Statistical Methods for Partially Controlled Studies

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): The objective of this project is to develop new statistical designs and methods to better evaluate partially controlled studies in public health. There are three specific aims. The first aim is to develop methods to evaluate treatments in studies that: do not directly control which subjects get which treatment, or which subjects provide outcomes; but do control the location of the sites that provide the treatments. Because existing methods, including standard instrumental variables, cannot handle the coexistence of the uncontrolled factors of taking treatment and of providing outcomes, the work under aim 1 is essential for proper treatment evaluation. This work will be applied to ongoing needle exchange programs in Baltimore, for preventing HIV transmission. The second aim is to develop new methods for clinical trials where treatment received is controlled, and outcomes are measured; a post-treatment surrogate endpoint is measured; and the goal is to estimate the treatments effects on the outcome that are concurrent and not concurrent with the surrogate. The methods under aim 2 use a new type of stratification for the uncontrolled surrogates that avoid the complications of standard stratifications on post-treatment variables. This work is essential for proper evaluation of early post-treatment symptoms as measures of later treatment effect, and we will apply it to cancer and ophthalmology trials. The third aim is to develop methods for study designs where: some persons drop out of a longitudinal follow-up, and where we can find (double-sample) a representative subset of the dropouts. This work is essential for preserving generalizability of results from long follow-up studies, and we will apply it to better evaluate time to potential drug dependence in youths.