March 7, 2002
The researchers describe a mathematical model of the impact of a partially effective AIDS vaccine in a population of IDUs in Bangkok. They constructed a dynamic compartmental, or subgroup, model for Bangkok's IDUs, adapted from previous simulations of the HIV epidemic in gay males. The model consisted of 14 population subgroups, varying by high- and low-risk injection behavior, and five levels of disease status (HIV-negative; HIV-positive, asymptomatic, unidentified; HIV-positive, asymptomatic, identified; HIV-positive, symptomatic; and AIDS). Two types of risk behavior change were analyzed: positive post-counseling behavior change and a false immunity-presumptive, negative change in behavior. HIV-positive individuals were not distinguished by vaccination status because researchers assumed it had no therapeutic effect.
Transmission parameters were based on previously developed models of injection drug use. Among IDUs, injection risk behavior and viral infectivity determine how likely an individual is to become infected with HIV. Injection risk behavior characteristics include contact rate, total number of injections per year, and the probability that a needle was shared before injection.
The analysis only examined transmission through shared needle use. "Restricted" mixing between high and low-risk groups with respect to sharing needles was assumed. Input data for the model were obtained from the published literature. Population- specific data were used where possible; in the absence of Bangkok-based research, estimates were taken from US research. For base-case analysis, key assumptions were made. First, 50 percent of the eligible population is vaccinated each year. Second, the vaccine provides complete protection for 75 percent, and no protection for 25 percent, of those vaccinated. Third, the duration of protection from vaccine is ten years on average, after which individuals are not re-vaccinated. Fourth, the annual probability of any risk behavior change is 5 percent. Finally, the initial HIV prevalence is 30 percent. The analysis simulated the course of the epidemic in a population of 36,600 male IDUs during both 40 and 150 years.
At a vaccine efficacy of 30 percent, HIV prevalence was 44 percent after 40 years and reached an equilibrium level of 42 percent. At a vaccine efficacy of 90 percent, equilibrium prevalence was 35 percent. Vaccination remained superior to no vaccination at all levels of expected efficacy.
Post-vaccine risk behavior changes were analyzed: positive post-counseling behavior change and a false immunity-presumptive, negative change in behavior. Sensitivity analysis showed that with a 75 percent efficacy, HIV prevalence is somewhat sensitive to these behavior changes. If 100 percent of high-risk individuals switched to low risk behavior in response to vaccine-related counseling, HIV prevalence decreased to 33 percent after 40 years, an improvement of 4 percent over base-case results.
Simultaneous, unfavorable change in contact rates, number of injections and an increase of per partner infectivity had a limited effect when vaccine efficacy was 75 percent. The beneficial effects of vaccination diminished slightly as unfavorable post-vaccination behavior increased, though vaccination was always superior to no vaccination. The benefit of vaccination with a low-efficacy vaccine can be mitigated significantly by unfavorable behavior changes.
Results suggested that the effectiveness of vaccination depends on the probability of vaccination-related behavior change only if vaccine efficacy is lower than the base-case assumption of 75 percent. The findings suggest that a low-efficacy AIDS vaccine is beneficial to IDUs in Bangkok and that this beneficial effect can be enhanced or moderated by risk behavior change. If IDU behavior is modifiable, the findings have significant policy and planning implications.
Journal of Acquired Immune Deficiency
02.01.02; Vol. 29; No. 2: P. 132-141; Elizabeth Bogard; Karen M. Kuntz