Accuracy of MR imaging for detecting epiphyseal extension of osteosarcoma

Background. Too few patients are receiving epiphyseal-sparing limb salvage procedures for osteosarcoma. Objective. To determine how magnetic resonance (MR) imaging can best predict the epiphyseal extension of osteosarcoma. Materials and methods. Forty children underwent complete pretreatment static and dynamic contrast-enhanced MR imaging (DEMRI). Static MR images T1-weighted and short tau inversion recovery (STIR)] of the epiphyses were read in three ways: (1) for suspicion of any abnormality (tumor or edema), (2) for suspicion of tumor, excluding suspected edema, and (3) validating the second method by using a scale to rate the likelihood of tumor. Presentation imaging was compared to histopathologic findings after chemotherapy and resection. The receiver operating characteristic (ROC) method was used to analyze the scaled ratings of static MR and DEMRI values. Results. At delayed resection, 20 of 40 children with osteosarcoma had confirmed epiphyseal tumor; however, 32 epiphyses were abnormal on STIR and 28 abnormal on T1. Differentiating suspected tumor from edema increased the accuracy to an A_z (area under the ROC curve) of 0.94 for both T1-weighted and STIR static sequences. T1-weighted MR had better specificity and STIR better sensitivity at any given rating. DEMRI was slightly less accurate (Az = 0.90). Conclusion. Static MR imaging most accurately detected epiphyseal extension of osteosarcoma when readers distinguished suspected tumor from edematous or normal tissue. Received: 20 August 1999 Accepted: 13 January 2000