Investigation of PD-L1 expression and response to pembrolizumab (pembro) in gastric cancer (GC) and cervical cancer (CC) using combined positive score (CPS) and tumor proportion score (TPS).

Karina Kulangara, Lindsay Guerrero, Alex Posch, Scott Boyer, Debra Hanks, Josette Carnahan, Jiangdian Wang, Jared Lunceford, Mary J. Savage, Matthew J. Marton,

Journal of Clinical Oncology · 2018 · 21 citations · 0 references

Concepts

Abstract

4065 Background: TPS, the percentage of viable tumor cells with partial or complete membrane staining at any intensity, has been invaluable for assessing PD-L1 expression in non–small cell lung cancer (NSCLC) and identifying patients (pts) likely to respond to anti-PD–1/PD-L1 therapy. However, TPS has limited utility beyond NSCLC. We investigated the predictive value of TPS and CPS and their association with response to pembro in pts with GC and CC. Methods: Tumor samples from pts with previously treated GC (KEYNOTE-059, NCT02335411) or CC (KEYNOTE-158, NCT02628067) were analyzed for PD-L1 expression per an investigational-use-only-labeled version of the PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies). Response was assessed per RECIST v1.1 by independent review. External reproducibility of CPS, the number of PD-L1–staining cells (tumor cells, lymphocytes, macrophages) divided by the total number of viable tumor cells, multiplied by 100, was assessed. Results: Response, prevalence, positive predictive value (PPV), and negative predictive value (NPV) for CPS and TPS at cutoffs of 1 (TPS1, CPS1) and 10 (TPS10, CPS10) are shown in the table. Response to pembro was significantly associated with CPS (P= 0.002) but not TPS (P= 0.224) in GC, whereas response to pembro was significantly associated with CPS (P= 0.008) and TPS (P= 0.023) in CC. Intersite and interobserver overall agreement assessments of external reproducibility for CPS were 92.0% (95% CI, 87.4-96.3) and 96.6% (95% CI, 94.0-98.7), respectively, in GC and 95.0% (95% CI, 90.3-99.2) and 99.7% (95% CI, 99.7-100.0), respectively, in CC. Conclusions: CPS is a robust, reproducible scoring method that identified more responders than did TPS in GC and CC. Further investigation of CPS in cancers beyond NSCLC is warranted. Clinical trial information: NCT02335411 and NCT02628067.Scoring and Cutoff Responders/Total n/N Prevalence of PD-L1 Expression, % PPV, % NPV, % Odds ratio for Tumor Response All Patients, N PD-L1 Positive PD-L1 Negative Gastric cancer CPS1 257 24/148 7/109 58 16 94 2.8 TPS1 5/32 26/225 12 16 88 1.4 Cervical cancer CPS1 96 13/81 0/15 84 16 100 ∞ TPS1 10/50 3/46 52 20 93 3.6 CPS10 10/45 3/51 47 22 94 4.6 TPS10 6/26 7/70 27 23 90 2.7