Assessment and Benchmarking of Model Informed Approaches in Drug Development for Hemoglobinopathies: A Review of Scientific Advices From January 2000 to December 2024

ABSTRACT

Despite the fact that modeling and simulation are now recognized as promising innovative methodologies, their use in the context of development of drugs for sickle cell disease and Thalassemia has not yet been reviewed. Considering the challenges of conducting clinical trials for hemoglobinopathies, our work aims at exploring the current status of use of modeling and simulation by drug developers and their endorsement by regulators, based on European Medicines Agency scientific advices given to industry by the Committee for Medicinal Products for Human Use from January 2000 to December 2024. The present study describes different aspects of how modeling and simulation methods have been used and assessed. Our results highlight the need for an openly accessible structured framework that details the standards/criteria that make a method acceptable for regulators. The use of the ICH M15 credibility assessment framework is suggested for this purpose. Bearing in mind the question‐centric approach, a repository of drug development questions and associated methods and data was built from 45 and 31 scientific advices for sickle cell disease and Thalassemia, respectively. The classification of the questions, methods and data based on the question‐centric approach enabled modeling and simulation methods to be easily identified, objectively assessed and benchmarked against alternative methods used to address the same questions to evaluate their performance for characterizing benefit/risk of the different drugs.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    Modeling and simulation are promising innovative methodologies and their results can constitute key evidence for drug approval. EMA's scientific advices are a useful resource for learning how the drug development process can be improved, particularly for rare diseases, using innovative methodologies for which regulatory guidelines are not available.
• What question did this study address?

  • ○
    Considering the challenges of conducting clinical trials for hemoglobinopathies, our work aims at exploring the current status of use of modeling and simulation by drug developers and their endorsement by regulators, based on EMA scientific advice given to industry by the CHMP from January 2000 to December 2024.
• What does this study add to our knowledge?

  • ○
    Using the question‐centric approach, a repository of drug development questions and associated methods and data was built from 45 and 31 scientific advices for SCD and Thalassemia, respectively.
  • ○
    The reformulation and classification of questions based on the question‐centric approach enabled modeling and simulation methods to be easily identified, objectively assessed, and benchmarked against alternative methods used to address the same questions to evaluate their performance for characterizing benefit/risk of the different drugs.
• How might this change drug discovery, development, and/or therapeutics?

  • ○
    Our results show that modeling and simulation are still underused as alternative approaches to address key drug development questions (e.g., study design optimization, dose selection, extrapolation of efficacy, and safety) for hemoglobinopathies.
  • ○
    They also highlight the need for an openly accessible structured framework that details the standards/criteria that make a method acceptable for regulators. The use of the ICH M15 credibility assessment framework is suggested for this purpose.

1. Introduction

Sickle cell disease (SCD) and Thalassemia are common monogenic disorders known as hemoglobinopathies. Approximately 350,000 children are born with hemoglobinopathies annually and 3.4% of the deaths in children aged between 0 and 5 years of age are due to hemoglobinopathies. Although hemoglobinopathies are more prevalent in malaria‐prone areas due to the red blood cell abnormalities protecting against malaria, they have become a global concern due to migration [1]. Thalassemia is a genetic disease characterized by a reduced production of α‐ or β‐globin chains, which causes an imbalance in the globin chain ratio leading to fewer normal red blood cells, increased hemolysis, and altered iron homeostasis. SCD is also a genetic disease characterized by defective hemoglobin known as hemoglobin S (HbS), which polymerizes when deoxygenated. The red blood cells become distorted and damaged, and they cause vaso‐occlusive crises (VOCs) and chronic hemolytic anemia [1].

The current treatment options for SCD and Thalassemia are chronic blood transfusions, stem cell transplantation, and medicines [2]. As of 2024, the medicines licensed for SCD in the EU contain hydroxycarbamide, exagamglogene autotemcel, and thiotepa. Exagamglogene autotemcel and hydroxycarbamide/hydroxyurea increase the production of fetal hemoglobin (HbF), which prevents recurrent painful vaso‐occlusive crises and decreases chronic anemia. Thiotepa is an antineoplastic agent used as conditioning treatment before transplantation of blood‐making cells [3]. For the treatment of Thalassemia, approved products in the EU contain exagamglogene autotemcel, deferasirox, deferiprone, luspatercept, and thiotepa. Exagamglogene autotemcel's increase in the production of HbF decreases the dependence on blood transfusions in Thalassemia patients who do not produce enough hemoglobin. Deferasirox and deferiprone are iron chelators used to decrease the iron overload that results from chronic blood transfusions. Luspatercept enhances erythropoiesis and increases the release of mature erythrocytes [4]. Several other treatment options are currently under development for SCD and Thalassemia. That considered, hemoglobinopathies remain with unmet medical need due to limited curative treatment options [2].

The development of drug products for SCD and Thalassemia is limited by the difficulty of involving children in clinical trials [1]. New approaches are necessary to improve the efficiency of pharmaceutical drug development and decision making [5]. Since drug development has low success rates, there is a need to accurately characterize a drug's properties and its efficacy and safety as early as possible [6]. The use of modeling and simulation (M&S) can therefore be considered with high potential given their ability for efficient data analysis, predictive modeling and for adequate extrapolation, potentially resulting in shorter drug development times [7]. Moreover, benchmarking M&S methods against other better established methods used to address the same drug development questions enables the performance of model‐informed drug development (MIDD) to be compared to traditional drug development methods [8]. The benefits and limitations of using M&S for certain types of drug development questions can therefore be elucidated more easily.

The draft of the international council for harmonisation of technical requirements for pharmaceuticals for human use (ICH) M15 Guideline on general principles for MIDD details the proposed framework for the assessment of MIDD evidence [9, 10] and provides the basis for model assessment.

As part of the present work, the ICH M15 credibility assessment framework was applied to M&S methods used in the context of hemoglobinopathies drug development, as part of the ERAMET project (Horizon Europe grant agreement No. 101137141). ERAMET (Ecosystem for Rapid Adoption of Modeling and Simulation Methods to Address Regulatory Needs in the Development of Orphan and Pediatric Medicines) is EU funded and managed by 17 partners from Belgium, Norway, United Kingdom, Italy, Spain, France, and the Netherlands.

Based on scientific advice requests related to hemoglobinopathies submitted to the EMA, a repository of drug development questions and related methods and data was built. The questions involving M&S were mapped in the hierarchy for final benefit–risk assessment. Therefore, not only does the potential for the use of MIDD become more apparent, but there is also more clarity on the requirement for the efficient use and acceptability by regulators.

2. Methods

The methodology used to build a repository of questions from EMA scientific advices, to implement the ICH M15 framework for model assessment and therefore benchmark M&S versus more established methods is shown in Figure 1.

FIGURE 1.

Schematic representation of the process used to compile the repository of regulatory questions, associated methods and data related to the benefit/risk balance of the drug products for Sickle cell disease and Thalassemia, and to assess and benchmark modeling and simulation methods. CHMP, Committee for Medicinal Products for Human Use; EMA, European Medicines Agency; ICH, International Council for Harmonization; IRIS, Integrated Review of Information System; M&S, modeling and simulation.

2.1. Data Searching

Applicants from the pharmaceutical industry have the opportunity to request scientific advice (SA) from the Scientific Advice Working Party (SAWP) provided by EMA's Committee for Medicinal Products for Human Use (CHMP) during the drug development or the post‐authorization phase. Scientific advice is referred to as protocol assistance when it involves an orphan medicinal product developed for a rare disease. The scope of the questions is product or indication specific and is related to quality, non‐clinical, and clinical aspects of drug development, as well as methodological issues and the overall development strategy [11, 12].

IRIS [13] is an online platform that EMA uses to manage scientific information and regulatory procedures, including scientific advice. Documents are stored in a procedure‐related format. EMA Scientific Explorer [14] is an AI search engine launched in 2024 for knowledge mining on IRIS, initially for scientific advice and recently for other documents. It searches for filtered keywords “in‐document‐text” using standardized questions [15].

In EMA Scientific Explorer, the agreed conditions “sickle cell,” “Thalassemia,” and “thalassemia” were used to search for SAs for SCD and Thalassemia. The filters “modeling and simulation” and “extrapolation” were used to search for the procedure numbers of SAs in which the words “modeling and simulation” and “extrapolation” appeared in the text of the document. Procedure numbers were sought for initial scientific advice, follow‐up scientific advice, initial protocol assistance, and follow‐up protocol assistance documents.

2.2. Repository Building

The repository of regulatory questions built for SCD and Thalassemia was based on the question‐centric approach [8] and encompasses questions related to the benefit–risk balance of the concerned drug. Information related to the quality, primary pharmacodynamics, toxicology, clinical efficacy, clinical safety, and clinical pharmacology of the drug was comprehensively reviewed. The related questions from drug sponsors were subsequently translated into the so‐called regulatory questions, that are relevant questions to be addressed during drug development and therefore to be considered during regulatory assessment. These questions were classified in four categories (i.e., characterization of PK, efficacy, safety, or Benefit/risk) [8], organized in a hierarchical manner, that is, at cellular, molecular, tissular, individual, or population levels and for more transparency and improved granularity. The rationale is that presentation of the information in our work is structured differently from what is seen in scientific advice. The present work, based on the question‐centric approach [8], is aimed at reorganizing the information from a knowledge building perspective providing a structured integration of the data pertaining to the three categories from a knowledge building perspective that clarifies how different data and methods inform benefit–risk assessment, leading to and generating loops of knowledge flow spanning across data types and fields of expertise.

The structure of the repository relies on the benefit–risk assessment (BRA) that is supported by the characterization of efficacy, safety, and PK that will inform both efficacy and safety. For each drug, the characterization of efficacy, safety, and PK would therefore constitute successive sublayers for BRA of a particular combination of dose, formulation, and indication drug (see Figure 3 for overall structure of the repository). The three elements (dose/regimen, formulation, and indication) are called “the conditions of use.” Questions, data, and methods related to bridging across conditions of use are presented on top of the BRA, after the initial positive benefit/risk establishment. In this case, some extrapolation of efficacy and safety and some borrowing of information pertaining to the initial program is permitted.

FIGURE 3.

The overall structure of the repository. PK, pharmacokinetic.

Within each of the categories related to PK, efficacy, and safety characterization, respectively, the repository questions were subsequently scaled in molecular, cellular, tissular, individual, and population levels. Therefore, in each category, from top to bottom, the questions are arranged in layers with decreasing biological and statistical complexity levels (i.e., from population to molecular level). This scale, based on biology and statistical complexity, is intended to facilitate quantitative and mechanistic integration of the different data types including nonclinical and clinical data.

From left to right, the questions are presented from the root questions closest to the population benefit–risk assessment and successively branched to the (sub)questions related to the data and methodologies used to address them. Therefore, after the first layers on the left constituted by questions and sub‐questions related to efficacy, safety, or PK characterization, moving to the right, a specific set of sub‐questions related to methods and data sets used to inform the answer to those questions are progressively reported.

When questions asked by the applicants covered more than one of these categories, they were split into different questions during the reformulation to reflect each of the respective categories. Data and analytical methods used to address these questions for each drug were also collected and reported. Questions related to the types of data at different stages of drug development, the methodology for drug development including the design of studies, the methods for the measurement of the drug's effects, the impact of inter and intra individual variability on the effects of the drug, the methods for analysis of study data and the extension of a drug product's indication across different conditions of use were included in the repository.

For each question asked by the applicant, regulatory questions were identified and a repository of those questions was separately compiled for SCD and Thalassemia. This notably allowed to identify similar questions across advices and the methods and data proposed/used to address them. The latter were manually extracted by reading through the text or searching for relevant keywords such as “model,” “simulat,” “analysis,” “measure,” “extrapolat,” “endpoint,” “study,” “phase,” “PK,” “efficacy,” “safety,” “exposure,” “response,” “pa(e)diatric,” “adolescent,” “dose,” and “regimen”(−).

For compliance with the confidentiality of the information reviewed, the questions originally asked by the applicant were anonymized and rephrased, and identifiable drug product names, names of drug intermediates and company names were replaced with “drug product,” “intermediate,” and “the applicant,” respectively. Any other identifiable words were anonymized in a similar manner.

2.3. Implementation of the ICH M15 Framework for Model Assessment and Benchmarking of the Use of M&S Versus Established Methods

From the methods and data extracted and reported in the repository (Figure S1), the use of M&S as a method to answer regulatory questions was further investigated. The types of models used and their uses were described. The stages of drug development for which the built models were used to optimize drug development were investigated, including the number of drug products for which different models were described.

An investigation into regulatory requirements for the qualification of the M&S methods mentioned in the scientific advices was conducted. The keywords used to identify whether the use of a M&S method was supported by the CHMP included the words “supported,” “agreed,” “acceptable,” “endorsed,” and “suitable.” When keywords such as “unclear,” “unacceptable,” and “questionable.”

In the response to each M&S‐related question, the CHMP implicitly provided an assessment of the credibility of the model based on their expert knowledge, opinions, and experience, as well as standards in the field and related guidelines on M&S, clinical pharmacology, efficacy, and/or safety assessment. The different comments provided were analyzed in order to understand how to increase the acceptability of the M&S methods by regulators. We also carefully identified cases where the M&S methods were not used or mentioned by the applicant but rather recommended by the CHMP: this was identified using keywords/phrases such as “recommends,” “preferably using an xyz model/approach,” “could be of value,” and “would offer the advantage of.” The provided comments were noted. Comments about the quality of the data required to increase the acceptability of the model were also noted.

The regulatory assessments of the credibility of the models were subsequently correlated to the relevant aspects of the ICH M15 credibility assessment framework and illustrate the advantages of adopting and using the ICH M15 credibility assessment framework as a communication tool [9]. The ICH M15 is the recently published guideline on general principles of model informed drug development (MIDD). It provides a framework for assessment of M&S approaches in the context of regulatory interactions. The most important recommendation to the ICH M15 is that, to establish the appropriateness of an MIDD approach in a particular context, one should always consider the risk for the patient and the current expectations from the regulators to address the concern question. According to the M15 guideline, it is important that the technical criteria for assessment of M&S and model evaluation ensure that regulatory acceptance of the model will not put the patient at risk related to either lack of efficacy or safety. The ICH M15 is therefore explicitly putting the regulatory question (and context of use) in the center of the model evaluation and acceptance by the regulators. The link to the present work that explicitly lists the regulatory questions as well as data and methods proposed to address them is therefore obvious.

Using this framework is therefore expected to increase transparency during regulatory interactions such as scientific advices as different aspects of the MIDD strategy would be very explicit, making it easier to spell out aspects of agreement and disagreement between drug sponsors and regulators.

For questions that were identified as possible to be addressed by more than one methodology across development programs, and when M&S was one of the methods proposed, after the credibility assessment, the different methods and approaches to address the same drug development question were compared. For this purpose, the details about the methods and affiliated data used were retrieved. Where available, the criteria for the acceptability of the methods and the standards for the affiliated data were also included for the benchmarking exercise.

3. Results

3.1. Data Searching

The scientific advice and protocol assistance procedure numbers were found on Scientific Explorer and the documents corresponding to the procedure numbers were retrieved from the IRIS website. 45 and 31 SA documents were found for SCD and Thalassemia, respectively, for the period between January 2000 and December 2024.

295 and 215 questions were retrieved for SCD and Thalassemia, respectively, of which 15% and 28% were concerned with quality data, 15% and 6% with non‐clinical data, and 70% and 66% with clinical data for drug development. The types of questions included product‐specific and indication‐specific questions concerning the adequacy of data and the appropriateness of the methodology for collecting or handling that data.

3.2. Repository Building

The different steps taken from the questions as asked by the applicants in scientific advices to the final repository questions are shown in Figure 2.

FIGURE 2.

Steps taken to anonymize applicants' questions and reformulate them into repository questions. PK, pharmacokinetic.

Two repositories were separately built for SCD and Thalassemia, consistently with the general classification of the questions according to the Benefit/Risk balance and the presentation of different hierarchy levels, as schematized in Figure 3.

Of the translated questions and sub‐questions for both diseases, the higher proportion of questions was related to efficacy, followed by safety‐related questions. Pharmacokinetics‐related questions were only 3% of the bulk questions and sub‐questions for both SCD and Thalassemia. Figure 4 shows the total number of translated questions and sub‐questions for both diseases, and the relative proportions per hierarchical levels. It also shows the proportions of questions related to efficacy, safety, and PK that were addressed using M&S.

FIGURE 4.

A detailed overview of the numbers of translated questions and sub‐questions per benefit/risk classification and hierarchical level, including the questions addressed by modeling and simulation. PK, pharmacokinetic.

Subsequent layers of repository sub‐questions included questions related to the methodology used for data collection and analysis and sub‐questions related to data quality and informativeness are shown in the detailed repository (Figure S1).

3.3. Use of M&S Methods in Hemoglobinopathy Drug Development

Only 15 out of 45 scientific advices for SCD and 6 out of 31 scientific advices for Thalassemia referred to the use of M&S methods. 22 and 9 questions by applicants for SCD and Thalassamia, corresponding to 64 and 47 repository questions, respectively, involved M&S as analysis methods.

In general, M&S methods were used for approximately 6% and 5% of the questions for SCD and Thalassemia, respectively, with the highest proportion of M&S questions per repository questions being about PK of the concerned drugs. 55% of the M&S questions were related to small molecule drugs and 45% to biologics.

The most commonly used M&S methods identified for both SCD and Thalassemia were POPPK models and exposure‐response (E‐R) analyses. PKPD and E‐R models were also used several times for questions about Thalassemia. Mixed‐effect models with repeated measurements (MMRM) and (Generalized estimating equations) GEE models were less commonly used.

Some of the models were used to address more than one question. For example, the same POPPK model could be proposed to support dosing in populations underrepresented in clinical studies (e.g., renally impaired patients), to assess the impact of concomitant medications on PK and to support extrapolation of drug use in children. As shown in Table 1, M&S methods were mostly used to inform transitioning to subsequent phases of (clinical) drug development, informing the design of the next study for both SCD and Thalassemia. M&S methods were mostly used for optimizing drug development in Phase 2 and 3 clinical studies. POPPK was the most commonly used model.

TABLE 1.

The use of M&S at different drug development stages.

Stage of drug development	Non‐clinical studies	Phase 1 studies	Phase 2 studies	Phase 3 studies
Sickle cell disease
Number of questions addressed using M&S	2	5	27	30
Number of drug products	2	2	6	7
Types of models	PK/PD	POPPK E‐R MMRM	POPPK E‐R PK/PD MMRM GEE	POPPK E‐R PK/PD MMRM
Thalassemia
Number of questions addressed using M&S	0	15	6	26
Number of drug products	0	1	1	3
Types of models	N/A	POPPK PK/PD	POPPK E‐R PB/PK	POPPK E‐R PD MMRM

Note: M&S was used to optimize and advance non‐clinical and clinical drug development, but mostly Phase 2 and 3 studies. In most cases, M&S models addressed more than one drug development question per drug product, illustrating that one model can be used to address more than one question. POPPK was the methodology most commonly used.

Abbreviations: E‐R, exposure‐response analysis model; GEE, generalized estimating equations; MMRM, mixed models for repeated measures; N/A, not applicable; PB/PK, physiologically based pharmacokinetic modeling; PD, pharmacodynamic model; PK/PD, pharmacokinetic/pharmacodynamic model; POPPK, population pharmacokinetic model.

The main uses of modeling and simulation were the following:

1. POPPK model was used to describe the drug's PK/exposure, mostly in adult patients using data from either Phase 1, 2, or 3 studies. Covariate model component describing the impact of patient characteristics (e.g., body weight, maturation of renal or hepatic functions) was used to predict the PK/exposure in the target population.

2. PBPK model was used to simulate the drug's PK/exposure in the target population based on prior knowledge on physiology and drug chemical, physical, and pharmacological properties.

3. Most of the time, extrapolation based on exposure matching was the basis for selecting safe and effective doses and regimens in the target population.

4. POPPK/PD and E‐R models were used to describe the drug's response, mostly in adult patients using data from either Phase 2 or 3 studies. A covariate model component describing the impact of patient characteristics (e.g., disease status) was used to predict the response in the target population.

5. MMRM were used in adults for trial design optimization, including sample size estimation for pediatric studies.

6. MMRM was used in adults to characterize efficacy (mean treatment effect) by analyzing the efficacy endpoint.

7. MMRM models were used to assess the predictive value of an endpoint.

3.4. Implementation of the ICH M15 Framework for Model Assessment

Table S1 illustrates how the comments and recommendations by the CHMP to improve the acceptability of the M&S methods can be correlated to the key assessment elements per the ICH M15 credibility assessment framework. For each key assessment element of the ICH M15 framework, an illustrative example of response from the CHMP is provided. A column with model impact assessment was added in the table to make a direct link to ICH M15.

3.5. Method Benchmarking

9 and 5 questions were identified for SCD and Thalassemia, respectively, for which M&S and at least one other alternative method were proposed by applicants or recommended by the CHMP across advices. Figure 5 shows an example of the ICH M15 credibility assessment of a M&S model used for extrapolation of efficacy in adolescents alongside an alternative approach against which the MIDD could be benchmarked. Examples of the other questions for which M&S was benchmarked against an alternative method(s), including standard statistical (SSA), are included in Table 2.

FIGURE 5.

Credibility assessment of a model and benchmarking against alternative statistical analysis for addressing a drug development question related to extrapolation of efficacy in adolescents. AUCss, area under the curve at steady state; EMA, European Medicines Agency; PK, pharmacokinetic; POPPK, population pharmacokinetic model.

TABLE 2.

Questions addressed using M&S and an alternative method(s).

Question of interest	M&S model	Model impact	Alternative analysis method
What is the optimal dose for the proposed indication?	POPPK	Medium	SSA—primary analysis Weight‐based allometric scaling of the dose
Are the proposed non‐clinical data adequate to characterize the safety of the drug at individual level?	PK/PD	Low	NOAEL
What are the appropriate inclusion/exclusion criteria for the Phase 2/3 study?	POPPK E‐R	Medium	DSMB methods
Is the proposed study size adequate to characterize the efficacy of the drug in the target (sub)population?	POPPK E‐R	Medium	SSA—Sample size re‐estimation + multiplicity correction
Is the proposed endpoint appropriate for characterizing the efficacy of the drug?	MMRM	Medium	Sensitivity analysis
Is the proposed indication supported by the data?	POPPK	Low	Exposure matching Sample size re‐estimation E‐R
Is the proposed strategy for pooled analysis of data appropriate for characterizing the efficacy of the drug?	MMRM	Medium	SSA – ITT analysis
Is the proposed strategy appropriate for dose selection/escalation from a safety perspective?	POPPK E‐R	Medium	PROs
What is the adequate dosing regimen (number of doses and overall duration) to characterize efficacy?	POPPK	Low	Simulation + theoretical assumptions Interim analysis
What is the adequate dosing regimen (number of doses and overall duration) to characterize efficacy?	E‐R	Medium	PK/PD
Does the available evidence support extension of the indication?	POPPK E‐R	High	SSA—Hb‐AT87Q change from baseline
Is the proposed extrapolation approach appropriate for characterizing efficacy?	POPPK E‐R	High	SSA—Hb‐AT87Q change from baseline
Is the proposed strategy appropriate for dose selection/escalation given the variability in the population?	POPPK PB/PK	High	E‐R
Is the proposed statistical approach appropriate for characterizing the efficacy of the drug?	MMRM	Low	SSA—Cochran Mantel Haenszel chi‐squared test

Note: Each row represents a repository question and at least one M&S model that was used to address the question, as well as an alternative Non‐M&S method(s). The table illustrates how the question‐centric approach facilitates the benchmarking of M&S models against other methods used to address the same questions. The main uses of modeling and simulation were the following:

1. POPPK model was used to describe the drug's PK/exposure, mostly in adult patients using data from either Phase 1, 2, or 3 studies. Covariate model component describing the impact of patient characteristics (e.g., body weight, maturation of renal, or hepatic functions) was used to predict the PK/exposure in the target population.
2. PBPK model was used to simulate the drug's PK/exposure in the target population based on prior knowledge on physiology and drug chemical, physical, and pharmacological properties.
3. Most of the time, extrapolation based on exposure matching was the basis for selecting safe and effective doses and regimens in the target population.
4. POPPK/PD and E‐R models were to describe the drug's response, mostly in adult patients using data from either Phase 2 or 3 studies. Covariate model component describing impact of patient characteristics (e.g., disease status) was used to predict the response in the target population.
5. Models in adults were used for trial design optimization, including sample size estimation for pediatric studies.
6. MMRM models were used to characterize efficacy (mean treatment effect) by analyzing the efficacy endpoint.
7. MMRM models were used to assess the predictive value of an endpoint.

Abbreviations: DSMB, data and safety monitoring board; E‐R, exposure‐response analysis model; ITT, intention to treat analysis; MMRM, mixed models for repeated measures; NOAEL, no observed adverse effect level; PB/PK, physiologically based pharmacokinetic modeling; PD, pharmacodynamic model; PK/PD, pharmacokinetic/pharmacodynamic model; POPPK, population pharmacokinetic model; PROs, patient‐reported outcomes; SSA, standard statistical analysis.

4. Discussion

This research scrutinizes the drug development questions asked by applicants on SCD and Thalassemia in scientific advices between January 2000 and December 2024. The question‐centric approach was used to build a repository of benefit–risk related questions classified into hierarchical levels and categories, after they were translated to the perspective appropriate for regulatory decision making.

The scientific advices provided by the CHMP and related to M&S methods were analyzed and a clear link to the recommendations per the ICH M15 framework and to the use of the question‐centric approach was found (as described in Table S1). The proposed presentation in the repository therefore allows linking very explicitly the M&S approach to the root regulatory question it is related to. This is important because it allows characterizing the risk for the patient if the model goes wrong (so‐called model risk per ICH M15) and knowing the positioning of the model vis a vis what is the evidence requirement by the regulators for addressing the concerned question (so‐called model impact per ICH M15). An interesting dimension of the model impact is that it allows benchmarking the M&S with other alternative approaches considered acceptable by the regulators to address the same questions. Knowing the model risk and impact, it becomes easier to justify the appropriateness of the M&S approach, to set technical criteria and perform the evaluation of the model to inform the answer to the concerned question. This research therefore emphasizes the need for an openly accessible structured framework and transparent criteria for assessing M&S methods which can be used by both the applicants and the regulators, to reach a common understanding of the requirements for using M&S methods for regulatory submissions. The scope of the ERAMET project aligns with ICH M15, particularly with regards to identifying relevant drug development and regulatory questions whose answers drive key decision steps. Identification of the questions, as well as data and methods proposed or used to address them, should pave the way toward establishing the acceptable criteria for the M&S methods in the context of drug development or regulatory decision‐making. As mentioned previously, the framework is considered a powerful tool for communicating these criteria.

The ERAMET ambition is to expand beyond the ICH M15 and establish a quantitative framework to link model risk and impact to the acceptance criteria by regulators.

Reformulating the applicants' product‐specific and indication‐specific questions concerning the benefit/risk balance led to an explicit framework that could allow knowledge transfer and learning across the procedures. The classification of questions by benefit/risk and in hierarchical levels enabled a better understanding of how efficacy, safety, PK, and conditions of use are related and how they feed into the overall question of benefit/risk balance. This would also allow knowledge building using different data types (e.g., nonclinical and clinical data) that contribute to addressing the same question. The observation that three quarters of the repository questions built from scientific advices are efficacy and safety population level questions reflects the fact that the regulatory interactions happen quite late in the development of drugs for hemoglobinopathies. This also explains the positioning of the use of M&S as methods to generate robust population‐wide evidence for regulatory assessment. The use of most of the M&S methods proposed in the advices was located during Phase 2 and 3 studies.

The structure and granularity of the repository enabled the question‐centric approach to be used and to ease the identification of the use of M&S, as well as alternative methods to address the questions concerned. The fact that it was mentioned in 33% of the SAs for SCD and 20% of the SAs for Thalassemia shows that M&S has real potential for future drug development for hemoglobinopathies. Although the percentage of M&S questions as a total of all repository questions was only 6% for SCD and 5% for Thalassemia, it is interesting to note that questions involving M&S methods were found at the individual and population levels of the repository for efficacy, safety, and PK, including when bridging across conditions of use.

POPPK models and exposure–response analyses were the most popular types of models used, with cases where one M&S method could be used to answer more than one drug development question at a time.

The drug development questions that were identified for benchmarking of M&S methods against alternative approaches were questions related to study design, selection of endpoints, dose selection, generating evidence of efficacy and safety, and extension of indication from adults to pediatrics. Following benchmarking and the establishing of standards for the use of the models, M&S may be used as an alternative to these methods. The credibility assessment of the model remains an important part to support the methods benchmarking process.

Overall, the use of M&S methods to address drug development and regulatory questions for hemoglobinopathies is promising if sufficient data can be provided to develop and qualify the M&S models.

Author Contributions

Grace Shalom Govere, Jean‐Michel Dogné, and Flora Musuamba wrote the manuscript; Flora Musuamba designed the research; Grace Shalom Govere performed the research; Grace Shalom Govere, Jean‐Michel Dogné, and Flora Musuamba analyzed the data.

Funding

This work was supported by the ERAMET project, funded by the European Health and Digital Executive Agency (Brussels, Belgium), under Horizon Europe through grant number 101137141.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: psp470192‐sup‐0001‐AppendixS1.pdf.

Data S1: psp470192‐sup‐0002‐AppendixS2.xlsx.

Table S1: psp470192‐sup‐0003‐AppendixS3.docx.

Govere G. S., Dogné J.‐M., and Musuamba F., “Assessment and Benchmarking of Model Informed Approaches in Drug Development for Hemoglobinopathies: A Review of Scientific Advices From January 2000 to December 2024,” CPT: Pharmacometrics & Systems Pharmacology 15, no. 3 (2026): e70192, 10.1002/psp4.70192.