Background
The theoretical utility of duct lavage to detect breast cancer or breast cancer precursors is based on the premises that intraductal epithelial lesions are the precursors of invasive breast cancer and that sampling cells from the mammary ducts can detect these precursor lesions. However, for duct lavage to be able to detect these lesions: 1. lesions signifying high risk for breast cancer must be accessible from the nipple and 2. these lesions are found in the 'fluid yielding'ducts.

Prediction
A surrogate marker for lesions being accessible from the nipple is the relative rate of intraductal nipple and central duct disease (NCD) found in breast cancer cases. We retrospectively examined 801 mastectomy specimens removed for breast carcinoma. 17% did not have intraductal disease identified. NCD was demonstrated in only 179 mastectomies (22%). NCD was seen in 22% (115/522) of infiltrating carcinomas (IC) with duct carcinoma in situ (DCIS) and in 32% (7/22) IC with lobular intraepithelial neoplasia (LIN). NCD disease was present in 43% of infiltrating lobular carcinomas and 33% of infiltrating mixed ductal-lobular carcinomas, but in only 16% of infiltrating duct carcinomas. According to this data, examination of the ductal system by lavage or other methods would potentially have a sensitivity of 20-25%.

Retrospective re-examination of mastectomy specimens specifically for intraductal disease also found only a minority of mastectomy specimens had ductal disease extending centrally. This method did not identify any additional cases overlooked by routine gross examination.

Testing
To discover what portions of the breast are affected by duct lavage, we studied a group of mastectomy patients who agreed to undergo ductal lavage while under anesthesia. If their lavage specimens had adequate cellularity, their mastectomy specimens would be injected with dye in the lavaged ductal tree and grossly sub-sectioned. Mastectomy specimens from 24 of 40 patients who underwent lavage met criteria for study. The lavage specimens were independently assessed by two cytopathologists. Four lavage samples had marked atypia or malignant cells in 20 cases of carcinoma, a sensitivity that is close to that predicted by our mastectomy study. Positive lavage findings coincided with extensive intraductal carcinoma. For many cases, the geographic region of lavage did not coincide with either IC or DCIS, suggesting that disease containing and fluid yielding ducts are not the same.

Summary
For any new diagnostic procedure, a prediction of sensitivity can be made by analyzing existing data. Sensitivity of diagnostic test should be tested against known disease. Studies should be constructed carefully to both address hypotheses as well as hidden assumptions. Our retrospective mastectomy study suggested that duct lavage would have a detection rate or sensitivity of 20-25%. Cytologic assessment of lavage specimens yielded a positive diagnosis of 25%, a finding congruent with our prediction. Histologic assessment of the mastectomy specimens subjected to duct lavage showed that in most cases, the lavaged ductal systems did not coincide with disease.