2416 - Project 30: Institutional-Scale Harmonization of Three Decades of Pediatric Radiation Oncology Data

Purpose/Objective(s):

Longitudinal radiation oncology research is constrained by fragmented vendor systems, heterogeneous nomenclature, and labor-intensive dosimetric data extraction and curation. We developed and deployed a scalable, vendor-neutral platform to integrate, validate, and operationalize radiation therapy dosimetry and clinical metadata for a large pediatric cohort spanning more than three decades of practice.

Materials/Methods:

The platform integrates data from multiple commercial treatment planning and oncology information systems into a unified research environment. Standardized structure nomenclature enables uniform querying across disease sites and protocols. Automated workflows perform dose–volume histogram (DVH) extraction and standardized dose metric calculation following dosimetric review. Rule-based quality assurance (QA), tailored to treatment region, is then applied. Cases failing predefined QA criteria are routed for mandatory physics review, while a randomly selected subset of passing cases undergoes additional physics review to maintain quality oversight. A protocol-aware workflow engine assigns tasks based on clinical role and study protocol, coordinating clinical research associates, dosimetrists, and medical physicists. Real-time dashboards monitor task distribution, completion status, and longitudinal throughput. The system is embedded into routine departmental operations and supports both retrospective data curation and prospective patient accrual.

Results:

Since inception in early 2025, the platform includes over 6,600 pediatric patients treated across 30 years of clinical practice and multiple generations of treatment planning systems. New patients are integrated prospectively as prescriptions, protocol enrollments, and treatment plans are entered into clinical vendor systems. Clinical research associates have completed structured abstraction for 6,183 patients. Subsequent dosimetric processing has been completed for 2,354 patients, with physics-level review finalized for 1,236 of 1,456 assigned patients, reflecting structured progression across clinical roles. Dashboard analytics demonstrate sustained weekly task completion and stable retrospective throughput during large-scale cohort processing. Automated DVH extraction and QA-driven routing enable real-time cohort-level dose queries across harmonized data. Completion of the retrospective cohort is anticipated in 2026. The project remains ongoing, with prospective accrual continuing.

Conclusion:

This scalable infrastructure demonstrates the feasibility of harmonizing pediatric radiation oncology data at institutional scale. By integrating standardized dosimetric feature generation with rule-based QA and role-aware workflow orchestration, the platform establishes a reproducible foundation for large-scale outcome modeling and future clinical trial design.