OBJECTIVE: The pharmacokinetic analysis of biologic compounds is frequently disturbed by the presence of endogenous levels, which cannot be discerned from exogenous levels. The frequently used method of subtracting baseline levels from subsequent measurements was compared to a fully adjusted regression model in a simulation study. METHODS: Simulations (5,000 each) were carried out for a standard one-compartment model with rich (n = 10) and poor (n = 6) postdose sampling, using unweighted as well as two-weighted types of non-linear regression. RESULTS: Whereas the fully adjusted model performed properly across various scenarios, the subtraction method showed a noteworthy bias (up to 14%) for area under the curve (AUC) and elimination half-life with weighted non-linear regression. For estimation of the Cmax parameter using any weighting scheme, and of any parameter using unweighted non-linear regression, the two methods performed equally well. As expected, poor in contrast to rich sampling resulted in larger coefficients of variation, but also in increasing failures (4.4%) of the regression algorithm (failure to converge, negative Cmax or half-life) for the subtraction method when it was combined with the weighting scheme giving highest weight to small concentrations. CONCLUSION: The risk of biased results may result from the subtraction method, which may also affect the analysis of dose linearity, bioequivalence and population kinetic studies with biologic compounds. When background endogenous levels are not negligible, a fully adjusted model is recommended.
OBJECTIVE: The pharmacokinetic analysis of biologic compounds is frequently disturbed by the presence of endogenous levels, which cannot be discerned from exogenous levels. The frequently used method of subtracting baseline levels from subsequent measurements was compared to a fully adjusted regression model in a simulation study. METHODS: Simulations (5,000 each) were carried out for a standard one-compartment model with rich (n = 10) and poor (n = 6) postdose sampling, using unweighted as well as two-weighted types of non-linear regression. RESULTS: Whereas the fully adjusted model performed properly across various scenarios, the subtraction method showed a noteworthy bias (up to 14%) for area under the curve (AUC) and elimination half-life with weighted non-linear regression. For estimation of the Cmax parameter using any weighting scheme, and of any parameter using unweighted non-linear regression, the two methods performed equally well. As expected, poor in contrast to rich sampling resulted in larger coefficients of variation, but also in increasing failures (4.4%) of the regression algorithm (failure to converge, negative Cmax or half-life) for the subtraction method when it was combined with the weighting scheme giving highest weight to small concentrations. CONCLUSION: The risk of biased results may result from the subtraction method, which may also affect the analysis of dose linearity, bioequivalence and population kinetic studies with biologic compounds. When background endogenous levels are not negligible, a fully adjusted model is recommended.
Authors: Donald J Irby; Mustafa E Ibrahim; Anees M Dauki; Mohamed A Badawi; Sílvia M Illamola; Mingqing Chen; Yuhuan Wang; Xiaoxi Liu; Mitch A Phelps; Diane R Mould Journal: CPT Pharmacometrics Syst Pharmacol Date: 2021-04