Hall of Mirrors
Assessing Sensitivity to Measurement Error in a Bayesian Pharmacokinetic Model (304979)
*Hannah L Weeks, Vanderbilt UniversityMatthew Steven Shotwell, Vanderbilt University
Keywords: Bayesian, simulation, pharmacokinetics, pharmacodynamics
Pharmacokinetic analyses rely on an assumption that information, particularly time stamps associated with infusions and pharmacokinetic blood draws, is correctly recorded. Pharmacokinetic models are important clinical decision tools used estimate pharmacodynamic target attainment, which indicates whether a patient is receiving the appropriate amount of a given drug. Model results and estimates of patient response to a drug may be susceptible to errors if times are reported inaccurately. Using a Bayesian two compartment pharmacokinetic model, we conduct a simulation study to determine how sensitive estimates of individual pharmacokinetic parameters and pharmacodynamic target attainment are to errors in these recorded times. In particular, we investigate errors in the infusion start times and blood draw (concentration measurement) times. The effect of errors in infusion times is also being assessed, in addition to the impact of short versus extended dosing regimens. By quantifying the magnitude of this problem, we can apply statistical methodology to counter the systematic bias that results from time stamp recording errors.