PrEP as a feature in the optimal landscape of combination HIV prevention in sub-Saharan Africa

INTRODUCTION: The new WHO guidelines recommend offering pre-exposure prophylaxis (PrEP) to people who are at substantial risk of HIV infection. However, where PrEP should be prioritised, and for which population groups, remains an open question. The HIV landscape in sub-Saharan Africa features limited prevention resources, multiple options for achieving cost saving, and epidemic heterogeneity. This paper examines what role PrEP should play in optimal prevention in this complex and dynamic landscape. METHODS: We use a model that was previously developed to capture subnational HIV transmission in sub-Saharan Africa. With this model, we can consider how prevention funds could be distributed across and within countries throughout sub-Saharan Africa to enable optimal HIV prevention (that is, avert the greatest number of infections for the lowest cost). Here, we focus on PrEP to elucidate where, and to whom, it would optimally be offered in portfolios of interventions (alongside voluntary medical male circumcision, treatment as prevention, and behaviour change communication). Over a range of continental expenditure levels, we use our model to explore prevention patterns that incorporate PrEP, exclude PrEP, or implement PrEP according to a fixed incidence threshold. RESULTS: At low-to-moderate levels of total prevention expenditure, we find that the optimal intervention portfolios would include PrEP in only a few regions and primarily for female sex workers (FSW). Prioritisation of PrEP would expand with increasing total expenditure, such that the optimal prevention portfolios would offer PrEP in more subnational regions and increasingly for men who have sex with men (MSM) and the lower incidence general population. The marginal benefit of including PrEP among the available interventions increases with overall expenditure by up to 14% (relative to excluding PrEP). The minimum baseline incidence for the optimal offer of PrEP declines for all population groups as expenditure increases. We find that using a fixed incidence benchmark to guide PrEP decisions would incur considerable losses in impact (up to 7%) compared with an approach that uses PrEP more flexibly in light of prevailing budget conditions. CONCLUSIONS: Our findings suggest that, for an optimal distribution of prevention resources, choices of whether to implement PrEP in subnational regions should depend on the scope for impact of other possible interventions, local incidence in population groups, and total resources available. If prevention funding were to become restricted in the future, it may be suboptimal to use PrEP according to a fixed incidence benchmark, and other prevention modalities may be more cost-effective. In contrast, expansions in funding could permit PrEP to be used to its full potential in epidemiologically driven prevention portfolios and thereby enable a more cost-effective HIV response across Africa.