3513 - Dissemination Phenotypes after First-Line Immunotherapy in Metastatic NSCLC: Implications for Metastasis-Directed Therapy

Purpose/Objective(s): Dissemination patterns at first immunotherapy (IO) failure in metastatic NSCLC are poorly characterized in systemic therapy trials. Metastasis-Directed Therapy (MDT) candidacy is largely determined by lesion number at progression (oligoprogression, OP), yet whether this captures the biologic heterogeneity of failure remains unclear. We hypothesized that systemic progression (SP) reflects biologically distinct systemic escape with new-organ seeding, whereas OP represents spatially contained failure.

Materials/Methods: This was a retrospective cohort study of Stage IV NSCLC patients treated with first-line IO at a single institution (2016-2024). Inclusion required biopsy-proven disease, baseline next-generation sequencing and imaging, and =1 on-treatment scan. At first radiographic progression, phenotype was classified as OP (=3 progressing lesions in =2 organs, excluding pleural effusion) vs SP (all others). The primary endpoint was incident progression in previously uninvolved organs at first IO failure. Associations were tested with Fisher’s exact test and multivariable logistic regression. Trend in SP frequency across baseline burden strata was assessed with ordinal trend analysis. Sensitivity analyses included temporal cohorts and alternative OP thresholds.

Results: Of 152 patients treated with first-line IO, 109 met inclusion criteria and 87 developed radiographic progression (median follow-up 24.9 mo). Median baseline burden was 6 lesions (IQR 4–10) across 3 organs (IQR 2–4). Common genomic alterations included TP53 (56%), KRAS (38%), and STK11 (23%). Among 87 progressors (43 OP, 44 SP), progression in previously uninvolved organs occurred in 82% of SP and 30% of OP (OR 10.4, 95% CI 3.8–28.4; p<0.0001), and remained independently associated with SP after adjustment for baseline burden, genomic subgroup, and PD-L1 TPS (aOR 16.1, 95% CI 4.7–55.5; p<0.0001). Progression remained confined to baseline-involved sites in 38/87 patients (44%), predominantly in OP (70% vs 18%). SP was enriched for progression involving sanctuary sites (CNS/liver/pleura) (26/44 vs 4/43; aOR 17.1, 95% CI 4.3–46.4; p<0.0001). SP frequency increased with baseline lesion burden (<4, 31%; 4–6, 32%; 7–10, 62%; >10, 76%), with ordinal analysis confirming a stepwise increase (OR 2.21 per stratum, 95% CI 1.37–3.76; p=0.0008). Findings were robust across sensitivity analyses (p<0.0001). Baseline metastatic burden independently stratified PFS, with >5 lesions associated with shorter PFS than 1–3 lesions after adjustment for age, PD-L1 TPS, and genomic subgroup (HR 1.73, 95% CI 1.05–2.86; p=0.031). Using a CURB-like definition (=5 lesions), 76% were potentially eligible for MDT, yet only 41% received RT.

Conclusion: First IO failure comprises biologically distinct dissemination states. Baseline lesion burden and progression phenotype may inform MDT candidacy in addition to lesion number, and support phenotype-guided prospective trials of MDT versus systemic escalation.