Landscape of regulatory quantitative systems pharmacology submissions to the U.S. Food and Drug Administration: An update report

Abstract

The number of quantitative systems pharmacology (QSP) submissions to the U.S. Food and Drug Administration has continued to increase over the past decade. This report summarizes the landscape of QSP submissions as of December 2023. QSP was used to inform drug development across various therapeutic areas and throughout the drug development process of small molecular drugs and biologics and has facilitated dose finding, dose ranging, and dose optimization studies. Though the majority of QSP submissions (>66%) focused on drug effectiveness, QSP was also utilized to simulate drug safety including liver toxicity, risk of cytokine release syndrome (CRS), bone density, and others. This report also includes individual contexts of use from a handful of new drug applications (NDAs) and biologics license applications where QSP modeling was used to demonstrate the utility of QSP modeling in regulatory drug development. According to the models submitted in QSP submissions, an anonymous case was utilized to illustrate how QSP informed development of a bispecific monoclonal antibody with respect to CRS risk. QSP submissions for informing pediatric drug development were summarized along with highlights of a case in inborn errors of metabolism. Furthermore, simulations of response variability with QSP were described. In summary, QSP continues to play a role in informing drug development.

Study highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Since our landscape analysis of QSP submissions to the U.S. Food and Drug Administration was published in 2021, the number of QSP submissions has been more doubled. There is a need for an update to inform the public of the current landscape of regulatory QSP submissions.

• WHAT QUESTION DID THIS STUDY ADDRESS?

This landscape analysis addresses the questions of how QSP modeling is utilized in regulatory drug development.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This landscape analysis informed us of the following: 1. The landscape of QSP submissions to the U.S. Food and Drug Administration. 2. The therapeutic areas with a high percentage of QSP modeling submissions. 3. The contexts of use in QSP modeling submitted in NDAs and BLAs. 4. The use of QSP modeling in pediatric drug development. 5. The use of QSP in informing in‐born errors of metabolism drug development. 6. How QSP is used to optimize the dosing regimens of bispecific monoclonal antibodies while minimizing their safety risk.

• HOW MIGHT THIS CHANGE DRUG DISCOVERY, DEVELOPMENT, AND/OR THERAPEUTICS?

This report informs the public of the broad utility of QSP modeling in facilitating drug development with respect to drug efficacy and safety.

INTRODUCTION

Quantitative systems pharmacology (QSP) is a modeling approach that can quantitatively and mechanistically integrate the mechanism of action of a drug, including its on‐target and off‐target pathways, with the pathological mechanism/process of its target indication across the multiscale of human biology (from molecules to cells to organs to systemic level) to simulate patient responses to drug treatment. QSP models can be constructed to simultaneously model patient responses with respect to drug efficacy and safety. ¹ QSP is a model‐informed drug development tool used across various diseases ² including rare diseases. ³ Since our previous landscape analysis publication of QSP submissions to the U.S. Food and Drug Administration (FDA) as of December 2020, ² the total number of submissions has increased with more QSP models submitted in new drug applications (NDAs) and biologics license applications (BLAs). For example, QSP simulations of treatment response and disease mechanism were compared between pediatric and adult patients. ⁴ QSP was also applied to separately simulate the treatment durations for immune‐competent patients and for immunocompromised patients. ⁵ Given continued activities in this area, an update on the landscape of regulatory QSP submissions was warranted.

This updated report consists of (1) descriptive statistics of QSP submissions to the U.S. FDA as of December 2023; (2) an overview of QSP contexts of use in regulatory submissions with specific cases highlighted; (3) a summary of QSP submissions for mitigating drug safety including illustration with a case study of a bispecific monoclonal antibody; (4) highlights of QSP utilized for informing clinical trials and drug development that involved adult and pediatric patients along with a case study involving inborn errors of metabolism; and (5) a summary of virtual patient populations (VPops) use in QSP submissions.

METHOD

The method published previously ² was used for mining FDA's submission databases with a set of key words, curating the submissions, and counting the number of QSP submissions. An internal standard of operation document that details the method published previously ² was referenced to ensure that consistent criteria were applied to qualify a QSP submission. The objective(s) and simulations (outputs) of a QSP submission described in a clinical protocol, a clinical study report, a QSP report, or a meeting package were carefully reviewed to obtain the needed information for this report. Individual contexts of use submitted in NDAs and BLAs were carefully reviewed following a comprehensive review of the model report in the NDA or BLA submission and assessment of the model. All QSP submissions that contained the use of VPops were reviewed and summarized in this report.

RESULTS

Summary of statistics

QSP submissions continuously increased from 2013 to 2023 with the total number of submissions reaching 367 by December 2023. The number of QSP submissions in 2023 edged over 80 (Figure 1). Figure 1 included yearly submissions from 2021 to 2023, updating what was published previously from 2013 to 2020. ² QSP modeling and simulation were applied to inform drug development across various therapeutic areas (Figure 2). Applications in cancer drug development programs accounted for 50% of total submissions, with 28% and 22% of total submissions attributed to submissions for the development of drugs for treating solid tumors and hematologic malignancies, respectively. The use of QSP to inform drug development for treating cancers has increased from 32% to 50% since our last landscape analysis, ² with applications in hematologic malignancies increasing from 10% by 2020 to 22% by 2023. Submissions in endocrine diseases and infectious diseases accounted for 9% and 8% of total submissions, respectively. Submissions in each of non‐malignant hematology, rheumatology and transplant medicine, urology/obstetrics/gynecology, pulmonary/allergy/critical care, and inborn errors of metabolism, accounted for 3%.

FIGURE 1.

Yearly trend of quantitative systems pharmacology submissions from 2013 to 2023.

FIGURE 2.

Pie chart of quantitative systems pharmacology for informing drug development across various therapeutic areas. The submissions in dermatology and dentistry, medical imaging and radiation therapy, hepatology and nutrition, anesthesiology, addiction medicine and pain medicine, and ophthalmology all together contributed to 5% of all submissions and grouped into the “Others” category.

QSP was applied to facilitate drug development across all drug development phases, NDA/BLA submissions, and post‐approval activities (post‐marketing requirements, post‐marketing commitments, efficacy and safety supplements, or annual safety reports) (Figure 3). The majority of QSP submissions were for aiding Phase I and Phase II drug development, accounting for 65% of total submissions. NDA and BLA submissions plus post‐approval submissions (supplements) contributed 17% of total submissions (Figure 3). The number of Phase III submissions was almost one‐half (18% vs. 35%) that of Phase II submissions. Preclinical submissions accounted for less than 1%. Compared to our previous report, ² applications in Phase I drug development increased by approximately 4% while Phase II drug development applications did not change. Submissions in Phase III, NDA/BLA and post‐approval activities including supplements decreased by approximately 4%. In brief, QSP modeling was mostly used to aid designs of clinical studies with the percentage across Phase I to Phase III totaling 83%.

FIGURE 3.

Horizontal stacked bar graph of quantitative systems pharmacology modeling for informing drug development across the drug development phases.

Context of use

From the objectives stated in QSP modeling submissions, QSP was utilized to inform drug development with respect to drug effectiveness and safety. To optimize the benefit/risk profile of a drug product candidate for a proposed indication or of a marketed drug product, QSP modeling was performed or planned to inform the first‐in‐human studies, dosing‐finding studies, dose‐ranging studies, dose selection studies, and dosing regimen optimization trials (single agent or multiple agent clinical trials), to fulfill the post‐marketing requirements, to update drug safety in the post‐marketing reports, and so on. To improve drug safety, QSP was applied to predict liver toxicity, to simulate cytokine elevation and associated risk of CRS, to simulate risk of bone loss (bone density), and others.

Recently, sponsors have described the context of use for their model submitted in NDAs and BLAs. The context of use of a QSP model submitted in an NDA or a BLA was reviewed along with the model report summarized (see Table 1 for the list of contexts of use). In brief, QSP modeling is an important drug development tool and has played a supportive role in drug development programs across small molecules, peptides, and biologics.

TABLE 1.

List of contexts of use summarized from NDA and BLA submissions.

Drugs	Context of use (COU)
Biologic A	The quantitative systems pharmacology (QSP) model incorporated biologically interpretable processes and associated parameters to represent the molecular mechanism of an in‐born errors of metabolism The model was used as a tool to understand the qualitative similarity between pediatric and adult patients at the molecular/cellular level of disease pathophysiology; and compare the responses to Biologic A between these two patient groups. In conjunction with other clinical data, the QSP model was employed to support the proposed dose regimen for both adult and pediatric patients
Small molecule A	Using QSP modeling approach to evaluate the optimal dosage of an antiviral drug for general high‐risk patients infected with COVID‐19 and subsequently for immunocompromised cohort
Peptide A	The QSP model was employed as an auxiliary tool to mechanistically model responses to Peptide A following administration to patients with an endocrine disease
Biologic B	The QSP analysis was used to support dosing decision focusing on two specific issues: Assessing and comparing biochemical response rates under various dosing regimens and different baseline soluble B‐cell maturation antigen (sBCMA) levels: For one phase II and one phase II studies, QSP simulated the dose level for patients with both high and low baseline sBCMA levels Examining the maintenance of responses following transition from weekly (QW) to bi‐weekly (Q2W) dosing: Specifically for a phase II study, QSP analysis aimed to discern whether the simulated virtual patient maintenance response under a constant QW regimen (no switch to Q2W) differed from a regimen involving a switch from QW to Q2W dosing
Biologic C	The QSP model was utilized to conduct virtual patient simulations to support the applicant's understanding of target protein's effect on platelet dynamics. The virtual patients were aligned with the phase III subjects in terms of the variabilities in Biologic C pharmacokinetics (PK) and baseline patient platelet production rates and baseline von Willebrand factor (VWF) levels. Simulated total von Willebrand factor (VWF) levels exhibited qualitatively similar trends over time as the observed von Willebrand factor ristocetin cofactor measurements across the treatment arms. Additionally, virtual patient simulations were expanded to a larger population size to forecast the treatment effect size. These simulations were employed to offer further supportive evidence for the assessment of Biologic C, complementing the clinical observations
Biologic D	The QSP model delineated the fundamental mechanism of action (MOA) for T‐cell redirecting bispecific antibody and incorporates relevant parameters to replicate the formation of target cell‐bispecific‐effecter (TBE) complex and Biologic D‐induced anti‐tumor activities in individuals with relapsed/refractory multiple myeloma (RRMM). The QSP model was applied to quantify the formation of TBE, assess treatment response, and estimate the duration of response in RRMM patients undergoing various dosing regimens of Biologic D. These simulations complemented the observed efficacy results in clinical studies and the findings from exposure‐response analysis
Biologic E	QSP modeling incorporated biologically interpretable processes and relevant parameters to quantify the release of IL‐6 induced by Biologic E treatment in patients with follicular lymphoma or diffuse large B‐cell lymphoma. QSP simulations predicted which of several proposed step‐up dosing regimens would likely result in lower IL‐6 concentrations compared to the original step‐up dosing regimen. With the assumption that reduced IL‐6 concentrations correspond to a lower risk of cytokine release syndrome, the step‐up dosing regimen that yielded the lowest predicted IL‐6 concentrations was implemented as part of a phase II study

Mitigation of drug safety

QSP applications in predicting drug safety were briefly summarized above in the Section 3.2. The details of QSP for predicting drug safety are further discussed below. A total of 122 QSP submissions were to address drug safety issues accounting for approximately one‐third of total submissions. Among these 122 submissions, 48% were for simulating drug‐induced liver injury, 34% for simulating and mitigating the risk of CRS, and 19% for other drug safety issues such as adverse effects on bone density (Figure 4). Among the 64 submissions for simulating non‐liver safety issues, 43 submissions simulated both drug effectiveness and safety. Clearly, QSP modeling can incorporate both pharmacological and toxicological effects of a drug in the context of optimizing its dosing regimen with respect to its effectiveness and safety. Importantly, for drug products to treat rare diseases including many hematological malignancies, QSP seems to be increasingly used to inform their development programs. ³ Among the 43 submissions, 32 utilized QSP to model and simulate step‐up/split dosing schemes for the goal of optimizing drug effectiveness while reducing CRS risk. To provide the context, an anonymous case illustrating the utility of QSP for mitigating CRS risk to inform the development of bispecific antibodies (bsAbs) is described below.

FIGURE 4.

Left pie chart shows 66.8% of quantitative systems pharmacology modeling (QSP) applications for simulating drug effectiveness only; 33.2% for simulating safety only and simulating both effectiveness and safety. Right pie chart shows the breakdown of the 33.2% submissions for simulating liver toxicity, cytokine release syndrome risk, and other safety issues (including bone density).

An anonymous case—mitigation of cytokine release syndrome risk

Bispecific antibodies (bsAbs) have become an important component of therapeutic strategies to treat cancers. Based on the general workflow in submitted QSP models, a case study of an anonymous bsAb targeting CD3 and tumor associated antigen (TAA) (namely, CD3/TAA bsAb) is highlighted below. This case study illustrated how mechanistic modeling approaches could be used to support the optimization of step‐up dosing regimen and to mitigate CRS risk, which is one of the key safety concerns with CD3/TAA bsAb and is potentially life‐threatening.

Step‐up dosing strategy (i.e., lower initial dose(s) followed by a higher efficacious full dose) is commonly used to mitigate the risk of CRS. Such a strategy can be guided by mechanistic modeling. Based on the submitted models, the general scheme is that the model accounts for in vivo dynamic interactions among drug targets, T cells, and CD3/TAA bsAb, relevant disease biology (e.g., plasma cells and B cells in the bone marrow of patient with multiple myeloma), and the hypothesized cytokine modulation (e.g., target engagement kinetics and immune desensitization). The model describes activation of CD8+ T cells, target cell killing, and cytokine release (using IL‐6 elevations to simulate CRS risk). By incorporating population PK or a minimal physiologically based PK model, QSP simulations of cytokine levels following various step‐up dosing scenarios are then linked with statistical modeling to predict clinical CRS risk, thereby informing the design of clinical trials. The use of QSP modeling allows for a more informative design of the step‐up dosing strategy to mitigate CRS risk. It also reduces the number of doses to be evaluated and shortens the study duration in terms of dosage optimization.

Drug development for pediatric patients

In general, a drug product is developed to treat its target disease in adult patients prior to the initiation of its pediatric study plan, which discusses whether the same disease also occurs in children. Under the Pediatric Research Equity Act (PREA), the plan to develop a drug product for treating a disease in pediatric patients should be submitted no later than 60 days after the end‐of‐phase II meeting or by the date mutually agreed to between the FDA and the sponsor. ⁶ For treating rare diseases that affect both adult and pediatric patients, or for treating non‐rare diseases such as autism for which the disease is primarily recognized in pediatric patients more so than in adult patients, the clinical drug development program of a drug product often may include enrolling both adult and pediatric patients in pivotal clinical trials. QSP has been applied to aid these drug development programs. There were 29 QSP submissions out of 367 submissions (approximately 8% of all submissions) in which QSP modeling was planned or conducted to inform development of drug products for treating these diseases in both adult and pediatric patients (see the gray area of the pie chart in Figure 5). Of note, there was only one submission proposing the use of QSP in the pediatric drug development plan for PREA. Since the development program of a drug product for treating a rare disease often enrolls both adult and pediatric patients in its clinical trials, more than 80% of these 29 submissions were for informing rare disease drug development (see the green portion of the bar graph in Figure 5). Below is a case study to illustrate how QSP is utilized in a drug development program for both adult and pediatric patients.

FIGURE 5.

Pie chart shows most of quantitative systems pharmacology submissions were for informing development of drugs for treating adult indications (92.1%). Only 4.9% and 3% submissions informed drug development for treating both adult and pediatric patients and for treating pediatric patients only, respectively. Among these 7.9% submissions, approximately 83% submissions were submitted in rare diseases.

Informing inborn errors of metabolism drug development

Model‐informed approaches have been used to support drug development and regulatory evaluation for rare diseases such as inborn errors of metabolism (IEM). ⁷ IEM is a group of rare genetic disorders caused by deficiency of one or more key enzymes, cofactors, or transporters involved in a specific metabolic pathway. Given its computational ability to describe pharmacological activity such as pharmacodynamic (PD) biomarker responses at multiscale levels, QSP has the potential to address some critical gaps in IEM drug development. For example, multiscale QSP model approaches have been developed to simulate the therapeutic effect of eliglustat in treating Type1 Gaucher disease (GD‐1) ⁸ ; and that of olipudase alfa for the treatment of acid sphingomyelinase deficiency (ASMD). ⁹ ASMD is a rare lysosomal storage disorder that includes the most severe form of Nieman‐Pick type A, the less severe form of Type B, and the intermediate form of Type A/B. ASMD is caused by pathogenic variants in the sphingomyelin phosphodiesterase 1 gene which results in reduced activity of the enzyme acid sphingomyelinase (ASM) and subsequent accumulation of sphingomyelin in various tissues. Olipudase alfa provides an exogenous source of ASM to degrade sphingomyelin in lysosomes of the cells. In clinical studies in patients with ASMD, olipudase alfa treatment resulted in a gradual debulking of sphingomyelin and subsequent release of ceramide which produced a transient increase in plasma ceramide levels after each dose administration. Reductions in plasma ceramide and lysosphingomyelin were observed in the maintenance phase of treatment after repeated dose administration of olipudase alfa. ¹⁰ The ASMD QSP model was developed with a list of assumptions; and consisted of a PK sub‐model (i.e., a reduced physiologically‐based PK (PBPK) model, a molecular level sub‐model, a cellular level sub‐model, and an organ‐level sub‐model ⁹ ). The cellular level sub‐model and organ‐level sub‐model were not mechanistic. The PBPK model was calibrated and validated with adult data and then scaled for pediatric patients.

The molecular level sub‐model was constructed with several assumptions and information borrowed from other cell types and enzymes for modeling the key molecular events describing the mechanism of action of olipudase alfa in macrophages. The model was able to describe both the PK profiles and the time‐courses of the plasma levels of ceramide and lysosphingomyelin which were biomarkers reflecting the mechanism of action and pharmacological activity of olipudase alfa. The QSP model provided an understanding of the response variability in ASMD patients with varying disease severity following treatment with olipudase alfa and an understanding of the qualitative molecular‐level similarity in macrophages between adult and pediatric patients. ⁴

Use of virtual patient population

With various computational algorithms developed by QSP scientists, QSP has been used to create a VPop for the goal of modeling and simulating the extent of variability and uncertainty (herein variability) of PD responses. ¹¹ In general, a range of values of individual key sensitive parameters including those known to have biological variations in a QSP model are utilized to create a VPop to simulate the variability of drug response. In total, 42 proposed drug products included in their drug development programs used VPops to simulate response variability, with some of these development programs still at the IND stages, some having advanced to NDA/BLA submissions, and the others at the post‐approval stage. These 42 drug development programs each have more than one QSP submissions (see Section 2 for counting of submissions). Ten of these 42 drug development programs included QSP modeling submitted in BLA/NDA submissions. Most of the 10 development programs included validation of the Vpop with only two development programs did not include validation of VPop. Approximately three quarters of the 32 drug development programs at the IND stages included some information for calibration and/or validation of the VPop generated.

In general, visual predictive checks (VPCs) in which simulations are compared with observed data were utilized to support the qualities of their model and VPop simulations. Simulation results in VPCs were expressed in various ways, including (1) mean and/or median, min, max; (2) mean and/or median, 90% or 95% prediction interval or 90% or 95% or 97.5% or 99% confidence interval; (3) mean and/or median, 25th and 75th percentiles and/or 5th and 95th percentiles; (4) median, interquartile range (IQR), with whiskers showing results with the range of Q1−1.5*IQR and Q3+1.5*IQR; (5) mean, standard deviation. Different algorithms and optimization methods were used to generate VPops. VPCs do not provide the distance measures between the observed trial data and the model predictions per a statistical criterion or mean square errors between the observed trial data and prediction per a quantitative cutoff or other quantitative measures for accepting a model. In most submissions, the terminology of confidence interval (instead of prediction interval) was used; and different ranges of interval were applied. The number of virtual patients used seemed arbitrary without any justifications regarding its relevance to the number of subjects in the observed data. From the information provided in NDA/BLA QSP submissions, the quality of VPops varied.

DISCUSSION

QSP is increasingly used to inform drug development, as evidenced in the increasing number of submissions to the U.S. FDA over the last decade. QSP informs drug development across a wide range of therapeutic areas and the drug development process. Overall, QSP is used for dose selection, dosing regimen optimization, and informing clinical trials for diseases including rare diseases with respect of drug effectiveness and safety. QSP submissions shed light on the broad range of context of use. Notably, QSP is increasingly utilized to simulate CRS risks at different splitting/step up dosing regimens to optimize the dosing regimen of bsAb in hematological therapeutic areas. In addition to informing clinical trials involving adult patients, QSP also informs clinical trials that include both adult and pediatric patients. In conclusion, QSP is a useful tool for informing drug development.

AUTHOR CONTRIBUTIONS

J.F.P.B., G.L., M.Z., J.W., Y.X., G.J.B., and T.T. wrote the manuscript; J.F.P.B designed the research; J.F.P.B., G.L., M.Z., J.W., Y.X., J.C.E., and T.T. performed the research; Y.Y., J.L., and H.Z. contributed analytical tools.

FUNDING INFORMATION

No funding was received for this work.

CONFLICT OF INTEREST STATEMENT

The authors declared no competing interests for this work.

DISCLAIMER

This article only reflects the views of the authors and should not be construed to represent the views or policies of the U.S. Food and Drug Administration.

Bai JPF, Liu G, Zhao M, et al. Landscape of regulatory quantitative systems pharmacology submissions to the U.S. Food and Drug Administration: An update report. CPT Pharmacometrics Syst Pharmacol. 2024;13:2102‐2110. doi: 10.1002/psp4.13208