Validation of an in vitro diagnostic test for endometriosis: impact of confounding medical conditions and lesion location

With the aim to shorten the time for diagnosis and accelerate access to correct manage- 8 ment, a non-invasive diagnostic test for endometriosis was developed and validated. The IVD test 9 combines an ELISA test kit to quantify CA125 and BDNF concentrations in serum and a data treat- 10 ment algorithm hosted in medical software processing results from the ELISA test and responses to 11 six clinical variables. Serum samples and clinical variables extracted from psychometric question- 12 naires from 77 patients were collected from the Oxford Endometriosis CaRe Centre biobank (UK). 13 Case/control classification was performed based on laparoscopy and histological verification of the 14 excised lesions. Biomarkers serum concentrations and clinical variables were introduced to the soft- 15 ware, which generates the qualitative diagnostic result (“positive” or “negative”). This test allowed 16 the detection of 32% of cases with superficial endometriosis, which is an added value given the 17 limited efficacy of existing imaging techniques. Even in the presence of various confounding medi- 18 cal conditions, the test maintained a specificity of 100%, supporting its suitability for use in patients 19 with underlying medical conditions. 20

In vitro diagnostic test; endometriosis; validation; lesion location; superficial endome- 21 triosis; confounding conditions. 22 23 1. Introduction 24 Endometriosis is a progressive, estrogen-dependent disease that affects approxi- 25 mately 10% of women of reproductive age [1]. It is characterized by the presence of en- 26 dometrial-like tissue outside the uterus, commonly affecting the pelvic cavity, ovaries, 27 fallopian tubes, and other surrounding structures[2]. These lesions result in a chronic 28 inflammatory response, which can lead to the formation of scar tissue and adhesions[3]. 29 The clinical presentation of endometriosis can be very diverse, with a wide range of 30 symptoms, including chronic non-menstrual pelvic pain, dysmenorrhea, dysuria, infer- 31 tility, and many others; with the onset of symptoms usually occurring during adoles- 32 cence[1,4]. The severity and manifestation of symptoms can be influenced by various 33 factors, including the location and extent of the endometrial implants, hormonal fluctua- 34 tions, and individual pain thresholds[5,6]. Also, symptoms often overlap with those of 35 various other conditions [7,8]. Although imaging techniques such as transvaginal ultra- 36 sound (TVUS) and magnetic resonance imaging (MRI) have been shown to accurately 37 diagnose some endometriosis cases, these are usually limited to more severe stages of 38 the disease[9,10]. Laparoscopy, with or without histological confirmation, remains the 39 gold standard for diagnosing endometriosis, but its invasive nature contributes to diag- 40 nostic delays [1–3,11]. Therefore, despite its high prevalence, accurately diagnosing en- 41 dometriosis can be challenging, with an initial misdiagnosis in up to 65% of women and 42 a diagnostic delay of 4-11 years [7,12].This delay hinders the identification of early 43 Copyright: © 2024 by the authors. Submitted for possible open access publication under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/license s/by/4.0/). All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2 of 14 stages, allowing the condition to progress, leading to increased severity, fibrosis, and 44 potential infertility [13,14]. Developing a non-invasive diagnostic test for endometriosis 45 becomes crucial in order to obviate the delay in diagnosis [15,16]. 46 Prior studies have delved into an extensive array of biomarkers, highlighting the 47 complexity of understanding endometriosis. CA125, a widely recognized glycoprotein, 48 has been a focal point in research due to its association with various gynecological con- 49 ditions, including endometriosis [17–19]. Despite its usefulness, the lack of specificity 50 and limited sensitivity as a standalone marker, along with its utility being limited to late 51 stages of the disease, has underscored the need for complementary biomarkers. Brain - 52 derived neurotrophic factor (BDNF), known for its involvement in neuroplasticity and 53 neuronal survival, has emerged as a promising candidate, with studies demonstrating 54 elevated levels in patients with endometriosis compared to healthy controls [11,20,21]. 55 However, the lack of specificity among individual biomarkers emphasizes the necessity 56 of a comprehensive diagnostic approach integrating multiple markers to enhance accu- 57 racy and reliability in endometriosis detection. 58 Recently, we have developed a diagnostic treatment algorithm that combines 59 CA125 and BDNF measurements with six pertinent clinical variables: patient's surgical 60 history related to endometriosis, the manifestation of painful periods as a leading symp- 61 tom for endometriosis referral, the intensity of menstrual pain during the previous cycle, 62 the age at the onset of intercourse-related pain, the age at the initiation of regular pain- 63 killer usage, and the age at the initial diagnosis of an ovarian cyst. CA125, BDNF, and 64 the six clinical factors were integrated into the final logistic regression model, achieving 65 an AUC of 0.867, sensitivity of 51.5%, and specificity of 95.6% [22]. 66 The influence of confounding conditions on the final diagnosis of endometriosis 67 using this test was challenged. This is because multiple conditions, gynecological (for 68 instance, adenomyosis [23–26], pelvic inflammatory disease (PID) [27–29], uterine fi- 69 broids [29–31] and ovarian cysts [29,32]) and non-gynecological (for instance, inflamma- 70 tory bowel disease (IBD) [33] or rheumatoid arthritis [34–36], asthma [37], anxiety and 71 depression [38–40]) could affect the levels of CA125 and BDNF. 72 The primary aim of this study was to validate the diagnostic performance of the test in 73 endometriosis patients while also discerning the specific subgroup of patients in which 74 the test demonstrates superior performance. The secondary aim was to further investi- 75 gate how confounding conditions influence CA125 and BDNF and whether or not the 76 performance of the test is affected. 77 2. Results 78 2.1. Diagnostic performance by endometriosis lesion type 79 One hundred percent of controls from the validation dataset were correctly diagnosed 80 (negative) with the IVD test, based on the threshold established in the development da- 81 taset. With this, a sensitivity (after weighing for disease stages) of 46.2% (95% CI: 25.5 - 82 66.8%) and a specificity of 100% (95% CI: 86.7-100%) was obtained. The accuracy was 83 64.1% (95% CI: 50.4-77.8%) and the AUC was 0.758 (95% CI: 0.650-0.867). To understand 84 in which subgroup of endometriosis patients the test works best, i.e., is capable of detect- 85 ing the highest number of cases, patients were separated in subgroups by lesion types. 86 First, the association between the stages of endometriosis and the types of endometriosis 87 lesions was examined using Pearson's chi-squared test. The analysis revealed a signifi- 88 cant association (χ² = 765.76, df = 25, p < 0.001), indicating a strong relationship between 89 the rASRM stages classification of endometriosis and classification by types of lesions. 90 The contingency table (Table 1) provides insight on how lesion types are distributed by 91 endometriosis rASRM stage for patients of pooled development and validation datasets. 92 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 3 of 14 Superficial lesions are observed mostly in stage I (81.5%). Extended lesions (endometri- 93 oma+DIE) are as expected mostly observed in stage IV. 94 95 rASRM stage lesion type Stage I (n=81) Stage II (n=30) Stage III (n=34) Stage IV (n=41) Unclassified (n=2) C-section 0 0 0 0 2 DIE 7 17 7 9 0 Endometrioma 5 0 16 7 0 Endometrioma + DIE 0 3 10 20 0 Superficial 66 10 1 2 0 Unclassified 3 0 0 3 0 96 Table 1. Contingency table for the distribution of lesion types by endometriosis rASRM stages. 97 98 Sensitivity was investigated by lesion type. Results (as reported in table 2) indicate that 99 the IVD test successfully identified around half of the cases of DIE and endometrioma. 100 Furthermore, with a sensitivity of 69.70%, the IVD test demonstrate that the test works 101 best in identifying cases of DIE+endometrioma. Interestingly, 32% of cases of superficial 102 endometriosis were correctly identified with the test. As expected, the two cases of endo- 103 metriosis located within c -section scars could not be identified with the test (different 104 pathophysiology, as described above). 105 106 107 Table 2. Distribution of cases, number of true positive and sensitivity by lesion type in both devel- 108 opment and validation datasets. 109 110 An ANOVA was conducted to examine the differences in CA125 values among various 111 types of endometriosis lesions in the pooled datasets (development and validation da- 112 tasets, figure 1 ). The results revealed a significant effect of lesion type on CA125 levels 113 (F(5, 275) = 26.162, p < 0.001). Post hoc analyses indicated that the differences were statis- 114 tically significant (p < 0.001) across the various lesion types. The Tukey multiple compar- 115 ison of means at a 95% family -wise confidence level revealed several significant differ- 116 ences between the types of lesions in terms of CA125 levels: comparing endometrioma to 117 DIE, there was a statistically significant difference (p<0.01). Additionally, the mean CA125 118 level (56.05 IU/mL, SD=39.35) were higher for endometrioma than for DIE (32.28 IU/mL, 119 SD=32.69) (p=0.01). Moreover, the mean CA125 level for endometrioma + DIE (67.69 120 Gynecological Condition Number of controls in development data (n=68) Number of controls in valida- tion data (n=25) Total number of controls (n=93) Ovarian cysts 28 11 39 Uterine fibroids 7 3 10 Adenomyosis 0 1 1 PCOS 16 8 24 Pelvic inflammatory disease 4 2 6 At least one condition 40 19 55 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 4 of 14 IU/mL, SD=45.49) was higher than the mean CA125 level for DIE (mean=32.28 IU/mL, 121 SD=32.69) (p<0.001). Lower CA125 levels were observed for superficial lesions 122 (mean=19.55 IU/mL, SD=24.74) than for endometrioma (p<0.001), DIE (p=0.02) and endo- 123 metrioma + DIE (p<0.001). An ANOVA conducted on BDNF values across different lesion 124 types did not show any significant differences of BDNF across different types of lesions 125 (p=0.094). This suggests that the improved sensitivity for DIE+endometrioma lesions is 126 likely to be due to higher levels of CA125 in those lesions, contributing to a higher rate of 127 true positive results in cases with those lesions. 128 129 130 Figure 1. Comparison of CA125 levels between lesion types. 131 2.2. Interference of potentially confounding medical conditions 132 As shown in table 3, despite 76% (19 out of 25 controls) of controls in the validation da- 133 taset having at least one condition that could elevate CA125, the specificity of the diag- 134 nostic test was 100%. 135 136 Gynecological Condi- tion Number of controls in development data (n=68) Number of controls in validation data (n=25) Total number of controls (n=93) Ovarian cysts 28 11 39 Uterine fibroids 7 3 10 Adenomyosis 0 1 1 PCOS 16 8 24 Pelvic inflammatory disease 4 2 6 At least one condition 40 19 55 137 Table 3. Distribution of gynecological conditions known to elevate CA125 across controls. 138 139 Two-way ANOVA with EndoState (Cases/controls) and each confounding condition as 140 predictors was run on CA125 levels in the pooled datasets. For ovarian cysts, the ANOVA 141 revealed the main effect of EndoState (F = 32.97, p<0.001) and Ovarian cyst condition (F = 142 22.65, p<0.001) on CA125 levels. Individuals with ovarian cysts had higher CA125 values 143 than individuals without ovarian cysts (p<0.001). No interaction between both predictors 144 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 5 of 14 was reported. For uterine fibroids (UF), a main effect for condition on CA125 was ob- 145 served (F=11.22, p<0.001) as well as an expected main effect for EndoState (F=15.30, 146 p<0.001). No interaction between both predictors was reported. Individuals with uterine 147 fibroids had higher CA125 values than individuals without uterine fibroids (p<0.001). 148 Two-way analysis of variance (ANOVA) with EndoState (Cases/controls) and each con- 149 founding condition as predictors was run on BDNF in pooled datasets. For Chronic fa- 150 tigue only, a main effect was observed for EndoState (F=5.75, p=0.017) and an interaction 151 between EndoState and the condition (F=4.20, p=0.04). Pairwise comparisons revealed 152 that cases without chronic fatigue have higher BDNF values than controls with chronic 153 fatigue (mean difference=6.06, p=0.04). 154 The performance of the diagnostic test was determined in the validation dataset exclud- 155 ing each confounding condition at a time. Results, as shown in Table 4, indicate that the 156 sensitivity values when excluding conditions stay within the 95% CI of the original sen- 157 sitivity (all conditions included) between 34.3 and 62.9, meaning that no condition criti- 158 cally affects the ability of the test of detecting cases. 159 160 161 Table 4. Performance of the IVD test (validation dataset) excluding each medical condi- 162 tion at a time. 163 164 165 Left out condition Number of subjects by condition Sensi- tivity 95% CI lower limit 95% CI upper limit All (no data left out) 0 48,5 34,3 62,9 Ovarian cyst 27 50,8 34,1 67,4 Uterine fibroids 7 45,8 31,3 61 Adenomyosis 2 46,1 32 60,8 Inflammatory Bowel Disease 2 46,1 32 60,8 Depression requiring medication or ther- apy 29 34,3 18,9 53,4 Anxiety requiring medication or therapy 20 37,9 22,6 55,8 Pelvic Inflammatory Disease 6 49,8 35,8 63,9 Eczema 16 55,6 39,6 70,5 Polycystic Ovary Syndrome 17 48,6 33,1 64,3 Interstitial cystitis 7 48,9 33,9 64 Asthma 23 48,8 32,1 65,7 Chronic fatigue syndrome - Myalgic en- cephalomyelitis 1 47,6 33,4 62,2 Fibromyalgia 1 47,6 33,4 62,2 Irritable Bowel Syndrome 17 42,7 27,2 59,6 Migraine 22 42 26 59,6 Glandular fever 5 47,9 33,5 62,6 Ulcerative colitis 2 46,1 32 60,8 High blood pressure 4 47,9 33,5 62,6 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 6 of 14 3. Discussion 166 The newly developed test for endometriosis demonstrates a high specificity of 100%, 167 suggesting its potential use as a rule -in test in clinical practice. This diagnostic test could 168 significantly contribute to the initial diagnostic workup, effectively confirming the pres- 169 ence of endometriosis and providing clinicians with a reliable tool for early detection and 170 intervention. Moreover, the test demonstrates an encouraging ability to identify superfi- 171 cial lesions of endometriosis, as evidenced by the reported sensitivity of 32 %. This feature 172 is of particular significance considering the constraints associated with the ability of alter- 173 native diagnostic methods to detect superficial lesions: superficial endometriosis, charac- 174 terized by its subtle and less invasive nature, presents unique challenges for detection 175 using ultrasound or MRI. Peritoneal implants invading less than 5 mm of depth from the 176 peritoneal surface are often invisible on MRI [10]. These imaging techniques may struggle 177 to capture the nuanced characteristics of these lesions due to their limited ability to visu- 178 alize subtle changes in the peritoneum and pelvic surfaces [42]. Additionally, the lack of 179 specific imaging markers or distinguishing features that differentiate superficial lesions 180 from surrounding healthy tissue makes it difficult to accurately identify these lesions us- 181 ing standard imaging modalities. The intricate anatomical location of superficial lesions, 182 often nestled within complex pelvic structures, further contributes to the complexity of 183 their detection, as these areas may be challenging to access and visualize accurately using 184 traditional imaging approaches [43]. By enabling the identification of superficial lesions, 185 the test offers clinicians an essential means of identifying cases that would otherwise have 186 gone undetected, thereby facilitating a more comprehensive and accurate patient man- 187 agement. 188 Also, the test demonstrated a relatively high sensitivity of 69.70% in detecting endo- 189 metrioma+ DIE lesions, possibly correlated to patients with those lesions having the high- 190 est level of CA125 compared to other types of lesions. Endometrioma, an endometriosis - 191 related ovarian cyst, often exhibits elevated CA125 levels due to its involvement of the 192 ovaries and resulting inflammatory processes. The higher mean CA125 level observed in 193 this group aligns with prior studies [44]. The observed higher mean CA125 level in the 194 endometrioma + DIE lesions compared to the DIE alone, along with the lowest CA125 195 levels in the superficial endometriosis, suggest that CA125 expression increases with the 196 extent of the disease (i.e., the extent of tissue involvement and disease spread). 197 Even in the presence of various confounding medical conditions, the test maintains 198 its robustness and reliability, emphasizing its independence from potential confounding 199 factors with 100% of the controls being negative. This characteristic supports its suitability 200 for use in various clinical settings, irrespective of the patient's medical history, thereby 201 ensuring its applicability without contraindications. 202 4. Materials and Methods 203 4.1. Patients’ characteristics and classification 204 The current report is a prospective analysis study using biobank samples. A total of 205 281 samples extracted from the renowned Oxford Endometriosis CaRe Centre biobank in 206 the UK were included for the development (I) and external validation (II) studies. The 207 biobank's repository comprised meticulously curated serum samples and comprehensive 208 clinical information derived from pre-surgical assessments and post-operative procedures 209 of patients within reproductive age (18–50 years old) undergoing laparoscopy because of 210 a suspicion of endometriosis. Patients were classified as cases or controls based on lapa- 211 roscopy and thorough evaluation of histological findings. After undergoing laparoscopy, 212 patients diagnosed with endometriosis were categorized into stages according to the re- 213 vised American Society of Reproductive Medicine (rASRM) classification. Patients who 214 had not used hormones in the 3 months prior to surgery were selected. 215 136 endometriosis cases and 68 controls were included in the development study 216 (n=204). For the validation study (n=77), 52 cases and 25 controls were included. The 217 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 7 of 14 demographic characteristics of those patients are available in Table 5. The experimental 218 procedures received approval from the Ethics Committee of CEIm HM Hospitales (codes: 219 19.05.1411-GHM and 22.03.2001-GHM). 220 221 Development study (I) Validation study (II) Controls N=68 Cases N=136 Controls N=25 Cases N=52 Age years (mean ± SD) 33.5 (5.96) 35.6 (6.42) 35 (6.44) 35 (6.47) BMI (mean ± SD) 25.38 (4.63) 26.46 (5.32) 26 (5.23) 26 (5.14) rASRM classification I–II III–IV Missing information - - 68 (50%) 68 (50%) - - - - 42 (81%) 7 (13%) 3 (6%) 222 Table 5. Demographic characteristics and rASRM classification of the patients in the 223 development (I) and validation (II) studies. 224 225 4.2. Lesion location and subtyping 226 Imaging findings and surgical examinations have been reported for each subject in- 227 cluded in the study. Endometriosis lesions were investigated by location. From these find- 228 ings, endometriosis lesions were classified into subgroups according to their location in 229 the ovaries and the peritoneal cavity: superficial (< 5 mm depth), endometrioma, and/or 230 deep infiltrative endometriosis (DIE). Specifically, the designation "superficial" was as- 231 signed when only superficial endometriosis lesions were identified in the ovaries or peri- 232 toneal cavity. The classification of "endometrioma" was used when endometriomas were 233 detected in the ovaries, either with or without accompanying superficial endometriosis. 234 In cases where infiltrative lesions were observed in the peritoneal cavity, with or without 235 associated superficial endometriosis, lesions were classified as "DIE". Moreover, the "en- 236 dometrioma + DIE" classification was assigned when both DIE and endometriomas were 237 found in the peritoneal cavity, with or without superficial endometriosis. While endome- 238 triosis is thought to be caused by retrograde menstruation, the most likely cause of cae- 239 sarean section (c-section) scar endometriosis is iatrogenic implantation. Due to this differ- 240 ent aetiology, 2 patients with c -section scar endometriosis were misclassified as they 241 should fall under a different category than endometriosis with spontaneous implantation. 242 The distributions of cases of the development and validation studies by lesions type are 243 described in Table 6. 244 245 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 8 of 14 246 247 Endometriosis Classification Development study (I) N=136 Validation study (II) N=52 Superficial 54 (39.7%) 24 (46.2%) Endometrioma 25 (18.4%) 3 (5.8%) DIE 28 (20.6%) 13 (25%) DIE + endometri- oma 25 (18.4%) 8 (15.4%) Unclassified 4 (2.9%) 2(3.8%) C-section scar 0 2 (3.8%) 248 Table 6. Classification of endometriosis cases according to lesion location. 249 250 4.3. Confounding disease screening 251 Patients were asked to fill out a presurgical survey including a question to indicate 252 the absence/presence of confounding medical conditions from a list. They were asked: 253 please mark whether you have had any of the following medical conditions, and at what 254 age you were first diagnosed by a doctor (please tick all that apply)” and were given the 255 list of medical conditions. Patients were also asked to indicate whether they were affected 256 by other unlisted medical conditions. This survey was administered to patients in one of 257 its 3 versions: version #1 did not list 3 medical conditions: Anxiety (1), cardiovascular dis- 258 ease (2) and high blood pressure (3). These conditions were only listed in questionnaires 259 #2 and #3. Versions #2 and #3 were responded by 141 out of 190 patients included in the 260 development study (14 patients did not answer to this question out of 204) and 64 out of 261 77 patients included in the validation study. For completeness, imaging and surgical find- 262 ings were used to further identify patients with gynaecological conditions. 263 Table 7 depicts the prevalence of the confounding conditions in patients included in 264 the development and validation studies. 265 266 Confounding condition Prevalence in devel- opment study Prevalence in validation study Anxiety requiring medication or ther- apy 39/141 (28%) 20/64 (31%) Asthma 42/190 (22%) 23/77 (30%) Adenomyosis 7/190 (3.7%) 2/77 (2.6%) Cardiovascular disease 0 0 Crohn’s disease 0 0 Chronic fatigue syndrome - Myalgic encephalomyelitis 10/190 (5.2%) 1/77 (1.3%) Depression requiring medication or therapy 68/190 (35.8%) 29/77 (38%) Diabetes requiring diet control 3/190 (1.6%) 0 Diabetes requiring insulin or tablets 1/190 (0.5%) 0 Eczema 32/190 (16.8%) 16/77 (21%) Uterine fibroids 28/190 (9.5%) 7/77 (9.1%) Fibromyalgia 4/190 (2%) 1/77 (1.3%) All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 9 of 14 Glandular fever 17/190 (8.9%) 5/77 (6.5%) Graves’s disease 0 0 Hashimoto’s disease 0 0 High blood pressure 9/141 (6%) 4/64 (6.2%) Irritable bowel syndrome 43/190 (23%) 17/77 (22%) Interstitial Cystitis 12/190 (62.5%) 7/77 (9%) Migraine 51/190 (27%) 22/77 (28.6%) Mitral valve prolapse 2/190 (1%) 0 Multiple sclerosis 2/190 (1%) 0 Ovarian cysts 93/190 (49%) 27/77 (35.1%) Pelvic inflammatory disease 13/190 (6.84%) 6/77 (7.8%) Polycystic ovarian syndrome 32/190 (16.8%) 17 (22.1%) Rheumatoid arthritis 0 0 Sjogren’s syndrome 0 1/77 (1.3%) Systemic lupus erythematosus 0 0 Thyroid disease 3/190 (1.6%) 0 Ulcerative colitis 1/190 (0.5%) 2/77 (2.6%) 267 268 Table 7. Prevalence of confounding conditions in the development and validation 269 datasets. 270 271 272 273 274 4.4. Blood sample collection and biomarkers measurement 275 The specimens were gathered and managed with explicit patient consent, following 276 the guidelines outlined in the Standard Operating procedures of the World Endometriosis 277 Research Foundation [41]. Before the collection of blood, patients were instructed to main- 278 tain a minimum fasting period of 10 hours. The serum samples were then preserved in 279 the biobank at temperatures as low as -80 ºC for a duration of up to 5 years, after which 280 they were transferred to the designated laboratory for analysis. The ELISA utilized in this 281 in vitro diagnostic test functions as a solid-phase sandwich enzyme-immunoassay for the 282 precise determination of BDNF and CA125 levels within human serum [22]. 283 284 4.5. Data treatment algorithm 285 All the necessary input parameters, including serum CA125, serum BDNF, and clin- 286 ical variables were gathered. Subsequently, laboratory technicians input this data into the 287 IVD test diagnostic medical software, which houses the data treatment algorithm. The 288 algorithm processed the input and generated outcomes, classifying them as either positive 289 or negative based on whether the value exceeded or fell below the predetermined thresh- 290 old value, respectively. 291 292 4.6. Statistical analysis 293 Statistical analysis was conducted utilizing R software, version 4.1.3, provided by the 294 R Foundation for Statistical Computing in Vienna, Austria. The statistical significance 295 level was set at p < 0.05, indicating a threshold below which results were considered sta- 296 tistically significant. In the validation study, the IVD test software was utilized to compute 297 algorithm scores and their corresponding outcomes. These outcomes were delineated as 298 positive diagnosis when the score surpassed the defined cut -off, and negative diagnosis 299 when the score fell below the defined cut -off. Specifically, the validation study's 300 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 10 of 14 sensitivity and specificity were expected to align with or exceed the lower limits of the 301 sensitivity and specificity 95% confidence intervals outlined in the algorithm development 302 study: an AUC of 0.867 with a sensitivity of 51.5% (42.8 - 60.1) at a specificity of 95.6% 303 (86.8 - 98.9%) as reported by Herranz et al. To assess the IVD test clinical performance, the 304

of the primary performance parameters (sensitivity and specificity) were con- 305 trasted with the acceptance criteria values established during the development study. To 306 ensure equitable representation of both the low -stage and high -stage groups, the out- 307 comes in the validation were appropriately weighted. 308 To further elucidate the performance of the IVD test, the sensitivity for each endome- 309 triosis classification, with the values specified alongside their respective 95% CI are re- 310 ported for the distinct subgroups based on lesion types. For a more comprehensive assess- 311 ment of test’s efficacy over a larger sample size, development and validation datasets were 312 pooled. Analysis was run on pooled dataset. BDNF values in pooled datasets followed a 313 normal distribution and CA125 values were arithmetically transformed to follow a normal 314 distribution. To investigate the effect of confounding diseases on biomarkers levels and 315 the performance of the test, only conditions with >1% prevalence in both datasets were 316 considered. A two -way ANOVA analysis was conducted to assess the effect of medical 317 conditions and EndoState (Cases/controls) on CA125 and BDNF, respectively, including 318 an interaction term. Only conditions showing significant main effects or interaction will 319 be reported. Furthermore, the performance of the algorithm on validation data was eval- 320 uated after excluding each specific conditions, one at a time. 321 322 5. Conclusions 323 Overall, the high specificity of the test, coupled with its independence from potential 324 confounding medical conditions, position it as a valuable and reliable tool for the accurate 325 and timely diagnosis of endometriosis. 326 6. Patents 327 There is a patent resulting from the work reported in this manuscript. 328 329 Author Contributions: For research articles with several authors, a short paragraph specifying their 330 individual contributions must be provided. The following statements should be used “Conceptual- 331 ization, E.D. and B-H.-B; methodology, E.D. and B-H.-B; software, E.D. and B-H.-B; validation, E.D. 332 and B-H.-B.; formal analysis, E.D. and B-H.-B; investigation, E.D. and B-H.-B; resources, E.D and B- 333 H.-B.; data curation, E.D. and B-H.-B.; writing—original draft preparation, E.D.; writing—review 334 and editing, E.D., B.H.-B, D.A.; visualization, E.D..; supervision, B.H.; project administration, E.D. 335 and B-H.-B.; funding acquisition, E.D. and B -H.-B. All authors have read and agreed to the pub- 336 lished version of the manuscript.” Please turn to the CRediT taxonomy for the term explanation. 337 Authorship must be limited to those who have contributed substantially to the work reported. 338 Funding: Exeltis (represented by Chemo Research S.L.) has fully sponsored the studies. 339 Institutional Review Board Statement: The study was conducted in accordance with the Declara- 340 tion of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of CEIm 341 HM Hospitales (codes: 19.05.1411-GHM and 22.03.2001-GHM and date of approval: April 12th, 2022 342 Informed Consent Statement: Informed consent was obtained from all subjects involved in the 343 study. The experimental procedures received approval from the Ethics Committee of CEIm HM 344 Hospitales (codes: 19.05.1411-GHM and 22.03.2001-GHM). 345 346 Conflicts of Interest: The authors E.D. and B.H.-B. were employed by the company Exeltis (repre- 347 sented by Chemo Research S.L.). 348 349 350 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 11 of 14

351 352 [1] Zondervan KT, Becker CM, Missmer SA. Endometriosis. New England Journal of Medicine 353 2020;382:1244–56. https://doi.org/10.1056/NEJMra1810764. 354 [2] Bulun SE, Yilmaz BD, Sison C, Miyazaki K, Bernardi L, Liu S, et al. Endometriosis. Endocr Rev 355 2019;40:1048–79. https://doi.org/10.1210/er.2018-00242. 356 [3] Symons LK, Miller JE, Kay VR, Marks RM, Liblik K, Koti M, et al. The Immunopathophysiol- 357 ogy of Endometriosis. Trends in Molecular Medicine 2018;24:748 –62. 358 https://doi.org/10.1016/j.molmed.2018.07.004. 359 [4] Brosens I, Gordts S, Benagiano G. Endometriosis in adolescents is a hidden, progressive and 360 severe disease that deserves attention, not just compassion. Hum Reprod 2013;28:2026–31. 361 https://doi.org/10.1093/humrep/det243. 362 [5] Schliep KC, Mumford SL, Peterson CM, Chen Z, Johnstone EB, Sharp HT, et al. Pain typology 363 and incident endometriosis. Hum Reprod 2015;30:2427 –38. https://doi.org/10.1093/hum- 364 rep/dev147. 365 [6] Johnson NP, Hummelshoj L, Adamson GD, Keckstein J, Taylor HS, Abrao MS, et al. World 366 Endometriosis Society consensus on the classification of endometriosis. Hum Reprod 2017;32:315– 367 24. https://doi.org/10.1093/humrep/dew293. 368 [7] Agarwal SK, Chapron C, Giudice LC, Laufer MR, Leyland N, Missmer SA, et al. Clinical diag- 369 nosis of endometriosis: a call to action. Am J Obstet Gynecol 2019;220:354.e1 -354.e12. 370 https://doi.org/10.1016/j.ajog.2018.12.039. 371 [8] Speer LM, Mushkbar, Erbele T. Chronic Pelvic Pain in Women. American Family Physician 372 2016;93. 373 [9] Alborzi S, Poordast T, Askary E, Javedanmehr M, Shahraki MH, Alborzi S, et al. Evaluation 374 and comparison of the accuracy of transvaginal ultrasound and MRI for the diagnosis of deep rec- 375 tal endometriosis: A cross -sectional study. International Journal of Gynecology & Obstetrics 376 2023;161:586–93. https://doi.org/10.1002/ijgo.14554. 377 [10] Bruyere C, Maniou I, Habre C, Kalovidouri A, Pluchino N, Montet X, et al. Pelvic MRI for 378 Endometriosis: A Diagnostic Challenge for the Inexperienced Radiologist. How Much Experience 379 Is Enough? Academic Radiology 2021;28:345–53. https://doi.org/10.1016/j.acra.2020.02.023. 380 [11] Wang S, Duan H, Li B, Hong W, Li X, Wang Y, et al. BDNF and TrKB expression levels in 381 patients with endometriosis and their associations with dysmenorrhoea. Journal of Ovarian Re- 382 search 2022;15:35. https://doi.org/10.1186/s13048-022-00963-9. 383 [12] Taylor HS, Kotlyar AM, Flores VA. Endometriosis is a chronic systemic disease: clinical chal- 384 lenges and novel innovations. Lancet 2021;397:839 –52. https://doi.org/10.1016/S0140 - 385 6736(21)00389-5. 386 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 12 of 14 [13] Shim JY, Laufer MR. Adolescent Endometriosis: An Update. J Pediatr Adolesc Gynecol 387 2020;33:112–9. https://doi.org/10.1016/j.jpag.2019.11.011. 388 [14] Parasar P, Ozcan P, Terry KL. Endometriosis: Epidemiology, Diagnosis and Clinical Manage- 389 ment. Curr Obstet Gynecol Rep 2017;6:34–41. https://doi.org/10.1007/s13669-017-0187-1. 390 [15] Ahn SH, Singh V, Tayade C. Biomarkers in endometriosis: challenges and opportunities. Fer- 391 tility and Sterility 2017;107:523–32. https://doi.org/10.1016/j.fertnstert.2017.01.009. 392 [16] Nisenblat V, Bossuyt PMM, Farquhar C, Johnson N, Hull ML. Imaging modalities for the non- 393 invasive diagnosis of endometriosis. Cochrane Database Syst Rev 2016;2:CD009591. 394 https://doi.org/10.1002/14651858.CD009591.pub2. 395 [17] Hirsch M, Duffy JMN, Deguara CS, Davis CJ, Khan KS. Diagnostic accuracy of Cancer Antigen 396 125 (CA125) for endometriosis in symptomatic women: A multi-center study. European Journal of 397 Obstetrics & Gynecology and Reproductive Biology 2017;210:102 –7. 398 https://doi.org/10.1016/j.ejogrb.2016.12.002. 399 [18] Hirsch M, Duffy J, Davis C, Nieves Plana M, Khan K, on behalf of the International Collabo- 400 ration to Harmonise Outcomes and Measures for Endometriosis. Diagnostic accuracy of cancer 401 antigen 125 for endometriosis: a systematic review and meta-analysis. BJOG: An International Jour- 402 nal of Obstetrics & Gynaecology 2016;123:1761–8. https://doi.org/10.1111/1471-0528.14055. 403 [19] Shen A, Xu S, Ma Y, Guo H, Li C, Yang C, et al. Diagnostic value of serum CA125, CA19-9 and 404 CA15-3 in endometriosis: A meta -analysis. J Int Med Res 2015;43:599 –609. 405 https://doi.org/10.1177/0300060515583076. 406 [20] Perricos A, Ashjaei K, Husslein H, Proestling K, Kuessel L, Obwegeser R, et al. Increased se- 407 rum levels of mBDNF in women with minimal and mild endometriosis have no predictive power 408 for the disease. Exp Biol Med (Maywood) 2018;243:50–6. https://doi.org/10.1177/1535370217742600. 409 [21] Wessels JM, Wu L, Leyland NA, Wang H, Foster WG. The brain -uterus connection: brain de- 410 rived neurotrophic factor (BDNF) and its receptor (Ntrk2) are conserved in the mammalian uterus. 411 PLoS One 2014;9:e94036. https://doi.org/10.1371/journal.pone.0094036. 412 [22] Herranz-Blanco B, Daoud E, Viganò P, García -Velasco JA, Colli E. Development and Valida- 413 tion of an Endometriosis Diagnostic Method Based on Serum Biomarkers and Clinical Variables. 414 Biomolecules 2023;13:1052. https://doi.org/10.3390/biom13071052. 415 [23] Tang Y, Wen M-B, Xiang R-M, Yang M-T, Shu B, Xu F, et al. Serum CA125 as a biomarker for 416 dysmenorrhea in adenomyosis. Int J Gynaecol Obstet 2023;163:131 –9. 417 https://doi.org/10.1002/ijgo.14832. 418 [24] Chen Q, Li Y-W, Wang S, Fan Q -B, Shi H-H, Leng J-H, et al. Clinical Manifestations Of Ade- 419 nomyosis Patients With Or Without Pain Symptoms. J Pain Res 2019;12:3127 –33. 420 https://doi.org/10.2147/JPR.S212117. 421 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 13 of 14 [25] Kil K, Chung J-E, Pak HJ, Jeung I -C, Kim JH, Jo HH, et al. Usefulness of CA125 in the differ- 422 ential diagnosis of uterine adenomyosis and myoma. Eur J Obstet Gynecol Reprod Biol 423 2015;185:131–5. https://doi.org/10.1016/j.ejogrb.2014.12.008. 424 [26] Burghaus S, Drazic P, Wölfler M, et al. Multicenter evaluation of blood-based biomarkers for 425 the detection of endometriosis and adenomyosis: A prospective non-interventional study. Int J Gy- 426 naecol Obstet 2023. 427 [27] Park Y, Lee J-H, Hong DJ, Lee EY, Kim H -S. Diagnostic performances of HE4 and CA125 for 428 the detection of ovarian cancer from patients with various gynecologic and non -gynecologic dis- 429 eases. Clin Biochem 2011;44:884–8. https://doi.org/10.1016/j.clinbiochem.2011.04.011. 430 [28] Moley KH, Massad LS, Mutch DG. Pelvic inflammatory disease. Correlation of severity with 431 CA-125 levels. J Reprod Med 1996;41:341–6. 432 [29] Meden H, Fattahi-Meibodi A. CA 125 in benign gynecological conditions. Int J Biol Markers 433 1998;13:231–7. https://doi.org/10.1177/172460089801300411. 434 [30] Moss EL, Hollingworth J, Reynolds TM. The role of CA125 in clinical practice. J Clin Pathol 435 2005;58:308–12. https://doi.org/10.1136/jcp.2004.018077. 436 [31] Babacan A, Kizilaslan C, Gun I, Muhcu M, Mungen E, Atay V. CA 125 and other tumor mark- 437 ers in uterine leiomyomas and their association with lesion characteristics. Int J Clin Exp Med 438 2014;7:1078–83. 439 [32] Sharma D, Vinocha A. Benign Ovarian Cysts with Raised CA-125 Levels: Do We Need to Eval- 440 uate the Fallopian Tubes? J Lab Physicians 2020;12:276–80. https://doi.org/10.1055/s-0040-1722547. 441 [33] Ataseven H, Oztürk ZA, Arhan M, Yüksel O, Köklü S, Ibiş M, et al. Cancer antigen 125 levels 442 in inflammatory bowel diseases. J Clin Lab Anal 2009;23:244–8. https://doi.org/10.1002/jcla.20323. 443 [34] del Porto F, Aloe L, Laganà B, Triaca V, Nofroni I, D’Amelio R. Nerve growth factor and brain- 444 derived neurotrophic factor levels in patients with rheumatoid arthritis treated with TNF -alpha 445 blockers. Ann N Y Acad Sci 2006;1069:438–43. https://doi.org/10.1196/annals.1351.042. 446 [35] Tsavaris N, Mavragani CP, Dimitrios P. Rheumatoid arthritis: correlation between rheuma- 447 toid factor levels and CA-125 tumour marker elevation. Ann Rheum Dis 2007;66:980. 448 https://doi.org/10.1136/ard.2007.070052. 449 [36] Lai N-S, Yu H-C, Huang Tseng H-Y, Hsu C-W, Huang H-B, Lu M-C. Increased Serum Levels 450 of Brain -Derived Neurotrophic Factor Contribute to Inflammatory Responses in Patients with 451 Rheumatoid Arthritis. Int J Mol Sci 2021;22:1841. https://doi.org/10.3390/ijms22041841. 452 [37] Wang S, Sathish V, Freeman M, Thompson M, Pabelick CM, Prakash YS. Secreted Brain -De- 453 rived Neurotrophic Factor and Asthma Severity. Am J Respir Cell Mol Biol 2016;54:297. 454 https://doi.org/10.1165/rcmb.2015-0265LE. 455 [38] Shen Z, Zhu J, Yuan Y, Ren L, Qian M, Lin M, et al. The roles of brain -derived neurotrophic 456 factor (BDNF) and glial cell line -derived neurotrophic factor (GDNF) in predicting treatment 457 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint 14 of 14 remission in a Chinese Han population with generalized anxiety disorder. Psychiatry Res 458 2019;271:319–24. https://doi.org/10.1016/j.psychres.2018.08.111. 459 [39] Suliman S, Hemmings SMJ, Seedat S. Brain-Derived Neurotrophic Factor (BDNF) protein lev- 460 els in anxiety disorders: systematic review and meta -regression analysis. Front Integr Neurosci 461 2013;7:55. https://doi.org/10.3389/fnint.2013.00055. 462 [40] Satomura E, Baba H, Nakano Y, Maeshima H, Suzuki T, Arai H. Correlations between brain - 463 derived neurotrophic factor and clinical symptoms in medicated patients with major depression. J 464 Affect Disord 2011;135:332–5. https://doi.org/10.1016/j.jad.2011.06.041. 465 [41] Rahmioglu N, Fassbender A, Vitonis AF, Tworoger SS, Hummelshoj L, D’Hooghe TM, et al. 466 World Endometriosis Research Foundation Endometriosis Phenome and Biobanking Harmoniza- 467 tion Project: III. Fluid biospecimen collection, processing, and storage in endometriosis research. 468 Fertil Steril 2014;102:1233–43. https://doi.org/10.1016/j.fertnstert.2014.07.1208. 469 [42] Guerriero S, Ajossa S, Pagliuca M, Borzacchelli A, Deiala F, Springer S, et al. Advances in Im- 470 aging for Assessing Pelvic Endometriosis. Diagnostics (Basel) 2022;12:2960. 471 https://doi.org/10.3390/diagnostics12122960. 472 [43] Robinson AJ, Rombauts L, Ades A, Leong K, Paul E, Piessens S. Poor sensitivity of transvagi- 473 nal ultrasound markers in diagnosis of superficial endometriosis of the uterosacral ligaments. Jour- 474 nal of Endometriosis and Pelvic Pain Disorders 2018;10:10 –7. 475 https://doi.org/10.1177/2284026518767259. 476 [44] Kitawaki J, Ishihara H, Koshiba H, Kiyomizu M, Teramoto M, Kitaoka Y, et al. Usefulness and 477 limits of CA-125 in diagnosis of endometriosis without associated ovarian endometriomas. Hum 478 Reprod 2005;20:1999–2003. https://doi.org/10.1093/humrep/deh890. 479 480 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 19, 2024. ; https://doi.org/10.1101/2024.04.17.24305952doi: medRxiv preprint