Endometrial cancer is the most common gynecological cancer in high-income countries, affecting over 400,000 women worldwide each year. When caught early, it is highly treatable, but a meaningful minority of women present with aggressive or advanced disease and face poor outcomes.
The main symptom that triggers investigation is post-menopausal bleeding (PMB). However, current diagnostic tools are invasive and imperfect. Transvaginal ultrasound has low specificity, referring over half of women for further testing unnecessarily. Hysteroscopy and endometrial biopsy are accurate but painful, anxiety-provoking, and prone to technical failure.
Only 5-10% of women presenting with PMB actually have cancer - yet millions undergo invasive tests annually at great personal and financial cost. An accurate, non-invasive triage test that could identify the women who truly need invasive follow-up would be a major advance.
Urine is an ideal candidate biofluid for such a test. It is cheap, easy to collect, non-invasive, and acceptable to both patients and clinicians. Cancer biomarkers can reach urine either via kidney filtration of blood proteins or by contamination from the nearby uterus through the urethra.
This prospective study enrolled 104 post-menopausal women with abnormal uterine bleeding: 50 with confirmed endometrial cancer and 54 controls without cancer. Urine samples were self-collected before any clinical examination or treatment, ensuring unbiased collection.
SWATH-MS (Sequential Window Acquisition of All Theoretical Mass Spectra) is an advanced mass spectrometry technique that can accurately identify and quantify hundreds of proteins in a biological sample simultaneously. It has high reproducibility and the ability to re-interrogate data, making it well-suited for biomarker discovery.
The researchers used a novel two-pronged search strategy. Urine proteomic data were searched against both a human plasma library (to find kidney-excreted systemic cancer proteins) and a bespoke cervico-vaginal fluid library (to find uterine-derived proteins that contaminate urine via anatomical proximity). This dual approach is more likely to yield cancer-specific markers than a standard urine library.
Machine learning using Random Forest algorithms was then applied to identify which proteins best distinguished cancer from control samples. Top discriminatory proteins were combined into multi-marker diagnostic panels using logistic regression, with performance assessed by area under the receiver operating characteristic curve (AUC) - a measure of how well a test separates cases from controls.
From the plasma library search, 798 urinary proteins were quantified. Forty-nine showed significantly altered levels between cancer and control women, with 39 being statistically significant. The top single discriminators included Cystatin A (CSTA), Calcium Binding Protein A7 (S100A7), and Fatty acid-binding protein 5 (FABP5), each with AUC values around 0.73-0.77 - indicating moderate individual accuracy.
From the cervico-vaginal fluid library search, 316 proteins were quantified. The most significant uterine-derived discriminator was Thioredoxin (TXN), followed by SPRR1B and CRNN (Cornulin). Notably, proteins like Complement C9 were fourfold higher in cancer cases compared to controls, while SerpinB3 was almost threefold elevated.
Critically, individual proteins only performed moderately, but combining them dramatically improved accuracy. The best systemic biomarker panel (10 proteins including CSTA, S100A7, MMP9, and others) achieved an AUC of 0.91 - meaning it correctly classified over 91% of cases. The best uterine-derived panel (10 proteins including SPRR1B, CRNN, CALML3, TXN, FABP5, and others) achieved an AUC of 0.92.
These panels maintained strong performance specifically for early-stage disease (FIGO stage I/II), achieving AUCs of 0.90 and 0.92 respectively - a crucial finding since early detection offers the best chance of cure.
The top-performing diagnostic model combined SPRR1B, CRNN, CALML3, TXN, FABP5, C1RL, MMP9, ECM1, S100A7, and CFI - ten proteins drawing from both the systemic and uterine-derived categories. This panel predicted endometrial cancer with an AUC of 0.92, sensitivity of 83.7%, and specificity of 83.9%.
In practical terms, with sensitivity at 83.7% and specificity at 83.9%, the panel would correctly flag approximately 84 out of 100 women who have cancer and correctly reassure approximately 84 out of 100 women who do not. The positive predictive value (chance that a positive result means cancer) was 85.5%, and the negative predictive value (chance that a negative result means no cancer) was 82.0%.
The researchers intentionally capped the panel at 10 proteins to ensure clinical simplicity. More complex models with more proteins might achieve marginally higher accuracy but would be harder to translate into routine clinical tests like ELISA or lateral flow strips.
Several of the top biomarkers have mechanistic links to endometrial cancer. CRNN (Cornulin) is pro-proliferative and regulates cell cycle progression by inducing CCND1. It also activates NFkB and PI3K/AKT signaling pathways, both known to drive endometrial cancer development. CRNN was found at lower levels in cancer urine, suggesting it may act as a tumor suppressor in this context.
Cystatin A (CSTA) belongs to a family of proteins known as cysteine protease inhibitors. In other cancers like esophageal and lung cancer, CSTA acts as a tumor suppressor by inhibiting MAPK and AKT pathways and promoting epithelial stability. Its lower levels in endometrial cancer urine are consistent with loss of this suppressive function.
MMP9 (Matrix Metalloproteinase 9) degrades proteins in the tissue scaffolding and plays a role in tumor invasion. It has previously been found to be differentially expressed in hyperplastic versus normal endometrium. FABP5 modulates gene expression and cell signaling, and has been reported as a potential biomarker in endometrial tissue and plasma studies.
The fact that multiple proteins from different biological pathways contribute to the panel suggests that urine captures a rich signature of the tumor's impact on the body, making multi-marker panels far superior to single-protein tests.
Major strengths of this study include the use of SWATH-MS, which offers superior reproducibility compared to older proteomic platforms, and the clinically realistic control group - women with PMB who do not have cancer, the exact population in which the test would be used. All cancer diagnoses were confirmed by gold-standard histopathology.
Limitations include a relatively small sample size that prevented analysis of specific cancer subtypes and very advanced stages. It is also unknown how the biomarkers perform in pre-menopausal women or as a screening tool in symptom-free women. Urine biomarker levels can be confounded by hydration status, medications, renal function, and diet.
The current technology of SWATH-MS is not practical for routine clinical labs. The authors note that future translation will require development of simpler assay formats - such as ELISA (a common laboratory test) or lateral flow strips (similar to home pregnancy tests) - with validated reproducibility.
The James Lind Alliance - a UK organization that surveys patients and healthcare professionals about research priorities - identified non-invasive cancer detection tests as the single most important research priority. This study directly responds to that priority.
This study provides proof of principle that a urine-based proteomic test can detect endometrial cancer with clinically meaningful accuracy. The combination of both systemically excreted and uterine-derived proteins in a 10-marker panel offers AUC above 0.90 - a threshold generally considered sufficient to warrant further clinical development.
A community-based urine self-collection test could transform how symptomatic women are triaged. Women who test negative could potentially be safely monitored rather than immediately referred for invasive procedures, dramatically reducing unnecessary tests and their associated anxiety, pain, and cost.
The authors call for large multi-centre prospective studies to validate these findings in independent cohorts and to fully understand the biological roles of the identified proteins in endometrial cancer development. Such validation is a standard and necessary step before any biomarker can enter clinical use.