Soft Tissue Sarcoma Response to Two Cycles of Neoadjuvant Chemotherapy: A Multireader Analysis of MRI Findings and Agreement with RECIST Criteria and Change in SUVmax

Rationale and Objectives: When soft tissue sarcomas are treated with neoadjuvant chemotherapy, the number of cycles of chemotherapy is usually dependent on the tumor's initial response. Popular methods to assess tumor response include Response Evaluation Criteria in Solid Tumors (RECIST) criteria, which rely solely on tumor size, and maximum standardized uptake value (SUVmax) reduction in positron emission tomography (PET), which requires an expensive and high radiation test. We hypothesized that contrast enhanced magnetic resonance imaging (MRI) may offer a good alternative by providing additional information beyond tumor size. Materials and Methods: Following IRB approval, a retrospective review identified patients with soft tissue sarcomas who underwent both PET and MRI before and after two cycles of neoadjuvant chemotherapy. Five readers independently examined the MRI exams for: changes in size, T2 or T1 signal, necrosis and degree of enhancement. Readers then made a subjective binary assessment of tumor response to therapy. Each reader repeated the anonymized randomized reading at least 2 weeks apart. 18 F-FDG PET exams were interpreted by a nuclear medicine specialist. The maximum standardized uptake values (SUVmax) for pre and post-chemotherapy exams were compared. Intra- and inter-reader agreement was assessed using Cohen's kappa and Light's kappa, respectively. Results: Twenty cases were selected for this multireader study, of which 9 (45%) were responders and 11 were nonresponders by SUVmax. Using all MRI criteria, 43% were classified as responders based on MRI and 1.5% were classified as responders by RECIST criteria. Using PET as the reference, the sensitivity and the specificity of the MRI diagnosis for response using all findings were 50% and 63%, respectively. There was fair to moderate intrareader (kappa = 0.37) and inter-reader (kappa = 0.48) agreement for the MRI diagnosis of response. None of the individual MRI signal characteristics were significantly different between the PET responders and nonresponders. Additionally, no MRI findings were significantly different between those with and without good clinical responses. Conclusion: By our assessment, there is a poor correlation between tumor response by RECIST criteria and PET SUVmax. In addition, varying MR features did not help in diagnosing tumor response. Imaging of tumor response remains a challenging area that requires further research.