If a preventive vaccine was combined with ART using DHHS criterion for initiation of treatment, the combined impact on HIV incidence is potentially substantial (Table 3). A preventive vaccine with only 25% efficacy, combined with treatment which covered 75% of the population at risk, could reduce R0 to 0.82, resulting in abatement of the epidemic. Vaccines of greater efficacy could achieve a comparable impact on HIV incidence and R0 at even lower levels of population coverage. The combination of a preventive vaccine with ART (Fig. 4) has a somewhat lesser impact on the HIV-infected population and HIV prevalence than a vaccine alone (Fig. 3), because of ART-associated improved survival of HIV infected persons.
To assess behavioral disinhibition, we assumed an increase in the numbers of sexual partners among persons receiving treatment. If patients were treated according to DHHS guidelines, and disinhibition led to an increase in the number of sexual partners among those on therapy, the impact of treatment on HIV incidence would be attenuated, and it would be impossible to reduce R0 even close to 1.0 (Fig. 5). If availability of treatment resulted in generalized disinhibition with a 25% increased number of sexual partners in the whole population (treated and untreated), then HIV incidence would rise markedly and the public health benefits of therapy would be lost (results not shown). Analogous simulations of disinhibition among recipients of a preventive vaccine with 50% efficacy suggest that disinhibition could counteract the public health impact, and even result in an increased HIV incidence (Fig. 5). The effects of disinhibition among vaccine recipients on HIV incidence is more marked than the impact of disinhibition among HIV-infected persons receiving therapy, because the population of HIV-negative vaccinees is much larger than the population of HIV-positive persons eligible for treatment.
This simulation model based on empirical data from Rakai required minimal underlying assumptions, and model generated estimates of HIV incidence and transmission probabilities per coital act were similar to the empirically observed estimates [10,17], suggesting that the model closely approximated the HIV epidemic in the Rakai population. Simulations of the effects of ART based on data derived from the two US programs [20,21], also allowed realistic estimates of the effects on HIV incidence that might accrue from reductions of HIV viral load and continuity of treatment. It was assumed that the experience of these US clinics would reasonably reflect the optimal circumstances that might pertain in the Rakai population, if a high quality antiretroviral program were to become available and sustainable.
Of the two scenarios for initiation of ART, the most appropriate and feasible would be adherence to criteria for initiation of treatment compatible with the current DHHS guidelines . As shown in Table 2, an ART program, comparable to that at the Johns Hopkins Clinic, could reduce but not interrupt transmission in Rakai. However, it is unlikely that such high standards of care are feasible in this resource-poor rural setting. A treatment program with viral load reductions comparable to those observed in the WIHS study would, in all likelihood, not be sufficient to control the epidemic. Among HIV infected persons in the Rakai cohort, 20% had HIV viral loads > 55 000 copies/ml, and would thus be eligible for initiation of therapy under DHHS guidelines. It is estimated that 28.3 million people in sub-Saharan Africa are currently HIV infected , and assuming viral load distributions similar to those in Rakai, approximately 5.7 million HIV-positive persons would be currently eligible for ART. If, as suggested by these simulation, feasible treatment is insufficient to reduce R0 to < 1.0, and has minimal effects on the number of HIV-infected persons (Fig. 2), then the numbers of patients requiring treatment will increase substantially in the future, because currently treated persons will survive longer, persons presently in the latent stage of infection will progress to immunodeficiency, and continued HIV transmission will result in new infections which will progress to a disease stage where therapy is indicated. Thus, the need for treatment and the financial and health service burdens of therapy will increase substantially for the foreseeable future.
Although less stringent criteria for initiation of treatment could potentially control the epidemic (Table 2), and avoid the costs and difficulty of assessing viral load or CD4 cell counts to determine treatment eligibility, such widespread use of antiretroviral drugs would be extremely expensive and logistically difficult. Moreover, universal provision of therapy would be ethically questionable, because early treatment is unlikely to provide significant benefit to patients, and would expose them to risks of drug toxicity and selection of resistant virus [26–30], thus depriving the infected individual of effective therapeutic options during late stage disease . We conclude, therefore, that ART alone is unlikely to curtail the growth of the HIV epidemic in sub-Saharan Africa.
There is a concern that poor compliance with use of HAART may result in treatment failures due to selection of resistant virus [19,29,30]. We did not formally simulate such a reduction in treatment efficacy because we cannot predict compliance in African populations. However, the majority of patients who discontinued therapy at the Johns Hopkins Clinic, did so as a result drug resistance and the absence of alternative therapeutic options . Thus, the empirical data on discontinuation of treatment incorporated in our models implicitly accounts for emergence of resistant virus or toxicity as the main reasons for therapeutic termination.
The simulations of the potential impact of an HIV vaccine lack empirical data on efficacy, since no phase III trial had been completed at the time of waiting . We therefore assumed a range of vaccine efficacies from a weakly protective (25% efficacy) to a highly protective vaccine (75% efficacy), based on animal studies or preliminary trial data. A vaccine of low preventive efficacy could not, by itself, control the epidemic (Table 3). However, if there was concomitant provision of ART to reduce viral load in HIV-positive individuals, there could be substantial reductions in HIV incidence and R0 could be reduced well below 1.0, even with incomplete coverage (Table 3). Therefore, consideration should be given to combined strategies of treatment and vaccination, even if future vaccines have relatively low preventive efficacy. As illustrated in Fig. 3, a 50% efficacious vaccine could markedly reduce the numbers of HIV-infected persons in the population, and this will diminish the need for ART over time, thus potentially saving substantial resources. More efficacious vaccines would obviously have the maximum impact, but unfortunately no vaccine of proven efficacy is currently available.
Other models have simulated the impact of HAART on HIV incidence among homosexual men in San Francisco , and Australia , and suggest that treatment could have a substantial impact on incident HIV, but these models did not initiate treatment using current DHHS guidelines. There is growing evidence that homosexual men in Europe and North America are adopting higher risk practices since HAART became widely available, and that such disinhibition is occurring both among the HIV-positive treated patients and among HIV-negative individuals [24,25]. If behavioral disinhibition were to occur in response to treatment or a vaccine in the Rakai population, it could markedly offset the public health benefits (Fig. 5). Moreover, the adverse effects of disinhibition are dependent on the size of the population adopting high-risk behaviors. For example, disinhibition among recipients of ART results in moderate increases in incidence because the number of persons receiving ART constitute a small proportion of the total population. However, if disinhibition occurred among vaccine recipients, who constitute a potentially large proportion of the total population, then the increase in HIV incidence with higher risk behaviors is much more marked and results in rising HIV incidence, abolishing all benefits of vaccination. Thus, the promotion of safe sex is critical to the public health benefits that might accrue from ART or vaccines.
In conclusion, improved access to ART is needed for eligible HIV-infected persons in developing and developed countries to improve survival and quality of life. However, there is also a need to consider the societal benefits of therapy in terms of control of the HIV epidemic. ART initiated only for persons with advanced disease is unlikely to reduce HIV transmission sufficiently to control the epidemic. Moreover, benefits of therapy could be markedly offset by behavioral disinhibition. A preventive vaccine, were it to become available, offers the best hope of controlling the epidemic in the long-term, particularly if there was a concurrent ART program, or if the vaccine itself could reduce viral load in HIV-infected persons.
Sponsorship: Supported by grants RO1 AI34826 and RO1 AI3426S, National Institute of Allergy and Infectious Diseases, grant 5P30HD06826, National Institute of Child Health and Development, grant 5D43TW00010 from the Fogarty Foundation, US National Institutes of Health, and the World Bank STI Project, Uganda. The WIHS is funded by the National Institute of Allergy and Infectious Diseases, with supplemental funding from the National Cancer Institute, the National Institute of Child Health & Human Development, the National Institute on Drug Abuse, and the National Institute of Craniofacial and Dental Research. U01-AI-35004, U01-AI-31834, U01-AI-34994, U01-AI-34989, U01-HD-32632 (NICHD), U01-AI-34993, U01-AI-42590, M01-RR00079, M01-RR00083.
1. Steinbrook R. Providing antiretroviral therapy for HIV infection.
Editorial. New Engl J Med
2. UK NGO AIDS Consortium Working Group Access to Treatment for HIV in Developing Countries. Access to treatment for HIV in developing countries; statement from international seminar on access to treatment for HIV in developing countries, London, June 5 and 6, 1998.Lancet
3. Binswanger HP. Public health: HIV/AIDS treatment for millions.Science
4. Lee L, Karon JM, Selik R, Neal JJ, Fleming PL.Survival after AIDS diagnosis in adolescents and adults during the treatment era, United States, 1984–1997.JAMA
5. Veugelers PJ, Cornelisse PGA, Craib KJP, Hogg MRS, Strathdee SA, Montaner JS, et al
. Models of survival in HIV infection and their use in the quantification of treatment benefits.Am J Epidemiol
6. Weidle PJ, Holmberg SD, DeCock KM. Changes in HIV and AIDS epidemiology from new generation antiretroviral therapy.AIDS
7. Freedberg KA, Losing E, Weinstein MC, Paltiel D, Cohen CJ, Seage GR, et al
. The cost effectiveness of combination antiretroviral therapy for HIV disease.New Engl J Med
8. Bozzette SA, Joyce G, McCaffrey DF, Leibowitz AA, Morton SC, Berry SH, et al.
, for the HIV Cost and Services Utilization Study Consortium. Expenditures for the care of HIV-infected patients in the era of highly active antiretroviral therapy.New Engl J Med
9. Quinn TC, Wawer MJ, Sewankambo NK, Serwadda D, Li C, Wabwire-Mangen F, et al.
, Viral Load and heterosexual transmission of Human Immunodeficiency Virus Type 1.New Engl J Med
10. Gray RH, Wawer MJ, Brookmeyer R, Sewankambo NK, Serwadda D, Wabwire-Mangen F, et al
. Probability of HIV-1 transmission per coital act in monogamous, heterosexual, HIV-1 discordant couples in Rakai, Uganda.Lancet
11. Blower SM, Bershengorn HB, Grant RM. A tale of two futures: HIV and antiretroviral therapy in San Francisco.Science
12. Law MG, Prestage G, Grulich A, Van de Ven P, Kippax S. Modeling the effect of combination antiretroviral treatments of HIV incidence.AIDS
13. Garnett GP. An introduction to mathematical models in sexually transmitted disease epidemiology.Sex Trans Infect
14. Morgan D, Mahe C, Mayanja B, Okongo JM, Lubega R, Whitworth JAG. HIV-1 infection in rural Africa: is there a difference in median time to AIDS and survival compared with that in industrialized countries?AIDS
15. Anderson RM, Garnett GP. Low-efficacy HIV vaccines: potential for community-based intervention programmes.Lancet
16. Wawer MJ, Sewankambo NK, Serwadda D, Quinn TC, Paxton LA, Kiwanuka N, et al.Control of sexually transmitted diseases for AIDS prevention in Uganda: a randomized community trial.Lancet
17. Wawer MJ, Gray RH, Sewankambo NK, Serwadda D, Paxton LA, Berkley S, et al.A randomized, community trial of intensive sexually transmitted disease control for AIDS prevention Rakai, Uganda.AIDS
18. Davidson AC, Hinkley DV. Bootstrap Methods and their Application
. New York: Cambridge University Press; 1997.
20. Moore RD. Understanding the clinical and economic outcomes of HIV therapy: the Johns Hopkins HIV clinical practice cohort
. J Acquir Immune Def Syndr
21. Barkan SE, Melnick SL, Preston-Martin S, Weber K, Kalish LA, Miotti P, et al.The Women's Interagency HIV Study.Epidemiology
22. Hatano H, Vogel S, Yoder C, Metcalf JA, Dewar R, Davey RT Jr, et al
. Pre-HAART HIV burden approximates post-HAART viral level following interruption of therapy in patients with sustained viral suppression.AIDS
23. Esparza J, Bhamarapravati N. Accelerating the development and future availability of HIV-1 vaccines: why, when, where, and how?Lancet
24. Miller M, Meyer L, Boufassa F, Persoz A, Sarr A, Robain M, et al.Sexual behavior changes and protease inhibitor therapy.AIDS
25. Katz MH, Schwarcz SK, Kellog TA, Klausner JD, Dilley JW, Gibson S, McFarland W. Impact of highly active antiretroviral treatment on HIV seroincidence among men who have sex with men: San Francisco.Am J Public Health
26. Arriaga EE and Associates. Population Analysis with Microcomputers.
Washington, DC: U.S. Bureau of the Census. International Programs Center; 1994.
27. Sewankambo NK, Gray RH, Ahmad S, Serwadda D, Wabwire-Mangen F, Nalugoda F, et al.Mortality associated with HIV infection in rural Rakai District, Uganda.AIDS
28. UNAIDS/WHO AIDS Epidemic update December 2002. Joint United Nations Programme on HIV/AIDS (UNAIDS) World Health Organization (WHO), Geneva, Switzerland 2002:1–29.
29. Carr A, Cooper DA. Adverse effects of antiretroviral therapy.Lancet
30. Gallant JE. Strategies for long-term success in the treatment of HIV infection.JAMA
Rakai Project Group Kalisizo, Uganda
N. Kiwanuka, G. Kigozi, F. Nalugoda.
Data by the Women's Interagency HIV Study (WIHS) Collaborative Study Group with centers (Principal Investigators) at New York City/Bronx Consortium (K. Anastos); Brooklyn, NY (H. Minkoff); Washington DC Metropolitan Consortium (M. Young); The Connie Wofsy Study Consortium of Northern California (Ruth Greenblatt, Herminia Palacio); Los Angeles County/Southern California Consortium (A. Levine); Chicago Consortium (M. Cohen); Data Coordinating Center (A. Munoz, S. J. Gange).
Parameters for model estimates of the probabilities of HIV transmission per coital act
Rakai data from HIV-discordant couples were used to estimate the model parameters, with frequencies of intercourse reported by HIV-infected persons in the population as an offset term . The analyses showed that the HIV-1 viral load in the HIV-positive partner was the main determinant of transmission risk, and the probabilities of transmission per coital act decreased with older age, and differed by gender. The likelihood ratio test of nested models indicated an improved fit with linear, quadratic and cubic log10 viral load terms (χ2, 7.01; P = 0.0081). The parameter estimates for the latter model were as follows: log (−log [1−γ]) = −97.5 + 61.1[log10 viral load] + 13.7 [log10 viral load]2 + 1.02 [log10 viral load]3 + 0.35 I [age 25–29] + 0.59 I [age 30–34] − 0.65 I [age 35–59] + 0.56 I [female] where I is an indicator function. It was assumed that the probabilities of transmission per coital act derived from studies of HIV-discordant couples applied to all HIV-infected persons, and that the distribution of viral loads in the discordant couples were representative of those in the general population of HIV-infected persons.
Variance of simulation models
The simulated probability of transmission per coital act was 0.0012 (SD, 0.000055, coefficient of variation 4.6%), and the mean HIV incidence was 1.57 per 100 PY (SD, 0.061, coefficient of variation 3.9%). Because of this precision in our estimates we have not included variance statistics in the results presented.
Data on the impact of anitretroviral therapy on HIV viral load
The effects of therapy on HIV viral load were determined from information on 492 patients attending the HIV Clinic at Johns Hopkins Hospital  and 999 women enrolled in the Women's Interagency HIV Study (WIHS) . The HIV Clinic at Johns Hopkins Hospital specializes in management of HIV infection; 79% of eligible patients received (HAART) between 1996 and 2000 . Among 275 treatment-naive patients with viral loads > 55 000 copies/ml attending the Johns Hopkins Clinic, the mean viral load declined from 5.32 to 3.06 log10 copies/ml, a 43.6% reduction within 1 month of therapy, and remained low thereafter (3.21 log10 copies/ml at 12 months). We therefore assumed a linear decrease in HIV viral load by 44% within 1 month of initiating therapy and an approximately constant viral load while on therapy. The participants in the WIHS study obtained care from a variety of service providers, 35% had received treatment prior to enrollment and 65.3% received HAART during follow up . In 316 women with initial viral loads > 55 000 copies/ml in the WIHS population, the mean log10 HIV viral load declined from 5.23 to 3.82 copies/ml within 3 months of initiation of HAART (a 27.0% reduction), and the viral load remained relatively stable thereafter (3.75 log10 copies/ml at 15 months). The differences in the viral load response to HAART in these two clinic populations probably reflects the more specialized management of the Johns Hopkins Clinic patients, compared with patients receiving care from diverse service providers in the WIHS data. These proportionate declines in HIV viral load were used to adjust the HIV-1 viral load distributions in the simulated populations receiving treatment.
Discontinuation of HAART, defined as interruption of therapy for more than 6 months for all reasons (drug toxicity, treatment failure and personal choice) was available from both studies for up to 3 years. At the Johns Hopkins Clinic, the mean discontinuation rate per month was 1.8%, and the cumulative proportion discontinuing treatment was 56.2% over 3 years. In the WIHS data, the mean discontinuation rate was 1.6% per month, with a cumulative discontinuation of therapy of 55.2% at 3 years. Thus, sensitivity analyses were conducted to assess the effects of better compliance assuming no terminations over 3 years, or half the observed Johns Hopkins termination rates (i.e., 0.9% per month). The effects of poorer compliance were assessed by assuming monthly discontinuation rates of 3.6% and 7.2% (i.e., a two- and fourfold increase in discontinuation rates). The reductions in viral load and continuity of treatment can be considered as measures of use-effectiveness of HAART in the USA, and these were applied to persons entering therapy in the model. It was assumed that the viral load returned to baseline levels within 1 month of termination of treatment , recursively returning terminating individuals to their baseline transmission risk. Persons continuing therapy could contribute to ongoing transmission, depending on their estimated HIV-1 viral loads. Treatment of acute seroconverters, as recommended by DHHS  was not simulated, because it is not feasible to identify newly infected persons during or shortly after the acute seroconversion illness in the rural Rakai setting. Seroconverters could become eligible for treatment if they met the criterion for initiation of therapy specified by the models.
Component projection model
In the model, HIV-negative individuals could remain uninfected when moving to next 5-year age group, become HIV infected within a 5-year age interval, or be removed from the population by death. For the HIV-positive populations, persons could progress to the next older 5-year age group with or without receiving treatment (depending on eligibility and assumed treatment coverage levels). Among those receiving therapy, we projected continuation or termination of treatment over 5-year age intervals, using data from the Johns Hopkins Clinic. The cumulative 5-year continuation of therapy was 34%, implying that 66% of patients would terminate therapy either because of drug toxicity or treatment failure, and would return to the untreated HIV-positive population or die. Five-year age-specific mortality rates by HIV status were obtained from Rakai data  to calculate 5-year survival probabilities using exponential assumptions. For treated patients, we assumed that the relative mortality risk was 0.51 compared to untreated patients, based on reductions in mortality due to AIDS observed in the USA . HIV-positive persons who discontinued treatment were assumed to have the same mortality rates as untreated individuals. Population growth rates were derived from Rakai data and we estimate that the population aged 15–19 years will increase by 3% per annum. HIV incidence estimated from the stochastic models was used as an input to calculate the HIV incidence over 5-year age intervals for the HIV-negative population. The age-specific probabilities of seroconversion over 5 years were obtained from the model derived incidence rates by an exponential distribution (seroconversion probability = 1−exp (−incidences within 5-year age group), allowing for declines in incidence, but assuming constant rate ratios of age-specific incidence rates. The proportional changes in the HIV-infected population estimated from the component projection model are presented.
Sensitivity analyses for ART
Sensitivity analyses examined lower and higher discontinuation rates than those observed in the Johns Hopkins Clinic. Using DHHS guidelines for therapy and the Johns Hopkins clinic data, if no patients discontinued treatment, the incidence would be 1.15 per 100 PY, approximately 10% lower than the simulations for complete coverage shown in Table 2. Thus, improved continuation of therapy beyond that currently observed at Johns Hopkins does not appear to yield substantially greater public health impact. However, if discontinuation rates were substantially higher than those observed in the Johns Hopkins Clinic, the impact on HIV incidence would be markedly attenuated. For example, if the termination rate was four-fold higher (i.e., 7.2% per month), HIV incidence would be approximately1.45 per 100 PY with complete coverage of eligible patients, which is only 7.6% less than the estimated incidence in the absence of treatment (1.57 per 100 PY). Thus, poor continuity of therapy could undermine the public health impact of HAART.