(B-225) Utility of Restricted Mean Duration of Response as an Intermediate Outcome in Advanced Non-small Cell Lung Cancer

Background: Valid intermediate endpoints can expedite the identification of effective cancer treatments, enabling prioritization of promising therapies for further development where long-term outcomes such as overall survival are evaluated. In oncology trials, the most common intermediate endpoint is progression-free survival (PFS). However, recent studies suggest that restricted-mean duration of response (rmDOR), the average time spent in response for the cohort as a whole, may be more sensitive than PFS at detecting intermediate treatment effects [Hu, 2021].

Objectives: To evaluate the correlation between rmDOR, objective response rate (ORR) and restricted mean PFS (rmPFS) with restricted mean overall survival (rmOS) in metastatic non-small cell lung cancer (mNSCLC) using real-world data.

Methods: Using retrospective electronic health records from COTA NSCLC from 2015-2023, three mNSCLC cohorts were identified: 1) 2L+ docetaxel (n = 276); 2) 1L anti-PD-(L)1 monotherapy (n = 479); 3) 1L anti-PD-(L)1 + platinum chemotherapy (n = 793). Intermediate outcomes rmDOR and rmPFS were measured at 9 and 12 months, and rmOS was tabulated at 18, 24, and 30 months. Overall response rate (ORR) was assessed during the treatment period. The rmDOR, was defined as the area between the PFS and response/ progression-free survival (RPFS) curves for each cohort, calculated using R-package PBIR [Luo, 2020]. RPFS was defined as the time from start of treatment to the first of complete or partial response, disease progression, or death. A bootstrap sampling technique was utilized to create 1000 samples of 150 patients in each cohort to compute Pearson correlation statistics between effectiveness measures.

Results: Across the three mNSCLC cohorts and time periods of interest, rmDOR was strongly to very strongly correlated with rmPFS (range: r= 0.619 – 0.818) and ORR (range: r=0.711 – 0.825). However, rmDOR was only moderately correlated with rmOS (range: r= 0.461 – 0.610), whereas rmPFS was strongly correlated with rmOS (range: r=0.591 – 0.702). Finally, overall response rate (ORR) was only weakly to moderately correlated with rmOS (range: r = 0.357– 0.467). Results were similar across the three treatment cohorts and rmOS time periods. To our knowledge, this was the first study to evaluate relationships between rmDOR and rmOS using real-world cohort of individual treatments.

Conclusions: In real-world data, rmDOR was moderately correlated with overall survival in three mNSCLC cohorts across different therapeutic regimens. More research is needed to understand differences in rmDOR performance in clinical trial versus real-world data and in a single cohort versus comparative cohort setting.