Timing of antiretroviral therapy. Use of Markov modeling and decision analysis to evaluate the long-term implications of therapy

Background: The timing of initiation of antiretroviral therapy is controversial. Current guidelines are heavily based on the principle of 'hit early, hit hard', although the long-term implications of this approach are unknown. Methods: Using Markov modeling and decision analysis we modeled the virologic outcomes over 10 years in a hypothetical population of 10 000 HIV-infected patients in which therapy (with the possibility of three sequential regimens before the development of multidrug-resistant virus) is started immediately versus starting progressively at rates of 5, 10, 15, 20 or 30% of the original population each year. The model used inputs from available clinical trial data: virologic success rate and durability of the response of the first and subsequent regimens. We performed one-way and two-way sensitivity analysis to evaluate changes in the input variables. Results: If therapy is started immediately in all patients, by 10 years 57% would be undetectable, but 38% would have detectable multidrug-resistant virus. In contrast, the population as a whole would have had better virologic outcomes if one waited before starting treatment at any progression rate; for example, initiating therapy in 10% of the subjects per year results in 64% of patients being undetectable and 24% with multidrug-resistant virus. Two-way sensitivity analysis demonstrates that immediate initiation should be at least 15 to 20% better than delayed antiretroviral therapy to justify immediate initiation of therapy over a wide range of success rates of the delayed start. Conclusion: Our analysis, utilizing optimistic outcomes based on short-term clinical trials, provides a theoretical basis for questioning the current aggressive early use of therapy and should help prompt studies that look at when and how to start antiretroviral therapy.