HIV-1 infection and AIDS remain one of the major global health challenges, with an estimated 2.7 million new infections alone in 2010.

Two innovative strategies have recently been proposed to curb HIV-1 spread:

1) Pre-exposure prophylaxis (PrEP) aims to protect uninfected individuals ‘at risk’ by decreasing the probability of infection upon virus exposure.

2) ‘Treatment for prevention’ aims to put infected individuals on therapy as early as possible. This rapidly decreases virus burden within the infected individual and thereby reduces the number of viruses per transmission event, which in turn reduces the probability of infecting the exposed partner. While ‘treatment for prevention’, was shown in 2011 to be highly efficient (preventing 96% infections relative to standard of care), its efficacy in preventing disease spread is intimately connected with viral suppression. However, HIV inevitably develops drug resistance upon treatment, which leads to virus rebound and nullifies the effect of ‘treatment for prevention’ for the time it remains unrecognized. In resource-constraint countries, where the potential for ‘treatment for prevention’ is maximal, fewer drugs are available and infrastructure and diagnostic facilities are less developed, constituting a risk for unrecognized drug resistance break-through and resistance spread.

We investigated optimal treatment- and diagnostic strategies against HIV-1, in terms of durable viral suppression, patient health and by economic means in resource-rich and resource-poor settings: We first developed a novel control framework (‘Markov decision processes with rare state observation’, RO-MDP) that allows patient-specific optimal treatment after diagnostic testing in the context of markovian disease dynamics. For each disease state, it computes the optimal treatment and the next time of medical examination, minimizing viral burden as well as treatment- and diagnostic costs from a national economic perspective. We then analyzed and compared the developed model with an open-loop switched system, which allows pro-active treatment switching in a cross-sectional manner without patient examination. Both frameworks were applied to a coarse-grained stochastic model of within-host HIV dynamics in the context of resource-rich and -poor settings, and monetary as well as patient-specific outcomes were assessed.

We discovered, based on the RO-MDP framework, that available drugs may not be utilized efficiently in resource-poor settings due to exorbitant diagnostic costs. Furthermore, we suspected that allowing treatment change only after diagnostic confirmation of treatment failure (i.e. some time after drug resistance has occurred) may limit future treatment options for the patient. In contrast, pro-active treatment switching without patient examination, as modelled by the open-loop switched system, yielded better outcomes in the resource-poor setting in terms of monetary costs and viral suppression.

While treatment is highly subsidized in resource-poor settings, the same is not true for diagnostics, which precludes a patient-specific, diagnostics-driven optimal use of available therapies. Our models predict that cross-sectional approaches may be used instead. The developed control framework is applicable to many medical phenomena. Further developments may improve its applicability to even more complex biomedical processes.