OBJECTIVES:  Guidance for HTA in several health jurisdictions mandates the use of EQ-5D in the measurement of health benefits.  In the UK a standard scoring system for EQ-5D-3L was established in a 1993 survey using a subset of 45/243 directly observed states. Significant advances in data analysis and modelling over the past 20 years might now yield a different scoring system. In this paper, we assess the circumstances under which modifications of the valuation set for EQ-5D-3L would produce changes in marginal QALYs and whether such changes would compromise the status of existing cost-effectiveness decisions. METHODS: The 1993 scoring system was compared with a number of alternatively derived value sets that explored the influence of parameter and structural uncertainty. The influence of these methodological factors on the distribution of the (243x242)/2 pairwise differences in health state utilities was tested using a range of values for incremental cost and cost-effectiveness threshold to establish the conditions under which past cost-effectiveness decisions might need to be revised. RESULTS: Our analysis shows important variation in the ICERs obtained using different value sets. In particular, when the relevant incremental QALY corresponds to a health improvement where one dimension changes from level 3 (“extreme” problems) to level 2 (“moderate” problems), ICERs are particularly sensitive to the inclusion of the “N3” constant in the model. For a range of baseline QALYs and health improvements, the decisions that have been made based on ICERs generated with the 1993 MVH value set may have been drastically different had another specification been used. CONCLUSIONS: The specification of the model used to estimate values for EQ-5D-3L health states has a pronounced impact on the scoring system that can compromise the legitimacy of economic evaluation of interventions targeting severe ill-health. This study has implications for the transitional use of the 5L and 3L versions of EQ-5D.