Evaluation of effect of below-limit of quantitation (BLQ) data due to nonadherence in population pharmacokinetic (POPPK) modeling

Abstract

Purpose: Data below lower limit of quantitation of analysis method is generally handled using likelihood-based approaches (M3 method) in POPPK. Medication nonadherence can cause BLQ data. Objective of current work is to evaluate the efficiency of M3 method on handling BLQ data generated through non-adherence. Methods: Stochastic simulation and estimation method was used. A two-compartment model with first order absorption was used to simulate a clinical trial with plasma samples collected at steady-state. Non-adherence was simulated with a priori fixed probability in dose omissions resulting in 100%, 70% and 50% of doses taken before plasma sample collection. The simulated dataset was used for estimating parameters with full adherence assumption and M3 method using NONMEM (version 7.4). Process was repeated 200 times for each non-adherence scenario. Parameter bias and precision were calculated for both fixed and random effects parameters using mean estimation error and root mean square error, respectively. Results: The %BLQ data in 100%, 70% and 50% nonadherence scenarios were 10%, 22% and 34%, respectively. The mean estimation error for fixed effect parameters were within the acceptable limits of bias in 100% adherence scenario except Vp showed mild bias (6%). The bias was >5% in fixed effect parameters with 70% and 50% adherence scenarios and increased with level of nonadherence. Between subject variability parameters showed bias >10% in all scenarios. Conclusion: Non-adherence induced BLQ data caused parameter bias although M3 method was used. Further research is warranted to find solutions to this problem.