Noninvasive imaging biomarkers for survival risk stratification with tarlatamab in ES-SCLC.

Abstract

8020 Background: Subsequent treatment for extensive-stage small cell lung cancer (ES-SCLC) following platinum-etoposide chemotherapy +/- an immune checkpoint inhibitor was historically limited to alternative forms of chemotherapy with relatively poor clinical outcomes. Bispecific delta-like ligand 3 (DLL3)-targeting T-cell engagers, such as tarlatamab, have since demonstrated clinically meaningful activity with notable improvements in survival. Although there was a modest improvement in median progression-free survival (PFS) with tarlatamab in the phase 3 DeLLphi-304 trial, 20% of patients remained free from disease progression at 12 months. Unfortunately, conventional clinical factors and early radiographic responses are unable to distinguish transient from sustained benefit; thus underlying the need for validated biomarkers to identify patients likely to derive a durable response from tarlatamab. Methods: This study sought to evaluate whether radiomic texture and quantitative vessel tortuosity (QVT) metrics derived from baseline CT imaging are associated with PFS and overall survival (OS) in patients with ES-SCLC receiving tarlatamab.50 patients from Cleveland Clinic with ES-SCLC who received tarlatamab after front-line chemoimmunotherapy were included in this study. Radiomic texture and QVT features characterizing intratumoral heterogeneity and vascular architecture were extracted from CT scans obtained before the first dose of tarlatamab (median scan-to-treatment interval: 33.5 days). A least absolute shrinkage and selection operator Cox regression model with cross-validation identified prognostic features for PFS and OS. A radiomic risk score (RRS) was computed as a weighted linear combination of selected features, and patients were stratified into high- and low-risk groups using the median RRS. Cox regression and Kaplan–Meier analyses with log-rank testing evaluated associations with PFS and OS. Results: On univariable analysis, RRS was significantly associated with PFS (HR = 2.3, 95% CI 1.36–3.94, P = 0.0019) and OS (HR = 3.9, 95% CI 1.67–9.1, P = 0.0016). Kaplan–Meier analyses demonstrated significantly shorter PFS and OS in the high-risk group. In multivariable models adjusting for age, sex, race, smoking history, COPD, and prior lines of therapy, RRS remained the only factor independently associated with PFS (HR = 2.1, 95% CI 1.16–4.0, P = 0.014) and OS (HR = 5.87, 95% CI 2.0–17.0, P = 0.001). Conclusions: Baseline CT-derived radiomic and vascular features represent noninvasive biomarkers for predicting PFS and OS in patients with ES-SCLC receiving tarlatamab. Further validation of these novel features can improve patient selection, treatment sequencing, and optimized utilization of DLL3-targeted therapy in this unique cancer population with otherwise limited treatment options.