{"paper_id":"f0d69d39-54e7-415c-83a5-f010996dd709","body_text":"1 \n \nRunning title:  In-vitro diagnostic test for endometriosis  \nArticle title: Development and Validation of a novel in vitro diagnostic test for \nendometriosis \nAuthors: Bárbara Herranz-Blanco PhD. a*, Elza Daoud PhD. a, Paola Viganò PhD b, Juan Antonio \nGarcía-Velasco MD., PhD.c, Enrico Colli MD.a \n \na Exeltis Healthcare, Madrid, Spain. \nb Infertility Unit, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milano, Italy \nc IVIRMA, Madrid, Spain, Department of Obstetrics and Gynaecology, Rey Juan Carlos University, \nMadrid \n \n \nCorresponding author:  \nAutho r Name :  Bárbara Herranz-Blanco  \nAf fili ati on:  Exeltis Healthcare, Madrid, Spain.  \nMa iling A d dres s : MA N UEL PO M BO A N G ULO, 28, PLANTA 3, MA DRI D 28050 , Spai n \nP hone : +34  680269 484 \nE mail:  barbara.herranz@exeltis.com \n \nArticle type: Study of Screening and Diagnostic Tests \nFunding Statement: Exeltis (represented by Chemo Research S.L.) has fully sponsored the \nstudies. \nDisclosure Statement:  \nAttestation statements: \n• The subjec t s in thi s tri al ha ve n ot  conc o mitantly be en inv olved in othe r rand omized tri als .  \n• Dat a r e gar ding a ny of th e subje ct s  in th e stu dy ha s  no t be en previ ou s ly  publi she d u nle ss  \ns p e cif i e d.   \n•  D a t a  will  be ma d e av ailabl e to th e edi t or s o f the jour nal  fo r  r e view  or qu ery upo n r e que st.  \nDisclosure Statement:  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n2 \n \nWord count abstract: 311 words  \nWord count text : 3700 words \n \n \n  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n3 \n \nCapsule \nA novel in-vitro diagnostic test for endometriosis was developed and tested. \nThis excellent rule-in test is intended to speed up the diagnosis, which was, until \nnow, significantly delayed in time.  \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n4 \n \nSTRUCTURED ABSTRACT  \nObjective: To develop a non-invasive diagnostic test for endometriosis. \nDesign: We conducted two studies: the development study (study 1) aimed at (i) evaluating the ability \nof CA125, brain-derived neurotrophic factor ( BDNF) and clinical variables in segregating between \ncases and controls and (ii) developing a diagnostic algorithm based on those results. In the validation \nstudy (study 2), the clinical performance of the developed in vitro diagnostic (IVD) test, in diagnosing \nendometriosis was validated. \nInterventions: Serum samples and clinical variables extracted from psychometric questionnaires were \ncollected from the Oxford Endometriosis CaRe Centre biobank (UK). Case/control classification was \nperformed based on laparoscopy and histological verification of endometrial glands and/or stroma in \nthe excised lesions.  \nMain outcome measures: Study 1 and 2 included n=204 and n=79 patients, respectively.  CA125 and \nBDNF concentrations were determined using the endometriosis IVD ELISA kit. In study 2, serum \nCA125 and BDNF concentrations and clinical variables were introduced into the IVD test software \nhosting the data treatment algorithm, which generates the qualitative diagnostic result (“positive” or \n“negative”). \nResults:\n \nResults from study 1 showed that, for both biomarkers levels, a statistically significant difference was \nfound between cases and controls. Among all clinical variables related to patients’ medical history \nconsidered, six were significantly different between cases and controls: record of a previous surgery to \ninvestigate endometriosis presence, painful periods leading to referral for endometriosis presence, \nseverity of menstrual pain during last cycle, age at first experience of intercourse pain, age at first \nregular use of painkillers and age at first diagnosis of ovarian cyst. In study 2, the novel endometriosis \nIVD test demonstrated sensitivity and specificity values of 46.2% (25.5-66.8%) and 100% (86.7-\n100%), respectively.  \nConclusion:\n \nBDNF and CA125, together with patient’s clinical variables allowed efficient segregation between \ncontrols and endometriosis cases. Due to its high specificity, the novel endometriosis IVD test is an \naccurate rule-in non-invasive method, potentially contributing to diagnose endometriosis. \nKey words: endometriosis, CA125, biomarker, diagnosis, BDNF, in vitro.   \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n5 \n \nArticle length target:  \n \nIntroduction: \n \nEndometriosis is an estrogen-dependent disease characterized by the growth of endometrial-like \ntissue outside the uterus (1,2). These lesions cause a chronic inflammatory reaction, which can \nlead to the generation of scar tissue and adhesions (3). Clinical symptoms include chronic pelvic \npain, dysmenorrhea, and infertility (4). Endometriosis may increase a woman’s risk for chronic \ndiseases such as cancer or autoimmune disorders and overall morbidity \nas well (5–7). \nEndometriotic lesions can occur at different locations, including the pelvic peritoneum and the \novary, or infiltrate pelvic structures below the peritoneal surface (deep endometriosis). According \nto these locations, three primary types of endometriosis have been defined: superficial peritoneal \nlesions (typically located on the pelvic organs or pelvic peritoneum), ovarian endometriomas and \ndeep infiltrating endometriosis (DIE) (8).Endometriosis affects at least 10% of women of \nreproductive age and is associated with a high societal and economic burden: the average annual \ncost of healthcare and loss of productivity due to pain from endometriosis was $11,300 for \naffected women from the United States and nine European countries (9).\n \n \nThe clinical examination of symptomatic women does not reliably predict the presence of \nendometriosis. Imaging techniques often fail to diagnose the disease, mainly in early-stages or \nwhen only superficial peritoneal lesions are present. Imaging techniques are recognized to be \nuseful in the identification of endometriomas, and in some cases of deep endometriosis.(1). \nMoreover, regardless of the lesion type and location, the interpretation of imaging findings is \nhighly dependent of clinician’s experience and skills, which hinders diagnosis using those \nmethods (10). Generally, women for whom there is high suspicion of endometriosis receive \nanalgesics and hormonal medication without a prior definitive diagnosis (11). Diagnosing the \ndisease in these cases becomes only possible via an invasive laparoscopy accompanied with \nhistologic confirmation of lesions(10, 12). This may result in a delay in diagnosis ranging from 4 \nto 11 years on average between the first appearance of symptoms and the final confirmation of the \ndiagnosis (13). In this context, the development of a non-invasive diagnostic tool is essential for \nfaster diagnosis, appropriate treatment, and triaging potential surgical patients (14,15). \n \nMultiple biomarkers have been studied as screening and triage tests for endometriosis (16,17). \nHowever, none of them has been implemented routinely in clinical practice (10). Out of all \ncandidate biomarkers, Cancer Antigen 125 (CA125) has been extensively studied in \nendometriosis.  CA125 is a high-molecular-weight glycoprotein expressed on the cell surface of \nsome derivatives of embryonic coelomic epithelium, which are believed to be the precursors of \nendometriotic lesions (18). Studies found  CA125 levels to be higher in patients with \nendometriosis, indicating that CA125 can be a useful marker for diagnosing endometriosis, \ndistinguishing the severity of the disease, monitoring the effect of treatment and reflecting \nmalignant transformation (19–21).\n A meta-analysis on the diagnostic accuracy of CA125 for \nendometriosis pooling 22 studies, including 3626 participants, showed that CA125 performed \nwell as a rule-in test, but that a negative test result is unable to rule out endometriosis. In addition, \nthe study showed that CA125 was significantly mo re sensitive for the diagnosis of moderate or \nsevere endometriosis (stages III and IV) compared with minimal disease (22–24). Besides, CA \n125 \n≥  30 U/m was highly predictive of endometriosis in women with symptoms of pain and/or \nsubfertility, but CA 125/i2 <30 U/ml was unable to rule out the disease (25). \n \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n6 \n \nAnother interesting biomarker, brain-derived neurotrophic factor (BDNF), has been found to be \nlinked with several pathways that are disturbed in women with endometriosis. BDNF is a \nneurotrophin (26) with a high-a ffinity to neurotrophic tyrosine receptor kinase 2 (NTRK2) also \nknown as Tropomyosin receptor kinase B (TrKB), and this ligand-receptor pair participates in \nsome aspects of uterine physiology (27). BDNF and NTRK2 were more significantly expressed in \nthe uterus of women with endometriosis compared to disease-free controls (28); BDNF has been \nfound to be dysregulated and overexpressed in ectopic but not in eutopic endometrial tissue. \nInterestingly, BDNF is a down-stream effector of estrogens mediating the pro-proliferative effects \nof estrogens promoting nociceptive pain (29,30). Estrogens strongly induce BDNF production by \nmacrophages, and BDNF promotes neurogenesis through its binding to NTRK2 receptors on \nnerves. Release of pro-inflammatory mediators from mast cells, also triggered by estrogens, \nsensitizes peripheral nerve endings in endometriotic lesions, contributing to pain (31,32). Several \nresearch groups demonstrated that BDNF appears to be a good biomarker for early-stages (I-II) \nendometriosis (33–35). \nWith the aim of combining CA125, BDNF and clinical variables in order to develop a diagnostic \ntool that could identify all stages of endometriosis, two studies were conducted. The development \nstudy (study 1) aimed at (i) confirming the ability of BDNF, CA125 and patients’ clinical \ninformation to differentiate between cases and controls and (ii) developing a diagnostic algorithm \nbased on the results. The validation study (study 2) aimed at establishing the clinical performance \nof the developed endometriosis IVD test, combining an ELISA kit to measure serum \nconcentrations of BDNF and CA125 and a diagnostic software hosting the diagnostic algorithm. \nMethods:  \nStudy populations \nStudies were conducted on serum samples and data acquired from the Oxford Endometriosis \nCaRe Centre biobank (UK). This biobank emerged from the World Endometriosis Research \nFoundation (WERF) Endometriosis Phenome and Biobanking Harmonisation Project (EPHect) \nconsensus on standardization and harmonization of phenotypic surgical/clinical data and biologic \nsample–collection methods in endometriosis research. Patients included in this biobank were of \nreproductive age (18 – 50 years old) and were undergoing a laparoscopy due to suspicion of \nendometriosis. This biobank comprised serum samples and patient’s clinical information (from \nquestionnaires) collected before surgery along with surgical information collected during the \nprocedure. Patients were classified as controls and cases and anonymized in the biobank. Patients \nwere classified as cases if endometriosis was confirmed by laparoscopy and histological \nevaluation of excised lesions, and as controls if endometriosis lesions could not be visualized \nduring the procedure or confirmed by laparoscopy and histology. The patients with endometriosis \nwere classified in stages after laparoscopy according to the revised American Society of \nReproductive Medicine (rASRM) classification. Also, endometriosis was classified per lesion \nlocation; superficial, endometrioma, and/or DIE depending on imaging and surgical findings. \nEndometriosis was classified as “superficial” if superficial endometriosis lesions only were found \nin the ovaries or in the peritoneal cavity. Endometriosis was classified as “endometrioma” if \nendometriomas were found in the ovaries with or without superficial endometriosis. \nEndometriosis was classified as “DIE” if infiltrative lesions were reported in the peritoneal cavity \nwith or without the presence of superficial endometriosis. Endometriosis is classified as \n“endometrioma + DIE” if DIE was found in the peritoneal cavity along with endometriomas (with \nor without superficial endometriosis). For 5 patients only, this classification was not available.  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n7 \n \nThe experimental protocols were approved by the Ethics committee of CEIm HM Hospitales. \nTwo cohorts of patients were considered: \n \n• Development cohort  \nSerum samples from n=204 patients were included in the development study: n=136 patients with \nendometriosis and n =68 controls. Table 1 depicts the demographic characteristics of those \npatients. In this study, low- and high-stage endometriosis were equally represented in the cases \ngroup (stages I-II, 50% and stages III-IV, 50%). \n \n \n \nTable 1. Demographic characteristics of the patients in the development cohort. \n \n \n• External Validation cohort  \nSerum samples from n=79 patients were included in the validation study: n = 52 patients with \nendometriosis and n=25 controls. Table 2 depicts the demographic characteristics of those patients.  In \nthis study, low stage (I-II) endometriosis patients represented 81% of the cases. \n \n \n Controls  \nN=68 \nCases \nN=136 \nAge years (mean ± SD) 33.5 (5.96) 35.6 (6.42) \nBMI (mean ± SD) 25.38 (4.63) 26.46 (5.32) \nrASRM classification \nI–II  \nIII–IV \n \n- \n-  \n \n68 (50%) \n68 (50%) \nEndometriosis Classification  \nSuperficial  \nEndometrioma  \nDIE  \nDIE + endometrioma \nUnclassified \n \n- \n- \n- \n- \n- \n \n54 (39.7%) \n26 (19.1%) \n29 (21.3%) \n25 (18.4%) \n2 (1.5%) \nOther gynaecological \nconditions  \nOvarian cysts \nOvarian cancer  \nUterine fibroids \nAdenomyosis \n \n \n28 \n1 \n7 \n0 \n \n \n 66 \n6 \n25 \n7 \nNote. BMI= Body Mass Index; rASRM= revised American Society for Reproductive Medicine, \nDIE = Deep Infiltrative Endometriosis \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n8 \n \n \nTable 2. Demographic characteristics of the patients in the external validation cohort. \n \nSample collection. \nThe specimens were collected and handled following the World Endometriosis Research Foundation \nStandard Operating procedures (Rahmioglu et al. 2014) after receiving patients’ consent. Patients were \nasked to fast for at least 10 hours prior to blood collection. Serum samples were stored in a biobank at  \n-80 ºC for up to 5 years and were transferred to the laboratory analysis site.  \nELISA method: CA125 and BDNF concentrations \nThe IVD test ELISA (Enzyme-Linked Immunosorbent Assay) is a solid-phase sandwich enzyme-\nimmunoassay for the quantitative determination of BDNF and CA125 in human serum. Each \nbiomarker was determined in a different set of wells. The ELISA plate was coated with an antibody \ndirected against either BDNF or CA125. BDNF or CA125 from samples and standards bind to the \nantibodies and were immobilized on the plate. Unbound biotin conjugate was washed off with \nwashing solution. In a further step, streptavidin-HRP conjugate was added, and bound to the biotin. \nUnbound streptavidin-HRP was washed off with washing solution. Finally, a substrate solution was \nadded, and the existing complex catalyzed the chemical reaction of the substrate into a colored \nchemical entity. The enzymatic color reaction was stopped after a defined period of time. The \nconcentration of the colored chemical correlating proportionally to the concentration of the antibody \nwas measured photometrically. \nSoftware input and score calculation  \nIn the validation study only, upon collection of all the essential input parameters (serum CA125, \nserum BDNF and clinical variables), these data were introduced by the laboratory technicians into  the \n Controls  \nN=25 \nCases \nN=52 \nAge years (mean ± SD) 35 (6.44) 35 (6.47) \nBMI (mean ± SD) 26 (5.23) 26 (5.14) \nrASRM classification \nI–II  \nIII–IV \nMissing information \n \n- \n-  \n \n42 (81%) \n7 (13%) \n3 (6%) \nEndometriosis Classification  \nSuperficial  \nEndometrioma  \nDIE  \nDIE + endometrioma  \nUnclassified  \n \n- \n- \n- \n- \n- \n \n25 (48.1%) \n3 (5.8%) \n14 (26.9%) \n8 (15.4%) \n3 (5.8%) \nOther conditions  \nOvarian cysts \nOvarian cancer  \nUterine fibroids \nAdenomyosis \n \n11 \n0 \n3 \n1 \n \n16 \n4 \n4 \n1 \nNote. BMI= Body Mass Index; rASRM= revised American Society for Reproductive Medicine, \nDIE = Deep Infiltrative Endometriosis. \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n9 \n \nIVD test diagnostic medical software` hosting the data treatment algorithm. The algorithm outcomes \nwere calculated and classified as positive or negative depending on whether the value was above or \nbelow the threshold value, respectively.  \nStatistical analysis: \nStatistical analyses were performed using the software R, version 4.1.3 (R Foundation for Statistical \nComputing, Vienna, Austria), blinded to the surgical and imaging findings. Normal distribution was \nchecked using the Shapiro-Wilk test. Because BDNF and CA125 levels did not follow a normal \ndistribution, Mann-Whitney U analysis was used to compare BDNF and CA125 values between cases \nand controls. Sample sizes were chosen so that the 95% confidence interval does not exceed 0.3 for \nsensitivity and specificity outcomes around the expected value. To evaluate the importance of \nincluding both BDNF and CA125 in a diagnostic model, three logistic regression models with CA125 \nand BDNF as predictors were generated: one comparing the controls with all the cases, one comparing \nthe controls with low-stage disease (S1-S2) and one comparing the controls with high-stage disease \n(S3-S4). Upon generation of these models, the Akaike information criterion (AIC) was applied during \nbackward stepwise regression to identify whether or not BDNF and CA125 could identify \nendometriosis cases in the model. \nBased on the results, CA125, BDNF and selected clinical variables were combined into a \nmultivariable logistic regression model. Missing data were estimated by imputation: a threshold of \n10% for each predictor was used as the maximum proportion of missing data for imputation. At each \ncut-off, sensitivity and specificity were computed together with the 95% confidence interval (CI). To \ncompare the performance of the different regression models, we used ROC (Receiver Operating \nCharacteristic) curves (Delacour et al., 2005). These allow comparison of specificity (proportion of \nnegatives, i.e., controls, correctly identified as negatives) and sensitivity (proportion of positives, i.e. \nendometriosis cases, correctly identified as positives) of different models for different cut-off values. \nThe higher the AUC (Area Under Curve) of these curves, the better the method. The maximum \npossible AUC is 1, which would indicate a perfect classifier. The Wilson score with continuity \ncorrection (Newcombe, 1998) was used to estimate 95% confidence intervals for accuracy, specificity, \nand sensitivity results. After selecting the most accurate model, the score was derived based on the \nfinal predictors and the corresponding regression coefficients. Rule-in cut-off and associated \nsensitivity were derived in the development cohort based on a specificity \n≥ 90%. \nIn total, 122 clinical variables were considered for inclusion in the multivariable diagnostic algorithm. \nThe predictors with a significant number of missing data points, with a significant correlation with age \nat time of surgery or a significant association with another candidate predictor (with a more significant \nassociation with endometriosis) were excluded for multivariable analysis. Chi-squared and Cochran-\nArmitage tests were used to determine which categorical variables were most strongly associated with \nendometriosis. Mann-Whitney U analysis was used for numerical variables.  \nIn the validation study, algorithm scores and associated outcomes (positive diagnosis if the score was \nhigher than the defined cut-off and negative diagnosis if the score was lower than the defined cut-off) \nwere computed by the IVD test software. Based on these results, the primary (sensitivity and \nspecificity) and secondary (accuracy, and AUC) performance parameters were calculated and reported, \ntogether with their 95% confidence intervals. The primary performance parameters results were \ncompared with the values of the acceptance criteria established in the development study, to conclude \nwhether the clinical performance of the device meets the criteria, i.e., whether the device can \nadequately classify the study subjects as positive or negative for endometriosis. In concrete, the \nsensitivity and specificity in the validation study should not be lower than the lower limits of the \nsensitivity and specificity 95% confidence intervals in the algorithm development study. Because the \nprevalence of stage I-II in the validation study was significantly higher than in the development study \n(Chi-square= 18.06, p<0.001)., the outcomes in the validation were weighted to give equal \nrepresentation to the low-stage and high-stage groups.  \n \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n10 \n \n  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\nResults: \n \nDiagnostic performance of CA125 and BDNF in endometriosis \n \n \nFigure 1. Serum concentration of CA125 and BDNF in endometriosis patients and controls. A st erisk sig ns a b o v e b oxplots \nin d i c a te  a statisti cally si gnifi c ant  diff e rence in  m ed ian val u e betwe en the in d ica te d po p u l a ti on ( ** : p  < 0 . 0 1; ** * : p < \n0.001)  a s es ta blished by a Ma nn - Whitney  U test.  \n \nFigure 1 displays BDNF and CA125 values in cases and controls. The Mann-Whitney U analysis \nshowed that both BDNF and CA125 were significantly higher in cases than in controls ( p < 0.01 and \np<0.001, respectively).  \nIn backwards stepwise regression analysis based on AIC, for the comparisons of the control group \nwith all cases and with the high-stage disease cases, both CA125 and BDNF were retained, meaning \nthey were both independently informative as predictors of endometriosis. For the comparison of the \ncontrol group with low-stage disease, only BDNF was retained, meaning that only BDNF was \nindependently informative as predictor of low-stage endometriosis. \nTaken together, both CA125 and BDNF are able to distinguish controls from endometriosis cases, \nwith the former performing very well in the high-stage group and the latter the better contributor for \nthe low-stage group. Therefore, both parameters were justified for inclusion in a multivariable model \nfor endometriosis diagnosis. \n \n \n \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n12 \n \n \nDevelopment of prediction model for endometriosis  \n  \nAmong all clinical variables related to patients’ medical history that were considered, chi-square \nanalysis showed that only three qualitative variables were significantly different between cases and \ncontrols. Most significantly, a previous history of surgery to examine endometriosis (even if the \ndisease was not diagnosed) was more common among patients tested positive for current \nendometriosis (54.6% in cases, 10.6% in controls, p < 0.001). Painful periods as a symptom leading to \na referral for endometriosis was also associated with a positive diagnosis with strong statistical \nsignificance (76.4% in cases, 36.8% in controls, p < 0.001). Other significant variables  was the \nseverity of the last menstrual cycle pain, with moderate/severe pain being more frequent in cases than \nin controls (78.0% in cases, 47.1% in controls, p < 0.01). For numerical (quantitative) variables, the \nmedian of three of them were significantly different between cases and controls in the Mann-Whitney \nU test: age at first regular use of painkiller (U = 343.5, p = 0.038), age at first diagnosis of ovarian cyst \n(U = 334.5, p = 0.023), and age at first experience of intercourse pain (U = 1201, p = 0.009). \nIn the final revised model, the eight above discussed variables were considered:CA125, BDNF and the \nsix clinical variables, i.e. record of a previous surgery to examine for endometriosis, painful periods as \na symptom leading to referral for endometriosis, severity of menstrual pain during last cycle, age at \nfirst experience of intercourse pain, age at first regular use of painkillers and age at first diagnosis of \novarian cyst.  \nTo estimate its performance on independent data, a logistic regression model was repeatedly generated \non 80% of the data and evaluated on the remaining 20%. The final model, generated from all \nalgorithm development data, was optimized for high specificity to render a rule-in test with a low rate \nof false positives (36). This model has an AUC of 0.867 with a sensitivity of 51.5% at a specificity of \n95.6%. \nModel Area Under Curve Youden’s index Accuracy Sensitivity Specificity \nAt 95% specificity 0.867 \n(0.819 – 0.915) \n47.1% \n(37.3 - 56.8%) \n66.2% \n(59.2 - 72.5%) \n51.5% \n(42.8 - 60.1) \n95.65% \n(86.8 - 98.9%) \nAt 95% sensitivity 0.867 \n(0.819 – 0.915) \n44.1% \n(31.7 - 56.5%) \n79.9% \n(73.6 -85%) \n95.6% \n(90.2 - 98.2) \n48.5% \n(36.4 - 60.9%) \nAt maximum \nYouden’s index \n0.867 \n(0.819 – 0.915) \n58.8% \n(46.7 - 70.9%) \n82.4% \n(76.3 - 87.2%) \n88.2% \n(81.3 - 92.9%) \n70.6% \n(58.1 - 80.7%) \nAt maximum \naccuracy \n0.867 \n(0.819 – 0.915) \n58.1% \n(45.9 - 70.3%) \n82.4% \n(76.3 - 87.2%) \n89% \n(82.2 - 93.5%) \n69.1% \n(56.6 - 79.5%) \n \nTable 3. Performance characteristics of the IVD test in the development study.  \nClinical performance evaluation (validation of the IVD test)  \n \nThe diagnostic performance of the IVD test, comprising the ELISA kit method for the determination \nof BDNF and CA125 together with the diagnostic algorithm established in the development study was \nevaluated in an independent sample cohort. The endometriosis IVD test had a sensitivity (after \nweighing for disease stages) of 46.2% (95% CI: 25.5-66.8%) and a specificity of 100% (95% CI: 86.7-\n100%). The accuracy was 64.1% (95% CI: 50.4-77.8%) and the AUC was 0.758 (95% CI: 0.650-\n0.867).  With an observed diagnostic specificity in this clinical performance study of 100%, the target \nspecificity of 86.8% (or higher) is met. A good specificity was the primary objective because this \nassay is primarily intended to aid in identifying individuals with endometriosis. For the sensitivity, a \nmid-range sensitivity rather than a low sensitivity was desired to ensure that a significant proportion of \nthe test population will test positive.  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n13 \n \n \nDifferential diagnosis \n \nConfounding conditions \nFurther, we investigated whether other gynaecological conditions could interfere with the performance \nof the IVD test, rendering a positive test result when endometriosis is not present (false positive). The \nconditions considered as potentially confounding were non endometriosis benign ovarian cysts, \novarian cancer, uterine fibroids and adenomyosis. In this respect, out of 93 controls included in the \ndevelopment (n=68) and validation ( n=25) studies, 42% ( n=39) had other ovarian cysts and 11% ( n= \n10) had uterine fibroids. Only 2 patients (2%) had a positive test result (false positive), and thus the \neffect of these potentially confounding conditions on the test is considered very limited.  \nDetection of superficial endometriosis \nWe have analysed the capacity of the endometriosis IVD test to identify the cases presenting just with \nsuperficial endometriosis. In the development cohort, endometriosis could be classified into different \ngroups (into superficial endometriosis, endometrioma, endometrioma + DIE and DIE) for 134 out of \n136 cases. In the validation cohort, the classification could be done for 50 out of 53 cases. Out of those \nn=184 patients in total (both cohorts), n=79 patients had superficial endometriosis (43%). With the \nendometriosis IVD Test, n=25 of the n=79 (32%) cases with superficial endometriosis were detected. \n \n4. Discussion  \n \nWe developed a non-invasive in vitro diagnostic (IVD) test for endometriosis using a step-by-step \napproach. In the development study, the abilit y of BDNF and CA125 to differentiate between \ncases and controls was confirmed. Based on those results, the IVD test, consisting of an ELISA \nkit for the determination of serum concentrations of BDNF and CA125 and a data treatment \nalgorithm hosted in a diagnostic medical software was developed. In the validation study, the \nclinical performance of The IVD test in diagnosing endometriosis was established.  Main results \nare discussed below.  \nFirst, although no cut-off values were found, CA125 and BDNF levels were demonstrated to be \nelevated in patients with endometriosis; with CA125 mostly able to identify high-stage \nendometriosis and BDNF performing well for both low- and high-stage disease. This confirms \nwhat was previously found by other research groups: BDNF concentrations are higher in \nendometriosis patients than in controls in plasma (34,37,38) and serum (33,35,39). We chose to \nmeasure BDNF concentration in serum because, as previously shown, during centrifugation, all of \nthe BDNF content is released from platelets, reducing measurement errors related to blood \nhandling, storage, and analysis encountered with plasma samples. (40,41).  Although there is \nmuch confounding evidence on the validity of CA125 as a biomarker for endometriosis, two \nmeta-analyses showed that it could be used in conjunction with clinical information (21,24).  \nA number of controls in both development and validation studies had other gynecological \nconditions that could elevate the CA125 concentration in serum (e.g., benign ovarian cysts, \nuterine fibroids, ovarian cancer and adenomyosis) (23) and had a negative diagnosis (classified as \ntrue negative) using the IVD test. The presence of such confounding factors did not lead to any \nfalse positive result in the validation study. This is likely because the IVD test does not rely solely \non CA125 but also on BDNF and the patient’s clinical information.  \n \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n14 \n \nIn the validation study, as the algorithm was optimized for specificity during development, the \nnovel endometriosis IVD test showed a limited sensitivity (46.2%) but a very high specificity of \n100%, making it an excellent rule-in test able to minimize the risk of false positives. A rule-in test \nis considered the most appropriate approach given the chronic and non-life-threatening nature of \nthe disease. A positive test result would aid the clinician in the diagnosis, when considered \ntogether with other clinical information.\n The diagnosis of women presenting only with superficial \nlesions by a non-invasive test is of special interest due to the limited value of existing imaging \ntechniques for their identification  (1,10), possibly leading to numerous misdiagnoses.  Considering \nthat the endometriosis IVD test was able to detect 32% of cases presenting with superficial lesions \nrecruited in the studies, this diagnostic tool can provide an added value for the diagnosis of this \ntype of disease. When the test is negative, the clinician may consider other causes for the \nsymptoms or symptomatic treatment for pain, according to their usual practice. If the suspicion of \nendometriosis persists after a follow up consultation, the women can be re-tested at the discretion \nof the clinician.   \nOur diagnostic test compares well with other benchmark diagnostic tests, such as prostate-specific \nantigen (PSA) to detect prostate cancer which has a sensitivity of 93% (95% CI 88%, 96%) and a \nspecificity of 20% (95% CI 12%, 33%) (42). \nAn essential strength of this study is that all the participants underwent laparoscopy (gold standard \ndiagnosis), a necessary component of algorithm development to provide the true clinical state of \neach participant.\n The diagnostic algorithm was developed based on n=204 patients in the \ndevelopment cohort. A total of 8 predictors were included in the multivariate logistic regression \nmodel: CA125, BDNF, record of previous surgery for endometriosis, painful periods leading to \nreferral for endometriosis, age at first intercourse pain, age at first painkillers use, age at first ovarian \ncyst symptom and severity of menstrual pain during last cycle. After performing the IVD ELISA \ntest, laboratory technicians can introduce CA125 and BDNF results in a diagnostic medical \nsoftware along with patients’ medical information. The software hosting the algorithm calculates \na score, which according to a certain cutoff value, provides a diagnosis.  \n \nThis novel endometriosis IVD test is a medical device that has been CE marked under the IVD \nDirective 98/79/EC. The test could be included in early workup to aid clinicians in the diagnosis \nof endometriosis when the disease is suspected, in conjunction with other clinical information, in \norder to facilitate timely access to a correct disease management. \n5. Conclusion \nWe have developed and validated a non-invasive in vitro  diagnostic test for endometriosis. The \nexcellent rule-in performance of this test could provide a significant value in the clinical management \nof this disease. \n \n6. Bibliography  \n \n1.  Zondervan KT, Becker CM, Missmer SA. Endometriosis. New England Journal of Medicine \n2020;382(13):1244–56.  \n2.  Zondervan KT, Becker CM, Koga K, Missmer SA, Taylor RN, Viganò P. Endometriosis. Nat \nRev Dis Primers 2018;4(1):1–25.  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n15 \n \n3.  Symons LK, Miller JE, Kay VR, Marks RM, Liblik K, Koti M, et al. The \nImmunopathophysiology of Endometriosis. Trends in Molecular Medicine 2018;24(9):748–62.  \n4.  Taylor HS, Kotlyar AM, Flores VA. Endometriosis is a chronic systemic disease: clinical \nchallenges and novel innovations. The Lancet 2021;397(10276):839–52.  \n5.  Rossi H-R, Uimari O, Terho A, Pesonen P, Koivurova S, Piltonen T. Increased overall morbidity \nin women with endometriosis: a population-based follow-up study until age 50. Fertility and \nSterility 2023;119(1):89–98.  \n6.  Sinaii N, Cleary SD, Ballweg ML, Nieman LK, Stratton P. High rates of autoimmune and \nendocrine disorders, fibromyalgia, chronic fatigue syndrome and atopic diseases among women \nwith endometriosis: a survey analysis. Hum Reprod 2002;17(10):2715–24.  \n7.  Somigliana E, Vigano’ P, Parazzini F, Stoppelli S, Giambattista E, Vercellini P. Association \nbetween endometriosis and cancer: a comprehensive review and a critical analysis of clinical and \nepidemiological evidence. Gynecol Oncol 2006;101(2):331–41.  \n8.  Simoens S, Dunselman G, Dirksen C, Hummelshoj  L, Bokor A, Brandes I, et al. The burden of \nendometriosis: costs and quality of life of women with endometriosis and treated in referral \ncentres. Human Reproduction 2012;27(5):1292–9.  \n9.  Agarwal SK, Chapron C, Giudice LC, Laufer MR, Leyland N, Missmer SA, et al. Clinical \ndiagnosis of endometriosis: a call to action. Am J Obstet Gynecol 2019;220(4):354.e1-354.e12.  \n10.  Becker CM, Bokor A, Heikinheimo O, Horne A, Jansen F, Kiesel L, et al. ESHRE guideline: \nendometriosis. Hum Reprod Open 2022;2022(2):hoac009.  \n11.  Dunselman G a. J, Vermeulen N, Becker C, Calhaz-Jorge C, D’Hooghe T, De Bie B, et al. \nESHRE guideline: management of women with endometriosis. Hum Reprod 2014;29(3):400–12.  \n12.  Jenkins TR, Liu CY, White J. Does response to hormonal therapy predict presence or absence of \nendometriosis? J Minim Invasive Gynecol 2008;15(1):82–6.  \n13.  Ahn SH, Singh V, Tayade C. Biomarkers in endometriosis: challenges and opportunities. \nFertility and Sterility 2017;107(3):523–32.  \n14.  Irungu S, Mavrelos D, Worthington J, Blyuss O, Saridogan E, Timms JF. Discovery of non-\ninvasive biomarkers for the diagnosis of endometriosis. Clin Proteomics 2019;16:14.  \n15.  Nisenblat V, Bossuyt PMM, Farquhar C, Johnson N, Hull ML. Imaging modalities for the non-\ninvasive diagnosis of endometriosis. Cochrane Database Syst Rev 2016;2(2):CD009591.  \n16.  Nisenblat V, Bossuyt PM, Shaikh R, Farquhar C, Jordan V, Scheffers CS, et al. Blood \nbiomarkers for the non/i2 invasive diagnosis of endometriosis. Cochrane Database of Systematic \nReviews [Internet] 2016 [cited 2023 Jan 26];(5). Available from: \nhttps://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD012179/full \n17.  Barbieri RL, Niloff JM, Bast RC, Schaetzl E, Kistner RW, Knapp RC. Elevated serum \nconcentrations of CA-125 in patients with advanced endometriosis. Fertility and Sterility \n1986;45(5):630–4.  \n18.  Chen Y, Pan M, Zuo Y, Yang B, Wang S. Research progress of CA125 in endometriosis: \nTeaching an old dog new tricks. Gynecology and Obstetrics Clinical Medicine 2022;2(4):191–8.  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n16 \n \n19.  Nagamani M, Kelver ME, Smith ER. CA 125 levels in monitoring therapy for endometriosis and \nin prediction of recurrence. Int J Fertil 1992;37(4):227–31.  \n20.  Shen A, Xu S, Ma Y, Guo H, Li C, Yang C, et al. Diagnostic value of serum CA125, CA19-9 \nand CA15-3 in endometriosis: A meta-analysis. J Int Med Res 2015;43(5):599–609.  \n21.  Kafali H, Artuc H, Demir N. Use of CA125 fluctuation during the menstrual cycle as a tool in \nthe clinical diagnosis of endometriosis; a preliminary report. European Journal of Obstetrics & \nGynecology and Reproductive Biology 2004;116(1):85–8.  \n22.  Karimi-Zarchi M, Dehshiri-Zadeh N, Sekhavat L, Nosouhi F. Correlation of CA-125 serum level \nand clinico-pathological characteristic of patients with endometriosis. Int J Reprod Biomed \n2016;14(11):713–8.  \n23.  Hirsch M, Duffy J, Davis C, Nieves Plana M, Khan K, on behalf of the International \nCollaboration to Harmonise Outcomes and Measures for Endometriosis. Diagnostic accuracy of \ncancer antigen 125 for endometriosis: a systematic review and meta-analysis. BJOG: An \nInternational Journal of Obstetrics & Gynaecology 2016;123(11):1761–8.  \n24.  Hirsch M, Duffy JMN, Deguara CS, Davis CJ, Khan KS. Diagnostic accuracy of Cancer Antigen \n125 (CA125) for endometriosis in symptomatic women: A multi-center study. European Journal \nof Obstetrics & Gynecology and Reproductive Biology 2017;210:102–7.  \n25.  Chao MV. Neurotrophins and their receptors: A convergence point for many signalling \npathways. Nat Rev Neurosci 2003;4(4):299–309.  \n26.  Wessels JM, Wu L, Leyland NA, Wang H, Foster WG. The brain-uterus connection: brain \nderived neurotrophic factor (BDNF) and its receptor (Ntrk2) are conserved in the mammalian \nuterus. PLoS One 2014;9(4):e94036.  \n27.  Wang S, Duan H, Li B, Hong W, Li X, Wang Y, et al. BDNF and TrKB expression levels in \npatients with endometriosis and their associations with dysmenorrhoea. Journal of Ovarian \nResearch 2022;15(1):35.  \n28.  Lim W, Bae H, Bazer FW, Song G. Brain-derived neurotrophic factor improves proliferation of \nendometrial epithelial cells by inhibition of endoplasmic reticulum stress during early pregnancy. \nJournal of Cellular Physiology 2017;232(12):3641–51.  \n29.  Dong F, Zhang Q, Kong W, Chen J, Ma J, Wang L, et al. Regulation of endometrial cell \nproliferation by estrogen-induced BDNF signaling pathway. Gynecological Endocrinology \n2017;33(6):485–9.  \n30.  Godin SK, Wagner J, Huang P, Bree D. The role of peripheral nerve signaling in endometriosis. \nFASEB BioAdvances 2021;3(10):802–13.  \n31.  Greaves E, Temp J, Esnal-Zufiurre A, Mechsner S, Horne AW, Saunders PTK. Estradiol Is a \nCritical Mediator of Macrophage-Nerve Cross Talk in Peritoneal Endometriosis. The American \nJournal of Pathology 2015;185(8):2286–97.  \n32.  Perricos A, Ashjaei K, Husslein H, Proestling K, Kuessel L, Obwegeser R, et al. Increased serum \nlevels of mBDNF in women with minimal and mild endometriosis have no predictive power for \nthe disease. Exp Biol Med (Maywood) 2018;243(1):50–6.  \n33.  Wessels JM, Kay VR, Leyland NA, Agarwal SK, Foster WG. Assessing brain-derived \nneurotrophic factor as a novel clinical marker of endometriosis. Fertil Steril 2016;105(1):119-\n128.e1-5.  \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint \n\n17 \n \n34.  Ding S, Zhu T, Tian Y, Xu P, Chen Z, Huang X, et al. Role of Brain-Derived Neurotrophic \nFactor in Endometriosis Pain. Reprod Sci 2018;25(7):1045–57.  \n35.  Trevethan R. Sensitivity, Specificity, and Predictive Values: Foundations, Pliabilities, and \nPitfalls in Research and Practice. Frontiers in Public Health [Internet] 2017 [cited 2023 Mar 6];5. \nAvailable from: https://www.frontiersin.org/articles/10.3389/fpubh.2017.00307 \n36.  Giannini A, Bucci F, Luisi S, Cela V, Pluchino N, Merlini S, et al. Brain-Derived Neurotrophic \nFactor in Plasma of Women with Endometriosis. Journal of Endometriosis 2010;2(3):144–50.  \n37.  Rocha AL, Vieira EL, Ferreira MC, Maia LM, Teixeira AL, Reis FM. Plasma brain-derived \nneurotrophic factor in women with pelvic pain: a potential biomarker for endometriosis? \nBiomark Med 2017;11(4):313–7.  \n38.  Liang YF, Huang XM, Wen LL, Kang H, Tao MH, Ye MZ. [Relationship between serum brain-\nderived neurotrophic factor and clinical stage and dysmenorrhoea of enodmetriosis]. Zhonghua \nYi Xue Za Zhi 2020;100(10):771–4.  \n39.  Polacchini A, Metelli G, Francavilla R, Baj G, Florean M, Mascaretti LG, et al. A method for \nreproducible measurements of serum BDNF: comparison of the performance of six commercial \nassays. Sci Rep 2015;5:17989.  \n40.  Polyakova M, Schlögl H, Sacher J, Schmidt-Kassow M, Kaiser J, Stumvoll M, et al. Stability of \nBDNF in Human Samples Stored Up to 6 Months and Correlations of Serum and EDTA-Plasma \nConcentrations. Int J Mol Sci 2017;18(6):1189.  \n41.  Merriel SWD, Pocock L, Gilbert E, Creavin S, Walter FM, Spencer A, et al. Systematic review \nand meta-analysis of the diagnostic accuracy of prostate-specific antigen (PSA) for the detection \nof prostate cancer in symptomatic patients. BMC Medicine 2022;20(1):54.  \n \nAll rights reserved. No reuse allowed without permission. \n(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. \nThe copyright holder for this preprintthis version posted March 29, 2023. ; https://doi.org/10.1101/2023.03.29.23287909doi: medRxiv preprint","source_license":"CC0","license_restricted":false}