Tumor Cell PD-L1 Score Predicts Longer Survival in Gastric Cancer Treated with Chemotherapy Alone

A large Japanese analysis of patients with advanced gastric cancer has found that a specific way of measuring PD-L1, a protein long associated with immunotherapy response, may also carry independent prognostic weight when patients are treated with chemotherapy alone. In an ancillary study of the randomized phase III trial JCOG1013, published in the British Journal of Cancer, researchers report that a tumor proportion score of at least 1% was linked to significantly longer overall survival and progression-free survival among patients receiving first-line chemotherapy without an immune checkpoint inhibitor. By contrast, the combined positive score, the measure most often used today to guide immunotherapy decisions, showed no such association.

Gastric cancer remains a major cause of cancer-related death, particularly in East Asia, even as its prevalence has been declining. For patients whose tumors cannot be surgically removed or have recurred, and whose tumors are HER2-negative, the standard first-line treatment backbone has long been a doublet of a platinum drug and a fluoropyrimidine. In the original JCOG1013 trial, conducted between April 2012 and March 2016 across 56 Japanese hospitals, 741 chemotherapy-naive patients aged 20 to 75 years were randomly assigned to receive either the triplet docetaxel, cisplatin and S-1 or the doublet cisplatin and S-1. That trial demonstrated no overall survival or progression-free survival benefit from adding docetaxel, leaving the doublet as the reference regimen.

The treatment landscape has since shifted with the arrival of immune checkpoint inhibitors targeting the PD-1/PD-L1 axis, which in combination with doublet chemotherapy have shown survival benefits in this population. PD-L1 expression, assessed either as the combined positive score or the tumor proportion score, serves as a predictive marker for the efficacy of PD-1 inhibitors. A meta-analysis of first-line phase III trials suggested that a combined positive score of 5 or higher represents an acceptable cut-off that optimizes the risk-benefit balance of immunotherapy, and chemotherapy alone remains an important option for patients below that threshold, particularly those with scores below 1.

Yet the prognostic significance of PD-L1 expression, independent of immunotherapy, has remained unclear in gastric cancer. Prior studies have produced inconsistent results, complicated by heterogeneous study populations that sometimes included earlier-stage disease, differences in antibody clones, scoring systems and cut-off values, and inter-observer variability among pathologists. Studies in lung cancer have found associations between PD-L1 expression and favorable prognosis, but few investigations have explicitly examined the relationship between PD-L1 status and outcomes in patients with unresectable or recurrent gastric cancer receiving first-line chemotherapy without an immune checkpoint inhibitor. The new analysis was designed to fill that gap using data from a randomized phase III trial.

The researchers analyzed pretreatment tumor specimens using the Dako 28-8 pharmDx assay on formalin-fixed endoscopic biopsy samples of the primary tumor. Two experienced gastrointestinal pathologists independently evaluated both scores while blinded to clinical data, resolving discrepancies by consensus. The combined positive score counts all PD-L1-positive cells, including tumor cells, lymphocytes and macrophages, relative to the total number of viable tumor cells, multiplied by 100. The tumor proportion score counts only PD-L1-positive tumor cells as a percentage of viable tumor cells. The cut-off of 5 for the combined score was prespecified based on the CheckMate-649 trial, while the 1% threshold for the tumor proportion score followed the ATTRACTION-4 trial. Because the analysis was exploratory, no alternative cut-offs were evaluated.

Of the 741 patients enrolled in JCOG1013, tumor sections were available for 550, and PD-L1 expression was assessable in 540, who formed the analysis population. Among these, 350 patients (65%) had a combined positive score below 5 and 190 (35%) scored 5 or higher, while 416 patients (77%) had a tumor proportion score below 1% and 124 (23%) scored 1% or above. The prevalence of combined positive score positivity of 35% was relatively low compared with the roughly 60% reported in CheckMate-649, though prevalence has varied widely from 20% to 52% across other studies. The tumor proportion score positivity rate of 23% fell squarely within the previously reported range of 16% to 27%. The greater variability for the combined score may reflect inter-observer variability, which previous studies have demonstrated.

With a median follow-up of 42.8 months, the analysis found no significant difference in median overall survival between patients with a combined positive score of 5 or higher and those below that threshold, 14.6 versus 15.3 months, with a hazard ratio of 0.99. Progression-free survival likewise showed no significant difference, 6.4 versus 7.1 months. For the tumor proportion score, unadjusted comparisons also showed no significant differences, with median overall survival of 15.5 versus 14.6 months and median progression-free survival of 7.2 versus 7.0 months, although the overall survival curves separated after three years. Treatment outcomes were similar between the docetaxel triplet and the cisplatin doublet irrespective of either PD-L1 measure, confirming that the addition of docetaxel conferred no survival advantage in any expression subgroup.

The picture changed in multivariable analyses adjusted for established prognostic factors. A combined positive score of 5 or higher was not significantly associated with overall survival, with a hazard ratio of 0.86 and a p-value of 0.127, or with progression-free survival, with a hazard ratio of 0.92 and a p-value of 0.392. In contrast, a tumor proportion score of 1% or higher emerged as a significant independent favorable prognostic factor for both endpoints, with a hazard ratio of 0.78 for overall survival and 0.76 for progression-free survival. Sensitivity analyses using alternative covariate sets and inverse probability of treatment weighting yielded consistent results, with hazard ratios ranging from 0.76 to 0.80 for overall survival and 0.76 to 0.77 for progression-free survival across all models. In the adjusted models, an Eastern Cooperative Oncology Group performance status of 1, two or more metastatic sites, and diffuse-type histology were each significantly associated with shorter survival, as expected from established clinical factors.

The authors propose a biological explanation for the divergence between the two scoring systems. The tumor proportion score reflects PD-L1 expression on tumor cells themselves, whereas the combined positive score also captures PD-L1 on immune cells within the tumor microenvironment. Tumor-cell PD-L1 can suppress antitumor T-cell responses by increasing apoptosis of antigen-specific T cells, and cytotoxic chemotherapy may diminish this inhibitory effect, conferring a more favorable antitumor immune milieu. The plateau observed at the end of the survival curves in the tumor proportion score-positive group suggests that chemotherapy might induce a durable immune response in a subset of patients. By contrast, the immunosuppressive microenvironment reflected in the combined score, which includes immune-cell PD-L1 that negatively modulates T-cell activation, may not be reversed by cytotoxic agents. Consistent with this interpretation, five-year overall survival rates in the chemotherapy-alone arm of CheckMate-649 did not differ meaningfully by combined positive score status, at 5%, 6% and 7% for scores of 1 or higher, 5 or higher and 10 or higher, respectively.

The findings come with important caveats. JCOG1013 was not designed to compare outcomes according to PD-L1 expression, which may limit statistical power, and the analysis was restricted to patients with available and assessable specimens, so selection bias related to tissue availability cannot be fully excluded, even though baseline characteristics were generally comparable between included and excluded patients. PD-L1 was assessed only on primary tumor biopsies, and spatial heterogeneity between primary and metastatic lesions may have influenced the results. Information on subsequent immune checkpoint inhibitor use was unavailable, though such exposure was likely limited because the trial predated widespread adoption of these drugs in gastric cancer. The authors emphasize that the tumor proportion score is not currently used to guide treatment selection and that these hypothesis-generating results require confirmation in further studies before any change in clinical practice. For now, the study offers the first randomized-trial-based evidence that, in the absence of immunotherapy, tumor-cell PD-L1 expression may mark a genuinely more favorable prognosis.

Subject of Research: Prognostic value of PD-L1 expression in unresectable or recurrent gastric cancer treated with first-line chemotherapy without immune checkpoint inhibitors

Article Title: PD-L1 expression and survival in unresectable/recurrent gastric cancer treated by 1st-line treatment without immune checkpoint inhibitors: JCOG1013A1

Article References: Hirano, H., Yamada, Y., Nagashima, K., Hiraoka, N., Sekine, S., Takahashi, N., Azuma, M., Iwasa, S., Kanato, K., Machida, N., Kinoshita, T., Hata, H., Kawakami, H., Takahari, D., Ojima, T., Kadowaki, S., Boku, N., Kurokawa, Y., Terashima, M., & Yoshikawa, T. (2026). PD-L1 expression and survival in unresectable/recurrent gastric cancer treated by 1st-line treatment without immune checkpoint inhibitors: JCOG1013A1. British Journal of Cancer. https://doi.org/10.1038/s41416-026-03607-7

Image Credits: AI Generated

DOI: 10.1038/s41416-026-03607-7

Keywords: PD-L1, gastric cancer, JCOG1013, combined positive score, tumor proportion score, chemotherapy, prognosis, immune checkpoint inhibitors, biomarker, oncology, expression, survival