Missing Data Methods in HIV Clinical Trials: Regulatory Guidance And Alternative Approaches
- PDF / 609,302 Bytes
- 9 Pages / 504 x 720 pts Page_size
- 14 Downloads / 184 Views
0092-8615/2001 Copyright 0 2001 Drug Information Association Inc.
MISSING DATA METHODS IN HIV CLINICAL TRIALS: REGULATORY GUIDANCE AND ALTERNATIVE APPROACHES* THOMASKELLEHER,PHD Senior Research Biostatistician, Biostatistics and Data Management
THIRY,PHD ALEXANDRA Biostatistician, Biostatistics and Data Management
RICHARDWILBER,MD Executive Director, Infectious Diseases Clinical Research
ANNECROSS,PHD Director, Biostatistics and Data Management Pharmaceutical Research Institute, Bristol-Myers Squibb Company, Wallingford, Connecticut
Efficacy in H N clinical trials is measured by changes in HIV RNA levels over time as well as the proportion of subjects with HIV RNA levels below an assay’s threshold of reliable quantification at a single time point. Missing data arise naturally due to missed visits and premature discontinuations of treatment. The available data are then analyzed using repeated measures models and univariate comparisons of proportions, assuming missing data occur at random or considering missing values as treatment failures (worst case scenario). These and other methods recently proposed by regulatory authorities are presented along with alternative approaches. Advantages and disadvantages of each method are discussed. Data from a recent comparison of ‘standard-of-care’ triple combination regimens are used for illustration. Key Words: Missing data; HIV, Clinical trials; Regulatory guidance
INTRODUCTION THERAPY TRIALS in are at least 48 weeks in duration and utilize serial measurements of HIV RNA in
*Presented at the FDNlndustry Workshop “Statisticdly Sound Decision Making,” Bethesda, MD, September 14-15.2OOO. Reprint address: Thomas Kelleher, PhD. Biostatistics and Data Management, Dept. 703, Bristol-Myers Squibb Company, 5 Research Parkway, Wallingford, CT 06492-7660. E-mail: [email protected].
plasma as endpoints. Problematic subject compliance with the length of the trial and incomplete attendance at all scheduled visits lead to missing data. Placebo or ineffective regimens are ethically unacceptable in this life-theatening disease. Thus, these trials often employ a noninferiority design to show that the experimental treatment as part of a triple drug combination is at least as effective as the triple drug combination that is currently the standard of care. The optimal single substitution design for comparisons of multiple-dmg regimens is not always possible. Regardless
1363
Downloaded from dij.sagepub.com at NANYANG TECH UNIV LIBRARY on April 26, 2015
1364
T. Kelleher, A. Thiry, R. Wilber, and A. Cross
of regimen, the primary efficacy measure in This type of missing visit may carry unmeathese trials is the level of HTV RNA. The sured information regarding the HIV RNA intervals for measurements of HIV RNA de- measurements or treatment effects, and should pend on study design, but are usually every be considered informative. Other reasons for missing HIV RNA meaeight or 12 weeks during periods of longer follow-up, as is current practice in the treat- surements may also be consid
Data Loading...
