The Optimal Tumor Mutational Burden Cutoff Value as a Novel Marker for Predicting the Efficacy of Programmed Cell Death-1 Checkpoint Inhibitors in Advanced Gastric Cancer

Programmed cell death-1 (PD-1) checkpoint inhibitors have proven to be effective in the treatment of locally advanced unresectable or metastatic gastric cancer (advanced gastric cancer [AGC]). Nivolumab combined with capecitabine or 5-fluorouracil plus oxaliplatin (XELOX or FOLFOX) is currently indicated as the first-line treatment for programmed death ligand-1 (PD-L1) combined positive score (CPS) ≥5 and human epidermal growth factor receptor 2-negative AGC [1, 2]. In Korea, pembrolizumab was approved as the second-line or subsequent treatment for microsatellite instability-high (MSI-H) or mismatch repair-deficient (dMMR) AGC based on the KEYNOTE-158 trial [3, 4]. Nivolumab monotherapy is also recommended as a third-line therapy based on a randomized phase 3 ATTRACTION-2 trial [2, 5].

However, the predictive markers for the efficacy of PD-1 checkpoint inhibitor monotherapy in AGC have not yet been sufficiently established [6, 7, 8]. There is no clear standard for identifying patients who are likely to benefit from PD-1 checkpoint inhibitors does not exist. Although PD-L1 and MSI are considered novel markers for PD-1 checkpoint inhibitors, their roles have not yet been determined [9, 10, 11].

Recent advances in whole-exome sequencing (WES) and next-generation sequencing (NGS) of multiple genes have defined the tumor biology of AGC [12, 13, 14]. Furthermore, molecular targeted therapies and immunotherapies, including PD-1 checkpoint inhibitors, have been developed based on the NGS findings [14, 15, 16, 17]. In addition, a high level of tumor mutational burden (TMB), the total number of non-synonymous mutations in the tumor-coding regions, is a novel predictive marker for immunotherapy or a prognostic marker in various tumor types [18, 19, 20]. Some somatic mutations in tumor DNA are associated with tumor neoantigens, which are targets of the immune system [21, 22, 23]. Therefore, tumors with a high TMB are likely to have more neoantigens and respond better to immunotherapies that activate T cell immunity. Although TMB is regarded as a biomarker for the response to immunotherapy in several solid tumors [24, 25, 26], the optimal value for defining TMB-high (TMB-H) is insufficient for AGC.

Therefore, in the present study, the optimal TMB cutoff value that could predict the efficacy of PD-1 checkpoint inhibitor monotherapy as a salvage treatment for AGC was determined.

The ROC curve analysis showed that the TMB cutoff value for predicting tumor response in patients with AGC who received PD-1 checkpoint inhibitors was 13.31 mt/Mb. Furthermore, when applying TMB-H, in addition to PD-L1 expression, patients who would benefit more from PD-1 checkpoint inhibitors could be selected. These findings may be helpful for the use of PD-1 checkpoint inhibitors in routine clinical practice.

The reported optimal threshold for defining TMB-H differed with various studies, tumor types, pre-analytic variables, bioinformatics, and test methods [18, 20, 24, 25, 26, 30]. In addition, most studies were mainly conducted in patients with lung cancer and melanoma and relatively few in AGC [30]. In the present study, the TMB of 53 patients was evaluated using the same NGS method to decrease heterogeneity. The NGS diagnostic platform, Illumina’s TSO 500 assay, has a diagnostic value similar to that of WES [31].

However, the clinical significance of the TMB in AGC remains unclear. Wang et al. [32] selected a cutoff of the top 20% TMB (12 mt/Mb) assessed using WES and toripalimab, a humanized PD-1 antibody. Patients with TMB-H showed improved clinical outcome. In a subset analysis of the ATTRACTION-2 phase III trial, TMB-H and TMB-L were classified as having a median value of 8.2 mt/Mb using an NGS panel, and the 2 groups showed no difference in the efficacy of nivolumab as third- or later-line therapy for AGC [11]. Mishima et al. [33] defined TMB-H as >10 mt/Mb based on an NGS panel in AGC and reported that TMB-H was not associated with response to nivolumab. In the present study, the cutoff value for TMB-H was determined to be 13.31 mt/Mb. Patients with TMB-H showed better ORR (71.4% vs. 8.7%, P=0.001), PFS (18.6 vs. 1.9 months, P=0.040), and OS (20.8 vs. 3.3 months, P=0.049) than patients with TMB-L. Compared with previous studies, a relatively larger number of patients were included in the present study, and the cutoff value was identified using the ROC curve analysis by calculating sensitivity and specificity.

The patients with TMB-H and positive PD-L1 expression had a higher ORR (66.7%; 95% CI, 22.3–95.7), which was in agreement with a previous study [32]. TMB status and PD-L1 expression are independent biomarkers for predicting the efficacy of PD-1 checkpoint inhibitors [18, 30, 34]. In gastric cancer, a cutoff value of PD-L1 CPS ≥5 is associated with an increased efficacy of PD-1 checkpoint inhibitors in combination with oxaliplatin-based chemotherapies [1, 2, 35]. However, in PD-1 checkpoint inhibitor monotherapy as a salvage treatment, the effect of PD-L1 is not clear. In the present study, defining the PD-L1 positive as a CPS of ≥1, the PD-L1-positive and TMB-H group exhibited better efficacy than the PD-L1-positive and TMB-L group. This means that among the patients for whom PD-1 checkpoint inhibitors were not indicated because the PD-L1 CPS was <5, additional patients who were expected to be effective could be found using TMB. Furthermore, the limitation of low sensitivity when the cutoff value of TMB is set to 13.31 mt/Mb could be corrected when used with PD-L1.

A higher mutational load is associated with higher neoantigens in tumor immunity, resulting in the recognition of the tumor as foreign and causing an immune response. Thus, hyper-mutated tumors, such as MSI-H or dMMR tumors, show a better response to PD-1 checkpoint inhibitors. MSI-H is associated with high TMB, which accounts for the majority of MSI-H tumors [36]. In the present study, all MSI-H tumors were TMB-H; however, only 57% (4 of 7) of the TMB-H tumors were MSI-H. In addition, one of 3 patients with both TMB-H and microsatellite stable showed a partial response as the best response and 11.8 months of PFS. Therefore, TMB should be assessed independently of other biomarkers such as MSI status when considering PD-1 checkpoint inhibitors.

The present study has several limitations. First, this was a retrospective study conducted at a single center. There were more female than male patients, which is different from previous reports, suggesting a potential for selection bias. Second, the type (i.e., pembrolizumab or nivolumab) and lines of PD-1 checkpoint inhibitors, as well as sequential treatments, were heterogeneous. Third, only patients who received PD-1 checkpoint inhibitors as monotherapy were included. Finally, because the study cohort was small, additional analyses, such as ROC curve analysis using a combination of PD-L1 and TMB, were not conducted.

In conclusion, in the present study, the TMB cutoff value for predicting the tumor response in patients with AGC who received PD-1 checkpoint inhibitor monotherapy as a salvage treatment was 13.31 mt/Mb based on the ROC curve analysis. Furthermore, when PD-L1 status is combined with TMB-H, patients who could benefit from PD-1 checkpoint inhibitors can be selected. These results may be helpful in the use of PD-1 checkpoint inhibitors in routine clinical practice.