Efficient and unbiased methods to utilize
HIV-RNA data from two different PCR assays
Chunpeng Fan
Dept of Statistics
UW Madison
Friday, January 20, 2006, 12:00pm
5235/75 MSC
| ABSTRACT |
Plasma HIV RNA is measured by PCR assays which usually have limits of reliable quantification (LoQ). For example, Amplicor Standard assay has a reliable range of 400-750000 copies/mL. If a value $<$ 400 copies/mL, blood sample may be re-tested by Ultrasensitive assay which has a lower LoQ of 50 copies/mL. For the calculation of change from baseline in log10 HIV RNA, values from Standard assay are used while Ultrasensitive assay results are typically ignored because of different variations. In the early stage of the HIV integrase program, important decisions (e.g., dose selection) will be made based upon limited data. Probability of success may be improved if this once ignored information can be utilized to aid the decisions.
We focus on change from baseline of plasma HIV RNA value in log10 scale. By comparing several approaches, we conclude that, 1) excluding substantial amount of data measured by the Ultrasensitive assay might waste precious information, and 2) naive approach to use such information might cause serious bias. We also proposed a new approach, which utilized information from Ultrasensitive assay to estimate corresponding Standard assay value when it was unobserved. Simulation results show that this new approach is efficient and unbiased beyond simplicity. Merck protocol 103/104 was used as an example to further compare our new approach with other traditional approaches.
Joint work with: Josh Chen, CBARDS (Clinical Biostatistics and Research Decision Sciences), Merck & Co. Inc., Blue Bell, PA |
Return to seminar list
