QSP in the Real World: Where Has Quantitative Systems Pharmacology (QSP) Been Used to Support Drug Development and With What Impact?

Authors: Shoda L

Quantitative systems pharmacology (QSP) is gaining acceptance and utilization in many drug development programs, but even those who appreciate its value may not fully recognize the extent to which it has already achieved impact and/or its range of potential applications. In part, this is because many QSP model results inform internal decision making and are not made public. Amongst publicly available examples of QSP models, it’s clear that QSP methodology has been applied to multiple diseases or health conditions, but the impact of QSP simulation results on drug development is often unstated.

In this blog post, I will focus on a set of case studies (and links to the journal articles) that demonstrate QSP impact on clinical trial design and/or decision making to concretely illustrate how your industry colleagues are gaining an edge using QSP.


Accelerating phase II/III dosing strategy based on predicted efficacy

Nirmatrelvir (PF-07321332), a coronaviridae family inhibitor, was under development for treatment of coronavirus disease 2019 (COVID-19). To accelerate drug development, population pharmacokinetic (popPK) and QSP modeling were leveraged in the analysis of phase I data. The popPK modeling allowed for selection of phase II/III dose, while the QSP model allowed for selection of treatment duration. The QSP model incorporated nirmatrelvir pharmacokinetics and preclinical pharmacology and assessed different dosing regimens. Simulation results predicted that a five-day regimen would significantly decrease viral load (as a surrogate of efficacy) in SARS-CoV-2 infected patients, whereas a ten-day regimen provided no additional benefit. Overall, modeling and simulation were estimated to expedite drug development with at least a 10-fold reduction in timeline relative to industry standards.


Optimizing dose regimen to improve phase I safety, while maintaining efficacy

Mosunetuzumab, a T cell dependent bispecific antibody that binds CD20 on B cells and CD3 on T cells to facilitate T cell mediated B cell killing, was under development for non-Hodgkin lymphoma (NHL). Cytokine release syndrome (CRS) was a recognized toxicity risk for mosunetuzumab, which had been shown to potently activate T cells. To mitigate CRS risk, QSP modeling was applied, including mosunetuzumab PK, nonclinical mosunetuzumab pharmacology, and clinical data from blinatumomab, a bispecific T cell engager molecule binding CD19 on B cells and CD3 on T cells, in a multi-compartmental (peripheral blood, bone marrow, secondary lymphoid tissues, and tumor) model. Simulations compared non-fractionated, flat fractionated, and double-step or single-step fractionated dosing and predicted that step fractionated dosing limits cytokine release. Further, simultaneous assessment of tumor response demonstrated a similar change in tumor size by day 84 across all dosing regimens. Simulation results supported step fractionated dosing regimen for phase I clinical trial design.


Optimizing dose regimen to substantiate phase I safety using lower doses

Emvodostat, a dihydroorotate dehydrogenase (DHODH) inhibitor initially in development to treat solid tumors, was terminated based on observed liver injury. Subsequently, preclinical data suggested emvodostat might effectively treat acute myeloid leukemia (AML) at lower doses. To facilitate translation and provide additional safety evidence, QSP modeling was applied, including emvodostat PK and nonclinical data on mechanisms of liver toxicity. Simulations evaluated eight different dosing regimens to identify dosing protocols that achieved target concentrations for AML treatment. Retrospective simulations reproduced liver toxicity, consistent with liver signals observed in the initial solid tumor trials. Prospective simulations predicted that the dosing regimens selected to achieve target concentration for AML treatment would not result in liver toxicity. Simulation results supported the design of a phase 1B clinical trial, and clinical results later confirmed liver safety.


De-risking next generation compound selection

After two first-generation CGRP receptor antagonist, telcagepant and MK-3207, failed in clinical trials due to hepatoxicity, several next-generation compounds were engineered to avoid hepatotoxicity. Despite the chemical modifications, questions remained about whether these next generation molecules might encounter similar liver safety issues. To de-risk next generation compounds, QSP modeling was applied, including PK and nonclinical data on mechanisms of toxicity for telcagepant, rimegepant, atogepant, ubrogepant, and zavegepant. Retrospective simulations reproduced telcagepant liver toxicity. Prospective simulations predicted that the next generation compounds had a dramatically lower risk of hepatotoxicity. Simulation results supported the decision to advance a next generation compound, which was subsequently shown to be efficacious and safe and received FDA approval.


Informing post-marketing trial design

Natpara, human parathyroid hormone (PTH) was in development to control hypocalcemia in patients with hypoparathyroidism. The registration trial demonstrated that human PTH given once daily, reduced the requirements for exogenous calcium and vitamin D, while retaining normal serum calcium levels. These data supported approval. However, once daily dosing with human PTH did not provide control of urinary calcium excretion, which under chronic conditions puts patients at risk for calcium deposition, kidney stones, and progressive renal impairment. To address whether alternate dosing regimens might allow for simultaneous control of serum and urine calcium levels, QSP modeling and simulation was applied, including PTH PK, oral vitamin D and exogenous calcium dosing, and calculation of 24-hour urinary calcium output. Simulations of alternate dosing regimens or formulations revealed that BID dosing or a slow‑release formulation could allow for control of both serum and urine calcium levels. Simulation results supported post-marketing requirements to conduct a clinical trial of alternative dose regimens, including assessment of hypercalciuria-associated long‑term safety.


Analyzing post-market carcinogenic hazard

When the California Office of Environmental Health Hazard Assessment (OEHHA) decided to review the carcinogenic hazard posed by acetaminophen (APAP), a detailed investigation of variability in patient characteristics, including baseline levels of glutathione (GSH) and hepatic antioxidants, and subsequent interaction with acetaminophen, was conducted in a QSP model. Simulations across a wide range of patient variability demonstrated that therapeutic doses of APAP (4g per day) were insufficient to deplete GSH and therefore would not induce oxidative stress leading to carcinogenic hazard. Supra-therapeutic dosing (12g per day) could deplete GSH stores in a subset of patients, leading to mitochondrial dysfunction, and cell death. The mechanistic sequelae of hepatic stress and death indicated that conditions would not lead to stress-induced DNA damage and potential carcinogenicity. In the acute overdose setting (>15g), simulations demonstrated a more extreme form of the mechanistic sequelae observed with supra-therapeutic dosing but also illustrated the hepatic regenerative response, leading to replacement of hepatocytes lost to APAP toxicity, with healthy hepatocytes. Together, the modeling and simulation supported the conclusion that APAP was not carcinogenic in therapeutic, supra-therapeutic, or acute overdose scenarios, adding to the overall weight of evidence in the assessment.


Identifying a narrow safety margin that supports program discontinuation

BAL30072, a novel antibiotic, was in development for the treatment of infections caused by multi‑drug‑resistant Gram‑negative bacteria but was halted due to liver toxicity in repeat dosing studies in normal healthy volunteers. To better characterize the safety margin and determine whether the safety margin could be improved with alternative dosing strategies, a QSP model was applied, including BAL30072 PK and nonclinical data on mechanisms of toxicity. Retrospective simulations reproduced dose‑dependent toxicity, as was observed in clinical trials. Toxicity parameters were optimized to reproduce clinical data more closely, specifically less toxicity in single dose protocols and more toxicity with prolonged dosing. With optimized parameters, novel dosing regimens were evaluated for comparative safety margins. Simulations demonstrated modestly improved safety margins with stringent monitoring protocols or with weight‑adjusted dosing. Simulation results added to the weight of evidence supporting the decision to halt development of BAL30072.


There are many ways QSP can be utilized to help answer critical questions during drug development. If you have questions about how QSP could be used in your program, we’d be happy to chat.