Background: This study tested the null hypothesis that computed tomography (CT) and magnetic resonance imaging (MRI) have the same diagnostic performance characteristics for triage of suspected scaphoid fractures.

Methods: Thirty-four consecutive patients with a suspected scaphoid fracture (post-injury tenderness of the scaphoid and normal radiographs) underwent CT and MRI within ten days after trauma. CT-reconstructions were made in planes defined by the long axis of the scaphoid. The reference standard for a true fracture of the scaphoid was 6-week follow-up radiographs in four views, based on current literature. A panel including surgeons and radiologists came to a consensus diagnosis for each type of imaging considered in a randomized and blinded fashion, independent of the other types of imaging. We calculated sensitivity, specificity and accuracy as well as positive (PPV) and negative predictive values (NPV) for both imaging modalities.

Results: According to the reference standard there were six true fractures of the scaphoid (prevalence 18%). CT diagnosed fracture of the scaphoid in five patients (15%), with one false positive, two false negative and four true positive results. MRI diagnosed a fracture in seven patients (21%), with three false positive, two false negative and four true positive results. Sensitivity, specificity and accuracy for CT were 67%, 96% and 91%; and for MRI 67%, 89% and 85% respectively. According to the McNemar test for paired binary data for each imaging modality these differences were not significant. The positive predictive values using Bayes’ formula were 76% for CT and 54% for MRI. Negative predictive values were 94% for CT and 93% for MRI.

Conclusions: CT and MRI had comparable diagnostic characteristics. Both were subject to both false positive and false negative interpretations. They were better to rule out a fracture than to rule one in. The best reference standard for a true fracture is debatable