Accuracy of plain radiographs in diagnosing biopsy-proven malignant bone lesions.

Background: The diagnosis of primary bone tumours is a three-fold approach based on a combination of clinical, radiological and histopathological findings. Radiographs form an integral part in the initial diagnosis, staging and treatment planning for the management of aggressive/malignant bone lesions. Few studies have been performed where the radiologist's interpretation of radiographs is compared with the histopathological diagnosis.

Objectives: The study aimed to determine the frequency of bone tumours at a tertiary hospital in South Africa, and, using a systematic approach, to determine the sensitivity and specificity of radiograph interpretation in the diagnosis of aggressive bone lesions, correlating with histopathology. We also determined the inter-observer agreement in radiograph interpretation, calculated the positive and negative predictive values for aggressive/malignant bone tumours and computed the cumulative effect of multiple radiological signs to determine the yield for malignant bone tumours.

Method: A retrospective, descriptive and correlational study was performed, reviewing the histopathological reports of all biopsies performed on suspected aggressive bone lesions during a 3-year period from 2012 to 2014. The radiographs were interpreted by three radiologists using predetermined criteria. The sensitivity and specificity of the readers' interpretation of the radiograph as 'benign/non-aggressive' or 'aggressive/malignant' were calculated against the histology, and the inter-rater agreement of the readers was computed using the Fleiss kappa values.

Results: Of the 88 suspected 'aggressive or malignant' bone tumours that fulfilled the inclusion criteria, 43 were infective or malignant and 45 were benign lesions at histology. Reader sensitivity in the diagnosis of malignancy/infective bone lesions ranged from 93% to 98% with a specificity of 53% - 73%. The average kappa value of 0.43 showed moderate agreement between radiological interpretation and final histology results. The four radiological signs with the highest positive predictive values were an ill-defined border, wide zone of transition, cortical destruction and malignant periosteal reaction. The presence of all four signs on radiography had a 100% yield for a malignant bone tumour or infective lesion.

Conclusion: The use of a systemic approach in the interpretation of bone lesions on radiographs yields high sensitivity but low specificity for malignancy and infection. The presence of benign bone lesions with an aggressive radiographic appearance necessitates continuation of the triple approach for the diagnosis of primary bone tumours.