endometrioid tumors and grade 3 endometrioid disease, carries mark-edly poorer prognosis [3]. However, due to the much higher incidenceof endometrioid tumors, the absolute number of recurrences is signifi-cant also in this group [4,5]. Identification of biomarkers that can aid se-lection of patients for optimal surgical and adjuvant treatmentindependent of histological parameters is vital to improve outcome.Several prognostic and diagnostic biomarkers have been identifiedin tumor biopsies [6] but few have so far been implemented in the clinicto improve treatment for endometrial cancer patients. In some institu-tions, hormonereceptor status is assessedasa supplementto traditionalhistological evaluation, and the design and validation of combined mo-lecular classifiers is ongoing, driven by initiatives like ProMisE andTransPORTEC [7,8]. Compared to these tissue-based biomarkers,blood-based biomarkers do not require a biopsy and thus representless invasive clinical tools to predict prognosis and to plan patient treat-ment, and pose little technical challenge in implementation. Severalblood biomarkers have alreadybeen investigatedin endometrialcancer,e.g. Ca-125 is shown to have prognostic value and identifies advanceddisease and lymph node metastasis [9]. Other blood based biomarkerssuch as HE4, GDF-15, and DJ-1 have also been found promising[10–12], but lack validation in a prospective implementation setting.Although the influence of hormone receptor expression on progno-sis has been extensively researched [13–15], few studies have evaluatedthe importance of endogenous steroid levels other than estrogen me-tabolites [16–18], and to some extent androstenedione (A4) and testos-terone (T) [19–22]. These studies have mostly focused on risk ofacquiring disease rather than biomarker properties. We have previouslydemonstrated differences in levels of several steroids in blood samplesin a matched patient series of long vs short surviving endometrial can-cer patients [23].For clinical implementation of a blood-based test, easy and reliablemethods for detection are vital. In this study, we measured levels of cir-culatingsteroids in endometrial cancer patients by Liquid Chromatogra-phy Tandem Mass Spectrometry (LC-MS/MS). We selected a panel ofrelevant steroids related to sex hormonesynthesis, used clinically for di-agnosing endocrinological disorders. This panel was supplementedwith measurements of estrone (E1) and estradiol (E2) using a novelsensitive protocol, properly quantifying postmenopausal estrogenlevels in plasma. We explored the relationship to body mass, fat distri-bution variables, associations to clinicopathologic characteristics of thedisease and patientsurvival. Finally, we analyzed differences in geneex-pression to identify links between host steroid levels and tumor biology.The aim of the study was to evaluate the prognostic value of circulatingsteroid levels in endometrial cancer patients.2. Methods2.1. Ethical considerationsThe study has been approved according to Norwegian legislation bythe Western Regional Committee for medical and health ResearchEthics (REK 2009/2315, REK 2014/1907, REK 2018/594, REK 2019/1020). All included patients gave written informed consent.2.2. Patient seriesA population based endometrial cancer patient series was prospec-tively collected from 2001 to 2015 in Hordaland County (Norway). Pa-tients were surgically staged according to the International Federationof Gynecology and Obstetrics (FIGO) 2009 criteria. Clinical and patho-logical variables including age at diagnosis, FIGO stage, histological sub-type, grade, and follow up data were collected by review of medicalrecords as previously described [24]. From this series, 100 postmeno-pausalpatients included from 2009 to 2013 were selected to reflect clin-icopathological characteristics of the whole prospective cohort(Supplementary Table S1). Median follow-up was 67 months (range1–116) and minimum 60 months for all survivors. During follow-up,20 cancer specific deaths were registered, and six patients were cen-sored due to non-cancer deaths.Immunohistochemical staining and evaluation of hormone recep-tors has been performed previously for this cohort [14,25–27]. Expres-sion data for estrogen receptor (ERα), progesterone receptor (PR),Androgen Receptor (AR) and Glucocorticoid Receptor (GR) was avail-able for 93, 94, 87 and 76 patients, respectively.2.3. Steroid analysisEDTA-blood was obtained before primary surgery, and prior to ad-ministration of any anesthetic medications, from 100 patients with en-dometrial cancer. The blood samples were centrifuged at 1600gfor15min and the plasma was stored at−80°C. Median storage time be-fore analysis was 66 months (range 47–104 months). Steroids weremeasured at the Hormone Laboratory, Haukeland University Hospital,Bergen Norway, using the routinely applied LC-MS/MS method forplasma analysis of 17-hydroxyprogesterone (17-OHP), 11-deoxycortisol (11-DOC), cortisol (CORT), androstenedione (A4) andtestosterone (T) previously described [28]. Briefly, isotope-labeled in-ternal standards were added to 85μL plasma and processed by liquid-liquid extraction. The steroids were resolved by ultra-high-pressurechromatography on a reverse phase column, and detected by triple-quadrupole mass spectrometry. For analysis of estrone (E1) and estra-diol (E2), a recently developed optimized protocol was used, with limitsof quantification of 0.3 pmol/L and 0.6 pmol/L respectively, thusallowing quantification within the postmenopausal ranges of these hor-mones [29]. Both LC-MS/MS methods are accredited according to ISO15189:2012.17-OHP, CORT, A4 and T were measured for all 100 patients. For 11-DOC, 98 patient samples were measured, two samples were not ana-lyzed due to technical difficulties. For six patients plasma level of 17-OHP was below the detectable threshold of 0.2 nmol/L. For one ofthese patients A4 and T were also below the threshold (0.2 nmol/Land 0.1 nmol/L respectively). These levels were set to the lowest detect-able value for each steroid and included in non-parametric analyses.Plasma level of progesterone was measured, but was below measure-ment threshold for all patients (b0.5 nmol/L) and was subsequentlyexcluded.E1 and E2 plasma levels were obtained from 96 patients. No valueswere below analytic range, measurements above the analytic range(above the highest calibrator,n=7forE1andn= 3 for E2) were ana-lyzed as ranked values in non-parametric analysis.2.4. Estimation of fat distribution from CT scansComplete diagnostic abdominal contrast-enhanced Computer To-mography (CT) scans were available for 83 patients and evaluated forassessment of abdominal fat volumes as previously described [23]. Thesoftware iNtuition (TeraRecon Inc.; San Mateo, CA, USA), was used toanalyze cross-sectional CT images from the upper right diaphragm toL5/S1-level, segmenting pixels with values for Hounsfield units (HU)corresponding to adipose tissue (−195 to−45 HU). If necessary, thecorrect segmentation between visceral and subcutaneous fat compart-ments was adjusted by the operator. Both the visceral abdominal fatvolume (VAV; cm3) and the subcutaneous abdominal fat volumes(SAV; cm3) were estimated, and the sum of these was the total abdom-inal fat volume (TAV; cm3). The percentage of visceral fat was calculated([VAV/TAV] x 100; VAV%). In addition, waist circumference was mea-sured in an axial image at the L3/L4 level.2.5. Gene expression analysisGene expression data from tumor tissue was available for all in-cluded patients and has been published previously [25]. Briefly, RNA401D. Forsse et al. / Gynecologic Oncology 156 (2020) 400–406