Estimated deaths from human immunodeficiency virus (HIV) over 10 years in South Africa for different HIV testing and treatment strategies. A comparison of the total number of HIV-related deaths over 10 years, by strategy, scaled to South Africa. The error bars represent the 95% confidence bounds from the probabilistic sensitivity analysis. LTFU indicates loss to follow-up.
Projected human immunodeficiency virus (HIV) prevalence in South Africa for different HIV testing and treatment strategies. LTFU indicates loss to follow-up.
Sensitivity analysis of the mortality benefits and reduction in transmission from gradual improvements in linkage to care and prevention of loss to follow-up (LTFU). A, The mortality benefits; B, the reduction in new human immunodeficiency virus (HIV) infections. Each line represents the benefit compared with universal testing and treatment alone. The relative benefits for improved linkage range over 67% to 100% linkage, and for LTFU from 20% to 0% loss. The comprehensive strategy represents simultaneous improvements in linkage and prevention of LTFU. For example, universal testing and treatment with improved linkage to 80% results in a gain of about 2.5 months of life expectancy, or 15% improvement, over universal testing and treatment alone.
Bendavid E, Brandeau ML, Wood R, Owens DK. Comparative Effectiveness of HIV Testing and Treatment in Highly Endemic Regions. Arch Intern Med. 2010;170(15):1347-1354. doi:10.1001/archinternmed.2010.249
Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010
Universal testing and treatment holds promise for reducing the burden of human immunodeficiency virus (HIV) in sub-Saharan Africa, but linkage from testing to treatment sites and retention in care are inadequate.
We developed a simulation of the HIV epidemic and HIV disease progression in South Africa to compare the outcomes of the present HIV treatment campaign (status quo) with 4 HIV testing and treating strategies that increase access to antiretroviral therapy: (1) universal testing and treatment without changes in linkage to care and loss to follow-up; (2) universal testing and treatment with improved linkage to care; (3) universal testing and treatment with reduced loss to follow-up; and (4) comprehensive HIV care with universal testing and treatment, improved linkage to care, and reduced loss to follow-up. The main outcome measures were survival benefits, new HIV infections, and HIV prevalence.
Compared with the status quo strategy, universal testing and treatment (1) was associated with a mean (95% uncertainty bounds) life expectancy gain of 12.0 months (11.3-12.2 months), and 35.3% (32.7%-37.5%) fewer HIV infections over a 10-year time horizon. Improved linkage to care (2), prevention of loss to follow-up (3), and comprehensive HIV care (4) provided substantial additional benefits: life expectancy gains compared with the status quo strategy were 16.1, 18.6, and 22.2 months, and new infections were 55.5%, 51.4%, and 73.2% lower, respectively. In sensitivity analysis, comprehensive HIV care reduced new infections by 69.7% to 76.7% under a broad set of assumptions.
Universal testing and treatment with current levels of linkage to care and loss to follow-up could substantially reduce the HIV death toll and new HIV infections. However, increasing linkage to care and preventing loss to follow-up provides nearly twice the benefits of universal testing and treatment alone.
The human immunodeficiency virus (HIV) epidemic in many sub-Saharan African countries has stabilized in the past few years, with a few countries reporting reductions in incidence, prevalence, and mortality.1- 3 However, the epidemic is still an unsustainable and disproportionate challenge to southern Africa, responsible for more than20% of adult mortality in some countries, an increasing number of orphans, and possible reversals in economic growth.1,4 Reducing the burden of the epidemic is a major goal of HIV testing and treatment programs, but difficulties in linking infected individuals to treatment sites and retaining them in care, low testing rates, and resource constraints challenge the capacity to achieve universal access.5- 9
Recent studies suggest that universal testing and treatment may decrease HIV prevalence in highly endemic regions through reduced incidence (infected patients who receive treatment are less likely to infect others), while at the same time markedly reducing HIV mortality.10 This strategy has considerable appeal, and clinical trials looking at the effectiveness of early antiretroviral therapy (ART) for HIV prevention are currently under way.11 However, previous estimates of the benefits of universal testing and treatment did not take into account the poor linkage between testing and treatment sites and the high rates of attrition from care after treatment initiation. Among patients who receive a diagnosis of HIV infection in South Africa, one-third to two-thirds never return for follow-up care.12,13 In addition, many clinics report high rates of loss to follow-up (LTFU)—4% to 39% in a recent systematic review (after accounting for mortality following ART initiation).14 These factors are increasingly recognized as central barriers to scale-up of ART programs in sub-Saharan Africa.
Given the limited resources for scaling up HIV testing and treatment in Africa, assessing the role of improving linkage to care and reducing loss to follow-up is critical. In this study, we assess the epidemiologic and health effects of 4 strategies to increase access to antiretroviral therapy: universal testing and treatment without substantial changes in rates of linkage to care and LTFU; universal testing and treatment with improved linkage to care; universal testing and treatment with reduced LTFU; and comprehensive HIV care with universal testing and treatment, improved linkage to care, and reduced LTFU.
We developed a stochastic HIV disease and transmission model in an adult population similar to that in South Africa, where HIV transmission is predominantly heterosexual. We used the model to evaluate the relative effectiveness of different strategies for scaling up access to ART through expanded testing, improved linkage to care, earlier treatment initiation, and reduced rates of loss to follow-up. The model follows groups of uninfected and HIV-infected individuals over time, and aggregates individual health outcomes as well as epidemiologic measures of HIV burden, such as incidence and prevalence.
We designed the model to reflect the current pace of scale-up in South Africa, including the rate of HIV testing, rate of linkage to care, treatment initiation thresholds, and rates of loss to follow-up. We describe the model structure, key assumptions, and the scale-up strategies.
The model follows groups of 10 000 individuals representative of the population of South Africa by age, sex, HIV status, circumcision status, and number of sexual partners.15- 18 Individuals enter the population at age 15 years and leave the population when they die from HIV or other causes. Baseline demographic parameters are shown in Table 1. The population is followed in 1-month intervals over a period of 10 years, and the health status of individuals is evaluated monthly based on their age, sex, HIV status, and HIV testing and treatment history. Risk factors for infection and disease progression are presented in Table 1 and a supplement (available at the authors’ Web site, http://healthpolicy.stanford.edu/publications/comparative_effectiveness_of_hiv_testing_and_treatment_strategies_in_highly_endemic_regions/ and on request).
Human immunodeficiency virus status in the initial time period of the model is determined by current estimates of age- and sex-specific HIV prevalence in South Africa. Individuals who are uninfected may become infected if they have at least 1 HIV-infected sexual partner. The number of sexual partnerships at any point in time is determined by estimates of concurrency in South Africa.20,38 The risk of infection per month is calculated based on the sex of the individual, the number of infected partners, and the viral load of any infected partners. The viral load of infected partners may be either suppressed or elevated, based on the distribution of viral loads in the population. Uninfected partners of individuals who recently seroconverted have a much higher risk of acquiring HIV infection. Risk of infection is reduced for men who are circumcised and for individuals who received HIV counseling. The magnitude of risk reduction is shown in Table 1.
Infected individuals are identified through HIV testing or when they develop a severe opportunistic disease. Individuals who develop a severe opportunistic disease are linked to care and initiate treatment, even if they were lost to care between testing and treatment or lost to follow-up. We assumed that those who are identified through testing are referred to an HIV treatment facility, where they are followed prior to and after starting ART. The imperfect linkage between testing and treatment sites, the treatment initiation criteria, and the LTFU after treatment initiation are shown in Table 2. We assumed that individuals were monitored on average every 6 months for clinical symptoms of disease progression and with CD4 lymphocyte counts.
All individuals in our population are eligible to be tested. In the status quo strategy, testing rates reflect the current mix of voluntary testing and counseling or health care provider–initiated testing, while the universal testing and treatment strategies assume that 90% of the population is tested and that individuals undergo testing every 2 years on average. In the status quo strategy, we assumed that 67% of individuals who tested positive for HIV are linked with an HIV treatment facility.39 Without targeting linkage, we assume that this rate of presentation to treatment sites remains unchanged.
We used our previous work to model the progress of infected individuals.40,41 Individuals who are linked to a treatment program are monitored regularly to determine the appropriate timing for initiating ART. We assumed that individuals are monitored for clinical signs of advanced disease and CD4 lymphocyte counts to guide timing of treatment initiation. Timing of treatment initiation is a topic of substantial debate42; our status quo strategy estimates reflect current guidelines (Table 2). Finally, we tracked whether individuals remained in treatment over time. We estimated rates of LTFU based on clinical experience from Cape Town but considered the range of reported experience in southern Africa. Base-case estimates of each component of HIV care are shown in Table 2.
We projected the course of the epidemic using the status quo strategy and 4 additional strategies of universal testing and treatment. The strategies (Table 2) are as follows:
Status quo strategy: The number of people tested for HIV, accessing treatment centers, and starting treatment continues to increase, but the pace of scale-up remains similar to that observed since 2007.
Universal testing and treatment: In this strategy, there is a shift from the current rate of testing, in which 18% of the adult population have been tested previously and about 10% of the adult population are tested per year, to a strategy in which 90% of the adult population are tested in the first 2 years, and individuals seek an additional HIV test every 2 years on average. About 67% of those who have positive HIV test results are linked to a treatment facility and start treatment with first-line ART within 6 months after their diagnosis; and about 20% of individuals who start treatment are lost to follow-up in the first year of treatment. Compared with the status quo strategy, this strategy illustrates the benefits of universal testing and treatment with the current approaches to patient linkage and retention in care.
Universal testing and treatment with improved linkage to care: This is the same as strategy 2, except that linkage to care is now 100%; that is, every individual who has a positive HIV test result is linked with a treatment facility and starts treatment within 6 months of diagnosis. Rates of LTFU are unchanged.
Universal testing and treatment with prevention of LTFU: This is the same as strategy 2, except that retention in care is 100%. As with strategy 2, 33% of individuals with positive HIV test results are not linked with treatment facilities.
Universal testing, improved linkage to care, early treatment, and prevention of LTFU (comprehensive): This includes universal testing, improved linkage to care, early treatment, and prevention of LTFU as noted for strategies 2 to 4. While perfect linkage and retention in care are not likely to be possible, this strategy is illustrative of the maximal benefits of a universal testing and treatment approach.
We measured 2 primary outcomes: gains in life expectancy and reduction in new HIV infections at the end of 10 years compared with the status quo strategy. The former is a measure of health benefits for individuals, whereas the latter is an important determinant of the burden of disease. While the benefits of scaling up HIV care apply most directly to the infected population, the benefits spill over to the uninfected population through reduced HIV transmission. Thus, we measured the gains in life expectancy across the entire population. We estimated 4 other important epidemiologic and health outcomes: reduction in HIV-related deaths, HIV death rates, adult HIV prevalence, and population growth. Life expectancy gains are reported using a 3% discount rate.
Recognizing the challenges in achieving 100% linkage and retention in care, we varied the rates from the status quo levels to perfect linkage and retention and report the results in percentage of improvement over universal testing and treatment without changes to linkage or retention in care. For example, improving linkage from 67% to 90% may provide a 25% greater improvement in life expectancy compared with universal testing and treatment without improvements in linkage. In probabilistic sensitivity analysis we varied all parameters simultaneously (detailed in the supplement). We drew each input parameter from a random distribution (normal, beta, gamma, or uniform) and repeated the analysis 1000 times. We report the results of this analysis as a 95% uncertainty bound around our estimates. These bounds are reported throughout the “Results” section and in the Tables and Figures.
In the status quo strategy, we estimate that the prevalence of HIV in South African adults will decrease from 18.0% to 17.2% over the next decade. This is consistent with current epidemiologic projections and recent trends.43 We estimate the HIV-specific mortality rate in the first year of our analysis to be 1140 deaths per 100 000 individuals older than 15 years, consistent with World Health Organization estimates from vital registration data and demographic projections.22,44
Compared with the status quo strategy, universal testing and treatment alone (strategy 2) was associated with a per-person gain in life expectancy of 12.0 (11.3-12.2) months averaged over the entire population. Universal testing and treatment with improved linkage to care (strategy 3) and prevention of LTFU (strategy 4) were associated with greater life expectancy gains: 16.1 and 18.6 months per person over the entire population, respectively, while comprehensive HIV care (strategy 5) was associated with an average life expectancy gain of 22.2 months per person. Two related events account for the gains in life-years: a decrease in HIV mortality from improved case detection and care, and an increase in the size of the population owing to the decreased mortality in the population of childbearing age. At the end of 10 years, we estimate that the comprehensive strategy resulted in 61.6% fewer deaths from HIV compared with the status quo strategy and an HIV death rate of 415 (318-515) per 100 000 person-years, 63.5% lower than the status quo strategy. Universal testing and treatment alone (strategy 2) yielded a 27.7% decrease in deaths from HIV and an HIV mortality rate of 802 (664-948) per 100 000 person-years. The mortality benefits are summarized in Table 3 and Figure 1.
While scaling up HIV care is likely to improve survival of HIV-infected individuals, the effect on HIV transmission and prevalence is less obvious. Increasing treatment coverage may reduce infectivity per coital act. However, the longer survival of HIV-infected individuals and reduction in deaths may increase the opportunity for individuals to infect others and may increase prevalence, which is in turn a determinant of infection. We estimate the number of new infections in the adult South African population over the next 10 years to be 4.5 (3.8-5.1) million in the status quo strategy, and 1.2 (0.9-1.6) million in a comprehensive strategy, a 73.2% reduction.
The decrease in new infections led to reductions in adult HIV prevalence compared with the status quo strategy of 1.6%, 2.5%, 1.9%, and 3.4% with universal testing and treatment alone, universal testing and treatment with improved linkage, universal testing and treatment with reduced LTFU, and the comprehensive strategy, respectively, at the end of 10 years. The reduction in new infections and HIV deaths, especially among individuals of childbearing age, was associated with an increase in the size of the population. We estimated 11.5% population growth over 10 years in the status quo strategy (about 1.1% per year), and 15.2% with universal testing and treatment alone (strategy 2). Figure 2 and Table 4 show the transmission and demographic benefits.
Figure 3 shows the sensitivity of our results to gradually improving linkage to care and reducing LTFU. The figure shows that even relatively modest improvements in linkage to care and prevention of LTFU provide substantial mortality and prevention benefits. A 10% higher linkage and 6% reduction in LTFU (67%-77% and 20%-14%, respectively) are associated with life expectancy improvements that are 30% higher than with universal testing and treatment alone. Similar improvements in linkage and prevention of LTFU are associated with 36% fewer HIV infections compared with universal testing and treatment alone. We performed 2-way sensitivity analysis on rates of testing and treatment initiation threshold, which showed that as long as treatment is initiated early, substantial reductions in prevalence and new infections can be accomplished with testing rates as low as 30% to 40% of the population. These results are discussed in the supplement.
Our comparative evaluation of the mortality and transmission benefits of scaling up HIV testing and treatment in sub-Saharan Africa provides several important insights. First, universal testing and early treatment alone have important health and epidemiologic benefits, but we estimate that they provide about half of the benefits of a comprehensive scale-up strategy that also includes improved linkage to care and prevention of loss to follow-up. Second, the mortality and transmission benefits of scaling up HIV testing and treatment have implications for population growth. Finally, we estimate that even under a strategy of comprehensive HIV care it will take longer than a decade to substantially reduce the burden of South Africa's widespread epidemic.
Our results support the notion that universal testing and treatment have significant mortality benefits in South Africa. Recent estimates from South Africa suggest that ART may prolong life expectancy of infected individuals by 12.5 years.45 We estimate that a comprehensive HIV testing, treatment, and care strategy will increase the average life expectancy of the entire population by 22 months compared with the status quo strategy. From 1990 to 2006, life expectancy in South Africa declined by about 12 years. Scaling up HIV testing and treatment may go a considerable way toward reversing that trend.19 Although a comprehensive plan with perfect linkage and full retention in care is not realistic, it provides an important bound for the possible benefits.
However, over a decade, the benefits of universal testing and treatment alone are much lower than the benefits of universal testing and treatment with improved linkage to care and prevention of loss to follow-up. This underscores the role of increasing the number of people who initiate treatment early: each individual who starts treatment early decreases the number of downstream infections by more than 1. Insights from mathematical epidemiology suggest that, in the absence of ART, each individual infected with HIV transmits the infection to more than 1 person, on average, over his or her lifetime. Thus, the benefits of long-term effective ART are multiplied, and so are the losses from having individuals forego ART because of poor linkage to care or LTFU.
Despite all these benefits, we find that even under the comprehensive strategy of HIV care, the burden of disease over the next decade is expected to remain substantial. Some researchers suggest that it would take as long as 50 years to reduce HIV prevalence in South Africa to below 1%.10 Our estimates, which include a detailed microsimulation of HIV disease and treatment, demographic changes, and multiple transmission risk factors, agree with these estimates: we show a nearly linear decrease in prevalence of 4.2% over a decade, suggesting that it would take more than 4 decades at the estimated rate of decline to decrease prevalence to around 1%.
Our estimates of benefits rely on several important assumptions. Most important, we assumed that HIV transmission risk is reduced for individuals receiving ART. Although much evidence supports this phenomenon, to our knowledge it has not been verified in a major clinical trial to date.46,47 We made several assumptions about the benefits of treatment in South Africa that affect our estimates of the longevity benefits of ART but do not change the comparative effectiveness of the strategies we examined. We also assumed no behavioral risk modification with decreasing disease burden; for example, as HIV mortality and prevalence drop, individuals may perceive the disease as less threatening and increase risk behaviors, such as multiple concurrent partnerships.48 However, we had no basis for assuming the type or extent of behavior risk modification. Finally, we assumed that the fertility rate will remain stable (ie, the number of children per woman will not change over the next decade). A decrease in the fertility rate may slow the decline in prevalence, because the growth in population size will slow down while the number of infected individuals may not change appreciably.
Our analysis uses a detailed epidemiologic simulation model to estimate the mortality and transmission benefits of HIV testing, treatment, and care in South Africa and quantifies the comparative effectiveness of alternative strategies for universal testing and treatment. We find that scaling up all aspects of HIV care nearly doubles the benefits of universal testing and treatment alone. An economic and operational evaluation of these strategies would further help in clarifying priorities.
Correspondence: Eran Bendavid, MD, MS, Division of General Internal Medicine, Stanford University, 117 Encina Commons, Stanford, CA 94305-8526 (email@example.com).
Accepted for Publication: February 10, 2010.
Author Contributions:Study concept and design: Bendavid, Wood, and Owens. Acquisition of data: Bendavid. Analysis and interpretation of data: Brandeau and Owens. Drafting of the manuscript: Bendavid and Owens. Critical revision of the manuscript for important intellectual content: Brandeau and Wood. Statistical analysis: Bendavid and Owens. Obtained funding: Owens. Study supervision: Brandeau, Wood, and Owens. Modeling expertise: Brandeau.
Financial Disclosure: None reported.
Funding/Support: This research is supported in part by the National Institute of Allergy and Infectious Diseases (K01-AI084582), the Department of Veterans Affairs, and the National Institute on Drug Abuse (R01 DA15612).
Role of the Sponsors: The funding agencies had no part in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.