Understanding variation in treatment sequences and outcomes in metastatic colorectal cancer: using real world data to answer real world questions

Abstract

Objectives The wide range of possible treatment combinations and sequences available for metastatic colorectal cancer (mCRC) treatment presents a major challenge to clinicians, who need to determine the optimal approach for an individual patient or patient subset. Real world data are a valuable resource to understand variation in treatment patterns and outcomes in routine practice. This study aimed to develop a data visualization tool to improve understanding of treatment complexity in mCRC and target further research efforts. Methods Real world data from an Australian mCRC registry were used to develop an online tool that visualizes variation in treatment sequences using interactive Sankey and sunburst diagrams. These diagrams were customizable to specific patient subsets based on patient and disease characteristics. To allow for different levels of detail, treatments were recoded according to different levels of abstraction. Results Of 2694 patients, 2057 (76%) started first-line treatment with chemotherapy or a biological agent, 1087 (40%) and 428 (16%) received second and third-line therapy, respectively. Combined, these three lines of treatment accounted for 733 unique sequences. After recoding treatment to the most intensive chemotherapy and the first exposed biologic, 472 unique sequences remained. The most frequent treatment decision by clinicians was whether to initiate first-line doublet chemotherapy with or without bevacizumab (n=1481, 72%), with median progression-free survival (95% confidence interval) of 10.1 months (9.5, 10.8) and 9.2 days (8.5, 10.7), respectively. Conclusions This initial exploration of the use of data visualization tools to inform understanding mCRC treatment practice, showed the potential for such tools to define variation in practice patterns and to identify opportunities to improve care and outcomes. Ultimately, clinicians and health system providers may use such tools to improve the delivery of personalized cancer care, where other applications such as health economic simulation models may be useful.