Evaluation of Methods for Estimating Haplotype Effects in Survival Studies

描述(由申请人提供)：随着对在疾病病因和遗传与环境因素之间的相互作用中起作用的遗传位点的识别的高度兴趣，在病例对照和队列研究中对单倍型的相对风险的估计已经投入了极大的关注。在遗传流行病学研究中，来自队列设计的经审查的生存数据通常被分析。在这种情况下，通常使用Cox比例风险回归模型来估计遗传因素与某些感兴趣事件的发生时间之间的关联。虽然半参数最大似然估计器已经被开发用于未观察到的单倍型的建模，但这种方法计算复杂，目前在标准统计软件包中不可用。因此，如果能够使用标准软件来估计和得出单倍型关联的有效推论，遗传流行病学家将非常感兴趣。此外，由于大多数回溯性遗传队列研究首先研究样本病例和相关家族的对照，因此，当特定的基因类型过度抽样和存在等级聚类时，方法理想地允许进行有效的估计和推断。这项研究的目的是在删失生存数据的背景下提供单倍型相对风险的简单估计，并通过广泛的模拟研究来验证这些方法。此外，我们将考虑等级聚类下标准误差的估计，以及在特定子组已被过抽样的队列研究中纳入抽样权重的逆概率的影响。

DESCRIPTION (provided by applicant): With heightened interest in the identification of genetic loci that play a role in the causation of disease and interactions between genetic and environmental factors, a great deal of attention has been devoted to the estimation of relative risks for haplotypes in both case-control and cohort studies. Censored survival data stemming from cohort designs are commonly analyzed in genetic epidemiology studies. In such cases, the Cox proportional hazards regression model is typically employed to estimate the association between genetic factors and the time to some event of interest. Although semi-parametric maximum likelihood estimators have been developed for modeling of unobserved haplotypes, such methods are computational complex and are not currently available in standard statistical software packages. As such it would be of great interest to genetic epidemiologists if standard software could be used to estimate and draw valid inference for haplotype associations. In addition, because most retrospective genetic cohort studies sample cases first and controls from the relevant family, methods would ideally allow for valid estimation and inference when particular genotypes are over-sampled and hierarchical clustering exists. The goal of the research proposed here is to provide simple estimates of haplotype relative risks in the setting of censored survival data and to validate such methods via extensive simulation studies. In addition, we will consider the estimation of standard errors under hierarchical clustering and the impact of incorporating inverse probability of sampling weights for cohort studies where particular subgroups have been oversampled.