Association between molecular subtype membership or hypoxia-associated gene expression signatures and clinical outcomes in the CALGB 90601 (Alliance) phase 3 clinical trial of gemcitabine and cisplatin (GC) plus bevacizumab (B) or placebo (P).

4562 Background: Our previous work showed that basal tumors were associated with the best clinical outcomes in a Phase 2 clinical trial of neoadjuvant dose-dense MVAC plus B, and in other work we showed that basal tumors were enriched with hypoxia-associated gene expression signatures. Here we attempted to validate these findings in the C90601 Phase 3 clinical trial of GC plus B versus GC plus P. Methods: Whole transcriptome RNAseq was performed on all available tumors using Ion Torrent’s Ampliseq platform (n = 189). Tumors were assigned to molecular subtypes using 3 different classifiers - BASE47 (k=2), MDA oneNN (k=3), and the Consensus classifier (k=6). Tumor hypoxia signature enrichment was determined using 2 different gene expression signatures and gene set variation analysis (GSVA). The proportional hazards model was used to correlate molecular subtype calls and hypoxia signature enrichment with overall survival (OS) and progression-free survival (PFS) adjusting for stratification factors and treatment arm (for PFS). Results: The median OS & PFS by different signatures and the hazard ratios (HR) are presented in the Table. Conclusions: Predefined signatures associated with clinical benefit in the Phase-2 neoadjuvant clinical trial were not associated with benefit in C90601. Possible explanations include the lack of strong therapeutic effects of the treatments, potential heterogeneity (“subtype plasticity”) between the profiled tissue samples and the metastatic lesions under treatment pressure, and differences in biology associated with the disease states (muscle-invasive vs advanced/metastatic disease). Support: U10CA180821, U10CA180882, Department of Defense (CA160312), Genentech; ClinicalTrials.gov Identifier: NCT00942331. [Table: see text]