Measurement of breast tissue estradiol levels could provide a powerful method to predict the risk of developing breast cancer but obtaining sufficient amounts of tissue from women is difficult from a practical standpoint. Assessment of aromatase in ductal lavage fluid or fine needle aspirates from breast might provide a surrogate marker for tissue estrogen levels but highly sensitive methods would be required. These considerations prompted us to develop an ultra-sensitive, "nested" PCR assay for aromatase which is up to one million fold more sensitive than standard PCR methods. We initially validated this assay using multiple tissues from the aromatase transgenic mouse and found that coefficients of variation for measurement of replicate samples averaged less than 5%. We demonstrated a 60-fold enhancement in aromatase message in the transgenic versus the wild type mouse breast but surprisingly, levels in the transgenic animals were highly variable, ranging from 0.4 to 27 relative units. The variability of aromatase expression in the transgenic breast did not correlate with the degree of breast development and did not appear to relate to hormonal manipulation of the MMTV promoter but probably related to lack of exhaustive inbreeding and mixed zygocity of transgenic animals. Extensive validation in mouse tissues provided confidence regarding the assay in human tissues, since nearly identical methods were used. The human assay was sufficiently sensitive to detect aromatase in a single human JAR (choriocarcinoma) cell, in all breast biopsies measured, and in 7/23 ductal lavage fluids.