Dr. Rosser points out an important problem: the accuracy of immunohistochemistry with keratin stains when it is used to evaluate the status of axillary lymph nodes thought to be negative by routine hematoxylin and eosin (H & E) staining. A key problem that arises, in some cases, is distinguishing a true metastasis from iatrogenic needle debris that has migrated to a lymph node. Making the distinction can be exceptionally difficult because some malignant cells, such as invasive lobular carcinoma cells, may appear to be benign. In addition, if a group of cells within a lymph node are clearly malignant cells, we have no way of being certain whether or not they were destined to be destroyed by that lymph node or whether they had the molecular genetic abilities required to become a metastasis. The Ludwig trial,1 in which negative axillary lymph nodes were serially sectioned at 6 levels and stained with H & E, revealed a 9% incidence of previously undetected micrometastasis. Patients who were converted to lymph node positive status by this technique had a 16% decrease in 5-year disease free survival and a 9% decrease in overall survival. Moreover, these data held up over 10 years of follow-up (R. Gelber, personal communication, December 1995). However, step sectioning of axillary lymph nodes is a tedious, time-consuming, and expensive process. Our hope is that immunohistochemical analysis will be as accurate as step sectioning in identifying these patients. Our concern, and clearly Dr. Rosser's concern, is that it may be "too accurate." The impact of "immunochemically only" positive lymph nodes has been examined for years (Table 1). In one of the largest series,2 12% of 343 routinely studied lymph node negative patients were found to have positive lymph nodes by immunohistochemical analysis. There was a 28% decrease in 5-year disease free survival but no statistically significant effect on overall survival. On the other hand, Trojani et al.3 demonstrated a 25% decrease in overall mortality among patients with immunochemically positive axillary lymph nodes. The harder one looks for positive lymph nodes, the more will be found. Current technology using polymerase chain reaction (PCR) may yield an even higher percentage of lymph node positivity4, 5 than immunohistochemistry. Clearly, patients with lymph nodes that are shown to be positive by one of these techniques and negative by standard H & E staining are not the same as patients with lymph nodes that are shown to be positive by standard pathology. In all likelihood, they have a better prognosis. Therefore, as these techniques become more widely used, lymph node positivity must be described by the method used to determine it. For example, a patient may be N0 (H & E) and N1 (immunohistochemistry). At this point in the evolution of our knowledge, we cannot be sure whether this patient has Stage I or Stage IIA disease. As even more sensitive techniques, such as polymerase chain reaction (PCR), become more widely available, we will have patients that are N0 (H & E), N0 (immunohistochemistry), N1 (PCR). As these new data accrue, they will have to be equated with outcome to determine the proper stage; in all likelihood, the current staging system will have to be amended to reflect this. Although we understand Dr. Rosser's point completely, "Don't ask, don't tell" is not acceptable for those of us studying and treating cancer patients. If that were our policy across the board, there would be little progress in cancer research. Rather, we must ask the hard questions. Immunohistochemical staining of axillary lymph nodes shown to be negative by routine H & E staining is but one attempt to do that. Should therapeutic decisions be made based on immunohistochemistry of axillary lymph nodes? Perhaps not today. More research, and our continued asking of the question, will determine whether it becomes standard policy in the future. We know that 30% of breast carcinoma patients who are lymph node negative (by routine H & E) ultimately die of their disease. How, then, do we identify this 30% so that they can receive more appropriate initial therapy? The correlation of patients who have immunohistochemically positive lymph nodes with outcome is but one attempt to do that. Melvin J. Silverstein M.D.*, Pamela H. Craig M.D.*