ESTRO 2020 Abstract book

S727 ESTRO 2020

PO-1288 R Shiny and Flexdashboards: Utilizing Open Source Tools for Interactively Tracking Clinical Trials M. Golafshar 1 , M. Petersen 1 , T. DeWees 1 1 Mayo Clinic, Division of Biomedical Statistics and Informatics, Scottsdale, USA Purpose or Objective Effectively tracking clinical trials is essential to ensuring trials meet accrual and stratification goals while monitoring adverse events, quality of life (QOL), and requirements that may lead to protocol deviations. Historically, trials are monitored through scheduled progress and safety reports; these are not continuously up- to-date or immediately accessible to the investigator. The financial and person-hour cost of utilizing clinical research staff for this purpose is impractical. Herein, we present online dashboards created from open-source platforms for monitoring clinical trials which provide continuously updated, on-demand access to the data; significantly aiding in successful research. Material and Methods To build dashboards for monitoring clinical trials, we utilize Flexdashboard, a free library from RStudio. Flexdashboard is an RMarkdown framework that cuts through the upfront cost and effort usually associated with dashboard deployment. By leveraging R, RMarkdown and Shiny simultaneously, Flexdashboard allows us to weave together all the elements necessary for building beautiful, interactive dashboards. The R programming language provides the backbone that allows users to pull data from a number of sources, transform and analyze that data, and then build the visual pieces needed for communicating data to stakeholders. Anything that can be built in R can be deployed from within Flexdashboard. Furthermore, by utilizing time-based schedulers (CRON) and application programming interfaces (APIs) dashboard updates can be automated. When deploying to a Shiny server, investigator-approved personnel can stay up-to-date through a secure web link. Results We have deployed the Flexdashboard framework in a number of prospective studies; witnessing increased protocol adherence by keeping investigators and clinical staff apprised on all study aspects, in one always up-to- date web link. The studies have benefitted from the on- demand access that dashboards provide. Monitoring accrual and stratification factors helps keep trials moving towards import milestones; immediate access to adverse events ensures the safety of the study; and accessing survey completion rates across key time-points allows for immediate patient consultation to retrieve protocol- related QOL forms. While these objectives are essential to effective clinical trials, dashboards can be extended to accomplish any investigator demand. Dashboards can track lab results, biomarker data, treatment plan parameters (e.g., DVH values) and any other aspect of a clinical trial; creating a hub for on-demand monitoring of not only clinical trials, but also registries and research databases. Predictive models can also be employed to aid in patient- level consultation of predicted clinical and QOL outcomes. Conclusion Flexdashboard is a tool that allows for rapid deployment of dynamic dashboards which provide investigators continuous, on-demand access to clinical trial data, yielding increases in successful accrual and protocol adherence.

Poster: Clinical track: Other

PO-1289 Patient Reported Outcomes: Using ESAS to screen for anemia P. Johnstone 1 , R. Alla 1 , H. Yu 1 , D. Portman 2 , R. Mitchell 3 , H. Jim 4 1 Moffitt Cancer Center, Radiation Oncology, Tampa, USA ; 2 Moffitt Cancer Center, Supportive Care, Tampa, USA ; 3 Moffitt Cancer Center, Biostatistics & Bioinformatics, Tampa, USA ; 4 Moffitt Cancer Center, Health Outcomes & Behavior, Tampa, USA Purpose or Objective Patient perspectives of their symptom burden provide valuable data to clinicians. We have investigated the Edmonton Symptom Assessment Scale (ESAS) extensively in our radiation oncology and supportive care clinics: reported levels of 7 or above for any symptom drives investigation and intervention if necessary. Since anemia is seen in up to 90% of cancer patients and is associated with disease recurrence, decreased functional status, impaired quality of life and reduced survival, we examined whether ESAS data could correlate with anemia. Material and Methods Our clinics have used a modified ESAS since 2015; patients now input data directly into the electronic medical record using a tablet interface. Of 9813 patients providing ESAS reports we retrieved hemoglobin (Hb) data from 8304. Of these, 1351 patients had both performed on the same day. Anemia existed if Hb was <13.0 g/L (man) or <12 g/L (woman). Results In these data, Tiredness had the highest reported mean severity at just over 4/10. Of the other reported symptoms, only Pain had a mean score exceeding 3/10. When self-reported scores for both Tiredness and Shortness of Breath were 7 or above, the positive predictive value (PPV) for anemia was 80% and specificity was 97.6%. Corresponding sensitivity was 8.2% and accuracy was 48.9%. This 2-item model could be a valuable screening tool for lack of anemia in cancer patients in the outpatient setting: if patients rate both