Background:  Progression of HIV infection and clinical ART treatment represent a complex system that influences disease outcome in patients. We proposed to develop a multi-state Markov model to describe the events of early HIV-infection and treatment initiation in a cohort of recently-infected, drug-naive HIV⁺ patients.

Methods:  We identified and included 195 ART-naive patients as part of a primary infection research study in this retrospective analysis at 120 days post-infection. Patients were categorized into 1 of 5 states:  4 drug-naive states were defined as having high or low viral load and CD4 counts using thresholds of 100,000 copies/mm³  and 350 cells/mm³. The fifth state represents the initiation of ART, and is the endpoint for this analysis. A transition rate matrix describing movement from state to state was defined by maximum likelihood methods for baseline data as well as in conjunction with covariates to describe factors that influence the natural history of HIV infection.

Results:  The multi-state Markov model converged to a solution for this cohort. Additionally, analysis with co-variates yielded a log-linear effect for each covariate entries in the transition matrix. This model describes the natural history of untreated early HIV infection, the decision to initiate ART in this cohort, and factors that influence the rate of transition from state to state. For instance, an untreated patient with well-controlled infection has a 23% probability of moving to a less favorable state within 1 year. Patients with poorly controlled infection are 2.3-fold more likely to initiate ART within a year. Finally, patients in the low-viral load/high-CD4 state who reported experiencing acute retroviral syndrome (ARS) symptoms transitioned to the high-viral load/high-CD4 state at a rate 1.8-fold higher than those with no ARS symptoms. Analysis of additional factors, including high resolution HLA genotypes and pol sequence polymorphisms are ongoing.

Conclusions:  We have constructed a multi-state Markov model to describe HIV progression, the decision to initiate ARV treatment in the early stages of HIV-infection, and covariates that influence these processes. Further exploration of this model promises to provide additional insights into factors associated with disease progression and clinical decision-making in early HIV infection.