Quantitative methods have played and will continue to play an important role in understanding AIDS epidemic and in evaluating therapies for HIV infected people. The primary objective of this proposal is to develop statistical methods and to provide new ideas and tools for design, monitoring and analysis of HIV/AIDS studies. Specifically, we will evaluate and develop statistical methods for analyzing HIV RNA data so that therapeutic studies involving the marker outcomes can be validly and efficiently evaluated. Like other laboratory-based outcomes, plasma HIV RNA measurements are subject to technical limitations in assays. When their values are below the assay's lower limit of detection or quantification, HIV RNA measures are censored. We will utilize surrogate/auxiliary information to supplement HIV RNA data in the analysis. Design issues on surrogate information collection will be discussed. Many recent HIV/AIDS clinical trials are designed to assess HIV RNA and CD4 over time. Sometimes a study objective is to evaluate how these markers change over time and how they co-vary with one another, or with another time dependent event. Our second specific aim is to provide robust statistical approaches to quantifying the time trend and the association over time from longitudinal data. We will develop methods that do not impose structural assumptions on the marker processes but take advantage of the correlation of the marker measures at different times. Informatively missing data will be addressed. Formal interim analyses of clinical trials based on pre-specified termination criteria (""""""""stopping rules"""""""") are now commonly used in clinical trials. Available therapies, biological knowledge and standard of care are rapidly evolving in HIV/AIDS study. It is often not clear how treatment effects and what kinds of effects are clinically meaningful and/or acceptable at the study design stage of some AIDS clinical trials. Our third specific aim is to continue our development of a methodology for formal interim analyses, the repeated confidence bands approach. The approach allows flexibility in design and monitoring process by not pre-specifying a stopping rule or even a metric on which inferences will be based on. With such confidence bands, we are able to examine the results of a trial during each interim analysis and, if it appears warranted, terminate the trial and provide a valid estimate.
|Zhao, Lihui; Hu, X Joan (2013) Estimation with Right-Censored Observations Under A Semi-Markov Model. Can J Stat 41:237-256|
|Zhao, Lihui; Hu, X Joan; Lagakos, Stephen W (2009) Statistical monitoring of clinical trials with multivariate response and/or multiple arms: a flexible approach. Biostatistics 10:310-23|
|Hu, X Joan; Schroeder, R Jason; Wang, Winfred C et al. (2007) Pseudoscore-based estimation from biased observations. Stat Med 26:2836-52|