Prognostic and Diagnostic Added Value of Medical Imaging in Gynecological Cancer (PRODIGYN)

Overview

The goal of this observational study is to learn about the added diagnostic and prognostic value of advanced medical imaging procedures in cervical cancer, endometrial cancer and ovarian cancer. The main questions it aims to answer are: * Does advanced medical imaging predict survival? * Can advanced medical imaging improve radiotherapy target planning? * Are advanced medical imaging results associated with risk markers found in tumor tissue? Participants will * Undergo four additional imaging procedures, as compared to clinical routine examinations, two at baseline and two after three months. * Be subject to clinical follow-up for five years.

This study has a retrospective and a prospective part, where the main aims are to: 1. Retrospectively validate the added value of radiological staging to clinical staging according to the International Federation of Gynecology and Obstetrics (FIGO) tumor classification system, in cervical cancer, endometrial cancer, and epithelial ovarian cancer. 2. Prospectively identify prognostic biomarkers with 18F-2-fluoro-2-deoxy-D-glucose fluorodeoxyglucose (FDG)-positron emission tomography (PET)/CT and FDG-PET/MRI in cervical cancer, endometrial cancer, and epithelial ovarian cancer 3. Assess the possible effect of PET/MRI on radiotherapy target delineation in cervical cancer 4. Improve non-invasive lymph node staging in endometrial cancer 5. Develop a machine learning decision support tool for characterization of ovarian lesions Material and methods (retrospective): All eligible patients from the multi-disciplinary gynecological tumor conference at Umea University Hospital during 2013-2022, with newly diagnosed cervical, endometrial, or epithelial ovarian cancer, known cFIGO, \>18 years old, and no other known current or previous malignancy within the last 10 years, will be included in a retrospective evaluation of radiological stage (rFIGO) based on all pre-operative imaging (MRI, CT and FDG-PET/CT), clinical stage (cFIGO) based on examination under anesthesia (EUA), and histopathological stage (pFIGO) based on available surgical and histopathological findings. The analysis will be carried out in two cohort groups - 2016-01-01-2018-05-31, and 2018-06-01-2022-06-01, before and after the implementation of the 2018 revised FIGO classification, after which the cFIGO may be influenced to larger extent by imaging results. For all epithelial ovarian cancer patients, Ovarian-Reporting and Data System (O-RADS) score will be annotated for each MRI examination. Agreement between rFIGO and cFIGO will be evaluated, and if feasible, compared to pFIGO. The investigators will thus be able to validate rFIGO in cervical cancer with cFIGO up to Ib2, and in endometrial and epithelial ovarian cancer treated with surgery. The added value of rFIGO in terms of metastasis assessment and change of therapy, as well as pattern and incidence of radiotherapy side effects will be evaluated in patients who were considered inoperable. Hypotheses (retrospective): 1. The degree of agreement is high between rFIGO T stage and cFIGO T stage in cervical, endometrial, and epithelial ovarian cancer. 2. There is high sensitivity, specificity, accuracy, and negative and positive predictive values of rFIGO to predict pFIGO in ovarian cancer of epithelial subtype. 3. There is an added value of rFIGO for metastasis assessment and change of patient management in cervical cancer stages \>Ib2, and in endometrial and epithelial ovarian cancer patients who are considered inoperable. Material and methods (prospective) All eligible patients with newly diagnosed cervical cancer stage \>1a, endometrial cancer type 2 and/or minimum stage 1, or strongly suspected epithelial ovarian cancer, consecutively referred to the gynecological- oncological department of Umea University Hospital, with written informed consent, will be included in a prospective study of the diagnostic and prognostic value of FDG-PET/CT and FDG-PET/MRI at baseline and at therapy response evaluation after 3 months. The subgroup of patients with cervical and endometrial cancer treated with radiotherapy, will undergo one additional stand-alone MRI with dedicated tumor protocol after one week of treatment for early response evaluation. Patient demographics and age of menarche, menopause and parity will be collected to characterize the study population. Furthermore, for epithelial ovarian cancer, levels of tumor markers cancer antigen (CA)-125 and CA-19-9 as well as risk of malignancy index will be collected. The FDG-PET/CT will be performed according to clinical routine with intravenous injection of FDG 3 megabecquerel (MBq)/kg, 60 minutes post-injection (with the addition of Sharp Inversion Recovery (IR) reconstruction to be used for comparison with the FDG-PET/MRI), but without intravenous iodine contrast agent, since the FDG-PET/MRI will be performed 120 minutes after the same FDG-injection and will be prioritized for administration of gadolinium-based contrast agent. The FDG-PET/MRI will be designed according to standard clinical MRI protocol, dedicated for each cancer type as described in detail below, with preparatory administration of 2 ml Buscopan 20 mg/ml and gadolinium-based contrast agent Dotarem 279.3 mg/ml, 0.2 ml/kg body weight (maximum 20 ml). If renal function is moderately impaired (relative GFR 45-59 ml/min/1.73 m2), the dose will be reduced to 0.1 ml/kg. If relative GFR is \<45 ml/min/1.73 m2 the examination will be performed without iv contrast agent. The total examination time is estimated to approximately 40 minutes. Cervical cancer: T2-weighted (T2W) (sagittal, axial, coronal oblique, axial oblique), T1 Dixon all (axial), diffusion-weighted imaging (DWI) (b 100, 800, axial), optional Gd+ T1 Dixon (axial). Endometrial cancer: T2W (sagittal, axial, axial oblique), T1Dixon all (axial oblique), DWI (b 100, 800, axial oblique), Gd+T1 Dixon (axial oblique, sagittal oblique). Ovarian cancer: T2W (sagittal, axial, coronal), T1 Dixon all (axial), DWI (b 100, 800, axial), Gd+T1Dixon (axial, sagittal). Clinical evaluation will take place at 3 months, 6 months, 1 year and 5 years after start of treatment with collection of clinical data progression-free survival (PFS, defined as the time from start of treatment to progression or specific cancer-related death), overall survival (OS, defined as the time from start of treatment to death from any cause), and pattern and incidence of any radiotherapy side effects. In FDG-PET/CT, pathological uptake of the suspected primary tumor will be visually categorized into 1 = uptake \< mediastinal background, 2 = uptake \> mediastinal background and \< liver background, 3 = moderate uptake \> liver background, or 4 = intense uptake \> liver background. From the PET/CT and PET/MRI examinations, primary tumor PET parameters will also be quantified in maximum standardized uptake value (SUVmax), mean standardized uptake value (SUVmean), functional tumor volume (FTV) and total lesion glycolysis (TLG). In addition, the categorical parameters tumor heterogeneity, suspected radiological lymph node metastases (present or not, N1 or N0) will be reported for both, and distant metastases (M1 or M0) will be reported for PET/CT. CT and MRI parameters volume, delineation, contrast enhancement and diffusion restriction, as well as tumor heterogeneity will also be assessed. Interpretation of rFIGO will be reported for both PET/MRI and PET/CT. At the 3 months´evaluation, the same imaging parameters will be evaluated and absolute differences in continuous parameters as well as up-grading or down-grading of categorical parameters will be analyzed. The patients treated with radiotherapy or chemotherapy will be categorized into responders, defined as complete or partial metabolic response, and non-responders, defined as stable metabolic disease or progressive metabolic disease, according to PERCIST criteria (see References). The feasibility of FDG-PET/MRI for radiotherapy dose planning guidance will be compared to standard imaging-based guidance regarding target delineation of gross tumor volume (GTV), and the prognostic difference between the group of early responders (any perceptible response) at one week´s stand-alone MRI evaluation, compared to non-responders (stable or progressive disease), will be assessed. In the histopathological analysis, prognostic factors will be recorded and if applicable, immunohistochemical stainings for P53, Ki-67, ER, D240 and CD31, as well as molecular analysis of microsatellite instability (MSI), breast cancer susceptibility gene (BRCA)-, and polymerase-epsilon (POLE)-mutations and possible additional genes of emerging interest will be performed. For the study participants with endometrial cancer scheduled for surgery with sentinel node algorithm, imaging characteristics of suspected lymph nodes will be described in terms of visually quantified pathological FDG-PET uptake according to the four previously mentioned categories, and PET parameters SUVmax, SUVmean, FTV, TLG and tumor heterogeneity. CT and MRI parameters size, shape, delineation, contrast enhancement, diffusion restriction and tumor heterogeneity will also be assessed. The lymph node with the highest metabolic activity (SUVmax) will be selected for each affected lymph node region: external iliac, internal iliac, common iliac, obturator and infrarenal paraaortic regions. In addition, the same parameters will be analyzed for the primary tumor to evaluate its predictive value of lymph node metastases. Regarding histopathology in this sub-study, as a starting point morphological patterns detected on hematoxylin-eosin stained glass will be recorded. These patterns will then guide further immunohistochemical and molecular analyses to highlight the changes that have occurred in the metastatic lymph nodes. For the ovarian cancer dataset, the investigators will develop a machine learning method for diagnostic decision support and prognostic prediction. The modeling data set will consist of the various MRI data from different MRI scanners and protocols, annotated with O-RADS (MRI), from ovarian cancer patients from the previous retrospective part of the PRODIGYN study. The matching dataset of controls will be acquired from the non-ovarian (cervical and endometrial) cancer patient cohort from the above-mentioned retrospective study. After training, validation and testing, the investigators will apply the machine learning method for O-RADS (MRI) risk categorization on the prospective study dataset and compare the diagnostic performance of the machine learning method with two radiologists, by area under the receiver operating characteristic curve (AUC-ROC) analysis, with ground truth histopathology. The prognostic predictive performance will be assessed using O-RADS 4 and 5 lesion labeling as markers of poor prognostic outcome, with ground truth PFS and OS. Hypotheses (prospective): 1. FDG-PET/CT and FDG-PET/MRI biomarkers can predict PFS and OS in cervical, endometrial, and epithelial ovarian cancer 2. FDG-PET/CT and FDG-PET/MRI metrics at follow-up of therapy response have higher prognostic impact than baseline 3. Early tumor response on MRI after radiotherapy predicts better prognosis 4. Early response patterns in organs at risk may predict serious adverse events 5. Target delineation of GTV in cervical cancer is significantly different with FDG-PET/MRI compared to local standard MRI 6. Degree of agreement, sensitivity, specificity and accuracy of FDG-PET/CT and FDG-PET/MRI are high for lymph node metastases on regional and on patient basis in endometrial cancer 7. Primary tumor FDG-PET/CT and FDG-PET/MRI imaging characteristics can predict aggressive histological type II, MSI phenotype and presence of lymph node metastases in endometrial cancer 8. FDG-PET/MRI can be used to distinguish BRCA-mutated from non-BRCA-mutated ovarian cancer by differences in growth pattern and metabolic activity 9. The histopathological immune response in sentinel nodes can predict prognosis and correlate with FDG-PET/CT and FDG-PET/MRI biomarkers in endometrial cancer 10. There is an added value of FDG-PET/CT and FDG-PET/MRI to the sentinel node algorithm, for detection of para-aortic lymph node metastases in endometrial cancer 11. The machine learning method performs similarly to radiologists in O-RADS 1-4 but is inferior in O-RADS 5