OBJECTIVES: Direct preference elicitation is uncommon in clinical trials of COPD treatments. Investigators have created an algorithm that estimates EQ-5D preference weights from the St. George's Respiratory Questionnaire (SGRQ) to permit the calculation of QALYs. METHODS: Using data from a placebo controlled randomized trial comparing three regimens: 1) tiotropium plus placebo; 2) tiotropium plus salmeterol; and 3) tiotropium plus salmeterol/fluticasone) in COPD, we examined the validity of estimated EQ-5D scores from the SGRQ. RESULTS: A total of 351 patients with complete SGRQ scores at each point in the trial were included in the analysis. The mean values of the SGRQ and the estimated EQ-5D revealed similar pattern across the three treatments over time. A scatterplot of the SGRQ scores versus the EQ-5D scores showed that although there was an underlying linear relationship, it was somewhat "stepped" as the conversion algorithm results in the estimation of the same EQ-5D utility value for a number of SGRQ scores. Given that the minimally clinical important difference (MCID) of the SGRQ is 4, it would be possiblethat one might lose an MCID response when converting to the EQ-5D (whose MCID is 0.03). This was not the case in 85% of the SGRQ MCID responders. Finally, to see if any of the above made a difference statistically, the SGRQ scores and the estimated EQ-5D scores were compared at the end of the study (52 weeks). For the predicted mean EQ-5D utilities, treatment arms 1 and 3 were significantly different. However, when using the SGRQ, treatment arms 1 and 3 and treatment arms 1 and 2 were significantly different. CONCLUSION: The conversion algorithm was judged to be sufficient. However,use of the algorithm resulted in a reduction in discriminatory ability of the estimated EQ-5D as compared to the SGRQ.