Elisa A.M. Calvier (1), Gilles Tuffal (1) , Christine Farenc (1) , François Donat (1) , David Fabre (1)
(1) Pharmacokinetics-Dynamics and Metabolism (PKDM), Translational Medicine and Early Development, Sanofi R&D, Montpellier, France
Objectives: Prediction of active dose range for phase II trials is often challenging due to differences in PK and/or PD between healthy volunteers (HV) and patients. SAR441344 is an antagonist of CD40L which is a central regulator of immune response. Relationship between PK and PD response could not be mechanistically extrapolated to multiple sclerosis (MS) and Sjögren’s syndrome (SjS) patients for the planning of doses for phase II clinical studies. Free target reduction estimated from total soluble CD40L (sCD40L) would guide for such extrapolation, but this data was not collected.
Therefore, the aims of this study were to predict free target levels:
1) in MS and SjS patients for SAR441344 using a model-based approach allowing for between-population extrapolation (HV to patients) and for between-drug extrapolation of the target kinetics using data from competitors sharing the same target.
2) and corresponding dose regimen that would lead to SAR441344 efficacy in MS and SjS patients.
Methods: Model development was performed in NONMEM 7.4.1 [1] with FOCE algorithm, using improvements in likelihood, and diagnostic plots as selection criteria.
A full Target Mediated Drug Disposition (TMDD) model structure for between-drug and between-population extrapolations was developed. This model consisted in 3 building blocks:
1) a drug-independent block: sCD40L kinetics was derived by fitting a full TMDD model to published data of VIB4920 [2], assuming that sCD40L can be used as a surrogate of the CD40L at the membrane level. Free VIB4920 and total (free and bound) sCD40L mean concentrations for each dose group (i.e., 3, 10, 30, 100, 300, 1000, or 3000 mg) from a Phase I study in HV were simultaneously fitted. This block was used as such for predictions in HV and, extrapolated to patients by changing the baseline sCD40L concentration (R0) to a literature value of 358 pM [3,4].
2) a drug-dependent block: the model describing the drug kinetics. VIB4920 PK model was replaced by an available preliminary PK model for SAR441344 and by published PK models for the competitors (i.e., VIB4920 and IDEC131). All PK models were dose linear.
3) a drug-target-dependent block composed of the association and dissociation rate constants kon and koff. Kon was extrapolated between drugs using in vitro measured KD, and a fixed koff of 11 day-1 (i.e., SAR441344 value).
TMDD block combinations were used to predict free target levels for the lowest doses leading to a pharmacodynamic response. These combinations were namely SAR441344 in HV, VIB4920 in HV and rheumatoid arthritis patients (RA). PD response was defined as TDAR ≥75% in HV, as DAS 28-CRP (+) and RF auto-Ab = -50% in RA patients. For IDEC131 in Systemic Lupus Erythematosus patients no clinical efficacy was observed, therefore a TMDD block combination was used to predict free sCD40L levels for the highest dose of the trial which were used as control. Based on these results, a free sCD40L level to achieve a PD effect was derived and used to define the minimum putative active dose of SAR441344 to administer in MS and SjS patients for the upcoming phase II clinical studies.
Results: VIB4920 model in HV was a full TMDD model with peripheral drug distribution and a proportional error of 31.8% and 25.7% for the drug and total target respectively. No inter-dose group variability was included in the model. KD, R0, target and drug-target complex degradation rates estimates were 305 pM, 21.8 pM, 0.931 day-1 and 0.0121 day-1 respectively. Sensitivity analysis for koff did not show any significant impact on the parameter estimates nor on model predictions of any of the TMDD block combinations.
Receptor occupancy was predicted to be 99% or higher even for doses leading to no PD effect. Free sCD40L predicted steady-state troughs following SAR441344 or VIB4920 administration in HV were 0.0636 or 1.18 pM respectively. Following VIB4920 administration in RA patients, a value of 1.3 pM was predicted. For IDEC131, a free sCD40L trough concentration of 1.51 pM was predicted for the highest dose (which did not lead to efficacy). The lower sCD40L trough predicted for SAR441344 in HV was likely due the absence of data for lower efficacious doses. Based on all results, the free sCD40L concentration leading to pharmacodynamic effect was assumed to be 1pM or lower to guide for dose selection.
Conclusions: This work enabled to fill in knowledge gaps to predict free target reduction by using competitor’s data to support dose selection for phase II clinical trials.
References:
[1] Beal S., Sheiner L.B., Boeckmann A., & Bauer R.J., NONMEM User’s Guides. (1989-2017), Icon Development Solutions, Ellicott City, MD, USA, 2017.
[2] Karnell et al. Sci Transl Med. 2019;11(489): eaar6584
[3] Ferro et al. Arthritis & rheumatism. 2004;50(5):1693–1694
[4] Kato et al. Clin Invest. 1999;104(7):947-955
Reference: PAGE 29 (2021) Abstr 9623 [www.page-meeting.org/?abstract=9623]
Poster: Drug/Disease Modelling - Other Topics