Abstract
Background: The aim is to propose a model for predicting occult paraaortic lymph node (PALN) involvement in locally advanced cervical cancer (LACC) patients by including parameters such as reconstruction detection technology (use of time-of-flight) and parameters related to the primary tumor. This model will then be compared with the scores used in routine clinical practice; Methods: This retrospective observational cohort study included patients diagnosed with LACC who underwent (18)F-FDG PET/CT prior to PALN surgical staging between February 2012 and May 2020. The following parameters were collected on PET/CT: tumor SUVmax, tumor MTV, number of common and distal pelvic node involvements. A multivariate regression analysis estimating the probability of PALN involvement was performed, with optimal thresholds determined via ROC curves; Results: In total, 71 patients met the inclusion criteria. Occult PALN involvement was detected in 12.7% of patients. A derived multivariate PET model selected four variables: number of common and distal iliac lymph nodes (OR 5.9 and 2.7, respectively), tumor-to-liver SUV ratio (OR 0.9) and the use of time-of-flight technology (OR 21.4 if no time-of-flight available). At the optimal threshold, a sensitivity of 77.8% and specificity of 88.7% was found. The model's performances varied significantly between patients whose PET/CT used time-of-flight and those whose PET/CT did not. No significant differences were found between our model and the one used in clinical practice (p = 0.55); Conclusions: This study shows that PET/CT technology influences the ability to detect occult PALN involvement in LACC. This parameter should be considered in the regular revision of PET-based scores.