Baseline predictors of early mortality in patients with PD-L1–high advanced NSCLC treated with first-line immunotherapy.

Abstract

e20691 Background: Single-agent immunotherapy (IT) is the standard first-line treatment for patients (pts) with advanced non–small cell lung cancer (aNSCLC) and PD-L1 ≥50%. Its favorable tolerability may increase use in frail pts or near the end of life. Identifying baseline features associated with lack of benefit is therefore crucial. Methods: We retrospectively analyzed pts with aNSCLC and PD-L1 ≥50% treated with first-line single-agent IT at Modena University Hospital (June 2020–February 2025). Data cut-off was June 30, 2025. Early mortality (EM) was defined as death within 3 months from treatment initiation. Group comparisons used logistic regression or Fisher’s exact test, as appropriate. Overall survival (OS) was estimated by Kaplan–Meier. Results: Fifty-four pts were included; 13 (24%) experienced EM, including 5 pts (9%) who died within the first month. 12 of 13 pts (92.3%) in the EM group received IT within the last 30 days of life. EM was significantly associated with poor performance status (PS) ≥2 (69% vs 34%; Odds Ratio (OR) 4.34, 95% CI 1.10–17.16; p = 0.032), and with ≥2 comorbidities (69% vs 37%; OR 3.90, 95% CI 0.99–15.35; p = 0.046). Never-smoking status was observed exclusively in the EM group (23% vs 0%; p = 0.022). Hospitalization at diagnosis (62% vs 49%) and baseline steroid use (62% vs 44%) were numerically higher in EM pts, without reaching statistical significance. EM was more frequent in pts with bone (85% vs 24%; OR 17.05, 95% CI 3.09–93.93; p = 0.001) and liver metastases (39% vs 12%; OR 4.50, 95% CI 1.01–20.01; p = 0.043), whereas the presence of ≥2 disease-related symptoms was not associated with EM. EM was also associated with poor Lung Immune Prognostic Index (LIPI) score (77% vs 32%; OR 7.31, 95% CI 1.32–40.59; p = 0.020) and higher Palliative Prognostic (PaP) score (mean 6.23 vs 3.71; p = 0.016). Age, sex, BMI, and PD-L1 expression level were not associated with EM. In the overall population, ECOG ≥2 and PaP score were significant predictors of OS. Median OS was 6.0 vs 23.2 months (ECOG ≥2 vs < 2; p = 0.003) and 4.8 vs 20.8 months (PaP B+C vs A; p = 0.003). Conclusions: In pts with aNSCLC and PD-L1 ≥50%, EM following initiation of first-line single-agent IT is not uncommon and may be driven by identifiable baseline clinical and prognostic factors. Poor PS, multiple comorbidities, selected metastatic patterns, and validated prognostic scores such as LIPI and PaP were statistically associated, in this study, with a high risk of EM. Reassessment of these readily available prognostic tools may improve patient selection, help avoid futile treatment, limit unnecessary toxicity and resource utilization, and reduce financial toxicity at the end of life. Prospective studies with larger cohorts are warranted to further refine selection strategies and preserve quality of life.