ABSTRACT
PURPOSE: To establish a grading system for mammographic fibrosis and correlate it with clinical fibrosis. PATIENTS AND METHODS: Analogous to the LENT/SOMA scale a four-tiered scoring scale of breast fibrosis in mammography (G0 = absent, G1 = barely increased density, G2 = definitely increased density to G3 = very marked density) was established by two observers in a group of 16 patients. Reference mammograms were selected. Independently and blinded to clinical results, three observers scored the fibrosis in mammograms of further 31 patients examined by one radiation oncologist in a cross-sectional follow-up study. Pretreatment parenchyma density was judged according to the American College of Radiology (ACR). Interobserver correlation of mammography scoring as well as correlation of mammography and clinical findings were calculated with Cohen's weighted kappa. All patients had breast-conserving surgery and axillary resection for breast carcinoma T1-2N0-1. The breast was irradiated to a median reference dose of 55 Gy (range 50-60 Gy) with 2 Gy five times weekly or 2.5 Gy four times weekly. Two patients received chemotherapy, 14 patients tamoxifen. Median age was 55 years, median follow-up 8 years (4-15 years). RESULTS: 14 of 31 patients had clinical fibrosis, twelve G1 and two G2. In mammography, mild fibrosis (G1) was seen in 12/12/18 patients (observer 1/2/3) and moderate fibrosis (G2) in 9/10/2 patients. Interobserver correlation for observers 1 and 2 who had developed the score was fair (Cohen's weighted kappa 0.64, 95% confidence interval 0.4-0.88). However, it was weak for observer 3 (0.36 and 0.42, respectively) who relied on reference mammograms only. Independent interobserver correlation of pretreatment breast density was good for all observers (Cohen's weighted kappa 0.73-0.8). The correlation of fibrosis by mammography and palpation was weak (Cohen's weighted kappa 0.32-0.42). CONCLUSION: Grading fibrosis as depicted by mammography is possible, especially if observers prepare by jointly analyzing a training group. It may be useful to study treatment effects, e. g., of fractionation or drugs, because retrospective and repeated analysis is possible. The correlation of mammography with clinical grading should be further evaluated with more objective clinical reference tools.
