McArdle John J, Small Brent J, Bäckman Lars, Fratiglioni Laura
Department of Psychology, University of Southern California, Los Angeles 90089, USA.
J Geriatr Psychiatry Neurol. 2005 Dec;18(4):234-41. doi: 10.1177/0891988705281879.
This article explores new statistical methodologies for using longitudinal data in the early prediction of Alzheimer's disease (AD). Specifically, the authors examine some new techniques that allow the joint or "shared" estimation of longitudinal components based on both duration (survival) and quantitative changes (growth curves). These new shared growth-survival parameter models may be used to characterize the declining functions that anticipate the onset of AD. The authors apply these models to data from the Kungsholmen Project, a longitudinal study of aging in Stockholm, Sweden. They examine age-based survival-frailty models for the onset of AD, latent growth-decline curve models for changes in cognition over age, and 3 alternative forms of models for the shared relationships of survival and early cognitive decline. The accuracy and reliability of this approach is considered for a better understanding of the developmental course of AD in these data, including the potential removal of biases due to subject selection.
本文探讨了在阿尔茨海默病（AD）早期预测中使用纵向数据的新统计方法。具体而言，作者研究了一些新技术，这些技术允许基于持续时间（生存）和定量变化（生长曲线）对纵向成分进行联合或“共享”估计。这些新的共享生长 - 生存参数模型可用于表征预期AD发病的功能衰退。作者将这些模型应用于来自瑞典斯德哥尔摩一项关于衰老的纵向研究—— Kungsholmen项目的数据。他们研究了基于年龄的AD发病生存 - 脆弱性模型、认知随年龄变化的潜在生长 - 衰退曲线模型，以及生存与早期认知衰退共享关系的3种替代形式模型。为了更好地理解这些数据中AD的发展过程，包括潜在消除因受试者选择导致的偏差，考虑了这种方法的准确性和可靠性。
