运用动态生存分析模型评价多类血液化验指标对胰腺癌病人生存的意义

Pancreatic cancer (PC) is a common type of malignant tumor. Because of the difficulty in early diagnosis as well as the slow progress in treatment, expanded searching of associated factors in survival of PC patients has become the emphasis and hot spot of study in the field of cancer epidemiology. The association between blood indicators and the survival of PC patients has gradually attracted the attention of researchers in recent years. However, only very few studies were published in this field. Moreover, nearly all currently available studies adopted traditional static Cox proportional hazards model to analyze the association between blood indicators and the survival of PC, nevertheless, during the whole survival period of PC patients, blood indicators always fluctuate, under this circumstance, in order to accurately estimate their associations with PC survival, dynamic survival analysis methods must be applied. In this study, we intend to use counting process method in combination with multiple failure-time Cox model to discuss the survival significance of multiple common blood indicators, and compare the differences in results of static and dynamic survival methods. Through the implementation of this study, we expect to fill the deficiency in the application of dynamic methods in cancer epidemiology field, provide a new angle in fully explanation of cancer survival results, and generate new thoughts in clinical treatment policies and measurements of PC.

胰腺癌是一种常见的恶性肿瘤，在早期诊断难以实现及治疗方法进展缓慢的背景下，全面寻找胰腺癌病人生存相关影响因素已成为当前肿瘤流行病学研究领域的重点和热点之一。血液化验指标与胰腺癌病人生存的关联近年来渐渐引发学界关注，但目前这一领域研究数量稀少，亟待充实。另外，现有的研究几乎都是采用静态的传统Cox比例风险模型分析血液化验指标与胰腺癌生存的关联，而在胰腺癌病人生存的全过程中，血液化验指标总是动态变化的，要正确估计动态指标与胰腺癌病人生存的关联，就必须采用动态生存分析的方法。本课题拟采用多结局Counting process法这一动态生存分析模型评价多类常见的血液化验指标对胰腺癌病人生存的意义，并比较动态和静态生存分析方法分析结果的差异。通过本项目的开展，可填补目前肿瘤流行病学领域动态生存分析方法运用的不足，调整生存分析结果全面解读的新视角，并为胰腺癌病人的临床治疗策略和措施提供新思路。

血液化验指标与恶性肿瘤病人生存的关联性研究是目前肿瘤流行病学领域的热点研究问题。在病人生存过程中，基本所有的血液化验指标都呈现随时间动态变化的特征，是典型的时依性变量，要评价它们与肿瘤病人预后的关联，只有运用考虑其变动特征的动态生存分析模型才可能得到更为可信的研究结论。本项目针对目前恶性肿瘤生存分析领域动态生存分析方法运用严重不足、亟待推广的现状，采用基于生存数据Counting process重构的动态多结局模型，评价了多类常见的血液化验指标与551名胰腺癌病人总体生存（Overall survival, OS）的关联。我们发现：系统炎症指标血小板-淋巴细胞比值（PLR）、碱性磷酸酶（ALP）和γ-谷氨酰转移酶（GGT）三个指标与胰腺癌病人总体生存显著关联，三个指标的升高，都与胰腺癌病人更差的总体生存相关。另外我们还比较了传统静态Cox比例风险模型和动态生存分析的多结局模型在评价血液指标与胰腺癌病人OS关联时分析结果的差异，结果显示：静态和动态生存分析模型分析结果存在较大差异，动态生存分析模型考虑了指标动态变化的特性，得到的结果更为真实可信。此外，针对生存分析时往往需要对连续型血液化验指标按照界值点进行二分类处理的情况，我们也自行研制了用于界值点敏感性分析的SAS宏程序。通过当前研究的开展，有效推进了动态生存分析方法在肿瘤流行病学领域的运用。此外，筛选出了PLR、ALP和GGT等三个与胰腺癌病人预后显著关联的指标，有望为实现胰腺癌病人生存改善提供重要临床治疗思路。