Inter-observer Agreement and Reproducibility of Pertrochanteric Fracture Classification Using Plain Radiograph Versus Computed Tomogram Images: A Study of 523 Patients.

Background A precise preoperative imaging classification system for pertrochanteric fractures is imperative due to the reported unreliability of the current classification system, which relies solely on plain radiographs. This study aims to achieve two primary objectives: (i) elucidate the reproducibility of pertrochanteric fracture evaluation based on the Revised Arbeitsgemeinschaft für Osteosynthesefragen/Orthopedic Trauma Association (AO/OTA) Classification, comparing plain radiographs and computed tomography (CT) scan images, and (ii) investigate the consistency of fracture classification between both imaging modalities. Methods A total of 523 patients (112 males and 411 females, mean age 85 years) who had both preoperative plain radiographic and three-dimensional CT images were enrolled in this study. Following the Revised AO/OTA Classification, three individual observers initially classified the fractures in plain radiograph images as either Stable (A1) or Unstable (A2). Subsequently, they further categorized them into five sub-categories (A1.1, A1.2, A1.3, A2.2, and A2.3). The same classification system was applied to the CT scan images. Inter-observer agreement and consistency of fracture classification between plain radiographs and CT scan images were assessed. Results The inter-observer agreement for fractures classified as stable or unstable using only plain radiographs was found to be fair among the three observers, with a mean κ of 0.397 (95% CI: 0.316-0.478). However, inter-observer agreement improved significantly when using CT scans, with a mean κ of 0.590 (95% CI: 0.518-0.662). Our results demonstrated a consistency level between two graphical modalities ranging from fair to moderate, with κ values of 0.581, 0.383, and 0.335, respectively. It's worth noting that plain radiographic classification occasionally resulted in underestimations, with each observer identifying 16.1%, 34.0%, and 37.9%, respectively, of cases as A1 in plain radiographs that were classified as A2 in CT scans. Conclusions This study reveals a moderate to substantial level of inter-observer agreement for fracture classification when using CT scan images, in contrast to plain radiographs. Fracture evaluation relying solely on plain radiographs sometimes underestimates fracture classification and exhibits less consistency compared to using CT scan images.