Abstract

Multiregional clinical trials have a hierarchical data structure because several regions form a patient population and individual patients are nested within their own regions. Data are obtained from two different levels: regions and patients. To incorporate such a hierarchical structure, hierarchical linear models were proposed for the response variables following a normal distribution by Kim and Kang. In this article, we extend the hierarchical linear models to propose hierarchical generalized linear models (HGLMs) so that the response variables can follow the exponential family. We describe the details of the model when the response variable follows the Bernoulli distribution and the Poisson distribution. Simulation studies show that the empirical powers of the HGLM are greater than random effects model when region-level covariates are incorporated.

摘要

多区域临床试验具有分层数据结构，因为几个区域形成一个患者群体，而个体患者嵌套在他们自己的区域内。数据来自两个不同的层次：地区和患者。为了结合这种层次结构，Kim 和 Kang 提出了响应变量遵循正态分布的层次线性模型。在本文中，我们扩展了层次线性模型以提出层次广义线性模型（HGLM），以便响应变量可以遵循指数族。当响应变量遵循伯努利分布和泊松分布时，我们描述了模型的细节。