Abstract
INTRODUCTION: There is currently limited information on the prognostic association of molecular aberrations in pleural mesothelioma. We evaluated the impact of common molecular alterations on overall survival (OS) in a three-institution cohort from the International Association for the Study of Lung Cancer Ninth Edition Staging Project as a pilot study to guide future data collection. METHODS: Biomarker data included programmed death-ligand 1 (PD-L1) and BAP1 immunohistochemistry (IHC), and somatic genomic aberrations with putative functional impact (pathogenic) revealed by next-generation sequencing. OS was calculated by the Kaplan-Meier method and compared between groups by Cox proportional hazard regression. Models were compared using pseudo-R(2) and Harrell's C-statistic. RESULTS: Pathogenic alterations in BAP1 were most common (55.0%, 143/260), followed by CDKN2A (36.2%, 94/260), NF2 (23.8%, 62/260), and TP53 (18.1%, 47/260). Loss of BAP1 expression by IHC was detected in 260 of 452 cases (57.5%). PD-L1 expression was positive (≥1% of tumor cells with membranous staining) in 89 of 200 cases (44.5%). On univariate analysis, pathogenic alterations in CDKN2A, NF2, and TP53 were associated with worse OS (p ≤ 0.001); BAP1 alterations and loss of expression were associated with better OS (p = 0.004 and p < 0.001, respectively); and PD-L1 expression was not associated with OS (p = 0.645). Multivariable analyses confirmed OS associations (hazard ratios: CDKN2A, 2.12; NF2, 1.65; TP53, 1.74; BAP1 next-generation sequencing, 0.58; BAP1 IHC, 0.58). Although clinical covariates performed better than molecular alterations alone (C-statistic: 0.67 versus 0.65; R(2): 25.3 versus 20.8), a model including BAP1 alterations plus any pathogenic alteration in CDKN2A, NF2, and TP53, along with clinical covariates, best predicted survival (C-statistic: 0.70; R(2): 37.0). CONCLUSION: In this large pleural mesothelioma cohort, CDKN2A, NF2, and TP53 alterations were associated with worse OS, whereas BAP1 alterations were associated with better prognosis, independent of clinical variables and histology. Modeling suggests that genomic alterations provide additional prognostic information beyond anatomic TNM and clinicopathologic features.