Chen Eric Bingshu, Cook Richard J
Department of Statistics and Actuarial Science, University of Waterloo, 200 University Avenue West Waterloo, Ontario, Canada N2L 3G1.
Lifetime Data Anal. 2003 Sep;9(3):275-91. doi: 10.1023/a:1025888820636.
In life history studies involving patients with chronic diseases it is often of interest to study the relationship between a marker process and a more clinically relevant response process. This interest may arise from a desire to gain a better understanding of the underlying pathophysiology, a need to evaluate the utility of the marker as a potential surrogate outcome, or a plan to conduct inferences based on joint models. We consider data from a trial of breast cancer patients with bone metastases. In this setting, the marker process is a point process which records the onset times and cumulative number of bone lesions which reflects the extent of metastatic bone involvement The response is also a point process, which records the times patients experience skeletal complications resulting from these bone lesions. Interest lies in assessing how the development of new bone lesions affects the incidence of skeletal complications. By considering the marker as an internal time-dependent covariate in the point process model for skeletal complications we develop and apply methods which allow one to express the association via regression. A complicating feature of this study is that new bone lesions are only detected upon periodic radiographic examination, which makes the marker processes subject to interval-censoring. A modified EM algorithm is used to deal with this incomplete data problem.
在涉及慢性病患者的生命历程研究中，研究标记过程与更具临床相关性的反应过程之间的关系通常很有意义。这种兴趣可能源于想要更好地理解潜在病理生理学的愿望、评估标记物作为潜在替代结局的效用的需求，或者基于联合模型进行推断的计划。我们考虑来自一项针对有骨转移的乳腺癌患者的试验的数据。在这种情况下，标记过程是一个点过程，它记录骨病变的发病时间和累积数量，反映转移性骨受累的程度。反应也是一个点过程，它记录患者经历由这些骨病变导致的骨骼并发症的时间。兴趣在于评估新骨病变的发展如何影响骨骼并发症的发生率。通过将标记物视为骨骼并发症点过程模型中随时间变化的内部协变量，我们开发并应用了一些方法，这些方法允许通过回归来表达这种关联。这项研究的一个复杂特征是，新骨病变仅在定期的影像学检查时才被检测到，这使得标记过程受到区间删失的影响。一种改进的期望最大化（EM）算法被用于处理这个不完全数据问题。
