MECHANISTIC MODELING OF BREAST CANCER SURVEILLANCE

这个项目的总体目标是考虑肿瘤发展和检测的机制模型在以人群为基础的乳腺癌筛查的影响分析中的效用。我们提出的癌症筛查的随机模型具有以下明显的优势：它提供了一个简单但仍然真实的癌症潜伏期描述；2. 它可以在保持其基本结构的同时以各种方式进行推广；3. 它为数据分析提供了生物学上有意义的解释；4. 它容纳了标准的基于人口的统计数据；它的实施并不严重依赖于筛选试验产生的数据的可用性；5. 采用严格的统计方法估计模型参数；6. 它可以用于设计最佳的癌症筛查和监测策略。该模型将通过犹他州人口数据库和犹他州癌症登记处的乳腺癌数据进行验证。利用这些资源，我们将获得相关估计算法的初始参数值，该算法是为国家卫生统计中心提供的乳腺癌死亡率分组数据设计的。这两步估计程序将通过计算机模拟和流行病学数据分析进行检验。此外，我们将探讨随机逼近技术在微模拟框架内估计模型参数的效用。

The general objective of this project is to consider the utility of mechanistic models of tumor development and detection in analysis of the impact of breast cancer screening in population- based settings. A stochastic model of cancer screening we propose offers the following distinct advantages: 1. It provides a simple but still realistic description of cancer latency; 2. It can be generalized in various ways while retaining its basic structure; 3. It furnishes a biologically meaningful interpretation of data analyses; 4. It accommodates standard population-based statistical data; its implementation does not depend heavily on availability of the data yielded by screening trials; 5. Rigorous statistical methods are available for estimating model parameters; 6. It can be used for designing optimal strategies of cancer screening and surveillance. The model will be validated with data on breast cancer from the Utah Population Data Base and the Utah Cancer Registry. Using these resources we will obtain initial parameter values for a pertinent estimation algorithm designed for grouped data on breast cancer mortality provided by the National Center for Health Statistics. This two-step estimation procedure will be tested by computer simulations and analyses of epidemiological data. In addition, we will explore the utility of stochastic approximation techniques in estimation of model parameters within the microsimulation framework.