Background: The need for cost effectiveness analyses in randomized controlled trials that compare treatment options is increasing. The selection of the optimal utility measure is important, and a central question is whether the two most commonly used indexes - the EuroQuol 5D (EQ5D) and the Short Form 6D (SF6D) - can be used interchangeably. The aim of the present study was to compare change scores of the EQ5D and SF6D utility indexes in terms of some important measurement properties. The psychometric properties of the two utility indexes were compared to a disease-specific instrument, the Oswestry Disability Index (ODI), in the setting of a randomized controlled trial for degenerative disc disease.
Methods: In a randomized controlled multicentre trial, 172 patients who had experienced low back pain for an average of 6 years were randomized to either treatment with an intensive back rehabilitation program or surgery to insert disc prostheses. Patients filled out the ODI, EQ5D, and SF-36 at baseline and two-year follow up. The utility indexes was compared with respect to measurement error, structural validity, criterion validity, responsiveness, and interpretability according to the COSMIN taxonomy.
Results: At follow up, 113 patients had change score values for all three instruments. The SF6D had better similarity with the disease-specific instrument (ODI) regarding sensitivity, specificity, and responsiveness. Measurement error was lower for the SF6D (0.056) compared to the EQ5D (0.155). The minimal important change score value was 0.031 for SF6D and 0.173 for EQ5D. The minimal detectable change score value at a 95% confidence level were 0.157 for SF6D and 0.429 for EQ5D, and the difference in mean change score values (SD) between them was 0.23 (0.29) and so exceeded the clinical significant change score value for both instruments. Analysis of psychometric properties indicated that the indexes are unidimensional when considered separately, but that they do not exactly measure the same underlying construct.
Conclusions: This study indicates that the difference in important measurement properties between EQ5D and SF6D is too large to consider them interchangeable. Since the similarity with the "gold standard" (the disease-specific instrument) was quite different, this could indicate that the choice of index should be determined by the diagnosis.