Statistical Methods--3D Conformal Radiation Therapy Data

描述（由申请人提供）：技术的进步使放射肿瘤学家能够使用3D适形放射治疗技术将更高剂量的放射输送到癌症的限定区域。对肿瘤的更高剂量的结果也可能是对关键正常组织和器官的更高剂量。然而，技术的进步在很大程度上先于对关键器官内的输送剂量分布与观察到的功能并发症之间的关系的适当定量理解。数学模型，无论是基于放射生物学现象或选择其数学上的方便，已被建议来描述的关系。虽然已经提出了一些模型，但很少有已发表的研究，其中这些模型已成功地拟合数据，使用现代统计方法，使有效地利用数据。本课题的研究融合了统计学、放射生物学和生物数学的概念。其目的是发展适当的生物数学模型来描述剂量分布和并发症之间的关系，并发展有效的统计估计方法。具体目标是：（一）建立剂量分布对器官损伤影响的数学模型;（二）建立有效的统计方法来估计这种关系;（三）将统计方法应用于头颈癌数据，并在模拟研究中对其进行评估。将开发和研究三种不同类型的模型，首先是放射生物学动机的剂量损伤损伤模型，然后将考虑一般的简约参数和半参数适应。估计方法将基于马尔可夫链蒙特卡罗方法。将制定评估拟合优度和表示结果的方法。该方法将被应用于重复测量唾液流量测量从腮腺与三维适形放射治疗头颈部癌症的患者。

DESCRIPTION (provided by applicant): The advancement in technology has enabled radiation oncologists to deliver higher doses of radiation to a defined region of cancer using 3D conformal radiation therapy techniques. A consequence of higher doses to the tumor can also be a higher dose to critical normal tissue and organs. However, the advancement in the technology has for the most part preceded a proper quantitative understanding of the relationship between the delivered dose distribution within the critical organs and the observed functional complication. Mathematical models, based on either radiobiological phenomena or chosen for their mathematical convenience, have been suggested to describe the relationship. Although a number of models have been suggested there have been few published studies in which these models have been successfully fit to data using modern statistical methods which make efficient use of the data. The research in this proposal is an integration of concepts in statistics, radiation biology and biomathematics. The aims are to develop appropriate biomathematical models to describe the relationship between the dose distribution and the complication and to develop efficient statistical estimation methods. The specific goals are (i) to develop mathematical models for the effect of the dose distribution on the injury to the organ; (ii) to develop efficient statistical methods to estimate the relationship and (iii) to apply the statistical methods to data from head and neck cancer and to evaluate them in simulation studies. Three different types of models will be developed and investigated, beginning with radiobiologically motivated dose-damage-injury models, then general parsimonious parametric and semi-parametric adaptations will be considered. Estimation methods will be based on Markov chain Monte Carlo methods. Methods of assessing goodness of fit and for representing the results will be developed. The methods will be applied to repeated measures saliva flow measurements from the parotid for patients treated with 3D conformal radiation therapy for head and neck cancer.