Figure 3

Sensitivity analyses of the BRCA-like^CGHstatus. We varied the previously determined thresholds defining the BRCA-like^CGH status (that is the threshold of the BRCA1-like^CGH and BRCA2-like^CGH patterns (13, 14)) and the empirically chosen thresholds of the aCGH quality (profile-quality status) and the tumor percentage, and evaluated the influence on results for overall survival. (A) The thresholds of the BRCA1- and BRCA2-like^CGH patterns, which define the BRCA-like^CGH status, were increased by 0.1 (from 0.63 to 0.73 and from 0.5 to 0.6, respectively (13, 14)); (B) similarly, thresholds were decreased by 0.1; (C) The threshold determining aCGH quality was increased (from 0.85 to 0.95 (14)), resulting in a subgroup of 225 patients; (D) The threshold of tumor percentage was increased from 60% to 70% resulting in a subgroup of 198 patients. All analyses were stratified for number of lymph nodes (4-9 vs. ≥10) and double-negative ER/PR status (ER <10% and PR <10% vs. other) and adjusted for pathologic tumor size (T1 vs. T2 vs. T3), histologic grade (I vs. II vs. III) and BRCA-like^CGH status. aCGH, array comparative genomic hybridization; ER, estrogen receptor; FE₉₀C, 5-fluorouracil, epirubicin, cyclophosphamide; HD-CTC, high-dose cyclophosphamide-thiotepa-carboplatin; PR progesterone receptor.