OBJECTIVES: Real world observational studies provide valuable data on key clinical endpoints but are often very complex, particularly in oncology. Collected data usually includes efficacy endpoints such as overall survival (OS) and progression-free survival (PFS) as well as information on treatment sequences and large volumes of adverse events (AEs) data. Despite significant software advances to describe such data, very little progress has been observed/published in terms of graphical representation. Our work aims at exploring various data visualization outputs to enable easier interpretation and dissemination of results within the scientific community.

METHODS: We demonstrate the use of relevant plots that can benefit data exploration by combining multiple dimensions. Data from an observational retrospective study in metastatic colorectal cancer were used for the purpose of illustration. Levels included conjunctions of main objectives, which were treatment sequence patterns, OS and PFS. In addition, adverse event visualization was a special focus.

RESULTS: Several visualizations were selected and included non-parametric survival curves for both OS and PFS based on treatment sequences. Visualizations such as sunburst plots or spatiotemporal graphics can provide a clear view of treatment lines or associate treatment line duration to survival outcomes. AE visuals of time to onset and volcano plots to help identify extreme events were also developed. All data visualizations will be presented with associated interpretation. Tips and software modules for visualization generation will also be provided.

CONCLUSIONS: There is limited guidance on data visualization tools during the review of data and results of real-world studies. We propose visual aids, based on the dataset’s multiple dimensions, to aid the assimilation and interpretation of the data. To summarize, we believe such visualizations could be of outstanding help throughout the whole study, particularly during the data review process, to support decisions relative to missing data mechanisms.