A platform quantitative systems pharmacology (QSP) model for preclinical to clinical translation of ADCs and clinical evaluation of thrombocytopenia


  1. Predicting clinical ADC efficacy and toxicity is a challenge. The mAb, linker, and payload all need to be optimized for a particular tumor target, indication, and patient population.
  2. Two mechanistic models were developed using preclinical Trastuzumab-emtansine (T-DM1) data to project clinical efficacy and tumor reduction and thrombocytopenia (TCP) incidence, a common ADC adverse event.
  3. The models were built to be generalizable for the study of any ADC.
  4. Combining efficacy and toxicity models allows us to explore common therapeutic metrics, such as therapeutic window and indexes.


  • Clinical outcomes of ADCs can be projected with this mechanistic platform model.
    • The efficacy model was validated with T-DXd in mBC patients, which successfully reproduced clinical PK, efficacy, and PSF.
  • Mechanistic TCP model can capture interpatient variability.
  • The models were built to be generalizable to other ADCs by substituting relevant parameters.

Future directions:

  • Continue validating efficacy model on other ADCs with different indications.
  • Use TCP rates from clinical data with higher doses to improve predictions. Determine whether the mechanistic TCP model can be used for preclinical to clinical translation.




In Vitro: MCF7/neoHER2 cells:

Observed (symbols) and calibrated in vitro model (lines) DM1 disposition in MCF7 cell lines.

Mouse: TGI:

Observed and calibrated TGI model of N87 tumor bearing xenograph mice in response to Q4D x 4 T-DM1 doses.



Platelets from 10 patients treated with with the approved dose of T-DM1 in a phase I study compared to TCP model simulations calibrated to each individual.


Download the Case Study