| Background:Breast cancer is a kind of systemic diseases. Metastasis is the leading cause of mortality in breast cancer patients. The metastasis of breast cancer is a polygene, multistep, multistage complex biological process. Tumor with the same clinical stage, same pathologic type, and same hormone receptor status may have different metastasis phenotype due to the different biologic characteristics. Currently, we can’t evaluate the metastasis potency of tumor cell accurately by the routine clinicopathologic and biologic factors. To discover new molecular markers for metastasis prediction and establish a prediction mathematic model by combining multifactors may predict the prognosis of breast cancer more accurately and provide the objective principle for clinical individualized therapy. Objective:To verify the metastasis-related gene which were discovered by high throughput screening in previous research in large clinical cases, combining clinicopathologic and routine biomarkers factors to optimize powerful prognostic factors by univariate analysis. To establish a useful molecular diagnosis method and a mathematic model for predicting the prognosis of breast cancer by multifactors. Methods:Real-time reverse transcription-polymerase chain reaction (RT-PCR) was used to detect the mRNA expression level of metastasis-related genes and genes of clinical routine molecular makers (ER、PR、HER-2、Ki-67、P53、E-cadherin) in primary invasive ductal carcinoma of251cases with three to five years follow-up. Receiver Operating Characteristic (ROC) curve was used to identify the cut-off value of gene mRNA level to group the patients. Chi square test was used to compare the differences between/among groups. Survival analysis was carried out according to Kaplan-Meier analysis. Log-rank test was used to compare the differences of survival between groups. Multivariate Cox proportional hazards regression model was used to evaluate the significance of each marker and find the independent factor for prognosis. The mathematical model for predicting the prognosis of breast cancer patients was build by combining metastasis-related genes, molecular marker in clinical routine and clinicopathologic factors. ROC curve was used to evaluate the prognosis value of the model. Calculate the the rate of predictive coincidence, sensitivity and specificity in the testing set to verify the accurancy for the prognosis index. Compare the rate of predictive coincidence between the training set and testing set to confirm the stability of the model. Compare the rate of predictive coincidence and risk ratio between the training set and testing set to evaluate its significance in clinical use.Results:1. The significance of metastasis-related gene for predicting the prognosis of breast cancer patients by univariate analysis:Lower ER, PR, HER-2, E-cadherin, P53, BRCA1, NDE, KIF1B, GFRA, FGD-3, GATM, GGT7, RUNX2, FOXF2and USP37mRNA level were correlated with poor prognosis in breast cancer patients (P<0.05). Higher Ki-67, NET02mRNA level were correlated with poor prognosis in breast cancer patients (P<0.05). The expression level of EGFR, VEGFR1, VEGFR2, IGFBP-5, KNSL4, LATS2, NCOA, NR4A1, LOC92689, VGLL3were not related to the prognosis of breast cancer (P>0.05).2. The significance of clinicopathologic factors for predicting the prognosis of breast cancer patients by univariate analysis:Tumor size, clinical stage, histological grade, lymph node status and positive lymph nodes number were related to the prognosis of breast cancer (P<0.05), but age was not (P>0.05).3. The establishment of the mathematical model for predicting the prognosis of breast cancer patients:Clinical stage, positive lymph nodes number, Ki-67, BRCA1, GFRA expression level were independent prognosis factors for five years disease-free survival of patients (P<0.05). Clinical stage, positive lymph nodes number and Ki-67were dangerous factors for prognosis, while BRCA1and GFRA were protective factors for prognosis. Disease-free-survival prognostic index model=0.73X Clinical stage+0.79X positive lymph nodes number-0.79×BRCA1-GFRA1-0.86X RUNX2+1.46X Ki-67.4. The validation of the mathematical model for predicting the prognosis of breast cancer patients:In the training set, the rate of predictive coincidence for prognostic index is76.79%. The five years disease-free survival of cases in high risk group were significantly lower than cases in low risk group (χ2=34.30, P=0.0000). The area under the ROC curve is0.822±0.040(95%CI:0.744-0.900, P=0.000). The rate of predictive coincidence for prognostic index was higher than clinical stage (66.96%). The hazard ratio (HR) of prognostic index (HR=5.640,95%CI:2.936-10.833, P=0.000) was higher than clinical stage (HR=2.826,95%CI:1.535-5.200, P=0.001). In the testing set, the rate of predictive coincidence for prognostic index is76.67%, sensitivity is76.92%, specificity is76.47%. The five years disease-free survival in high risk group were significantly lower than in low risk group (χ22=8.77, P=0.0031). The HR of the high risk compared to low risk groups were5.629(95%CI:1.541-20.562, P=0.009) for the endpoint of disease-free-survival (DFS). There was no difference between the rate of predictive coincidence for prognostic index in training set and training test (P=0.989).Conclusion:Clinical stage, positive lymph nodes number, Ki-67, BRCA1, GFRA expression level were independent factors for predicting prognosis of breast cancer patients. The mathematical model was accurate and stable in predicting the disease-free survival of breast cancer patients. It had good prognosis predict value and clinical significance in individualized therapy. |
PDF Full Text Request