We have stratified the cancer risk implications of lobular pattern in situ neoplasias of the breast by separating marked examples of this histologic spectrum (lobular carcinoma in situ [LCIS]) from lesser examples (atypical lobular hyperplasia). The lesser-developed examples have been shown previously to have a lower relative risk (RR) of later invasive carcinoma of the breast (IBC). Forty-eight examples of LCIS were found in 10,542 otherwise benign breast biopsies, representing an incidence of 0.5%. Nine patients were excluded from follow-up because of bilateral mastectomy within 6 months of entry biopsy, IBC within 6 months of entry biopsy, or prior IBC. Follow-up of the remaining 39 patients was complete, averaged 18 years, and revealed an RR of subsequent IBC of 6.9 (P less than .00001). Average overall follow-up for LCIS patients was 19 years; it was 25 years for those alive and free of IBC at the time of their follow-up interview. Neither family history of IBC nor postmenopausal estrogen therapy further affected risk. The absolute risk of IBC after LCIS was 17% at 15 years (adjusted for withdrawals), and the RR was 8.0 in the first 15 years of follow-up compared with the general population. An analysis based on a time-dependent hazards model found that during the first 15 years following biopsy women with LCIS had 10.8 times the risk of breast cancer compared with biopsied women of comparable age who lacked proliferative disease. Some previously published articles reporting lobular neoplasia (LN) suggest that those series with the greatest incidences of LN (whether termed LN or LCIS) have the lowest RR of subsequent breast cancer. Those series with higher incidences of LN include less well-developed histologic patterns of LN (atypical lobular hyperplasia). We conclude that our study of LN and studies performed by others support the higher risk of IBC after histologically flagrant examples (LCIS, about nine times higher) and a relatively lower but definable risk after more histologically subtle examples (atypical lobular hyperplasia, four to five times lower). This relative cancer risk is probably not constant over more than 15 years; thus, cancer risk 15 to 25 years after initial diagnosis of LCIS is uncertain.