(P 225) PAN-SARCOMA DATABASE (PSDB): A MODEL-VIEW-CONTROLLER FRAMEWORK UTILIZING REDCAP, SHINY, AND R TO CREATE A CONTINUOUS PIPELINE OF SARCOMA DATA

Objective: The Pan-Sarcoma Database (PSDB) is an innovative open-source research application developed to efficiently collect and manage data on sarcoma patients to create a reliable research pipeline. Our data process uses free software, REDCap, to collect, update, and store data and metadata, and uses open-source coding in R to transform, merge, explore, and analyze data. This process can be modified and customized to fit other institutions. We aim to share our data structure and process to establish basic standards in hopes of increased data sharing and collaborative research with institutions that treat and study patients with sarcoma.

Methods: PSDB implemented a Model-View-Controller (MVC) framework, a software design pattern used by many modern-day applications. In this framework, "model" is the database, including the structure and metadata, "view" is the user interface software, and "controller" is the coding language used to handle user input, update the model, and interact with the viewer. PSDB uses REDCap, shiny, and R, respectively. REDCap, a free software, provides a centralized location for data storage, strict user privileges, logged data modifications, and has a robust application program interface (API). Shiny is an open-source R package that provides a way to create a custom user interface to explore, transform, and visualize data. Lastly, the R programming language for statistical computing provides extensive packages for data manipulation, visualization, and analysis. R is also used to create a custom package of functions for sophisticated data quality checks and custom data uploads using the REDCap API. We integrated other concepts in our design such as normalization and standards from Clinical Data Interchange Standards Consortium (CDISC). For example, partial dates are saved as partial and imputed as needed during analysis. As another example, data that is singular to a patient such as the medical record number, birth date, demographics, and survival status are stored at the highest level and data that can repeat, such as a sarcoma diagnosis or treatments are stored at another level, similar to relational databases. R code is mapped to our “model” and its associated transformation process to produce updating tables, time-to-endpoint analyses, and other modifiable outputs.

Results: The PSDB successfully addresses the unique data challenges faced in sarcoma research since it is comprised of a heterogenous set of mesenchymal cancers. The use of the MVC framework with REDCap, shiny, and R, enabled the creation of an application that is secure, user-friendly, and adaptable. The database is currently comprised of 290 data fields to capture demographics, survival data, sarcoma data, and associated treatment data. We continue to integrate multiple data sources, such as an internal cancer database, pathology records, and prior research data. We captured data from 28,380 candidate sarcoma patients from 2010 to present, including 3,291 confirmed sarcoma cases, with 751 cases reviewed to date. By capturing key data and organizing an open-source code base around it, we automated time-to-endpoint analyses that are updated as the data changes. Amongst our confirmed cases we have a cohort that is 53.3% female and 32.1% Hispanic. Amongst the cases reviewed so far we have a diverse group of histopathologies: GIST (176), Leiomyosarcoma (127), Angiosarcoma (100), Unspecified/Other (93), Ewing Sarcoma (49), Liposarcoma (43), Synovial Sarcoma (40), Chondrosarcoma (35), Fibrosarcoma (34), Rhabdomyosarcoma (32), Desmoid Tumor (23), Myxofibrosarcoma (23), Osteosarcoma (23), Malignant Peripheral Nerve Sheath Tumor (20), Solitary Fibrous Tumor (20), Hemangioendothelioma (8), Alveolar Soft Part Sarcoma (6), Clear Cell Sarcoma (6), Dermatofibrosarcoma (5), Inflammatory Myofibroblastic Tumor (5), Perivascular Epithelioid Cell Neoplasm (4), Chordoma (Notocord), (2), Karposi Sarcoma (2), Lymphangiosarcoma (1), Phyllodes Tumor (1), Undifferentiated Pleomorphic Sarcoma (1). We have several active projects working to grow our database. Currently, 30.6% of our cases are metastatic at diagnosis; 57.7% are High grade, 28.9% are Intermediate grade, and 8.9% low grade; the median age at diagnosis is 55; 93.5% of reviewed cases received systemic therapy and 73.1% received radiation.

Conclusion: The Pan-Sarcoma Database (PSDB) presented in this study represents an innovative approach to address the data management needs of a clinical research team focusing on sarcoma oncology. By employing the MVC software design pattern and leveraging REDCap, shiny, and R, the PSDB provides a secure, flexible, version-controlled, reusable, and accessible clinical database application. The successful implementation of the PSDB facilitated efficient data collection, integration, and management, supporting collaborative research efforts and potentially enabling future applications of AI. This project emphasizes the importance of employing appropriate tools for effective data management and encourages similar initiatives to enhance collaboration and benefit both research teams and the patients they study.