Model-Informed Drug Development: Bang for the Buck?

Abstract

Model-Informed Drug Development (MIDD) has emerged as a foundational pillar of modern drug development, transforming how evidence is generated, integrated, and acted upon throughout the lifecycle. Once viewed as a complementary strategy, MIDD is now routinely embedded in regulatory and industry decision-making, offering powerful tools to optimize study design, inform dose selection, and support benefit–risk assessments. Its impact is both measurable, saving time and reducing costs by streamlining development, and qualitative, reflecting the critical, but less easily quantified, value of more informed labeling statements that support clinical decision-making and patient care. As MIDD continues to evolve, its influence can be understood across three core areas: (1) improving efficiency and generating cost savings, (2) mitigating risk across development programs, and (3) enhancing product labeling to inform real-world decisions. Together, these dimensions illustrate the value of MIDD not only as a technical approach, but also as a strategic framework central to the future of drug development.

One of the most evident and widely recognized impacts of MIDD is its ability to improve the efficiency and reduce the overall cost of drug development. By integrating quantitative models early and throughout the development process, MIDD enables more strategic decision-making, allowing sponsors to streamline programs, reduce redundancy, and target resources more effectively. For example, model-based bridging approaches can support smaller or fewer clinical trials by extrapolating existing data across populations or dosing regimens, while optimized dose selection reduces the risk of trial failure due to suboptimal exposure-response relationships. Moreover, model-based trial designs, including adaptive designs, can accelerate Investigational New Drug and New Drug Application timelines by informing dose-ranging studies, refining endpoints, or enabling innovative approaches such as dose-exposure extrapolation or the waiver of confirmatory trials in specific contexts.

These methodological efficiencies are translating into tangible economic value. In this issue of Clinical Pharmacology & Therapeutics (CPT), a recent analysis by Pfizer found that the use of MIDD approaches was associated with an average reduction of 10 months in development cycle time and $5 million in development costs per program.¹ To the best of our knowledge, this is the first retrospective analysis of internal research and development (R&D) programs quantifying the broader organizational value of MIDD by capturing the key development questions informed by MIDD, associated assumptions and risks, and the potential impact on cost, timelines, and decision-making. Methods based on per-subject approximations and trial size reductions were used to estimate cost and time savings, using benchmarks for study timelines and enrollment metrics across development phases. This work expands upon a 2013 CPT publication from Pfizer which first highlighted the potential for model-based strategies to generate cost savings in the range of $70 million per year.² Another prior analysis found that MIDD implementation can reduce development costs by $30 to $70 million, depending on the scope and integration of modeling approaches.³ These benefits reflect not only streamlined development timelines but also earlier and more confident go/no-go decisions that can be enabled by predictive simulations. This ever-evolving role of MIDD as a driver of efficient development is also highlighted on the cover of this issue (Figure 1).

Importantly, the economic benefits of MIDD extend beyond the point of regulatory approval. Traditionally, methods such as model-based meta-analysis (MBMA) have been known to strengthen the quality of clinical evidence. In one article in this issue of CPT, a less common application is highlighted, demonstrating how this methodology was used to improve the likelihood of favorable reimbursement decisions and facilitates more timely access for patients to innovative therapies.⁴ In this way, MIDD supports not only efficient development, but also more efficient delivery of therapeutic value to the healthcare system.

MIDD also plays a critical role in mitigating risk throughout the drug development lifecycle by leveraging computational tools and predictive models to optimize decision-making. This approach helps reduce the likelihood of failure and unforeseen complications, particularly as drug programs move from early to late-stage development. One of the primary applications of MIDD is in optimizing dose selection, directly addressing the risk of advancing a development program with suboptimal efficacy and/or safety profiles. Using population pharmacokinetic (popPK) modeling and exposure-response (ER) assessments, MIDD enables the identification of an optimal dose range that maximizes therapeutic benefits while minimizing adverse effects. This improves the likelihood of success in clinical trials, reducing both time and costs associated with development. Beyond dose optimization, MIDD also aids in forecasting the probability of technical, commercial, and regulatory success throughout the development process. By quantitatively assessing the likelihood of success at various stages, MIDD helps prioritize resources and focus efforts on high-probability programs, thus minimizing investments in projects with lower chances of success. In an environment where the costs of failure are high, this potential to assess and mitigate risk early in development is invaluable.

MIDD is especially valuable in mitigating risk for novel therapeutic modalities, where traditional approaches may be limited due to scientific uncertainty or lack of historical data. In these cases, predictive modeling can also guide the selection of dosing regimens and anticipate potential clinical outcomes. For example, in this issue of CPT, the FDA describes its experience with the development of chimeric antigen receptor (CAR) T-cell therapies, a class of “living drugs.” They focus on how the unique pharmacokinetic and pharmacodynamic behaviors present significant challenges for conventional methods.⁵ Learnings from the FDA demonstrate how popPK and ER modeling can be readily leveraged to overcome these challenges, particularly for dose selection and ensuring internal consistency of clinical data. Similarly, this issue of CPT explores how novel modalities like bispecific antibodies present their own set of challenges due to their unique mechanism of action.⁶ These antibodies bind to two distinct targets in vivo, forming a trimolecular complex that often results in a bell-shaped concentration-response curve. Traditional approaches may struggle to address these complexities, but the application of mechanistic physiologically based pharmacokinetic (PBPK) and response-optimized models enables more accurate predictions of clinical outcomes. These examples showcase how MIDD plays a pivotal role in mitigating risks throughout the drug development process, particularly for innovative and complex therapeutic modalities, by enabling more informed decision-making and optimization strategies.

MIDD has had a transformative impact on the way drug labeling supports clinical decision-making, particularly when it comes to dose optimization for specific patient populations. In the past, labeling often included generalized dosing recommendations that may not have initially accounted for the variability seen across different demographic groups, such as older adults, pregnant individuals, or pediatric patients. However, with the rise of MIDD, the ability to generate more informative and personalized labeling statements has been greatly enhanced, ensuring that the right drug, at the right dose, reaches the right patient.

One of the core values of MIDD lies in its patient-centricity. By utilizing advanced modeling approaches, MIDD facilitates more accurate dose selection, particularly for populations where clinical trial data may be challenging to generate. For example, the application of PBPK modeling has been instrumental in supporting dose recommendations for pregnant individuals, a group traditionally excluded from clinical trials. One article in this issue of CPT explores its application to therapeutic antibodies, which are often prescribed off-label during pregnancy, and shows how PBPK modeling can be used to assess drug exposure and guide dose adjustments in this population.⁷ Similarly, MIDD has proven invaluable in pediatric drug development, where dosing regimens often rely on extrapolations from adult data. Several studies in this area have utilized modeling and simulation to refine pediatric dosing recommendations. For instance, two articles in the current issue describe how MIDD supported apixaban dosing in pediatric patients, helping to optimize the therapeutic regimen while minimizing adverse effects.^{8, 9} In another example, popPK and ER modeling was used to optimize the dose of Brentuximab Vedotin in pediatric patients with advanced-stage newly diagnosed Hodgkin lymphoma.¹⁰ These cases demonstrate how MIDD contributes to better clinical decision-making by improving the precision and accuracy of labeling information. The ability to tailor dosing recommendations to specific populations is a critical advancement that supports more effective and safer treatments, ultimately benefiting both patients and healthcare providers.

Today, MIDD is considered a mainstay of modern development efforts. As the landscape of drug development continues to evolve, the future of MIDD lies in the integration of more advanced methodologies, particularly artificial intelligence (AI), machine learning (ML) and real-world data, across the entire development continuum. This vision aligns with a broader reframing of MIDD as MID3: Model-Informed Drug Discovery and Development. For example, AI and ML approaches can be used early in drug discovery to identify novel targets, predict compound activity, and prioritize candidates with favorable safety and efficacy profiles. Modeling and simulation methodologies play a critical role in every facet of development and regulatory decision-making, and the integration of AI/ML across this landscape has the potential to further strengthen these efforts.

Recent FDA draft guidance recognizes the potential of AI/ML in regulatory decision-making, signaling an increasing openness from regulatory bodies to embrace these technologies.¹¹ Early engagement from these agencies underscores the importance of staying ahead of technological advancements and leveraging them to enhance MIDD practices. While recent analyses, such as Pfizer's report in this issue of CPT that estimates cost savings associated with MIDD strategies,¹ have helped quantify the downstream impact of model-informed approaches, future work may further expand on this foundation by exploring the potential value of modeling and AI/ML in preclinical and discovery phases. For instance, AI-enabled target identification, lead optimization, and early safety prediction could generate additional efficiencies and cost savings upstream, further strengthening the overall case for investment in MID3.

Looking ahead, AI/ML techniques are likely to be incorporated into every stage of the development lifecycle. In the case of drug discovery, AI/ML can accelerate identification of effective therapies by enabling more precise target identification and validation. By leveraging data sources, such as multi-omic repositories and extensive literature-based datasets, these technologies can uncover novel disease targets and prioritize those with the highest translational potential. In clinical development, AI-driven patient stratification and trial enrichment will identify subpopulations most likely to respond to therapies, facilitating shorter trials with higher success rates. On the operational side, AI has the potential to automate clinical trial monitoring, streamlining data review, detecting safety signals in real-time, and dynamically adjusting trial protocols as needed, which could substantially reduce both operational costs and risks. Moreover, AI-enabled systems, such as digital twins, can simulate virtual populations and predict trial outcomes, providing invaluable support for regulatory submissions and post-market commitments. Post-market surveillance could likewise benefit from AI-powered pharmacovigilance systems, which can analyze vast amounts of real-world data, ranging from EHRs to social media and spontaneous reporting systems, to detect adverse event patterns more quickly and accurately than traditional methods.

In light of these advancements, the call for broader adoption and continued training in AI/ML tools within drug development is paramount. As the integration of these technologies becomes increasingly central to the drug development process, it is crucial for stakeholders to invest in the necessary infrastructure, training, and interdisciplinary collaborations to ensure that these tools are appropriately incorporated into a development program. Just as a concerted investment in pharmacometrics laid the groundwork for the rise of MIDD, we now stand at a similar inflection point where embracing AI/ML with the same level of commitment can redefine what is possible in the next era of drug development.

No funding was received for this work.

The authors declared no competing interests for this work.