(B-222) Comparison of time-to-next treatment and time-to-event endpoints using real-world data in checkpoint inhibitor-treated patients with advanced non-small cell lung cancer

Background:

Background: Progression-free survival (PFS), an endpoint in cancer trials, requires costly and time-consuming curation of clinical data not routinely available in real-world (RW) data. RW time-to-next treatment (rwTTNT) is used as a surrogate for rwPFS due to data availability. Agreement of rwPFS with rwTTNT hasn’t been well documented, and different rwTTNT calculations are used, commonly, time-to-next treatment or death (rwTTNT-D).

Objectives:

Objectives: Compare estimates and correlation between rwPFS and rwTTNT-D in adults with advanced non-small cell lung cancer (aNSCLC). Evaluate the associations of different TTNT definitions with rwPFS.

Methods:

Methods: aNSCLC adults treated with checkpoint inhibitors (CPI) and platinum chemotherapy in 1L and subsequent therapy from the nationwide Flatiron Health electronic health record-derived deidentified database with enhanced rwPFS extractions from 2011 to 2023 were included. rwTTNT-D was compared to rwPFS by line of therapy (LOT) using Kaplan-Meier analysis (median months with 95% confidence interval [CI]) and Spearman’s correlation (r). Comparisons to rwPFS were also conducted with alternative TTNT definitions handling death differently: censoring patients at last activity before death (rwTNTT-A); and excluding patients who died (rwTTNT-E).

Results:

Results: The dataset included 947 patients starting at 2L with a median age of 70 years, of whom 347 received treatment in 3L and 133 in 4L+. RwTTNT-D was highly correlated with rwPFS across LOT (2L: r=0.82, 3L: r=0.86, 4L+: r=0.89). RwTTNT-D and rwPFS had comparable medians (95% CI) (2L: 7.1 [6.4, 7.8] vs 6.9 [6.1, 8.1]; 3L: 5.3 [4.6, 6.2] vs 6.1 [4.8, 7.5]; 4L+: 4.1 [2.9, 5.1] vs 4.9 [3.2, 6.3] months, respectively). Of the 947 patients, 365 died and impacted rwTNTT-D, -A, and -E estimates. All rwTTNT definitions were highly correlated with rwPFS (r >0.80) across LOT. However, differences were observed in estimates at 2L (rwTTNT-A: 13.1 [12.2, 14.6]; rwTTNT-E: 11.0 [9.2, 12.1] vs. rwTTNT-D: 7.1 [6.4, 7.8]) and in later LOT.

Conclusions:

Conclusions: Among adults with aNSCLC treated with CPI and chemotherapy, rwTTNT-D estimateswere comparable to and strongly correlated with rwPFS. This confirms its validity as a clinically meaningful surrogate for rwPFS in RW data. While all rwTTNT definitions were highly correlated with rwPFS, rwTTNT-D was most comparable.