August 2026: AI in Pediatric Clinical Trials, Research, and Care
The CCDI workshop, "How Do We Effectively Use AI to Improve Pediatric Clinical Trials, and How Can AI Accelerate Cancer Research and Treatments?" was held on August 14, 2026.
AI has the potential to continue transforming many aspects of the clinical trial process and drive innovation to expedite treatment discovery. During this interactive workshop, Dr. Warren Kibbe, NCI’s Deputy Director for Data Science and Strategy, and Dr. Jack Shern, NCI’s CCDI Scientific Director, led a discussion on how AI can be used to accelerate pediatric cancer research and improve pediatric clinical trials.
Access the event recording.
Workshop Key Focus Area
This workshop included three brief presentations on the opportunities and use of AI in pediatric cancer research and care. After each presentation, Drs. Kibbe and Shern opened the floor to questions and insights on each topic.
Presentation 1: "AI to Advance Pediatric Clinical Trials"
Presenter: Michael Cohen-Wolkowiez, M.D., Ph.D. — Duke University / Duke Clinical Research Institute
Dr. Cohen-Wolkowiez described the unique challenges of pediatric clinical trials, including developmental differences across age groups, small and heterogeneous patient populations, dependence on caregivers, distinct risk-benefit considerations, and age-specific endpoints.
While stressing that expert human oversight will remain important, he emphasized that AI can help researchers:
- make better use of limited pediatric data by integrating clinical, physiologic, preclinical, and other information and by improving modeling, trial simulation, and evaluation of study-design tradeoffs.
- improve recruitment and site selection, identify potentially eligible participants from electronic health records, analyze new data streams such as wearables, and support development of age-appropriate digital endpoints.
- automate administrative and operational activities such as protocol and consent development, translations, case-report forms, training, and patient/family communications .
Presentation 2: "Data for the Common Good: GEARBOx"
Presenter: Sam Volchenboum, M.D., Ph.D. — University of Chicago / Data for the Common Good
Dr. Volchenboum focused on the difficulty of matching children with cancer to appropriate clinical trials, particularly because protocols and eligibility criteria are often written in nonstandard, non-computable formats, and relevant patient information can be scattered across structured and unstructured EHR data.
He presented Gearbox, a clinical trial matching platform that uses natural language processing and large language models to transform trial eligibility criteria into structured information and compare those criteria with patient characteristics to identify potential matches. He described a future in which matching systems incorporate EHR data, external laboratory and molecular data, hospital clinical-trial systems, and data obtained directly from patients, with conversational interfaces making the process easier for clinicians and families. A major recommendation was to create more standardized, computable clinical trial criteria (potentially through a national clearinghouse) and to improve access to timely information about where trials are actually open and enrolling.
Presentation 3: "Using AI to Accelerate Pediatric Cancer Research and Discovery"
Presenter: Javed Khan, M.D. — National Cancer Institute
Dr. Khan discussed how AI is expanding beyond clinical trial operations into laboratory and translational cancer research, including biomarker discovery, tumor biology, therapeutic-target identification, and drug development. He noted that emerging AI and agentic coding tools can enable scientists and clinicians with strong domain expertise to interrogate complex datasets more rapidly, while AI-based approaches may help identify drug targets, design therapeutic binders, explore previously “undruggable” cancer drivers, and generate new therapeutic hypotheses. As an example, he described work using AI to analyze H&E pathology images in rhabdomyosarcoma, where models showed potential to further stratify intermediate-risk patients into higher- and lower-risk groups; newer foundation models may improve on earlier Convolutional Neural Network (CNN)-based approaches. He emphasized, however, that the greatest limitation remains access to sufficiently large, high-quality, clinically annotated pediatric cancer datasets and advocated for broader digitization, data sharing, and ultimately international collaboration to generate the scale of data needed for reliable AI models.
November 2022: Rare Cancers
The CCDI workshop, Advancing a National Initiative for Rare Cancers in Children, Adolescents, and Young Adults, was held November 18, 2022.
Currently, there is an unmet need for a national effort to collect data on very rare childhood cancers to better understand their natural history and accelerate progress in the treatment of these cancers. The Childhood Cancer Data Initiative (CCDI) is looking to work with collaborating investigators, organizations, and advocates to coordinate this effort, but such a widespread initiative requires input and buy-in from the broader childhood cancer community.
The purpose of this workshop was to better understand the current landscape of observational and interventional studies in rare cancers to better inform how CCDI and collaborators can help create an infrastructure to collect and share existing and future data meaningfully, in addition to promoting collaboration and participation from the community.
Access the event presentation and event recording.
Workshop Key Focus Areas
The workshop included a morning and afternoon session. Each session had four presentations and a closing panel of scientists and advocates. The morning session provided a foundation for the meeting, patient perspective, and context on why this national effort is needed. In the afternoon session, discussions focused on the role of data collection, management, and sharing in this effort.
Morning Session
Presentation 1: Need for Coordinated Effort in Pediatric Rare Cancers
This presentation focused on why a coordinated effort to collect data on rare childhood cancers is needed, particularly given that rare tumors comprise a significant portion of childhood cancers. Challenges related to scaling individual efforts, selecting what rare tumors to focus on, and limited enrollment of patients with rare tumors in clinical trials were also discussed, along with the need for a unified framework and resources to study rare tumors.
Presentation 2: Patient Perspective
The workshop then transitioned to a presentation by a mother who had two children diagnosed with multiple rare cancers as a result of a p53 gene deletion. She shared her children’s stories and how her family has coped with their cancer journeys. Experiences like this family’s further indicate the need for a national effort to make faster progress in rare childhood cancers.
Presentation 3: Pleuropulmonary Blastoma/DICER1: Bedside to Bench and Back Again
Using pleuropulmonary blastoma (PPB) and tumors related to the DICER1 gene as examples, this presentation highlighted how much progress can occur in a rare childhood cancer when there is an organized effort to study it. The mission and goals of the International PPB/DICER1 Registry were discussed, along with the benefits of a multidisciplinary approach and collaboration in progressing research.
Presentation 4: Early Considerations of a Framework for a CCDI Coordinated National Rare Tumor Initiative
During this presentation, a concept for a coordinated national study of rare childhood cancers was shared. The study would allow for the enrollment of as many participants as possible so that multiple rare cancers can be studied simultaneously more rapidly, efficiently, and consistently. It would also enable building a rare cancer registry, state-of-the-art molecular profiling, meaningful comparisons across multiple rare cancers, facilitation of patient navigation and treatment recommendations, and providing the scientific rationale as well as identification of external controls and possible identification of therapy targets for clinical trials.
Afternoon Session
Presentation 5: Core Data Elements
Both core and expanded clinical data will need to be collected from patients with rare childhood cancers. This presentation highlighted what data elements are required to advance a national initiative, based on lessons learned from genomics projects being conducted by the Dana-Farber Cancer Institute, Boston Children’s Hospital, and the Broad Institute.
Presentation 6: Connecting the Data: CCDI Data Ecosystem
This presentation focused on what data capture, management, and federation in the CCDI Data Ecosystem could look like. A summary of the CCDI Data Ecosystem’s objectives, infrastructure, and components were provided. There was also additional discussion on how the ecosystem could help in a national effort to collect and share data on rare childhood cancers.
Presentation 7: Patient Portals: Role and Utility for Patients and Research
Patient portals that show coordinated rare childhood cancer data are not currently available and the utility for patients needs to be explored. Using patient portal research from the Leukemia & Lymphoma Society, this presentation delved into why such a portal would be beneficial for patients and families to help navigate their care and submit self-reported data, and for providers and researchers alike who may be seeking these coordinated data or information related to clinical trial participation. Work performed through Count Me In, a Broad Institute initiative, was also used to illustrate how to partner with patients and engage them during the portal development process.
Presentation 8: Potential Role for National Rare Cancer Tumor Boards
This presentation highlighted why coordinating and standardizing data from rare childhood cancer programs, networks, and clinical trials can ultimately optimize treatment and care for patients. The Texas Children’s Cancer and Hematology Centers’ Rare Tumor Board was used to demonstrate how a multi-institutional group of specialists could advise other clinicians on treating patients with rare cancers. Challenges related to this and similar efforts were also noted, including the need for dedicated resources, professional liability, data storage for future access, multidisciplinary participation, and research integration.
November 2022: Electronic Health Records
The CCDI workshop, The Importance of Electronic Health Record Data in Clinical Care and Research, was held November 2, 2022.
Electronic health records (EHRs) are systems that store health information in a digital format. They document important health data that can be shared or communicated between a patient and their provider. While EHRs are increasingly used in the United States to collect data and provide opportunities to better manage care for patients, these systems were not designed to readily share data with other systems or intended for research. Efforts to assess quality metrics in EHRs have been limited.
The focus of the workshop was to discuss ways to improve EHR data to make these data easier to use and share, with the goal to make discoveries faster in childhood cancer research. Insights from the workshop help inform Childhood Cancer Data Initiative activities and the development of NCI funding opportunities related to EHR data sharing and accessibility.
Access the event presentation and event recording.
Workshop Key Focus Areas
The workshop included four sessions focused on:
- EHR data portability and interoperability
- structured EHR data, data extraction, and translation
- EHR-directed clinical trials
- research use of EHR data
Session 1: EHR Data Portability and Interoperability
In this session, panelists discussed the issues surrounding equitable access to health data by patients and their providers and emphasized supporting rapid and seamless transfer of personal health data among care providers. Panel experts discussed the importance of empowering patients to provide their consent for how and where their data are used and the significance of access to personal data for a broad range of research uses.
Session 2: Structured EHR Data, Data Extraction, and Translation
Panelists in the second session outlined how data in EHR systems can best be structured and standardized to be most useful for patient care and across cancer research. Among other issues, experts noted challenges around variable data structures, a lack of consistency in using standards, and poor data accuracy for rare cancers, particularly in pediatric, adolescent, and young adult populations. Possible solutions and standards were also discussed.
Session 3: EHR-Directed Clinical Trials
The first afternoon session focused on how EHR data could facilitate new discoveries in clinical trials. Panelists discussed the critical need to obtain patient treatment and other relevant patient data from EHRs and use data to define structured eligibility criteria to match patients to clinical trials. They also discussed developing appropriate control arms and cohorts for new studies. Experts further outlined challenges related to regulatory and commercial requirements for data quality, particularly in rare cancers, before describing existing and possible new tools and approaches for potential solutions.
Session 4: Research Use of EHR Data
In the final session, panelists explored opportunities to use EHR data for validation and discovery in biomedical research and public health reporting, beyond clinical trials. They discussed possible approaches to address challenges around research access to EHR data, the structure and integration of these data, the variability of information available in EHRs, and institutional and regulatory limitations affecting research.