Diagnosis of femoral neck fracture associated with femoral shaft fracture: Blinded comparison of computed tomography and plain radiography

OBJECTIVE:: Femoral neck fractures that are associated with femoral shaft fractures have historically been associated with high rates of missed diagnosis. Despite the potentially serious consequences of a missed femoral neck fracture, little work has been conducted to rigorously evaluate the ability of commonly used imaging studies to detect such fractures. Our hypothesis was that axial-view computed tomography is superior to plain radiography at detecting femoral neck fractures. DESIGN:: Blinded assessment of a randomized image set that included axial-view computed tomographic scans of the pelvis, anteroposterior-view radiographs of the pelvis, and anteroposterior-view radiographs of the femur. SETTING:: Academic trauma center. PATIENTS:: Twenty-eight patients who had femoral shaft fractures with femoral neck fractures and 60 patients who had femoral shaft fractures without femoral neck fractures. INTERVENTION:: Images that were stripped of demographic data were independently viewed in random order on computer workstations by 5 trauma fellowship-trained orthopaedic surgeons who were blinded to treatment and diagnosis. MAIN OUTCOME MEASUREMENTS:: Sensitivity, specificity, positive post-test probability, and 1 minus negative post-test probability were calculated for diagnosing femoral neck fracture based on each image type. RESULTS:: Interobserver reliability showed "substantial agreement" (kappa > 0.65) for all imaging modalities, indicating that the surgeons had a high level of agreement. All 3 imaging sets had high specificity (>94%) and 1 minus negative post-test probability (>95%) but had poor sensitivity (<65%) and positive post-test probability (<58%). False positives and false negatives were equally likely for all imaging modalities (P > 0.2). CONCLUSIONS:: Plain radiography and computed tomography have rates of missed femoral neck fractures that are similar and substantial, with a sensitivity of only 56%-64%. Our data emphasize the importance of intraoperative and postoperative imaging in detecting nondisplaced femoral neck fractures in association with femoral shaft fractures.