Statistical methods for assessing stability of compounds in whole blood for clinical bioanalysis

The objective of a blood stability study is to assess the stability of a compound in whole blood after storage in collection vessels at room temperature for two to four hours for samples drawn in the clinic, and for six to eight hours for samples drawn elsewhere and transported to the clinic, in order to develop proper sample handling recommendations for clinical studies. Common practice is to transfer blood from a spiked whole blood pool into collection vessels at 24, 3, and 0 hours prior to analysis. A loss of 15% for a specific proportion of replicates, or for the observed mean, compared to the 0 hour control mean, is often used to decide whether there is any clinically relevant instability at 3 and 24 hours. This procedure is easy to implement, and meets Food and Drug Administration requirements, but it has unknown and uncontrolled risks of incorrect decisions (failing to detect a truly unstable compound, or falsely concluding instability). Investigations were conducted to evaluate various data analysis and decision procedures that control these risks of incorrect decisions, and to optimize the study design and sample size. The recommended data analysis procedure is a linear regression analysis, with an equivalence test for comparing predicted mean concentrations at time = x hours versus the zero hour control, with 15% specification limits. A study design with n = 9 replicates at 0, 3, 6, and 24 hours yields sufficient power for demonstrating stability.