Aims: To develop a software tool with the ability to rapidly compare Non-Compartmental Analysis (NCA) and Non-Linear Mixed-Effect Modelling (NLMEM) methods in determining bioequivalence of two-compartment kinetic drugs and to elucidate the impact of changes in (i) random unexplained variability (RUV) at the lower limit of quantification (LLOQ), (ii) the extent of censoring data below the LLOQ and (iii) the concentration sampling times.
Methods: A tool using both R and NONMEM was developed to allow the comparison of NCA and NLMEM in multiple scenarios. The R code generated sample data, ran the NCA analysis and called NONMEM to run NLMEM on the same data. Each scenario tested included 500 bioequivalence studies, comprising of 24 subjects each. Each study had concentration-time profiles simulated where relative bioavailability (Frel), RUV at the LLOQ and LLOQ differed between scenarios. NLMEM analyses employed the M1 and M3 methods for dealing with values below the LLOQ and two methods for determining Frel: using post-hoc estimates of Frel directly or estimated from post-hoc calculations of AUC from model parameters. Finally the test/reference ratios were used to determine bioequivalence.
Results: The tool allowed for NCA and NLMEM to be run sequentially for 105 scenarios. NLMEM showed a consistent 10-20% higher accuracy and sensitivity in identifying bioequivalent drugs when compared to NCA, while NCA was found to have a higher specificity than NLMEM (1-10% higher). Increasing the LLOQ resulted in censoring of larger amounts of data and a decrease in the accuracy and sensitivity of NCA by approximately 20%, but had minimal effects on NLMEM.
Conclusion: A highly robust and automated tool has been developed that provides a platform for comparing NCA and NLMEM methods and can be used to extend beyond the scenarios evaluated here. In the situations examined it is seen that NLMEM is more robust than NCA and may have a role to play in the determination of bioequivalence, despite its lower specificity.