A trial-level correlation analysis of progression-free survival as a surrogate for overall survival in the immunotherapy era.

Abstract

e18580 Background: Progression-free survival (PFS) is accepted as a surrogate endpoint for cancer drugs approval by the Food and Drug Association. In the expanding immunotherapy era, there is no convincing data for reliable surrogacy of PFS. Methods: We systematically searched PubMed and included all phase III trials of ICIs. Besides, we manually included phase III trials of ICIs presented in congresses. Trials including non-metastatic patients and hematologic malignancies were excluded. The primary endpoint was the relationship between treatment effects (i.e., hazard ratio (HR) of PFS and overall survival (OS)). The secondary endpoint was the relationship between the median PFS (mPFS) and median OS (mOS). We also performed subgroup analyses by ICIs subtypes (i.e., anti-programmed death-1 (PD-1) and anti-programmed death ligand-1 (PD-L1)), treatment lines (i.e., the first-line and subsequent treatments), ICI plus CT combination, and ICI monotherapy. To ensure parametric assumptions, we addressed logarithmic conversion to HRs by using log10. We used Pearson’s correlationand weighted linear regression for parametric variables and Spearman’s rank-order correlation for non-parametric variables. We assessed the relationship between variables by using correlation coefficient (r) and coefficient of determination (R2). Results: We included 57 phase III clinical trials with 39,525 patients. A total of 67 outcomes were compared. There was a good correlation between the logarithmic HR of PFS (log HRPFS) and OS (log HROS) (r=0.71, R2=0.50 p<0.001). In subgroup analyses, there was a weak or moderate correlation between the log HRPFS and log HROS in anti-PD-L1, ICI plus CT, and the first-line treatment subgroups (r=0.61, R2=0.38 p=0.004 for anti-PD-L1; r=0.65, R2=0.43 p=0.008 for ICI plus CT; r=0.70, R2=0.49 p<0.001 for the first-line treatment). Conversely, there was a good correlation between the log HRPFS and log HROS in the anti-PD-1, ICI monotherapy, and subsequent treatment line subgroups (r=0.81, R2=0.66 p<0.001 for anti-PD-1; r=0.71, R2=0.51 p<0.001 for ICI monotherapy; r=0.78, R2=0.62 p<0.001 for subsequent treatment line). Correlations between the mPFS and mOS were good or strong in all subgroups. (Table). Conclusions: PFS might be considered a reliable surrogate endpoint for OS, particularly in anti-PD-1 ICIs, ICI monotherapy, and subsequent metastatic cancer treatment lines.[Table: see text]