Thomas Frank, Wolfgang Schmider, Ashley Strougo, Andreas Kovar
Sanofi
Objectives: Insulin aspart Sanofi (IAS) is a biosimilar of Novo Nordisk’s rapid acting insulin aspart, approved as NovoLog® in the US and NovoRapid® in the EU. In a phase 1 study (ClinicalTrials.gov Identifier: NCT03202875), the primary pharmacokinetic (PK) parameters demonstrated bioequivalence (confidence intervals in the range of 0.8 to 1.25 between IAS, NovoRapid, and NovoLog). Likewise, the pharmacodynamic (PD) variables GIR-AUC0-12 and GIRmax (GIR = glucose infusion rate, AUC = area under the curve) were similar for all 3 insulin formulations. However, in the late phase (4 to 12 hours) of the PK and PD profiles, partial AUC 4-12 hours was significantly lower for IAS than for NovoRapid. Differences in the late phase could indicate differences in the absorption process due to flip-flop kinetics (i.e., elimination rate is faster than absorption rate). A similar absorption rate is essential to demonstrate bioequivalence between products. The main objective of the analysis was to use modeling and simulation to get a better understanding of potential differences in the absorption rates between IAS (test product) and NovoRapid/NovoLog (reference products) and their impact on the GIR.
Methods: To describe the insulin aspart concentrations, two empirical PK/PD models with similar structure were developed i.e., one for the test product (IAS) and one for the reference product. The two reference products (NovoRapid and NovoLog) were considered as one. The PK model that best described the time course of insulin aspart concentrations was a one-compartment model with first-order absorption and a transit compartment. The PK/PD model consisted of an effect compartment model linked to the central compartment for insulin aspart disposition in plasma as previously reported [1, 2]. Simulations considering the uncertainty in the population parameter estimates were performed to evaluate potential differences in insulin aspart PK (AUC) and PD effect (GIR-AUC) between the products.
Results: The PK and PK/PD models adequately described the concentration versus time profiles of IAS and NovoRapid/NovoLog as well as the relationship between insulin aspart and GIR as evaluated by likelihood of data, accuracy of parameter estimates (standard error), goodness of fit plots, and visual predictive check. Partially, model parameters were different among the products (ke, Vc, EC50). Simulations considering only population PK/PD parameter estimates and their uncertainties showed the insulin aspart and GIR time profiles to almost fully overlap. The simulations showed slight differences in the typical (median) values describing the late phases of the PK (AUC4-12) and PD (GIR-AUC4-12) profiles. However, the 90% confidence intervals formed by the uncertainty of the typical values were fully overlapping.
Conclusions: Models were successfully developed to describe the PK of the insulins and their PK/PD (GIR) relationship. The simulations clearly illustrated that neither the profiles nor the derived AUCs allow to conclude that they are different. Therefore, the differences observed in the study are most likely due to chance. This evaluation was also accepted by the regulators.
References:
[1] Landersdorfer CB, Jusko WJ. Pharmacokinetic/pharmacodynamic modeling of glucose clamp effects of inhaled and subcutaneous insulin in healthy volunteers and diabetic patients. Drug Metab. Pharmacokinet. 2010;25 (5):418–29.
[2] Rueppel D, Dahmen R, Boss A, et al. A Population Dose-Response Model for Inhaled Technosphere Insulin Administered to Healthy Subjects. CPT Pharmacometrics Syst Pharmacol 2017;6 (6):365–72.
Reference: PAGE 30 (2022) Abstr 9949 [www.page-meeting.org/?abstract=9949]
Poster: Drug/Disease Modelling - Other Topics