Distinction of long bone stress fractures from pathologic fractures on cross-sectional imaging: how successful are we?

Objective: The objectives of our study were to define CT and MRI features that distinguish pathologic fractures from stress fractures and to compare the performance of CT and MRI with radiography.
Materials and Methods: Two reviewers retrospectively reviewed 45 MR images, 37 CT scans, and 43 radiographs in 59 patients (30 biopsy-proven pathologic fractures and 29 stress fractures followed to resolution). The features observed on MRI were abnormal bone marrow (well-defined, ill-defined); intracortical, periosteal, or muscle T1 or T2 signal; endosteal scalloping; and a soft-tissue mass. The features seen on CT were marrow abnormality and character (well-defined, ill-defined, permeative, moth-eaten), endosteal scalloping, periosteal reaction (benign, aggressive), and a soft-tissue mass. Reviewers rated their confidence for diagnosing a pathologic fracture on a 1-3 scale (< 50%, 50-95%, > 95% sure, respectively) with each technique. Performance of each technique was defined by reviewer accuracy and area under the receiver operating characteristic curve (Az); the frequency with which the MRI and CT features were associated with pathologic and stress fractures was calculated.
Results: For both reviewers, accuracy for differentiating pathologic from stress fractures was highest on MRI (accuracy/Az: reviewer 1, 98%/0.97; reviewer 2, 93%/0.99); CT (reviewer 1, 88%/0.83; reviewer 2, 82%/0.90) was less accurate than radiography (reviewer 1, 94%/0.98; reviewer 2, 88%/0.96). On MRI, pathologic fractures compared with stress fractures exhibited well-defined T1 marrow signal (83% vs 7%, respectively; p < 0.001), endosteal scalloping (58% vs 0%, p < 0.001), muscle signal (83% vs 48%, p = 0.026), and a soft-tissue mass (67% vs 0%, p < 0.001). On CT, pathologic fractures compared with stress fractures exhibited marrow abnormality (84% vs 17%, respectively; p = 0.001), endosteal scalloping (44% vs 0%, p = 0.006), and aggressive periosteal reaction (36% vs 0%, p = 0.04).
Conclusion: MRI is useful for distinguishing pathologic from stress fractures, especially after inconclusive radiographic findings. Specifically, pathologic fractures exhibit well-defined T1 marrow alterations, endosteal scalloping, and adjacent soft-tissue abnormalities.