Correlation of gene expression signatures and clinical outcomes in patients with advanced gastric cancer treated with pembrolizumab (MK-3475).

Abstract

3026 Background: Pembrolizumab, an anti–PD-1 monoclonal antibody, has shown antitumor activity in advanced gastric cancers. Hypothesis testing of 4 prespecified multigene immune expression signatures enriched for T-cell mediated interferon signaling and antigen presentation genes was used to evaluate associations with clinical outcomes in patients with gastric cancer enrolled in KEYNOTE-012, a multicohort, phase 1 study of pembrolizumab in patients with PD-L1+tumors. Methods: 33 of 39 pembrolizumab-treated advanced gastric cancer patients had RNA expression profiling data of sufficient quality. RNA was extracted from formalin-fixed, paraffin-embedded samples and analyzed using the NanoStringnCounter system. Hypothesis testing was conducted on 4 prespecified signatures (“Interferon-gamma,” “TCR-signaling,” “Expanded-immune,” and “Denovo”) that were constructed based on analyses in patients with melanoma and prior to linking any NanoString data to clinical outcome data. Results: In the patients with NanoString data, ORR was 30%. For ORR, all 4 signatures showed a P value <0.10 (Table). For PFS, all signatures achieved P< 0.05. All signatures showed a general pattern such that pts with very low signature scores did not respond. ROC analyses suggest that cutoffs with high sensitivity and a resultant high negative predictive value (NPV) can readily be achieved. The cutoff associated with the Youden Index of the ROC curve for the Interferon-gamma signature yielded a positive predictive value (PPV) of 45% with a NPV at 92%. Conclusions: Immune response signatures were associated with clinical benefit from anti–PD-1 therapy with pembrolizumab in patients with advanced gastric cancer enrolled in KEYNOTE-012. Clinical trial information: NCT01848834. Signature Nominal One-Sided Pa ORR N = 33 PFS N = 33 IFNγ (6 gene) 0.077 0.032 TCR Signaling (13 gene) 0.034 0.024 Expanded Immune (18 gene) 0.062 0.049 Denovo (33 gene) 0.068 0.037 aFrom logistic or Cox regression for ORR and PFS, respectively.