BPS3012 Chap.2 Model-Informed Drug Development
Model-Informed Drug Development
Model-Informed Drug Development uses quantitative models and simulations to integrate evidence and support development decisions. The chapter compares non-compartmental analysis, population PK/PD and PBPK, then shows how dose, interaction and trial questions determine the appropriate model. Assumptions, validation and uncertainty remain visible throughout.
The starting point is a decision, not a software package.
A development team might need to choose a dose, anticipate variability, understand exposure–response, predict an interaction or design a trial. The right model depends on that question and on the available evidence. Non-compartmental analysis can summarise observed exposure without representing every physiological process. Population PK/PD can describe typical behaviour and between-person variability.
Physiologically based pharmacokinetic modelling represents organs, flows and compound properties to explore mechanistic scenarios. Each approach is useful within a defined purpose.
A disciplined workflow moves from question to model structure, assumptions, inputs, calibration, evaluation and decision.
Assumptions should be inspectable: which processes are simplified, which parameters come from experiments, which are inferred, and which populations are represented? Evaluation then asks whether the model can recover patterns relevant to its intended use. Agreement with the data used to fit a model is not enough.
Confidence strengthens when predictions are compared with informative observations outside calibration and when sensitivity analysis shows which uncertain inputs control the conclusion.
Models integrate evidence but do not manufacture certainty. A simulation result is conditional on structure, data quality and parameter choices.
When two plausible assumptions give different dose recommendations, the disagreement identifies a knowledge gap rather than a reason to select the preferred output. The practical response may be an additional study, a more conservative decision rule or explicit monitoring. State uncertainty in terms of the decision it could change.
This makes the limitation useful instead of appending a generic sentence that every model is imperfect.
For assessment answers, distinguish description, prediction and validation. First name the development decision. Then justify why the model represents the processes or variability that matter. Explain what data inform it, how performance would be challenged and what outcome would alter the decision.
When comparing approaches, use purpose, structure, data burden, interpretability and extrapolation risk as common criteria. The chapter is not asking you to perform calculations; it is asking you to reason clearly about how quantitative evidence can support—but not replace—pharmaceutical judgement.
What this chapter covers
- 01
Decision questions before model choice
- 02
Evidence, assumptions and simulation
- 03
Non-compartmental exposure summaries
- 04
Population PK/PD and variability
- 05
PBPK and physiological mechanisms
- 06
Dose, interaction and trial applications
- 07
Validation domains and uncertainty
Worked example · free
Choose a model for an interaction question
- +1State the decision: interaction risk and possible dose modification.
- +1Select PBPK because physiology and metabolic mechanism drive the scenario.
- +1Map in vitro, clinical and population inputs to evidence.
- +1Validate against interactions already observed.
- +1Report uncertainty and the population or dose domain covered.
- +1Use the result to prioritise a study or decision, not as proof beyond the validated domain.
Key terms
- MIDD
- Model-informed drug development, which integrates evidence through models and simulations to support development decisions.
- Non-compartmental analysis
- A method that summarises observed exposure without a detailed physiological model.
- Population PK/PD
- A modelling approach for typical behaviour, between-person variability and explanatory covariates.
- PBPK
- A physiological modelling approach representing organs, flows and biochemical mechanisms.
- Validation domain
- The population, dose, formulation and mechanism range in which a model has been adequately tested.
Model-Informed Drug Development FAQ
What makes a model useful in drug development?
It must answer a defined decision question, integrate relevant evidence, expose assumptions, report uncertainty and be checked against observations. A technically complex model without a decision fit is not automatically useful. Fitness therefore depends on intended use: dose selection, interaction prediction and trial design may demand different structures and checks.
State which output changes the decision, where inputs came from, which assumptions dominate and what external observation would show that the model has left its reliable domain.
How do population PK/PD and PBPK differ?
Population PK/PD focuses on typical behaviour, variability and covariates across people. PBPK represents physiology and mechanisms, making it useful for scenarios such as drug interactions or extrapolation when supported by evidence. Choose between them from the question rather than prestige.
Population analysis is valuable when variation and covariates drive the decision; PBPK is attractive when physiological mechanisms structure the extrapolation. Both remain conditional on data, assumptions, evaluation and the population or scenario in which the output will be used.
Does a simulation replace a clinical study?
No. A simulation can inform priorities, study design or a bounded decision. Its claim is limited by its structure, parameters, validation data and domain of use, and new observations should update it. Treat disagreement between plausible simulations as information about uncertainty. Identify which assumption controls the decision, then propose the most informative observation or sensitivity analysis.
A model supports judgement when its conditional nature remains visible; it does not transfer responsibility to an equation or software package.
Assessment move
Revise each model type with three prompts: what question does it answer, what evidence does it need and what can invalidate it? Practise short scenarios in which you justify one approach and reject another. Use the audit order of decision, inputs, assumptions, behaviour, external check, domain and next evidence.
