OBJECTIVES: Preference-based Quality-of-Life Indexes (QLI) represent the utility value related with health status. QLI consist of 1) a descriptive system of health states and 2) an algorithm to calculate values for each health states described in the system. When QLI are combined with life years, it is possible to obtain quality-adjusted life years (QALYs). For QALYs to be useful in the prioritization of resources and decision-making in health, QLI must reflect a cardinal order in the preferences of patients. In this regard, the theory that provides an axiomatic framework to reach a cardinal order in preferences was the von Neumann-Morgenstern Expected Utility Theory (EUT). According to the vN-M’s EUT patients prefer alternatives that provide them a maximum utility expected value (which, in the context of health, would be a higher QLI). To ensure the maximization of expected utility value, independence is a necessary condition. However, in 1953 Maurice Allais demonstrated the violation of the independence condition using an experimental design. The purpose of this study was to redesign the Allais’ experiment using 13 health states described with the EQ-5D system, in order to verify whether the paradox of Allais is also present in the QLI, which are calculated as vN-M utilities. METHODS: First, the QLI were calculated using the standard gamble approach. Second, we carried out the experiment in 20 individuals in Colombia. RESULTS: Results showed that maximization of the expected utility value is not fulfilled because the violation of independence condition. CONCLUSIONS: These results cast doubts on the accuracy of QALYs as a measure of patient preferences and their validity to support the process of decision-making in health, as suggested by other researchers.