Abstract
To compare the within-study discriminatory characteristics of established inflammatory back pain (IBP) criteria sets and individual items in predefined groups of patients with axial psoriatic arthritis (axPsA), non-psoriatic axial spondyloarthritis (axSpA), and chronic mechanical low back pain (cLBP). Data were drawn retrospectively from a multicentre PsA registry and a separate single-center axSpA/cLBP cohort. IBP criterion sets (Calin, ASAS, Berlin) and individual IBP items were evaluated using sensitivity, specificity, likelihood ratios, and area under the receiver operating characteristic curve (AUC). Because case and comparator groups were predefined rather than prospectively assembled from an unselected diagnostic population, the estimates represent classification-discrimination within the study sample, not real-world diagnostic accuracy. A total of 697 patients were analyzed, including 282 patients with axPsA, 292 with axSpA, and 123 with cLBP. In these selected comparisons, Calin and ASAS criteria showed high observed specificity, whereas Berlin criteria showed greater sensitivity (91.1% in axPsA and 86.7% in axSpA). Among individual IBP items, insidious onset showed the highest AUC in both comparisons (0.939, 95% CI 0.911-0.961 in axPsA and 0.913, 95% CI 0.881-0.939 in axSpA). In contrast, age at onset < 40 years and good response to NSAIDs showed little discrimination in axPsA (AUC 0.544 and 0.543, respectively). In multivariable analyses, Calin, ASAS, and Berlin criteria were associated with axPsA group membership, while Calin and ASAS criteria showed the strongest associations with axSpA group membership; conventional and Firth models yielded directionally similar estimates. Within these predefined cohorts, IBP criteria and individual items differed in their ability to discriminate group membership. IBP should be regarded as a supportive clinical feature, not a standalone screening or diagnostic tool. The high specificity and positive likelihood ratio estimates may reflect enrichment of symptom- and imaging-positive cases and use of an ASAS-negative cLBP comparator; prospective validation in unselected clinical populations is required.