Cost-effectiveness of pre-exposure prophylaxis targeted to high-risk serodiscordant couples as a bridge to sustained ART use in Kampala, Uganda

Introduction Despite scale-up of antiretroviral therapy (ART) for treating HIV-positive persons, HIV incidence remains elevated among those at high risk such as persons in serodiscordant partnerships. Antiretrovirals taken by HIV-negative persons as pre-exposure prophylaxis (PrEP) has the potential to avert infections in individuals in serodiscordant partnerships. Evaluating the cost-effectiveness of implementing time-limited PrEP as a short-term bridge during the first six months of ART for the HIV-positive partner to prevent HIV transmission compared to increasing ART coverage is crucial to informing policy-makers considering PrEP implementation. Methods To estimate the real world delivery costs of PrEP, we conducted micro-costing and time and motion analyses in an open-label prospective study of PrEP and ART delivery targeted to high-risk serodiscordant couples in Uganda (the Partners Demonstration Project). The cost (in USD, in 2012) of PrEP and ART for serodiscordant couples was assessed, with and without research components, in the study setting. Using Ministry of Health costs, the cost of PrEP and ART provision within a government programme was estimated, as was the cost of providing PrEP in addition to ART. We parameterized an HIV transmission model to estimate the health and economic impacts of 1) PrEP and ART targeted to high-risk serodiscordant couples in the context of current ART use and 2) increasing ART coverage to 55% of HIV-positive persons with CD4 ≤500 cells/µL without PrEP. The incremental cost-effectiveness ratios (ICERs) per HIV infection and disability-adjusted life year (DALY) averted were calculated over 10 years. Results The annual cost of PrEP and ART delivery for serodiscordant couples was $1058 per couple in the study setting and $453 in the government setting. The portion of the programme cost due to PrEP was $408 and $92 per couple per year in the study and government settings, respectively. Over 10 years, a programme of PrEP and ART for high-risk serodiscordant couples was projected to avert 43% of HIV infections compared to current practice with an ICER of $1340 per infection averted. This was comparable to ART expansion alone, which would avert 37% of infections with an ICER of $1452. Conclusions Using Uganda's gross domestic product per capita of $1681 as a threshold, PrEP and ART for high-risk persons have the potential for synergistic action and are cost-effective in preventing HIV infections in high prevalence settings. The annual cost of PrEP in this programme is less than $100 per serodiscordant couple if implemented in public clinics.

The rate of progressing from CD4 state to The rate of progressing from VL state to The reduction in HIV transmission due to circumcision ( ), PrEP ( ), ART ( ), or condom use ( )

Demography:
At each iteration, the force of infection and the number of births are calculated and then used to evaluate the ODEs along with mortality and disease progression. The numbers of incident infections, HIV-related deaths, and individuals entering CD4≤200 cells/µL are also calculated to determine QALYs.

Births:
The number of births, , determines how many newborns enter the population of gender , disease state , sexual risk group , and circumcision status ( for uncircumcised; for circumcised males). For simplicity, we assume only neonatal circumcision (the circumcision level is increased over time such that 30% of males are circumcised by 2013, as currently observed in Kampala and shown in Figure S2 [2]), that infected births enter the acute stage, and that women age 15-49 give birth. Fertility rates are stratified by age and stage of disease. Births from uninfected mothers, , and from HIVpositive mothers, , are: HIV-negative births for uncircumcised males, , are: HIV-negative births for circumcised males, , are: HIV-negative births for females, , are: HIV-positive births for males and females, , are: The equation variables are: The proportion of individuals in age , gender , and treatment status ( , no treatment; , PrEP; , ART) that is born into sexual risk group The proportion of births from HIV-positive females that result in vertical transmission The proportion of HIV-negative males that is circumcised at birth The annual fertility rate for females by age and disease state Each birth is multiplied by 0.5 given an assumed gender ratio at birth of 1:1.

Mortality:
People leave the population due to death or aging past age 59. Mortality is represented by disease-induced mortality [3,4]. The background mortality rate is estimated to be the population mortality rate in 1990, prior to the generalized HIV epidemic.

Disease Transmission:
Disease transmission is governed by the force of infection, , which determines the number of people who are infected at each time-step.
The equation variables are: The number of partners from age and sexual risk group that an individual has per year The mixing matrix which describes the distribution of partners from each age and sexual risk group The probability of HIV transmission per partnership between an HIVpositive person of stage and HIV-negative person of risk group The overall force of infection for a specific age-group is the sum of the risk of acquiring HIV from all possible partners.

Mixing Matrix:
Using methods similar to other models, the mixing matrix, , describes patterns of sexual contact by calculating the proportion of one's sexual partners that come from a specific age and sexual-risk group [5]. Mixing patterns vary between random and assortative, as determined by the parameter .

Random mixing (
) is mixing proportional to the relative sizes of all compartments and this method is consistent for both random mixing by risk and by age. However, assortative mixing ( ) is among groups with similar characteristics and differs for mixing by risk and age.
Assortative mixing by risk ( ) is defined by the identity matrix , whereas assortative mixing by age ( ) is defined by an off-diagonal matrix . The off-diagonal pattern results in females of age being more likely to form partnerships with males of age , which is consistent with reports of such age discrepancies in sub-Saharan Africa [6,7]. Although this off-diagonal method results in some age groups having fewer than 100% of their partnerships, those age-groups are and , which contribute relatively little to overall HIV transmission. We assume that this tendency for age-gaps diminishes in 2005. Furthermore, and shift from random to assortative over the course of the simulation, given the consistent government campaigns against risky sexual behavior [2].

Per-Partnership Probability of Transmission:
The

Rate of Partner Change:
Data on sexual behavior and specifically, sexual contact rates, , are often subject to biases leading to contact rate data that, when assuming solely heterosexual contact, are inconsistent between males and females [8]. We account for this variability by using an adjusted contact rate, , which equilibrates the reported number of sexual partners by males and females [5]. The adjusted contact rate can be male-or female-driven, as determined by the parameter , where for male-driven, for female-driven, and when compromised equally. We assume given the lack of data to assume otherwise. The adjusted contact rate for females is: For males, the adjusted contact rate is: The discrepancy between the two populations, , is defined as:

Population Aging
To age the population, one-fifth of each compartment enters the next age group of corresponding gender, sexual risk, and disease state. When individuals age, they also change sexual risk; therefore, they redistribute to a set sexual-risk profile, , that varies by age, gender, and treatment status. All compartments, except for the youngest and oldest age-groups, experience influx from the prior age and efflux into the next age. The 0 to 4 age-group only receives influx through births while the 55 to 59 age-group exits the population rather than entering the next age. Therefore, each state has a second ODE that occurs at each time step: For For

II. Interventions
Enrollment is represented by , while dropout is represented by .

ART Treatment Enrollment:
Coverage of ART treatment for HIV-positive persons increases from 0% in 2000 to 44% for persons with CD4 ≤200 cells/µL in 2003 as previously observed in Uganda [9], then to 50% coverage for persons with CD4 ≤350 cells/µL in 2013 as observed in the Home HTC study [10].
Treatment coverage then changes depending on the specified coverage and scale-up period. ART treatment is assumed to reduce the likelihood of HIV transmission by 96% as suggested by recent studies ( 6), and persons on ART are expected to have the same life expectancy as HIV-negative persons of similar age and gender, and thus, are assumed not to be subject to HIV-associated mortality [4,[11][12][13][14]. The annual drop-out rate is 6%, which is equally likely for all individuals regardless of their HIV state prior to treatment. Individuals who drop out of ART return to the infected stages at the same proportion with which they enrolled.

Circumcision:
This model includes a background level of circumcision of 30% as currently observed in Kampala [2]. Several studies show that circumcised males have a 60% ( ) lower risk of acquiring HIV, but are not at a reduced risk of transmitting HIV [15][16][17]. Therefore, the model does no track the circumcision status of HIV-positive persons. The ODE for HIV-negative circumcised males is: Other models have studied the impact of circumcision in-depth to include wound healing periods and sexual activity [18,19]. However, this model assumes that circumcision is instantaneous.