The assessment of DNA had demonstrated altered methylation in malignant compared to benign breast tissue. The purpose of our study was to (i) confirm the predictive ability of methylation assessment in breast tissue, and (ii) use the genes found to be cancer predictive in tissue to evaluate the diagnostic potential of hypermethylation assessment in nipple aspirate fluid (NAF) and mammary ductoscopic (MD) samples. Quantitative methylation specific (qMS)-PCR was conducted on three specimen sets: 44 malignant (CA) and 34 normal (NL) tissue specimens, 18 matched CA, adjacent normal (ANL) tissue and NAF specimens, and 119 MD specimens. Training and validation tissue sets were analyzed to determine the optimal group of cancer predictive genes for NAF and MD analysis. NAF and MD cytologic review were also performed. Methylation of CCND-2, p16, RAR-beta and RASSF-1a was significantly more prevalent in tumor than in normal tissue specimens. Receiver operating characteristic curve analysis demonstrated an area under the curve of 0.96. For the 18 matched CA, ANL and NAF specimens, the four predictive genes identified in cancer tissue contained increased methylation in CA vs. ANL tissue; NAF samples had higher methylation than ANL specimens. Methylation frequency was higher in MD specimens from breasts with cancer than benign samples for p16 and RASSF-1a. In summary, i) routine quantitative DNA methylation assessment in NAF and MD samples is possible, and ii) genes hypermethylated in malignant breast tissue are also altered in matched NAF and in MD samples, and may be useful to assist in early breast cancer detection.