Aline Fuchs (1), Mohammed Cherkaoui Rbati (1), Anne Kuemmel (2), Jeremy Burrows (1), Didier Leroy (1), Azrin N. Abd-Rahman (4), Jörg Möhrle (1,3), Nathalie Gobeau (1)
(1) Medicines for Malaria Venture, Geneva, Switzerland, (2) IntiQuan GmbH, Basel, Switzerland, (3) University of Basel, Basel, Switzerland, (4) QIMR Berghofer Medical Research Institute, Brisbane, Australia
Introduction:
Activity of antimalarial candidate compounds is first tested in severe combined immunodeficient (SCID) mice engrafted with human erythrocytes and infected with Plasmodium falciparum parasites, one of the five malaria parasites infecting humans but not mice. A range of doses are tested and the compound concentrations and parasitemia are measured in each mouse. A PKPD model is build and used, in combination with a prediction of human PK, to make an initial estimation of the efficacious dose in humans. This estimated dose is used to prioritize the compounds to be tested in humans. Before going into patients, the compound is administered at subtherapeutic doses to volunteers inoculated with P. falciparum infected red blood cells. From a PKPD analysis based on human data, the initially estimated dose is revised with the human PK model. Five compounds have now been tested in both mice and Volunteers Infected Studies (VIS) and a retrospective analysis to compare the PKPD models in mice and in humans has been initiated.
Objectives:
To assess if the PKPD relationship translates from SCID mice to humans
Methods:
For each compound and each study, in SCID mice and in volunteers, a population PKPD model was developed with a two-stage approach: first a population PK model was built; secondly, the individual PK parameters were used as regression parameters and the PD parameters were estimated. The PD model consisted of the balance between a parasite net growth rate and a drug killing rate. Systematically, four models linking the drug killing rate to the drug concentration were tested: a sigmoidal Emax model, an effect compartment model, a turnover model and a direct effect model model in which an additional clearance term accounts for the removal of dead parasites from the body (clearance model). This was introduced since it was believed that the mechanisms for clearing dead parasites were different between mice and humans. For this model, Emax was either fixed to the invitro value or estimated with SCID mice and VIS data. Of all models, the best model was selected based on model convergence, goodness-of-fit plots, reliability of parameters estimates and BIC.
The Emax values obtained in SCID and VIS were compared. Also, simulations of parasitemia profiles for a range of doses were undertaken with the VIS PKPD model on one hand; and with the VIS PK model combined with the SCID PD model where the net parasite growth rate and baseline parasitemia were replaced with the VIS values. The minimum dose needed to clear 6 log parasites, the criterion for selecting a discovery compound for clinical development, are compared for these two scenarios.
To test if the clearance model, aimed at accounting for possible differences in dead parasites clearing mechanisms between SCID and VIS, could improve the dose predicted, additional simulations were made where the Emax, EC50 and Hill values estimated in the SCID model were used in an Emax model combined with the VIS PK model – making the assumption that humans will clear the parasites very quickly. The minimum dose needed to clear 6 log parasites was estimated and compared with the dose from the VIS PKPD model.
Results:
The best PKPD model was never the same in SCID and in VIS. The Emax model was never the best model in SCID but was in VIS for 3 compounds out of 5. These differences in the structural model between SCID and VIS are believed to be due to the different nature of data collected in SCID mice experiments and VIS since the parasites targeted by the compounds in SCID and VIS are the same. Only the clearance of the dead parasites may be different across species.
The Emax parameters were within a 3-fold margin between SCID and human and within a 2-fold margin for 4 out of the 5 compounds. Doses predicted based on 6 log parasite clearance were within a 2-fold margin between SCID and human.
The clearance model did not improve the agreement between doses predicted to clear 6 log parasites.
Conclusions:
This ongoing analysis represents a unique opportunity to better understand and assess the translation from the preclinical model to human for antimalarial compounds. Human dose prediction assuming the PKPD relationship is the same in SCID and in humans shows reasonable predictions for all 5 compounds Further analysis is planned to investigate possible improvement
This work will help MMV to refine its methodology in predicting human dose to select the most promising compounds.
Reference: PAGE 28 (2019) Abstr 8998 [www.page-meeting.org/?abstract=8998]
Poster: Drug/Disease Modelling - Infection