Model-Based Drug Development vs. MIDD

1. What Are MBDD and MIDD?

Model-Based Drug Development (MBDD) and Model-Informed Drug Development (MIDD) describe closely related ways of using quantitative models and simulations during drug development.

The terminology has evolved. In contemporary regulatory usage, MIDD is the broader and more formally defined term. The 2026 ICH M15 guideline defines MIDD as the use of computational modeling and simulation methods that can integrate nonclinical data, clinical data, prior information, and knowledge to generate evidence for drug development and decision-making. :contentReference[oaicite:1]{index=1}

MBDD is often used to emphasize a development strategy in which models are central to planning, interpreting, and optimizing development activities. Depending on the organization and historical context, MBDD and MIDD may be used almost interchangeably. It is therefore important to define the terminology being used rather than assuming that the two labels always describe different technical methods.

2. Why Are There Two Terms?

The language of model-based drug development predates the current harmonized MIDD framework. Over time, organizations, academic groups, pharmaceutical companies, and regulatory agencies have used terms such as model-based drug development, model-informed drug development, model-informed drug discovery and development, pharmacometrics, and modeling and simulation.

Historically, MIDD has been defined by FDA as the development and application of exposure-based, biological, and statistical models derived from preclinical and clinical data to inform drug development and regulatory decision-making. :contentReference[oaicite:2]{index=2}

The current ICH M15 guideline provides a harmonized framework for MIDD evidence, including recommendations concerning planning, model evaluation, documentation, and regulatory interaction. FDA issued the final M15 guidance in June 2026, and the EMA lists the ICH M15 guideline as effective July 23, 2026. :contentReference[oaicite:3]{index=3}

3. What Is MIDD?

Under ICH M15, MIDD involves computational modeling and simulation that can integrate multiple sources of information, including nonclinical data, clinical data, prior information, and knowledge about the drug or disease. The resulting MIDD evidence can then be used to inform drug development and decision-making. :contentReference[oaicite:4]{index=4}

This definition is intentionally broad. MIDD is not a single model, statistical method, software package, or therapeutic area.

For example, MIDD can involve:

• Population pharmacokinetic models.
• Exposure-response models.
• PK/PD models.
• Physiologically based pharmacokinetic (PBPK) models.
• Quantitative systems pharmacology (QSP) models.
• Clinical trial simulations.
• Drug-disease-trial models.
• Mechanistic models of pharmacology or safety.
• Models used to evaluate dosing or therapeutic individualization.

FDA describes MIDD applications as potentially supporting clinical outcome prediction, trial design and efficiency, efficacy assessment, dose optimization, therapeutic individualization, safety evaluation, and product performance optimization. :contentReference[oaicite:5]{index=5}

4. What Does “Model-Based Drug Development” Mean?

The phrase model-based drug development emphasizes the role of models in the overall development process.

Instead of treating a model as an isolated analysis performed after data collection, an MBDD strategy can use modeling prospectively to influence what information is collected, how studies are designed, what doses are evaluated, and how subsequent development decisions are made.

For example, a development program might use an early population PK model to characterize variability, combine it with exposure-response information, and then use simulations to evaluate candidate dose regimens before conducting a larger clinical study.

These are not necessarily competing approaches. In practice, a model-based development strategy may generate MIDD evidence at multiple stages of development.

5. MBDD vs. MIDD: A Practical Comparison

The most useful practical point is that the underlying scientific methods can be identical. The distinction is primarily one of framing, terminology, and how the model-derived evidence is incorporated into development decisions.

6. MIDD Starts With a Question of Interest

A strong MIDD analysis begins with a clearly defined question of interest.

Examples include:

• What dose or dosing interval should be evaluated in the next study?
• What exposure range is associated with the desired pharmacologic effect?
• How should dosing be adjusted for renal or hepatic impairment?
• Can existing data support an alternative dosing regimen?
• How much information is required from a new clinical trial?
• Can a pediatric dosing regimen be informed using information from adults and younger populations?
• What clinical trial design provides adequate information under plausible assumptions?
• What exposure-response relationship should be considered when evaluating efficacy or safety?

FDA's current MIDD program specifically asks sponsors to articulate the development question underlying an MIDD approach and the context in which the model will be used. :contentReference[oaicite:6]{index=6}

7. Context of Use

The context of use describes how model-derived evidence is intended to be used to answer the question of interest.

For example, consider a population PK model developed to characterize the effect of renal function on drug clearance.

The same model could potentially be used for a different question, but the credibility and evaluation requirements may differ because the consequences of using the model differ.

This is one reason modern MIDD frameworks focus not only on whether a model is technically sophisticated, but also on whether the model is fit for its intended purpose.

8. What Types of Models Are Used in MIDD?

MIDD is a framework rather than a particular model class. Different scientific questions require different models.

FDA and EMA materials describe a broad range of model-informed approaches, including population PK, exposure-response, PBPK, drug-trial-disease, and mechanistic or systems pharmacology modeling. :contentReference[oaicite:7]{index=7}

9. Where Does MIDD Fit in Drug Development?

MIDD is not confined to a particular clinical phase. FDA states that MIDD approaches can add value throughout drug development and encourages early engagement when a development program has clearly defined questions for which MIDD may provide useful evidence. :contentReference[oaicite:8]{index=8}

10. Example: Population PK as MIDD

Suppose a drug is being developed for a heterogeneous patient population. A population PK model is fitted to concentration-time data from several clinical studies.

A simplified model might describe clearance as:

Here, \(CL_{\mathrm{pop}}\) represents a typical clearance, \(WT_i\) is patient body weight, \(\theta_{WT}\) describes the relationship with body weight, and \(\eta_i\) represents between-subject variability.

The model can then be used to simulate exposure under different dosing regimens.

The important MIDD question is not simply whether the population PK model fits the concentration data. The question might instead be whether the model provides adequate evidence to support a dosing recommendation for a particular patient subgroup.

11. Example: PBPK as MIDD

A physiologically based pharmacokinetic model represents drug disposition using physiological and drug-specific information. Depending on the purpose, the model can represent organs, blood flows, tissue partitioning, metabolic pathways, transport processes, and other determinants of exposure.

For example, a PBPK model may be used to explore a drug-drug interaction involving a metabolic enzyme.

EMA specifically identifies PBPK, physiologically based biopharmaceutics, and QSP among the mechanistic models used in the context of model-informed drug development. :contentReference[oaicite:9]{index=9}

12. Example: QSP as MIDD

Quantitative systems pharmacology (QSP) models extend the modeling framework from drug concentration and exposure toward biological mechanisms and disease systems.

A simplified conceptual model might contain interacting biological quantities:

where \(C\) represents drug exposure and \(X_1\) and \(X_2\) represent biological states.

A QSP model might therefore be used to investigate how target modulation propagates through a biological network and ultimately affects a disease-relevant endpoint.

QSP is not automatically MIDD simply because it is mechanistic. It becomes part of an MIDD application when the model-generated evidence is used to answer a defined drug development question.

13. Clinical Trial Simulation

One of the clearest examples of model-based development is clinical trial simulation.

A trial simulation can combine assumptions about:

• Patient characteristics.
• Disease progression.
• Drug exposure.
• Treatment effects.
• Dropout or missing data.
• Measurement variability.
• Trial design.

Repeated simulated trials can then be generated to explore operating characteristics under specified assumptions.

For example, a simplified simulation might estimate the probability of rejecting a null hypothesis under different sample sizes:

The simulation does not establish what will happen in the real trial. Rather, it explores what the trial might produce under explicitly specified assumptions.

14. How Does a Model Become Decision-Relevant Evidence?

A model becomes useful for development when its outputs can be connected to a decision.

This question-to-decision structure is central to modern MIDD thinking. ICH M15 establishes an assessment framework intended to evaluate MIDD evidence in relation to the question of interest, context of use, model influence, and consequences of an incorrect decision. :contentReference[oaicite:10]{index=10}

15. Model Risk and Consequences of Error

Not every modeling decision carries the same consequence.

Suppose a model is used to explore an early hypothesis. An imperfect model may still provide useful exploratory information. By contrast, a model used to support a major dosing decision may require substantially more rigorous evaluation.

A useful conceptual relationship is:

This does not mean that model evaluation can be reduced to a numerical formula. Rather, it illustrates why the amount and type of evaluation should be connected to the intended use of the model.

The current ICH M15 framework explicitly incorporates model risk and the potential impact of MIDD evidence on decisions. :contentReference[oaicite:11]{index=11}

16. How Are MIDD Models Evaluated?

Model evaluation depends on the model type and intended use, but common considerations include:

• Structural adequacy: Does the model represent the relevant mechanisms or relationships?
• Parameter identifiability: Can the available data adequately inform the parameters?
• Predictive performance: Does the model make useful predictions in appropriate validation settings?
• Data quality: Are the data sufficiently reliable for the intended application?
• External evaluation: Does the model remain useful when evaluated against independent or later information?
• Sensitivity: Which assumptions and parameters materially influence the conclusions?
• Uncertainty: How much uncertainty surrounds model parameters, assumptions, and predictions?
• Context of use: Is the evaluation adequate for the decision the model is intended to inform?

A model does not become credible merely because an optimization algorithm converged or because a statistical goodness-of-fit measure is favorable.

17. Model Uncertainty Is Part of the Evidence

Model-informed conclusions should distinguish between different sources of uncertainty.

Simulation can propagate some forms of uncertainty through a model. For example, if clearance follows a probability distribution, simulated concentration profiles can reflect that variability.

The resulting prediction distribution can be more informative than a single deterministic prediction when uncertainty is material to the decision.

18. MIDD and Regulatory Decision-Making

MIDD can generate evidence used in regulatory interactions and submissions, but model use does not automatically make a conclusion acceptable for a particular regulatory purpose.

FDA's current MIDD program supports sponsor-agency interactions concerning questions such as dose selection, clinical trial simulation, predictive or mechanistic safety, response measures, and other development issues. :contentReference[oaicite:12]{index=12}

The 2026 ICH M15 guideline provides recommendations for planning, model evaluation, documentation, and regulatory interactions involving MIDD evidence. :contentReference[oaicite:13]{index=13}

EMA also maintains guidance and scientific advice mechanisms for model-informed approaches and identifies modeling and simulation as tools that can support development and regulatory decisions. :contentReference[oaicite:14]{index=14}

19. Worked Example: Using MIDD to Select a Dose

Consider a hypothetical drug with a target exposure range of approximately 50–100 mg·h/L. Early clinical data suggest that clearance varies substantially between patients.

Step 1: Develop a population PK model

Suppose the estimated typical clearance is:

For a simplified IV dosing scenario, expected exposure can be approximated as:

Step 2: Evaluate candidate doses

For a 500 mg dose:

For a 250 mg dose:

Step 3: Incorporate variability

Suppose the population model indicates that clearance varies across patients. Instead of evaluating only the typical patient, simulations can generate an exposure distribution for each candidate dose.

Step 4: Link exposure to the development question

The question might be whether a proposed dose is expected to place an adequate proportion of patients within the exposure range associated with the desired benefit-risk profile.

Step 5: Use the model-informed evidence

The simulated exposure distributions can then inform dose selection for the next study.

20. What MIDD Is Not

Several common misconceptions can make MIDD sound either narrower or more powerful than it actually is.

• MIDD is not synonymous with population PK. Population PK is one important MIDD approach among many.
• MIDD is not synonymous with PBPK. PBPK is one class of mechanistic model that can contribute to MIDD.
• MIDD is not synonymous with QSP. QSP can provide MIDD evidence when it addresses a relevant development question.
• MIDD is not simply simulation. Simulation is a computational activity; MIDD concerns the generation and use of model-informed evidence.
• MIDD is not a replacement for all clinical trials. The value of modeling depends on the question, data, model credibility, and decision context.
• A sophisticated model is not automatically a better model. Complexity should be justified by the scientific question and available information.
• A good statistical fit does not automatically establish predictive credibility. Evaluation must be connected to the intended use.

21. MBDD as an Iterative Development Cycle

One useful way to understand the broader MBDD philosophy is as an iterative cycle:

22. Why Use Model-Informed Approaches?

FDA identifies several potential contributions of MIDD, including improved trial efficiency, dose optimization, therapeutic individualization, clinical outcome prediction, and evaluation of safety or efficacy evidence. :contentReference[oaicite:15]{index=15}

More generally, modeling can help development teams:

• Integrate information from multiple studies.
• Quantify relationships that are difficult to evaluate directly.
• Explore dose and exposure scenarios before conducting additional studies.
• Identify important sources of variability.
• Design informative clinical studies.
• Evaluate alternative development strategies.
• Translate information between populations or settings when scientifically justified.
• Make assumptions explicit so that their consequences can be examined quantitatively.

These benefits depend on model quality, data quality, appropriate context of use, and careful interpretation.

23. What Can Go Wrong?

Model-based development can fail when the model is asked to answer a question that the available data cannot support.

• Insufficient data: key parameters may not be identifiable.
• Structural misspecification: the model may omit an important mechanism.
• Biased inputs: incorrect assumptions can propagate through simulations.
• Overfitting: a model can describe existing data without predicting new observations adequately.
• Unvalidated extrapolation: predictions may extend beyond the evidence supporting the model.
• Misinterpretation: model outputs may be treated as observations rather than conditional predictions.
• Decision mismatch: evaluation may not be sufficient for the consequence of the decision being supported.

24. A Practical MIDD Workflow

1. Define the question of interest. State exactly what development problem needs to be addressed.
2. Define the context of use. Specify how the model evidence will be used.
3. Identify relevant information. Include clinical, nonclinical, prior, mechanistic, and disease information where appropriate.
4. Select an appropriate model. Choose a model that can answer the question without unnecessary complexity.
5. Pre-specify important assumptions. Make critical assumptions transparent.
6. Develop and evaluate the model. Assess parameter estimation, diagnostics, predictive performance, and uncertainty as appropriate.
7. Simulate relevant scenarios. Propagate important variability and uncertainty.
8. Assess model risk. Consider how strongly the decision depends on the model and the consequences of an incorrect decision.
9. Generate MIDD evidence. Translate model outputs into evidence relevant to the question.
10. Make the development decision. Use the evidence alongside other available information.
11. Update the model as new evidence becomes available. MBDD is an iterative process.

25. A Simple Mental Model

The relationship between the terminology can be summarized as follows:

MBDD emphasizes making this model-centered process part of the development strategy.

MIDD emphasizes the generation and use of model-informed evidence to answer a development question and support a decision.

The same population PK, PBPK, QSP, PK/PD, exposure-response, or trial-simulation model can therefore participate in both a model-based development strategy and a specific MIDD application.

Where to Go Next

A natural progression is to examine the individual modeling approaches that make up modern MIDD.

Start with Population PK and PK/PD modeling, then move to Exposure-Response Modeling, Clinical Trial Simulation, PBPK, and QSP. The next step is to see how each approach answers a different class of development question and how multiple models can be integrated into a model-informed development strategy.

For mechanistic modeling, PBPK and QSP are particularly useful examples because they show how biological knowledge can be incorporated into models that extend beyond traditional concentration-time analysis. EMA identifies PBPK and QSP among the mechanistic modeling approaches used in the MIDD context. :contentReference[oaicite:16]{index=16}

References

1. U.S. Food and Drug Administration. M15 General Principles for Model-Informed Drug Development. Guidance for Industry, June 2026. FDA guidance.
2. International Council for Harmonisation. ICH M15: General Principles for Model-Informed Drug Development. Final guideline, 2026. EMA / ICH M15.
3. U.S. Food and Drug Administration. Focus Area: Model-Informed Product Development. FDA MIPD overview.
4. U.S. Food and Drug Administration. Model-Informed Drug Development Paired Meeting Program. FDA MIDD Program.
5. U.S. Food and Drug Administration. Model-Informed Drug Development Paired Meeting Program Frequently Asked Questions. FDA MIDD FAQs.
6. European Medicines Agency. Modelling and simulation: questions and answers. EMA modelling and simulation Q&A.
7. European Medicines Agency. Guideline on assessment and reporting of mechanistic models used in the context of model informed drug development. EMA mechanistic-model guidance.