University of California, San Francisco Logo

University of California, San Francisco | About UCSF | Search UCSF | UCSF Medical Center

Predicting the impact of a partially effective HIV vaccine and subsequent risk behavior change on the heterosexual HIV epidemic in low- and middle-income countries: A South African example
Global Health Sciences Literature Digest
Published January 07, 2008
Journal Article

Andersson KM, Owens DK, Vardas E, Gray GE, McIntyre JA, Paltiel AD. Predicting the impact of a partially effective HIV vaccine and subsequent risk behavior change on the heterosexual HIV epidemic in low- and middle-income countries: A South African example. J Acquir Immune Defic Syndr. 2007 Sep 1;46(1):78-90.


To simulate with a mathematical model the impact of various partially effective preventive HIV vaccination scenarios and subsequent changes in risk behavior in a population at high risk for heterosexually transmitted HIV.

Study Design

Mathematical simulations using a deterministic differential equations model to explore how different vaccine efficacies combined with changes in post-vaccination condom use would affect HIV prevalence and total HIV infections over a 10-year period.


The hypothetical setting was the urban township population of Soweto, South Africa, used to represent a low- and middle-income country. Input parameters came from published data sources for mean age, life expectancy, HIV prevalence, and number of sexual partners in this population. Additional input parameters on per-partnership probability of HIV transmission in heterosexual partnerships and the lengths of disease stages, as well as the time from AIDS development to death, came from published Ugandan data.


The 823,000 adults >16 years in Soweto, South Africa were used as the hypothetical population receiving an HIV vaccine. This population had an estimated HIV prevalence of 11.6% in men and 20.0% in women (between ages 15 and 49), a mean age of 25.1 years, a life expectancy of 60.8 years from birth in the absence of deaths attributable to AIDS, and all adults >16 years old were assumed sexually active. The population was defined by gender, disease stage, and vaccination status to simulate the process of heterosexual HIV transmission and disease progression through the following six categories: 1. unvaccinated HIV-negative; 2. vaccinated HIV-negative; 3. unvaccinated asymptomatic HIV-positive; 4. vaccinated asymptomatic HIV-positive; 5. symptomatic HIV-positive; and 6. AIDS. There were separate states for men and women, for a total of 12 categories.


Mathematical simulations were run, in which rates of movement of individuals were specified by a set of 12 deterministic differential equations that govern allowable transitions between the 12 population subgroups. The inputs for these transition probabilities came from the published data from Uganda and South Africa. The model considered a population of sexually active adults who enter the simulation upon reaching sexual maturity and can leave the simulation because of death from non-AIDS-related or AIDS-related causes. The model assumed 75% vaccination coverage, 100% take (every vaccinated person has an immune response), a 10-year duration of protection, exclusive male negotiation of condom use, and current condom use of 50%. The model varied the number of sexual partners a man or woman has and the probability of HIV transmission within those sexual partnerships according to disease stage. The simulations were conducted as if a vaccine was currently available and a vaccination program was implemented for a period of 10 years. The effects of various vaccination scenarios were evaluated by examining the total number of HIV infections averted and the change in HIV prevalence in different model outputs. A base-case scenario was simulated, in which a vaccination program was implemented using a vaccine with 40% efficacy and assuming no post-vaccination change in risk behavior represented solely by male-negotiated condom use. Then, model predictions from the base-case scenario were compared to other vaccination scenarios with varying vaccine efficacies and post-vaccination changes in male condom use. A 40% effective vaccine means vaccinated HIV-negative individuals are protected from HIV infection in 40% of partnerships with infected heterosexual partners. Baseline condom use was estimated to be 50% in all sexual partnerships with a 14% condom failure rate. Subsequent analyses utilized values of 25% to 75% condom use to explore a wide range of potential changes in post-vaccination risk behavior.

Three sensitivity analyses were performed: 1. By varying vaccine efficacy; 2. By reducing vaccine coverage from 75% to 50% of the population; and 3. By varying the male-to-female (MTF) and female-to-male (FTM) infectivity probabilities.

Primary Outcomes

Primary outcomes were HIV prevalence and HIV infections averted 10 years after vaccine initiation.


In the absence of HIV prevention interventions in Soweto, the model predicted that HIV prevalence would increase over the next 50 years, from 16% currently to 26%, with 743,000 new HIV infections. Within the next 10 years alone, simulations predicted that the overall population prevalence would rise to 20% and that 161,000 new HIV infections would occur. For vaccination scenarios with no change in post-vaccination male condom use (50% use), the base-case vaccination scenario, using 40% effective vaccine, would prevent 61,000 new HIV infections and reduce HIV prevalence from 20% to 13%. In comparison, a 100% effective vaccine would avert 128,000 infections over 10 years and decrease HIV prevalence to 7%, whereas a 20% effective vaccine would avert 32,000 infections and decrease the HIV prevalence to 17%.

With an increase to 75% post-vaccination condom use and a 40% effective vaccine, 75,000 infections would be averted, resulting in 12% HIV prevalence; a 40% effective vaccine with a decrease in condom use to 25% would avert only 46,000 new infections and reduce the HIV prevalence to 15%. The combination of a less effective vaccine and a decrease in condom use could result in an increase in HIV infections. A 20% effective vaccine with 25% condom use would cause an additional 14,000 infections beyond what would be expected without any vaccination campaign, and would increase the HIV prevalence to 22%.

By estimating all combinations of vaccine efficacy and changes in post-vaccination condom-use scenarios that would prevent the same 61,000 HIV infections as in the base-case scenario, the following results were found: a. a vaccine only 40% effective would be less effective than a program that could increase condom use to at least 80%; on the other hand, b. any HIV vaccine that was more than 66% effective would prevent HIV infections even if all individuals no longer used condoms after being vaccinated. Likewise, by estimating which combinations of vaccine efficacy and changes in condom use would result in a net decrease in the number of HIV infections, it was found that, as the efficacy of the vaccine increased, HIV infections were still reduced even with decreasing condom use. It was found that a vaccine with an efficacy of at least 43% would be beneficial in terms of infections prevented, regardless of changes in condom use.


The authors conclude that even modestly effective HIV vaccines with 30% to 40% efficacy can have enormous benefits in terms of HIV infections averted and decreased HIV prevalence; however, vaccination must be coupled with behavior change programs.

Quality Rating

There is no quality rating for mathematical simulation models, but this study based the values of its parameters on geographically relevant data from published sources. The simulations were well thought through and provided an overall picture of the possibilities of partially preventative vaccines, sexual risk reduction, and HIV prevalence estimates for communities such as the township of Soweto. The authors do not show the data from their sensitivity analyses, but reported that their qualitative conclusions were not changed. The model is limited by only including an ARV-naive population, and only including the impact of a preventive vaccine, rather than disease-modifying vaccines that may decrease viral load and disease progression, and thus HIV transmission, in those already infected. Future analysis would benefit from the incorporation of migration and heterogeneous risk behavior groups, use of antiretroviral treatment, and vaccination scenarios varying other parameters, such as vaccine take and duration of protection.

In Context

Previous models in the South African setting had examined the impact of a preventive HIV vaccine in a cohort of adolescent girls with a Markov modeli and investigated the impact of disease-modifying HIV vaccines in a homogenous population with a dynamic transmission model.(2) Using a dynamic transmission model, this model goes a step further to simulate the impact of preventive HIV vaccines and changes in condom use on heterosexual HIV transmission in South Africa. This is the first time that differential condom use between men and women had been incorporated in a model. Control of condom use was restricted to males to reflect the cultural and economic barriers that limit female condom negotiation in South Africa. The findings of this model are consistent with previous models on the effects of preventive HIV vaccines(3,4,5,6,7) and their sensitivity to changes in risk behavior.(8,9,10)

Programmatic Implications

Risk behavior assessment must remain an important component of HIV vaccine clinical trials to provide a means of predicting changes in risk behavior after a mass vaccination campaign. Programs to reduce risk behavior are important components of successful vaccination campaigns, particularly when lower efficacy vaccines are used. However, the reductions in HIV prevalence and new HIV infections that can results from using partially effective vaccines are substantial, even in the presence of increased risk behavior. To that end, development of preventive vaccines should remain a high priority, despite concerns that the vaccines may have only moderate efficacy.


  1. Amirfar S, Hollenberg JP, Abdool Karim SS. Modeling the impact of a partially effective HIV vaccine on HIV infection and death among women and infants in South Africa. J Acquir Immune Defic Syndr. 2006;43:219-225.
  2. Blower SM, Bodine EN, Grovit-Ferbas K. Predicting the potential public health impact of disease-modifying HIV vaccines in South Africa: the problem of subtypes. Curr Drug Targets Infect Disord. 2005;5:179-192.
  3. Anderson RM, Garnett GP. Low-efficacy HIV vaccines: potential for community-based intervention programmes. Lancet 1996;348:1010-1013.
  4. Anderson RM, Swinton J, Garnett GP. Potential impact of low efficacy HIV-1 vaccines in populations with high rates of infection. Proc Biol Sci 1995;261:147-151.
  5. Owens DK, Edwards DM, Shachter RD. Population effects of preventive and therapeutic HIV vaccines in early- and late-stage epidemics. AIDS 1998;12:1057-1066.
  6. Blower SM, Koelle K, Kirschner DE, et al. Live attenuated HIV vaccines: predicting the tradeoff between efficacy and safety. Proc Natl Acad Sci USA. 2001;98:3618-3623.
  7. Anderson RM, Gupta S, May RM. Potential of community-wide chemotherapy or immunotherapy to control the spread of HIV-1. Nature 1991;350:356-359.
  8. Gray RH, Li X, Wawer MJ, et al. Stochastic simulation of the impact of antiretroviral therapy and HIV vaccines on HIV transmission; Rakai, Uganda. AIDS. 2003;17:1941-1951.
  9. Blower SM, McLean AR. Prophylactic vaccines, risk behavior change, and the probability of eradicating HIV in San Francisco. Science 1994; 265:1451-1454.
  10. Edwards DM, Shachter RD, Owens DK. A dynamic HIV-transmission model for evaluating the costs and benefits of vaccine programs. Interfaces 1998;28:144-166.