{"paper_id":"3701507c-513d-444e-9ce1-e617a8a5c827","body_text":"Clin. Exp. Obstet. Gynecol. 2026; 53(5): 47790\nhttps://doi.org/10.31083/CEOG47790\nCopyright: © 2026 The Author(s). Published by IMR Press.\nThis is an open access article under the CC BY 4.0 license .\nPublisher’s Note: IMR Press stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.\nOriginal Research\nIncremental Diagnostic Value of the Plasma Kynurenine/Tryptophan\nRatio for Deep Infiltrating Endometriosis and Its Correlation With\nEutopic Endometrial Matrix Metalloproteinase-9 Expression: A\nSingle-Center Prospective Study\nHubin Xu1, Haimin Jiang 1, Chenmei Ma 2,*\n1Hangzhou Normal University, 31112 Hangzhou, Zhejiang, China\n2Department of Obstetrics, Y ongkang Hospital of Traditional Chinese Medicine, 321300 Y ongkang, Zhejiang, China\n*Correspondence: Cm9498156@163.com (Chenmei Ma)\nAcademic Editor: Osamu Hiraike\nSubmitted: 29 October 2025 Revised: 8 December 2025 Accepted: 6 January 2026 Published: 26 May 2026\nAbstract\nBackground: Surgery for deep infiltrating endometriosis (DIE) is complex, and current clinical imaging has limited ability to identify\nand stratify atypical lesions. The kynurenine/tryptophan ratio (KTR) reflects immunometabolic activation, and matrix metalloproteinase-\n9 (MMP-9) is associated with tissue invasion. Methods: This was a single-center prospective study, with pathology as the gold standard.\nThree analysis populations were defined. KTR was measured by liquid chromatography-tandem mass spectrometry (LC-MS/MS), and\nMMP-9 by immunohistochemistry (IHC) derived histochemical score (H-score). Multivariable regression, DeLong comparison, cal-\nibration and decision curves, and nested F test were used. Results: Compared with controls, KTR differed in the DIE and non-DIE\ngroups, with β_diff = 0.28/0.15 (both p_adj < 0.05). Each 1 SD increase in natural log transformed KTR (lnKTR) was associated with\nhigher ENZIAN stage (OR_perSD = 1.62, p < 0.001). After KTR was added to the baseline model, the area under the receiver operating\ncharacteristic curve (AUC) increased from 0.79 to 0.83 ( ∆AUC = 0.04, 95% CI: 0.02–0.07, p = 0.003). Net benefit increased across the\n10%–30% threshold range, and calibration improved from α = –0.127, β = 0.861 to α = –0.039 and β = 0.946. Among imaging suspected\nbut atypical subjects, net correct reclassification was +14 for events and +17 for non-events. KTR was independently and positively asso-\nciated with MMP-9 (β_std = 0.29, 95% CI: 0.15–0.43, p < 0.001), with ∆R2 = 0.04 (F = 11.882,p < 0.001). Conclusions: KTR provides\nan independent and translatable diagnostic increment on current pathways and is associated with eutopic endometrial MMP-9, supporting\ncoupling between systemic immunometabolism and local remodeling. These findings support its use for preoperative stratification and\noptimization of surgical planning.\nKeywords: deep infiltrating endometriosis; kynurenine/tryptophan ratio; diagnostic gain; MMP-9\n1. Introduction\nEndometriosis is a common chronic disease in women\nof reproductive age, and the deep infiltrating subtype rep-\nresents a high burden phenotype with marked invasive-\nness, increased complications, and greater surgical com-\nplexity [ 1,2]. Current diagnosis and treatment rely on\nsymptoms, clinical examination, transvaginal ultrasound\n(TVUS), and magnetic resonance imaging (MRI). How-\never, detection of lesions in complex locations or with\natypical features remains unstable, and serum carbohydrate\nantigen 125 (CA-125) shows limited discriminative abil-\nity for phenotype and depth of infiltration [ 3]. Histolog-\nically, matrix metalloproteinase-9 (MMP-9) mediates ex-\ntracellular matrix degradation and fibrotic remodeling and\nis considered with a marker of tissue infiltration [ 4]. In\nterms of immunometabolism, interferon-related pathways\ncan activate the kynurenine pathway by directing trypto-\nphan through indoleamine 2,3-dioxygenase and tryptophan\n2,3-dioxygenase. Kynurenine, together with aryl hydrocar-\nbon receptor (AhR) signaling, is hypothesized to contribute\nto tissue invasion [ 5]. The plasma kynurenine/tryptophan\nratio (KTR) provides a systemic measure of this pathway\nhas potential as an easily accessible biomarker [ 6]. Sero-\nlogic studies targeting the deep infiltrating subtype mostly\nadopt cross-sectional designs and report receiver operating\ncharacteristic (ROC) curves for single biomarkers. They\nrarely evaluate the independent incremental value beyond\nreal-world clinical and imaging pathways and generally\nlack evidence of clinical net benefit demonstrated by re-\nclassification or decision curve analysis. A pervious study\nhas inadequately controlled key confounders, such as men-\nstrual cycle, exogenous hormones, inflammation, and renal\nfunction, and often provide incomplete pre-analytical and\nlaboratory quality-control (QC) information, limiting their\ngeneralizability [ 7]. Key gaps remain regarding whether,\nunder standardized procedures, kynurenine pathway mark-\ners show gradient associations with phenotype and infil-\ntration load, whether they provide calibratable incremental\nvalue with net benefit beyond a baseline model composed\nof symptoms, signs, CA-125, and imaging [ 8], and whether\n\nsystemic markers are independently associated with MMP9\nin the eutopic endometrium at the individual level. These\nkey gaps limit effective preoperative stratification and sur-\ngical planning [ 9]. In this single-center prospective study,\nwe aimed to clarify the gradient association of the plasma\nKTR with the phenotype and infiltration load of deep in-\nfiltrating endometriosis (DIE) based on existing clinical\nand imaging pathways, to evaluate its independent incre-\nmental contribution to diagnostic performance, and to ex-\nplore its relationship with MMP-9 expression in the eu-\ntopic endometrium. Using rigorous pre-analytical proce-\ndures and laboratory validation, plasma KTR was quan-\ntified by liquid chromatography-tandem mass spectrome-\ntry (LC-MS/MS). A baseline prediction model including\nsymptoms, signs, CA-125, and imaging indicators was con-\nstructed, and discrimination, calibration, reclassification,\nand decision curve analyses were used to evaluate the di-\nagnostic gain provided by KTR. MMP-9 expression in the\neutopic endometrium was quantified by immunohistochem-\nistry (IHC), and multivariable regression together with ro-\nbustness checks was used to examine the association be-\ntween the systemic immunometabolic marker to local ma-\ntrix remodeling. The results showed that KTR provided\nquantifiable diagnostic information with clinical net benefit\non top of the existing pathway and was independently and\npositively associated with MMP-9 expression in the eutopic\nendometrium, providing an actionable biological basis for\npreoperative risk stratification and kynurenine pathway-\ntargeted precision interventions.\n2. Materials and Methods\n2.1 Study Design and Sample Size Estimation\nThis study was a single-center, prospective, diagnos-\ntic gain evaluation with paired histological correlation. The\nprimary objective was to assess the independent incremen-\ntal value of the plasma KTR for the diagnosis and preoper-\native stratification of DIE beyond a baseline pathway com-\nprising symptoms, clinical signs, CA-125, and imaging.\nThe paired analysis of eutopic endometrial MMP-9 expres-\nsion aimed to provide evidence of biological plausibility for\nKTR as an auxiliary biomarker, rather than a systematic in-\nvestigation of pathogenesis.\nParticipants were consecutively enrolled along the\nclinical pathway. All testing was completed before surgery\nand pathological determination, and the study team main-\ntained mutual blinding across all components. Ethical ap-\nproval was obtained prior to study initiation (Approval\nNo. KT2022062), and all participants provided written in-\nformed consent. Participants were enrolled from January 1,\n2023 to December 31, 2024, and the data were finalized on\nMarch 31, 2025. Sample size was pre-calculated based on\ntwo primary objectives. Objective 1 considered the differ-\nence in the area under the receiver operating characteristic\ncurve (AUC) for KTR added to the “baseline model” (De-\nLong method, two-sided α = 0.05, effect ∆AUC = 0.04,\nevent proportion 0.35, correlation coefficient within the\nprediction population 0.60, power 0.80), requiring a total\nof 216 participants. Objective 2 assessed the standardized\ndifference in KTR between DIE and non-DIE (Cohen’s d =\n0.5, two-sided α = 0.05, power 0.90), requiring ≥85 partic-\nipants per group. Considering both objectives and an antic-\nipated 10% rate of loss to follow-up or unusable specimens,\nthe planned total sample size was 270, with consecutive en-\nrollment stratified into three categories, DIE, non-DIE en-\ndometriosis, and non-endometriosis controls, targeting an\napproximately balanced distribution to preserve the power\nof the primary analyses; no fixed quotas were set.\n2.2 Study Population\nThe study population consisted of women of repro-\nductive age scheduled to undergo laparoscopy or laparo-\ntomy. Inclusion criteria: age 18–50 years; planned deter-\nmination of the presence or absence of endometriosis in\nthe current surgery; completion of symptom quantification,\npelvic imaging, and blood sampling before surgery; and\nconsent to obtain eutopic endometrial samples at surgery.\nExclusion criteria: fever or active infection within the past\nfour weeks; use of systemic glucocorticoids or immunosup-\npressants within two weeks before surgery; chronic renal\ninsufficiency [estimated glomerular filtration rate (eGFR)\n<60 mL·min−1·1.73 m−2]; pregnancy or lactation; history\nof malignancy; inability to complete follow-up; or inade-\nquate specimen handling.\n2.3 Outcome Ascertainment and Disease Grading\nThis study used surgical exploration combined with\nhistopathological examination as the gold standard. Par-\nticipants were classified by lesion type into three mutu-\nally exclusive study groups: ¬ DIE group: suspicious le-\nsions were found during surgery in deep structures such\nas the uterosacral ligaments, rectovaginal septum, vaginal\nfornix, bladder, and bowel wall; histopathology confirmed\nthe presence of endometrial-like glands and/or stroma, and\nthe lesions involved subperitoneal tissue with an infiltra-\ntion depth ≥5 mm, consistent with the definition of DIE; ­\nnon-DIE endometriosis group: histopathology confirmed\nthe presence of endometriosis, but the lesions were confined\nto ovarian endometriomas and/or superficial peritoneal le-\nsions, without involvement of the above deep structures\nor not meeting the DIE criterion of infiltration depth ≥5\nmm; ® non-endometriosis control group: reproductive-age\nwomen scheduled to undergo laparoscopy or laparotomy\nfor benign non-inflammatory pelvic conditions such as uter-\nine fibroids or simple ovarian cysts. All patients underwent\nstandardized TVUS preoperatively, and, when necessary,\npelvic MRI within 90 days before surgery to assess for sus-\npicious signs of endometriosis. Intraoperatively, the lead\nsurgeon systematically explored typical sites of involve-\nment, including the pelvic peritoneum, ovaries, uterosacral\nligaments, rectovaginal septum, vaginal fornix, bladder,\n2\n\n\nand bowel wall. Suspicious lesions, such as ovarian cyst\nwalls and peritoneal pigmented or fibrotic areas, were rou-\ntinely excised or biopsied for pathology. Only when preop-\nerative imaging did not suggest endometriosis, no typical\nor suspicious endometriotic lesions were seen intraopera-\ntively, and all suspicious lesions were confirmed by pathol-\nogy to show no endometriosis at any site, the participants\nwere included in the non-endometriosis control group.\nAll pathological slides were independently reviewed\nby two senior pathologists blinded to the KTR results, and\nany discrepancies were resolved by discussion to reach con-\nsensus. The primary outcome was the presence or absence\nof DIE (binary), which was used for diagnostic modeling\nand gain analyses. At the same time, based on surgical\nrecords and imaging data, trained investigators completed\nthe ENZIAN classification, recording the involvement lev-\nels of zones A, B, C, and F (grades 0–3), as well as the total\nnumber of involved sites, and calculating the overall score\n[10], to quantify infiltration burden and disease severity.\n2.4 Specimen Collection and Laboratory Testing\n2.4.1 Plasma KTR Measurement\nWithin 14 days before surgery, 5 mL of EDTA-\nanticoagulated blood was collected from the antecubital\nvein in the early morning under fasting conditions. Within\n1 hour after collection, plasma was separated by centrifu-\ngation at 4 °C, 1500 g for 10 minutes, aliquoted into pre-\ncooled polypropylene tubes, stored at –80 °C, with a max-\nimum of one freeze-thaw cycle. The date of last men-\nstrual period, menstrual cycle phase (early follicular phase\nprioritized for sampling), exogenous hormone use, inter-\nval from blood draw to surgery, hemolysis index, and as-\nsay batch were recorded. A methodologically validated as-\nsay was used to quantify kynurenine and tryptophan, us-\ning stable isotope internal standards to correct for matrix\neffects and recovery. Calibration curves covered 0.5–20\nµmol·L−1 for kynurenine and 20–200 µmol ·L−1 for tryp-\ntophan, with the lower limit of quantification and linear\ncorrelation coefficients pre-specified in the validation re-\nport and meeting QC acceptance [11]. Low-, medium-, and\nhigh-level QC materials were included in each batch, and\nboth within-batch and between-batch coefficients of varia-\ntion were maintained at 10%. KTR was calculated as the\nratio of kynurenine/tryptophan and natural log transformed\n(lnKTR) to improve distributional characteristics. Labora-\ntory personnel were blinded to clinical groupings and out-\ncomes throughout the study.\n2.4.2 Eutopic Endometrial Sampling and MMP-9\nExpression Detection\nEutopic endometrial sampling was performed at the\nstart of surgery to avoid interference from intraoperative\nenergy devices. Tissues were immediately fixed in 10%\nneutral buffered formalin for 24 hours, routinely dehy-\ndrated and embedded, and sectioned at 4 µm. A laboratory-\nvalidated rabbit monoclonal anti-MMP-9 antibody was\nused with heat-induced antigen retrieval in a pH 9.0 buffer,\na polymer detection system with 3,3 ′-diaminobenzidine\n(DAB) chromogen, and hematoxylin counterstaining. Each\nbatch included both a negative control and a known pos-\nitive control. Two pathologists, blinded to KTR and out-\ncomes, independently evaluated staining intensity and the\npercentage of positive cells in glandular epithelium and\nstromal areas. They calculated the H-score (range 0–300)\nfor each sample. Slides with a difference of >30 points\nwere jointly reviewed to reach a consensus value. The team\nassessed inter-rater consistency quarterly using duplicate\nsamples and reported the intraclass correlation (ICC) coeffi-\ncient (target ≥0.80). For samples with suspected uncertain\ncycle phase, a third pathologist reviewed the endometrial\nphase to ensure consistency with the blood sampling time\nwindow.\n2.5 Baseline Clinical and Imaging Information\nOn the same day as blood sampling, symptom assess-\nment was completed, including 10-point visual analog scale\nscores for dysmenorrhea, chronic pelvic pain, and dys-\npareunia [ 12]. Cyclical bowel or urinary symptoms were\nrecorded as binary variables using a structured question-\nnaire. A senior physician performed a bimanual pelvic\nexamination and recorded posterior fornix tenderness and\npalpable nodules. Serum CA-125 was measured using a\nchemiluminescent immunoassay platform, traceable to na-\ntional reference materials and subjected to daily QC. The\nimaging protocol followed a TVUS-first strategy performed\nby trained sonographers, with pelvic MRI supplemented\nwithin 90 days before surgery when necessary. Whether\nDIE was suggested and the total number of involved sites\nwere recorded in a unified manner. The imaging readers\nand the surgical team were mutually blinded.\n2.6 V ariable Definitions, Time-Window Alignment, and\nConfounding Control\nThe primary exposure variable was lnKTR. The pri-\nmary outcome was pathology-confirmed DIE analyzed as\na binary variable. Secondary outcomes included the EN-\nZIAN grade and the count of involved sites. Pre-specified\nconfounders included age, body mass index (BMI), smok-\ning status, menstrual cycle phase or exogenous hormone\nuse, high-sensitivity C-reactive protein, creatinine, and es-\ntimated glomerular filtration rate. Time-window alignment\nrequirements were as follows: symptom scales and physi-\ncal examination were completed on the same day as blood\nsampling; the interval between imaging and surgery did not\nexceed 90 days; the interval between blood sampling and\nsurgery did not exceed 14 days; and the eutopic endome-\ntrial sampling was performed concurrently with surgery.\nRecords exceeding the time windows were not included in\nthe primary analyses.\n3\n\n2.7 Statistical Analysis\n2.7.1 Chain A: Between-Group Differences and Gradient\nTrend of KTR With DIE\nThe analysis first compared differences in KTR\namong the three groups. Depending on the data distri-\nbution, analysis of variance or generalized linear models\nwere used with inclusion of pre-specified confounders, and\nadjusted mean differences and 95% confidence intervals\n(CIs) were reported. An ordinal logistic regression model\nbased on ENZIAN grade was constructed to test the gradi-\nent trend of KTR. Stratified analyses were performed to as-\nsess the consistency of association direction by menstrual\ncycle phase and exogenous hormone use. Two sensitiv-\nity analyses tested the robustness of the results: excluding\nthose with high-sensitivity C-reactive protein >10 mg·L−1\nand excluding those with an estimated glomerular filtra-\ntion rate <60 mL ·min−1·1.73 m −2 [13]. Effect sizes and\np-values were two-sided, with a significance threshold of\n0.05 (p < 0.05).\n2.7.2 Chain B: Evaluation of Independent Diagnostic Gain\nA pre-specified baseline prediction model was con-\nstructed, including symptom quantification, clinical signs,\nCA-125, and imaging binary indicator and site count. KTR\nwas then incorporated into this model, the AUCs were\ncompared using the DeLong method with optimism cor-\nrection by bootstrap, and the optimism-corrected AUC dif-\nference was reported. Category-based net reclassification\nimprovement and integrated discrimination improvement\nwere calculated according to clinically relevant risk thresh-\nolds (10%, 20%, 30%) [ 14]. Decision curve analysis was\nused to evaluate changes in net benefit within the above\nthreshold range. Calibration plots and Brier scores were\nprovided, and calibration improvements after adding KTR\nwere reported. For the imaging-suspicious but atypical sub-\ngroup, likelihood ratios and changes in pre- versus post-test\nprobabilities were calculated to display the impact of KTR\non clinical classification.\n2.7.3 Chain C: Multivariable Correlation Between KTR\nand MMP-9 Expression in Eutopic Endometrium\nWith MMP-9 H-score as the dependent variable and\nKTR as the main independent variable, a multiple linear\nregression model was built with inclusion of pre-specified\nconfounders. Collinearity and residual distributions were\nexamined, and H-score was log transformed or robust re-\ngression was used when necessary. Standardized regres-\nsion coefficients and 95% CIs were reported. Interaction\nterms were specified for different menstrual cycle phases\nand exogenous hormone use to evaluate effect modifica-\ntions. Measurement consistency was verified by the ICC\ncoefficient between the two raters, and measurement error\nsensitivity analyses of the model were incorporated.\n2.7.4 Internal V alidation and Sensitivity Analyses\nAll discrimination models underwent internal valida-\ntion with 1000 bootstrap resamples to obtain optimism-\ncorrected AUC, calibration slope, and intercept. Three\ntypes of sensitivity analyses were conducted for the primary\nresults: including only those sampled in the early follic-\nular phase; excluding those with elevated high-sensitivity\nC-reactive protein and with renal insufficiency; and strati-\nfication by imaging modality. Consistency of results was\njudged by the direction of effects and the magnitude of\nchange in effect sizes.\n2.7.5 Data Management and QC\nAn electronic case report form was established, with\ndual independent data entry along with range and logic\nchecks. After source data verification, outliers were cor-\nrected. Samples were tracked throughout with barcodes\nand three-level QC samples. Duplicate samples were ran-\ndomly interleaved within assay batches to monitor batch\neffects. Bidirectional blinding was implemented among\nthe clinical team, laboratory, and pathology, and unblind-\ning was performed only after data lock. When the propor-\ntion of missing data was <5%, complete-case analysis was\napplied; when >5%, multiple imputation was performed\nfor covariates and baseline predictors (10 imputed datasets,\nchained equations), and the exposure and outcome were\nnot imputed. Statistical analyses were performed using R\nsoftware (version 4.3 or later; R Foundation for Statistical\nComputing, Vienna, Austria). The main packages included\nstable versions for ROC comparison, reclassification, and\ndecision curve analysis. All code and analysis logs were\narchived for future reference after data lock.\n3. Results\n3.1 Participant Characteristics and Assay QC\nA total of 356 candidate participants were assessed,\nof whom 272 enrolled. Of these, 257 provided samples\nsuitable for pathological evaluation, including 94 DIE, 87\nnon-DIE endometriosis, and 76 non-endometriosis con-\ntrols. The study identified three distinct analytical cohorts\n(Fig. 1). Analysis population A comprised 252 partici-\npants with valid KTR (92/85/75). Analysis population B\nincluded 234 cases were used to determine diagnostic gain\n(85/79/70). Analysis population C included 213 cases for\nthe KTR–MMP-9 correlation (79/71/63) (Fig. 1). Kruskal–\nWallis and Pearson χ2 tests showed no statistically signif-\nicant differences among the three groups in baseline de-\nmographics, cycle/hormones, renal function, and smoking\nstatus (all p > 0.05). Levels of high-sensitivity C-reactive\nprotein (hsCRP) and CA-125 differed significantly among\nthe three groups (both p < 0.001). Similarly, the imag-\ning results for suspected DIE and the total count of sus-\npected involved sites showed significant differences (both\np < 0.001) (Table 1). Method validation and descriptive\n4\n\n\nFig. 1. Flowchart of participant screening, inclusion, and exclusion. DIE, deep infiltrating endometriosis; KTR, kynure-\nnine/tryptophan ratio; IHC, immunohistochemistry; LC-MS/MS, liquid chromatography-tandem mass spectrometry; CA-125, carbo-\nhydrate antigen 125.\nTable 1. Baseline characteristics of the study population (n, %)/M [IQR].\nV ariable DIE group (n = 94) Non-DIE endometriosis\ngroup (n = 87)\nNon-endometriosis\ncontrol group (n = 76)\nStatistic p-value\nDemographic and behavioral\nAge (years) 33.7 [28.9–38.6] 32.1 [27.6–37.7] 31.4 [26.5–36.2] H = 2.713 0.258\nBMI (kg·m−2) 22.8 [21.1–25.0] 22.6 [21.0–24.3] 22.7 [21.1–24.6] H = 0.534 0.766\nSmoking status (yes) 13 (13.83%) 10 (11.49%) 9 (11.84%) χ2 = 0.594 0.743\nSmoking status (no) 81 (86.17%) 77 (88.51%) 67 (88.16%) — —\nCycle and hormones\nMenstrual cycle phase: follicu-\nlar phase\n52 (55.32%) 46 (52.87%) 40 (52.63%) χ2 = 0.847 0.932\nMenstrual cycle phase: early se-\ncretory\n24 (25.53%) 22 (25.29%) 18 (23.68%) — —\nMenstrual cycle phase: late se-\ncretory\n18 (19.15%) 19 (21.84%) 18 (23.68%) — —\nExogenous hormone use (yes) 22 (23.40%) 18 (20.69%) 13 (17.11%) χ2 = 2.506 0.286\nExogenous hormone use (no) 72 (76.60%) 69 (79.31%) 63 (82.89%) — —\nInflammation and renal function\nhsCRP (mg·L−1) 1.68 [0.92–3.21] 1.22 [0.74–2.15] 0.98 [0.61–1.62] H = 16.427 <0.001\nCreatinine (µmol·L−1) 67 [60–73] 65 [59–72] 65 [58–71] H = 1.057 0.589\neGFR (mL·min−1ꞏ1.73 m−2) 105 [97–114] 106 [98–114] 107 [99–115] H = 1.833 0.400\nDiagnostic pathway variables\nCA-125 (U·mL−1) 29.5 [17.6–46.9] 18.7 [12.5–27.9] 13.0 [8.5–18.1] H = 35.761 <0.001\nImaging modality: TVUS only 51 (54.26%) 46 (52.87%) 49 (64.47%) χ2 = 6.537 0.162\nImaging modality: MRI only 12 (12.77%) 11 (12.64%) 7 (9.21%) — —\nImaging modality: both 31 (32.98%) 30 (34.48%) 20 (26.32%) — —\nImaging suggesting DIE (yes) 72 (76.60%) 27 (31.03%) 4 (5.26%) χ2 = 136.984 <0.001\nImaging suggesting DIE (no) 22 (23.40%) 60 (68.97%) 72 (94.74%) — —\nCount of suspected involved\nsites on imaging\n2 [1–3] 1 [0–1] 0 [0–0] H = 91.538 <0.001\nM, median; IQR, interquartile range; BMI, body mass index; hsCRP , high-sensitivity C-reactive protein; eGFR, estimated glomerular\nfiltration rate; TVUS, transvaginal ultrasound; MRI, magnetic resonance imaging.\nstatistics, and an ICC two-way random-effects consistency\nmodel were used. The LC-MS/MS calibration range cov-\nered 0.50–200.00 µmol/L, with a lower limit of quantifica-\ntion (LLOQ) of 0.50 and 20.00, Bias ranged from –2.27% to\n–1.83%, while within-run and between-run CVs remained\nbetween 3.29–6.18%. The QC pass rate was ≥97.22% (Ta-\nble 2A). The IHC scoring demonstrated good to excellent\nconsistency, with overall H-score ICC = 0.892 (95% CI\n5\n\nTable 2A. LC-MS/MS performance.\nParameter Kynurenine Tryptophan\nCalibration range (µmol/L) 0.50–20.00 20.00–200.00\nLLOQ (µmol·L−1) 0.5 20\nLOD (µmol·L−1) 0.14 6.37\nAccuracy (%Bias) –1.83 –2.27\nWithin-run CV (%) 3.29 4.06\nBetween-run CV (%) 4.91 6.18\nQC pass rate (%) 98.06 97.22\nHemolysis index range 0–31 0–31\nFreeze-thaw cycles (times) 0–1 0–1\nLLOQ, lower limit of quantification; LOD, limit of detection;\nCV , coefficient of variation.\nTable 2B. IHC scoring consistency (MMP-9 H-score).\nRegion Rater 1 mean H-score Rater 2 mean H-score ICC (95% CI)\nGlandular epithelium 148.32 ± 39.14 151.07 ± 38.22 0.887 (0.851–0.916)\nStromal area 134.26 ± 35.79 136.11 ± 34.88 0.874 (0.835–0.906)\nOverall H-score 141.59 ± 36.42 144.05 ± 35.76 0.892 (0.858–0.920)\nMMP-9, matrix metalloproteinase-9; ICC, intraclass correlation.\n0.858–0.920), and ICCs of 0.887 and 0.874 for glandular\nepithelium and stromal areas, respectively (Table 2B).\n3.2 Phenotypic Differences and Gradient of KTR With DIE\n(Chain A)\nIn Analysis population A, the raw distributions of\nlnKTR overlapped significantly across the three diagnos-\ntic groups, with the median and upper quartile tending to be\nhigher in the DIE group (Fig. 2). Using multivariable lin-\near regression (ANCOV A) and ordinal logistic regression,\nafter adjustment for age, BMI, smoking status, menstrual\ncycle phase/exogenous hormones, hsCRP and eGFR, the\nadjusted geometric means of KTR in the DIE and non-DIE\nendometriosis groups were significantly higher than in the\nnon-endometriosis control group ( β_diff = 0.28/0.15; both\np_adj < 0.05) (Table 3A). Moreover, each 1 standard de-\nviation (SD) increase in lnKTR was associated with an in-\ncrease in ENZIAN grade (OR_perSD = 1.62, p < 0.001)\n(Table 3B).\n3.3 Independent Diagnostic Gain of KTR (Chain B)\nLogistic regression modeling was employed for Anal-\nysis population B. AUC values were compared using the\nDeLong method, while the net reclassification improve-\nment (NRI), integrated discrimination improvement (IDI),\nand their 95% CIs were obtained by 1000 bootstrap resam-\nples. Calibration was assessed with the Brier score accom-\npanied by calibration plots. After adding KTR, AUC in-\ncreased from 0.79 to 0.83 ( ∆AUC = 0.04, 95% CI: 0.02–\n0.07, p = 0.003). At the pre-specified 20% primary thresh-\nold, NRI_category = 0.16 ( p = 0.004), and these results\nwere consistent at the 10% and 30% thresholds. The Brier\nscore decreased from 0.185 to 0.173 ( ∆–0.012), indicat-\ning improvements in discrimination and overall error (Ta-\nble 4). Decision curve analysis was performed, with 95%\nCIs obtained by 1000 bootstrap resamples. Within Pt =\n0.10–0.30, the “Baseline + KTR” curve lay above “Base-\nline” across the entire range, showing a sustained and stable\nmagnitude increase in net benefit (Fig. 3A). A logistic cal-\nibration model with locally estimated scatterplot smooth-\ning (LOESS) smoothing was used to display predicted–\nobserved agreement. The calibration intercept and slope of\nthe baseline model were α = –0.127 and β = 0.861, indicat-\ning slight underestimation; after adding KTR,α approached\n0 (–0.039) and β approached 1 (0.946), with calibration\nmarkedly improved (Fig. 3B). A risk-band reclassification\nmatrix, a signed-rank test, and the Wilson method were ap-\nplied (threshold 20%). In the imaging-suspicious but atyp-\nical subgroup, after adding KTR the net correct reclassi-\nfication was +14 for events and +17 for non-events (Ta-\nble 5A). Pre- versus post-test probability changes showed\nan increase for events and a decrease for non-events, both\np < 0.001 (Table 5B). Based on the 20% threshold, the\nlikelihood ratios were positive likelihood ratio (LR+) 3.14\n(95% CI 2.07–4.76) and LR– 0.30 (95% CI: 0.17–0.54) (Ta-\nble 5C).\n3.4 Independent Association Between KTR and MMP-9 in\nEutopic Endometrium (Chain C)\nAnalysis population C (n = 213) underwent multiple\nlinear regression. MMP-9 H-score was defined as depen-\ndent variable, standardized the lnKTR as z-score, log trans-\nformed the hsCRP levels, and applied a nested F test to eval-\nuate incremental explained variance. lnKTR was positively\nassociated with MMP-9 in eutopic endometrium ( β_std =\n0.29, 95% CI: 0.15–0.43, p < 0.001). Among covariates,\n6\n\n\nFig. 2. Raw distribution of plasma KTR across the three diagnostic groups. lnKTR, log transformed KTR.\nTable 3A. Adjusted differences in plasma KTR across the three diagnostic groups.\nGroup KTR (×10−3) (95% CI) β_diff (ln ratio) (95% CI) Wald z p_adj\nDIE 40.21 (37.88–42.69) 0.28 (0.16–0.41) 4.39 <0.001\nNon-DIE endometriosis 35.22 (33.01–37.59) 0.15 (0.05–0.26) 2.8 0.01\nNon-endometriosis controls (reference) 30.31 (28.19–32.52) — — —\nNote: Geometric means are back-transformations of model marginal means of lnKTR and are presented as ( ×10−3);\nβ_diff denotes the difference in log means relative to the non-endometriosis control group.\nTable 3B. ENZIAN grade trend analysis.\nPredictor OR_perSD (95% CI) Wald z p-value Brant test χ2 Brant test p-value\nlnKTR (z) 1.62 (1.35–1.94) 4.883 <0.001 4.183 0.523\nNote: OR_perSD denotes the odds ratio for a 1 SD increase in lnKTR corresponding to a higher\nENZIAN grade.\nonly ENZIAN total score (β_std = 0.18), hsCRP (ln) (β_std\n= 0.13), and late secretory relative to follicular phase (β_std\n= 0.11) were significant (all p < 0.05). Model yielded R 2\n= 0.32, after adding lnKTR, ∆R2 = 0.04 (F = 11.882, p <\n0.001) (Table 6). The scatter plot, colored by menstrual cy-\ncle phase, and the partial regression plot was controlled for\nall covariates. The overall partial regression line was con-\nsistent with the previously described direction, and the point\nclouds largely overlapped across all phases (Fig. 4).\n3.5 Stratified Analyses and Missing Data Sensitivity\nAnalyses\nANCOV A, the DeLong method with NRI_category\n(20% primary threshold), and multiple linear regression\nconsistent with the main analyses were used. In stratifi-\ncations of “early follicular only”, excluding hsCRP >10\nmg·L⁻1/eGFR <60 mL ·min⁻1·1.73 m⁻ 2, and by imaging\nmodality (TVUS only versus MRI only), β_diff in Chain A,\n∆AUC and NRI in Chain B, and β_std in Chain C were all\npositive and significant (all p < 0.05). The directions and\nmagnitudes of effects were consistent with the main analy-\nses, indicating robustness (Table 7).\n7\n\nFig. 3. Clinical benefit and calibration assessment of the models. (A) Decision curve analysis: net benefit after adding KTR. (B)\nCalibration plot: predicted vs. observed (Baseline vs. Baseline + KTR).\nTable 4. Comparison of diagnostic performance between the baseline model and “Baseline + KTR”.\nMetric Baseline (symptoms + signs + CA-125 + imaging) Baseline + KTR\nAUC (95% CI) 0.79 (0.74–0.84) 0.83 (0.79–0.87)\n∆AUC (95% CI) — 0.04 (0.02–0.07)\np_DeLong — 0.003\nBrier score 0.185 0.173\nΔBrier — –0.012\nNRI_category\n(10%/20%/30%, estimate\n[95% CI])\n— 0.18 (0.07–0.29)/0.16 (0.05–0.28)/0.14 (0.03–0.26)\np_NRI (10%/20%/30%) — 0.001/0.004/0.009\nIDI (95% CI) — 0.04 (0.02–0.07)\np_IDI — 0.001\nAUC (optimism-corrected) 0.78 0.82\nNote: Thresholds were defined as 10%, 20%, and 30%; all comparisons were based on the same Analysis population B and covariate set.\nAUC, area under the receiver operating characteristic curve; NRI, net reclassification improvement; IDI, integrated discrimination improve-\nment.\nTable 5A. Risk-band reclassification matrix.\nRisk category before adding KTR/Risk\ncategory after adding KTR\n<10% 10– <20% 20– <30% ≥30% Subtotal (events/non-events)\n<10% 1/15 3/5 1/2 0/0 5/22\n10–<20% 0/12 3/10 7/6 2/0 12/28\n20–<30% 0/2 2/8 6/7 5/2 13/19\n≥30% 0/2 0/7 2/1 8/2 10/12\nSubtotal (events/non-events) 1/31 8/30 16/16 15/4 40/81\nNet correct reclassification numbers: events +14 (up-classified 18, down-classified 4); non-events +17 (down-classified 32,\nup-classified 15). Note: Rows indicate risk categories before adding KTR, columns indicate risk categories after adding\nKTR, and each cell is presented as events/non-events.\nTable 5B. Change in predicted probability before and after adding KTR ( ∆p = after adding KTR – before adding KTR).\nGroup ∆p M[IQR] Z-value p-value\nEvents (n = 40) +0.07 [+0.03, +0.13] 4.086 <0.001\nNon-events (n = 81) –0.05 [–0.10, –0.02] –4.732 <0.001\n8\n\n\nTable 5C. Likelihood ratios based on the 20% threshold (after adding KTR).\nMetric Point estimate 95% CI (Wilson)\nLR+ 3.14 2.07–4.76\nLR– 0.30 0.17–0.54\nNote: Risk-band thresholds were the same as in Table 4 (<10%, 10–\n<20%, 20– <30%, ≥30%). Table 5C was based on post-KTR binary\nclassification (threshold 20%): sensitivity = 31/40, specificity = 61/81.\nLR, likelihood ratio.\nTable 6. Multivariable association between KTR and MMP-9 expression in eutopic endometrium.\nIndependent variable β_std (95% CI) SE t value p-value\nlnKTR (z) 0.29 (0.15–0.43) 0.07 4.143 <0.001\nAge 0.06 (–0.04–0.16) 0.05 1.200 0.232\nBMI –0.04 (–0.14–0.06) 0.05 –0.800 0.424\nSmoking (yes = 1) 0.05 (–0.07–0.17) 0.06 0.833 0.406\nPhase: early secretory (vs. follicular) 0.08 (–0.02–0.18) 0.05 1.600 0.111\nPhase: late secretory (vs. follicular) 0.11 (0.01–0.21) 0.05 2.200 0.029\nExogenous hormones (yes = 1) –0.07 (–0.17–0.03) 0.05 –1.400 0.162\nhsCRP (ln) 0.13 (0.03–0.23) 0.05 2.600 0.010\neGFR –0.03 (–0.13–0.07) 0.05 –0.600 0.547\nENZIAN total score 0.18 (0.06–0.30) 0.06 3.000 0.003\nNote: Model R 2 (with KTR): 0.32; ∆R2 (increment when adding lnKTR to the baseline\nmodel without KTR): 0.04; nested model F test (without KTR vs. with KTR): F = 11.882,\np < 0.001.\nFig. 4. Scatter plot of KTR (log) versus MMP-9 H-score and partial regression line.\n9\n\nTable 7. Summary of robustness in stratified and sensitivity analyses.\nAnalysis scenario Chain A: adjusted β_diff of lnKTR for\nDIE vs. controls (95% CI), p\nChain B: ΔAUC (95% CI), p_DeLong Chain B: NRI_category (95% CI), p Chain C: β_std (lnKTR→MMP-\n9) (95% CI), p\nEarly follicular only 0.30 (0.16–0.45), p = 0.001 0.05 (0.02–0.08), p = 0.002 0.19 (0.07–0.31), p = 0.002 0.27 (0.12–0.42), p = 0.001\nExcluding hsCRP >10 mg·L−1 and eGFR\n<60 mL·min−1·1.73 m−2\n0.27 (0.14–0.41), p = 0.001 0.04 (0.01–0.07), p = 0.010 0.17 (0.05–0.28), p = 0.006 0.28 (0.14–0.41), p < 0.001\nBy imaging modality: TVUS only 0.26 (0.11–0.40), p = 0.001 0.03 (0.003–0.06), p = 0.047 0.14 (0.02–0.26), p = 0.022 0.27 (0.11–0.41), p = 0.001\nBy imaging modality: MRI only 0.31 (0.16–0.47), p = 0.001 0.05 (0.02–0.09), p = 0.004 0.21 (0.06–0.35), p = 0.006 0.30 (0.13–0.46), p = 0.001\nTable 8. Variable missingness and imputation settings (Analysis population B, n = 234).\nV ariable Missing count (n) Missing rate (%) Imputation method (MICE)\nAge (years) 2 0.85 PMM\nBMI (kgꞏm−2) 3 1.28 PMM\nSmoking status (yes/no) 5 2.14 Binary logistic\nMenstrual cycle phase (follicular/early secretory/late secretory) 8 3.42 Multinomial logistic\nExogenous hormone use (yes/no) 4 1.71 Binary logistic\nDysmenorrhea V AS (0–10) 7 2.99 PMM\nDyspareunia V AS (0–10) 9 3.85 PMM\nPosterior fornix tenderness (yes/no) 6 2.56 Binary logistic\nhsCRP (mg·L−1, ln used for modeling) 12 5.13 PMM (on ln values)\neGFR (mL·min−1·1.73 m−2) 6 2.56 PMM\nCA-125 (U·mL−1) 10 4.27 PMM\nImaging suggesting DIE (binary) 1 0.43 Binary logistic\nCount of suspected involved sites on imaging 3 1.28 PMM\nNote: Little’s MCAR test: χ2 = 18.742, p = 0.282. Because the missing rate of hsCRP was >5%, the diagnostic model used multiple\nimputation (MICE, m = 10, chained equations; including all covariates and imaging indicators, with exposure KTR and outcome DIE not\nimputed but entered as predictors in the imputation equations).\nV AS, visual analog scale; PMM, predictive mean matching.\n10\n\n\nTable 9. Comparison of sensitivity between complete-case analysis and multiple imputation (Analysis Population B, n = 234).\nMetric Complete cases (CC, n = 206) Multiple imputation (MI pooled, n = 234; m = 10)\nlnKTR (per ↑1 SD) → DIE, OR (95% CI), p 1.57 (1.23–2.04), p = 0.001 1.62 (1.29–2.05), p = 0.001\nAUC (Baseline model) 0.79 (0.73–0.84) 0.79 (0.74–0.84)\nAUC (Baseline + KTR) 0.83 (0.78–0.87) 0.83 (0.79–0.87)\nΔAUC (DeLong 95% CI), p_DeLong 0.04 (0.01–0.07), p = 0.009 0.04 (0.02–0.07), p = 0.003\nBrier score (Baseline → Baseline + KTR) 0.186 → 0.174 0.185 → 0.173\nCalibration intercept α (Baseline → Baseline + KTR) –0.121 → –0.043 –0.127 → –0.039\nCalibration slope β (Baseline → Baseline + KTR) 0.865 → 0.944 0.861 → 0.946\nCC, complete-case analysis; MI, multiple imputation.\nLittle’s MCAR test and MICE multiple imputation (m\n= 10) was assessed. Within Analysis population B, the\nmissing data rates for each covariate remained <5%, with\nthe exception of hsCRP , and the missingness pattern was\nconsistent with the MCAR assumption ( χ2 = 18.742, p =\n0.282) (Table 8). Using logistic regression and the DeLong\ntest, the odds ratio (OR) for DIE per 1 SD increase in lnKTR\nwas similar between complete case and imputed analyses\n(1.57 vs. 1.62). Moreover, the AUC increased from 0.79 to\n0.83, and the directions of improvement in the Brier score\nand calibration were consistent, indicating that multiple im-\nputation did not alter the main conclusions (Table 9).\n4. Discussion\nAlthough the distributions of lnKTR overlapped\namong the three groups, participants with the deep infil-\ntrating phenotype showed higher median and upper quar-\ntile levels when surgical pathology, quantified by ENZIAN,\nserved as the gold standard. After simultaneous adjustment\nfor age, BMI, smoking, menstrual cycle phase, exogenous\nhormones, hsCRP , and eGFR, this difference remained.\nThe non-DIE endometriosis group also showed higher\nlnKTR levels than the non-endometriosis controls, and each\nstandard deviation increase in lnKTR corresponded to a\nhigher ENZIAN grade, suggesting a clear dose–response\nrelationship. Stratified and sensitivity analyses showed no\nreversal of effect direction, and the linearity, precision, and\nQC pass rate of LC-MS/MS assay, as well as the consis-\ntency of IHC scoring, provided technical credibility for the\ninference. From an immunometabolic perspective, inter-\nferon signaling in the inflammatory microenvironment ac-\ncelerates tryptophan metabolism through indoleamine 2,3-\ndioxygenase and tryptophan 2,3-dioxygenase pathways,\nleading to tryptophan depletion and kynurenine accumula-\ntion, with KTR rising as a systemic readout of pathway ac-\ntivation [15,16]. As a ligand of AhR, kynurenine can syn-\nergize with proinflammatory transcriptional networks, up-\nregulate MMP expression, and enhance matrix degradation\nand fibrotic remodeling, conferring stronger adhesion and\ninvasive capacity to lesions [17]. Accordingly, systemic el-\nevation of KTR is consistent with the gradient of deeper tis-\nsue infiltration and broader involvement. Previous reports\nhave mostly observed abnormalities of the tryptophan path-\nway in mixed endometriosis populations [18], but evidence\nfor discrimination of the deep infiltrating subtype and for\nquantification of anatomical burden has been insufficient,\nand control of menstrual cycle and hormonal exposure has\noften been inadequate. In contrast, the present prospective\ndata, obtained under rigorous preanalytical procedures and\nmultivariable adjustment, demonstrate both phenotypic dif-\nferences and grading trends, indicating that KTR relates not\nonly to the presence of disease but also to the intensity of the\npathological phenotype, providing a testable framework for\nperioperative stratification and for subsequent mechanistic\nstudies targeting immunometabolism.\nAfter adding KTR to the baseline model composed\nof symptoms, signs, CA-125, and imaging, discrimination\nimproved, prediction error decreased, probability calibra-\ntion approached the ideal, and sustained net benefit ap-\npeared within the clinically relevant risk range of 0.10–\n0.30; ∆AUC = 0.04. Although this represents a numer-\nically moderate increment that commonly appears with\na strong baseline model, under the premise that symp-\ntoms, signs, and imaging information already integrate into\nthe model, this level of improvement usually corresponds\nto more substantive risk re-ranking among intermediate-\nrisk patients, especially in preoperative stratification sce-\nnarios around the 10%–30% decision thresholds. In the\nimaging-suspicious but atypical population, addition of\nKTR achieved net correct reclassification for both events\nand non-events, shifted individual post-test probabilities\nin the appropriate direction, and improved likelihood ra-\ntios at the 20% threshold. These findings suggest reduc-\ntions in unnecessary interventions and missed diagnoses in\npreoperative stratification and management pathways. All\nthese improvements remained robust after internal boot-\nstrap correction. The immunometabolic dimension repre-\nsented by KTR provides a biological basis for the above\nphenomena. The conversion of tryptophan to kynurenine\nthrough indoleamine 2,3-dioxygenase and tryptophan 2,3-\ndioxygenase pathways is regulated by inflammatory sig-\nnals, and an elevated KTR reflects systemic pathway ac-\ntivation [19]. Its information content does not overlap with\nthe structural features captured by imaging or the humoral\ncharacteristics of CA-125. When incorporated into the\nmodel, it can correct risk ranking and probability calibra-\n11\n\ntion, allowing individuals at intermediate risk near deci-\nsion thresholds to be adjusted more accurately upward or\ndownward, thereby translating into net benefit within clin-\nically relevant trade-off ranges [ 20]. High-quality preana-\nlytical control and laboratory validation provided method-\nological assurance for this gain. A pervious serologic study\nhas largely remained at the level of ROC curves for single\nindicators, have seldom reported calibration metrics and de-\ncision curves concurrently, and have lacked evidence of re-\nclassification and likelihood ratios in imaging-atypical sce-\nnarios [ 21]. The present results, evaluated at prespecified\nthresholds, show concordant improvements in discrimina-\ntion, calibration, and decision performance, with clinical\ninterpretability presented by the reclassification matrix and\nlikelihood ratios. These findings fill key gaps in the prior\nevidence chain, indicating that KTR, as an effective supple-\nment to the traditional pathway, has clear clinical translata-\nbility.\nWith surgical pathology as the reference and after ad-\njustment for age, BMI, smoking, menstrual cycle phase, ex-\nogenous hormones, high-sensitivity C-reactive protein, es-\ntimated glomerular filtration rate, and lesion burden, lnKTR\nshowed an independent positive association with MMP-9\nH-score in eutopic endometrium. Partial regression plots\nshowed a stable slope, with point clouds across cycle strata\nthat largely overlapped and stratified, as well as sensitiv-\nity analyses pointing in the same direction, suggesting that\nthis association was not driven by cycle differences or co-\nvariates such as inflammation or renal function. At the\nlevel of immunometabolism, inflammatory signals induce\nactivation of indoleamine 2,3-dioxygenase and tryptophan\n2,3-dioxygenase, increasing the flux of tryptophan toward\nkynurenine [22]. Kynurenine, as a ligand of AhR, can acti-\nvate downstream transcriptional networks, enhance MMP-\n9 transcription and secretion, and promote changes in cell\nadhesion, extracellular matrix degradation, and stromal re-\nmodeling, conferring stronger invasive and fibrotic capac-\nity to lesions [ 23]. The concordant changes in systemic\nKTR elevation and local MMP-9 enhancement can be in-\nterpreted as a coupling marker between immunometabolic\nsignaling and tissue microenvironmental activity [ 24], sup-\nporting the discriminative value of the plasma biomarker\nfor the infiltrative phenotype and providing a measurable\nbridging index for perioperative risk stratification and po-\ntential targeted research. Previous multi-source disease\nmodels have suggested that kynurenine and AhR upregu-\nlate MMP-9, but population-level evidence in human eu-\ntopic endometrium remains scarce [ 25]. Paired blood–\ntissue data, combined with consistent robustness checks, fill\nthis key gap by enabling the systemic marker and histologic\nprocesses to be compared and verified within the same sub-\njects.\nLimitations\nTo avoid overstating the conclusions, several limita-\ntions should be acknowledged. The single-center prospec-\ntive cohort may be affected by referral and spectrum bias,\nlacks external and temporal validation, and the stability\nof risk thresholds has not yet been prospectively tested.\nNRI, IDI, and decision curves are sensitive to sample size\nand threshold selection, and their CIs carry inherent un-\ncertainty. KTR is influenced by concomitant inflamma-\ntion, diet, and metabolic status. Although adjustments\nwere made for hsCRP and renal function, residual con-\nfounding may remain. MMP-9 was semi-quantified by\nIHC, and despite good ICC, batch effects and reader drift\nmay occur. The maximum 90-day interval from imag-\ning to surgery may lead to disease progression. Future\nwork should conduct multicenter, pre-registered external\nand temporal validation; establish cross-platform standard-\nization of LC-MS/MS and IHC with reference materials;\nprospectively confirm thresholds according to clinical use\nscenarios; evaluate decision impact and cost-effectiveness\nin imaging-atypical and preoperative stratification path-\nways; combine plasma markers, symptoms, and structural\nimaging to develop deployable multimodal models; refine\npreanalytical standard operating procedures (SOPs) and\ninter-laboratory consistency; and, within the cohort, track\nlongitudinal changes and modifiability of the kynurenine–\nAhR–MMP-9 pathway to clarify causal associations with\nclinical outcomes.\n5. Conclusions\nThis study demonstrates that plasma KTR provides in-\ndependent and quantifiable diagnostic value beyond symp-\ntoms, signs, CA-125, and routine imaging. It improves dis-\ncrimination, calibration, and clinical net benefit, and is par-\nticularly useful for preoperative stratification in imaging-\nsuspicious but atypical cases. KTR shows an indepen-\ndent positive association with MMP-9 expression in eu-\ntopic endometrium, suggesting an association between sys-\ntemic immunometabolic activation and local matrix remod-\neling with clear biological plausibility. Overall, KTR can\nserve as a practical auxiliary biomarker for DIE to optimize\nreferral and surgical planning and provides a foundation\nfor kynurenine pathway–related precision interventions and\nmultimodal diagnostic models.\nAvailability of Data and Materials\nThe datasets used and analyzed during the current\nstudy are available from the corresponding author on rea-\nsonable request, subject to institutional data sharing policies\nand ethical approval.\nAuthor Contributions\nHX: Conceived and designed the study, participated\nin patient enrollment and data collection, performed pre-\n12\n\n\nliminary data analysis, and drafted the first version of\nthe manuscript. HJ: Assisted in study design, performed\nstatistical analysis and data verification, prepared figures\nand tables, and contributed to revising and polishing the\nmanuscript. CM: Conceived and supervised the overall\nstudy, coordinated patient recruitment and clinical data in-\nterpretation, provided critical revision of the manuscript\nfor important intellectual content, and serves as the corre-\nsponding author. All authors read and approved the final\nmanuscript. All authors have participated sufficiently in\nthe work and agreed to be accountable for all aspects of the\nwork.\nEthics Approval and Consent to Participate\nThis study was conducted in accordance with the Dec-\nlaration of Helsinki. The protocol was reviewed and ap-\nproved by the Ethics Committee of Zhejiang Provincial\nPeople’s Hospital, Hangzhou, Zhejiang, China (approval\nNo. KT2022062). Written informed consent was obtained\nfrom all participants before enrolment in the study.\nAcknowledgment\nThe authors sincerely thank the clinicians, sonogra-\nphers, and nursing staff of the Department of Obstetrics,\nY ongkang Hospital of Traditional Chinese Medicine, for\ntheir support in participant recruitment, perioperative man-\nagement, and imaging examinations. We are also grateful to\nthe pathology and laboratory teams for their assistance with\nspecimen processing, LC-MS/MS assays, and immunohis-\ntochemical evaluations, as well as to all women who gen-\nerously participated in this study.\nFunding\nThis research received no external funding.\nConflicts of Interest\nThe authors declare no conflicts of interest.\nDeclaration of AI and AI-Assisted\nTechnologies in the Writing Process\nDuring the preparation of this manuscript, the authors\nused ChatGPT to assist with language editing (including\ngrammar, wording, and style). The authors carefully re-\nviewed and revised all AI-assisted suggestions and take full\nresponsibility. No AI tools were used for data analysis, data\ngeneration, or scientific interpretation.\nSupplementary Material\nSupplementary material associated with this article\ncan be found, in the online version, at https://doi.org/10.\n31083/CEOG47790.\nReferences\n[1] Allaire C, Bedaiwy MA, Y ong PJ. Diagnosis and management\nof endometriosis. CMAJ: Canadian Medical Association Journal\n= Journal De L’Association Medicale Canadienne. 2023; 195:\nE363–E371. https://doi.org/10.1503/cmaj.220637.\n[2] Grube M, Castan M, Drechsel-Grau A, Praetorius T, Greif K,\nStaebler A, et al . Diagnostics and Surgical Treatment of Deep\nEndometriosis-Real-World Data from a Large Endometriosis\nCenter. Journal of Clinical Medicine. 2024; 13: 6783. https:\n//doi.org/10.3390/jcm13226783.\n[3] Becker CM, Bokor A, Heikinheimo O, Horne A, Jansen F, Kiesel\nL, et al. ESHRE guideline: endometriosis. Human Reproduction\nOpen. 2022; 2022: hoac009. https://doi.org/10.1093/hropen/hoa\nc009.\n[4] Ke J, Y e J, Li M, Zhu Z. The Role of Matrix Metalloproteinases\nin Endometriosis: A Potential Target. Biomolecules. 2021; 11:\n1739. https://doi.org/10.3390/biom11111739.\n[5] Y ang HL, Li, MQ. Indoleamine 2,3-dioxygenase in endometrio-\nsis. Reproductive and Developmental Medicine. 2019; 3: 110–\n116. https://doi.org/10.4103/2096-2924.262391.\n[6] Gáspár R, Halmi D, Demján V , Berkecz R, Pipicz M, Csont T.\nKynurenine Pathway Metabolites as Potential Clinical Biomark-\ners in Coronary Artery Disease. Frontiers in Immunology. 2022;\n12: 768560. https://doi.org/10.3389/fimmu.2021.768560.\n[7] Gibbons T, Rahmioglu N, Zondervan KT, Becker CM. Crimson\nclues: advancing endometriosis detection and management with\nnovel blood biomarkers. Fertility and Sterility. 2024; 121: 145–\n163. https://doi.org/10.1016/j.fertnstert.2023.12.018.\n[8] Vickers AJ, van Calster B, Steyerberg EW. A simple, step-by-\nstep guide to interpreting decision curve analysis. Diagnostic\nand Prognostic Research. 2019; 3: 18. https://doi.org/10.1186/\ns41512-019-0064-7 .\n[9] Brulport A, Bourdon M, V aiman D, Drouet C, Pocate-Cheriet\nK, Bouzid K, et al . An integrated multi-tissue approach for\nendometriosis candidate biomarkers: a systematic review. Re-\nproductive Biology and Endocrinology: RB&E. 2024; 22: 21.\nhttps://doi.org/10.1186/s12958-023-01181-8 .\n[10] Keckstein J, Saridogan E, Ulrich UA, Sillem M, Oppelt P ,\nSchweppe KW, et al . The #Enzian classification: A compre-\nhensive non-invasive and surgical description system for en-\ndometriosis. Acta Obstetricia et Gynecologica Scandinavica.\n2021; 100: 1165–1175. https://doi.org/10.1111/aogs.14099.\n[11] International Council for Harmonisation (ICH). M10 Bioanalyt-\nical Method V alidation and Study Sample Analysis. 2022. Avail-\nable at: https://database.ich.org/sites/default/files/M10_Guidel\nine_Step4_2022_0524.pdf (Accessed: 2 January 2026).\n[12] Cozzolino M, Coccia ME, Lazzeri G, Basile F, Troiano G. V ari-\nables Associated with Endometriosis-related Pain: A Pilot Study\nusing a Visual Analogue Scale. Revista Brasileira De Ginecolo-\ngia E Obstetricia: Revista Da Federacao Brasileira Das So-\nciedades De Ginecologia E Obstetricia. 2019; 41: 170–175.\nhttps://doi.org/10.1055/s-0039-1679879 .\n[13] Kidney Disease: Improving Global Outcomes (KDIGO) CKD\nWork Group. KDIGO 2024 Clinical Practice Guideline for the\nEvaluation and Management of Chronic Kidney Disease. Kid-\nney International. 2024; 105: S117–S314. https://doi.org/10.\n1016/j.kint.2023.10.018.\n[14] Wang Z, Cheng Y , Seaberg EC, Becker JT. Quantifying diagnos-\ntic accuracy improvement of new biomarkers for competing risk\noutcomes. Biostatistics (Oxford, England). 2020; 23: kxaa048.\nhttps://doi.org/10.1093/biostatistics/kxaa048.\n[15] Seo SK, Kwon B. Immune regulation through tryptophan\nmetabolism. Experimental & Molecular Medicine. 2023; 55:\n1371–1379. https://doi.org/10.1038/s12276-023-01028-7 .\n[16] Midttun Ø, Ulvik A, Meyer K, Zahed H, Giles GG, Manjer J,\n13\n\net al. A cross-sectional study of inflammatory markers as deter-\nminants of circulating kynurenines in the Lung Cancer Cohort\nConsortium. Scientific Reports. 2023; 13: 1011. https://doi.org/\n10.1038/s41598-023-28135-9 .\n[17] Griffith BD, Frankel TL. The Aryl Hydrocarbon Receptor: Im-\npact on the Tumor Immune Microenvironment and Modulation\nas a Potential Therapy. Cancers. 2024; 16: 472. https://doi.org/\n10.3390/cancers16030472.\n[18] Murgia F, Angioni S, D’Alterio MN, Pirarba S, Noto A, San-\ntoru ML, et al . Metabolic Profile of Patients with Severe En-\ndometriosis: a Prospective Experimental Study. Reproductive\nSciences (Thousand Oaks, Calif.). 2021; 28: 728–735. https:\n//doi.org/10.1007/s43032-020-00370-9 .\n[19] Sorgdrager FJH, Naudé PJW, Kema IP , Nollen EA, Deyn PPD.\nTryptophan Metabolism in Inflammaging: From Biomarker to\nTherapeutic Target. Frontiers in Immunology. 2019; 10: 2565.\nhttps://doi.org/10.3389/fimmu.2019.02565.\n[20] Piovani D, Sokou R, Tsantes AG, Vitello AS, Bonovas S. Op-\ntimizing Clinical Decision Making with Decision Curve Anal-\nysis: Insights for Clinical Investigators. Healthcare (Basel,\nSwitzerland). 2023; 11: 2244. https://doi.org/10.3390/healthca\nre11162244.\n[21] Dolińska W, Draper H, Othman L, Thompson C, Girvan S, Cun-\nningham K, et al . Accuracy and Utility of Blood and Urine\nBiomarkers for the Noninvasive Diagnosis of Endometriosis: A\nSystematic Literature Review and Meta-analysis. F&S Reviews.\n2023; 4: 116–130. https://doi.org/10.1016/j.xfnr.2022.12.001.\n[22] Zhou Q, Shi Y , Chen C, Wu F, Chen Z. A narrative review of\nthe roles of indoleamine 2,3-dioxygenase and tryptophan-2,3-\ndioxygenase in liver diseases. Annals of Translational Medicine.\n2021; 9: 174. https://doi.org/10.21037/atm-20-3594 .\n[23] Wang Z, Zhang Y , Liao Z, Huang M, Shui X. The potential of\naryl hydrocarbon receptor as receptors for metabolic changes in\ntumors. Frontiers in Oncology. 2024; 14: 1328606. https://doi.\norg/10.3389/fonc.2024.1328606.\n[24] Wang Y , Jiao L, Qiang C, Chen C, Shen Z, Ding F, et al .\nThe role of matrix metalloproteinase 9 in fibrosis diseases and\nits molecular mechanisms. Biomedicine & Pharmacotherapy =\nBiomedecine & Pharmacotherapie. 2024; 171: 116116. https:\n//doi.org/10.1016/j.biopha.2023.116116.\n[25] Huang Q, Song Y , Lei X, Huang H, Nong W. MMP-9 as a clinical\nmarker for endometriosis: a meta-analysis and bioinformatics\nanalysis. Frontiers in Endocrinology. 2024; 15: 1475531. https:\n//doi.org/10.3389/fendo.2024.1475531.\n14","source_license":"CC0","license_restricted":false}