This paper offers an indirect measure of patient welfare based on whether patients comply with the prescription they receive. Adherence behavior is supposed to reveal patients' subjective valuations of particular therapies. We write a simple theoretical model of patient adherence behavior, that reflects the trade-off between perceived costs and observed regimen efficacy. A discrete choice framework is then used for the estimation, i.e. the comparison of the incremental benefit of drug intake between two regimens. Consequently, the empirical analysis is based on the identification of patient and drug characteristics associated with adherence. The econometric approach is implemented through a bivariate panel two-equation simultaneous system studying jointly the factors associated with adherence and response to treatment. The data come from a randomized clinical trial conducted in France between 1999 and 2001 and comparing the efficacy of two tritherapy strategies in HIV disease. Both the theoretical and empirical results suggest that, for comparable clinical efficacy and toxicity levels, a higher adherence level is associated with higher patient welfare, thus adding valuable information to conclusions drawn by a mere biostatistical analysis. Therefore, from the perspective of the patient, the adherence-enhancing drug must be favored. Our results based on panel data also stress that unobserved patient characteristics account substantially for drug valuation and that the assessment evolves during the course of the treatment. Furthermore, we provide a new framework for the analysis of adherence data. The microeconometric framework highlights that non-adherence is an endogenous behavior, thus suggesting new ways for improving adherence.