长期生存者资料的参数混合模型

发布时间：2019-03-08 19:12

【摘要】： 在肿瘤临床试验中,由于病人总体的异质性,有些对处理敏感的患者不会出现任何癌症症状或迹象,则他们被认为是“长期生存者(long term survivors)”或治愈者(cured individuals),即在充足的随访时间内,这类个体不会发生某一特定的终点事件,通常具有较长的删失生存时间。长期生存者意味着直到研究结束他们也不发生或根本就不会发生终点事件(死亡、疾病的复发等)。对这种资料的分析,首先要判断资料中是否有长期生存者存在?如果资料没有长期生存者,或者没有证据证实有长期生存者存在,则采用传统生存分析方法;若有证据证实资料中有长期生存者存在,且随访充足,采用传统分析方法不恰当,因为它将长期生存者都作为截尾观察对象来处理,显然这是不合理的。盲目采用传统生存分析方法,有可能造成结果解释不合理,甚至得到相反的结论。本文介绍适用于长期生存者资料的统计模型——参数混合模型,包括指数混合模型、Weibull混合模型、Burr XII混合模型。这三种模型在分析长期生存者资料中都有广泛应用。我们利用拟合优度检验选出拟合实际资料的最佳模型,并用相应的模型作进一步分析。参数估计采用最大似然法,通过软件S-PLUS6.0来实现。通过实例分析证实,对有长期生存者存在的资料,若采用传统生存分析方法,由于不能很好地利用资料所提供的信息,得出的结论不仅不全面,甚至是相反的;如果采用适于长期生存者资料的参数混合模型分析,可得到全面、正确的结论,从不同侧面提供更多的信息,能有效合理地指导临床治疗,进而弥补和完善了经典生存分析方法应用之不足。随着医疗技术的不断提高,越来越多的临床肿瘤资料属于长期生存者资料。本文介绍的适用于长期生存者资料的统计分析模型适用性和实用性都很强,值得推荐。
[Abstract]:In cancer clinical trials, because of the overall heterogeneity of patients, some patients who are sensitive to treatment do not show any cancer symptoms or signs, so they are considered to be "long-term survivors (long term survivors)" or healing (cured individuals), That is, in sufficient follow-up time, such individuals do not have a specific end-point event, and usually have a longer censored survival time. Long-term survivors mean that they do not occur or never end-point events (death, relapse of disease, etc.) until the end of the study. In the analysis of this kind of data, we should first determine whether there are long-term survivors in the data. If there is no long-term survival in the data or there is no evidence to confirm the existence of long-term survival, the traditional survival analysis method is used. If there is evidence that there are long-term survivors in the data, and the follow-up is sufficient, the traditional analytical method is not appropriate because it treats the long-term survivors as amputation observation objects, which is obviously unreasonable. Blind use of traditional survival analysis may lead to unreasonable interpretation of the results and even to the opposite conclusion. In this paper, we introduce a statistical model for long-term survivor data-parametric hybrid model, including exponential mixed model and Weibull hybrid model, Burr XII hybrid model. These three models are widely used in the analysis of long-term survival data. We use the goodness-of-fit test to select the best model to fit the actual data, and use the corresponding model to make further analysis. The maximum likelihood method is used to estimate the parameters, and the software S-PLUS6.0 is used to realize the estimation. Through the analysis of examples, it is proved that if the traditional survival analysis method can not make good use of the information provided by the data, the conclusion is not comprehensive, even the opposite, if the traditional survival analysis method is used. If the parameter mixed model is used to analyze the long-term survival data, a comprehensive and correct conclusion can be obtained, and more information can be provided from different sides, which can guide the clinical treatment effectively and reasonably. Furthermore, it makes up for and consummates the deficiency of the classical survival analysis method. With the continuous improvement of medical technology, more and more clinical tumor data belong to long-term survival data. The statistical analysis model for long-term survival data introduced in this paper is of great applicability and practicability, and is worthy of recommendation.
【学位授予单位】：山西医科大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：R311

【相似文献】