Objective The objective of the present study was to compare the SF-6D and EQ-5D dimension scores and utilities of patients with hypertension, ankylosing spondylitis, panic disorder and osteoarthritis. Methods The data set was taken from four separate studies (867 cases). Both the EuroQol and the SF36 were completed at baseline. The EQ-5D and the SF-6D utilities were derived using previously described algorithms (Dolan, 1995; Brazier et al, 2002, respectively). Only cases who completed both questionnaires were included in the analysis (hypertension N=233, ankylosing spondylitis N=111, panic disorder N=184, osteoarthritis = 155). The within population agreement was examined by paired-samples T-tests and Spearman Rank Correlation coefficients. The agreement within cases was assessed by the Intra Class Correlation coefficient (ICC). Discriminative ability, and correlations and differences in frequencies between the dimension scores were evaluated. Results The mean EQ-5D utilities were 0.84 (sd 0.20) in hypertension, to 0.67 (sd 0.20) in ankylosing spondylitis, 0.61 (sd 0.27) in panic disorder, and 0.34 (sd 0.33) in osteoarthritis. The mean SF-6D utilities were 0.77 (sd 0.12) in hypertension, 0.62 (sd 0.11) in panic disorder, 0.65 (sd 0.10) in ankylosing spondylitis, and 0.58 (sd 0.11) in osteoarthritis. The EQ-5D utilities covered a range of 0.50, whereas the SF-6D utilities covered a range of 0.19. The EQ-5D and SF-6D utilities were statistically significantly different in hypertension and osteoarthritis (p<0.0001). The relation between the utility estimates was moderate to strong (correlation coefficients from 0.48 to 0.59). The ICCs showed that agreement within cases was low to moderate (0.22 in osteoarthritis to 0.46 in hypertension). In contrast to the SF-6D utilities, the EQ-5D utilities did not discriminate between patients with ankylosing spondylitis and panic disorder. The correlations between similar dimensions of the EQ-5D and SF-6D were predominantly moderate (0.30-0.50). The lowest correlations were observed in osteoarthritis between SF-6D physical functioning and EQ-5D mobility (0.19), and EQ-5D usual activities (0.29), and in ankylosing spondylitis between SF-6D pain and EQ-5D pain and discomfort (0.28). Disparities between frequencies of the scores on similar dimensions were observed in all conditions. The disparities between the frequencies on similar dimensions were particularly conspicuous in ankylosing spondylitis and panic disorder, since the mean utilities for these conditions did not differ significantly. Conclusion The EQ-5D utility was 0.17 higher compared to the SF-6D utility in the population with a mild condition (hypertension), the utilities were alike in the populations with a moderate condition (ankylosing spondylitis and panic disorder), and the EQ-5D estimate was 0.24 lower in the population with a severe condition (osteoarthritis). Clearly, EQ-5D utilities and SF-6D utilities cannot be used interchangeably. Whether the differences between the two utility estimates follow a pattern in which healthier states have a higher utility on the EQ-5D, whereas poorer health states have a higher utility on the SF-6D deserves further investigation. While in our dataset the range of the EQ-5D utilities was 2.6 times larger than the range of SF-6D utilities, another question that remains is whether the larger range of EQ-5D utilities translates into larger utility gain
