Objective: To develop criteria for symptomatic improvement in patients with ankylosing spondylitis (AS), using outcome domain data from placebo-controlled clinical trials of nonsteroidal antiinflammatory drugs (NSAIDs).

Methods: Patient data from 5 short-term, randomized, controlled trials were used to assess equivalence, reliability, and responsiveness of multiple items in the 5 outcome domains for AS treatment: physical function, pain, spinal mobility, patient global assessment, and inflammation. At least one measure per domain was responsive (standardized response mean of > 0.5), except for the spinal mobility domain, which was omitted from the criteria. We developed and tested candidate improvement criteria in a random two-thirds subset from the 3 largest trials and used the remaining one-third for validation. These 3 largest trials included 923 patients (631 receiving NSAIDs, 292 in placebo groups). We selected the multiple domain definition that best distinguished NSAID treatment from placebo by chi-square test and that had a placebo response rate of < or = 25%.

Results: Candidate definitions were changes in single domains and in multiple measure indices, as well as combinations of improvements in multiple domains. Worsening in a domain was defined as a change for the worse of > or = 20% and a net change for the worse of > or = 10 units on a scale of 0-100. Partial remission (for comparison purposes) was defined as an end-of-trial value of < 20/100 in each of the 4 domains. Among 20 candidate criteria, change of > or = 20% and > or = 10 units in each of 3 domains and absence of worsening in the fourth discriminated best in the development subset (51% of patients improved with NSAIDs, 25% with placebo; chi2 = 36.4, P < 0.001). Results were confirmed in the validation subset. Almost all patients satisfying the definition of partial disease remission at the end of the trial had also improved by this criterion. Among all 923 patients, improvement rates using this criterion were 49% for NSAID-treated patients and 24% for placebo-treated patients.

Conclusion: Although further validation using data from new trials is still needed, we conclude that we have developed a clinically valid, easy-to-use measure of short-term improvement in AS.