OBJECTIVES Where data collected using a preferred utility instrument are not available, mappings may be available to link either disease specific or generic quality of life instruments to the preferred measure. The objective of this study was to investigate the similarity of a widely used older mapping from SF-36 to EQ-5D and a more recently published mapping linking the same two instruments. METHODS Patient level SF-36 data from the PSUMMIT trials of ustekinumab versus placebo in psoriatic arthritis (PsA) were used to generate EQ-5D utilities based upon the older algorithm (Gray) and the newer algorithm (Rowen). A regression linking disease specific instruments to the resulting EQ-5D utilities, used in a published economic evaluation of anti-TNFα treatments for PsA, was also replicated to investigate the difference the choice of mapping would make if the results were used in economic analysis. RESULTS The mapping algorithms showed similar mean values (Gray = 0.572, Rowen = 0.568) and a large degree of agreement in estimates (R= 0.95). Replicating a published regression predicting EQ-5D based on Health Assessment Questionnaire (HAQ) and Psoriasis Area and Severity Index (PASI) scores, however, led to subtle yet statistically significant differences in coefficients (Gray EQ-5D = 0.782 -0.196xHAQ -0.00227xPASI vs. Rowen EQ-5D = 0.816 -0.220xHAQ –0.0231xPASI). In this analysis the Rowen algorithm was more sensitive to changes in disease status, which is potentially important in a disease area with multiple available therapies. CONCLUSIONS The choice of mapping algorithm has the potential to affect the results of analyses performed. Although mean values obtained from different mapping algorithms may be similar, regression coefficients were shown to differ by mapping algorithm. Analyses using all available mappings should be presented to avoid accusations of ‘cherry picking’ the most favourable estimate. When suggesting the most appropriate analysis, clear justification should be provided.