeding compared to artificial feeding

The impact of antiretroviral (ARV) therapy on mother-to-child transmission (MTCT) will depend on levels of breastfeeding in the population. It is estimated that breastfeeding contributes between one-third and half of all cases of MTCT in breastfeeding populations (7, 8). In a breastfeeding population, therefore, ARV therapy that is only provided perinatally may be less effective in preventing MTCT than in a non-breastfeeding population, owing to this additional risk of vertical transmission. In recording modes of feeding it is important to distinguish mixed feeding from exclusive breastfeeding, since there is some evidence that the former may entail a higher risk of transmission than exclusive breastfeeding (9).

2. Prevalence of HIV infection in treated women

In countries with a very high prevalence of HIV infection, strategies to provide cheap one- or two- dose regimens of ARV drugs to pregnant women may be cost-effective. In these cases the prevalence of HIV infection in the women to be treated will be of key importance in determining the impact of the intervention on the incidence of perinatal cases of HIV infection.

3. Viral load of mothers

Viral load is a direct determinant of the probability of MTCT in the perinatal period and during breastfeeding (10, 11). If the average viral load of pregnant women varies between populations owing to the stage of the HIV/AIDS epidemic, or if genital shedding of the virus varies because of differences in nutritional status (12), ARV therapy may be more effective in those populations with a higher mean maternal viral load. Measurement of the viral load or, as a proxy, CD4 levels in a sample of mothers may therefore give an indication of programme effectiveness. However, it is unclear whether there is likely to be significant variation between populations in terms of maternal viral load, in particular as prolonged HIV infection results in infertility (13).

4. Mode of delivery and rates of premature delivery

Perinatal transmission of HIV is on average more probable during vaginal delivery than Caesarean section , and is enhanced in pre-term delivery (14). ARV therapy to prevent MTCT may therefore be more effective in populations with low rates of Caesarean section and high rates of premature delivery — features of lower-income countries.

5. STI prevalence

The presence of certain STIs during labour may promote MTCT through increased genital shedding of HIV (12, 15). ARV therapy in populations with a high STI prevalence may therefore be more effective. Currently it is likely that STI treatment facilities will be in place prior to the existence of health services capable of supplying ARV drugs. However, this may change with the advent of cheap, one- or two-dose drug regimens, e.g. with nevirapine, to prevent MTCT (16).

Alternatives to breastfeeding

1. Viral load of mothers; availability of ARV drugs

The mother's viral load, determined by the natural history of the disease and the availability of ARV therapy, is a determinant of postnatal transmission (11). Thus interventions promoting alternatives to breastfeeding may be more effective in populations with high viral loads. As noted previously, the relevance of this indicator for the epidemiological context is not clear, since viral load may not vary at the population level for pregnant women.

2. Levels of mastitis and severe vitamin A deficiency

Mastitis and severe vitamin A deficiency may increase the probability of HIV transmission during breastfeeding (17, 18). If these factors vary significantly between populations, they will be important indicators for the epidemiological context of artificial feeding interventions.

3. Mode of delivery and rates of premature delivery; and STI prevalence

Mode of delivery, rates of premature delivery and prevalence of STIs have an impact on HIV transmission during labour. They therefore determine rates of intrapartum compared to postpartum HIV transmission, and hence efficacy of measures aimed at reducing postpartum transmission (the mother's viral load may also affect this balance).

References

1. Wasserheit JN, Aral SO. The dynamic topology of sexually transmitted disease epidemics: implications for prevention strategies. Journal of Infectious Diseases, 1996, 174 (suppl. 2): S201–S213.

2. Fleming DT, Wasserheit JN. From epidemiological synergy to public health policy and practice: the contribution of other sexually transmitted diseases to sexual transmission of HIV infection. Sexually Transmitted Infections, 1999, 75: 3–17.

3. Rottingen J-A, Cameron DW, Garnett GP. A systematic review of the epidemiological interactions between classical STDs and HIV: How much is really known? Sexually Transmitted Diseases, 2001, 28: 579–597.

4. Augenbraun M et al. Increased genital shedding of herpes simplex virus type-2 in HIV seropositive women. Annals of Internal Medicine, 1999, 123: 845–847.

5. Morris M et al. Bridge populations in the spread of HIV/AIDS in Thailand. AIDS, 1996, 10: 1265–1271.

6. Grosskurth H et al. Control of sexually transmitted disease for HIV- 1 prevention: understanding the implications of the Mwanza and Rakai trials. Lancet, 2000, 355: 1981–1987.

7. De Cock KM et al. Prevention of mother-to-child HIV trans-mission in resource-poor countries: translating research into policy and practice. JAMA, 2000, 283: 1175–1182.

8. Nicoll A et al. Infant feeding and HIV-1 infection. AIDS, 2000, 14 (suppl. 3): S57–S74.

9. Coutsoudis A et al. Influence of infant-feeding patterns on early mother-to-child transmission of HIV-1 in Durban, South Africa: a prospective cohort study. Lancet, 1999, 354: 471–476.

10. Weiser B et al. Quantitation of human immunodeficiency virus type 1 during pregnancy: relationship of viral titer to mother-to- child transmission and stability of viral load. Proceedings of the National Academy of Sciences of the USA, 1994, 91: 8037–8041.

11. Dickover RE et al. Identification of levels of maternal HIV-1 RNA associated with risk of perinatal transmission: effect of maternal zidovudine treatment on viral load. JAMA, 1996, 275: 599– 605.

12. John GC et al. Genital shedding of human immunodeficiency virus type 1 during pregnancy: association with immunosuppression, abnormal cervical or vaginal discharge and severe vitamin A deficiency. Journal of Infectious Diseases, 1997, 175: 57– 62.

13. Zaba B, Gregson S. Measuring the impact of HIV on fertility in Africa. AIDS, 1998, 12 (suppl. 1): S41–S50.

14. The European Collaborative Study. Vertical transmission of HIV-1: maternal immune and obstetric factors. AIDS, 1996, 10: 1675–1681.

15. Rotchford K, Strum AW, Wilkinson D. Effect of coinfection with STDs and of STD treatment on HIV shedding in genital-tract secretions – systematic review and data synthesis. Sexually Transmitted Diseases, 2000, 27: 243–248.

16. Guay LA et al. Intrapartum and neonatal single-dose nevirapine compared with zidovudine for prevention of mother-to-child transmission of HIV-1 in Kampala, Uganda: HIVNET 012 randomised trial. Lancet, 1999, 354: 795–802.

17. Semba RD et al. Human immunodeficiency virus load in breast milk, mastitis, and mother-to-child transmission of human immunodeficiency virus type 1. Journal of Infectious Diseases, 1999, 180: 93–98.

18. Nduati RW et al. Human immunodeficiency virus type 1-infected cells in breast milk: association with immunosuppression and vitamin A deficiency. Journal of Infectious Diseases, 1995, 172: 1461–1468.