Seventy-eight examples of intraductal carcinoma (DCIS) were identified after pathologic review of 2072 specimens obtained from National Surgical Adjuvant Breast Project protocol 6. This randomized clinical trial compares the therapeutic merit of total mastectomy (TM) with lumpectomy (L), with (LX) and without (LO) postoperative irradiation. All patients were subjected to axillary lymph node dissection. Seven (14%) of the 51 patients with DCIS treated by L exhibited breast recurrence within or close to the site of the initial lesion 4 to 53 months (average, 16 months) after L. Only 2 (7%) of these events occurred in the 29 women treated by LX, as opposed to 23% in the LO group. No pathologic features were noticed that might have been considered predictive of local breast recurrence. The three DCIS recurrences and the four invasive forms noted are considered to represent overlooked or incompletely excised foci of cancer because of the multifocality (not multicentricity) of some breast cancers. The possibility that DCIS may represent a marker of risk for the development of cancer rather than a precursor lesion per se is suggested. Despite apparent difficulties in the pathologic diagnosis of DCIS as well as uncertainty concerning its natural history, no evidence was found to indicate that it represents a more ominous disease than invasive cancer. Indeed, treatment failure occurred in only one patient treated by LX and a similar number subjected to TM (4% versus 2%). Although these observations are short term (average follow-up, 39 months), estimates of the probability of local recurrence or survival suggest that they will not be significantly altered after longer periods of surveillance. Thus, there are no compelling reasons why DCIS may not be treated in a cosmetically acceptable manner by LX. A randomized clinical trial addressing this issue is now in progress.