Performance of cut-offs of the ASAS Health Index to discriminate between treatment groups in patients with axial spondyloarthritis in the TICOSPA trial

Objective: To test trial and longitudinal known group discrimination of thresholds of meaning for improvement and health states of the ASAS Health Index (ASAS HI) in patients with active axSpA treated in a randomized study. Methods: Data from baseline and week 48 from the tight-controlled, treat-to-target trial TICOSPA study were used. The performance of different thresholds to assess change or health states of the ASAS HI were evaluated between arms and against changes in patients' relevant outcomes and various external responder criteria. Analyses were performed by comparing the mean values t-tests or proportion of responders of continuous and dichotomous external criteria respectively. Trial discrimination of the ASAS HI thresholds were assessed by odds ratios and Phi coefficient in a large number of potential ASAS HI thresholds. Differences in health states in relevant external outcomes between ASAS HI responders and non-responders was assessed by comparing the best performing improvement and state thresholds by using t-tests and chi-square, as appropriate. Missing data on outcomes was handled by non-responder imputation (NRI). Results: All 160 patients had available ASAS HI data. Trial discrimination was larger for absolute ASAS HI change of >= 2.0, >= 2.5, and >= 3.0 points followed by ASAS HI 20 % improvement. Odds ratio ranged between 1.27 and 1.75 for absolute and between 1.0 and 1.64 for relative improvement outcomes. Longitudinal discrimination of ASAS HI improvement >= 30 % or >= 3.0 points had a larger reduction in patient global and disease activity and reached more often remission compared to patients with no significant improvement in global functioning. Patients who achieved ASAS HI <= 5.0 compared with patients who did not achieve such states were more likely to have ASAS partial remission, ASDAS inactive disease or ASDAS low activity at week 48. Conclusions: The data-driven thresholds of the ASAS HI identified in a longitudinal observational setting perform well in the context of a randomized trial.