Oncology Early Development Technology to Facilitate Early Analysis of Safety and Efficacy Data.

e14116 Background: The lack of a standard data review platform for the surveillance of early oncology trials led to the formation of a cross-functional initiative with the objective of creating a data surveillance framework that would provide clinical study teams with access to near real-time data in order to be able to generate fast insights into the safety and efficacy findings of oncology early development trials so that early signal detection or clinical proof-of-concept could take place. Methods: To determine the requirements that would drive the framework, the cross-functional team first conducted user interviews with study physicians to gather information on the types of tables, graphs, and visual outputs that would be most useful to the clinical study team, as well as most useful to re-create as a standard across the early oncology portfolio, focusing at this stage, on solid tumor trials using RECIST 1.1 for the efficacy output. Once these requirements were determined, specification details, including mock-ups of the visual output and guidelines to follow, were created. Programming and development of the pilot dashboard then began using R and Spotfire as the tools of choice. Results: The resulting dashboard branded “OED Fast Insights” included a standard set of pages composed of dynamic tables and graphs organized around 3 components: study disposition, disease response, and treatment-emergent adverse events. The output closely resembled tables and graphs the clinical study team were familiar with through statistical programming output using SAS. However, leveraging the interactive functionality of Spotfire allowed the team to produce dynamic tables and graphs that could be readily filtered by different pre-specified criteria (e.g. cohort, dose level, gender, lines of therapy, best overall response, type of cancer, biomarker status, etc.) or by subject or aggregate-level views, changed by different color indication, or adjusted in other ways by the user. The dynamic aspect of spider plots, waterfall plots, and swimmer lanes were of special interest in identifying early signal detection. Conclusions: The cross-functional initiative began in 2018 and by the end of 2019 the data surveillance framework was successfully scaled and rolled-out to the entire early oncology portfolio of existing trials utilizing RECIST 1.1 solid tumor criteria. The next phase of the initiative will be to extend the framework beyond RECIST 1.1 efficacy surveillance to include hematologic cancers.