Mahy, Mary ScD, MHSc; Warner-Smith, Matthew MPH; Stanecki, Karen A MPH; Ghys, Peter D MD, PhD, MSc
HIV prevalence data have been the most commonly used information for determining the scale and evolution of HIV epidemics. However, because HIV prevalence is a function of HIV incidence and AIDS mortality, changes in HIV prevalence need to be interpreted in the light of current information on incidence and mortality, as countries can have very different patterns of mortality and incidence and yet have the same prevalence curve over time (Fig. 1). Successful prevention interventions lead to fewer new infections, whereas successful treatment would lead to steady or increasing numbers of people living with HIV. A perfectly successful HIV program would therefore result in no net change in prevalence. A reduction in prevalence is only desirable if it results from a reduction in incidence, whereas an increase in prevalence does not represent failure if it results from a decrease in mortality for persons on antiretroviral treatment (ART).
Program evaluation is based on knowledge or presumptions that certain interventions lead to particular changes in behaviors that in turn lead to changes in incidence and mortality. Using a logical framework, we expect inputs (such as money and staff time) to lead to outputs (such as accessible treatment programs), which lead to outcomes (such as people receiving ART), which culminate with an impact on the disease such as a reduction in AIDS-related mortality (Fig. 2). Evidence underpinning these logical frameworks exists for some interventions. For example, there is strong evidence that ART delays mortality from AIDS-related diseases.2,3 Similarly, good evidence exists that needle and syringe exchange programs for injecting drug users lead to reduced use of contaminated injection equipment and that reduced use of contaminated injecting equipment leads to reductions in HIV incidence.1-6
This article reviews the availability of measures of HIV incidence (among adults and children) and AIDS mortality since 2000. Specifically, it examines countries' abilities to measure progress toward the United Nations General Assembly Special Session on HIV/AIDS (UNGASS) Declaration of Commitment on HIV/AIDS (DoC) targets and proposes improvements necessary for measuring progress against Millennium Development Goal 6 in 2015.7
Measuring Incidence Among Adults
Ideally, countries would directly measure changes in the number of new infections to assess the performance of prevention programs. Although some countries have community-based cohorts in place that provide a direct measure of HIV incidence in small areas, it is not currently possible to measure directly the incidence rate or the annual number of new infections at the national level.
A potential way to measure the annual number of new infections is through case reports of new HIV diagnoses. However, there are wide variations in the completeness of case reporting systems depending largely on testing availability and uptake.8,9 In addition, newly diagnosed infections could have occurred several years previously. For those reasons, in most countries, case reports of new HIV diagnoses severely underestimate the true number of new HIV infections.
Much hope has been placed on deriving a measure of incidence from cross-sectional studies through the application of laboratory assays that can distinguish recent infections from longstanding infections. However, although this approach remains promising, the current assays and associated statistical methods present several challenges.10 These stem mainly from an overall lack of specificity of the assay related to HIV-infected individuals who never cross the threshold for the assay and therefore falsely seem to have been recently infected. The magnitude of this lack of specificity varies across settings.10 In addition, both severe immunosuppression (low CD4 counts) and ART are associated with cases in which individuals are misclassified as have been recently infected. New assays, assay algorithms and appropriate analytic methods need to be developed and extensively validated before this approach can be used for surveillance.
A third approach, which is described in this article, is to use HIV prevalence trends in subsets of the population as proxy measures for incidence trends. The proxy measures would differ depending on the nature and scale of the HIV epidemic in the country (whether it is a generalized epidemic or a concentrated/low-level epidemic).
Measuring Incidence Among Adults in Generalized Epidemics
In generalized epidemics (where HIV transmission is spreading in the general population mainly through heterosexual sex), trends in incidence can be deduced from trends in HIV prevalence among young people aged 15-24 years. Because infections in this age group are likely to have been relatively recent (and therefore not much affected by mortality or treatment), trends in HIV prevalence in this age group are presumed to be similar to trends in incidence.11 The ideal source for data on HIV prevalence among young men and women are national household surveys. However, these are typically carried out infrequently (usually every 5 years) because of high costs. Most countries that have undertaken such a survey have carried out only 1 to date; the data therefore do not yet allow for trends in prevalence to be measured. In addition to the direct trends in HIV prevalence among 15-year to 24-year olds, specific methods have been developed for deriving point estimates of age-specific HIV incidence from age-specific HIV prevalence data collected in consecutive national surveys.12 Again, this method will be more useful for deriving incidence trends once countries have conducted at least 3 such surveys.
As an alternative to using national household surveys to determine HIV prevalence trends among young people, it is possible to use HIV prevalence data from young pregnant women attending antenatal clinics (ANCs). HIV prevalence trends among this group can be used as a proxy for HIV incidence trends in the general population (although HIV prevalence trends among ANC attendees do not provide information on prevalence trends among men and they may not match exactly the trend in all women). Thus UNAIDS has recommended using prevalence among pregnant women aged 15-24 years attending ANCs.13
Accurate trends can only be assessed if data from consistent ANC sites are used over time. Depending on the coverage of the surveillance system, these consistent sites may or may not be representative of the entire country. For HIV prevalence among young ANC attendees to be comparable, the sites used to calculate the prevalence need to be consistent over time, any over sampling or weighting needs to be consistent between rounds and the quality of laboratory procedures needs to remain high. In a review of the quality of HIV serosurveillance systems in low-income and middle-income countries, the majority (24 of 40) of generalized epidemic countries had functional surveillance systems, 2 countries had systems that were not functional and the remaining 14 had partially functioning systems.14 This is reflected in the fact that the majority of generalized epidemic countries were able to report HIV prevalence among young women for UNGASS.
The number of countries with generalized HIV epidemics reporting HIV prevalence among young women attending ANCs increased over the 3 rounds of UNGASS reporting in 2003, 2005 and 2007 (Fig. 3). In the most recent round, 22 countries with generalized epidemics provided data based on surveillance among ANC attendees, 5 countries provided data from household surveys, 3 countries estimated this indicator using models and 10 countries used alternative data sources [usually prevention of mother-to-child transmission (PMTCT) program sites, blood donation sites or mandatory military testing]. The sources were not necessarily comparable for the different reporting periods, and therefore, could not be used to establish trends. An additional 48 concentrated or low-level epidemic countries submitted data for this indicator. Unfortunately, the majority of those countries did not use consistent sites over time, which hindered the calculation of trends based on the data submitted in their UNGASS reports.
Measuring HIV Incidence Among Adults in Concentrated/Low-Level Epidemics
Measuring incidence in concentrated and low-level epidemics poses additional challenges. In such epidemics, HIV prevalence in the general population, including women attending ANCs, is very low, making it difficult to measure trends over time. Instead, HIV prevalence among population groups with high-risk behaviors (sex workers, injecting drug users and men who have sex with men) is used to monitor the HIV prevention response. To report on their progress toward the UNGASS DoC, countries with concentrated or low-level epidemics have provided HIV prevalence rates among population groups with high-risk behaviors that were deemed relevant to their respective epidemics.
The number of countries able to report HIV prevalence in these populations has increased over time. In the first round of UNGASS (2003), indicators on HIV prevalence among population groups with high-risk behaviors were not included as required core UNGASS indicators; those indicators were added to the UNGASS core indicators in 2005. Figure 3 shows the increase in numbers of countries reporting data for these population groups. The largest increase occurred among countries reporting data from sex workers (Fig. 3).
Although these data are critical for national programming and for targeting interventions, they are not necessarily an accurate basis for establishing trends. As is the case for the analysis of data from pregnant women in generalized epidemic countries, accurate prevalence trends require data collected from the same sites and with the same sampling strategy. This is complicated because of several factors. In many countries, the behaviors that put populations at increased risk to HIV are illegal or highly stigmatized (such as injecting drugs, selling sex, or men having sex with men) making it difficult to reach those populations for surveillance purposes. A review of the surveillance systems in concentrated and low-level epidemic countries found that only 16 of the 87 concentrated epidemic countries had fully functional surveillance systems. Almost half were classified as poorly functional (42 of 87) and the remaining countries were classified as partially functional.14 Second, variations in sampling strategies and in population definitions limit countries' abilities to make conclusions about trends in HIV prevalence in most concentrated and low-level epidemic countries. For some population groups that present high population turnover (ie, on average, individuals stay in the population group for a relatively short period), trends in prevalence may mirror trends in incidence. However, the increasing coverage of ART makes interpretation of trends in prevalence more difficult. One approach for overcoming this difficulty is to limit the analysis of prevalence to individuals who recently initiated behaviors that put them at increased risk to HIV.
Measuring Impact of PMTCT
The UNGASS DoC also requires gauging the impact of PMTCT by measuring the percentage of infants who are born to HIV-infected women and who themselves are infected with the virus. The indicator is calculated based on modeled estimates of the number of HIV-infected pregnant women and the coverage of PMTCT services.
Alternative measures of the impact of PMTCT programs are challenging. With few exceptions, children have not been included for measurement of HIV prevalence in national population-based surveys. Many low-income and middle-income countries do not have the capacity to routinely test infants for HIV. Countries have only recently begun to put systems in place to detect HIV through polymerase chain reaction that can detect a child's HIV status a few weeks after birth. Infection of infants also takes place during breastfeeding, which can last 2 years or longer after birth in some settings and which further complicates the measurement of HIV incidence in children.
Measuring Impact of Treatment Programs
Two UNGASS indicators track progress in the provision of ART. The first is the percent of individuals still on treatment 12 months after starting treatment (this indicator was not required in 2003). In the 2005 round of UNGASS reporting, 10 countries reported on this indicator. This increased to 104 countries in the 2007 reporting round. In the 2007 reporting round, the median percentage of individuals still on treatment 12 months after starting treatment was 85% globally, ranging from 78% in Eastern Europe and Central Asia to 98% in Oceania. However, there are limitations to these data. Most countries do not have data management systems sophisticated enough to conduct the cohort analysis necessary for reporting against this indicator. Specifically, countries that request each of their ART treatment facilities to report on this indicator are faced with the difficulty of aggregating those data into a single national value for the indicator. Those countries may also underestimate the success of treatment because patients who transfer from one facility to another are not part of the numerator for that facility even though they are still on ART. Work is ongoing to establish more comprehensive and centralized data management systems.
The other UNGASS indicator related to ART programs is the percentage of people with advanced HIV infection receiving ART. The numerator (number of people on ART) is obtained from program records, whereas the denominator (number of persons in need of ART) is estimated by mathematical models (below). Among the UNGASS Country Progress Reports submitted in 2007, 100 countries reported ART coverage estimates.
The number of individuals on ART has increased dramatically from approximately 400,000 in 2003 to 3 million in 2007.15,16 Although research studies have shown the success of ART in terms of increased survival globally and in low-income and middle-income countries,2,3 only a few studies have demonstrated a change in AIDS mortality at the community level. Indeed, there is very little reliable data on AIDS mortality in low and middle-income countries.
Mathematical Modeling of Trends in Prevalence, Incidence, and Mortality
The data collected from surveillance systems, along with demographic data and assumptions derived from research studies, is used to model estimates of different HIV-related and AIDS-related epidemiological indicators. UNAIDS, with support from many other organizations, has developed and used the Workbook Method, the Estimation and Projection Package (EPP), and the Spectrum software to perform such modeling. EPP fits a smooth curve of adult HIV prevalence to a time series of observations of HIV prevalence using a mathematical formulation that reflects the dynamics of the epidemic.17 The Workbook Method is a simpler tool that fulfils the same function for countries with sparser data.18 The Spectrum package uses the HIV prevalence over time together with demographic information, epidemiological assumptions and information about the ART and PMTCT programs to model HIV prevalence, incidence and mortality indicators, and the number of people in need of ART.19 EPP and Spectrum are attuned to capturing long-term changes in incidence that would arise as the result of gradual and sustained behavioral modification and can capture the interplay between prevalence, incidence and mortality. This is important because the natural course of HIV epidemics includes a peak and a short-term decline in HIV prevalence followed by a plateau, after which prevalence tends to level. Such leveling reflects the mortality rate among infected persons (which lags some 9-11 years behind the peak in new infections20) gradually coming into balance with the incidence rate. EPP and Spectrum therefore allow for the examination of trends in incidence that includes an allowance for the natural dynamics of the epidemic. Other modeling tools, such as Asian Epidemic Model21 and Actuarial Society of South Africa Model,22 can be used for similar analysis.
By applying such models to epidemiological data through 2007, it has been possible to determine and publicize trends in adult HIV prevalence, the numbers of people living with HIV, HIV incidence and AIDS mortality. In 2008, estimates were published for 153 countries. Thanks to improvements in the available surveillance and surveys data, the accuracy of the epidemiological assumptions and the structure of the mathematical model itself, these estimates have become more accurate in recent years.
The global annual number of new HIV infections among adults peaked around 1996 at almost 3 million and has been stable in recent years at around 2.3 million (Fig. 4A). Given the continued growth of the world's population, this implies a decline in the incidence rate (the number of persons infected annually as a proportion of the susceptible population). The models suggest that HIV prevalence among children is declining more rapidly, from a peak of around 450,000 in 2001 to around 370,000 in 2007 (Fig. 4B). It is estimated that mortality due to AIDS peaked in 2005 for both adults and children at around 2.2 million and declined slightly subsequently, to around 2 million in 2007 (Fig. 4C).
These same models have also been used to estimate the number of life years gained due to ART. For example, between 2002 and 2006, an estimated, 2 million life years were saved in low-income and middle-income countries,23 a number that is increasing rapidly with the continued rollout of ART.
Despite the limitations to the data available and the technical challenges in measuring incidence, it is possible to assess trends in measures of the impact of the response to the HIV epidemic.
Preventing HIV in Generalized Epidemics
Data from 15-year to 24-year olds attending consistent ANC sites suggest a reduction of HIV incidence has occurred in 14 of 16 high-burden countries with sufficient data for the analysis.16,24 At the same time, results from models also suggest a stabilization or reduction in incidence in recent years. Taken together, this evidence suggests that there is indeed a reduction in incidence, especially in countries with generalized epidemics.
A reduction in incidence does not necessarily reflect the success of national programs, but could be due to the natural history of the epidemic and the saturation of infections among those at most risk for infection.25 Recent attempts to overcome this limitation have used triangulation or modeling to link changes in incidence to changes in behaviors.
A number of recent studies have compared the outcomes (behavior) and impact (prevalence among 15-year to 24-year olds) using available data for a select set of countries. A recent analysis shows that of 16 countries with sufficient data, 14 showed a decline in prevalence among 15-year-old to 24-year-old pregnant women in either urban or rural areas, with 7 countries showing a statistically significant trend.16,24 Six of those 7 countries had sufficient data to analyze trends in sexual behavior; in them, significant declines in risky behavior were observed.
Modeling of both biological and behavioral data has confirmed declines in incidence that are temporally associated with reductions in risk behavior in several countries,26 including in Uganda and Zimbabwe where additional evidence for incidence declines associated with behavioral change is available from cohort studies in small areas.27-30 Uganda is widely credited with having achieved the first successful national response to HIV. However, to explain adequately “what worked” to achieve HIV incidence declines, robust epidemiological estimates must be triangulated with equally robust evaluations of interventions and policies.31,32
Preventing HIV in Concentrated and Low-Level Epidemics
There is still very little concrete evidence on possible declines in national HIV incidence in low-level and concentrated epidemic countries. There have been a few examples of countries that have shown trends in HIV prevalence among consistently sampled population groups with high-risk behavior. In other countries, data from mismatched sites with imperfect sampling techniques provide a rough description of the change in the epidemic in those countries.
Among the countries that were able to show a national decline in incidence, a few have provided strong examples of monitoring trends. In Thailand, the 100% condom promotion campaign resulted in lower prevalence levels among sex workers, military recruits and the general population.33,34,16 Cambodia has also seen a reduction in prevalence over time,16 with specific recent data from brothel-based sex workers also suggesting a decline in HIV incidence among sex workers.35 In Ukraine, reductions in HIV among drug users who had recently begun injecting seem to correspond to improvements in HIV prevention programs.36
Preventing HIV Among Infants
Mathematical modeling suggests that there is a decline in incidence among children younger than 15 in most countries (approximately 90% of these infections are a result of mother-to-child transmission). However, there is little empirical data in low-income and middle-income countries that confirm such declines in incidence. In Thailand-where there is very high follow-up to children exposed to HIV during pregnancy, delivery or breastfeeding-the number of children diagnosed with HIV has decreased between 2001 and 2005.37 Other evidence of the impact of PMTCT programs comes from developed countries where the rate of HIV transmission from mother to child has decreased to approximately 1%.38 Although there is evidence that the coverage of PMTCT services is increasing,15,16 few countries currently are able to directly monitor the impact of this intervention (whether fewer infants are becoming infected) at a national level.
Reducing AIDS-Related Mortality
Mathematical modeling shows a decline in mortality reflecting the large-scale rollout of ART. Improved data on the survival of people on ART and better information on what happens to individuals that are lost to follow-up could improve modeling further.
Several studies provide empirical evidence that ART programs are leading to a reduction in AIDS mortality. A recent study in rural Malawi documented a 10% decrease in overall mortality coincident with the early scale-up of ART.39 In Addis Ababa, Ethiopia, based on data from burial societies, ART was estimated to have reduced AIDS deaths in 2007 by 57%-63%.40 Although many of the most-affected countries do not have functional vital registration systems, there is underreporting of AIDS as the cause of death even in countries where these systems are functional because of stigma and underdiagnosis.41-44 In South Africa, a country with very high HIV prevalence, the rise in mortality (from all causes) began leveling off in 2005 as ART provision expanded.45 Several countries have sentinel demographic surveillance that is expected to contribute data on overall mortality and AIDS-specific mortality through verbal autopsies. Some countries are beginning to implement mortality monitoring systems, which are linked either to census or other national surveys.46,47
Looking Ahead to 2015
Modeling suggests that both HIV incidence and AIDS mortality are declining globally. More specific evidence for high-burden countries substantiates these global trends. However, additional information is needed to confirm the trends and refine our understanding of the changing epidemiology of HIV. Specifically, what is needed are (1) more and better surveillance data, including mortality data and a direct measure of incidence; (2) more and improved analysis of these data; (3) more and better monitoring data showing the quality and coverage of interventions; and (4) high-quality evaluations that take into consideration all factors that influence the epidemics, including the programmatic responses to HIV.
Logical frameworks provide a basis for determining the plausibility of whether changes in the epidemics are the result of HIV interventions. However, before such conclusions can be arrived at, reliable evidence is required that programs have been implemented to a sufficiently high quality and on a sufficiently large scale to be able to cause the desired effects. In short, we must ask: Are we doing the right things? Are we doing them correctly? And, are we doing them on a large enough scale to make a difference? Unless these 3 questions can be unambiguously answered, there is no rational basis for seeking to attribute changes in epidemiology to programs. Addressing these questions requires evidence of the validity of logical frameworks underpinning the program (ie, the ability of each intervention to produce the desired effect), evidence of the extent to which programs have been implemented as planned (program fidelity) and the quality of services provided and measures of program coverage.
When changes in impact measures are observed that are geographically and temporally consistent with reliable evidence of high-quality program coverage, it becomes plausible to assume that there is a relationship between the program and the epidemic. However, before attribution can be made, an understanding of the dynamics of HIV epidemics and the variety of factors that contribute to HIV transmission and AIDS mortality other than HIV interventions is required, to adjust for potentially confounding factors. Therefore, in addition to better quality data on HIV incidence and mortality and greatly enhanced monitoring of programmatic responses to HIV, sound evaluation studies are required to begin to ascribe attribution.
Much progress has been made in tracking trends in HIV epidemics. Stronger measures of incidence and mortality are needed, as are more complete data from a larger number of countries. Inspite of these limitations, it seems probable that by 2015, it will be possible to make a reasonable objective determination as to whether or not the HIV epidemic has been “halted and reversed” at the global level. It will also be important for each individual country to assess this. Many countries urgently need to improve their epidemiological data collection to allow such assessment. Whether programmatic responses to HIV have been successful will be considerably more difficult to determine. This will require more and better data on program quality and coverage and a better understanding of the full range of factors that influence HIV epidemics.
1. Brinkhof MWG, Boulle A, Weigel R, et al. Mortality of HIV-infected patients starting antiretroviral therapy in Sub-Saharan Africa: comparison with HIV unrelated mortality. PLoS Med. 6(4): e1000066. doi:10.1371/journal.pmed.1000066.
2. The Antiretroviral Therapy in Lower Income Countries (ART-LINC) Collaboration and ART Cohort Collaboration (ART-CC) groups. Mortality of HIV-1-infected patients in the first year of antiretroviral therapy: comparison between low-income and high-income countries. Lancet. 2006;367:817-824.
3. Des Jarlais DC, Marmor M, Paone D, et al. HIV incidence among injecting drug users in New York City syringe-exchange programmes. Lancet. 1996;348:987-991.
4. Drucker E, Lurie P, Wodak A, et al. Measuring harm reduction: the effects of needle and syringe exchange programs and methadone maintenance on the ecology of HIV. AIDS. 1998;12(Suppl A):S217-S230.
5. Hartgers C, Buning E, van Santen G, et al. The impact of the needle and syringe-exchange programme in Amsterdam in injecting risk behaviour. AIDS. 1989;3:571-576.
6. Lurie P, Sorensen J, Lane S, et al. The public health impact of needle exchange programs (NEPS). Int Conf AIDS. 1994;10:72.
7. United Nations. Declaration of Commitment on HIV/AIDS. New York, USA: United Nations; 2001.
8. EuroHIV. HIV/AIDS Surveillance in Europe. End-year report 2006. Saint-Maurice, France: Institut de veille sanitaire; 2007. No. 75.
10. Hargrove JW, Humphrey JH, Mutasa K, et al. Improved HIV-1 incidence estimates using the BED capture enzyme immunoassay. AIDS. 2008;22:511-518.
11. Zaba B, Boerma T, White R. Monitoring the AIDS epidemic using HIV prevalence data among young women attending antenatal clinics: prospects and problems. AIDS. 2000;14:1633-1645.
12. Hallett TB, Zaba B, Todd J, et al, on behalf of the ALPHA Network. Estimating incidence from prevalence in generalized HIV epidemics: methods and validation. PLoS Med. 2008;5(4):e80.
13. UNAIDS. Monitoring the Declaration of Commitment on HIV/AIDS: guidelines on construction of core indicators: 2010 reporting. Geneva, Switzerland: Joint United Nations Programme on HIV/AIDS (UNAIDS); 2009.
14. Lyerla RL, Gouws E, Garcia-Calleja JM. The quality of sero-surveillance in low- and middle-income countries: status and trends through 2007. Sex Transm Infect. 2008;84(Suppl I):i85-i91.
15. WHO. Towards Universal Access: Scaling up Priority HIV/AIDS Interventions in the Health Sector: Progress Report 2008. Geneva, Switzerland: World Health Organization; 2008.
16. UNAIDS. 2008 Report on the Global AIDS Epidemic. Geneva, Switzerland: Joint United Nations Programme on HIV/AIDS (UNAIDS); 2009.
17. Brown T, Salomon JA, Alkema L, et al. Progress and challenges in modelling country-level HIV/AIDS epidemics: the UNAIDS Estimation and Projection Package 2007. Sex Transm Infect. 2008;84(Suppl 1):i5-i10.
18. Lyerla R, Gouws E, Garcia-Calleja JM, et al. The 2005 Workbook: an improved tool for estimating HIV prevalence in countries with low level and concentrated epidemics. Sex Transm Infect. 2006;82(Suppl 3):iii41-iii44.
19. Stover J, Johnson P, Zaba B, et al. The Spectrum projection package: improvements in estimating mortality, ART needs, PMTCT impact and uncertainty bounds. Sex Transm Infect. 2008;84(Suppl 1):i24-i30.
20. Todd J, Glynn JR, Marston M, et al. Time from HIV seroconversion to death: a collaborative analysis of eight studies in six low and middle-income countries before highly active antiretroviral therapy. AIDS. 2007;21(Suppl 6):S55-S63.
21. Brown T, Peerapatanapokin W. The Asian Epidemic Model: a process model for exploring HIV policy and programme alternatives in Asia. Sex Transm Infect. 2004;80(Suppl I):i19-i24.
22. Dorrington RE, Bradshaw D, Johnson L, et al. The Demographic Impact of HIV/AIDS in South Africa: National and Provincial Indicators for 2006. Joint publication by the Centre for Actuarial Research, the Burden of Disease Research Unit (Medical Research Council) and the Actuarial Society of South Africa. 2006. Available at: http://aids.actuarialsociety.org.za/Models-3145.htm
. Accessed October 2, 2009.
23. UNAIDS. 2006 AIDS Epidemic Update. Geneva, Switzerland: Joint United Nations Programme on HIV/AIDS (UNAIDS); 2006.
24. Gouws E, Stanecki KA, Lyerla R, et al. The epidemiology of HIV infection among young people aged 15-24 years in southern Africa. AIDS. 2008;22(Suppl 4):S5-S16.
25. Hallett TB, White PJ, Garnett GP. Appropriate evaluation of HIV prevention interventions: from experiment to full-scale implementation. Sex Transm Infect. 2007;83;55-60.
26. Hallett TB, Aberle-Grasse J, Bello G, et al. Declines in HIV prevalence can be associated with changing sexual behaviour in Uganda, urban Kenya, Zimbabwe, and urban Haiti. Sex Transm Infect. 2006;82(Suppl I):i1-i8.
27. Stoneburner RL, Low-Beer D. Population-Level HIV declines and behavioral risk avoidance in Uganda. Science. 2004;304:714-718.
28. Gregson S, Garnett GP, Nyamukapa CA, et al. HIV decline associated with behavior change in Eastern Zimbabwe. Science. 2006;211:664-666.
29. Mahomva A, Greby S, Dube S, et al. HIV prevalence and trends from data in Zimbabwe, 1997-2004. Sex Transm Infect. 2006;82(Suppl 1):i42-i47.
30. Mbulaiteye S, Mahe C, Whitworth J, et al. Declining HIV-1 incidence and associated prevalence over 10 years in a rural population in south-west Uganda: a cohort study. Lancet. 2002;360:41-46.
31. Parkhurst J. “What worked?”: the evidence, challenges in determining the causes of HIV prevalence decline. AIDS Educ Prev. 2008;20:275-283.
32. Kirby D. Changes in sexual behaviour leading to the decline in the prevalence of HIV in Uganda: confirmation from multiple sources of evidence. Sex Transm Infect. 2008;84:ii35-ii41.
33. Nelson KE, Celentano DD, Eiumtrakol S, et al. Changes in sexual behaviour and a decline in HIV infection among young men in Thailand. N Engl J Med. 1996;86:1713-1716.
34. Rajanapithayakorn W, Hanenberg R. The 100% condom program in Thailand. AIDS. 1996;10:1-7.
35. Ly Penh. Proxy indicator for HIV incidence measure. Presented at: 2nd Global HIV/AIDS Surveillance Meeting; March 5, 2009; Bangkok, Thailand.
36. Saliuk T, Varetska O, Skoropatska Y, et al. First indications of impact of prevention programs among injecting drug users on the HIV epidemic in Ukraine: results based on data triangulation from different sources. Presented at: 2nd Global HIV/AIDS Surveillance Meeting; March 5, 2009; Bangkok, Thailand.
37. Plipat T, Naiwatanakul T, Rattanasuporn N, et al. Reduction in mother-to-child transmission of HIV in Thailand, 2001-2003: results from population-based surveillance in six provinces. AIDS. 2007;21:145-151.
38. Paintsil E, Andiman WA. Update on successes and challenges regarding mother-to-child transmission of HIV. Curr Opin Pediatr. 2009;21(1):94-101.
39. Jahn A, Floyd S, Crampin AC, et al. Population-level effect of HIV on adult mortality and early evidence of reversal after introduction of antiretroviral therapy in Malawi. Lancet. 2008;371:1603-1611.
40. Reniers G, Araya T, Davey G, et al. Steep declines in population-level AIDS mortality following the introduction of antiretroviral therapy in Addis Ababa, Ethiopia. AIDS. 2009;23:511-518.
41. Dorrington R, Bourne D, Bradshaw D, et al. The impact of HIV/AIDS on adult mortality in South Africa. 2001 September. Medical Research Council. Available at: http://www.mrc.ac.za/bod/
. Accessed October 2, 2009.
42. Anderson BA, Phillips HE. Adult Mortality (Age 15-64) Based on Death Notification Data in South Africa: 1997-2004. Report No. 03-09-05. Pretoria, South Africa: Statistics South Africa; Actuarial Society of South Africa; 2005.
43. Medical Research Council. South African national burden of disease study 2000. Cape Town, South Africa: Medical Research Council; 2005. Available at: http://www.mrc.ac.za/bod/estimates.htm
. Accessed October 2, 2009.
44. Bradshaw D, Laubscher R, Dorrington R, et al. Unabated rise in number of adult deaths in South Africa. S Afr Med J. 2004;94:278-279.
45. Smith A. Vital registration in South Africa. Presented at: 2nd Global HIV/AIDS Surveillance Meeting; March 4, 2009; Bangkok, Thailand.
46. Mazive E. Mozambique: Post-Census Mortality Survey. Presented at: 2nd Global HIV/AIDS Surveillance Meeting; March 4, 2009; Bangkok, Thailand.
47. Central Statistical Office Zambia. Sample Vital Registration with Verbal Autopsy. Presented at: 2nd Global HIV/AIDS Surveillance Meeting, March 4, 2009; Bangkok, Thailand. (presented by Lorraine West).
© 2009 Lippincott Williams & Wilkins, Inc.