Although EMA allows a dose correction in the bioequivalence guideline (for chemically-derived products) if the difference exceeds 5?%, the NCA assumes linearity in its correction, which is not appropriate for MAbs, that display nonlinear pharmacokinetics. identified as statistically significant covariate on any parameter in the combined model, and the addition of drug product as covariate Metoclopramide HCl did not improve the model fit. A similar structural model described both the test and reference data best. Only minor differences were found between the estimated parameters from these separate models. Conclusions PPK can also be used to support a biosimilarity claim for a MAb. However, in Metoclopramide HCl contrast to the standard non-compartmental analysis, there is less experience with a PPK approach. Here, we describe two methods of how PPK can be incorporated in biosimilarity testing for complex therapeutics. Electronic supplementary material The online version of this article (doi:10.1007/s00228-016-2101-6) contains supplementary material, which is available to authorized users. (predicted) individual concentrations at the original sampling times. AUC from administration (time 0) to the time of the last concentration? ?LLOQ (AUClast) was calculated using the linear trapezoidal method. AUC extrapolated to infinity (AUCinf) based on the apparent terminal elimination rate constant was calculated as well. Rabbit Polyclonal to MLH3 Biosimilarity statistics were performed on AUCinf or AUClast of all participants who were exposed to 6?mg/kg, comparing T to R in an unpaired test, using the software package R. AUCs were natural log (ln)-transformed prior to statistical analysis. The estimated difference in means and the corresponding 90?% confidence interval (CI) were back-transformed to obtain the relative geometric mean ratio (GMR) of T over R (T/R). These results were then compared to those calculated in a standard NCA. To correct for the difference between actual (5.96 and 6.44?mg/kg) and labelled dose (6?mg/kg), a linear normalisation to 6?mg/kg was applied to the individual AUCs in the NCA. In the PPK, individual profiles were simulated with the actual and labelled dose. Both corrected and uncorrected AUCs were calculated and statistically compared. Results Population Pharmacokinetic data were gathered from 110 healthy male volunteers, whose demographics are presented in Table ?Table1.1. In total, 1247 serum trastuzumab concentrations were available for the test product (T), of which 143 were LLOQ (64 pre-dose). In the 6?mg/kg test group, 60/906 observations were LLOQ (46 pre-dose) and for the reference product (Herceptin?), 51/912 observations (44 pre-dose). Table 1 Demographics lean body, body surface area, HER2 extracellular domain Model development First step: combined model Initial exploration of the data suggested that a two- or three-compartment model would describe the data best. Based on the observed non-linear kinetics, Michaelis-Menten kinetics was incorporated, described in terms of maximum rate of elimination (is the concentration which produces half of the is the concentration. V1, V2 and V3 are Metoclopramide HCl the distribution volumes; Q1 and Q2 are the inter-compartmental clearances to the peripheral compartments After identification of the structural model, individual estimates of random effects for between-subject variability were identified for the parameters V1, and and in the model. Significant correlations were found between lean body weight (LBW), body weight (WT), body surface area (BSA), height (HT) and body mass index (BMI) vs. V1, with correlation coefficients of 0.61, 0.55, 0.60, 0.54 and 0.28, respectively. Linear regression analysis of LBW vs. BSA resulted in a coefficient of 1 1 and for LBW vs. WT in 0.96. Furthermore, significant correlation coefficients were observed between BMI and (0.60), between serum concentrations HER2 ECD and (0.29), and between serum concentrations HER2 ECD and (0.18). Implementing LBW as a linear covariate on V1 (Online Resource Eq. 1) significantly improved the objection function value (OFV) and was added to the model. Incorporating other weight-related covariates (WT, HT and BMI) separately in the model did not result in a significant improvement compared to LBW; accordingly, they were not implemented in the model. Covariate analyses identified BMI as the one most significantly correlated to.