Computed Tomography of Suspected Scaphoid Fractures

Abstract

Computed tomography (CT) can be used to triage suspected scaphoid fractures. This study assessed intraobserver and interobserver reliability and positive and negative predictive values of CT for the diagnosis of a nondisplaced scaphoid fracture. Eight observers evaluated CT scans from 30 patients (13 with nondisplaced scaphoid fractures, 17 with no scaphoid fractures) for the presence or absence of a fracture. Five observers evaluated the scans a second time. Statistical analyses included intraobserver and interobserver reliability and diagnostic characteristics. Computed tomography had substantial intraobserver and interobserver reliability for the diagnosis of a nondisplaced scaphoid fracture. The average sensitivity, specificity, and accuracy of CT for a nondisplaced scaphoid fracture were 89%, 91%, and 90% for the first round and 97%, 85%, and 88% for the second round of observations, respectively. Based on an estimated prevalence of 5% true fractures among patients with suspected scaphoid fractures, the average positive predictive value for the detection of radiographically occult scaphoid fractures with tomography of the wrist was 0.28. The average negative predictive value was 0.99. Computed tomography should be used with caution for triage of nondisplaced scaphoid fractures because false-positive results occur, perhaps from misinterpretation of vascular foraminae or other normal lines in the scaphoid. Given the relative infrequency of true fractures among patients with suspected scaphoid fractures, CT is better for ruling out a fracture than for ruling one in.