Operational definitions and observer reliability of the plain radiographic features of psoriatic arthritis

To determine the standardization and the observer reliability of potentially diagnostic plain radiographic features of psoriatic arthritis (PsA).

Radiographic features were chosen on the basis of a systematic review of the literature. One hundred sixty-four radiographs from 62 patients were selected from various sources by a musculoskeletal radiologist. Radiographs were read independently by 2 observers (WJT, PSH) and scored for the presence or absence of each evaluated feature. Cohen's kappa was used to determine observer agreement beyond chance, and the accuracy of each observer (with reference to the radiologist's judgment) was determined by likelihood ratios.

The 2 observers demonstrated similar accuracy, although WJT tended to be more accurate for items classed as absent and PSH more accurate for items classed as present. The following features showed sufficient reliability to be reasonably included in further testing of their discriminatory value (intra- and interobserver kappa values): marginal syndesmophyte (0.68, 0.69), non-marginal syndesmophyte (0.75, 0.59), paravertebral ossification (0.89, 0.79), destructive discovertebral lesion (0.85, 0.65), Romanus lesion (0.64, 0.43), sacroiliitis (0.99, 0.86), entheseal erosion (0.80, 0.71), entheseal ossification (0.69, 0.76), distal interphalangeal erosive disease (0.58, 0.52), joint osteolysis (0.62, 0.47), juxtaarticular bony proliferation (0.43, 0.42), bony ankylosis (0.53, 0.54), tuft osteolysis (0.51, 0.36). The features that showed inadequate reliability were: loss of cortical definition of terminal tuft (0.33, 0.31) and periosteal new bone formation (0.42, 0.03).

A number of plain radiographic features of PsA have sufficient reliability to justify inclusion in diagnostic classification criteria sets for further testing.