(W-108) Incorporating High Dimensional Gut Microbiome Data into Population Pharmacokinetic Modeling of Mycophenolate Mofetil

Mahmoud Al-Kofahi, PhD – Adjunct Professor, Experimental and Clinical Pharmacology, University of Minnesota; Shen Cheng, PhD – Assistant Professor, Experimental and Clinical Pharmacology, University of Minnesota; Christopher Staley, PhD – Assistant Professor, Department of Surgery, University of Minnesota; Pamala Jacobson, PharmD – Distinguished Professor, Experimental and Clinical Pharmacology, University of Minnesota

Objectives: Mycophenolate mofetil (MMF) is a common immunosuppressant in transplantation. Mycophenolic acid (MPA), the active metabolite, undergoes enterohepatic circulation (EHC) characterized by secondary peak(s). The microbiome affects MPA exposure by altering EHC through deglucuronidation of mycophenolic acid glucuronide (MPAG), an inactive but abundant metabolite[1]. While the high dimensionality of microbiome data presents a challenge for incorporating it into population pharmacokinetic (PopPK) modeling, a semi-mechanistic PopPK model involving EHC was developed incorporating microbiome communities as covariates. This study aims to utilize simulation to evaluate the impact of the gut microbiome on the MPA exposures, and to evaluate the contribution of the gallbladder (GB) and EHC to the MPA exposure.

Methods: PK sampling of MPA and its glucuronide metabolites (MPAG and ACMPAG) from IV MMF dosing, stool microbiome and clinical data were analyzed in 20 hematopoietic cell transplant (HCT) recipients [1]. A PopPK model was developed using NONMEM v7.5. The microbiome data was reduced from 241 to 64 species (spp) by excluding spp. present in < 20% of individuals. A correlation network was created using SparCC and the Louvain algorithm identified microbiome communities [2,3]. Microbiome covariates were evaluated using a full fixed effects covariate modeling approach. Model-based simulations were conducted after validated translation from NONMEM to mrgsolve in R to evaluate 1) the contribution of GB and EHC on MPA exposures at first-dose and steady-state; 2) the impact of microbiome communities on drug exposures

Results: Analysis of 160 MPA, MPAG, and ACMPAG concentrations revealed that a one-compartment model for MPA and MPAG and a two-compartment model for ACMPAG, with linear elimination and GB and gut compartments, showed the best fit. Typical values of MPA, MPAG and ACMPAG CL were 35.2 L/h/70 kg (13.1% CV), 3.7 L/h/70 kg (56.4% CV) and 73.8 L/h/70 kg (49.9% CV), respectively. The fraction of MPAG undergoing EHC (FrEHC) in a typical patient was 0.387 (89.1% CV). Low relative abundance of certain communities containing Bacteroides spp reduced FrEHC, MPA AUC and trough concentration (Co) by 80, 6 and 83%, respectively. EHC significantly contributed to MPA accumulation, increasing AUC and Co by 33.5% and 100% with every 8 h dosing. GB removal didn’t affect MPA AUC but altered MPA concentration fluctuation over the dosing interval leading to 39% higher and 11% lower Co at first-dose and steady-state, respectively.

Conclusions: Bacteroides spp abundance affects FrEHC and MPA exposure. EHC contributes to MPA accumulation. GB removal impacts MPA Co but not AUC. This study shows an approach to integrate complex microbiome data into PopPK modeling. Future steps involve simulations to optimize MMF dosing considering various microbiome compositions and impaired GB function in virtual populations.

Citations: 1. Saqr, A. et al. Transplant. Cell. Ther. 28, 372.e1-372.e9 (2022). 2. Fabbrini M. et al. Microbiome Research Reports. 2023;2(4). 3. Blondel, V. D. et al. J. Stat. Mech. Theory Exp. 2008, P10008 (2008).