节点文献

淋巴结阳性乳腺癌预后因素分析及决策树预后模型的构建

Analysis of Prognostic Factors of and to Establish a Prognostic Decision Tree Model for Patients with Nodes Positive Invasive Breast Cancer

【作者】 张斌

【导师】 郝希山;

【作者基本信息】 天津医科大学 , 肿瘤学, 2008, 博士

【摘要】 研究目的:对浸润性乳腺癌患者临床-病理指标-预后之间关系进行数据挖掘,探讨影响乳腺癌预后的临床病理参数。通过分层解析的方式分析不同阳性淋巴结情况下,腋窝淋巴结外浸润(ECE)现象和阳性-切检淋巴结比率(PER)对患者的影响及不同淋巴结状态下的不同比例风险以及和放疗及化疗等保护性因素之间的交互作用。拟合腋淋巴结阳性乳腺癌患者预后数学模型。材料与方法:对我院1988-1995年间1230例腋淋巴结阳性乳腺癌患者进行随访,建立大规模数据库,探索TNM系统之外对患者预后明显影响的临床病理指标,主要探索腋窝淋巴结外浸润(ECE)现象和阳性-切检淋巴结比率(PER)对患者的影响,及在不同淋巴结状态下的不同比例风险以及和放疗及化疗等保护性因素之间的交互作用。并且用一种新的统计学方法——决策树法,建立腋窝淋巴结阳性乳腺癌预后数学模型,并且用379例腋窝阳性淋巴结患者进行验证。结果:(1)1230例腋淋巴结阳性乳腺癌患者中486例(39.5%)出现ECE,死亡598例(48.62%),中位生存期113.7月,5年生存率66.29%,10年生存率55.28%,随肿物直径增大、阳性淋巴结数目和检查淋巴结数目增加及PER增高ECE阳性率也增高;Logistic回归模型单因素分析表明阳性淋巴结、受检淋巴结、PER和肿瘤直径是ECE出现的危险因素,多因素分析表受检淋巴结不是独立危险因素。ECE阳性患者较易转移,但与非内脏转移更相关。(2)ECE+是未化疗患者生存期缩短的危险因素,无论ECE状态所有患者均可从化疗中获益,ECE+患者获益更大但化疗不能平衡ECE+的死亡风险。1~3个阳性腋淋巴结(LNM)ECE对总生存期影响不显著,LNM>4时影响显著,LNM4-9时更显著。(3)分层分析,LNM1~3时ECE+不是未化疗患者生存缩短的危险因素,ECE不影响化疗效果,化疗能平衡ECE+的死亡风险;LNM4-9时ECE+是未化疗患者生存缩短的危险因素,ECE+患者化疗获益较多,化疗能减小ECE+带来的死亡风险,化疗的保护不足以平衡ECE的危险;LNM>10时ECE+不是未化疗患者生存缩短危险因素,所有患者均可从化疗中获益,ECE—患者获益多,化疗保护效应能平衡ECE+的风险.(4)ECE+是未放疗患者复发的危险因素;所有患者均从放疗中获益,放疗不能平衡ECE+带来的局部复发风险;ILNM1~3时ECE对患者LRF影响不显著;LNM≥4时,ECE对患者LRF影响具有显著性;LNM4-9时,对患者复发影响最为强烈。(5)分层分析LNM1~3时ECE和放疗对LRFFS均无影响,所有患者均未放疗中获益;LNM4-9时ECE+是未放疗患者复发危险因素,ECE+患者放疗获益较大,因此放疗能平衡ECE带来的复发危险;LNM≥10时ECE+不是未放疗患者LRFFS缩短的危险因素,所有患者均未从放疗中获益,ECE+患者因死亡率增加复发率降低。≥10阳性例巴结患者各组中均有53-78.1%患者发生无复发死亡,局部复发率的降低由于较高死亡率所致,预防性放疗收益有限。(6)PER是总生存期缩短的预后不良因素(HR=1.724),对于LNM1~3患者分割点为PER>0.2,对于LNM4-9患者为PER>0.4,对于LNM>10患者为PER>0.6,采用不同分割点后PER在各淋巴结分层中的作用相似。(7)亚组分析表明,各组中PER不是未化疗患者生存缩短的危险因素,LNM1~3时PER≤0.2患者化疗获益,PER>0.2患者不获益,化疗后PER>0.2患者生存期比PER≤0.2患者短;LNM4-9PER≦0.4患者化疗获益较大,PER>0.4患者获益较小;LNM≧10患者,PER<0.6患者化疗获益较大,PER>0.6患者化疗获益小。(8)LNM1~3患者,肿瘤直径是预后次级因素;LNM4~9患者ECE是预后次级因素;LNM≧10患者,PER是预后次级因素;(9)利用决策树技术拟合乳腺癌预后模型,准确率为91.56%,误差率已明显小于Logistic和Cox风险比例回归模型拟合同类模型(82.4%和86.1%),提示决策树模型对异质性较强的腋窝淋巴结阳性乳腺癌更为适用结论:(1)ECE和化疗存在交互作用,但ECE影响更大,不同LNM下ECE作用不同;LNM1-3时化疗对生存的保护大于ECE的危险;只有LNM4-9未化疗患者ECE影响患者生存,LNM1~3时所有患者化疗效果相似,LNM4-9时ECE+患者化疗获益更大,LNM>10时ECE—患者化疗获益更大。(2)LNM1~3时所有患者不从放疗中获益,ECE不是放疗指证,LNM4-9时放疗能平衡ECE复发危险, ECE+患者放疗获益大,LNM≥10时ECE+不是未放疗患者复发危险因素,53-78.1%患者复发前死亡,预防性放疗收益有限。随阳性淋巴结数目增加,死亡率增加,因为复发需要放疗的患者,ECE+是危险因素。(3)对于不同阳性淋巴结状态采用不同的分割点分析更合适。PER较高均是患者生存期缩短的危险因素,对于PER较高并且已经死亡的患者,PER的EF均在50%左右,PER较大的患者,化疗收益减小;(4)决策树模型对异质性较强的腋窝淋巴结阳性乳腺癌更为适用,是一种简单实用的乳腺癌预后拟合数学模型。

【Abstract】 Purpose:To data mine, by the methods of multivariate statistics, on the clinical-pathological parameters and the result of systematic follow-up of the patient suffering invasive breast carcinoma, and to clarify the prognostic impact of these parameters on the survival of invasive breast cancer. We also intend to explored the prognostic impact of extra capsular extension (ECE) and positive-examined rate (PER) on maxillary lymph node involved invasive breast carcinoma and stratified analysis the impact by axillary lymph node status, and also to explored the different hazards proportion of the parameters and the interaction with protective factor, such as chemotherapy and radiotherapy, in different lymph node status.Materials and methods:We have established a huge data base as the result of systematic follow-up of1230patients with lymph node positive invasive breast carcinoma, who had admined in the Tianjin cancer hospital during1988to1995.By the method of data mining, we explored several significant prognostic factors besides TNM system, especially focusing on the impact of the ECE and PER on the survival data and the interaction of the implact posed by the risk factors, such as lymph node involved, ECE and PER, and the protective factors, such as chemotharapy and radiotharapy. In addition, we have tried to established a prognostic model by the method of decision tree, which was trained by1230cases of positive nodes invasive breast cancer and confirm by379partner with similar clinical scenarios.Result:Among the1230nodes positive breast cancer cases,486(39.5%) patients founded presented with ECE and598(48.6%) cases died. The median survival period is113.71months, while the5years overall survival rate is66.29%and the10years rate is55.28%. The patients who were41-50years old have the lowest ECE rate, and the patients younger than40years of older than50years have higher ECE rates. The ECE rates correlated with tumor size, the number of positive nodes, PER and the number of examined nodes. All of the factors are prognostic factors in the univariate Logistic regression analysis, while only tumor size, the number of positive nodes, PER are independent prognostic factors in the multivariate Logistic regression analysis, the number of examined nodes is not an independent prognostic factors. The ECE positive cases inclined to metastasis but more associated with no-viscera metastasis. ECE positive is risk factor for the non-chemotherapeutic patients. All of the patients could gain benefit from chemotherapy regardless of ECE status, but the ECE positives cases gained more. Despite this, the benefit from chemotherapy could not balance the risk from ECE. For the patients with1-3positive nodes, ECE has not a significant impact on overall survival, but has a significant impact when the patients have more than4positive nodes, the impact more significant in the4-9LNM cases. Stratified analysis according to the number of lymph nodes proved that ECE was not risk fact for overall survival in the LNM1-3patients, did not influenced the effect of the chemotherapy, and chemotherapy could balance the risk brought by ECE. On the contrary, in the LNM4-9patients ECE was risk fact for OS,ECE positive patients gained more benefit from chemotherapy, and chemotherapy could only reduce but not balance the risk brought by ECE completely. When the LNM≥10, ECE was not risk fact for OS, all of the ECE positive patients gained more benefit from chemotherapy, but the ECE negative cases gained more, and chemotherapy could only reduce but could not balance the risk brought by ECE completely. All of the patients could gain benefit from chemotherapy regardless of ECE status, but the ECE positives cases gained slightly more. The effect of chemotherapy was similar, and the protective impact from chemotherapy was stronger than the risk impact from ECE. Despite this, the overall prognoses of the patients in these two groups were not good. ECE positive is risk factor of local regional recurrence for all patients had not received radiotherapy. All of the patients could gain benefit from radiotherapy regardless of ECE status, the benefit from radiotherapy could not balance the recurrent risk from ECE. For the patients with1-3positive nodes, ECE has not a significant impact on local-regional failure free survival (LRFFS), but has a significant impact when the patients have more than4positive nodes, the impact more significant in the4-9LNM cases. Stratified analysis according to the number of lymph nodes proved that neither ECE nor radiotherapy was risk fact for LRFFS in the LNM1-3patients, all of the patients could not gain benefit from radiotherapy regardless of ECE status. On the contrary, in the LNM4-9patients ECE was risk fact for LRF, ECE positive patients gained more benefit from radiotherapy, and radiotherapy could balance the risk brought by ECE completely. When the LNM≥10, ECE was not risk fact for LRFFS, none of the ECE positive patients could gained more benefit from chemotherapy. ECE positive patients have less LRF rate because higher mortality rate, for, there were53-78.1%mortality without metastasis. The benefit of radiotherapy was limit for this group of patients. PER was a risk factor of overall patients (HR=1.724). The cutting-point is0.2for the patients with1-3positive nodes, while0.4for4-9positive nodes and0.6for LNM>10cases. The impacts of PERs were similar when adopted different cutting-point in different groups of patients. Stratified analysis according to the number of lymph nodes proved that PER>0.2was not risk fact for OS in the LNM1-3non-chemotherapy patients. Patients with PER>0.2gain benefit from chemotherapy, but the patients could not gain benefit from chemotherapy when the PER≤0.2. If receiving chemotherapy, the OS of patients with PER≤0.2shorter than those PER>0.2. PER>0.4was not risk fact for OS in the LNM4-9non-chemotherapy patients. All patients gained benefit from chemotherapy, while the patients gain less benefit from chemotherapy when the PER≤0.4. If receiving chemotherapy, the OS of patients with PER≤0.4shorter than those PER>0.4. PER>0.6was not risk fact for OS in the LNM≥10non-chemotherapy patients. Patients with PER<0.6gain more benefit from chemotherapy, while the patients could gain less benefit from chemotherapy when the PER≤0.6. If receiving chemotherapy, the OS of patients with PER>0.6shorter than those PER>0.6.In the decision tree algorithm, Tumor size was second main prognostic factor in the patients with1-3positive nodes after number of positive nodes, which was the first important prognostic factor. However, ECE was the second in the cases with4-9positive nodes and the PER for the LNM>10. We constructed a decision tree prognostic regression model for nodes positive breast cancer patients with the veracity is91.56%, which was significant higher than the algorithm of Logistic regression and Cox hazard proportion regression model.Conclusion:The impact of ECE contra-interacted with chemotherapy, but the ECE more influent than chemotherapy. The impacts of ECE were different according to the numbers of positive nodes. The chemotherapy has more impact than ECE for the patients with1-3positive nodes. ECE only reduced the OS in the patients with4-9positive nodes. All of the patients obtain equally benefit from chemotherapy, while the ECE positive patients gain more in4-9nodes patients and the ECE negative patients gain more if the patients have more than10positive nodes. The ECE relative dead evens increased according to the number of involved nodes. All of the patients could not beneficial from radiotherapy,1-3positive node is not the indication for radiotherapy. For the patients with4-9positive nodes, radiotherapy could balance the local-regional recurrence risk brought by ECE, ECE positive patients will gain more from radiotherapy. ECE was not recurrence risk factor for the patients with more than10positive nodes, because53-78.1%patients would dead before local recurrence, that mean the benefit of prophylactic radiotherapy was limited. However, in the patients needed radiotherapy for local-regional failure, the ECE was a significant risk factor (increasing4%recurrent rate). The result from the research suggested that it should adopt different cutting-points for different number of positive lymph-nodes. A higher PER was a risk factor for reducing overall survival. For the patients who died in the follow period and with a higher PER, the etiological factor were50%in all groups of patients, the patients with higher PER would obtain less from chemotherapy. The decision tree algorithm was more suitable prognostic model for the nodes positive patients, which have more heterogeneity, and was a single and practical prognostic model.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络