(074) Progression-Free Survival (PFS) as a Surrogate for Overall Survival (OS) in Advanced Non-Small Cell Lung Cancer (aNSCLC): Analysis of Patient-Level Pooled Randomized Clinical Trials (RCTs)

Background: Progression-free survival (PFS) is often used as a primary endpoint in randomized clinical trials (RCTs) for advanced non-small cell lung cancer (aNSCLC). While previous studies have indicated modest trial-level surrogacy of PFS for overall survival (OS) with first-line (1L) cytotoxic chemotherapy (chemo) (R2 < 0.70) [Buyse et al. J. Clin. Oncol. 2008; Laporte et al. BMJ Open 2013], the emergence of novel treatments (txs) such as immune checkpoint inhibitors (ICIs) necessitates re-evaluation.

Objectives: To evaluate whether PFS is a valid surrogate endpoint (SE) for OS in aNSCLC patients (pts) receiving 1L tx with an ICI-containing regimen (ICI+) or chemo at both the pt-level and trial-level.

Methods: Pt-level data from 2 RCTs (response assessed by RECIST v1.1) comparing ICI+ to chemo for the 1L tx of aNSCLC were sourced from the Medidata Enterprise Data Store and pooled. An unadjusted Cox model using landmark (LM) analysis at 24 weeks assessed the pt-level correlation between PFS and OS. The correlation between the Tx effects (TxEs) on PFS and on OS was analyzed between each ICI+ and chemo arm using weighted linear regression (R2WLS) and bivariate Copula (Copula R2), with the country of pt enrollment as the unit of analysis. A high correlation was defined as an R2 value ≥ 0.85. The surrogate threshold effect (STE) was also reported.

Results: Among 1862 pts (50.2% aged ≥ 65 years (y), 70.0% male, 72.4% White, 25.4% Asian, and 2.2% Other/Missing, median (IQR) follow-up of 1.1 y (1.8), study txs ICI+ (60.6%) and chemo (39.4%), the overall response rate was 31.4%. The median OS and PFS were 12.6 [95% confidence interval: 11.9-13.4] and 4.6 [4.3-5.3] months (mos), respectively. 1400 pts without death or censoring before the LM date were included in the LM analysis. At the pt-level, LM analysis showed pts with PFS beyond the LM date had better OS (HR, 0.40 [0.36, 0.46]). For the evaluation of trial-level surrogacy, pts were divided into 46 subunits (median: 48.1; range: 11- 116 pts). At the trial subunit-level, PFS showed weak surrogacy for OS (R2WLS = 0.51 [0.30, 0.72]; Copula R2 = 0.68 [0.41, 0.81]). The STE was 0.47, indicating that only a new tx with a major impact on PFS (at least a 53% risk reduction) would be expected to have a significant effect on OS.

Conclusions: This study found a robust pt-level correlation between PFS and OS, and trial-level R2 values that were similar to values previously reported with chemo. These findings suggest that PFS is, at best, a modest trial-level surrogate for OS in the 1L tx of aNSCLC with ICI+ vs chemo, despite a high pt-level correlation. Few RCTs and a limited range of TxEs constrained this investigation. Further investigation with more RCTs and a greater variety of txs may be warranted.