Postoperative women with breast cancer but without histopathological evidence of metastases to the axillary lymph nodes or clinical evidence of metastases were studied. Six hundred fifty-five "good-risk" patients who were estrogen receptor positive (ER+) with primary tumors less than 3 cm were registered for observation. Twenty-four of these patients were treated with chemotherapy. Five hundred thirty-six "poor-risk" patients who were either ER+ with primary tumors greater than or equal to 3 cm or estrogen receptor negative (ER-) with any primary tumor size were randomly assigned between chemotherapy and observation. Randomization was stratified by type of surgical procedure, number of lymph nodes examined, menopausal status, tumor size, and ER status. The chemotherapy (CMFP) consisted of six 4-week cycles of cyclophosphamide, 100 mg/m2 orally days 1-14; methotrexate, 40 mg/m2 intravenously (IV) days 1 and 8; fluorouracil, 600 mg/m2 IV days 1 and 8; and prednisone, 40 mg/m2 orally days 1-14. Treatment arms in the randomly assigned patients were balanced with respect to pretreatment characteristics. This analysis includes 445 eligible patients entered in the registration arm and 425 eligible patients entered into the randomized treatments. The median follow-up is 4.5 years in the randomly assigned cohort and 4.8 years in the registered cohort. The overall 5-year disease-free survival (DFS) among the randomly assigned patients was 83% with CMFP and 61% with observation (P less than .0001). A DFS treatment benefit was observed in premenopausal and postmenopausal patients as well as in patients with ER+ or ER- tumors. There were fewer local-regional and distant relapses among the CMFP-treated patients.(ABSTRACT TRUNCATED AT 250 WORDS)