A Signature That May Be Predictive of Early Versus Late Recurrence After Radiation Treatment for Breast Cancer That May Inform the Biology of Early, Aggressive Recurrences.

摘要: Purpose Unmet clinical needs in breast cancer (BC) management include the identification of patients at high risk local failure despite adjuvant radiation and an understanding biology these recurrences. We previously reported a response signature here extend those studies to identify predictive recurrence timing (before or after 3 years). Methods Materials Two independent patient cohorts were used. The training cohort included 119 with in-breast tumor (343 total), validation testing had 16 recurrences (112 total). All received treatment breast-conserving surgery. Initial feature selection used Spearman rank correlation, linear model was trained locked before validation. Cox regression for univariate multivariable analyses (UVA MVA, respectively). Biologically related concepts identified using gene set enrichment analysis. Results correlation 485 genes whose expression significantly associated time (early vs late). Feature reduction further refined list 41 retained within signature. In training, score 0.85 (P value Conclusions report that may unlikely respond be predict implications potential intensification duration follow-up women treated radiation.