Despite extensive study, the optimal imaging modality for the diagnosis of a suspected scaphoid fracture remains controversial. Traditionally, patients with clinical findings consistent with scaphoid fracture but with negative initial radiographs have been managed with splint or cast immobilization followed by repeat clinical and radiographic examination two weeks following injury. However, given the relatively low prevalence of true fractures in this population, the classical approach has been criticized for overtreating a majority of these patients. To avoid diagnostic delay and unnecessary immobilization, advanced imaging techniques have gained popularity to facilitate early definitive diagnosis of scaphoid fractures.
Studies advocating the use of magnetic resonance imaging (MRI), computed tomography (CT), bone scintigraphy, and ultrasound in the evaluation of suspected scaphoid fractures are plentiful. Among these imaging modalities, MRI has emerged as a relative gold standard, due to both its excellent reported diagnostic performance as well as its ability to identify alternate causes of wrist pain. Criticisms of MRI have been that it is susceptible to misinterpretation of injuries such as bone bruises, potentially resulting in false positive results, and that it has limited capacity to demonstrate subtle degrees of fracture displacement. The use of CT in the acute diagnosis of scaphoid fractures is less well documented but has been advocated because of its lower costs and widespread availability, although the diagnostic accuracy of CT in this situation has been questioned1. Although current institutional protocols vary, an often recommended algorithm is to first proceed with MRI when there is a clinically suspected scaphoid fracture but negative radiographic results, reserving CT to assess for potential displacement or angulation when a fracture has been demonstrated2. The appeal of a single diagnostic modality that can simultaneously identify fracture and provide a clear visualization of displacement is apparent.
Recent systematic reviews demonstrate only marginally superior diagnostic performance of MRI over CT in noncomparative studies3,4. To my knowledge, to date there has been only a single prospective trial comparing the performance of MRI and CT against a reference standard. Memarsadeghi et al.5 found improved diagnostic performance of MRI (sensitivity 100%, specificity 100%) compared with CT (sensitivity 73%, specificity 100%) with regard to fracture diagnosis on six-week radiographs, although this difference did not reach significance. This study has been criticized for the use of CT images obtained relative to the axis of the wrist rather than the recommended long axis of the scaphoid.
In this issue of The Journal, Mallee et al. present a prospective comparative evaluation of the use of MRI and CT in the diagnosis of occult scaphoid fractures. The authors have modified the CT imaging protocol utilized by Memarsadeghi et al.5 to examine reconstructed images relative to the long axis of the scaphoid, not the wrist. This study also incorporates a subtle but notable alteration in the MRI criteria for diagnosis of fracture with regard to the interpretation of marrow edema. Thirty-four consecutive patients with clinically suspected scaphoid fractures were examined prospectively with CT and MRI. Six-week follow-up radiographs were used as the reference standard, demonstrating true fracture in six patients. The sensitivity, specificity, and accuracy were 67%, 96%, and 91%, respectively, for CT, and 67%, 89%, and 85%, respectively, for MRI. Prevalence-adjusted positive and negative predictive values, estimating a prevalence of true fracture of 16%, were 0.76 and 0.94, respectively, for CT, compared with 0.54 and 0.93, respectively, for MRI. The study was designed to achieve 80% power to detect proportional differences of 20%. Within this framework, these performance characteristics were not statistically different. The authors conclude that CT provides a comparable alternative to MRI in the diagnosis of suspected scaphoid fractures.
This well-designed prospective study follows the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) guidelines and contributes quality evidence to the literature. However, some limitations are apparent. As the authors acknowledge, a fundamental difficulty in the design and interpretation of comparative imaging studies is the absence of an established gold standard for the definitive diagnosis of scaphoid fractures. The commonly utilized reference standard—conventional radiographs obtained six weeks after injury—remains controversial and has potential shortcomings, as noted in their discussion. Ultimately, however, the reliance on this standard by the present study unavoidably limits the strength of its conclusions. While it may be argued that CT scanning at six weeks may represent a more appropriate tool to define the presence or absence of scaphoid fracture, the use of this modality as a reference standard remains unvalidated.
A related limitation in the evaluation of advanced imaging studies resides in the precise criteria used for diagnosing a fracture. A concern with both MRI and CT is the potential for false-positives. This effect is magnified, given the low prevalence of true fracture in these patients. On CT imaging, the significance of unicortical lucent lines remains unknown; it has been speculated that these lucencies may represent vascular channels rather than fracture1. The CT criteria utilized by Mallee et al. may have been susceptible to overdiagnosis of fracture in these patients. On MRI images, the significance of bone-marrow edema without clear cortical fracture lines, similar to the appearance of a stress fracture, continues to be debated. The current study, by means of a consensus opinion among the reviewing panel, decided that a focal zone of edema was considered a fracture. Potential overinterpretation of this finding may have also contributed to false-positive results. On a larger scale, variations in the treatment of these indeterminate findings across studies limit comparative analysis and may account for some of the inconsistency in diagnostic performance that has been reported in the literature. Ultimately, the clinical significance of these findings must be further characterized to determine which imaging findings should or should not be considered to represent a fracture. Consistent diagnostic criteria and uniform treatment of these cases will also allow more rigorous comparison between studies.
A common theme among second-line imaging techniques is that they are all better at excluding rather that confirming fracture, and this finding was confirmed in the present study. Reported negative predictive values for MRI, CT, and bone scintigraphy in the literature are sufficient such that a negative result with any one of these modalities should exclude the diagnosis with reasonable certainty4. Both CT and MRI in the present series compared poorly relative to these reported diagnostic characteristics. Although the number of true-positive fractures in accordance with the reference standard was small, the fact that both CT and MRI failed to diagnose one-third of these fractures certainly invites further investigation.
Although the present study is unlikely to definitively settle the debate surrounding the optimal imaging modality for the diagnosis of scaphoid fracture, it does contribute quality evidence to this discussion. Even though there may be no perfect imaging technique, this report challenges the perceived superiority of MRI over CT. At the same time, this study highlights many of the inherent limitations in the comparative analysis of imaging modalities in this setting and emphasizes the need for more rigorous assessment of CT and MRI findings.