Abstract / Summary
Background: The prognostic relevance of PDCD-related genes in non-small cell lung cancer remains uncertain. We examined PDCD2, PDCD4, PDCD5, PDCD10, and AVEN separately in lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC). Methods: Matched TCGA expression cohorts comprised 492 LUAD and 484 LUSC tumors. Primary Cox models evaluated continuous expression per standard deviation using separate gene-specific models adjusted for age, sex, smoking history, and pathologic stage. Benjamini–Hochberg correction covered ten gene–histology tests. Proportional hazards, multicollinearity, model specification, and the incremental value of PDCD4 were assessed. GSE30219 provided 83 LUAD and 61 LUSC patients for unadjusted and clinically adjusted external analyses. Hallmark enrichment was evaluated using legacy RSEM expression and a STAR TPM sensitivity analysis. Results: The primary TCGA models included 469 LUAD patients with 167 deaths and 467 LUSC patients with 205 deaths. AVEN was associated with poorer LUAD survival after correction (HR 1.272, 95% CI 1.091–1.483; q = 0.022). PDCD4 was not significant in the primary LUAD model (HR 0.875, 95% CI 0.749–1.021; q = 0.150). Adding PDCD4 to clinical variables yielded a C-index increase of 0.014 (bootstrap 95% CI −0.004 to 0.031). No adjusted external association survived correction. PDCD4-associated enrichment differed between expression pipelines. Conclusions: AVEN showed an exploratory adverse association in TCGA LUAD, but external prognostic replication was not established. These analyses do not support a robust independent prognostic claim for PDCD4 or the clinical use of any of the five evaluated genes.