Progression-Free Survival and Objective Response Rates As Surrogate Endpoints for Overall Survival Among Patients With ROS1+ Locally Advanced or Metastatic Non-Small Cell Lung Cancer Receiving ROS1 Tyrosine Kinase Inhibitors
Author(s)
Ellen E. Korol, MSc1, Yong Yuan, PhD2, Juan Pablo Diaz-Martinez, PhD3, Adam Lee, MSc4, Christophe Y Cavet, PhD5, David Ross Camidge, PhD6, Sarah Goring, MSc1.
1Broadstreet HEOR, Vancouver, BC, Canada, 2Bristol Myers Squibb, Princeton, NJ, USA, 3Broadstreet, Vancouver, BC, Canada, 4Bristol Myers Squibb, Uxbridge, United Kingdom, 5Bristol Myers Squibb, Princeton, BC, Canada, 6Cancer Center, University of Colorado, Aurora, CO, USA.
1Broadstreet HEOR, Vancouver, BC, Canada, 2Bristol Myers Squibb, Princeton, NJ, USA, 3Broadstreet, Vancouver, BC, Canada, 4Bristol Myers Squibb, Uxbridge, United Kingdom, 5Bristol Myers Squibb, Princeton, BC, Canada, 6Cancer Center, University of Colorado, Aurora, CO, USA.
OBJECTIVES: European Joint Clinical Assessment documentation specifies criteria for establishing surrogate relationships: level 1) evidence of corresponding treatment effects; level 2) association between outcomes; and level 3) biological plausibility. To support decision-making for repotrectinib, we aimed to estimate the relationship between overall survival (OS) and response-based endpoints (progression-free survival [PFS] or objective response rate [ORR]) among patients with ROS1+ locally advanced or metastatic non-small cell lung cancer (aNSCLC) who are ROS1 tyrosine kinase inhibitor (TKI)-naïve or -experienced.
METHODS: A systematic literature review was conducted, capturing clinical trials among patients with ROS1+ aNSCLC, treated with TKIs, and reporting OS and either PFS or ORR. No randomized controlled trials (RCTs) were identified, precluding comparison of treatment effects. Aggregate arm-level correlations were conducted using weighted linear regression to compare median OS versus: median PFS; and percent ORR. TKI-naïve and -experienced populations were analyzed separately. R2 was calculated along with 95% confidence intervals (CIs) using a bootstrapping method.
RESULTS: Twelve cohorts from non-randomized clinical trials involving treatment with crizotinib, entrectinib, lorlatinib, brigatinib or ceritinib were identified; repotrectinib cohorts were excluded. Nine cohorts consisting of 540 patients contributed to the analysis. Median OS ranged from 12.2 to 51.4 months. Among TKI-naïve patients, a strong correlation was observed between median OS and median PFS (6 cohorts; R2 = 0.97, 95% CI: 0.84, 1.00), but low correlation was observed versus percent ORR (5 cohorts; R2 = 0.22: 0.00, 1.00). The evidence base among TKI-experienced patients (3 cohorts) was considered insufficiently robust for formal analysis.
CONCLUSIONS: The current analysis demonstrated a strong association between OS and PFS among TKI-treated patients with ROS1+ aNSCLC, lending support to previous real world evidence evaluating surrogacy of PFS. However, OS and ORR demonstrated a weak association with substantial uncertainty. The lack of head-to-head evidence precluded assessment of the correlation between treatment effects.
METHODS: A systematic literature review was conducted, capturing clinical trials among patients with ROS1+ aNSCLC, treated with TKIs, and reporting OS and either PFS or ORR. No randomized controlled trials (RCTs) were identified, precluding comparison of treatment effects. Aggregate arm-level correlations were conducted using weighted linear regression to compare median OS versus: median PFS; and percent ORR. TKI-naïve and -experienced populations were analyzed separately. R2 was calculated along with 95% confidence intervals (CIs) using a bootstrapping method.
RESULTS: Twelve cohorts from non-randomized clinical trials involving treatment with crizotinib, entrectinib, lorlatinib, brigatinib or ceritinib were identified; repotrectinib cohorts were excluded. Nine cohorts consisting of 540 patients contributed to the analysis. Median OS ranged from 12.2 to 51.4 months. Among TKI-naïve patients, a strong correlation was observed between median OS and median PFS (6 cohorts; R2 = 0.97, 95% CI: 0.84, 1.00), but low correlation was observed versus percent ORR (5 cohorts; R2 = 0.22: 0.00, 1.00). The evidence base among TKI-experienced patients (3 cohorts) was considered insufficiently robust for formal analysis.
CONCLUSIONS: The current analysis demonstrated a strong association between OS and PFS among TKI-treated patients with ROS1+ aNSCLC, lending support to previous real world evidence evaluating surrogacy of PFS. However, OS and ORR demonstrated a weak association with substantial uncertainty. The lack of head-to-head evidence precluded assessment of the correlation between treatment effects.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
CO58
Topic
Clinical Outcomes
Topic Subcategory
Relating Intermediate to Long-term Outcomes
Disease
SDC: Oncology