Abstract PD14-04: Contribution of tumor and immune cells to PD-L1 as a predictive biomarker in triple-negative breast cancer (TNBC): Analysis from KEYNOTE-119

Abstract

Background: Pembrolizumab monotherapy did not significantly improve overall survival (OS) as second- or third-line treatment for metastatic TNBC vs chemotherapy in the randomized, open-label, phase 3 KEYNOTE-119 study (NCT02555657; N = 622). However, the benefit of pembrolizumab compared with chemotherapy appeared to be greater with increasing PD-L1 expression as quantified by combined positive score (CPS; defined as the number of PD-L1–staining cells [tumor cells, lymphocytes, macrophages] divided by the total number of viable tumor cells, multiplied by 100). In the current exploratory analysis, we aimed to determine whether expression of PD-L1 on tumor cells contributes to the value of PD-L1 as a predictive biomarker in metastatic TNBC.Methods: Patients with centrally confirmed TNBC and 1 or 2 prior systemic treatments for metastatic disease were enrolled in KEYNOTE-119. Patients were randomly assigned 1:1 to pembrolizumab 200 mg Q3W or investigator’s choice of single-agent chemotherapy (capecitabine, eribulin, gemcitabine, or vinorelbine). PD-L1 expression in tumor samples was assessed using PD-L1 IHC 22C3 pharmDx and quantified per tumor proportion score (TPS; defined as the percentage of PD-L1–expressing tumor cells [partial or complete membrane staining] relative to total number of tumor cells) and CPS. Quantitative immune cell density (QID) was defined as CPS minus TPS. QID isolates immune cells but may be truncated when TPS is high. The ability of each scoring method (TPS, CPS, and QID) to predict objective response rate (ORR) with pembrolizumab, including receiver operating characteristics (ROC) analysis, and OS hazard ratios (HRs; pembrolizumab vs chemotherapy) was evaluated.Results: Tumor samples were available for 601 patients (pembrolizumab, 309; chemotherapy, 292) in KEYNOTE-119. ORR was 9.7% (30/309) with pembrolizumab and 11.3% (33/292) with chemotherapy when PD-L1 expression status was not considered. In the pembrolizumab arm, the area under the ROC curve (AUROC; 95% CI) was 0.69 (0.58-0.80) for tumor samples scored for CPS, 0.66 (0.55-0.77) for QID, and 0.55 (0.46-0.64) for TPS. ROC analysis is shown in the Table. At each cutoff, QID had lower estimated sensitivity (ie, missed responders) and a lower Youden Index compared with CPS. The number of missed responders (of 30 total in the pembrolizumab arm) for QID relative to CPS were 2, 5, 5, 6, 2, and 2 at cutoffs of 1, 10, 20, 30, 40, and 50, respectively. Across all practical cutoffs, the OS HR tended to be slightly smaller for CPS than QID. At cutoffs corresponding to the upper percentiles of 10, 20, 40, and 60, OS HRs were 0.497, 0.658, 0.758, and 0.850, respectively, for CPS vs 0.572, 0.712, 0.814, and 0.863, respectively, for QID. QID appeared to be orthogonal to TPS (r = -0.03 for all 601 observations; r = -0.04 after eliminating 7 potentially truncated values).Conclusions: Trends estimated using KN119 suggest that tumor cell expression is an important component of PD-L1 as a predictive biomarker of pembrolizumab efficacy in metastatic TNBC. In this exploratory analysis, when immune cells alone were used to measure PD-L1 expression, a meaningful number of responders was missed and OS benefit trended toward higher HR estimates. Tumor and immune cell PD-L1 expression may represent distinct (presumably negative modulatory) mechanisms. Table. ROC AnalysisCutoffCPS SensCPS SpecCPS YICPS PrevQID SensQID SpecQID YIQID PrevTPS SensTPS SpecTPS YITPS Prev010011001100110.8330.3660.1990.6540.7670.4160.1820.6020.3000.7890.0890.220100.5670.7170.2840.3110.4000.8140.2140.2070.2000.8920.0920.117200.5000.8490.3490.1840.3330.9250.2580.1000.1670.9320.0990.078300.3670.8920.2590.1330.1670.9570.1240.0550.1000.9430.0430.061400.2000.9350.1350.0780.1330.9860.1190.0260.0670.9530.0200.049500.2000.9570.1570.0580.1330.9930.1260.0190.0670.9680.0340.036 Citation Format: Eric P. Winer, Oleg Lipatov, Seock-Ah Im, Anthony Goncalves, Eva Muñoz-Couselo, Keun Seok Lee, Zbigniew Nowecki, Peter Schmid, Kenji Tamura, Laura Testa, Isabell Witzel, Shoichiro Ohtani, Stephanie Hund, Karina Kulangara, Vassiliki Karantza, Jaime A. Mejia, Junshui Ma, Petar Jelinic, Lingkang Huang, Kenneth Emancipator, Javier Cortes. Contribution of tumor and immune cells to PD-L1 as a predictive biomarker in triple-negative breast cancer (TNBC): Analysis from KEYNOTE-119 [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PD14-04.

Immune-related biomarkers in triple-negative breast cancer

“…In addition, Winer EP et al explained the advantages of CPS which was tested by 22C3 over other testing methods at SABCS 2020. PD-L1 expression in tumor and immune cells both are important in metastatic TNBC which could be as a predictive biomarker of pembrolizumab efficacy in KEYNOTE-119 cases [ 47 ]. So based on the researches published, we prefer the two approaches, SP142 by Ventana and 22C3 by Dako Agilent, which have been approved by FDA [ 32 ].…”

Prognostic Significance of CSF-1R Expression in Early Invasive Breast Cancer

“…Immune checkpoint inhibitor therapy in combination with chemotherapy has emerged as a promising therapeutic modality for early and metastatic triple-negative breast cancer [ 15 , 16 , 17 ]. Clinical benefit has been demonstrated with regards to higher rates of pathological complete responses [ 15 , 18 , 19 ] and a modest increase in objective response rates [ 16 ], though improvement in progression-free survival and overall survival has been predominantly achieved in PD-L1-enriched, treatment-naïve metastatic tumors [ 16 , 17 , 20 ]. Durable responses with immune checkpoint blockade have been reported in diverse tumor types [ 21 ]; however, such evidence is limited in breast cancer [ 16 , 22 ].…”