Predictive criteria for MRI-based evaluation of response both during and after radiotherapy for cervical cancer

PURPOSE For cervical carcinoma, the presence of persistent disease after radiotherapy (RT) is a significant predictor for survival. To date, no standard protocol is available to evaluate a response. This study was performed to assess magnetic resonance imaging (MRI) to evaluate presence of local residual disease during and after RT for Federation of Gynecology and Obstetrics (FIGO) stage Ib1-IVa cervical cancer. MATERIAL AND METHODS Forty-two patients were included. Patients underwent MRI before external beam RT, at final intracavitary brachytherapy (BCT) and 2-3 months after completion of RT. Two blinded radiologists (observer 1: experienced, observer 2: less experienced) scored the likelihood of residual tumor. Magnetic resonance imaging was evaluated by means of (a) 'subjective' visual evaluation of T2 weighted MRI images, and (b) 'objective' visual evaluation of T2 weighted MRI images according to predefined imaging criteria. RESULTS Seven patients had residual disease. Area under the receiver operating characteristics curve (AUC) for 'subjective' visual assessment was 0.79/0.75 (observer 1/observer 2) after RT and 0.75/0.43 at final BCT. The combined 'objective' MRI criteria (isointense, nodular, and irregular) resulted in improved prediction of residual tumor (AUCs of 0.91/0.85 after RT). For the less experienced observer, the MRI criteria set significantly improved prediction of residual tumor compared to 'subjective' visual assessment. Observer dependency decreased, kappa of 0.41 compared to 0.84 for the MRI criteria set after RT. CONCLUSION Compared to 'subjective' visual assessment, predefined 'objective' MRI criteria increase diagnostic performance and decrease observer dependency for assessing residual tumor after RT in cervical cancer.