Programmed Cell Death Ligand-1 Blockade in Urothelial Bladder Cancer: To Select or Not to Select.

Abstract

The recent accelerated approval of the anti–programmed cell death ligand-1 (PD-L1) monoclonal antibody atezolizumab for metastatic urothelial bladder cancer (UBC) progressing during or after platinum-based chemotherapy signified a milestone in this disease, for which no new treatments had been approved since 1990. Approval was based on a single-armed, multinational study (IMvigor 210) in which 315 patients with UBC who had archived or fresh tissue available for PD-L1 staining were treated; the data showed an overall response rate (ORR) of 15% with significant durability, and it is noteworthy that the current approval does not require biomarker analysis before treatment. In the article that accompanies this editorial, Massard et al present data from a phase I/II trial of a second anti–PD-L1 monoclonal antibody (durvalumab); these data show that durvalumab also has significant activity in UBC, with an ORR of 31% in a population that was enriched for PD-L1–positive patients. Although the durvalumab trial included fewer patients (61) versus those in the phase II atezolizumab study (315), both studies included analyses of response on the basis of PD-L1 expression on either tumor cells (TCs) or immune cells (ICs), providing a unique opportunity to examine data regarding the potential predictive value of this biomarker. It is important to recognize at the onset that the two studies used different antibodies for staining; IMvigor used the rabbit monoclonal antibody SP142 (Spring Biosciences, Pleasanton, CA), whereas the durvalumab study used the SP263 antibody (Ventana Medical Systems, Tucson, AZ). This is noteworthy because recent data show that different anti–PD-L1 antibodies may give different results, with a discordance of approximately 25% between the SP142 and the E1L3N antibodies. To date, cross-comparison of the SP142 and SP263 antibodies has not been performed. Studies of atezolizumab have generally found a better correlation between response and staining on ICs, and the IMvigor analysis focused on IC staining. Within the IMvigor study, the ORR was 26% in patients positive for PD-L1 in immune cells (IC2/3), versus approximately 10% in the IC0/1 patients. An ORR of 10% is in the range observed for chemotherapy regimens in this patient population, so, on the basis of its good tolerability profile, atezolizumab is likely to be administered to patients with both PD-L1–positive and PD-L1–negative bladder cancer in its on-label second-line setting. Similarly, durvalumab responses in UBC correlated with PD-L1 expression on either tumor cells or immune cells. Considering TC only, the ORR was 47% for positive versus 22% for negative staining. IC staining also correlated with ORR, with responses of 57% versus 13%, respectively. The authors of the durvalumab study also explored a novel approach to the PD-L1 biomarker—integrating TC and IC expression to define composite positive and negative categories. In this new definition, PD-L1 positive was designated as $ 25% staining in either ICs or TCs, whereas PD-L1 negative specimens were negative (# 25%) in both the IC and TC compartments. Using this new composite definition led to a potential improvement in negative predictive value; patients whose samples were negative for PD-L1 in both compartments had an ORR of 0% (0 of 14), and patients positive for PD-L1 in either compartment had an ORR of 46%. To contextualize these data, it is important to note that the durvalumab trial was enriched for PD-L1–positive patients; the first 20 patients were enrolled regardless of PD-L1 status, and the next 41 patients were required to have$ 5% staining on either TCs or ICs. The small number of patients involved adds further caution to interpreting these results. Indeed, the first data correlating PD-L1 status to response, obtained in a phase Ib study of the anti– programmed death-1 (PD-1) antibody nivolumab, also showed a dramatic concordance between the absence of PD-L1 expression on TCs and response, with 0 of 17 PD-L1–negative patients experiencing an objective response. Subsequent larger trials showed this apparently striking negative predictive value to be premature; for example, in the pivotal first-line trial of nivolumab in melanoma, the ORR in PD-L1–negative or indeterminate patients was 33%, versus 53% in PD-L1 positive. Similar results were obtained in second-line nonsquamous non–small-cell lung cancer, with ORRs 14% in PD-L1–negative patients versus 34% in PD-L1– positive patients (using a cutoff of . 5% TC staining). Taken together, these data suggest that the low response rate seen in the PD-L1–negative patients in the accompanying durvalumab bladder study is broadly consistent with prior results but is unlikely to be quite so clear cut in larger trials. On the other hand, given that antibodies to PD-1 or PD-L1 are postulated to function by blocking the inhibitory signal transmitted from TCs or ICs to the CD8 T cells that infiltrate a tumor, it might seem somewhat puzzling that PD-L1 expression is not more