Development and preselection of criteria for short term improvement after anti-TNFα treatment in ankylosing spondylitis

Objective: To develop and compare candidate improvement criteria for anti-TNFalpha treatment in ankylosing spondylitis with optimal discriminating capacity between treatment and placebo. Methods: Data from two randomised controlled trials which included 99 patients treated with infliximab or etanercept were used to evaluate 50 candidate improvement criteria. These were developed on the basis of pain, patient's global assessment, function, morning stiffness, spinal mobility, and C reactive protein. Different levels of improvement in each domain (20-60%) were used to define Boolean type criteria. These criteria were compared with different percentages of improvement on the BASDAI and with modified ASAS improvement criteria. Bootstrap methods were applied to calculate 95% confidence intervals (CI) of the chi(2) test values to select the best candidate improvement criteria. Results: The best performing improvement criteria were "20% improvement in five of six domains'' (chi(2) = 31.9 (95% CI, 18.0 to 46.9)) with a low placebo response of 2.9% and a high response to infliximab of 67.7%; and "ASAS 40% improvement'' (chi(2) = 26.5 (13.3 to 41.1)), with response to placebo of 5.7% and response to infliximab of 64.7%. The good discriminating capacity of the two improvement criteria was confirmed by the combined dataset of the infliximab and etanercept trial. Conclusions: The "five of six'' improvement criterion has the advantage of including the objective domains spinal mobility and acute phase reactants, but requires only 20% improvement. The ASAS 40% improvement criterion has the advantage of setting a high threshold, but only in patient reported outcomes. The choice between these improvement criteria needs to be based on further validation from upcoming trials.