Multicenter randomized phase II trial of atezolizumab with or without cobimetinib in biliary tract cancers

Abstract

BACKGROUNDMEK inhibitors have limited activity in biliary tract cancers (BTCs) as monotherapy but are hypothesized to enhance responses to programmed death ligand 1 (PD-L1) inhibition.METHODSThis open-label phase II study randomized patients with BTC to atezolizumab (anti-PD-L1) as monotherapy or in combination with cobimetinib (MEK inhibitor). Eligible patients had unresectable BTC with 1 to 2 lines of prior therapy in the metastatic setting, measurable disease, and Eastern Cooperative Oncology Group (ECOG) performance status less than or equal to 1. The primary endpoint was progression-free survival (PFS).RESULTSSeventy-seven patients were randomized and received study therapy. The trial met its primary endpoint, with a median PFS of 3.65 months in the combination arm versus 1.87 months in the monotherapy arm (HR 0.58, 90% CI 0.35-0.93, 1-tail P = 0.027). One patient in the combination arm (3.3%) and 1 patient in the monotherapy arm (2.8%) had a partial response. Combination therapy was associated with more rash, gastrointestinal events, CPK elevations, and thrombocytopenia. Exploratory analysis of tumor biopsies revealed enhanced expression of antigen processing and presentation genes and an increase in CD8/FoxP3 ratios with combination treatment. Patients with higher baseline or lower fold changes in expression of certain inhibitory ligands (LAG3, BTLA, VISTA) on circulating T cells had evidence of greater clinical benefit from the combination.CONCLUSIONThe combination of atezolizumab plus cobimetinib prolonged PFS as compared with atezolizumab monotherapy, but the low response rate in both arms highlights the immune-resistant nature of BTCs.TRIAL REGISTRATIONClinicalTrials.gov NCT03201458.FUNDINGNational Cancer Institute (NCI) Experimental Therapeutics Clinical Trials Network (ETCTN); F. Hoffmann-La Roche, Ltd.; NCI, NIH (R01 CA228414-01 and UM1CA186691); NCI's Specialized Program of Research Excellence (SPORE) in Gastrointestinal Cancers (P50 CA062924); NIH Center Core Grant (P30 CA006973); and the Passano Foundation.

Keywords: Cancer immunotherapy; Immunology; Liver cancer; Oncology.

Conflict of interest statement

Conflict of interest: MY received grant and research support from Incyte, Bristol-Myers Squibb, Exelixis, and Genentech/Roche and acted in a consulting or advisory role for Eisai, Exelixis, AstraZeneca, Genentech/Roche, and Geneos. NSA acted in a consulting or advisory role for QED, AstraZeneca, Mirati, and Taiho; received research funding from Celgene, Genentech, Astex Pharmaceuticals, Agios, Merck, Bristol-Myers Squibb, Syndax, Array BioPharma, Intensity Therapeutics, Atlas, Taiho, LOXO, and EMD Serono. GBL has consulted for and received compensation from ProDa Biotech, LLC and has received research funding from Merck and Co., Bristol-Myers Squibb, Boerhinger-Ingelheim, and Vaccinex.

Figures

Figure 1. Consort diagram.

Figure 2. Unstratified Kaplan-Meier plot of PFS for atezolizumab monotherapy (Arm A) and atezolizumab plus cobimetinib (Arm B).

Figure 3. Best response by RECIST 1. 1 among the evaluable patients treated with atezolizumab monotherapy and atezolizumab plus cobimetinib.

(A) Monotherapy. (B) Combination therapy.

Figure 4. Exploratory gene expression profiling.

Tumor biopsies obtained at approximately treatment day 21 revealed differences in the expression of multiple genes in the combination treatment arm versus the monotherapy arm. Enhanced expression of genes involved in antigen processing and presentation (including TAP-associated glycoprotein [TAPBP], proteasome subunit beta type-8 [PSMB8], and human leukocyte antigen [HLA]) as well as an increase in interferon signaling pathway member interferon-induced transmembrane protein 2 (IFITM2) in the combination arm, are consistent with preclinical models of MEK inhibition. Quality control assessment and data normalization were performed using the default settings for positive controls and the housekeeping genes in nSolver Analysis Software (NanoString Technologies) without P value adjustment, n = 6 samples per group.