Quantitative prediction of human pharmacokinetics for monoclonal antibodies: retrospective analysis of monkey as a single species for first-in-human prediction

Abstract

Background and objectives: Prediction of human pharmacokinetics for monoclonal antibodies (mAbs) plays an important role for first-in-human (FIH) dose selection. This retrospective analysis compares observed FIH pharmacokinetic data for 16 mAbs to those predicted in humans based on allometric scaling of Cynomolgus monkey pharmacokinetic data.

Methods: Ten mAbs exhibited linear pharmacokinetics in monkeys based on non-compartmental analysis. For these, simple allometric scaling based on bodyweight was applied to predict human clearance (CL) and volume of distribution (V(d)) from those obtained in monkeys. Six mAbs exhibited nonlinear pharmacokinetics in monkeys based on population modelling. For these, a population modelling approach using nonlinear mixed-effects modelling software, NONMEM, was applied to describe monkey data by a two-compartment pharmacokinetic model with parallel linear and nonlinear elimination from the central compartment. The pharmacokinetic parameters in monkeys were then scaled to humans based on simple allometry. Human concentration-time profiles of these mAbs were then simulated and compared with those observed in the FIH studies.

Results: Antibodies with linear elimination in monkeys also exhibited linear elimination in humans. For these, observed CL and V(d) were predicted within 2.3-fold by allometry. The predictability of human peak serum concentration (C(max)) and area under the serum concentration-time curve (AUC) for mAbs with nonlinear pharmacokinetics in monkeys was, however, concentration dependent. C(max) was consistently overestimated (up to 5.3-fold higher) when below the predicted Michaelis-Menten constant (Km; range 0.3-4 μg/mL). The prediction of human C(max) was within 2.3-fold when concentrations greatly exceeded Km. Similarly, differences between predicted human AUCs and those observed in the FIH studies were much greater at low doses/concentrations. Consequently, predicted drug exposure in humans at low starting doses (range 0.01-0.3 mg/kg) in FIH studies was poorly estimated for three of six mAbs with nonlinear pharmacokinetics.

Conclusions: Allometric prediction of human pharmacokinetics may be sufficient for mAbs that exhibit linear pharmacokinetics. For mAbs that exhibited nonlinear pharmacokinetics, the best predictive performance was obtained after doses that achieved target-saturating concentrations.