Real-world characterization of PD-L1 expression in patients with advanced esophagogastric adenocarcinoma in Spain

Abstract

Purpose

Epidemiological data from Sapin are lacking, especially in a publication relating Spanish data and prevalence of programmed cell death ligand-1 (PD-L1) expression. This study aimed to evaluate the prevalence of PD-L1 expression by Combined Positive Score (CPS) ≥ 5 in patients with advanced esophagogastric adenocarcinoma (aEGAC) in Spain.

Methods

This observational, retrospective, and multicenter study collected sociodemographic and clinical data from adult patients with locally advanced unresectable, recurrent, or metastatic EGAC across 21 centers in the AGAMENON-SEOM registry. CPS PD-L1 expression was centrally analyzed using the IHC 28–8 pharmDx.

Results

A total of 166 patients were included, with 144 valid and evaluable samples. The primary tumor locations were stomach (70.1%), gastroesophageal junction (19.4%), and esophagus (10.4%). At diagnosis, 84.0% had synchronous metastatic disease, 4.9% had unresectable locally advanced disease, and 11.1% had recurrent disease. PD-L1 CPS ≥ 5 was observed in 41.7% of the patients (95%CI, 34.3–50.4%). Prevalence by tumor type was 41.6% (95% CI, 32.0–51.2%) in gastric adenocarcinoma, 35.7% (95% CI, 18.0–53.5%) in gastroesophageal junction adenocarcinomas, and 60.0% (95% CI, 35.2–84.8%) in esophageal adenocarcinoma. PD-L1 CPS ≥ 5 was found in 42.9% (95% CI, 6.2–79.5%) of patients with unresectable locally advanced disease, 43.0% (95% CI, 34.2–51.8%) of those with metastatic disease, and 37.5% (95% CI, 13.8–61.2%) of those with recurrent disease.

Conclusion

This is the first study to report the real-world prevalence of PD-L1 expression in patients with aEGAC in clinical practice in Spain. These findings underscore the importance of understanding PD-L1 prevalence in the Spanish population and investigating discrepancies with previous studies.

Introduction

Gastric and esophageal cancers are the fourth and sixth leading causes of cancer-related deaths globally, accounting for a combined total of 1.3 million deaths annually [1]. According to GLOBOCAN 2020, the age-standardized mortality rates for these cancers are estimated at 7.7 and 5.5 per 100,000 individuals, respectively. Gastric cancers can be categorized into two types based on their location: gastric adenocarcinomas (GACs) and gastroesophageal junction adenocarcinomas (GEJACs) [2]. Esophageal cancer is divided into two major subtypes: esophageal squamous cell carcinoma and esophageal adenocarcinoma (EAC) [3]. GEJACs are also considered in part esophageal cancers, at least for tumor, node, and metastasis (TNM) definition [4]. From a molecular perspective, esophageal adenocarcinomas (EACs) share similarities with GEJACs or GACs. These differ significantly from other esophageal cancer subtypes, such as esophageal squamous cell carcinoma [5]. Recently, immunotherapy has opened a new era for treating esophagogastric adenocarcinomas (EGACs) [3]. The programmed death receptor 1/programmed death ligand 1 (PD-1/PD-L1) pathway has been shown to have prognostic significance in EGACs. In fact, several studies have demonstrated a relationship between PD-L1 expression and overall survival (OS) [6, 7]. PD-L1 expression is assessed in those tumors using the tumor proportion score (TPS) and combined positive score (CPS) [6]. More recently, tumor area positivity (TAP) score has emerged as a novel scoring method to evaluate tumor and immune cells simultaneously using visual estimation [8]. The PD-L1 CPS scoring is a strong predictor of benefit for patients with EGACs treated with immune checkpoint inhibition (ICI)-based therapies [9, 10].

In the randomized, placebo-controlled, double-blind, phase III study KEYNOTE-590, patients with adenocarcinoma or esophageal squamous cell carcinoma and PD-L1 CPS ≥ 10 who received pembrolizumab plus chemotherapy showed a superior median OS (13.9 months) than those receiving placebo plus chemotherapy (8.8 months) [11]. In the study, 50% and 52% of the patients on pembrolizumab plus chemotherapy (n = 373) or placebo plus chemotherapy (n = 376), respectively, showed PD-L1 CPS ≥ 10. In the multicenter, randomized, open-label, phase III trial CheckMate 649, nivolumab plus chemotherapy resulted in significant improvements in OS [hazard ratio (HR) 0.71 (98.4% CI 0.59–0.86); p < 0.0001] and progression-free survival [PFS; HR 0.68 (98% CI 0.56–0.81); p < 0.0001] versus chemotherapy alone in patients with previously untreated, unresectable, non-human epidermal growth factor receptor 2 (HER2)-positive GAC, GEJAC, or EACs patients with PD-L1 CPS ≥ 5. Of patients, 81% and 83% of those on nivolumab plus chemotherapy (n = 789) or chemotherapy alone (n = 792), respectively, showed PD-L1 CPS ≥ 1 (60% and 61% with PD-L1 CPS ≥ 5). Nivolumab is the first PD-1 inhibitor to show superior OS, along with PFS benefit and an acceptable safety profile, in combination with chemotherapy versus chemotherapy alone, showing a 3.3-month improvement in median OS [14.4 months (95% CI 13.1–16.2) vs 11.1 months (10.0–12.1)] in patients with a PD-L1 CPS ≥ 5 [12]. This study has the longest minimum follow-up to date, 5 years, and the proportion of patients alive at 60 months was numerically higher with nivolumab plus chemotherapy [16% (95% CI 12–19)] than with chemotherapy alone [6% (95% CI 4–9)] [13]. Similarly, KEYNOTE-859 trial demonstrated that patients with human epidermal growth factor receptor 2 (HER2)-negative and advanced gastric cancer who received pembrolizumab plus chemotherapy achieved a longer median OS, especially in those with PD-L1 CPS ≥ 1 (13.0 months) and PD-L1 CPS ≥ 10 (15.7 months), compared to placebo plus chemotherapy (11.4 and 11.8 months, respectively) [14]. In the study, 78% of the patients on pembrolizumab plus chemotherapy (n = 790) or placebo plus chemotherapy (n = 789) showed PD-L1 CPS ≥ 1, and 35% and 34%, respectively, had PD-L1 CPS ≥ 10. In the RATIONALE-305 randomized, double-blind, phase III trial, tislelizumab combined with chemotherapy as a first-line treatment for advanced or metastatic GAC or GEJAC improved the median OS (17.2 months) compared to placebo plus chemotherapy (12.6 months) in patients with a PD-L1 TAP ≥ 5% [15]. Among patients, 55% of those on tislelizumab plus chemotherapy (n = 501) or placebo plus chemotherapy (n = 496) showed PD-L1 TAP ≥ 5%. Recent studies have also demonstrated a benefit as PD-L1 CPS increases, with typical cut-offs defined as CPS ≥ 5 or CPS ≥ 10 [16]. Given that PD-L1 expression may be used as a predictive tool for identifying patients who are likely to benefit from ICIs, knowledge of PD-L1 expression has become crucial in managing EGACs. Although real-world studies have examined PD-L1 expression in advanced EGAC (aEGAC) [17–22], none have been conducted in Spain. Therefore, this study aims to fill that gap by evaluating the prevalence of PD-L1 CPS ≥ 5 in clinical practice samples from Spanish patients with aEGAC.

Methods

Study design

The AGAMENON-SEOM registry is a Spanish registry of esophageal and gastric cancer promoted by the Spanish Society of Medical Oncology (SEOM), involving 45 Medical Oncology Departments. CARPATOS is an observational, retrospective, multicenter subproject that included patients with aEGAC from 21 hospitals throughout Spain. The inclusion criteria were: adult patients with a histologically confirmed diagnosis of unresectable locally advanced, recurrent, or metastatic EGAC; confirmed HER2-negative status; biopsies obtained within 2.5 years before the study inclusion; treated in first line with chemotherapy; and providing written informed consent study (informed consent was waived for deceased patients). Samples had to meet quality criteria, including being obtained from the primary tumor, containing a minimum of 4 neoplastic tissue fragments, and having at least 100 viable cells. Exclusion criteria included patients with no diagnosis of aEGAC; confirmed HER2-positive status; biopsy obtained more than 2.5 years before the study inclusion; not treated in first line with chemotherapy; not providing written informed consent; or with samples not meeting the predefined quality criteria. Patients were consecutively recruited at each participating center to minimize potential confounding biases and assigned an anonymized coding number in the AGAMENON-SEOM registry. This coding number was also used to anonymize all information related to patients’ biopsy samples, ensuring data confidentiality and consistency across the study. The study was approved by the Ethical Committees of the participating centers and was conducted following the principles of the Declaration of Helsinki and local Spanish regulations.

Samples and PD-L1 expression

PD-L1 expression was analyzed using CPS scoring in diagnostic samples collected following routine clinical practice for advanced EGACs within 2.5 years prior to study inclusion. Archival samples required sufficient tumor tissue fixed in neutral, phosphate-buffered, 10% formalin, and embedded in paraffin for immunohistochemical studies. Slides stained with hematoxylin–eosin were reviewed by an expert pathologist (MI-C) to ensure they met quality criteria. Immunohistochemical staining for PD-L1 expression were performed on 4 µm-thick paraffin sections using the Agilent PD-L1 immunohistochemistry 28–8 pharmDx assay (Dako, Santa Clara, CA, USA). The PD-L1 CPS assessment was centralized at the reference Laboratory in Hospital del Mar, Barcelona, Spain by an experienced pathologist (MI-C). The PD-L1 CPS was determined as the number of PD-L1 stained cells (tumor and mononuclear immune cells) divided by the total number of viable tumor cells multiplied by 100 [23].

Patient’s data collection and variables

CARPATOS study gathers detailed sociodemographic and clinical data from patients included in the AGAMENON-SEOM registry. Baseline characteristics were recorded immediately prior to initiation of first-line systemic therapy for advanced disease. Variables included sex (male or female), age, body mass index (BMI, calculated as weight [kg] divided by height squared [m²], expressed as kg/m²), and Eastern Cooperative Oncology Group Performance Status (0–3, assessed by the treating oncologist at baseline). Primary tumor location at diagnosis (GAC, GEJAC, or EAC), disease stage (metastatic, recurrent, or unresectable locally advanced), and metastatic burden, including metastatic sites and the number metastases (defined as the total count of individual metastatic lesions visible on baseline chemotherapy, irrespective of their anatomical site). Additional baseline tumor and biomarker variables were captured as recorded in local pathology reports, including Lauren classification (intestinal, diffuse, or not specified), signet ring cell carcinoma (presence, absence, or not studied), and microsatellite instability-high (MSI-H)/deficient mismatch repair (dMMR) status (presence, absence, or not studied). Previous treatments at diagnosis of the advanced disease (none, adjuvant, or neoadjuvant), first-line chemotherapy backbone (monotherapy, doublet, or triplet, based on the number of cytotoxic agents), and use of immunotherapy in first-line (yes or no; any PD-1/PD-L1 inhibitor) were also collected. Data are collected using standardized electronic case report forms (eCRFs: www.agamenonstudy.com) and are routinely monitored to ensure accuracy and completeness. Variables are extracted directly from the medical records of registered patients. The registry is managed by the AGAMENON-SEOM Steering Committee (ACB, JGP, PJF), which oversees data quality, analysis, and publication. Its primary objectives include enhancing esophageal and stomach cancer knowledge to optimize patient management, treatment, and outcomes; characterizing these cancers at the molecular and pathological levels; enabling comparisons with other datasets; and supporting high-quality scientific publications [24].

Sample size determination and data analysis

The sample size was calculated based on an expected prevalence of 60% for PD-L1 CPS ≥ 5 in samples from patients with unresectable advanced, recurrent, or metastatic EAC/GEJAC/GAC, as reported in the CheckMate 649 trial [12]. The required sample size was calculated and estimated in 200 patients using a formula for proportions with an alpha value of 5% and a precision of 7%. However, the final sample size was adjusted to 223 patients to account for an expected 10% drop-out rate (commonly used in studies). The primary objective was to estimate the prevalence of PD-L1 CPS ≥ 5 in evaluable samples. Quantitative variables were described using the median and interquartile range (IQR; 25th–75th percentiles); whereas qualitative ones as absolute and relative frequencies (n, %). The 95%CI was calculated using normal Wald statistics. The prevalence of PD-L1 CPS ≥ 5 was estimated overall and stratified by stage (unresectable locally advanced, metastatic, or recurrent) and by primary tumor location (GAC, GEJAC, or EAC). Analyses were performed on available cases, without imputation of missing values, and the corresponding denominators were reported for each estimate. All analyses were descriptive and were performed using SAS statistical software (v9.4, SAS Institute Inc., Cary, NC).

Results

Study population

Between August 2023 and January 2024, out of the estimated 223 patients, 166 patients were identified for inclusion in the study; however, 22 were excluded for the following reasons: lack of diagnosis of unresectable locally advanced, recurrent, or metastatic EGAC (n = 1); HER2-positive status (n = 8); biopsy obtained more than 2.5 years before the study inclusion (n = 2); and samples not meeting quality criteria (n = 11). Thus, 144 patients’ samples were evaluable for the present analysis (Fig. 1). The study population was predominantly males (72.2%), with a median age of 66.0 years (IQR, 58.5–72.0) and an Eastern Cooperative Oncology Group Performance Status 1 in 64.6% of patients (Table 1). Most patients (70.1%) had GAC, followed by GEJAC (19.4%) and EAC (10.4%). The disease stage was predominantly metastatic (84.0%), with a median of 2 organs affected by metastases (IQR, 1.0–3.0). Metastases were mainly located in distant lymph nodes (43.1%), liver (41.6%), and peritoneum (41.6%). According to Lauren’s classification, 55.6% of the patients had intestinal subtype cancer. The majority of samples (81.9%) were MSS (microsatellite stable) or pMMR (proficient mismatch repair). At diagnosis of the advanced disease, most patients (89.6%) had not received any prior therapy. For first-line treatment, 88.2% of patients received doublet chemotherapy (fluoropyrimidine and platinum). At the time of data collection, the majority of patients did not receive immunotherapy-based treatment, as it was not yet approved in Spain.

Fig. 1.

Patient flow chart. EGAC, esophagogastric adenocarcinomas; HER2, human epidermal growth factor receptor 2

Table 1.

Sociodemographic and clinical characteristics of patients

Baseline characteristics	Patients (N = 144)
Sex male, n (%)
Male	104 (72.2)
Female	40 (27.8)
Age (years), median (IQR)	66.0 (58.5–72.0)
BMI, median Kg/m2 (IQR)	25.1 (22.1–27.9)
ECOG PS, n (%)
0	26 (18.1)
1	93 (64.6)
2	20 (13.9)
3	5 (3.5)
Location of the primary tumor at diagnosis, n (%)
GAC	101 (70.1)
GEJAC	28 (19.4)
EAC	15 (10.4)
Disease stage, n (%)
Metastatic	121 (84.0)
Recurrent	16 (11.1)
Unresectable locally advanced	7 (4.9)
Number of organs with metastasis, median (IQR)	2.0 (1.0–3.0)
Location of metastasis, n (%)*
Distant lymph nodes	59 (43.1)
Liver	57 (41.6)
Peritoneum	57 (41.6)
Ascites	29 (21.2)
Lung	16 (11.7)
Bone	15 (11.0)
Other location†	11 (8.0)
Lauren classification, n (%)
Intestinal	80 (55.6)
Diffuse	59 (41.0)
Not specified	5 (3.5)
Signet ring cell carcinoma, n (%)
No	86 (59.7)
Yes	51 (35.4)
Not studied	7 (4.9)
MSI-H/dMMR, n (%)
No	118 (81.9)
Yes	9 (6.3)
Not studied	17 (11.8)
Previous treatments (at diagnosis of the advanced disease), n (%)
No	129 (89.6)
Yes, adjuvant	13 (9.0)
Yes, neoadjuvant treatment	2 (1.4)
Chemotherapy regimens for first-line advanced disease, n (%)
Monotherapy	4 (2.8)
Doublet	127 (88.2)
Triplet	13 (9.0)
Immunotherapy for first-line advanced disease, n (%)
Yes	14 (9.7)
No	130 (90.3)

IQR interquartile range, BMI body mass index, ECOG PS Eastern Cooperative Oncology Group Performance Status, GAC gastric adenocarcinoma, GEJAC gastroesophageal-junction adenocarcinoma, EAC esophageal adenocarcinoma, MSI-H microsatellite instability-high, dMMR deficient mismatch repair

^*A patient could be included in more than one category. Calculated over 137 patients with metastasis

^†Other location included: adrenal gland (n = 3), ovaries (n = 2), muscle (n = 2), local recurrence (n = 1), pancreas (n = 1), adrenal gland and brain (n = 1), and testicle (n = 1)

Expression of PD-L1 CPS

Overall, 41.7% of patients (95%CI, 34.3–50.4%) had a PD-L1 CPS ≥ 5 (Table 2). By disease stage, PD-L1 CPS ≥ 5 was observed in 42.9% of patients with unresectable locally advanced disease (95%CI, 6.2–79.5%), 43.0% with metastatic disease (95%CI, 34.2–51.8%), and 37.5% with recurrent disease (95%CI, 13.8–61.2%). By tumor location, PD-L1 CPS ≥ 5 was found in 41.6% of patients with GAC (95%CI, 32.0–51.2%), 35.7% with GEJAC (95%CI, 18.0–53.5%), and 60.0% with EAC (95%CI, 35.2–84.8%). The median PD-L1 CPS was 2.5 (IQR, 0.0–7.0). The real-world distribution of PD-L1 CPS in this cohort of patients with HER2-negative, locally advanced unresectable, recurrent, or metastatic aEGAC in the study samples is shown in Fig. 2. Specifically, 26.4% of patients had a PD-L1 CPS < 1, 31.9% had a CPS of 1–4, 24.3% had a CPS of 5–9, and 17.4% had a CPS ≥ 10.

Table 2.

PD-L1 expression in overall patients, and considering tumor stage and location

PD-L1 expression	n (%)	95%CI
Overall patients
PD-L1 CPS ≥ 5	61 (41.7)	34.3–50.4
PD-L1 CPS < 5	83 (58.3)	49.6–65.7
By tumor stage
Unresectable locally advanced
PD-L1 CPS ≥ 5	3 (42.9)	6.2–79.5
PD-L1 CPS < 5	4 (57.1)	20.5–93.8
Metastatic
PD-L1 CPS ≥ 5	52 (43.0)	34.2–51.8
PD-L1 CPS < 5	69 (57.0)	48.2–65.9
Recurrent
PD-L1 CPS ≥ 5	6 (37.5)	13.8–61.2
PD-L1 CPS < 5	10 (62.5)	38.8–86.2
By tumor location
GAC
PD-L1 CPS ≥ 5	42 (41.6)	32.0–51.2
PD-L1 CPS < 5	59 (58.4)	48.8–68.0
GEJAC
PD-L1 CPS ≥ 5	10 (35.7)	18.0–53.5
PD-L1 CPS < 5	18 (64.3)	46.5–82.0
EAC
PD-L1 CPS ≥ 5	9 (60.0)	35.2–84.8
PD-L1 CPS < 5	6 (40.0)	15.2–64.8

95% CI 95% confidence interval, GAC gastric adenocarcinoma, GEJAC gastroesophageal-junction adenocarcinoma, EAC esophageal adenocarcinoma

Fig. 2.

Real-world distribution of PD-L1 CPS in samples from patients with locally advanced unresectable, recurrent, or metastatic, HER2-negative esophagogastric adenocarcinoma in Spain

Discussion

PD-L1 expression is one of the most studied predictive biomarkers of ICI efficacy [9, 10]. In this study of Spanish patients with aEGAC, PD-L1 CPS expression was similar to that observed in other real-world clinical practice series, but lower than the prevalence reported in clinical trials. Specifically, 31.9% of patients had a CPS of 1–4, 24.3% had a CPS of 5–9, and 17.4% had a CPS ≥ 10. This prevalence varied according to the location of the primary tumor, being higher in EAC (60.0%) than in GAC (41.6%) or GEJAC (35.7%).

Most data on PD-L1 CPS in aEGAC comes from clinical trials such as CheckMate 032 and CheckMate 649 [12, 25–28]. In the phase I/II, open-label, dose-escalation CheckMate 032 involving 104 patients with advanced GAC, GEJAC, or EACs, the prevalence of PD-L1 CPS ≥ 5 and CPS ≥ 10 was 50% and 33%, respectively [28]. In the CheckMate 649 trial, which included 1,581 patients with untreated advanced GAC, GEJAC, or EAC and HER2-negative, PD-L1 CPS ≥ 5 was observed in 60.4% [12].

Real-world data on PD-L1 CPS prevalence in aEGAC is limited but provides critical insights [17–21]. The summary of the PD-L1 expression data in studies of aEGACs is shown in Table 3. A retrospective, observational study conducted in the United States utilized a nationwide, de-identified electronic health record database of 1142 adult patients with aEGAC who initiated systemic treatment between 2017 and 2020. Of these, 571 patients underwent PD-L1 testing, predominantly with the commercial 22C3 assay (used in over 70% of cases). Samples were obtained from the primary tumor (74.6%) or metastatic sites (25.4%). Among the 406 patients with available PD-L1 CPS data, 88.2% had a CPS ≥ 1, 49.3% had a CPS ≥ 5, and 33.3% had a CPS ≥ 10. By tumor subtype, PD-L1 CPS ≥ 5 was present in 48.4% of the 128 GAC samples, 45.8% of the 96 GEJAC samples, and 43.2% of the 139 EAC samples [19]. Another retrospective, observational, multicenter study conducted in Japan evaluated 389 patients with advanced GAC and GEJC after surgical resection (stage II/III) or with unresectable advanced cancer (stage IV). PD-L1 expression was assessed using the 22C3 assay in biopsy samples (81.2%) or surgical resection specimens (18.8%). Among these patients, 62.0% had PD-L1 CPS ≥ 1, 40.1% had CPS ≥ 5, and 27.0% had CPS ≥ 10 [17]. A retrospective, unicenter study from China analyzed clinical data from 930 patients with GAC (stages I–IV) with available PD-L1 CPS. PD-L1 expression was also assessed using the 22C3 assay in biopsy or metastatic tumor samples. In this cohort, 62.3% had CPS ≥ 1, 49.2% had CPS ≥ 5, and 33.3% had CPS ≥ 10 [18]. A similar retrospective, unicenter study in Korea determined PD-L1 expression in 399 patients with metastatic GAC, also using the 22C3 assay on biopsy or metastatic tumor samples. This study found that 69.2% of patients had CPS ≥ 1, while 30.3% had CPS ≥ 5 [21]. A retrospective, unicenter study from Thailand analyzed PD-L1 expression in 268 patients with GAC (stages I–IV) using the 22C3 assay on resection specimens. In this cohort, 21.6% of patients had PD-L1 CPS ≥ 1, and only 6.7% had CPS ≥ 5 [20]. A retrospective observational study using electronic health records from the US Flatiron Health database recently evaluated the real-world treatment and testing patterns of adults diagnosed with aEGAC. Among 790 patients treated with first-line chemotherapy-only therapies, 54.2% underwent PD-L1 CPS testing. Of them, 21.1% had PD-L1 CPS ≥ 5. Among 362 patients on immuno-oncology-based treatments, 74.0% underwent PD-L1 CPS testing, with 38.4% having a PD-L1 CPS ≥ 5 [22].

Table 3.

Summary of PD-L1 expression data in studies of advanced esophagogastric adenocarcinoma

Study	Number of cases	Tumor location	Percentage of PD-L1 expression			Additional comments
		(Stomach, GEJ, esophagus)	≥ 1	≥ 5	≥ 10
CARPATOS-AGAMENON/SEOM (present study)	144	Mixed (GAC, GEJAC, EAC); Breakdown: EAC 60.0%, GAC 41.6%, GEJAC 35.7%	73.6%	41.7%	17.4%	Real-world Spanish study
CheckMate 032 (Lei et al. 2021) [28]	104	Mixed (GAC, GEJAC, EAC)	N/A	50%	33%	Phase I/II clinical trial
CheckMate 649 (Janjigian et al. 2021) [12]	1,581	Mixed (GAC, GEJAC, EAC)	N/A	60.4%	N/A	HER2-negative; phase III clinical trial
US EHR study (Mehta et al. 2023) [19]	406	Mixed (GAC, GEJAC, EAC); Subgroup breakdown: GAC 48.4%, GEJAC 45.8%, EAC 43.2%	88.2%	49.3%	33.3%	Retrospective study; 22C3 assay used
Japanese study (Yoshida et al. 2022) [17]	389	GAC and GEJC (Stomach and GEJ)	62.0%	40.1%	27.0%	Multicenter retrospective study
Chinese study (Chen et al. 2022) [18]	930	GAC (Stomach)	62.3%	49.2%	33.3%	Unicenter retrospective study
Korean study (Shin et al. 2023) [21]	399	GAC (Stomach)	69.2%	30.3%	N/A	Unicenter retrospective study
Thai study (Chitapanarux et al. 2023) [20]	268	GAC (Stomach)	21.6%	6.7%	N/A	Unicenter retrospective study
US Flatiron Health study (Chemo-only) (Kakehi et al. 2024) [22]	790	Mixed	N/A	21.1%	N/A	54.2% underwent PD-L1 testing
US Flatiron Health study (Immuno-oncology) (Kakehi et al. 2024) [22]	362	Mixed	N/A	38.4%	N/A	74.0% underwent PD-L1 testing

GEJ gastroesophageal junction, GAC gastric adenocarcinoma, GEJAC gastroesophageal-junction adenocarcinoma, EAC esophageal adenocarcinoma, EHR electronic health record, GEJC gastroesophageal junction cancer

N/A indicates that the specific percentage was not provided or not applicable. The “Additional Comments” column includes extra details relevant to each study, such as the assay used, study design, or other pertinent notes

Our study, based on a cohort of Spanish patients with unresectable advanced, recurrent, or metastatic EAC, GEJAC, or GAC, revealed a prevalence of PD-L1 CPS ≥ 5 in 41.7% of cases and 31.9% having a PD-L1 CPS between 1–4. These findings highlight the heterogeneity in PD-L1 CPS expression among aEGAC patients and underscore the importance of individualized approaches in selecting candidates for ICIs. Our PD-L1 CPS ≥ 5 result agrees with some previous real-world data, such as rates of 49.3% [19] and 49.2% [21], but differs from others: 40.1% [17], 38.4%, and 21.1% [22], and it is slightly lower than those reported in clinical trials, especially in the CheckMate 649, which reported a prevalence of 60.4% [12]. Differences in PD-L1 CPS prevalence between studies may arise from methodological disparities, such as specimen location (e.g., biopsy vs. resection) or immunohistochemistry assay used, sample adequacy, inter-observer interpretation, or tumor heterogeneity (for example Lauren classification). For example, PD-L1 CPS in gastric mucosa biopsy has been shown to be significantly lower than in surgical specimens [18]. Additionally, comparisons of PD-L1 expression using the 28–8 and 22C3 pharmDx assays have yielded inconsistent results, with some studies reporting higher positivity rates with the 28–8 assay [29, 30], while others report the opposite [31]. These discrepancies underscore the need to standardize methodologies for assessing PD-L1 expression to ensure consistency across studies and enhance the utility of PD-L1 CPS as a predictive biomarker for ICI treatment in clinical practice. It is important to note that, out of the eligible patients for receiving immunotherapy, either PD-L1 ≥ 1 (73.6%) or PD-L1 ≥ 10 (17.4%), only 9.7% of patients were finally receiving it in our series due to lack of indication in Spain at that time.

A potential limitation of the present study may stem from the selection of patients, which is an inherent challenge in retrospective observational studies. However, the use of consecutive recruitment across centers aimed to reduce selection bias. Additionally, the inclusion of biopsy samples from deceased patients further mitigates this issue by providing a broader and more representative dataset. Another limitation is the reliance on archival biopsy samples collected up to 2.5 years before the study inclusion. While this approach was necessary to ensure sufficient sample size, the potential impact of sample age on PD-L1 CPS determination cannot be entirely excluded. Nevertheless, existing evidence suggests that this time of biopsy collection does not significantly affect PD-L1 CPS evaluation. In contrast, the main strength of the study is that it has a specific design to evaluate the real-world distribution of PD-L1 CPS in aEGAC, addressing a critical knowledge gap in this area. This is particularly relevant as such data are scarce globally and entirely absent in Spain prior to this study.

In this multicenter Spanish real-world cohort of HER2-negative advanced esophagogastric adenocarcinoma with centralized PD-L1 assessment using the 28–8 pharmDx assay, PD-L1 CPS ≥ 5 was observed in 41.7% of patients, and 17.4% had CPS ≥ 10. PD-L1 positivity varied by primary tumor site (highest in esophageal adenocarcinoma) and showed comparable rates across metastatic, recurrent, and unresectable locally advanced disease. These data provide the first Spain-specific estimates of PD-L1 CPS distribution in clinical practice. Differences from clinical-trial prevalences likely reflect population and methodological factors (e.g., specimen type and assay), highlighting the need for further harmonization of PD-L1 testing. Prospective studies in larger cohorts are warranted to confirm these findings and to better characterize determinants of PD-L1 expression in Spanish patients with aEGAC.

Author contributions

Conception or design: PJ-F, AC-B, JGP, MRG, MEMG, DV, MI-C. Data acquisition: PJ-F, MDG, JLC-V, PP-W, LV, AC-B, MLL, AF-M, MR, RV-T, JGP, GEM, SL, AT, LG-P, AN, JLD-M, MLL-V, MR-G, FJI-G, EMdC, CH, MG-R, MI-C. Data analysis: PJ-F, AC-B, JGP, MRG, MEMG, DV, MI-C. Data interpretation: PJ-F, MDG, JLC-V, PP-W, LV, AC-B, MLL, AF-M, MR, RV-T, JGP, GEM, SL, AT, LG-P, AN, JLD-M, MLL-V, MR-G, FJI-G, EMdC, CH, MG-R, MRG, MEMG, DV, MI-C. Manuscript review: PJ-F, MDG, JLC-V, PP-W, LV, AC-B, MLL, AF-M, MR, RV-T, JGP, GEM, SL, AT, LG-P, AN, JLD-M, MLL-V, MR-G, FJI-G, EMdC, CH, MG-R, MRG, MEMG, DV, MI-C.

Funding

This work was funded by Bristol-Myers Squibb. Medical writing support was provided by Evidenze Health España S.L.U. and funded by Bristol-Myers Squibb.

Data availability

Data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Conflict of interest

PJ-F declares consultant, advisory, or speaker roles from BMS. JLC-V has received research support/funding/grants from Advanz Pharma, and Escuela Andaluza de Salud Pública; speaker honorarium from AstraZeneca, Nutricia, PharmaMar, and Servier; travel support from Amgen, AstraZeneca, BMS, Esteve, Gilead, Ipsen, Jazz Pharmaceuticals, Lilly, Merck, MSD, Novartis, Nutricia, Pfizer, Pierre Fabre, Roche, Sanofi, Servier, Takeda. He also owns stocks in PharmaMar. AC-B has have received research support/funding/grants from Astra-Zeneca; speaker honorarium from Novartis and Astellas; consultant honorarium from Astellas; and travel support from Amgen, Astellas, and Servier. MLL has have received speaker honorarium from Bristol Myers Squibb, consultant honorarium from Astra Zeneca, and travel support from LEO FARMIS. AF-M has received speaker honorarium from Amgen, Merck, Sanofi, Servier, Astra Zeneca, Daichi-Sainchi, Takeda, MSD, BMS, BeOne, Asofarma, and Leo Pharma. She also has received consultant honorarium from Astra Zeneca, BeOne, BMS, Amgen, LeoPharma, and travel support from Roche, Takeda, and Astra Zeneca. RV-T declares consultant, advisory, or speaker roles for Amgen, Servier, BMS, Merck, MSD, Pierre-Fabre, Astellas, Bayer, GSK, and educational, scientific activities and travel and accommodation support from Roche, Servier, MSD, Merck, BMS and Amgen. JGP declares advisory role for Astellas, Beigene, BMS and MSD and speaker role for BMS. AT has received speaker honorarium from Be one, MSD and Astrazeneca. LG-P has received speaker honorarium from BMS. EMdC has received speaker honorarium from BMS, Astra Zeneca, MSD, Rovi, LeoPharma, Sanofi Aventis, Roche, Servier; consultant honorarium from BMS, Incyte, MSD, Servier; and travel support from Roche, Servier. CH declares consultancy roles (speaker and advisory roles) for Lilly, AstraZeneca, Bristol-Myers Squibb (BMS), MSD and Astellas. She has received research grants and support via Institution from Merck. She is involved as principal investigator in clinical trials from BMS, MSD, Jazz Pharmaceuticals, ALX Oncology, AstraZeneca, Relay Therapeutics, Abbie and Medicenna, and has received travel and accommodation expenses from Amgen, Roche, BMS, MSD, Jazz Therapeutics and Merck. MG-R has received speaker honorarium from Bristol, Servier, Astellas, Astrazeneca, and Merck; consultant honorarium from Bristol; travel support from Bristol, MSD, MErck, Amgen, and Servier; and has participated in advisory boards from Bristol. MI-C declares consultant, advisory, or speaker roles for Roche, Astellas, BMS, MSD, BeiGene, Gilead, AstraZeneca, Servier, Merck and Daichii-Sankyo. MRG, MEMG, and DV are employees at Bristol Myers Squibb. MDG, PP-W, LV, MR, GEM, SL, AN, JLD-M, MLL-V, MR-G, and FJI-G declare no conflict of interest.

Ethics approval

The study was conducted following the Declaration of Helsinki and approved by the Ethical Committees of the participating hospitals (Approval number: 2023/10865/I from the first hospital, i.e. the Hospital del Mar Research Institute, Barcelona, Spain).

Informed consent

Except for deceased patients (in whom informed consent was waived), all patients provided a written informed consent to participate in the study.