Table 7

 Pairwise comparisons between and within direct and indirect methods

ComparisonNumber of statesMean difference (SD)Number of independent groupsSign testRegression slope (SE)Predictive P value
Current patients
Time trade off – EQ-5D400.17 (0.13)26P<0.0010.54 (0.07)P<0.001
Standard gamble – EQ-5D390.20 (0.15)27P<0.0010.50 (0.05)P<0.001
Time trade off – HUI-3240.19 (0.09)13P=0.0230.48 (0.06)P<0.003
Standard gamble – HUI-3290.15 (0.11)16P=0.0040.65 (0.08)P<0.002
Standard gamble – Time trade off280.02 (0.06)16P=0.804
HUI-3 – EQ-5D250.03 (0.13)14P=0.424
Hypothetical patients
Time trade off – EQ-5D100.13 (0.12)2NA0.81 (0.01)P<0.994
Standard gamble – EQ-5D110.15 (0.11)2NA0.50 (0.06)P<0.042
Standard gamble – Time trade off60.22 (0.05)1NA

NA=not applicable. Second column shows numbers of states contributing to mean and SD of the difference. Sign tests and regression analyses are based on aggregate health states obtained from averaging data within independent groups of participants. Regression slopes refer to least squares lines for predicting direct utility from indirect utility, constrained to pass through the point (1,1). Predictive P-values assess the goodness-of-fit of the lines, taking variation within studies into account.