Tutorials › Pharmacometrics › Model-Informed Pediatric Drug Development
Pharmacokinetics · Pediatric Drug Development

Model-Informed Pediatric Drug Development

Learn how pharmacokinetic, pharmacodynamic, population PK, PBPK, exposure-response, and simulation models can help translate knowledge across ages, select pediatric doses, design efficient studies, and reduce uncertainty in pediatric drug development.

Intermediate Pediatric PK MIDD Modeling & Simulation Drug Development
01 · The big picture

1. What Is Model-Informed Pediatric Drug Development?

Model-informed pediatric drug development uses quantitative models and simulations to integrate information from adults, older children, younger children, neonates, nonclinical studies, prior clinical studies, pharmacology, and other relevant sources.

The central idea is straightforward: pediatric development should not treat every age group as if it were an entirely new drug-development problem. Existing knowledge can often be used to reduce uncertainty, while pediatric data are collected where they are genuinely needed.

Model-informed approaches are particularly valuable in pediatrics because the available sample sizes may be small, blood sampling may be constrained, developmental changes can alter drug disposition, and it may be ethically or practically undesirable to expose children to unnecessary studies or dose levels.

Existing data adult · pediatric · nonclinical Models PK · PD · PBPK · PopPK Simulation dose · exposure · trials Development decision dose selection · study design · extrapolation · uncertainty Model-informed development is an iterative evidence-integration process.

Model-informed pediatric development combines existing knowledge with targeted pediatric data and quantitative predictions to support development decisions.

Core idea: modeling does not replace pediatric clinical research. It helps determine what information is already sufficient, what uncertainty remains, and which additional studies or measurements are most informative.
02 · Why modeling matters

2. Why Is Modeling Especially Useful in Pediatric Development?

Pediatric drug development spans a population undergoing rapid biological change. Weight, organ function, enzyme expression, transporter activity, body composition, and other physiological characteristics can change substantially from infancy through adolescence.

At the same time, pediatric trials often have practical constraints that are less prominent in conventional adult development. Blood volume can limit sampling, patient numbers may be small, and investigators may need to select a useful dose range before enough pediatric PK data are available to characterize every age group independently.

ChallengePotential model-informed contribution
Limited pediatric sample size Borrow information from adults and other pediatric age groups while accounting for developmental differences.
Sparse blood sampling Use population PK or hierarchical models to estimate individual and population exposure from limited observations.
Rapid physiological maturation Represent age-, weight-, size-, organ-function-, or maturation-related changes quantitatively.
Limited dose-ranging information Simulate candidate regimens before exposing large numbers of children to multiple doses.
Uncertainty about adult-to-child similarity Use PK, PD, exposure-response, disease, and pharmacology information to assess whether extrapolation is scientifically reasonable.
Rare diseases or very small populations Integrate prior information and simulations to make efficient use of limited clinical data.

ICH E11A describes pediatric extrapolation as an iterative process: understand the existing information, identify gaps and uncertainties, and generate additional information where needed. Quantitative modeling and simulation can be part of that process. ICH E11A guideline.

03 · Pediatric extrapolation

3. Pediatric Extrapolation: What Are We Actually Trying to Carry Forward?

Pediatric extrapolation is not simply the act of taking an adult dose and scaling it by body weight. The scientific question is whether existing evidence can reduce the amount or type of new information required in a pediatric population.

The information being extrapolated may include efficacy, safety, pharmacokinetics, pharmacodynamics, exposure-response relationships, disease progression, or other aspects of the benefit-risk assessment.

Reference population adults / older children Evidence PK · PD · disease · efficacy safety · pharmacology Pediatric target population

Extrapolation asks which aspects of existing evidence remain applicable to the pediatric target population and what additional information is required.

ICH E11A emphasizes consideration of the similarity of disease, drug pharmacology, and treatment response between a reference population and the pediatric target population. The amount of new pediatric evidence should reflect the remaining uncertainty. EMA summary of ICH E11A.

Important distinction: pediatric extrapolation is a scientific strategy, not a single statistical method. Modeling and simulation are tools that can support the extrapolation strategy.
04 · Growth and maturation

4. Why Age Is Not Enough

A child's chronological age is informative, but age alone is rarely a sufficient mechanistic description of drug disposition.

Two children of the same age can have different body weights, organ function, developmental stages, disease characteristics, or concomitant medications. Conversely, children of different ages can have similar physiological characteristics relevant to the disposition of a particular drug.

Model-informed approaches therefore frequently consider several covariates or physiological processes simultaneously.

FactorPotential relevance to PK
Body weight or sizeMay influence clearance, volume, and dose requirements through size-related scaling.
AgeCan serve as a proxy for developmental processes but does not uniquely define maturation.
Organ maturationChanges in renal and hepatic function can alter drug clearance.
Enzyme ontogenyDevelopmental changes in metabolic enzyme activity can affect clearance.
Renal maturationGlomerular filtration and tubular processes develop substantially after birth.
Body compositionChanges in water, fat, and lean mass can influence distribution.
Disease statusDisease-related changes can modify physiology and drug disposition independently of age.

These factors are one reason that pediatric PK models may use mechanistic maturation functions, allometric relationships, covariate models, or PBPK representations rather than relying on a simple age-based dose rule.

05 · Population PK

5. Population Pharmacokinetics in Children

Population pharmacokinetics describes typical PK behavior while simultaneously modeling variability between individuals and residual variability in observations.

A simplified population model can be written as:

$$CL_i=CL_{\mathrm{pop}}\left(\frac{WT_i}{70}\right)^{\theta}\exp(\eta_i)$$

Here, \(CL_i\) is the clearance for individual \(i\), \(CL_{\mathrm{pop}}\) is a typical value, \(WT_i\) is body weight, \(\theta\) describes the size relationship, and \(\eta_i\) represents unexplained between-subject variability.

The exact model used in practice can be much more sophisticated. Age, renal function, maturation, genotype, disease status, concomitant medications, and other covariates may be incorporated when scientifically justified.

Why PopPK matters in pediatrics: sparse sampling can still be informative when observations from many individuals are analyzed jointly through a hierarchical model.

Population PK models can therefore support estimation of individual exposure, identification of covariates, dose selection, simulation of alternative regimens, and evaluation of whether pediatric exposures are consistent with a reference population.

06 · PBPK

6. Physiologically Based Pharmacokinetic Modeling

Physiologically based pharmacokinetic (PBPK) models represent drug disposition using physiological compartments and mechanistic descriptions of processes such as blood flow, tissue partitioning, enzyme activity, transporter activity, and organ function.

In pediatric applications, PBPK can explicitly represent developmental changes in physiology. A conceptual representation is:

$$\text{Drug properties}+\text{physiology}+\text{maturation}\rightarrow C(t)$$

The model can incorporate information about tissue volumes, organ blood flows, renal function, enzyme ontogeny, transporter activity, and other physiological characteristics. The specific implementation depends on the drug and modeling platform.

Population PKPBPK
Typically estimates population and individual PK parameters statistically from observed concentration data. Uses a more mechanistic representation of physiology, drug properties, and disposition processes.
Covariates are incorporated when supported by data and model structure. Physiological covariates and maturation processes can be represented explicitly.
Particularly useful when clinical PK data are available across individuals. Can be useful for prediction in populations or situations where direct clinical data are limited, provided the model is appropriately qualified.

These approaches are not mutually exclusive. The appropriate modeling strategy depends on the scientific question, available information, model assumptions, and intended use.

Regulatory agencies have specific expectations for PBPK model documentation and qualification. The EMA's PBPK reporting guideline describes expectations for reports included in regulatory submissions, including pediatric investigation plans. EMA PBPK reporting guideline.

07 · Exposure matching

7. Exposure: A Central Bridge Between Populations

One of the most important uses of model-informed pediatric development is to compare drug exposure between populations.

Suppose an adult development program has established an exposure range associated with a desired pharmacologic response. A pediatric model can be used to investigate what dose or regimen would produce comparable exposure in children, subject to the assumptions of the model and the scientific evidence supporting the exposure-response relationship.

$$\text{Dose}\rightarrow\text{PK model}\rightarrow\text{Exposure}\rightarrow\text{Response}$$

Common exposure metrics include:

  • AUC — area under the concentration-time curve.
  • Cmax — maximum observed or model-predicted concentration.
  • Cmin — minimum or trough concentration.
  • Average concentration — often relevant for repeated dosing.
  • Time above or below a concentration threshold — useful for drugs where duration of exposure matters.

Exposure matching does not automatically establish equivalent efficacy or safety. The relevance of an exposure target depends on the drug, disease, endpoint, mechanism of action, and quality of the exposure-response evidence.

08 · PK/PD

8. Connecting Pediatric Exposure to Pharmacodynamic Response

A pediatric dose is ultimately intended to produce an appropriate therapeutic effect while maintaining an acceptable safety profile. PK therefore often needs to be considered together with PD or exposure-response relationships.

A simple pharmacodynamic model might be:

$$E(C)=E_0+\frac{E_{\max}C}{EC_{50}+C}$$

Here, \(E_0\) is baseline effect, \(E_{\max}\) is the maximum drug-related effect, and \(EC_{50}\) is the concentration associated with half of the maximum drug-related effect under this model.

The pediatric question is not necessarily whether children have exactly the same concentration-response relationship as adults. Instead, the development program should evaluate whether available evidence supports the assumptions required to use adult or older-pediatric response information for the target population.

PK versus PK/PD: PK modeling asks whether the appropriate exposure can be achieved. PK/PD or exposure-response modeling asks whether that exposure is expected to correspond to the desired pharmacologic or clinical response.
09 · Dose selection

9. Model-Informed Pediatric Dose Selection

Once a pediatric PK model has been developed, simulations can be used to evaluate candidate doses before a study begins.

For example, suppose the target is an exposure distribution established from adults or older pediatric patients. A model can simulate children across a range of body weights and developmental characteristics and calculate the probability that each candidate regimen achieves the desired exposure range.

$$PTA=\Pr(L\leq E\leq U)$$

where \(L\) and \(U\) define a target exposure interval and \(PTA\) is the probability of target attainment.

Candidate regimenPossible modeling question
Fixed doseDoes the same dose produce acceptable exposure across the pediatric weight range?
Weight-based doseDoes mg/kg dosing control exposure variability adequately?
Body-surface-area doseDoes BSA scaling improve exposure prediction for this drug?
Age-adjusted doseIs an additional developmental adjustment supported by the model?
Weight-band dosingCan practical dosing bands provide adequate target attainment?
Loading + maintenance regimenCan different initial and maintenance doses achieve appropriate exposure over time?

The objective is not to identify the mathematically most complex dosing rule. A practical regimen should balance exposure control, formulation constraints, dosing accuracy, adherence, safety, and feasibility.

10 · Simulation

10. Why Simulate Pediatric Patients?

Simulation translates model parameters and uncertainty into predicted outcomes under hypothetical scenarios.

A simulation might generate thousands of virtual pediatric patients with different weights, ages, clearances, maturation states, and residual variability. Candidate dosing regimens can then be evaluated across those virtual populations.

Model parameters + uncertainty Virtual patients age · weight · physiology Regimens dose + interval Simulated exposure AUC · Cmax · Cmin · PTA · variability Development decision select, modify, or reject candidate regimens

Simulation allows candidate pediatric regimens to be evaluated over realistic distributions of patient characteristics and model uncertainty.

Simulation is especially useful when direct experimentation with every candidate dose would be impractical or ethically undesirable.

11 · Trial optimization

11. Using Models to Optimize Pediatric Trials

Model-informed methods can influence not only the dose but also the design of the pediatric study.

  • Sampling design: identify time points that provide the most information about PK parameters.
  • Sample size: explore how many participants may be needed for the intended model or decision.
  • Dose selection: choose doses expected to provide informative and appropriate exposures.
  • Weight bands: evaluate practical dosing categories.
  • Adaptive decisions: update the model as new information becomes available.
  • Trial simulation: evaluate operating characteristics under alternative assumptions.
  • Sparse PK: determine whether limited samples can provide sufficient information.

Optimal sampling designs are particularly useful when blood collection is constrained. Rather than collecting many poorly informative samples, a model can help identify a smaller set of strategically timed measurements.

Practical principle: in pediatrics, the goal is often not to collect more samples but to collect the most informative samples that can reasonably be obtained.
12 · Bridging across ages

12. Bridging Across Pediatric Age Groups

Pediatric development is itself heterogeneous. Adolescents, school-age children, toddlers, infants, and neonates may have substantially different PK characteristics.

A model can provide a quantitative framework for determining how information can be transferred from one pediatric age group to another.

Reference groupTarget groupPotential model-informed question
AdultsAdolescentsAre disease, pharmacology, exposure-response, and PK sufficiently similar to support extrapolation?
Older childrenYounger childrenCan maturation and size explain differences in exposure?
ChildrenInfantsAre developmental changes in clearance and distribution adequately represented?
InfantsNeonatesAre rapidly changing physiological processes and limited clinical data adequately characterized?

The scientific strength of the bridge depends on what is known about the disease, pharmacology, treatment response, and developmental differences. Modeling can quantify some of these differences, but it cannot eliminate uncertainty that is fundamentally unsupported by data.

13 · Worked example

13. Worked Example: Selecting a Pediatric Dose

Consider a hypothetical drug with a well-characterized adult exposure-response relationship. Suppose the adult development program suggests that a daily AUC between 80 and 120 mg·h/L is an appropriate target exposure range.

A pediatric population PK model predicts the following exposure distributions for three candidate regimens:

RegimenMedian AUC10th–90th percentilePTA within 80–120
Regimen A68 mg·h/L42–10131%
Regimen B96 mg·h/L70–12772%
Regimen C135 mg·h/L98–18134%

Step 1: Define the target

The model uses \(80\) to \(120\) mg·h/L as the target exposure interval:

$$80\leq AUC\leq120$$

Step 2: Simulate candidate regimens

For each candidate regimen, the model generates exposure predictions across a simulated pediatric population that represents the intended age and weight distribution.

Step 3: Evaluate target attainment

The probability of target attainment is estimated as:

$$PTA=\frac{\text{number of simulated patients with }80\leq AUC\leq120}{\text{total simulated patients}}$$

Step 4: Interpret the result

The simulation shows that the candidate regimens produce materially different exposure distributions. The modeling exercise therefore provides quantitative evidence about how the dosing options behave across the simulated population.

Important: this example demonstrates the mechanics of model-informed dose evaluation. A real pediatric dose decision would additionally consider efficacy, safety, uncertainty in the model, formulation constraints, developmental biology, and the quality and relevance of the exposure target.
14 · Uncertainty

14. Model Uncertainty Is Part of the Evidence

Model-based predictions are not single deterministic truths. They depend on parameter estimates, covariate distributions, residual variability, structural assumptions, and assumptions about how information transfers between populations.

A useful model-informed analysis therefore asks not only:

“What does the model predict?”

but also:

“How sensitive is the decision to reasonable alternative assumptions?”

Source of uncertaintyExample
Parameter uncertaintyClearance or maturation parameters are estimated imprecisely.
Structural uncertaintyOne- and two-compartment models provide different predictions.
Covariate uncertaintyThe relationship between weight and clearance is uncertain.
Extrapolation uncertaintyAdult exposure-response information may not fully apply to a pediatric target population.
Physiological uncertaintyOntogeny or developmental functions may be incompletely characterized.
Simulation uncertaintyThe simulated population may not perfectly represent the future study population.

Sensitivity analyses can show whether a development decision remains similar across plausible assumptions. This is often more informative than presenting a single model result without its uncertainty.

15 · Model evaluation

15. How Should a Pediatric Model Be Evaluated?

Model evaluation should be linked to the intended use of the model. A model being used to describe existing PK data has different requirements from a model being used to select a pediatric dose or support an extrapolation decision.

Common evaluation activities include:

  • Goodness-of-fit assessment: compare observed and predicted concentrations.
  • Residual diagnostics: evaluate systematic patterns in unexplained variability.
  • Visual predictive checks: compare observed data with prediction distributions.
  • Bootstrap or resampling: evaluate parameter stability where appropriate.
  • External validation: test predictions using data not used for model development when feasible.
  • Predictive checks across age groups: examine whether the model adequately represents relevant developmental strata.
  • Sensitivity analysis: determine which assumptions materially affect predictions.
  • Simulation-based evaluation: examine performance under the intended decision scenario.
Model evaluation should be purpose-specific. A model can be adequate for describing observed concentrations but inadequate for extrapolating into an unstudied neonatal population.

The FDA's 2026 ICH M15 guidance provides general principles for planning, evaluating, documenting, and communicating evidence generated through model-informed drug development. FDA ICH M15 guidance.

16 · Regulatory context

16. Model-Informed Approaches and Regulatory Decision-Making

Model-informed pediatric development is increasingly integrated into regulatory development programs. The purpose is not to create a separate modeling exercise disconnected from the clinical program. Instead, models can become part of the evidence used to justify development decisions.

Important regulatory questions include:

  • What is the intended use of the model?
  • What decision will the model inform?
  • What data support the model?
  • What assumptions are required?
  • How well has the model been evaluated for its intended use?
  • How is uncertainty represented?
  • What additional clinical data are still needed?
  • How will the model be updated when new pediatric information becomes available?

ICH E11A provides a harmonized framework for pediatric extrapolation and emphasizes an iterative approach to identifying evidence gaps and generating additional information. The guideline became effective as an ICH Step 5 guideline in January 2025. ICH E11A at EMA.

ICH M15, finalized in June 2026, establishes general principles for planning, model evaluation, and documentation of MIDD evidence and addresses regulatory interactions and submission considerations. FDA ICH M15.

17 · Practical workflow

17. A Practical Model-Informed Pediatric Development Workflow

  1. Define the development question. Is the goal dose selection, extrapolation, trial optimization, exposure matching, or another decision?
  2. Define the reference and target populations. Specify which population provides information and which pediatric population is the target of inference.
  3. Assemble existing evidence. Include clinical PK, PD, efficacy, safety, disease, pharmacology, nonclinical, and physiological information as appropriate.
  4. Identify developmental differences. Determine which physiological or disease-related processes could change drug exposure or response.
  5. Select the modeling approach. Consider population PK, PK/PD, exposure-response, PBPK, disease models, or combinations of approaches.
  6. Develop and evaluate the model. Assess parameter estimates, diagnostics, predictive performance, plausibility, and uncertainty.
  7. Simulate relevant pediatric scenarios. Include realistic age, weight, physiological characteristics, variability, and candidate dosing regimens.
  8. Evaluate sensitivity. Determine whether reasonable alternative assumptions materially change the development conclusion.
  9. Identify remaining evidence gaps. Decide what additional PK, PD, efficacy, or safety information is needed.
  10. Design the pediatric study. Use model results to inform dose levels, sampling schedules, weight bands, and other design elements.
  11. Update the model. Incorporate new data as the pediatric program progresses.
  12. Use the model for the intended decision. Clearly distinguish observed evidence from model-based inference and communicate uncertainty.

This workflow is iterative rather than strictly linear. New pediatric data can change the model, and a revised model can change what additional data are considered necessary.

18 · Common mistakes

18. Common Mistakes in Model-Informed Pediatric Development

1. Treating weight-based dosing as a model

Simply expressing a dose as mg/kg does not establish that weight is the correct predictor of exposure. A quantitative analysis should evaluate whether weight scaling adequately describes the relevant PK process.

2. Assuming exposure matching proves efficacy

Exposure matching is only sufficient for efficacy extrapolation when the scientific assumptions linking exposure to response are supported.

3. Using age as the only developmental covariate

Age may correlate with maturation, but it does not necessarily represent the physiological process responsible for a change in clearance or distribution.

4. Ignoring uncertainty

A simulated exposure distribution is conditional on the model. Uncertainty in parameters and structural assumptions should be considered when the model informs a consequential decision.

5. Building a model that is more complicated than the data support

Additional compartments, covariates, or mechanistic pathways are useful only when they are identifiable and relevant to the intended purpose.

6. Confusing model fit with predictive performance

A model can fit the data used for development and still perform poorly when predicting a new population or age group.

7. Treating a model as a substitute for clinical evidence

Model-informed development is designed to make better use of available evidence and target new data efficiently. It does not remove the need for pediatric evidence when uncertainty remains clinically important.

19 · Beyond dose selection

19. What Else Can MIDD Support?

Although pediatric dose selection is a major application, model-informed approaches can contribute to many stages of development.

ApplicationExample question
Pediatric dose selectionWhich regimen is expected to produce an appropriate exposure distribution?
ExtrapolationHow much adult or older-pediatric evidence can be transferred to a younger population?
Trial designWhich sampling times and dose levels provide useful information?
Formulation developmentHow might formulation changes alter exposure?
Drug-drug interactionsHow could concomitant medications affect pediatric exposure?
Special populationsHow might renal or hepatic impairment alter pediatric PK?
Exposure-responseWhat exposure range is associated with efficacy and safety?
Clinical pharmacology strategyWhich studies are most informative for reducing remaining uncertainty?

The EMA specifically describes MIDD applications in pediatrics as including dose and regimen selection, clinical trial optimization, extrapolation, and support for pediatric posology. EMA Modelling and Simulation Q&A.

20 · References

20. References

  1. ICH E11A: Pediatric Extrapolation. International Council for Harmonisation, Step 5 guideline, effective January 25, 2025. ICH E11A guideline.
  2. FDA/ICH M15: General Principles for Model-Informed Drug Development. U.S. Food and Drug Administration, Final Guidance, June 2026. FDA M15 guidance.
  3. EMA. Modelling and simulation: questions and answers. European Medicines Agency. Pediatric applications of model-informed drug development, including dose selection, trial optimization, extrapolation, and posology. EMA Modelling and Simulation Q&A.
  4. EMA. Extrapolation of efficacy and safety in paediatric medicine development. Reflection paper on the use of extrapolation in pediatric medicine development. EMA pediatric extrapolation guidance.
  5. EMA. Guideline on the role of pharmacokinetics in the development of medicinal products in the paediatric population. European Medicines Agency. EMA pediatric PK guideline.
  6. EMA. Guideline on the reporting of physiologically based pharmacokinetic modelling and simulation. European Medicines Agency. EMA PBPK guideline.

21. Key Takeaways

  • Model-informed pediatric drug development integrates existing clinical, pharmacological, physiological, and quantitative information to guide pediatric development.
  • Pediatric extrapolation asks how much existing evidence can be transferred to a target pediatric population and what additional evidence remains necessary.
  • Modeling and simulation can reduce unnecessary experimentation while helping identify the pediatric studies that are most informative.
  • Population PK models can quantify typical PK, between-subject variability, and relationships between exposure and patient characteristics.
  • PBPK models can represent developmental physiology and mechanistic drug-disposition processes when the underlying information and model are sufficiently supported.
  • Body weight, maturation, organ function, disease, and other factors may influence pediatric exposure; chronological age alone is not necessarily sufficient.
  • Exposure matching can be an important component of pediatric extrapolation, but exposure matching does not automatically establish equivalent efficacy or safety.
  • Simulation can evaluate candidate pediatric doses and estimate quantities such as exposure distributions and probability of target attainment.
  • Model-informed methods can optimize not only dose selection but also sampling schedules, trial design, and the overall pediatric development strategy.
  • Model evaluation should be appropriate for the intended use, especially when predictions are made in pediatric populations that were poorly represented in the model-development data.
  • Model uncertainty is part of the evidence and should be considered explicitly in consequential development decisions.
  • Model-informed development does not eliminate the need for pediatric clinical evidence when important uncertainties remain.
  • ICH E11A provides a framework for pediatric extrapolation, while ICH M15 provides general principles for model-informed drug development evidence.
Next step

Where to Go Next

A natural progression is to study Model-Based Pediatric Dose Selection, followed by Pediatric Population PK Modeling, Pediatric PBPK Modeling, Developmental Maturation Models, Pediatric Exposure-Response Modeling, and Pediatric Extrapolation Strategies.

The next tutorial can build directly on the concepts introduced here by showing how a population PK model is constructed for a pediatric dataset, how weight and maturation covariates are evaluated, and how the resulting model is used to simulate candidate dosing regimens.

← Back to Pharmacokinetics Tutorials