State‐of‐the‐Art on Model‐Informed Drug Development Approaches for Pediatric Rare Diseases

ABSTRACT

Pediatric rare diseases present unique challenges for drug development due to small patient populations, ethical constraints on clinical trial design, and limited prospectively defined natural history data. Model‐Informed Drug Development (MIDD) has emerged as a powerful paradigm to address these challenges by leveraging quantitative methods to enhance decision‐making across all stages of drug development. This paper reviews the state‐of‐the‐art MIDD approaches being applied to pediatric rare disease therapeutics, including the traditional pharmacometrics methodologies of population pharmacokinetic/pharmacodynamic (PK/PD) modeling, physiologically based pharmacokinetic (PBPK) modeling, disease progression modeling, and more future‐facing Bayesian trial designs, and real‐world data integration. We highlight how these methods facilitate dose optimization, support extrapolation from adult or other pediatric data, and enable more efficient and ethical clinical trial strategies. Case studies from recent regulatory submissions illustrate the growing acceptance of MIDD in pediatric rare disease contexts. Finally, we discuss the technological and regulatory advances driving this field forward, as well as current limitations and future opportunities for expanding the impact of MIDD on accelerating safe and effective treatments for children with rare diseases.

1. Introduction

MIDD has become a cornerstone of drug discovery, development, and regulatory approval [1, 2]. Drug development for pediatric rare diseases presents unique and complex challenges that require innovative approaches to optimize therapeutic outcomes. Conventional methods often fall short due to limitations such as small patient populations, ethical concerns restricting extensive clinical trials, lack of consensus on specific clinical measures for quantitative assessment at various stages of disease, considerable variability in the rate and non‐linearity of disease progression, and the physiological differences between children and adults [3]. MIDD uses mathematical and computational models to integrate multidisciplinary data, enabling informed decision‐making with regard to drug efficacy, safety, and dosing in specialized patient populations like children with rare diseases [4]. These models serve as tools to guide experimental design, predict outcomes, and assess risks, thereby reducing uncertainties associated with drug development while improving efficiency and the likelihood of success. The primary objective of MIDD is to enable informed decision‐making by synthesizing data and integrating knowledge from various scientific disciplines into predictive frameworks. By harnessing predictive capabilities, MIDD supports the design of rational therapeutic strategies and accelerates development timelines, ensuring that effective and much‐needed treatments reach pediatric patients.

MIDD relies on several types of models, each addressing specific components of drug action, pharmacology, and clinical outcomes. While a diverse range of modeling approaches is used in MIDD, each offering distinct capabilities depending on the development stage and scientific questions being addressed, the critical types of modeling approaches central to MIDD, particularly in the context of pediatric rare disease drug development, include pharmacokinetic (PK), population (pop) PK, pharmacodynamic (PD), physiologically based PK (PBPK), PK/PD, quantitative systems pharmacology (QSP), model‐based meta‐analysis (MBMA) and disease progression models (DPM) ([5], See Section S1). Figure 1 illustrates the importance of viewing MIDD as an ecosystem and highlights the key components of the data continuum.

FIGURE 1.

Harnessing the value of the data continuum in MIDD.

Collectively, these methodologies encompassing MIDD allow the development and application of exposure‐based, biological, and statistical models derived from preclinical and clinical data sources to inform drug development and decision‐making. MIDD can integrate and leverage all available information from various data sources to answer questions, bridge gaps, and inform decision‐making along the orphan drugs development journey to reduce uncertainty and lower failure rates. FDA's MIDD paired meeting program is a great avenue to engage the agency to discuss and seek advice on the applications of MIDD during drug development [6]. With the plethora of challenges in rare disease drug development, the application of MIDD is particularly compelling to address critical drug development issues including dose optimization, therapeutic individualization, risk assessment and mitigation, preclinical to clinical translation, clinical trial design, end point selection, product labeling, and marketing registration [4]. Therefore, contributing to the data package that supports evidence of effectiveness. The drug development challenges are compounded in the pediatric population because of scientific, logistical, ethical, and regulatory complexities [7]. MIDD is a powerful toolset that can streamline pediatric drug development by integrating and leveraging all available existing knowledge to help bridge knowledge gaps and facilitate development and decision‐making processes. A comprehensive review on this topic [8] broadly laid out the applications of MIDD into three categories—support extrapolation of effectiveness and safety from adult studies, dose selection and optimization, and informing pediatric trial design.

In this review, the development and applications of MIDD in rare pediatric diseases are discussed. This includes an in‐depth background of several MIDD approaches, case studies on MIDD applications, regulatory considerations, challenges and limitations of MIDD in pediatric rare diseases, and finally future considerations.

2. Applications of MIDD in Pediatric Rare Diseases

We begin by highlighting how each of these tools played a role in the development of novel therapeutics through a series of case studies. The case studies are ordered by the nature of complexity.

2.1. Case Study 1: Spinal Muscular Atrophy (SMA)

SMA is an inherited neurodegenerative disorder caused by mutations in the SMN1 gene, which encodes for survival motor neuron protein (SMN). The reduced expression of SMN leads to loss of motor neurons, severe muscle weakness, loss of motor function and ambulation, and often early death [9]. Risdiplam is an orally administered small molecule that increases the level of functional SMN protein [10]. Risdiplam was approved by the US FDA in 2020 as the first orally administered drug for the treatment of SMA for patients ≥ 2 months old, followed by the EMA. Risdiplam has been approved for the treatment of SMA in pediatric and adult patients based on three studies—(1) Firefish, a pivotal Ph 2/3 study in patients 1–7 months old, (2) Sunfish, a pivotal Ph 2/3 study in patients 2–25 years old, and (3) Jewelfish, a supportive Phase 2 study in patients 6 months—60 years old [11].

Risdiplam development and approval was supported by physiologically based PK (PBPK) and population PK (popPK) modeling as detailed in the summary basis of approval. A PBPK model was developed to predict DDI potential of risdiplam as a perpetrator of CYP3A4 based DDI in pediatric population [12]. Risdiplam exhibits time‐dependent inhibition (TDI) of CYP3A in vitro and thus had the potential to be a perpetrator of CYP3A DDI. However, as a clinical DDI study was not feasible in pediatric patients with SMA, a PBPK model‐based strategy was applied to extrapolate DDI risk from healthy adults to children with SMA. The modeling was based on clinical DDI data generated in healthy adults at relevant risdiplam exposures observed in children. An 18‐fold lower in vivo CYP3A inactivation constant compared with the in vitro value was required for risdiplam to capture the observed DDI in adults. This value was applied prospectively in the pediatric risdiplam PBPK model and used to simulate the DDI with midazolam. The impact of various intestinal and hepatic CYP3A ontogenies in children relative to adults was also investigated. The PBPK model demonstrated comparable DDI effect of risdiplam on midazolam in adults and pediatric patients. The model showed that there was low potential of clinically relevant interactions between risdiplam and CYP3A substrates (midazolam AUC ratio with and without risdiplam of 1.09–1.18) in any age group (2 months–18 years).

As an extension to the above PBPK model, a mechanistic popPK model was developed by integrating popPK and PBPK models of risdiplam to derive the in vivo flavin‐containing monooxygenase3 (FMO3) ontogeny and investigate its impact on DDI in children [13]. Risdiplam is primarily eliminated through hepatic metabolism by flavin‐containing monooxygenase3 (FMO3) and CYP3A, by 75% and 20%, respectively. While the FMO3 ontogeny is critical input data for the prediction of risdiplam PK in the pediatric population, it was mostly studied in vitro, and robust in vivo FMO3 ontogeny was not available. This model successfully estimated in vivo FMO3 ontogeny from risdiplam data collected in 2 months–61 years old. Additionally, the use of in vivo FMO3 ontogeny function improved the prediction of risdiplam PK in children compared to in vitro FMO3 ontogeny functions. Simulation using this model with refined FMO3 ontogeny using a range of theoretical CYP3A‐FMO3 substrates demonstrated comparable or lower CYP3A‐victim DDI propensity in children compared with adults. A comparable or lower CYP3A‐victim DDI and TDI potential were predicted for risdiplam in children aged ≥ 2 months compared with adults. The derivation of a robust in vivo FMO3 ontogeny function through this MIDD approach has significant implications for the prospective prediction of PK and DDI in children for other FMO3 substrates. This showcases the inclusion of refined FMO3 ontogeny to support the modeling of substrates metabolized by both CYP3A and FMO3.

This case study demonstrates a novel application of mechanistic modeling in a rare pediatric disease to assess DDI risk in children by bridging from adults. A popPK model was developed for risdiplam using pooled data from 26 healthy volunteers and 301 patients with SMA [11]. A 2‐compartment PK model with three transit absorption compartments and first‐order elimination adequately described PK of risdiplam. The model included data from infants, adolescents, and adults with varying ages and body weights. The popPK model‐based analysis suggested that age and body weight influenced risdiplam PK. Based on these analyses, weight‐based dosing in patients who were ≤ 2 years and those ≥ 2 years but had body weight < 20 kg was recommended. But fixed dose was recommended in patients ≥ 2 years old and weighed > 20 kg.

Another example of a drug in SMA is nusinersen, a synthetic antisense oligonucleotide administered that is given intrathecally. Nusinersen works by modifying splicing of SMN2 pre‐mRNA, resulting in exon 7 inclusion and increased production of full‐length and functional SMN protein. Nusinersen is approved for treatment of SMA in many countries with a recommended dosing regimen of four 12‐mg loading doses, followed by maintenance doses every 4 months. The PK of nusinersen was assessed in both the cerebrospinal fluid (CSF) and plasma of infants (n = 29) and children (n = 43) with spinal muscular atrophy (SMA) from across five different trials which formed the basis of the role of MIDD approaches in dose selection. The summary basis of review [14] revealed that the Sponsor had determined target tissue concentrations in a variety of models. For example, in SMA transgenic mice, Biogen determined the target tissue concentrations needed to produce 50% to 90% SMN2 exon 7 inclusion as between 1 and 10 μg/g in spinal cord tissue. Furthermore, the target tissue concentrations of spinal cord tissue (lumbar, thoracic, and cervical regions) taken from 3 deceased patients in Phase 2 Study CS3A were also determined and found to be > 11 μg/g (range: 11.9 to 31.8 μg/g) at nusinersen doses of 6–12 mg. PK/PD modeling and E–R analyses supported the dose of 12 mg [14].

2.1.1. Lessons Learned From a MIDD Perspective

Risdiplam and nusinersen represent two differently delivered drugs for the same indication. For risdiplam, MIDD was applied for dose selection using PBPK and popPK methods across the covariates of age and weight, whereas for nusinersen, CSF exposure modeling was the primary approach. In both cases, natural history data were leveraged, with modeling and some external controls applied for risdiplam, whereas for nusinersen, there was stronger reliance on SMA natural history data. In both cases, SMN protein was used as a biomarker for exposure/response. For risdiplam, bridging for pediatrics was modeled from adults to infants, and for nusinersen, they were extrapolated from older to younger patients.

2.2. Case Study 2: Post‐Transplant Cytomegalovirus (CMV) Infection

Maribavir was approved by the US Food and Drug Administration for the treatment of patients aged ≥ 12 years and weighing ≥ 35 kg with posttransplant cytomegalovirus infection/disease refractory (with/without resistance) to valganciclovir, ganciclovir, cidofovir, or foscarnet, with an oral dose of 400 mg twice daily [15, 16]. With no pediatric clinical data available and difficulty in trial recruitment, population pharmacokinetic modeling and simulations were conducted to predict the pharmacokinetics and inform maribavir dosing in adolescents.

The popPK analysis aimed to address the challenge of determining the appropriate dosing of maribavir for adolescents with posttransplant refractory CMV infection, given the absence of pediatric clinical data and difficulties in recruiting adolescents for clinical trials. MIDD‐based approach was employed to support this approval and inform dosing recommendations for adolescents.

Initially, a popPK model was developed for adults using data from multiple clinical studies, including phase I, II, and III trials. This model utilized nonlinear mixed effects modeling to describe maribavir's plasma concentration over time in both healthy volunteers and transplant recipients with CMV infection. The model considered various covariates, such as body weight, which influenced clearance and volume of distribution parameters. However, due to imprecise estimates when weight exponents were estimated from the data, fixed allometric exponents were used instead. The adult PopPK model was then adapted for adolescents through simulations. Virtual adolescent participants aged 12 to 18 years were created, with weights ranging from 25 to less than 100 kg. The simulations predicted concentration‐time profiles for maribavir following a 400 mg twice‐daily dosing regimen. The model incorporated interindividual variability but excluded residual variability in predictions. Key pharmacokinetic measures, such as steady‐state area under the plasma concentration curve (AUC_0‐τ ), peak plasma concentration (C _max,ss), and trough plasma concentration (C _min,ss), were calculated using noncompartmental methods.

The simulations revealed trends in maribavir exposure relative to weight, with exposure generally increasing as body weight decreased. For efficacy considerations, the simulated exposures in virtual adolescents were compared to those in adults receiving the same dose. It was found that at least 50% of virtual adolescents had AUC_0‐τ and C _min,ss values above the geometric means observed in adults. For safety considerations, exposures in adolescents weighing less than 40 kg were higher than those in adults, but still within acceptable limits compared to higher adult doses, indicating that a 400 mg twice‐daily dose of maribavir in adolescents weighing at least 35 kg is likely to provide similar safety and efficacy profiles as observed in adult patients with CMV. This approach, relying on popPK modeling and simulations, allowed extrapolation of adult data to adolescents, addressing the unmet need for dosing recommendations in this population without direct pediatric clinical trials. This supports extending maribavir's use to adolescents through PK‐matching, highlighting popPK modeling's potential in pediatric dose selection when clinical data are limited.

2.2.1. Lessons Learned From a MIDD Perspective

In the case of maribavir, a plethora of MIDD tools were used in the development program. PBPK modeling helped predict pediatric exposure and guided dose selection. PopPK modeling used sparse PK data to describe variability in exposure, and exposure matching was applied to bridge efficacy from adults to pediatrics.

2.3. Case Study 3: Leveraging Natural History Data in Regulatory Decision Making

The FDA relied on historical studies to approve Carbaglu (carglumic acid) for treating hyperammonemia caused by N‐acetylglutamate synthase (NAGS) deficiency [17]. This disease is characterized by clinically significant increases in plasma ammonia levels that relate closely with manifestations of encephalopathy. Ammonia and plasma citrulline are monitorable biomarkers indicative of disease progression. In the carglumic acid development program, there were no Phase 3 studies done due to the small number of patients. In early trials of clinical pharmacology characterization, ammonia concentrations were not studied because of the healthy volunteer population. In the FDA summary basis of approval [17], the weight of evidence was a dose–response relationship in 6 patients. Therefore, the reviewers indicated that the clinical efficacy of carglumic acid relied on data from retrospective cases in 23 patients who received carglumic acid over 16 years. Carglumic acid is indicated for the treatment of acute hyperammonemia in pediatric and adult patients with a suspected or confirmed diagnosis of propionic acidemia or methylmalonic acidemia.

2.3.1. Lessons Learned From a MIDD Perspective

Pharmacometric analysis of carglumic acid data included popPK modeling of drug exposure with covariates of age and weight. Exposure/response modeling involved the use of relating drug concentrations with reduction in ammonia. PK/PD simulations supported dose selection and titration strategies. However, the prominent MIDD methods relied heavily on the use of pediatric extrapolation to inform dosing in neonates and infants with a heavy integration and dependence on natural history data, which supplemented evidence in the absence of a randomized clinical trial (RCT).

2.4. Case Study 4: Pozelimab for the Treatment of CHAPLE Disease

Veopoz (pozelimab‐bbfg) injection is the first drug approved for CHAPLE disease [18]. It is a recombinant human‐monoclonal IgG4 antibody that binds with high affinity to human complement component 5 (C5). Pozelimab works by binding to the C5 protein, resulting in the inhibition of its cleavage to C5a and C5b. It is indicated in the treatment of patients 1 year of age and older with CD55‐deficient protein‐losing enteropathy (PLE), commonly referred to as CHAPLE disease. The FDA reviewers weighed on the substantial evidence of effectiveness as has been shown in one adequate and well‐controlled trial plus confirmatory evidence [18]. This approval is unique in that the basis of evidence on the efficacy and safety of pozelimab in 10 subjects with confirmed CD55‐deficient PLE relied on a single‐arm, open‐label trial design. The agency found this approach acceptable based on pre‐treatment data on each trial subject, which allowed an understanding of the subject's disease course, clinical status, and severity as a function of time in the absence of treatment; therefore, allowing each subject to serve as his/her own control. This approach suggests that in certain situations, a full RCT may not be needed.

There is also literature evidence in the summary basis that in 23 additional subjects with symptomatic CHAPLE disease, there were longitudinal data on low serum albumin as a prominent feature of subjects with active disease and which does not normalize in the absence of efficacious treatment. Moreover, the sponsor collected valuable data on multiple biomarkers in the single arm study. These supportive data was complimented by the reviewer in that the sponsor provided “strong mechanistic support.” The approval was based on an understanding of the pharmacological impact of the single gene defect, use of drug effect data using a plethora of in vitro and ex vivo assays on clinically relevant pharmacodynamic biomarkers of complement activity leading to the confirmatory evidence providing substantial evidence of effectiveness to support approval.

2.4.1. Lessons Learned From a MIDD Perspective

MIDD was applied in the development of pozelimab by informing weight‐based dose strategy using popPK and exposure/response relationships to complement inhibition using CH50 suppression as a surrogate endpoint. E/R relationships supported dose justification across pediatric age groups. Natural history data served as an external comparator to support clinical benefit decisions.

This case study raises an interesting question on how to integrate natural history data, whether as a standalone study or embedding natural history data within a clinical trial. A standalone study is a preferred option for a pediatric rare disease that is fraught with uncertain disease progression and highly variable signals. A well characterized natural history cohort can serve as an external control. Such studies can support accelerated approvals based on FDA and EMA mechanisms. Integrating natural history assessment into a clinical trial may be more appropriate when the disease has a well characterized progression, and the purpose of the data integration is to provide additional longitudinal insights.

2.5. Case Study 5: Leveraging Natural History Data for the Approval of Mirdametinib for Neurofibromatosis Type 1 (NF1)

Mirdametinib is a small molecule inhibitor of mitogen‐activated protein kinases 1 and 2 (MEK1/2), which received recent FDA approval for neurofibromatosis type 1 (NF1), an autosomal dominant disorder prevalent in pediatric patients and adults [19]. Plexiform neurofibromas (PNs) are benign tumors of the peripheral nerve sheath associated with significant morbidity including pain, neurologic deficits, and functional impairment. The sponsor showed that mirdametinib inhibited ERK phosphorylation and reduced neurofibroma tumor volume and proliferation iIn a mouse model of NF1. Mirdametinib is indicated in adult and pediatric patients 2 years of age and older with neurofibromatosis type 1 (NF1) who have symptomatic plexiform neurofibromas (PN) not amenable to complete resection. The recommended dosage of mirdametinib is 2 mg/m2 orally twice daily with or without food for the first 21 days of each 28‐day cycle, until disease progression or unacceptable toxicity.

The review package indicated that approval was based primarily on overall response rate (ORR) and duration of response (DOR) results from the ReNeu study [19]. Mirdametinib showed a clinically meaningful effect on ORR that was durable and associated with improvements in disease‐related morbidities. Because this is a direct clinical benefit to patients, reviewers recommended regular approval. The safety profile was considered in relation to other MEK inhibitors and found to be acceptable because of the serious and life‐threatening disease by the reviewers. The ReNeu study is a multicenter, single‐arm trial evaluating mirdametinib in adult and pediatric patients with NF1 and progressive or symptomatic plexiform neurofibromas not amenable to complete surgical resection. The primary efficacy population consisted of 114 patients, including 56 pediatric and 58 adult patients. There was radiographic tumor shrinkage, but the ReNeu study also evaluated mirdametinib effects on PN‐related functional impairment, symptoms, and disfigurement using various clinical outcome assessment (COA) measures as well as longitudinal photographic evaluations. This approval is yet another example where a single‐arm design, small sample size, and disease heterogeneity were not limiting to FDA approval because functional reduction in tumor volume and durable improvement in clinical outcomes were evidenced.

The sponsor also provided literature evidence showing a key characteristic of NF1 PN is the uncommon occurrence of spontaneous regression, which showed that the tumor response in the ReNeu study was due to drug effect. This supportive information included the placebo control arm of the NCI study evaluating tipifarnib in pediatric patients with NF1 and progressive PNs, as well as another NCI study which was a natural history study of children, adolescents, and adults with NF1.

2.5.1. Lessons Learned From a MIDD Perspective

PopPK modeling of mirdametinib allowed the selection of doses for pediatrics, with exposure‐matching supporting extrapolation approaches from adult data. Natural history integration involved the use of informed endpoints and study design.

2.6. Case Study 6: MIDD in Duchenne Muscular Dystrophy (DMD)

DMD is the most common and severe form of pediatric muscular dystrophy, with an incidence of approximately 1 in 3500–5000 live male births globally [20]. DMD is an invariably fatal, X‐linked, monogenic, degenerative, neuromuscular disease caused by mutations within the dystrophin gene, resulting in an absence or deficiency of functional dystrophin, an important structural protein critical to muscle health and function. In patients with DMD, progressive and irreversible muscle damage is ongoing at birth due to the lack of functional dystrophin, leading to ambulatory difficulties, cardiomyopathy, the need for assisted ventilation, and premature death [21, 22, 23].

Advancement in disease management and elucidation of disease pathology led to the emergence of novel therapeutics that target the underlying pathobiological cause and significantly extend the life span of patients. Of these therapies, eteplirsen (Exondys 51) and delandistrogene moxeparvovec (ELEVIDYS) are two examples where MIDD served an influential and prominent role in their pediatric development and approvals.

Eteplirsen is an antisense oligonucleotide (ASO) and the first approved exon‐skipping therapy in patients without age restriction with exon 51 skip‐amenable DMD. Eteplirsen binds to RNA targets, alters pre‐mRNA splicing to restore the reading frame, and enables the production of a shortened but functional dystrophin protein. Eteplirsen is in the therapeutic class of phosphorodiamidate morpholino oligomers (PMOs) that provide different targeted skipping of exons within the DMD gene (VYONDYS 53, AMONDYS 45, VILTEPSO). The DMPK and ADME characteristics of the PMOs have been extensively evaluated and demonstrated consistent PK and favorable safety profiles [24]. Clinical evaluations showed eteplirsen in patients > 4 years old is safe and well‐tolerated, as well as having consistent PK characteristics and attenuated pulmonary and ambulatory decline compared with mutation‐matched natural history controls [25]. However, clinical evaluations in the youngest DMD population (< 4 years old) are ongoing. This deficit is addressed via population PK modeling of eteplirsen, which helped characterize PK profiles and identified clinically meaningful covariates across a broad age range to assess the dosing regimen that maintained consistent efficacious exposure [26]. These results supported a weight‐based dosing regimen and extrapolation of the uniform dosing at 30 mg/kg/week to the youngest patients that is implemented in clinical use today.

Delandistrogene moxeparvovec is the first approved recombinant adeno‐associated virus (AAV) gene therapy designed to treat the proximate cause of DMD in patients ≥ 4 years old by replacing dysfunctional or missing dystrophin protein with a functional shortened dystrophin (micro‐dystrophin) in cardiac, pulmonary, and skeletal muscle, tissues most affected in DMD. The clinical development and dose selection for Elevidys (one‐time dosing therapy) were guided by the extensive nonclinical PK/PD evaluation and modeling performed in the DMD^MDX mouse model [27]. These findings were pivotal in characterizing the biodistribution of the AAV drug in target and off‐target tissues and in elucidating the exposure–response relationship of key efficacious biomarkers that guided the clinical translation and study designs in patients. Importantly, vector load‐expression‐response analysis provided insights into the relationship between surrogate biomarkers and functional improvement. This body of work served as one of the first examples demonstrating the application of PK/PD evaluation to AAV‐based gene transfer therapies.

Parallel to the advancement made in the treatment of DMD, exciting progress has been made in the development of DMD disease progression models on various key clinical measures across a broad DMD age range, disease severity, and trajectories in both ambulant and non‐ambulant populations [28]. These models accurately described the observed longitudinal data from both clinical trials and natural history studies and identified key prognostic factors that helped explain the population variability in disease progression. Notably, these models can be leveraged through clinical trial simulations to optimize clinical trial design, inform placebo arms through historical controls, as well as enable clinical assessment of therapeutic effect during all phases of drug development. Delandistrogene moxeparvovec dose decisions were also supported by caregiver‐reported outcomes, wherein, on average, caregivers of delandistrogene moxeparvovec‐treated patients reported improvements compared with those receiving placebo on one or more global item(s). While Clinical Outcome Assessments were used in the development of delandistrogene moxeparvovec, they were not the primary endpoint for FDA approval. As such, this is likely the first report evaluating a caregiver‐reported outcome [29].

2.6.1. Lessons Learned From a MIDD Perspective

MIDD was significantly leveraged in the development of both eteplirsen and delandistrogene moxeparvovec. In both cases, the primary endpoint of dystrophin (or in the case of delandistrogene, microdystrophin) was integrated with endpoints of 6MWT or NSAA and functional decline, respectively. Natural history data allowed the use of an external control, and biomarker/outcome correlations were achieved by linking dystrophin or microdystrophin with functional outcomes. For pediatric dosing, a combination of trials and modeling was used for eteplirsen, and for delandistrogene, AAV dosing was scaled by weight. Vector load‐expression‐response analysis was undertaken for delandistrogene moxeparvovec.

3. Lessons Learned From the Case Studies: Integrated Models for Pediatric Rare Diseases

These case studies highlight that, with limited pediatric trial data, PK modeling reduces the need for extensive invasive studies by extrapolating results from adult data or other age groups. At the very least, PK modeling helps bridge knowledge gaps and optimize drug dosing for pediatric populations. Because pediatric patients, especially neonates and infants, have immature organ systems, such as liver enzymatic activity and renal clearance capacity, factors underpin the need for age‐stratified parameterization in PK models to accurately reflect pediatric physiology. The extrapolation of adult PK data to pediatric populations generally employs a combination of empirical scaling methods and mechanistic modeling approaches. While allometric scaling is common in pediatric patients [30], there is greater adoption of PopPK modeling in neonatal drug development studies. PopPK uses nonlinear mixed‐effects modeling and allows for the analyses of sparse PK data from pediatric populations integrated with adult PK data. This approach allows for the identification of covariates (e.g., age, weight, renal function) affecting PK parameters and supports dose individualization in different pediatric age groups. The PopPK model that includes influential covariates can be used for simulations to enable extrapolation of adult PK data to the pediatric population.

Age‐dependent physiological differences (e.g., organ maturation, enzymatic activity, changes in tissue composition, and body size) on drug disposition are also incorporated in PBPK modeling, which integrates detailed physiological and anatomical data with drug‐specific properties to predict PK outcomes across age groups [31]. Pediatric PBPK modeling can be used as complementary to other modeling techniques as well. For example, a population PK model supplemented with a PBPK model was used to support the regulatory approval of valganciclovir in infants < 4 months of age. With this complementary approach, an existing PBPK model was extended to neonates and used to confirm physiological similarity between young infants and older children with parameters important for valganciclovir.

Children represent a unique subset due to their age‐related variations in organ maturity, enzyme activity, immune system responsiveness, and receptor sensitivity. PD modeling can contribute to substantial evidence of effectiveness in situations where clinical endpoints or surrogate markers are well‐established. PopPD modeling plays a key role by identifying appropriate dose regimens and sources of variability in drug response. Population‐based PD models have influenced the design of pediatric clinical studies, optimizing pediatric drug regimens before large‐scale clinical trials are conducted.

Mechanism‐based PD models are particularly important in pediatric drug development because they incorporate age‐specific processes on the causal pathway between drug administration and pharmacological effects. These processes may vary due to changes in target affinity and activation during development or due to variability in biologic cascades governing the pharmacological response, which evolve as children age. Such models outperform traditional empirical approaches by accounting for these developmental processes, enabling more reliable extrapolation and prediction of drug responses in pediatric subgroups. In addition, PD modeling in pediatrics requires age‐appropriate endpoints and biomarkers to assess drug efficacy and safety accurately.

DPM uses mathematical functions to represent the time‐dependent changes in disease status quantitatively. Since the last two decades, DPMs have become instrumental in understanding disease progression, optimizing drug dosing regimens, and assessing the efficacy of novel therapies. These models integrate biomarkers that reflect disease severity or clinical outcomes, enabling researchers to understand the natural progression of a condition. There are three main types of DPMs: (1) Empirical Models: data‐driven without detailing biological processes; (2) Semi‐Mechanistic Models: combine data with basic biological understanding; and (3) Systems Biology Models: highly detailed, focusing on cellular and biochemical interactions [4]. In the context of rare disease drug development, DPMs are often coupled with PK‐PD models to evaluate the impact of drug treatment on disease progression. Additionally, DPMs are used alongside natural history data to predict the trajectory of a disease in scenarios where no therapy is administered [32]. DPMs enhance clinical trial designs by identifying key factors like baseline characteristics and biomarkers that could serve as surrogate endpoints. For example, in childhood‐onset dystrophinopathy, DPMs linked genetic variations to loss of ambulation, highlighting the importance of genetic information in trial designs [33]. Innovative approaches to data sharing are critical for maximizing the utility of individual patient data [34]. Developing systems that allow qualified researchers access to robust datasets will enhance the efficiency and accuracy of DPMs, thereby elevating their role in facilitating clinical advancements. Such initiatives will not only improve modeling outcomes but also enable greater collaboration within the scientific community.

4. The Core Strengths of MIDD in Pediatric Rare Diseases

The case studies exemplified in the section above highlight some of the inherent complexities, either in data or in the integration of information. We posit that MIDD integrates quantitative modeling and simulation techniques to enhance decision‐making throughout the drug development process. In the context of rare pediatric diseases, MIDD offers significant advantages by addressing challenges such as limited patient populations, ethical considerations, and gaps in understanding disease biology and treatment effects. Table 1 summarizes additionally selected drugs approved in pediatric rare diseases under each of those categories where MIDD played a pivotal role.

TABLE 1.

Selected examples of MIDD advantages in pediatric rare diseases.

Category	Drug	Indication	Challenges and solutions	MIDD approach	References
Dosing and treatment optimization	Adalimumab	Indication expansion for treatment of moderately to severely active ulcerative colitis (UC) to include pediatric patients aged ≥ 5 years with a body weight of ≥ 20 kg	Transition from body weight–based (mg/kg) dosing to fixed‐dose tiers by weight ranges. Incorporation of a less frequent maintenance dosing option. Adjustments to dosing schedule beyond those studied in clinical trials	Approved pediatric UC dosing regimen based on model‐informed bridging strategy. Avoided additional clinical trial through robust scientific and regulatory evidence enabled by MIDD	[35]
Dosing and treatment optimization	Naxitamab	Indicated in combination with granulocyte‐macrophage colony–stimulating factor for the treatment of pediatric patients aged ≥ 1 year and adult patients with relapsed or refractory high risk neuroblastoma in the bone or bone marrow	FDA recommended a dosing cap of 450 mg per cycle for patients weighing > 50 kg. PK simulations ensured exposure comparability across weight groups (> 50 kg vs. 30–50 kg). Weight‐based dosing of 3 mg/kg/day (9 mg/kg per cycle) used for patients 30–50 kg	MIDD approach utilized clinical PK, efficacy, and safety data integration for dosage selection. Dosing regimen for high body weight patients (> 50 kg) determined without additional trials	[36, 37]
Dosing and treatment optimization	Selpercatinib	Treatment of (1) adult patients with metastatic rearranged during transfection (RET) fusion‐positive non–small cell lung cancer; (2) adult and pediatric patients aged ≥ 12 years with advanced or metastatic RET‐mutant medullary thyroid cancer who require systemic therapy; and (3) adult and pediatric patients aged ≥ 12 years with advanced or metastatic RET fusion‐positive thyroid cancer who require systemic therapy and who are radioactive iodine refractory	Proposed dosing regimen was 160 mg twice daily for the general population. Modeling predicted QT‐prolongation risk with 160 mg in patients weighing < 50 kg. Dose reduced to 120 mg twice daily for patients < 50 kg to mitigate QT‐prolongation risk. Reduced dose maintained steady‐state exposure consistent with clinical data. Final recommendation: 120 mg twice daily for patients < 50 kg and 160 mg twice daily for patients ≥ 50 kg	MIDD utilized modeling and simulation to optimize dosing, ensuring safety by mitigating QT‐prolongation risk while maintaining therapeutic exposure levels	[36, 38]
Dosing and treatment optimization	Selumetinib	Treatment of pediatric patients aged ≥ 2 years with neurofibromatosis type 1 who have symptomatic, inoperable plexiform neurofibromas	Recommended dosing: 25 mg/m2 twice daily, rounded to 10‐ and 25‐mg capsule strengths. Population PK modeling revealed lower exposure in the smallest BSA group (0.55–0.69 m2). FDA recommended a daily dose adjustment to 30 mg for BSA 0.55–0.69 m2 (20 mg morning, 10 mg evening). Adjusted dosing ensures exposure consistency across BSA groups	MIDD utilized population PK modeling to identify exposure discrepancies and supported dose adjustments to ensure consistent therapeutic levels across all BSA groups	[36, 39]
Streamlining efficient trial design	Tecovirimat	First antipox viral small‐molecule drug approved in the United States	Tecovirimat developed under the Animal Rule due to the infeasibility of clinical efficacy trials for smallpox. Efficacy evaluated using lethal animal models (cynomolgus monkeys and New Zealand white rabbits). PK and safety studies conducted in healthy human adults. MIDD approaches guided dose selection for the general population and subgroups (e.g., pediatrics, individuals > 120 kg). Population PK modeling predicted human exposures exceeding effective doses in monkeys while staying below safety limits based on the most sensitive species	MIDD facilitated optimal human dose selection by bridging efficacy data from animal models and human PK/safety data, addressing limitations of traditional clinical trials for smallpox	[36, 40]
Streamlining efficient trial design	Adalimumab	Expansion of indication to adolescent patients aged ≥ 12 years with hidradenitis suppurativa (HS), weighing at least 30 kg	HS prevalence: 0.1% in U.S. adults, significantly lower in pediatric patients (< 0.114% across age groups). Clinical trials for adolescent HS patients would be highly time‐consuming due to low prevalence. Disease similarity and comparable adalimumab response between adults and adolescents enabled efficacy extrapolation. Approval supported primarily by a model‐informed approach without new clinical PK or efficacy data for adolescents	MIDD enabled efficacy extrapolation from adult data to adolescents with HS, avoiding the need for new clinical trials and expediting approval despite low patient prevalence	[36, 41]
Streamlining efficient trial design	Avalglucosidase alfa	Treatment of patients aged ≥ 1 year with late‐onset Pompe disease (LOPD)	Pivotal trial eligibility included patients aged ≥ 3 years, but youngest enrolled was 16 years. Pediatric patients with LOPD underrepresented in clinical data due to extremely low prevalence (< 0.003%). Separate pediatric clinical trials for LOPD deemed infeasible. MIDD approaches enabled extrapolation of efficacy for pediatric patients using PK data. Similar treatment responses across ages justified pediatric indication without new clinical trials	MIDD enabled pediatric LOPD indication avoiding the need for new pediatric trials	[36, 42]
Bridging disease and treatment gaps	Fosdenopterin	Approved for reducing the risk of mortality in patients with molybdenum cofactor deficiency (MoCD) type A.	Rare fatal disease with ≈50 U.S. patients (mostly < 10 years); multiple trials infeasible. Evidence of effectiveness supported by one trial plus data from biomarkers and animal models. Exposure–response analysis linked fosdenopterin exposure to reduced urinary SSC levels. Biomarker trends confirmed dose/exposure‐dependent effectiveness of fosdenopterin	MIDD leveraged biomarker data, exposure–response analysis, and animal model evidence to demonstrate fosdenopterin's effectiveness, overcoming the challenge of limited patient population for multiple trials	[36, 43]
Bridging disease and treatment gaps	Lumasiran	Indicated for primary hyperoxaluria type 1 to lower urinary oxalate levels in pediatric and adult patients	Modeling and simulation guided maintenance dose selection and dosing for patients < 20 kg. PD activity linked more closely to liver PK than plasma PK. Human liver PK estimated using allometric PK/PD scaling from animal data	MIDD utilized allometric scaling to estimate liver PK and guided dose optimization for patients < 20 kg, aligning dosing with PD activity	[36, 44]

4.1. Optimizing Rare Disease Trials Using the MIDD Approach

The MIDD approach has significantly reduced the reliance on large, resource‐intensive clinical trials by enabling innovative design strategies, particularly for rare and pediatric diseases. In the case of pediatric trials for rare diseases, the use of modeling and simulation is crucial to optimizing clinical studies, such as determining sample sizes, starting doses, timing of sampling, and the number of samples. The FDA's draft guidance on pediatric rare diseases highlights the importance of integrating modeling strategies to streamline the rationale for dosing decisions and overall study design [45]. Population PK and PBPK models are frequently employed to establish initial dose regimens for pediatric trials, aiming to achieve drug exposure levels that mirror those found effective and safe in adult populations.

One foundational assumption underlying this “exposure‐matching” strategy is that pediatric patients share similarities with adults regarding disease pathophysiology, the drug's mechanism of action, and exposure–response relationships for efficacy and safety [46]. Additionally, optimizing sampling techniques within pediatric trials helps reduce burdens on the small pool of available participants. By leveraging clinical trial simulations, ideal PK sampling windows can be identified to minimize the volume of collected blood samples while ensuring accurate estimates of PK parameters [47]. These innovations enhance logistical flexibility, reduce patient burden, and ensure robust data generation.

As randomized, placebo‐controlled trials are often infeasible for rare diseases due to limited patient availability, natural history data becomes an invaluable asset. Such data informs drug development at every stage, from initial discovery through clinical trial design, and even into post‐marketing phases. Disease progression models integrated with natural history datasets are powerful tools to simulate trial designs. In certain cases, untreated natural history data can serve as external control groups, reducing the number of trial participants exposed to control treatments and lowering ethical concerns around placebo administration.

Recognizing the role of quantitative modeling and simulation in rare disease drug development, under specific conditions outlined by the Food and Drug Administration Modernization Act, the FDA may grant drug approval based on one robust clinical trial supplemented by additional confirmatory evidence [48]. PD biomarkers and exposure–response (E–R) relationships strengthen this evidence and provide key insights into a drug's effectiveness. PD markers serve as a supplementary indicator of efficacy, while validated E–R models further bolster assumptions supporting pediatric extrapolation for drug approvals [49]. An application of E–R modeling to simplify trial designs is exemplified in Paricalcitol, a drug used in adults and pediatric patients (aged 10 years or older) for the prevention and treatment of secondary hyperparathyroidism associated with chronic kidney disease stage. Pediatric evaluations were performed through a single‐arm study involving 13 patients, showing a meaningful response rate of 61.5%. Though the trial lacked a control arm, the exposure–response relationships of biomarkers (intact parathyroid hormone, serum calcium, and serum phosphorus) supported the effectiveness of Paricalcitol in pediatric patients, ultimately contributing crucial evidence for regulatory approval [50].

4.2. Informing Dose Selection, Regimen Optimization, and Individualized Treatment Approaches

The MIDD approach is instrumental in optimizing drug dosing regimens, especially for scenarios where clinical trial data is insufficient, such as rare disease studies with limited patient populations. Its application extends to diverse therapeutic areas, addressing challenges like dose refinement in pediatric subgroups, individuals with organ impairment, or patients on multiple medications, where conducting extensive dose‐finding studies is often infeasible.

In these instances, MIDD facilitates data‐driven decision‐making by integrating modeling and simulation strategies that leverage clinical and nonclinical evidence. Late‐phase dose optimization often employs model‐based analyses to validate dosing regimens that were not directly studied in trials, ensuring that the chosen regimen optimizes the benefit/risk profile, enhances adherence, or simplifies treatment complexity. Population PK modeling is frequently employed to identify dosing strategies that achieve target therapeutic exposures while minimizing safety risks by maintaining exposure within safe limits. Additionally, PBPK modeling proves valuable in optimizing doses for scenarios involving drug–drug interactions.

MIDD approaches have become increasingly pivotal for emerging therapeutic modalities like antisense oligonucleotides and small interfering RNA therapies, approved for rare diseases such as Duchenne muscular dystrophy and hereditary transthyretin amyloidosis [51]. These methods informed dose selection and regimen optimization, underscoring their value in modern drug development.

In pediatric dosing, exposure‐matching techniques are often utilized to extrapolate from adult data; however, these methods have shown limitations, with variability in PD frequently exceeding that observed in PK, thereby compromising efficacy predictions [46]. Alternative model‐based approaches integrate adult and pediatric PK/PD data to simulate optimal dosing strategies, offering flexible solutions that account for interstudy variability and individual differences.

As pediatric drug development advances, regulatory frameworks increasingly recognize the importance of mechanism‐based modeling, such as PBPK or disease‐specific PK/PD models, for dose characterization. E–R analysis serves as a critical tool to optimize dosing strategies based on efficacy and safety thresholds [50]. E–R assessments also aid in developing new formulations, routes of administration, or dosing regimens, ensuring comparable clinical outcomes without extensive new trials. For example, E–R analysis has supported transitions from body surface area‐based dosing to simplified weight‐based regimens by demonstrating equivalent efficacy and safety profiles (Table 1).

While E–R evaluation is invaluable, challenges persist, including variability caused by intrinsic and extrinsic factors like age or disease status, narrow exposure ranges, and small sample sizes common in pediatric trials. Accurate placebo response assessments are also crucial for certain conditions. Despite these limitations, MIDD, combined with advanced pharmacometric tools, continues to enhance drug development efficiency, guiding dosing decisions and pediatric trial designs with the robustness of data‐driven predictions.

4.3. Role of MIDD in Bridging Gaps in Understanding Disease Biology and Treatment Effects

PK and PD profiles derived from adult clinical trials can be leveraged in Clinical Trial Simulations to evaluate the potential impact of study designs on virtual pediatric populations. By integrating pharmacometric and statistical models, these simulations provide valuable insights to optimize pediatric trial designs [4].

Efficacy extrapolation plays a critical role in pediatric drug development, regulatory approvals, and labeling [46]. This approach relies on data from adult studies or other pediatric populations to model and predict responses within specific pediatric subpopulations. Advanced pharmacometric techniques, which model the interactions between developmental physiology, PK, PD, and disease progression, are increasingly utilized by clinicians, drug developers, and regulatory authorities to streamline pediatric studies when direct data are scarce.

Ideally, clinical outcomes used in adult and pediatric studies for the same disease are consistent and robust, enabling the definition of PD dose–exposure–response relationships with fewer pediatric patients [50]. However, challenges such as limited patient availability or infrequent clinical events in children often make it impractical to directly apply adult endpoints in pediatric trials without necessitating untenably large sample sizes to achieve statistical power. MIDD methodologies effectively address these gaps by leveraging existing data to enhance the success and feasibility of pediatric studies.

Given the limited availability of pediatric patients and the logistic challenges of collecting comprehensive data, pharmacometric modeling and simulation techniques have proven indispensable. These approaches improve clinical trial efficiency, increase the likelihood of regulatory approval, and enable individualized therapy design even in the absence of dedicated clinical trials. Historically, pediatric dose adjustments were often based on simple linear scaling from adult populations using weight or body surface area as a scalar. However, advanced PK/PD modeling now provides a more nuanced approach to address these limitations.

PK/PD modeling also contributes to various aspects of clinical trial design. It supports optimal design strategies by identifying the most informative sampling schedules, the appropriate number of samples per participant, and optimal sample sizes [47]. Dedicated software tools aid in these processes. Aspects like patient acceptability and trial logistics are critical considerations during the design phase. Simplified trial designs are often preferred, provided they maintain scientific rigor, although opportunistic sampling should be approached cautiously to ensure methodological appropriateness for the drug or analyte being studied.

Incorporating prior knowledge, such as anticipated placebo effects and study attrition rates, further refines PK/PD modeling applications. Early involvement of pediatric patients and their families in the trial planning process—commonly referred to as patient‐public involvement—also plays a key role in guiding study design and ensuring ethical considerations are upheld. Such engagement provides valuable insights not only for planning but also when obtaining approval from Institutional Review Boards or ethics committees, ultimately facilitating trials that are both scientifically robust and ethically sound.

5. Areas for Further Development

5.1. Age‐Related Differences in Pharmacokinetics and Pharmacodynamics

Pediatric PK and PD present unique challenges due to age‐related physiological and developmental differences affecting drug disposition and action. Postnatal growth impacts processes such as enzyme maturity in metabolism (e.g., phase I and II enzyme activities), body composition changes (e.g., water and lipid partitioning), receptor expression, growth rate, and organ functionality [52]. These factors collectively influence PK variability across pediatric age groups. Population PK/PD modeling plays a key role in pediatric trial design, enabling reliable estimation of parameters from sparse or rich sampling. It enhances extrapolation by leveraging knowledge from adult trials or older pediatric populations to predict drug exposure and response in younger populations. Similarly, simulation scenarios support the exploration of optimal dose regimens across age ranges, addressing developmental changes in PK and PD. Extrapolation relies on the assumption that the underlying disease mechanism and drug action are similar across adult and pediatric groups, which requires consideration of developmental PK/PD changes to improve the quality and success of extrapolation.

5.2. Models to Address Variations in Disease Presentation and Progression in Rare Diseases

Rare pediatric diseases add unique complexity due to small patient populations, uncharacterized disease progression, and age‐dependent drug responses. For example, DMD involves the same primary genetic defect in all patients, but its progressive muscle‐wasting nature significantly impacts the pharmacodynamics of drugs targeting skeletal muscle, such as etiplersen or ataluren [53]. The effectiveness may decline with age as target tissue is gradually lost.

Model‐informed approaches aim to overcome such challenges by integrating natural history studies, predictive clinical endpoints, and biomarker development to assess drug response characteristics at early disease stages. These collaborative efforts identify sensitive and reliable outcome measures tied to future quality‐of‐life improvements, providing strong PD endpoints for modeling.

In eosinophilic esophagitis, disease presentation varies significantly across different pediatric age groups. For instance, infants may experience feeding dysfunction, school‐aged children may exhibit reflux‐like symptoms, and adolescents may report dysphagia. Such variability complicates the development of patient‐reported (PRO) and proxy‐reported outcome measures essential to capturing PD efficacy across age ranges. Validated instruments addressing these developmental differences are necessary to improve extrapolation and optimize treatment evaluation in pediatric populations [54].

5.3. Use of Caregiver‐Reported Outcomes for Accelerated Approvals

There is a clear need for developing additional therapies for rare pediatric diseases. However, numerous challenges remain in conducting conclusive and confirmatory clinical trials. These challenges include small and dispersed patient populations, high disease heterogeneity, limited natural history data, and complex logistical issues. Regulatory decisions face the challenge of balancing the need for accelerated registration and patient access to novel therapies with promising efficacy and safety from early clinical trials, often based on surrogate biomarkers, against the necessity for confirmatory benefit–risk characterization from long‐term and late‐stage trials. These later trials can delay patient access, even though earlier intervention is crucial for modifying disease trajectory and reducing fatality.

Approaches that consider the totality of evidence, particularly one informed by caregiver‐reported treatment outcomes, are beneficial when considering the patient at the center of the drug development process. This approach integrates various sources of evidence, including clinical trials, real‐world data, and patient‐reported outcomes (PROs), to provide a comprehensive understanding of treatment effects. Caregiver‐reported outcomes are especially valuable in evaluating patients' physical functioning and response to treatment, as caregivers often observe daily changes and challenges that patients face [55, 56]. These insights significantly contributed to the weight of evidence in the case of delandistrogene moxeparvovec gene therapy, where the caregivers' perceived change in patient disease status or severity was quantified using the Caregiver Global Impression of Change and Severity (CaGI‐C and CaGI‐S, respectively) [29]. Such caregiver‐reported outcomes should be encouraged in drug development programs, wherever possible.

5.4. Use of Longitudinal Data and Patient‐Reported Outcomes in Modeling

Longitudinal data and PROs are increasingly central to improving pediatric drug development and fostering patient‐centric approaches. PROs allow direct communication of patients' experiences, symptoms, and disease burdens, providing insights into quality‐of‐life impact that traditional clinical measures may miss. For example, registries capturing real‐world PRO data on diseases like DMD enable the assessment of natural disease history and its interaction with therapeutic interventions. These registries help inform dose selection and confirm effectiveness [4].

Longitudinal data (both individual patient and aggregated population data) brings additional value by tracking disease progression over time, thereby improving predictive models summarizing trial outcomes and efficacy measures [57]. In accelerated development programs, where phase I trials may include patient populations, PRO integration helps identify phase II dose selection while reflecting symptom severity and quality‐of‐life metrics.

Innovative trial designs leveraging sparse data collection and population PK/PD modeling address ethical barriers and small sample sizes prevalent in pediatric trials. Although sparse sampling minimizes biosample collection, its disadvantages include the need for larger sample sizes and reliance on modeling for analysis. Incorporating PROs into modeling frameworks results in a more comprehensive understanding of treatment effects, enabling tailored dose optimization and enhancing trial success rates.

From a data analysis perspective, longitudinal clustering is a new family of computational methods that organize patients into groups (clusters) based on similarity in their time‐series data [58]. Unlike simple cross‐sectional clustering that uses single time‐point data, longitudinal clustering captures the temporal dynamics of disease progression and uncovers patterns in trajectory shapes. These methods lend themselves well to identifying subgroups that otherwise may not be identified with standard methods and present an opportunity in pediatric rare diseases where limitations of sample size prevent the application of standard signal detection tools. By identifying clusters associated with different progression rates or severity, we can better predict individual patient outcomes.

Pediatric rare disease data are limited by small sample sizes, irregular visit schedules, and missing values. Longitudinal clustering methods are designed to handle such real‐world clinical data complexities, making them well suited for rare disease research. Longitudinal clustering represents a transformative approach in pediatric rare disease research, allowing the detection of distinct disease progression phenotypes from complex, temporal clinical data.

By addressing these challenges—age‐related PK/PD differences, variations in rare disease presentation, and the integration of PROs and longitudinal data—modeling approaches improve the efficiency and effectiveness of pediatric drug development while minimizing exposure to unnecessary trials and optimizing therapeutic strategies.

5.5. MBMA for Pediatric Rare Diseases Remains Work in Progress

MBMA is a powerful quantitative methodology that can integrate data from multiple sources, including clinical trials and observational studies, in order to model disease progression (placebo or stand‐of‐care response) or drug efficacy and safety over time. MBMA can be used as a benchmarking model and as a comparison tool to visualize differentiation amongst other treatment options. It allows for the prediction of dose–response relationships, optimization of clinical trial designs, and informed decision‐making throughout the drug development process. However, while MBMA has proven valuable in many therapeutic areas, its application to rare diseases presents unique and significant challenges. The first issue is in the scarcity of datasets. In order to quantify sources of variability and make reliable predictions, MBMA (like all modeling tools) relies on robust data comprising numerous clinical trials or studies with comparable endpoints. In rare diseases, clinical outcomes data are often scarce because these patients are inherently rare, which constrains the number and size of clinical trials. MBMA has been more impactful in more common diseases, where many clinical study results are reported. A lack of historical studies or failed trials in rare diseases means there is limited data on which to develop an MBMA. Thus, any variability in rare disease study design, patient characteristics, and endpoints is difficult to explain with a parametric model. Without sufficient, standardized data, the assumptions underlying MBMA become fragile, increasing uncertainty in model outputs and reducing confidence in their predictive value. In that context, even when multiple studies are available across a given treatment, for example, heterogeneity in disease criteria can be a significant obstacle. In rare diseases, variability is more pronounced because of differences in diagnostic criteria and disease definitions across regions and time, a lack of universally accepted clinical endpoints or biomarkers, and variability in treatment regimens, including off‐label use and compassionate care protocols. Such heterogeneity makes it difficult to pool data meaningfully or to apply standard MBMA models without introducing bias or requiring substantial model adjustments, which can reduce reliability and interpretability.

In newer approvals, rare disease trials often involve single‐arm studies, historical controls, or adaptive designs, as placebo control arms may be unethical or infeasible. This limits the number of comparative trials that MBMA typically utilizes. However, MBMA may still be a valuable tool to develop a synthetic control arm for comparison. Nevertheless, due to the severity and heterogeneity of many rare diseases, outcomes can be highly variable and influenced by patient‐specific factors not accounted for in traditional MBMA frameworks.

In our opinion, MBMA holds promise for accelerating and optimizing drug development, and while its utility in rare diseases is constrained by limited data, heterogeneous study designs, and ethical constraints on trial design, it still remains a powerful tool. To overcome these challenges, innovative data sharing and standardization practices (See Section S2), harmonization of outcome measures, and collaboration across industry, academia, and regulatory bodies are essential. Until then, the application of MBMA in rare diseases must be approached with caution, rigorous validation, and clear communication of assumptions and uncertainties.

6. Vision for the Future: Rethinking Randomized Clinical Trials (RCT) in Pediatric Disease Drug Development

RCTs are widely regarded as the gold standard for evaluating the safety and efficacy of new drugs [59]. Their structured design, involving random assignment and control groups, aims to minimize bias and ensure high internal validity. However, in the context of ultra‐rare pediatric diseases, RCTs present a host of ethical, logistical, and scientific challenges. These diseases often have small, heterogeneous patient populations, high morbidity, and no existing treatments. We believe that RCTs may not be the optimal approach for drug development in this setting, and we focus on ethical considerations, practical limitations, scientific constraints, and the availability of alternative study designs.

The ethical issues surrounding the use of RCTs in ultra‐rare pediatric diseases relate to the use of placebo. In diseases with no approved therapies and life‐threatening or severely debilitating outcomes, assigning a child to a placebo group may be perceived as denying access to potentially life‐saving treatment. Conducting RCTs requires a sufficiently large and relatively homogeneous population to ensure statistical power and meaningful subgroup analyses. However, ultra‐rare pediatric diseases often affect only a few dozen individuals worldwide. Recruiting enough participants for a statistically valid RCT is therefore nearly impossible, particularly if inclusion criteria are narrowly defined to ensure comparability across subjects.

Additionally, the infrastructure and cost associated with multicenter international trials, often the only way to reach the sample size needed, pose another significant hurdle. These trials require regulatory coordination across countries, ethical approval from multiple institutions, and extensive financial investment. For many small biotech firms or academic institutions developing treatments for ultra‐rare diseases, these requirements are prohibitive. Furthermore, the heterogeneity in disease progression and phenotype in ultra‐rare pediatric disorders complicates efforts to standardize endpoints and compare outcomes meaningfully across trial arms. This limits the interpretability and generalizability of results even if an RCT could be conducted.

6.1. Scientific Limitations and Risk of Inconclusive Results

The scientific validity of RCTs in ultra‐rare pediatric disease contexts is frequently undermined by sample size limitations. Small trials have limited power to detect statistically significant differences, increasing the risk of Type II errors—failing to detect a true effect. Even when treatment is effective, an underpowered trial may not demonstrate its benefit, potentially delaying or denying approval. Moreover, RCTs rely on clearly defined clinical endpoints, which may not exist for many ultra‐rare diseases due to their novelty or variable presentation. In such cases, the trial may rely on surrogate endpoints that are not well‐validated, weakening the trial's conclusions. The rigid structure of RCTs can also prevent adaptive learning during the course of a study. In the context of rapidly evolving scientific understanding of a rare disease, the inflexibility of a traditional RCT may hinder integration of new insights that could improve study design or patient care.

6.2. Alternatives to RCTs and Path Forward

Given these challenges, alternative trial designs should be considered for drug development in ultra‐rare pediatric diseases. These include single‐arm trials with historical controls, N‐of‐1 trials, adaptive designs, and platform trials. While these designs may have lower internal validity compared to RCTs, they can offer credible and ethically sound evidence, particularly when supplemented with rigorous data collection and robust statistical methodologies. Regulatory bodies such as the FDA and EMA have increasingly recognized the limitations of traditional RCTs in this context. Initiatives like the FDA's Accelerated Approval Program (https://www.fda.gov/drugs/nda‐and‐bla‐approvals/accelerated‐approval‐program) and the EMA's PRIME scheme (https://www.ema.europa.eu/en/human‐regulatory‐overview/research‐development/prime‐priority‐medicines) encourage the use of alternative data sources and real‐world evidence to expedite drug development for rare conditions.

Ultimately, drug development for ultra‐rare pediatric diseases requires a paradigm shift that values ethical imperatives and practical realities alongside scientific rigor. The field must embrace methodological flexibility, patient‐centric design, and global collaboration to ensure that children with these devastating conditions are not left behind due to outdated regulatory expectations.

6.3. Call for Action: Defining the MIDD Approach to Weight of Evidence as a Substitute of RCT

In the development of drugs for ultra‐rare diseases, synthetic data can serve as a crucial substitute for control arms in randomized clinical trials (RCTs), helping to overcome challenges like small patient populations and ethical constraints. Table 2 lists the types of synthetic data that can be used. Some of the key considerations for advancing synthetic data include data standardization, where data must be curated, standardized, and validated to ensure reliability. They should also address bias and confounders, which will require discussion and harmonization requiring statistical techniques (e.g., propensity score matching) to mitigate bias. Finally, there must be alignment on regulatory acceptance.

TABLE 2.

Approaches to consider a “Reduced RCT” using MIDD.

Type	Definition	Use	Source	Key consideration
Historical control data	Data from past patients who received standard of care or no treatment	Serves as a comparator group when a new drug is tested in a single‐arm trial	Natural history studies, medical records, disease registries, published literature	Differences in data collection methods, changes in standard of care over time, and population mismatch
Natural history data	Longitudinal observational data documenting disease progression in untreated patients	Establishes a baseline trajectory of disease, useful for demonstrating that a new therapy alters the natural course	Patient registries, research consortia, or prospective cohort studies	Especially useful in pediatric conditions with predictable decline or mortality
External control arms (real‐world data)	Patient‐level data from external sources used to construct a non‐randomized control group	Enables comparisons without enrolling a placebo group	Electronic health records (EHR), insurance claims, or patient‐reported outcomes from prior studies	Increasingly accepted by the FDA and EMA in certain contexts, especially with rigorous matching
Digital twins/simulated control arms	Statistical or AI‐generated “digital replicas” of patients that simulate what would happen under standard care	Model the disease trajectory without intervention for a virtual control arm	AI/machine learning trained on historical or observational data	No need to withhold treatment from real patients
Bayesian priors and hybrid designs	Prior information (e.g., historical or expert data) is used to inform statistical models in a trial	In adaptive or Bayesian trial designs, this prior information can reduce the number of control patients needed		Increases efficiency and ethical acceptability of trials
Patient‐generated health data	Health‐related data created and recorded by patients outside of clinical settings	Supplements or supports synthetic control arms, especially in tracking disease symptoms and quality of life	Data from wearable devices, mobile apps, or caregiver logs

7. Summary

Quantitative medicine, including MIDD, systems pharmacology, and advanced statistical modeling, is poised to play a transformative role in the future of pediatric rare diseases drug development. While many of the challenges of small, heterogeneous patient populations and limited clinical data remain, quantitative approaches continue to offer powerful tools to maximize the value of every data point. Looking ahead, we believe that quantitative medicine will enable more precise characterization of disease progression and drug response in pediatric populations by integrating real‐world evidence, patient registries, and caregiver outcomes. Model‐based simulations will support ethically responsible, efficient trial designs, such as extrapolation from adult data, adaptive protocols, and synthetic control arms. These methods will be especially critical where randomized trials are not feasible or ethical. Our vision remains that RCTs are replaced by responsible data‐integrated alternatives.

Conflicts of Interest

All authors are employees of consulting or pharmaceutical companies, and in addition to employment‐based salary may hold stock or stock options in their companies.

Supporting information

Data S1.