Validity of conventional radiography in determining scaphoid waist fracture displacement.

Abstract

This study examines interpreter accuracy and interobserver agreement in evaluating conventional radiographs for scaphoid waist fracture displacement. Six fresh-frozen cadaver arms were obtained transected above the elbow. A waist fracture was created in each scaphoid. In a random fashion, three of the fractures were displaced 1 mm in the radial-ulnar plane, whereas the other three were reapproximated to anatomic position before all fractures were stabilized with radiolucent adhesive glue. A three-view conventional radiography series consisting of a posterior-anterior, lateral, and ulnar-deviated elongated scaphoid view was obtained for each wrist. Each radiograph was then presented in the same sequence for interpretation to six independent observers: two hand surgeons, two musculoskeletal radiologists, and two senior orthopaedic surgery residents who were all blinded to the actual fracture pattern. In 14 of the 18 (78%) displaced fracture radiographic series, the interpreters correctly recognized displacement being present. However, displacement was also reported in six of the 18 (33%) nondisplaced fracture series. The estimated overall accuracy of all readings for distinguishing between displaced and nondisplaced fractures was 72%. Of the 90 total possible pairwise agreements between interpreters regarding fracture displacement, there were 54 actual agreements (60%), and kappa was estimated to be 0.31. Taken together, these two measures of agreement can be interpreted as indicating poor to moderate agreement. Our results suggest that conventional radiography is not reliable in determining 1-mm scaphoid waist fracture displacement in the radioulnar plane and also indicated a lack of strong interobserver agreement.