Background/Purpose: A standardized validated definition for gout flares (or attacks) is not available. Two provisional definitions published in 2012 were based on patient-reported elements (patient-defined attack, pain at rest greater than 3 in a 0-10 numeric rating scale, presence of at least one swollen joint, presence of at least one warm joint). These definitions had acceptable sensitivity and specificity but lacked external validation which would facilitate its adoption in gout clinical studies. Our objective was to perform external validation of previously published preliminary gout flare definitions in patients with gout.

Methods: We enrolled 509 participants with gout from 17 international sites in a cross-sectional study performed during routine clinical care. All patients met the 2015 ACR/EULAR classification criteria for gout. Criteria from the previously published gout flare definitions were collected by a site investigator and the final adjudication of a gout flare status was done by a local expert rheumatologist, through an evaluation independent from that of the site investigator. Logistic regression, Bayesian statistics, and receiver-operator curves were used to calculate the final diagnostic performance of the gout attack definitions which were based on number of criteria and a classification and regression tree (CART) approaches.

Results:   The mean age of participants was 57.5 years (standard deviation [SD] 13.9) and 89% were men. Mean disease duration was 12.3 (SD 10.3) years, 35.4 % had tophi, and 75% were taking urate-lowering therapies. The previously published and favored number of criteria definition requiring the presence of 3 or more out of 4 criteria was found, using the current study data, to be 85% sensitive and 95% specific in confirming the presence of flare in patients with gout (Table). The concurrent logistic regression model had an area under the curve of 0.97.  The previously published definition based on a CART algorithm (entry point pain at rest > 3 followed by patient-defined attack “yes”) was 73% sensitive and 96% specific using the current study data (Table). The number of criteria approach with a cut-point at 3 or more out of 4 criteria had higher diagnostic accuracy using the current study data than in its initial 2012 description (92% versus 84%).  Finally, using current study data the number of criteria approach at 3 or more out of 4 criteria had higher accuracy to the CART algorithm based approach (92% versus 89%) but with a much better sensitivity (85% versus 73%).

Conclusion: The definition requiring the presence of 3 or more out of 4 patient-reported criteria is validated to be sensitive, specific, and accurate in identifying flares in patients with gout using an independent large international sample. Having a validated gout flare definition will improve ascertainment of outcomes in gout clinical studies.