OBJECTIVES: Measuring progression-free survival (PFS) in real-world is challenging; due to the variability in assessments’ timing and missing data. In aNSCLC, patient-level analysis conducted by the FDA showed that TTD may represent a pragmatic alternative endpoint for PFS. To date, the relationship between TTD and OS is yet to be tested in experimental and real-world settings. Therefore, we aimed to assess if TTD can be used as a pragmatic predictor of OS in real-world.

METHODS: For this application, we limited our analysis to first-line systemic treatment in a German claims-based dataset for aNSCLC. The correlation between TTD and OS was estimated using Pearson's r, Spearman's rho, and Kendall’s τ. Kendall’s τ was computed using methods for doubly censored data. Landmark analyses were also performed at several timepoints to estimate OS in two groups of patients, conditional on TTD at a specific timepoint: the landmark time. Multivariate and univariate Cox proportional hazard models were fitted to estimate the effect of TTD on OS. Hazard ratios (HRs) together with the 95% confidence interval (CI) were estimated.

RESULTS: 1,672 patients were included in the analysis. Median TTD was 5.75 months (95%CI:5.62-5.98) and median OS was 9.90 months (95%CI:9.24-10.59). Correlation between OS and TTD was statistically significant (Pearson's r=0.44 (p<.001), Spearman's rho=0.61 (p<.001) and Kendall’s τ=0.32 (p<.001). TTD was consistently found to be a strong predictor of OS at all timepoints. At a landmark timepoint of 6 months, the HR for death was 0.717 (95%CI:0.623-0.826) for patients who were on treatment compared to those who discontinued treatment.

CONCLUSIONS: TTD is strongly correlated with OS in real-world, may predict OS and be used as a patient-relevant endpoint in the first-line aNSCLC. Prior to the usage of TTD in clinical trials as a key endpoint, further investigation of this relationship in different oncology settings is warranted.