PURPOSE: To investigate the incidence of and variables associated with clinically evident fat necrosis in women treated on a protocol of high-dose-rate (HDR) brachytherapy alone without external-beam whole-breast irradiation for early-stage breast carcinoma. METHODS AND MATERIALS: From 6/1997 until 8/1999, 30 women diagnosed with Stage I or II breast carcinoma underwent surgical excision and postoperative irradiation via HDR brachytherapy implant as part of a multi-institutional clinical Phase I/II protocol. Patients eligible included those with T1, T2, N0, N1 (< or = 3 nodes positive), M0 tumors of nonlobular histology with negative surgical margins, no extracapsular lymph-node extension, and a negative postexcision mammogram. Brachytherapy catheters were placed at the initial excision, re-excision, or at the time of axillary sampling. Direct visualization, surgical clips, ultrasound, or CT scans assisted in delineating the target volume defined as the excision cavity plus 2-cm margin. High activity (192)Ir (3-10 Ci) was used to deliver 340 cGy per fraction, 2 fractions per day, for 5 consecutive days to a total dose of 34 Gy to the target volume. Source position and dwell times were calculated using standard volume optimization techniques. Dosimetric analyses were performed with three-dimensional postimplant dose and volume reconstructions. The median follow-up of all patients was 24 months (range, 12-36 months). RESULTS: Eight patients (crude incidence of 27%) developed clinically evident fat necrosis postimplant in the treated breast. Fat necrosis was determined by clinical presentation including pain and swelling in the treated volume, computed tomography, and/or biopsy. All symptomatic patients (7 of 8 cases) were successfully treated with 3 to 12 months of conservative management. Continuous variables that were found to be associated significantly with fat necrosis included the number of source dwell positions (p = 0.04), and the volume of tissue which received fractional doses of 340 cGy, 510 cGy, and 680 cGy (p = 0.03, p = 0.01, and p = 0.01, respectively). Other continuous variables including patient age, total excised tissue volume, tumor size, number of catheters, number of days the catheters were in place, planar separation, dose homogeneity index (DHI), and uniformity index (UI) were not significant. Discrete variables including the presence/absence of DCIS, sentinel versus full axillary nodal assessment, receptor status, presence/absence of diabetes, and the use of chemotherapy or hormone therapy were not found to have a significant association with the risk of fat necrosis. CONCLUSIONS: In this study of HDR brachytherapy of the breast tumor excision cavity plus margin, treatment was planned and delivered in accordance with the dosimetric parameters of the protocol resulting in a high degree of target volume dose homogeneity. Nonetheless, at a median follow-up of 24 months, a high rate of clinically definable fat necrosis occurred. The overall implant volume as reflected in the number of source dwell positions and the volume of breast tissue receiving fractional doses of 340, 510, and 680 cGy were significantly associated with fat necrosis. Future dosimetric optimization algorithms for HDR breast brachytherapy will need to include these factors to minimize the risk of fat necrosis.