Abstract
BACKGROUND: Current predictive model is not developed by inflammation-related genes to evaluate clinical outcome of breast cancer patients. METHODS: With mRNA expression profiling, we identified 3 mRNAs with significant expression between 15 normal samples and 669 breast cancer patients. Using 7 cell lines and 150 paraffin-embedded specimens, we verified the expression pattern by bio-experiments. Then, we constructed a three-mRNA model by Cox regression method and approved its predictive accuracy in both training set (n = 1095) and 4 testing sets (n = 703). RESULTS: We developed a three-mRNA (TBX21, TGIF2, and CYCS) model to stratify patients into high- and low-risk subgroup with significantly different prognosis. In training set, 5-year OS rate was 84.5% (78.8%-90.5%) vs 73.1% (65.9%-81.2%) for the low- and high-risk group (HR = 1.573 (1.090-2.271); P = 0.016). The predictive value was similar in four independent testing sets (HR>1.600; P < 0.05). This model could assess survival independently with better predictive power compared with single clinicopathological risk factors and any of the three mRNAs. Patients with both low-risk values and any poor prognostic factors had more favorable survival from nonmetastatic status (HR = 1.740 (1.028-2.945), P = 0.039). We established two nomograms for clinical application that integrated this model and another three significant risk factors to forecast survival rates precisely in patients with or without metastasis. CONCLUSIONS: This model is a dependable tool to predict the disease recurrence precisely and could improve the predictive accuracy of survival probability for breast cancer patients with or without metastasis.