This is a prospective, single-center diagnostic study testing whether a new, minimally invasive analysis of nipple fluid can distinguish benign from malignant causes of pathologic nipple discharge. Many patients with bloody or blood-tinged nipple discharge undergo surgery to make a diagnosis, yet most are ultimately found to have benign disease. The investigators aim to develop a laboratory test that analyzes DNA in nipple fluid to help avoid unnecessary operations while still identifying cancers. Approximately 30 adults with spontaneous, single-duct, unilateral bloody or serosanguinous nipple discharge who are already scheduled for standard diagnostic surgery will be enrolled at Hubei Cancer Hospital. Before surgery, the investigators will collect a small sample of nipple fluid (or gently obtain nipple aspirate fluid using a soft suction cup if needed) and one tube of blood. The investigators will analyze the fluid's DNA using two approaches: low-pass whole-genome analysis to look for copy number changes and fragmentation patterns, and genome-wide DNA methylation profiling. Surgical pathology will serve as the reference standard. Using these data, the investigators will build and validate a model to classify lesions as benign or malignant. The primary outcome is diagnostic accuracy (area under the ROC curve, sensitivity, and specificity). Secondary outcomes include positive and negative predictive values, model calibration, subgroup performance (e.g., ductal carcinoma in situ vs invasive cancer), and an estimate of potential clinical impact (for example, how many benign cases might safely avoid surgery at a high-sensitivity threshold). Study test results will not affect current clinical care; all participants will receive usual evaluation and surgery. Risks are minimal and may include brief nipple discomfort or skin irritation from gentle suction and routine blood-draw risks (bruising, lightheadedness). There is no direct benefit to participants, but the findings may support a future noninvasive test to guide care and reduce unnecessary surgery. Data will be de-identified and stored securely. Expected enrollment is from September 2025 to May 2026.
Study Type
OBSERVATIONAL
Enrollment
60
Wu Xinhong
Wuhan, Hubei, China
AUC of the combined lcWGS-methylome model for distinguishing malignant vs benign lesions
Area under the ROC curve (AUC) with 95% CI in the independent validation cohort. Reference standard is postoperative pathology. Model integrates preoperative nipple discharge features (CNV/fragmentomics from low-coverage WGS and genome-wide DNA methylation) with a locked threshold selected on training.
Time frame: From preoperative sampling to receipt of final surgical pathology (per participant); primary analysis at validation lock (≈0-3 months post-enrollment).
Sensitivity and specificity at a prespecified threshold (target sensitivity ≥95%)
Sensitivity (true positive rate) and specificity (true negative rate), with 95% confidence intervals, estimated in the independent validation cohort at a threshold fixed on the training set to target ≥95% sensitivity. Surgical pathology is the reference standard.
Time frame: Perioperative/Periprocedural: preoperative sampling (Day 0); outcome assessed at final surgical pathology report, up to 30 days postoperatively.
Positive and negative predictive values (PPV, NPV)
PPV and NPV, with 95% confidence intervals, calculated in the independent validation cohort at the prespecified decision threshold, using surgical pathology as the reference standard.
Time frame: preoperative sampling (Day 0); outcome assessed at final surgical pathology, up to 30 days after surgery.
Model calibration (calibration slope, intercept, and Hosmer-Lemeshow test)
Agreement between predicted probabilities and observed outcomes in the independent validation cohort; report calibration slope and intercept (with 95% confidence intervals) from logistic calibration, and Hosmer-Lemeshow goodness-of-fit p-value (by deciles). Reference standard: final surgical pathology.
Time frame: Perioperative/Periprocedural: preoperative sampling (Day 0); outcome assessed at final surgical pathology report, up to 30 days postoperatively.
Proportion of model-negative results in benign cases at a prespecified decision threshold (independent validation cohort)
Among participants with benign final surgical pathology, calculate the proportion with a model-negative result when applying a decision threshold prespecified on the training set to achieve \>=95% sensitivity for malignancy. Report the percentage and 95% confidence interval. Reference standard: final surgical pathology.
Time frame: Preoperative sampling (Day 0); outcome assessed at the time of the final surgical pathology report, up to 30 days postoperatively.
Subgroup performance by pathology subtype and imaging status
Area under the ROC curve (AUC), sensitivity, and specificity (with 95% confidence intervals) in the independent validation cohort, stratified by pathology subtype (DCIS vs IDC) and by presence vs absence of preoperative imaging abnormalities (mammography/ultrasound/MRI). Sensitivity and specificity estimated at a prespecified threshold fixed on the training set to target ≥95% sensitivity. Reference standard: final surgical pathology.
Time frame: Perioperative/Periprocedural: preoperative sampling (Day 0); outcome assessed at final surgical pathology report, up to 30 days postoperatively.
Technical success rate and sample adequacy
Proportion of nipple discharge samples yielding sequencing libraries that pass prespecified QC thresholds and are analyzable; report percentage and 95% confidence interval. Summarize cfDNA yield (ng) and sequencing depth metrics (aligned reads and mean unique coverage) with medians and interquartile ranges.
Time frame: Baseline (Day 0): sample collection; laboratory QC and adequacy assessed up to 14 days post-collection.
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.