A disease progression model is uniquely positioned to answer a question of interest in which longitudinal data to evaluate the natural history, rate of progression, and/ or factors that might influence disease progression is a significant component.

Potential question of interest topics include to:

• determine the optimal patient population (e.g., selection/enrichment strategies)
• select an endpoint or biomarker
• evaluate the target and therapy performance
• benchmark against competition or previously developed compounds
• select the dosage
• contribute to evidence of treatment effectiveness
• enhance trial design (e.g., informing study power, duration, predicting and quantifying dropout rates)
• understand covariate effects
• predict probability of success
• create virtual control arms/digital twins
• support cross population extrapolation or pooling
• Other – should be able to briefly describe

Question to ask clinical and modeling experts within your organization: Is knowledge about the time course or trajectory of the disease critical to answer a question of interest for the development of this medical product?

If the answer to the above question is ‘yes’, then a DPM should be considered to derive insights and add greater certainty when making medical product development decisions. In Recommendation #2 we further outline when a DPM should be considered.

Value Scenario:
A DPM can identify fast progressors (patients whose disease progresses rapidly) from slow progressors (patients whose disease progresses slowly) which can better inform patient enrichment in your trial design. For example, enriching for fast progressors may help with designing a shorter duration trial, while enriching for slow progressors may require a longer trial to allow more time for a drug of interest to take effect.

Seeking Examples?
See Appendix for a few examples from the literature for which DPM has been used to answer a question of interest. In addition, a list of 14 disease models used in clinical trials and that informed regulatory decision making is located in Table 2 on FDA’s Division of Pharmacometrics webpage. A list of disease specific qualification responses from the EMA are located in the Appendix.

The use of a DPM can allow for a simplified and potentially cost-effective way to answer a question of interest in the medical product development process, with flexibility to be reused and customized for future trials.

Developing and implementing a DPM can be time and resource-intensive but conducting an early assessment of feasibility and value over other approaches will help to realize long term benefits.

To identify when DPM should be considered over other approaches that might exist, use the CTTI DPM Considerations Framework. At a high level, evaluate:

• The value and benefit versus risk of using a DPM at the selected point(s) in the medical product development lifecycle
• The availability of data and resources (See Recommendations 3 & 4)
• The anticipated investment for data, talent, and technological capabilities
• Whether the investment into the DPM can be used beyond the intended context of use (COU)*.

* The context of use is a description of the model and its specific role and scope to answer the question of interest (adapted from V&V40 Assessing Credibility of Computational Modeling through Verification and Validation: Application to Medical Devices and the General Principles for Model-Informed Drug Development M15, ICH Harmonized Guideline, draft version, Nov 2024)

The subsequent recommendations outline tactical considerations when incorporating disease progression modeling into a medical product development program.

Disease progression models are often informed by prior work. A good starting point is to identify whether models for the disease already exist.

Search for disease models published in the literature and model repositories such as the Innovative Health Initiative’s Drug Disease Model Resources (DDMoRe) repository or Critical Path’s Tools and Platforms.

Disease specific simulator programs can also be leveraged, either freely available or at cost. It is also important to explore the literature, especially literature based on contemporaneous data, to assess the clinical measures of progression that are relevant to your question of interest.

Input from patients and patient organizations, clinicians, regulators, and model developers should be included early to decide whether the measures to be tracked and described are meaningful and clinically relevant. Patients, for example, can weigh in on informing the selection of endpoints and the disease progression points or longitudinal factors of greatest importance. Patient associations/patient advocacy groups/patient foundations may also have knowledge about existing models for their disease and they can provide input to ensure model results are understandable and actionable. Model developers can provide insight on the optimal model method.

This collected information can inform the DPM method and how it will be used in the medical development process (i.e. context of use). Table 1 (below) offers examples of measures of progression and model method types.

Question to ask clinical and modeling experts within your organization: What is the most up-to-date understanding of the disease and measure(s) of progression to inform the disease progression modeling approach?

As with any data sources, maintain appropriate levels of data privacy.

Value Scenario:
Disease progression modeling can describe different progression measures to give a more in-depth picture of how a disease’s physiologic underpinnings connect to its outcomes. Different models may be used in the process. For example, a disease progression model of Type II diabetes could track changes in glucose in relation to changes in HbA1C and then to cardiovascular outcomes. This in turn can inform critical decisions around medical product development and its value for population health.

Seeking an Example?
See the Appendix for an example from the literature for which DPM has been used to link disease pathophysiology to medical product development decisions, in this case for dosing decisions.

Accumulate and integrate data from multiple sources, as varied data can improve the model’s representativeness for its intended application. Data sources may include previous clinical trials in the disease, data from pre-competitive consortia, registries, literature data, and multimodal data such as electronic health records (EHRs), imaging (MRI, CT scans, pathology slides), genomics and other omics data (proteomics, metabolomics), wearable and digital biomarkers, and patient-reported outcomes. The integration of these diverse data types enhances the robustness and predictive accuracy of the disease progression model.

The context for which the DPM is being used in the medical product development process will dictate where attention should be given when assessing the data. For example, recognize that if a DPM is being used to support regulatory decision making, the data will need to meet the regulatory standards of reliability and relevance.

Consider the following set of questions to better evaluate the reliability and relevance of the data available.

• Are the available sources of data consistent?
• What level of data variability exists across countries?
• How precise and sufficient are the available data? (The precision of the data will inform methodology.)
• Is the data of high enough quality relative to the DPM goals?
• How much data and what type of missingness is present in the data? Is the data amenable to imputation (which can dilute the info/process)?
• Is there bias in the data?
• Are data about intervention effects available?

Collaborative input, including from patients and patient organizations, clinicians, regulators, and model developers, will also help to assess whether the sources of data are relevant and reliable.

Example resources to support the use of data for the development and implementation of a DPM, including for regulatory purposes, are provided in the Appendix.

Question to ask modeling experts within your organization: Do we have access to data that is relevant and reliable to develop a DPM for the context of its use?

Beyond availability of relevant and reliable data, it is important to have the appropriate talent and technological capabilities in place to support DPM development and use in your organization.

The necessary skills include technology, clinical, and regulatory expertise. Modeling experts skilled in the expected model method provide the technological expertise. Clinical expertise includes those with domain knowledge about the disease. Based on the context of use of the DPM, those with regulatory expertise can provide input around expectations if the DPM is meant to support a regulatory submission.

Additional expertise to speed the development and implementation of a DPM includes data management expertise, specialized project managers who are knowledgeable about modeling, and biostatisticians.

Explore public-private partnerships and other organizations that are well versed in developing models such as Critical Path Institute’s Quantitative Medicine Program, academic institutions that have developed disease progression models, and Contract Research Organizations (CROs) specialized in modeling.

In addition, consider whether there is access to technological resources required to develop and/or implement the model. This may include computational and analytical platforms for the anticipated model.

Question to ask modeling experts within your organization: Do model, technological resources, and skill sets exist, or can they be obtained to support DPM development and implementation?

DPM models can make predictions for and bridge gaps between medical product development phases, a valuable asset for strategic planning and decision making.

Disease progression models may be adapted and validated as the medical product development lifecycle progresses, saving time in later stages of development compared to having to develop an entirely new model.

Knowledge gleaned from earlier trials (phase I or phase II) using disease progression models, such as for endpoint or dose selection, allows medical product developers to make “Go/No-Go” decisions and inform later-stage trials with confidence.

Starting early will allow for the appropriate time to build and gain confidence in the model to be ready for and inform later phase trials.

Seeking Examples? See the Appendix for a few literature examples where DPM has been used to bridge gaps between early-stage and late-stage development.

Question to ask modeling experts within your organization: Can we implement the disease progression model early on in the medical product development lifecycle to build confidence in the model for later stages of development?

To optimally implement a DPM, including reuse beyond what it was originally trained for, it is important to continually assess and update the model as contemporaneous data is collected and the understanding of the disease evolves. Be cognizant that the performance of the disease progression model may need to be re-evaluated if the context of use changes, with consideration of model robustness and potential limitations.

Below are potential questions to consider when assessing the disease landscape for ongoing performance of a DPM:

• Has the disease diagnostic approach changed?
• Has the disease prognosis changed?
• Is there new knowledge of disease progression?
• Are there changes to the population’s underlying characteristics?
• Has the medical practice (standard of care) or treatment landscape changed?
• Has there been a change in the anticipated endpoint based on the context of use?

Example resources that assist with assessing the performance of a DPM are included in the Appendix.

Question to ask modeling experts within your organization: Do we have the resources to continually assess the performance and ensure that the DPM is qualified for the intended application?

The final recommendations provide general considerations for advancing the use of DPM.

Rapidly advancing technological capabilities are creating new opportunities for DPM in medical product development. These opportunities include the use of artificial intelligence (AI) to support DPM efficiency and knowledge enhancement for precision medicine.

AI is increasingly being used in mechanistic disease models to better understand disease and causality. AI is also supporting operational efficiency of empirical models to analyze multi-modal sources of data for models. AI-driven approaches can leverage multimodal machine learning to integrate and analyze large-scale patient data, such as real-world evidence (electronic health records, claims data), wearable sensor data, and patient-reported outcomes. For example, deep learning on longitudinal electronic health records and digital biomarker data can predict mobility decline in Parkinson’s disease.

The Food and Drug Administration released guidance on Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products providing a framework of responsible use of AI.

Value Scenario:
A research team uses AI-powered external validation to assess a Parkinson’s Disease progression model, integrating multimodal data (electronic health records, imaging, and wearable sensors). AI detects bias in patient subgroups, identifies model drift, and quantifies uncertainty in predictions, ensuring robustness. Explainable AI (XAI) makes insights interpretable for clinicians, while regulatory oversight ensures risk mitigation and trustworthiness.

Seeking an Example? See the Appendix for an example in the literature of machine learning in DPM used to enhance precision medicine.

Information sharing will help with alignment in terminology and highlight the unique role that DPM plays in decision making throughout medical product development.

CTTI encourages involvement in precompetitive initiatives, professional societies, case examples, webinars, publications, and conferences to share:

• Data used to train and develop disease progression models
• Synthetic data (which can facilitate transparency and reproducibility with privacy built in)
• Code and other open source tools (e.g., Python or R language)
• Models, including reduced order models that lessen model complexity
• Validation and cross validation approaches

Dissemination of information should occur beyond the modeling and technical sector and into conferences where medical product development decision-makers, therapeutic area clinical experts, and clinical trial investigators engage. Insights should be shared with patients and patient groups, particularly those that contributed to model development, to build trust and improve impact.

See the Appendix for a few examples of collaborative groups and case study exchanges where information can be shared.