Patterns of Care and Outcomes of Adjuvant Radiation Therapy for Stage II Endometrioid Uterine Cancer: A Surveillance, Epidemiology, and End Results (SEER) and Medicare Linked Analysis

Abstract

We investigated whether a biological and radiological score, derived from neutrophilia and quantitative imaging biomarkers extracted from FDG-PET scans, could predict outcome in locally advanced cervical cancer (LACC) patients treated with definitive chemoradiation plus image-guided adaptive brachytherapy (IGABT). Materials/Methods: We included retrospectively and consecutively patients treated in our institution between 2006 and 2013 with both baseline FDG-PET/CT and complete blood count prior to any treatment. We divided patients into two groups depending on the PET device used: the training set (TS) and the validation set (VS). Tumors were semi-automatically delineated by applying a threshold of 40% of maximum SUV (Standardized Uptake Value) in a manually drawn region of interest. Images from TS were resampled to VS-voxel grid to reduce differences between devices. LIFEx software was used to calculate 42 conventional and textural imaging features. The optimal index for recurrence prediction was selected using time-dependent Area Under the Curve (td-AUC) for 3-year Local Control (LC) and cutoff using Youden Index. We defined the "Neutrophil SUV Grade" (NSG) score according to the number of risk factors (i) neutrophilia; (ii) high-risk radiomic index: 0 if none, 1 if one, 2 if both. The prognostic value of the proposed score was evaluated for LC and overall survival (OS). Results: 108 patients were identified. Estimated 3-year LC was 72% (95% CI: 62e84%) in TS (nZ69), 65% (95%CI: 51e82%) in VS (nZ39). Using 7 G/L cutoff to define neutrophilia, 19 patients (27%) and 12 patients (31%) had neutrophilia in TS and VS respectively (pZ0.72). Td-AUC for neutrophil count to predict 3-year LC was 0.61 in TS. In TS, SUV peak (defined as the maximum average SUV within a 1 mL sphere located in the tumor) was selected as the strongest index associated with 3year LC (td-AUCZ0.75), and optimal cutoff was 10. There was no interaction on LC between neutrophilia and high-risk SUV peak (pZ0.074). In univariate analysis, hazard ratio (HR) for worse LC was 2.6 for neutrophilia, 4.4 for high-risk SUV peak , and 9.2 for both (NSGZ2) in TS. In multivariate analysis, both neutrophilia (HRZ3.0, pZ0.017) and high-risk SUV peak (HRZ1.4, pZ0.007) independently predicted poor LC in TS. NSG scoreZ2 independently decreased LC (HRZ7.5, p<0.001) and OS (HRZ5.8, pZ0.001) in TS. NSG scoreZ2 was validated in VS (HRZ5.2 for OS and 3.5 for LC, p<0.02). Conclusion: In our series, SUV peak was superior to other individual radiomic indices. In LACC, NSG score combining neutrophilia and FDG-PET-derived SUV peak predicts worse OS and LC. FDG-PET-derived SUV peak combined with neutrophilia may be an independent biomarker that could allow the early identification of patients with poor outcome to drive personalized therapeutic strategies.