Can Orthopaedic Oncologists Agree on What Is and Is Not a Pathologic Fracture?

BACKGROUND: Orthopaedic oncologists have not agreed on a common definition of pathologic fracture despite abundant reports on diagnosis and management. A standard definition is important for treatment decisions, outcomes assessment, and administrative categorization; therefore, it is first necessary to understand areas of agreement and disagreement on the imaging findings exclusively before considering nonradiologic features of pathologic fracture.

QUESTIONS/PURPOSES: (1) Based on images (radiographs, CT scans, and MRIs), to what degree do orthopaedic oncologists agree on whether or not a pathologic fracture is present? (2) What imaging features lead orthopaedic surgeons to believe that a pathologic fracture is present?

METHODS: We developed a survey of images from 20 patients with metastatic bone disease and distributed them to 30 orthopaedic oncologists holding committee or leadership positions in the Musculoskeletal Tumor Society (MSTS). Images were reviewed twice, at least 6 months apart. Each question consisted of a PowerPoint slide of a patient with metastatic bone disease, with representative plain radiographs, CTs, and MRIs when available. On the first survey, there was only one question for each patient scenario: "Is there a pathologic fracture?" Respondents were asked to respond simply "yes" or "no." In the second survey, consisting of the same PowerPoint slides in a different order, we asked the participants to provide additional insight and rationale behind their decision for each case regarding whether or not there was a fracture to further investigate areas of differences in interpretation or assessment. The Fleiss kappa (κ) was used to evaluate interrater agreement, and the Cohen kappa assessed intrarater agreement for the binary classification of pathologic fractures. The strength of agreement was classified on an interval scale based on a kappa value from 0.0 to 1.0.

RESULTS: The Fleiss kappa for interrater reliability was 0.54 (p < 0.001) for the first survey, indicating moderate agreement, and 0.64 (p < 0.001) for the second, reflecting substantial agreement according to established definitions of agreement. The Fleiss kappa was 0.47 (p < 0.001) for the description of imaging findings, reflecting moderate agreement. We found that nearly every surgeon agreed that images from patients with a complete visible fracture line or bone displacement represented a pathologic fracture. Similarly, patients whose images showed bone erosion but no obvious complete cortical disruption were determined not to represent a pathologic fracture. The areas of major disagreement among orthopaedic oncologists in image interpretation occurred when there was (1) a cortical perforation with a soft tissue component from the underlying tumor, (2) a severely compromised cortex without a clear discontinuity, or (3) an incomplete fracture line or callus in a portion of the cortex.

CONCLUSION: Often raters agreed on the imaging findings but drew opposite conclusions regarding whether a pathologic fracture was present based on sets of images that replicate common clinical scenarios. This disproportionality of responses is too great to have common language for communication and clear surgical indications. This indicates that further review and consensus aimed at agreement on imaging findings based on these data are needed to allow for a clear definition of a pathologic fracture that can be universally applied. Our findings have stimulated early work for a multispecialty consensus project within the MSTS, which aims to classify skeletal lesions for use in clinical communication, research, and registries.

LEVEL OF EVIDENCE: Level IV, diagnostic study.