Analysis and reporting of platelet kinetic studies

Authors


Larry J. Dumont, MBA, PhD Dartmouth-Hitchcock Medical Center, Department of Pathology, One Medical Center Drive, Lebanon, NH 03756; e-mail: Larry.J.Dumont@hitchcock.org.

Abstract

BACKGROUND: For platelet (PLT) kinetic studies in the healthy subject, autologous transfusion models are standardly used for the evaluation of new or modified PLT collection or treatment methods. The investigator is usually seeking to demonstrate that the new product is not inferior to a control condition. Consistent, statistically sound analysis and reporting methods will enable clear interpretation and communication of the results.

STUDY DESIGN AND METHODS: Noninferiority hypotheses are constructed in light of a recent FDA workshop. One- and two-stage analysis methods are described to evaluate the paired differences in control and test arms with calculation of upper confidence intervals. Sample size estimates are presented for various experimental design options. The sensitivity of sample size to the difference in test and control recoveries is performed.

RESULTS: The noninferiority hypothesis test was applied to a recently published study. Seven-day, plasma-stored apheresis PLTs were shown to be noninferior to freshly prepared PLTs with 95 percent confidence that the difference in recovery was not more than 11.4 percent and survival not more than 2.6 days, with maximum acceptable differences of 19.8 percent and 4.4 days. As the test recovery and survival approach the minimum acceptable value, the sample size required to demonstrate noninferiority increases nonlinearly. Preliminary sample size calculations presented here suggest that for most studies such as this, a minimum sample size of 7 should be anticipated. Because most of these studies are used to support regulatory submissions, this is usually executed in at least two centers for a total minimum of 14.

CONCLUSION: The randomized, paired design is the most efficient approach to these studies. Consistent reporting of summary statistics for each test arm and the paired differences, the upper confidence limit for the differences in recovery and survival, and the maximum acceptable difference for each is recommended.

Ancillary