Inferring target occupancy from fitting nonlinear-PK data with mechanistic PK-RO model for Pembrolizumab


Pembrolizumab (Pembro), an anti-PD-1 antibody, has been approved for several cancer indications at 200 mg or 2 mg/kg IV every 3 weeks (Q3W). Understanding the percent target receptor occupancy (RO) for PD-1 at this approved dose can help with dose selection for anti-PD-1 combination studies and other anti-PD-1 drugs in development. However, RO can be challenging to determine through direct experimental measurements.


  1. A semi-mechanistic PKRO model was developed to fit Pembro phase 1 PK data.  The model was then used to predict RO.


  2. For the clinically approved dose of 2 mg/kg Q3W, the model predicted that the range of sPD-1 concentration has no significant impact on PD-1 RO in tumor.


  3. The impact of Kd and drug:PD-1 turnover rate on predicted RO was explored. When either of the parameters varied by 10-fold higher or lower from its nominal value, the model still predicted greater than 90% RO at 2 mg/kg dose.

Model Diagram of Pembrolizumab

The Model

The QSP model was based on first principles as a system of elementary mass-action, mechanistic PK/PD, ordinary differential equations. The model used three compartments to capture target-drug interactions in tumor, circulation (Central) and other organs/tissues (Peripheral), and captures overall membrane bound PD-1 expression and soluble PD-1 (sPD-1) in all three compartments.  PD-L1 is only expressed in the tumor compartment. Mechanisms captured by the model are 1) drug binding to PD-1 receptor and sPD-1, 2) PD-1 receptor turnover, 3) drug:PD-1 internalization and clearance, and 4) clearance of free drug and drug:sPD-1 complex.


  1. The model predicted PD-1 expression level in circulation to be approximately 10.8 pM with no soluble PD-1. In the presence of soluble PD-1, PD-1 expression is predicted to be higher to capture the non-linear PK.


  2. For 2 mg/kg IV Q3W dosing, Pembro > 98% RO was predicted in the tumor compartment with different sPD-1 concentrations tested.


  3. When Kd is varied by 10-fold higher or lower from its nominal value (29 pM), the model predicts > 95% PD-1 occupancy at steady state with 2 mg/kg Q3W. When the half-life of mAb-PD1 is varied by 10-fold higher or lower from its nominal value (2 hours), the model predicts > 90% PD-1 occupancy at steady state with 2 mg/kg Q3W.

Impact of sPD-1 on PD-1 RO in Tumor

Download the full PDF

Download Poster