Novel Multiplex Immunofluorescence-based Tumor Inflammation Score Provides Apparent Predictive Biomarker in a Phase I/II Study of Pembrolizumab With Gemcitabine in Patients with Previously-Treated Advanced Non-Small Cell Lung Cancer (NSCLC)

Retrospective characterization of cell-cell relationships in the tumor microenvironment provides significantly better predictive power than PD-L1 expression, tumor mutational burden (TMB), or gene expression profiles. In this small study assessing the safety and possible efficacy of gemcitabine and pembrolizumab in immunotherapy-naive patients with NSCLC who have received prior treatment, we investigated both standard and novel immune parameters on 16 enrolled patients. The combination of gemcitabine and pembrolizumab could be administered safely but did not demonstrate synergism compared with historical controls. Novel findings of this study are that elevated frequencies of Tregulatory cells near CD3 T cells at baseline was associated with improved outcome to treatment (p<0.05). Integrating this Treg / T cell relationship metric together with overall T cell density yielded a tumor inflammation score which correlated (p<0.002) with disease response. We postulate this is indicative of ongoing anti-cancer immune response. Additionally, while prior studies documented that IgG Ab responses to TAA can identify targets of a coordinated T and B cell response and evidence of immune surveillance, this study found that high autoantibody responses, while not statistically significant, trended towards a worse outcome (p=0.06). This suggests to us that tumors may have developed mechanisms to escape the immune response to these TAAs.