Methods for Long-Term Follow-Up of HIV-Infected Patients

点击翻译按钮获取中文摘要

DESCRIPTION: (provided by applicant) We aim to develop methods to help in understanding the long-term consequences of treatment with antiretroviral drugs, for control of virus, for development of antiviral resistance, and for clinical outcome. Antiretroviral treatment choices have longer-term consequences lasting well beyond the period of treatment because of the development of resistant strains of virus under drug pressure and because of cumulative toxicities. Understanding the relationship between treatment history, onset of resistant mutations, and response to subsequent therapy can help in determining longer-term consequences of therapy and in making treatment choices. Sources of information such as rollover studies and observational databases exit. However such investigations require development of new statistical methods because of the high dimensionality of the genetic information; this dimensionality is increased when genetic sequences are obtained over time and because problems arising from dropout and selective willingness to participate in long-term studies/sequential randomizations must be addressed. Our specific aims are: to develop new methods for investigation of antiviral resistance, including: 1) Model-based methods to aid in drug selection by characterizing the relative performance of different antiviral drugs by genotype; and 2) Nonparametric methods for predicting HIV drug susceptibility phenotype or patient response to treatment from genotype. We also aim to develop new methods of estimation and testing that handle informative dropout, with focus on issues of special relevance to analyses of data from longer-term AIDS clinical studies and rollover studies. The specific areas are: 3) Analyzing the effects of sequences of treatments from rollover studies subject to dropout, self-selected treatment discontinuation, lack of consent to rollover, and selective accrual; and 4) Covariate-adjusted K-sample tests for censored repeated measures and time to event data with informative dropout.