EXPERIMENTAL DESIGN AND DATA ANALYSIS METHODOLOGY FOR ANIMAL EXPERIMENTS

这个项目关注的是统计方法问题 实验动物实验的设计、分析和解释。 具体的研究工作包括：(1)解决假阳性问题 在长期的啮齿动物致癌性研究中，使用统计学方法 研究了对多重比较进行调整的方法学。一个 利用三种不同的统计决策规则进行实际案例研究 对25项NCI致癌性研究进行评估。这些人之间的协议 决策规则被证明比最初报告的更多。(2) 国家毒理学计划检测结果的评价 致癌性研究导致制定和出版了 肿瘤历史对照数据库的建立及潜能鉴定 肿瘤发生率的变异性来源。(3)不同的实验设计 对雄性兔生殖功能的评估进行了研究，以确定 他们是否有足够的力量检测到一种合适大小的有毒物质 效果。已经确定，使用预曝光期的研究 将是合适的。(4)开发了统计方法，用于 从疾病死亡率数据中提取有关疾病发病率的信息。 在某些简化的假设下，疾病死亡率可以是 以发病率和致死率表示。建议数 该方法适用于存在两个种群的情况， 一个有发病率和疾病致死率的数据，另一个 有关于疾病死亡率的数据。基于关联事件的模型 在两组人群中，第二人群的发病率可以 通过最大化仅涉及死亡率数据的可能性来估计。 未来的研究包括统计方法的比较 评估不需要确定死因的肿瘤数据和 评估饮食和笼养方案对变异性的影响 控制肿瘤发病率。

This project is concerned with statistical methodology issues in the design, analysis, and interpretation of laboratory animal experiments. Specific research efforts include (1) To address the false positive issue in long term rodent carcinogenicity studies, the use of statistical methodology that adjusts for multiple comparisons was investigated. An actual case study utilizing three different statistical decision rules to evaluate 25 NCI carcinogenicity studies was examined. Agreement among these decision rules was shown to be greater than originally reported. (2) Evaluation of results from the National Toxicology Program's carcinogenicity studies has resulted in the formulation and publication of an historical control tumor database and the identification of potential sources of variability in tumor rates. (3) Various experimental designs for assessing male reproductive function in rabbits were studied to determine whether they had sufficient power to detect a reasonably sized toxic effect. It was determined that studies which employed a preexposure period would be appropriate. (4) Statistical methods have been developed for extracting information on disease incidence from data on disease mortality. Under certain simplifying assumptions, the disease mortality rates can be expressed in terms of the incidence and lethality rates. The proposed approach is appropriate for the case in which there are two populations, one having data on disease incidence and disease lethality and the other having data on disease mortality. Based on a model linking the incidence rates in the two groups, the incidence rates in the second population can be estimated by maximizing a likelihood that involves only mortality data. Future research includes a comparison of statistical methodologies for evaluating tumor data that do not require cause of death determinations and assessing the effects of diet and caging protocols on variability in control tumor incidence.