The goal of this research is the development of new statistical methods for several problems that rise in AIDS clinical trials, focusing on three specific areas involving the analysis of repeated measurement data. Pharmacokinetic studies are carried out in all phases of AIDs clinical trials to provide guidance on effective administration of drugs for the HIV-infected population. Characterization of individual and population pharmacokinetics requires accurate recording of dose and sampling times and precise measurement of subject attributes such as organ function and disease status; however, often in these trials, such covariates are measured with non-negligible error. The usual framework for analysis, the nonlinear mixed effects model, assumes covariates are error-free. The effects of measurement error on inference are not understood, and methods to take such error into account are not available. We will study the consequences of covariate measurement error in this context and develop statistical methods to address the problem. Analysis of immunological, virological, and growth measures collected over time is often carried out in a linear mixed model framework; however, measures such as CD4 are widely variable within a given subject, thus, the usual assumption of normality of transformed CD4 may be inappropriate, and conclusions may be sensitive to intra-subject response outliers. Moreover, the standard normal model for the random effect distribution characterizing the population may be unrealistic and limiting. We propose methods that allow nonlinearity of the response model, possibly incorporating transformations, that are robust to individual outlying observations, and that permit more flexible modeling of the population. Procedures to carry out such robust inference in the presence of informative censoring will be developed. A continuing challenge in AIDS research is the development of new assays for quantification of markers of disease progression and drug concentrations. Although measurements determined by such assays often constitute the data for longitudinal or pharmacokinetic analysis, little attention has been paid by statisticians involved in clinical trials to the statistical issues in assay development. We will develop statistical methods for accurate characterization of the precision of calibration.
| Yeap, B Y; Davidian, M (2001) Robust two-stage estimation in hierarchical nonlinear models. Biometrics 57:266-72 |
| Ko, H; Davidian, M (2000) Correcting for measurement error in individual-level covariates in nonlinear mixed effects models. Biometrics 56:368-75 |
| Oberg, A; Davidian, M (2000) Estimating data transformations in nonlinear mixed effects models. Biometrics 56:65-72 |
| Hu, P; Tsiatis, A A; Davidian, M (1998) Estimating the parameters in the Cox model when covariate variables are measured with error. Biometrics 54:1407-19 |
| Higgins, K M; Davidian, M; Chew, G et al. (1998) The effect of serial dilution error on calibration inference in immunoassay. Biometrics 54:19-32 |
| Zeng, Q; Davidian, M (1997) Calibration inference based on multiple runs of an immunoassay. Biometrics 53:1304-17 |
