Kattan Michael W, Heller Glenn, Brennan Murray F
Department of Urology, Memorial Sloan-Kettering Cancer Center, New York, USA.
Stat Med. 2003 Nov 30;22(22):3515-25. doi: 10.1002/sim.1574.
The majority of staging systems focus on the definition of stage, and, therefore, prediction of prognosis. In the current era of clinical trial research, it has become apparent that the clinical stage alone is not sufficient to assess patient risk of treatment failure. As the number of biological markers increases, our ability to partition the traditional disease classification system improves, and our ability to predict patient success continues to increase. One approach to quantifying individual patient risk is through the nomogram. Nomograms are graphical representations of statistical models, which provide the probability of treatment outcome based on patient-specific covariates. We will focus on the use of the nomogram when the response variable is time to failure and there are multiple, possibly dependent, competing causes of failure. In this setting, estimation of the failure probability through direct application of the Cox proportional hazards model provides the probability of failure (for example, death from cancer) assuming failure from a dependent competing cause will not occur. In many clinical settings this is an unrealistic assumption. The purpose of this study is to illustrate the use of the conditional cumulative incidence function for providing a patient-specific prediction of the probability of failure in the setting of competing risks. A competing risks nomogram is produced to estimate the probability of death due to sarcoma for patients who have already developed a local recurrence of their initially treated soft-tissue sarcoma.
大多数分期系统侧重于阶段的定义，因此也侧重于预后的预测。在当前的临床试验研究时代，很明显仅靠临床分期不足以评估患者治疗失败的风险。随着生物标志物数量的增加，我们划分传统疾病分类系统的能力得到提高，预测患者治疗成功的能力也在不断增强。一种量化个体患者风险的方法是通过列线图。列线图是统计模型的图形表示，它根据患者特定的协变量提供治疗结果的概率。当响应变量是至失败时间且存在多个可能相互依赖的竞争失败原因时，我们将重点关注列线图的使用。在这种情况下，直接应用Cox比例风险模型估计失败概率时，假设不会发生因依赖竞争原因导致的失败，从而提供失败（例如，死于癌症）的概率。在许多临床情况下，这是一个不现实的假设。本研究的目的是说明在竞争风险情况下，使用条件累积发病率函数为患者提供特定的失败概率预测。制作了一个竞争风险列线图，以估计已发生初始治疗的软组织肉瘤局部复发的患者因肉瘤死亡的概率。
