Presentation (CTI)

OBJECTIVES: The “Real wOrld PROgnostic score (ROPRO)” (Becker 2020) is a pan-cancer composite biomarker learned from 122,694 oncology patients from Flatiron Health® (Flatiron Health). Its prognostic power for overall survival (OS) was later confirmed in 17 independent clinical trials (Bauer-Mehren 2020) and 64,233 independent RWD patients (Becker 2022). The score is constituted from blood laboratory and vital parameters which are routinely available in clinical trials and real world data (RWD), also longitudinal. ROPRO can be interpreted as a measure of overall patient health and is not treatment-specific, which qualifies it as a tool for comparing efficacy signals across different therapies. An endpoint model to predict OS efficacy from ROPRO time course​ is available for NSCLC (Loureiro 2023).​
METHODS: From a de-identified, multimodal database of routine clinical care and treatment RWD (Tempus AI, Inc), we analyzed metastatic colorectal cancer (mCRC) patients, n=959 and n=509 at index line of treatment (LoT) 2 and 3, respectively. In a landmark analysis at 18 weeks, we correlated (real-world) objective response rate (ORR) and ROPRO-on-treatment values (RoT) to final OS. Next, we bootstrapped emulated trials and correlated differences of ORR/RoT between treatment arms to the final OS of the simulated trials (n=3,000).
RESULTS: In the landmark analysis, ORR and RoT at 18 weeks both correlated with post-landmark OS, HR-ORR = 0.66 [0.50;0.88], HR dichotomized RoT = 0.42 [0.33;0.55], for instance, at LoT2. In the simulation study, a combined model of early PFS, early OS, ORR and RoT had best correlation with final OS (r=0.82). Results from application to clinical trial data will be presented as available by the time of the meeting.
CONCLUSIONS: The analysis highlights the potential of leveraging RoT for early decision making. Combination with ORR or PFS/OS until a landmark optimizes performance.