半参数时依AUC方法在弥漫大B细胞淋巴瘤患者预后评价中的应用

发布时间：2018-05-26 12:03

本文选题：IRF4/MUM1 + 弥漫大B细胞淋巴瘤　；参考：《山西医科大学》2017年硕士论文

【摘要】：目的:本研究的目的是引入半参数时依AUC方法(semi-parametric time-dependent AUC method)来评估IRF4/MUM1的表达对弥漫大B细胞淋巴瘤(DLBCL)患者的预后价值。本研究分为两部分,第一部分Meta分析,全面和定量的研究了IRF4/MUM1表达与DLBCL患者预后之间的关系。另一部分是方法应用,使用半参数时依AUC方法来评估,添加一个肿瘤标志物IRF4/MUM1后是否会改善国际预后指数(IPI)对DLBCL患者预后的预测准确度。方法:检索Pubmed,万方和CNKI数据库从建库到2016年12月已发表的关于IRF4/MUM1与弥漫性大B细胞淋巴瘤的预后有关的所有中英文文献,过程中由两名评价者独立完成文献阅读。提取IRF4/MUM1正常组和升高组总体生存率(OS)的HR及HR95%CI,并采用固定效应模型对其进行合并,考虑到多种因素影响进行亚组分析,采用敏感性分析估计合并风险比的变化情况。在本文中,我们采用半参数时依AUC方法构建关于时间t的AUC(t)函数,使用灵活的分数多项式方法(fractional polynomials model)建模。该方法可以跳过ROC曲线估计,直接得到时依AUC估计量。对于模型参数的估计采用伪偏似然估计法(pseudo partial-likelihood procedure),同时采用总的一致性指标IAUCC(τ)比较两生物标志物间总的预测准确度。结果:Meta分析纳入19篇文章,累积共有DLBCL病例1699例。各研究间无明显的异质性(2X=24.79,自由度df=18,P=0.131,2I=27.4%),合并总效应量HR的值为2.04(95%CI:1.75-2.38),对结果进行检验有统计学意义(Z=9.24,P0.001)。随后经亚组分析和敏感性分析结果显示,IRF4/MUM1是DLBCL患者预后的重要危险因素。收集2014年1月初至2016年9月底山西省肿瘤医院血液科诊断明确、随访资料齐全的DLBCL病例315例。将IRF4/MUM1加入IPI中,并建立一种新的0-6分的预后评分系统称为肿瘤标志物调整的IPI(T-IPI),并且探讨T-IPI对DLBCL患者预后评价的预测准确度。从IPI和T-IPI的AUC(t)图形及其可信区间可见,随着时间的变化,两预后指标的AUC(t)曲线都在AUC(t)=0.5的直线上方且变化较为平稳,并且明显可见T-IPI的AUC(t)曲线在IPI之上,提示T-IPI对DLBCL的预测准确度要好于IPI。为了确定这种优势是否具有统计学意义,我们检验了总的一致性指标IAUC C(τ)在IPI和T-IPI之间的差异,IAUCC(τ)的估计差异为0.138,标准误差为0.059,得出T-IPI明显具有比IPI更好的预测准确度的结论(P=0.026)。结论:Meta分析结果和半参数时依AUC方法分析结果均显示IRF4/MUM1阳性表达是DLBCL患者预后的危险因素。将IRF4/MUM1加入IPI中可以有助于提高其对DLBCL患者预后的预测准确度。
[Abstract]:Objective: to evaluate the prognostic value of IRF4/MUM1 expression in patients with diffuse large B-cell lymphoma (DLBCL) by semi-parametric time-dependent AUC method (AUC method). This study is divided into two parts. The first part, Meta analysis, comprehensively and quantitatively studied the relationship between IRF4/MUM1 expression and prognosis of DLBCL patients. The other part is the application of the method. The semi-parametric AUC method is used to evaluate whether the addition of a tumor marker IRF4/MUM1 will improve the prediction accuracy of international prognostic index (IPI) for the prognosis of DLBCL patients. Methods: all published Chinese and English literatures on IRF4/MUM1 and prognosis of diffuse large B-cell lymphoma were searched in Pubmed, Wanfang and CNKI databases from the establishment of the database to December 2016. The literature was read independently by two reviewers. HR and HR95CIof IRF4/MUM1 normal group and elevated group were extracted. Fixed effect model was used to combine HR and HR95 CI.The sensitivity analysis was used to estimate the change of risk ratio. In this paper, we use the semi-parametric AUC method to construct the AUC t function for time t, and use the flexible fractional polynomial method to model the fractional polynomials model. This method can skip the ROC curve estimation and get the AUC estimator directly. Pseudo partial-likelihood procedure is used to estimate the parameters of the model, and the total accuracy of prediction between the two biomarkers is compared by using the general consistency index IAUCC( 蟿). Results A total of 1699 cases of DLBCL were included in 19 articles. There was no obvious heterogeneity between the two studies, and there was no obvious heterogeneity between the two studies. The degree of freedom (dftl) was 0.131 ~ 2i ~ (2) I ~ (27. 4), and the total effect HR was 2.0495% CI: 1.75-2.38. The results were statistically significant. The subgroup analysis and sensitivity analysis showed that IRF4 / MUM1 was an important prognostic factor in DLBCL patients. From the beginning of January 2014 to the end of September 2016, 315 cases of DLBCL with definite diagnosis and complete follow-up data were collected. IRF4/MUM1 was added to IPI, and a new prognostic scoring system with 0-6 scores was established, which was called IPIT-IPIN adjusted by tumor markers. The predictive accuracy of T-IPI in evaluating the prognosis of DLBCL patients was discussed. From the IPI and T-IPI AUC t) patterns and their confidence intervals, the AUC t) curves of the two prognostic indicators were both above the straight line of AUC(t)=0.5 and changed steadily with the change of time, and it was obvious that the AUC t) curves of T-IPI were above IPI. The results suggest that the prediction accuracy of T-IPI to DLBCL is better than that of IPI. In order to determine whether this advantage is statistically significant, we tested the difference between IPI and T-IPI of the total consistency index IAUC C (蟿). The estimated difference of IAUCC (蟿) is 0.138 and the standard error is 0.059. It is concluded that T-IPI has better prediction accuracy than IPI. Conclusion the positive expression of IRF4/MUM1 was found to be a risk factor for the prognosis of patients with DLBCL. Adding IRF4/MUM1 to IPI can improve the accuracy of predicting the prognosis of DLBCL patients.
【学位授予单位】：山西医科大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R733.1

【参考文献】