Tissue Claudin 18.2 levels as a prognostic factor in gastric adenocarcinomas: A retrospective cohort study

Abstract

Gastric cancer continues to be a worldwide health crisis because it results in many deaths. The protein Claudin 18.2 (CLDN18.2) exists at higher levels in gastric adenocarcinoma tumors which makes it a promising therapeutic target. This study evaluated CLDN18.2 protein levels in gastric cancer tissues and their relationship with patient and tumor characteristics. This retrospective cohort study analyzed 112 gastric adenocarcinoma patients together with 62 healthy controls. The research team used ELISA to determine CLDN18.2 concentrations in tumor tissue and adjacent normal tissue and control specimens. Patients were stratified based on their Helicobacter pylori infection status, tumor site, tumor grade, and presence of metastasis. Diagnostic performance was assessed using receiver operating characteristic analysis, and Cox regression analysis was performed to evaluate prognostic factors for survival. The cancer tissue contained elevated CLDN18.2 levels which measured at 2.385 ng/mL (median), whereas normal tissue and controls displayed 1.349 ng/mL and 1.260 ng/mL (median) respectively (P < .001). The receiver operating characteristic analysis for diagnosing gastric adenocarcinoma showed 92.4% sensitivity and 100% specificity at 1.675 ng/mL (area under the curve: 0.956). The study found that CLDN18.2 levels were higher in patients with Helicobacter pylori infection and in cases with poor tumor differentiation and metastasis. The risk of metastasis increased by 4.82 times for every 1 ng/mL rise in CLDN18.2 levels (P < .001). This study demonstrates that CLDN18.2 exhibits strong diagnostic performance for gastric adenocarcinoma and shows strong correlation with cancer progression. The Cox regression analysis established CLDN18.2 as an independent prognostic factor for patient survival outcomes.

1. Introduction

Stomach cancer represents a major public health concern because it results in 1 million new cases and 7,00,000 to 7,69,000 deaths each year worldwide, making it the third leading cause of death.^[1,2] The survival rates for patients with advanced cancer remain low because their 5-year survival rates fall between 5% and 10%.^[3,4] The high death rate from this disease indicates that current treatments remain inadequate, highlighting the urgent need for novel therapeutic strategies.

The serum tumor markers CEA, CA19-9, and CA72-4 serve as established markers for gastric cancer diagnosis, staging, and disease monitoring.^[5] The conventional biomarkers show poor diagnostic performance because they detect gastric cancer with sensitivities of 20.1% for CEA, 21.4% for CA19-9, and 27.6% for CA72-4, and their combined sensitivity reaches only 48.2%.^[6] The markers fail to provide sufficient data to support decisions about targeted therapy and immunotherapy treatment.^[7] Molecular biomarkers have become increasingly important for developing individualized treatment plans in gastric cancer patients during the last few years. The first molecular biomarker in gastric cancer which is human epidermal growth factor receptor 2 (HER2) exists in 8% to 20% of gastric cancer cases and provides a target for trastuzumab treatment.^[5,7] The target for trastuzumab therapy exists in 8% to 20% of gastric cancer patients who show HER2 protein overexpression.^[5] The immune checkpoint inhibitor treatment for gastric cancer depends on the presence of PD-L1 expression which occurs in 65% of gastric cancer cases.^[5] The immunotherapy treatment for gastric cancer patients depends on their MSI-H/dMMR phenotype which affects 10% of all gastric cancer cases.^[5] The selection of targeted therapy for gastroesophageal adenocarcinoma requires 4 key biomarkers, including HER2, dMMR/MSI, PD-L1, and Claudin 18.2 (CLDN18.2).^[7]

The transmembrane protein CLDN18.2 exists in 40% to 80% of gastric cancer patients who have diffuse-type tumors and shows its highest expression in these cases.^[1,8,9] The protein CLDN18.2 is predominantly expressed in differentiated gastric mucosal cells during physiological conditions but can be targeted after cancer cells destroy their tight junctions.^[5] The restricted normal tissue expression of this protein along with its widespread presence in cancer tissues makes it suitable for targeted therapeutic approaches. The development of anti-CLDN18.2 monoclonal antibodies including zolbetuximab during recent years demonstrates the therapeutic value of this molecule.^[10] Zolbetuximab was approved for treating CLDN18.2-positive HER2-negative gastric cancer in Japan during March 2024 and in the United States and Europe during October 2024. The treatment requires tumors to exhibit at least 75% of cells showing moderate-to-strong membranous staining.^[7]

Multiple studies in the current literature present conflicting data about CLDN18.2 expression levels and prevalence and its prognostic value in gastric cancer.^[3] The main reason for these differences stems from different research approaches which include different criteria for CLDN18.2 positivity, different scoring systems, antibody specificity, staining protocols, and quantification methods.^[2,3] Research studies present conflicting data about CLDN18.2 prevalence because they used different standardized testing methods which resulted in reported rates ranging from 28% to 80%.^[2,5] The current understanding of CLDN18.2 expression in relation to clinicopathological parameters and its prognostic value remains insufficient.^[4,11] The scientific community lacks studies that examine how CLDN18.2 is associated with Helicobacter pylori infection together with tumor localization and survival outcomes in patients.^[9] This study investigates the prognostic value of tissue CLDN18.2 levels in gastric adenocarcinomas as well as their relationship to clinicopathological parameters and their diagnostic capabilities to fill an essential knowledge gap in the literature.

2. Materials and methods

2.1. Study design

The research followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational studies and employed a retrospective cohort design, as illustrated in Figure 1.

Figure 1.

Study flow diagram. ELISA = enzyme-linked immunosorbent assay, GIST = gastrointestinal stromal tumor, H pylori = Helicobacter pylori, ROC = receiver operating characteristic.

2.2. Study population and sample

The research studied patients who received gastric adenocarcinoma diagnoses at Istanbul Medipol University Faculty of Medicine Hospital during January 2020 through June 2025. The researchers obtained all samples by December 2023 before they analyzed survival information from 2025 after receiving ethical approval.

The study included patients who received primary gastric adenocarcinoma diagnosis through histopathological examination and who were at least 18 years old and had all necessary medical information. The study excluded patients who had lymphoma or gastrointestinal stromal tumor (GIST) or previous chemotherapy or radiotherapy treatment or additional cancers or severe organ deterioration or liver cirrhosis or inadequate tissue material or absent follow-up information (Fig. 1). The researchers excluded 12 patients from H pylori-related subgroup analyses because their H pylori status remained unknown.

CLDN18.2 expression levels were measured in cancer tissue and normal mucosa (≥10 cm from tumor). The control group consisted of dyspepsia patients with benign endoscopic findings. Our region has high H pylori prevalence, potentially affecting control subjects’ gastric mucosa. A significant age difference existed between groups (P < .05, Table 1), which was statistically adjusted. Metastasis was assessed per tumor-node-metastasis staging,^[12] differentiation grade per World Health Organization classification,^[13] and tumor location (cardia, corpus, antrum, or diffuse) based on endoscopic/pathological reports.

Table 1.

Patient characteristics and Claudin 18.2 expression levels.

Characteristic	Case group (n = 112)	Control group (n = 62)	P value
Age (yr)	63.4 ± 14.6	60.3 ± 9.8	.147†
Age groups n (%)			.039‡
40–49	13 (11.6%)	5 (8.1%)
50–59	30 (26.8%)	21 (33.9%)
60–69	20 (17.9%)	22 (35.5%)
70–79	26 (23.2%)	12 (19.4%)
≥80	17 (15.2%)	0 (0.0%)
Gender n (%)			.194‡
Male	73 (65.2%)	34 (54.8%)
Female	39 (34.8%)	28 (45.2%)
H pylori status n (%)
Positive	76 (67.9%)	–
Negative	24 (21.4%)	–
Unknown	12 (10.7%)	–
Tumor location n (%)
Cardia	18 (16.1%)	–
Corpus	30 (26.8%)	–
Antrum	58 (51.8%)	–
Diffuse	6 (5.4%)	–
Degree of differentiation n (%)
Well-differentiated	32 (28.6%)	–
Moderately differentiated	19 (17.0%)	–
Poorly differentiated	61 (54.5%)	–
Metastasis status n (%)
No metastasis	38 (33.9%)	–
Locally advanced	31 (27.7%)	–
Distant metastasis	43 (38.4%)	–
Claudin 18.2 levels (Median [IQR]) ng/mL*
Normal tissue	1.349 (1.276–1.410)	1.260 (1.090–1.408)	<.001§
Cancerous tissue	2.385 (2.123–3.648)	–

Post hoc Dunn: Control vs Case-normal P = .086, Control vs Case-cancer P < .001, Case-normal vs Case-cancer P < .001.

IQR = interquartile range.

^*CLDN18.2 levels expressed as ng/mL (5 mg tissue in 1:9 w/v PBS).

^†Student’s t test.

^‡Chi-square test.

^§Kruskal–Wallis test.

2.3. Intervention protocol

The research included 112 participants with gastric adenocarcinoma in the case group while the control group consisted of 62 participants with benign endoscopic pathology who matched the case group by age and sex. The researchers collected 2 tissue samples from cases which included tumor tissue and macroscopically normal mucosa that was at least 10 cm away from the tumor using separate biopsy forceps in individual containers. The researchers collected normal mucosa samples from all control participants. The laboratory processed all samples through the same procedure to determine CLDN18.2 levels in ng/mL. CLDN18.2 levels were measured in ng/mL, and age was recorded in years.

2.4. Tissue sampling and biochemical analyses

Patients were identified by including those who presented to the hospital outpatient clinic and underwent upper gastrointestinal endoscopy (SonoScape HD500; China) during which a mass was detected in the stomach, and subsequently diagnosed with adenocarcinoma based on standard histopathological examination with hematoxylin and eosin (H&E) staining. Clinical and survival data were retrospectively collected from patient records.

During the endoscopy procedure, 2 to 3 punch biopsies were taken from patients and stored at −80°C (Nuvo DF 490; Turkey) without further processing. The manufacturer states that phosphate-buffered saline (PBS) buffer (0.01 M at pH 7.4) should be used to extract blood residues from tissue samples before performing CLDN18.2 tissue determination. The tissue processing involved weighing 5 mg of tissue from each sample before using Qiagen Tissue Lyser to homogenize the tissue with 9 parts of cold PBS (Sigma-Aldrich, St. Louis) by weight. The Bandelin Sonopuls HD 2070 ultrasonic homogenizer (Bandelin Electronic GmbH & Co. KG, Berlin, Germany) was used to sonicate the samples until they reached complete cell lysis. The homogenates were then centrifuged at +4 °C at 800g (5000 rpm) for 5 minutes, and the supernatants were collected.

Tissue CLDN18.2 concentration in the obtained supernatants was measured using KRISHGEN BioSystems KRIBIOLISA™ Human Claudin 18.2 enzyme-linked immunosorbent assay (ELISA) Kit (Cat No#KBGT903). This kit employs an indirect capture antibody sandwich ELISA technique where monoclonal Anti-Claudin 18.2 is pre-coated onto microwells. Samples were pipetted into microwells, and CLDN18.2 present in samples was bound by the antibodies. Anti-Human IgG:HRP Conjugate was then added and incubated to form a complex. The CLDN18.2 amount determined the amount of TMB substrate solution which produced color development. The amount of CLDN18.2 protein caused the solution to develop color which directly correlated to its concentration. The stop solution stopped color development so the instrument measured absorbance at 450 nm. The kit detects CLDN18.2 specifically according to manufacturer specifications which includes tissue homogenates as an accepted sample type.

The researchers used weight-based standardization through exact tissue weight measurements of 5 mg per sample and they kept a constant buffer volume ratio instead of using total protein normalization. The method establishes uniform tissue starting points which reduces variability in experimental results between different samples. Since gastric cancer tissues often have variable protein content due to necrosis, fibrosis, and inflammatory cell infiltration, weight-based standardization directly relates CLDN18.2 levels to tissue mass, which is clinically more interpretable.

The presence of H pylori bacteria was established through histopathological examination which used Giemsa staining to detect spiral bacteria as a positive indicator. The research team retrieved survival data from hospital records and performed telephone interviews to verify and update the information. The researchers monitored patients for 28 months as the study period ranged from 6 months to 60 months. The ELISA kit evaluation process involved determining intra-assay coefficient of variation through plate testing of 10 identical samples and inter-assay coefficient of variation through testing the same samples across 10 different days which resulted in 5.8% and 7.2% respectively.

2.4.1. Pathological assessment and tissue cellularity validation

The 2 experienced pathologists examined all tissue specimens from cancer patients in the prospective phase to confirm adenocarcinoma diagnosis and assess their readiness for quantitative protein analysis. The pathologists confirmed that cancer tissue and normal tissue samples had matching cell numbers and structural patterns which made them suitable for protein analysis through tissue weight normalization.

The research team removed samples that showed major cellular density differences and deep stromal invasion and large areas of necrosis and tumors with irregular cellular arrangements because these factors would affect the accuracy of CLDN18.2 measurements. The pathological assessment revealed that tumor tissue contains slightly more cells than normal mucosa because of different levels of differentiation and invasion, but the cell number difference remains within 20 to 30% according to histological assessment without impacting tissue-based ELISA quantification accuracy.

The consulting pathologists verified that ELISA-based CLDN18.2 protein measurement through 5 mg tissue samples with standardized cell density assessments allows successful protein expression evaluation between gastric tumors and normal mucosa tissues. The method generates accurate results for tissue samples which have undergone full pathological evaluation to verify their cellular structure.

The research team conducted all experiments twice according to their established protocol.^[14] The authors used World Health Organization classification standards for gastrointestinal tumors to maintain uniform terminology across all sections of the manuscript.

The research team started their independent evaluation by placing cancer tissue and normal mucosa samples into separate containers. The researchers conducted ELISA tests on separate samples to maintain the unique expression patterns of CLDN18.2 protein levels. The analytical process followed a particular separation protocol which produced complete separation between cancerous and normal tissue samples. The tissue homogenization process used a 1:9 (w/v) ratio to achieve equal results between all samples that were tested.

2.5. Statistical analysis

The research team used SPSS version 25.0 (IBM Corporation, Armonk) for data analysis. The researchers used Kolmogorov–Smirnov test to check for normal distribution in the data. The research presented normally distributed data as mean (Standard deviation) but non-normal data as median (interquartile range [IQR]). The research team applied Student t test for parametric data comparison but used Mann–Whitney U and Kruskal–Wallis tests for nonparametric data. The researchers conducted chi-square tests to analyze categorical data and Spearman tests to determine the strength of correlations. The researchers performed survival analysis through Kaplan–Meier method with log-rank test. The researchers used Cox regression to identify prognostic factors and logistic regression to determine metastasis risk factors by defining metastasis as the combination of locally advanced and distant metastasis cases versus no metastasis cases. The researchers conducted receiver operating characteristic (ROC) analysis to determine the diagnostic performance of CLDN18.2 for differentiating gastric adenocarcinoma from nonmalignant gastric tissue. The researchers removed <5% of missing data points from the analysis. The study considered all results with P values < .05 as statistically significant. P values were reported using fixed values as examples (e.g., P = .04 for values ≥ .01 and P = .008 for values < .01), whereas smaller values were reported as P < .001.

This study was not based on a formal a priori sample size calculation; the sample size was determined by the availability of eligible patients during the study period. The investigators conducted a post hoc power analysis through GPower 3.1 software.^[15] The study included 112 participants for its main analysis, comparing CLDN18.2 expression between cancer and normal tissue. The study achieved more than 99% statistical power for its primary analysis, which compared cancer and normal tissue CLDN18.2 levels at α = 0.05 with an observed effect size of Cohen’s d = 1.35. For the independent comparison between cancer tissue (n = 112) and control group (n = 62), achieved power was also >99% (Cohen’s d = 1.81). The size of these effects demonstrates that CLDN18.2 expression levels between cancerous and normal tissues exhibit substantial biological variations.

2.6. Ethical considerations

The research received approval from Istanbul Medipol University Non-Interventional Clinical Research Ethics Committee on May 29, 2025 through decision E-10840098-202.3.02-3380. The researchers obtained permission to use patient data without consent because the study used existing records and protected all information by anonymization. The study followed Helsinki Declaration principles.

3. Results

3.1. Patient demographics and clinical characteristics

The tissue sampling protocol was validated by the successful analysis (cancer tissue and normal mucosa) from all 112 cancer patients. The independent processing of these samples was confirmed by the distinct CLDN18.2 expression patterns observed between tissue types, with no evidence of sample cross-contamination or pooling artifacts.

The study included 112 patients with gastric adenocarcinoma and 62 healthy controls. The mean age of patients was 63.4 ± 14.6 years compared to 60.3 ± 9.8 years in controls, with no significant difference between groups (P = .147). The patient population consisted of 65.2% males and 34.8% females, but the control group consisted of 54.8% males and 45.2% females. The highest prevalence rate of 26.8% occurred among patients in the 50 to 59 age group according to Table 1. However, there was a statistically significant difference in age group distribution between the case and control groups (P = .039), particularly with 17 patients aged ≥80 years in the case group versus none in the control group, which represents a potential confounding factor.

The clinical assessment of patients produced the following findings:

• The H pylori test showed positive results in 67.9% of patients.

• The tumor locations were mostly found in the antrum (51.8%) among patients.

• The tumor locations were distributed as follows: corpus (26.8%), cardia (16.1%) and diffuse-type (5.4%).

The histopathological grading results showed the following distribution:

• Poorly differentiated adenocarcinoma: 54.5%;

• Well-differentiated adenocarcinoma: 28.6%;

• Moderately differentiated adenocarcinoma: 17.0%.

The evaluation of metastasis status showed that distant metastasis occurred in 38.4% of patients while 27.7% had locally advanced disease and 33.9% showed no metastasis (Table 1).

3.2. Claudin 18.2 expression analysis

3.2.1. Tissue expression levels

The research on CLDN18.2 expression showed that different tissue samples contained varying levels of this protein. The cancer tissue showed the highest CLDN18.2 expression at 2.385 ng/mL (IQR: 2.123–3.648) while adjacent normal tissue showed 1.349 ng/mL (IQR: 1.276–1.410) and control group normal tissue showed 1.260 ng/mL (IQR: 1.090–1.408; P < .001). The tissue processing followed standardized methods which combined 5 mg of tissue with PBS at a 1:9 (w/v) ratio to achieve equal ng/mL values that represented CLDN18.2 content per tissue weight. The uniform treatment of all samples produced equal protein amounts in every specimen. The post hoc analysis showed that cancer tissue expressed CLDN18.2 at different levels than both control group tissue and adjacent normal tissue, but control group tissue showed no difference with adjacent normal tissue from patients (Table 1).

The 2.19-fold increase in CLDN18.2 levels was consistently observed in all 112 patients, indicating a true biological difference rather than a measurement artifact due to tissue composition variability. The standardized ELISA method enabled quantitative analysis of all samples although immunohistochemical validation would have added extra confirmation.

3.2.2. Demographic distribution

The control group showed higher CLDN18.2 levels in females (1.326 ng/mL) than in males (1.215 ng/mL) according to gender-based analysis (P = .042). The research found no meaningful gender differences in tissue samples from patients with normal or cancerous conditions. The research analyzed different age brackets to show CLDN18.2 levels surpassed normal tissue amounts in all cancerous tissue samples from participants of every age group (Table 2). The study results showed CLDN18.2 levels exceeded normal tissue levels in cancerous tissue across all age ranges with significant statistical results (P < .001 for ages 50–79; P = .002 for 80 + years; Table 2).

Table 2.

Distribution of Claudin 18.2 levels according to demographic and clinical-pathological characteristics.

Parameter	n	Claudin 18.2 (Median [IQR]) ng/mL*	P value
Distribution by gender
Control group			.042†
Male	34	1.215 (1.023–1.373)
Female	28	1.326 (1.205–1.448)
Case group - normal tissue			.316†
Male	73	1.343 (1.273–1.403)
Female	39	1.361 (1.298–1.428)
Case group - cancer tissue			.428†
Male	73	2.310 (2.130–3.560)
Female	39	2.710 (2.110–4.130)
Distribution by age groups
40–49			.018‡
Control	5	1.100
Case-normal	13	1.320
Case-cancer	13	2.420
50–59			<.001‡
Control	21	1.230
Case-normal	30	1.342
Case-cancer	30	2.265
60–69			<.001‡
Control	22	1.285
Case-normal	20	1.360
Case-cancer	20	2.200
70–79			<.001‡
Control	12	1.315
Case-normal	26	1.358
Case-cancer	26	2.835
80+ yr			.002‡
Control	0	–
Case-normal	17	1.386
Case-cancer	17	2.310
Clinical-pathological parameters (Cancer tissue)
H pylori status			.002†
Positive	76	2.540 (2.170–3.690)
Negative	24	1.950 (1.798–2.310)
Tumor location			.016‡
Cardia	18	2.180 (2.083–3.185)
Corpus	30	2.702 (2.243–3.763)
Antrum	58	2.245 (2.015–2.755)
Diffuse	6	5.735 (5.612–6.123)
Degree of differentiation			<.001‡
Well-differentiated	32	2.215 (2.118–2.345)
Moderately differentiated	19	2.130 (2.020–2.190)
Poorly differentiated	61	3.560 (2.470–4.420)
Metastasis status			<.001‡
No metastasis	38	2.100 (1.853–2.228)
Locally advanced	31	3.780 (3.560–4.490)
Distant metastasis	43	2.310 (2.160–3.180)

Median values in subgroups with n < 10 should be interpreted with caution.

IQR = Interquartile range.

^*CLDN18.2 levels expressed as ng/mL (5 mg tissue in 1:9 w/v PBS).

^†Mann–Whitney U test.

^‡Kruskal–Wallis test.

3.3. Association with clinicopathological parameters

The research of CLDN18.2 expression in cancer tissue in relation to clinicopathological parameters revealed several important findings. The H pylori-positive patients showed significantly higher expression (2.540 ng/mL) than H pylori-negative patients (1.950 ng/mL, P = .002). The highest expression of CLDN18.2 was observed in diffuse-type tumors (5.735 ng/mL) followed by corpus (2.702 ng/mL), antrum (2.245 ng/mL) and cardia (2.180 ng/mL) locations (P = .016). A strong correlation was observed between differentiation grade and CLDN18.2 levels, with significantly higher expression in poorly differentiated tumors (3.560 ng/mL) compared to moderately (2.130 ng/mL) and well-differentiated tumors (2.215 ng/mL; P < .001). The highest levels of CLDN18.2 were observed in patients with locally advanced disease (3.780 ng/mL), followed by distant metastasis (2.310 ng/mL) and no metastasis (2.100 ng/mL; P < .001; Table 2). The research shows that CLDN18.2 levels reach their peak in patients with locally advanced disease instead of patients who have distant metastases (Table 2).

3.4. Expression change analysis

The cancerous tissue displayed a 2.19-fold average increase and a 1.77-fold median increase when compared to normal tissue. The expression levels in cancer tissue exceeded normal tissue in all 112 patients (100%) while no patient showed reduced or stable expression. The expression change rate averaged 128.4% ± 88.2%, ranging from 36.8% to 435.7%. The Spearman correlation analysis between the 2 tissues showed a weak nonsignificant positive relationship (r = 0.121, P = .185; Table 3).

Table 3.

Claudin 18.2 expression levels and change analysis in normal and cancer tissue.

Parameter	Value	P value
Comparison between normal and cancer tissue
n	112
Normal tissue Claudin 18.2
Mean ± SD (ng/mL)	1.318 ± 0.115
Median (IQR; ng/mL)	1.349 (1.276–1.410)
Cancer tissue Claudin 18.2
Mean ± SD (ng/mL)	2.889 ± 1.106	<.001
Median (IQR; ng/mL)	2.385 (2.123–3.648)	<.001
Difference (Cancer/Normal)
Mean basis	2.1
Median basis	1.77
Spearman correlation analysis
R value	0.12
P value		.185
Expression change
Change from normal tissue to cancerous tissue
Number of patients showing an increase	112 (100.0%)
Number of patients showing a decrease	0 (0.0%)
Number of patients showing no change	0 (0.0%)
Change rate (%) statistics
Mean ± SD	128.4% ± 88.2
Median (IQR)	94.1% (73.3–187.5)
Minimum	36.8
Maximum	435.7
Relationship between the change rate and clinical-pathological parameters
Metastasis present vs absent	156.3% vs 82.1%	.002
Poorly differentiated vs well-differentiated	162.8% vs 72.4%	<.001
H pylori positive vs negative	138.7% vs 96.2%	.034

Paired t test Mann–Whitney U test.

IQR = interquartile range, SD = standard deviation.

The relationship between expression change rate and clinicopathological characteristics demonstrated:

• The expression change rate was approximately double in patients with metastasis (156.3%) than in patients without metastasis (82.1%, P = .002);

• The change rate was higher in poorly differentiated tumors (162.8%) than in well-differentiated tumors (72.4%, P < .001);

• The change rate was higher in H pylori-positive patients (138.7%) than in H pylori-negative patients (96.2%, P = .034; Table 3).

3.5. Subgroup analysis by H pylori status

The analysis of H pylori status subgroups produced different results regarding clinicopathological parameters. The analysis of H pylori-positive patients showed that CLDN18.2 expression levels correlated with tumor localization (P = .024) and differentiation grade (P < .001) and metastasis status (P = .008). The highest expression levels of CLDN18.2 were detected in diffuse-type tumors (5.890 ng/mL) and poorly differentiated tumors (3.615 ng/mL) and metastatic cases (2.780 ng/mL). The analysis of CLDN18.2 expression levels in H pylori-negative patients did not show any significant correlations with these parameters (Table 4).

Table 4.

Subgroup analyses of Claudin 18.2 expression in cancer tissue according to H pylori status.

	H pylori positive (n = 76)			H pylori negative (n = 24)
Parameter	n	Claudin 18.2 (Median) ng/mL*	P value	n	Claudin 18.2 (Median) ng/mL*	P value
Tumor localization			.024†			.142†
Cardia	12	2.710		4	2.075
Corpus	18	2.863		6	2.145
Antrum	45	2.230		13	1.820
Diffuse	3	5.890		1	2.310
Degree of differentiation			<.001†			.867†
Well-differentiated	19	2.180		9	2.210
Moderately differentiated	11	2.180		7	2.020
Poorly differentiated	46	3.615		8	2.080
Metastasis status			.008‡			.236‡
None	21	2.100		13	2.100
Available	55	2.780		11	1.820

The tissue processing involved 5 mg tissue samples in 1:9 (w/v) ratio with PBS to enable direct assessment of protein content per tissue weight.

^*Claudin 18.2 levels are expressed as ng/mL of tissue homogenate.

^†Kruskal–Wallis test.

^‡Mann–Whitney U test.

3.6. Multivariate analysis for metastasis risk

The multivariate logistic regression analysis revealed multiple independent factors which affect metastasis presence. The risk of metastasis increased by 4.82 times (95% confidence interval [CI]: 2.13–10.89, P < .001) for every 1 ng/mL increase in CLDN18.2 levels. The risk of metastasis increased by 8.76 times (95% CI: 3.12–24.58, P < .001) in male patients and each 10-year age increase provided protection (odds ratio: 0.42, 95% CI: 0.24–0.73, P = .002). The risk of metastasis was 4.31 times higher in poorly differentiated tumors than in well-differentiated tumors (95% CI: 1.52–12.23, P = .006). The presence of H pylori and tumor location did not affect the risk of metastasis. The model showed good fit with Nagelkerke R² = 0.487 and correct classification rate of 78.6% (Table 5).

Table 5.

Multivariate logistic regression analysis for predictors of metastases.

Independent variables	OR	95% CI	P value
Claudin 18.2 level (per 1 ng/mL increase)	4.82	2.13–10.89	<.001
Age (per 10-year increase)	0.42	0.24–0.73	.002
Gender (male vs female)	8.76	3.12–24.58	<.001
H pylori (positive vs negative)	1.03	0.36–2.94	.956
Localization (Antrum = reference)
Cardia	3.21	0.87–12.06	.080
Corpus	2.56	0.84–7.80	.098
Diffuse	1.12	0.18–6.95	.904
Differentiation (well-differentiated = reference)
Moderately differentiated	1.87	0.48–7.27	.367
Poorly differentiated	4.31	1.52–12.23	.006

Dependent variable: Presence of metastasis (Yes: n = 74, No: n = 38); “locally advanced” (n = 31) and “distant metastasis” (n = 43) combined as “metastasis present” vs “no metastasis” (n = 38). Model: Nagelkerke R² = 0.487, Hosmer–Lemeshow P = .642, classification rate 78.6%. P < .05 considered significant.

CI = confidence interval, OR = odds ratio.

3.7. Survival analysis

The survival analysis of CLDN18.2 expression levels at 2.385 ng/mL showed that patients with low expression achieved an 82.1% survival rate but patients with high expression achieved a 75.0% survival rate without reaching statistical significance (log-rank P = .417). However, multivariate Cox proportional hazards regression analysis identified CLDN18.2 level as a continuous variable as an independent prognostic factor (hazard ratio: 1.28, 95% CI: 1.09–1.51, P = .003), along with presence of metastasis and poor differentiation.

The observed difference between Kaplan–Meier and Cox regression results demonstrates a common statistical issue which occurs when researchers divide continuous variables into groups using random threshold values (such as the median) because this approach decreases statistical power and hides actual dose-response patterns. The Cox model maintains all available expression level data through continuous treatment of CLDN18.2 which results in better prognostic accuracy and sensitivity. The results show CLDN18.2 functions as a risk factor which produces worsening survival levels progressively instead of acting as a basic prognostic indicator. The median-based dichotomization method used for patient grouping lacks sufficient discriminatory power for risk stratification in this study population (Fig. 2).

Figure 2.

Survival analysis of gastric adenocarcinoma patients by Claudin 18.2 expression levels. (A) Kaplan–Meier curves comparing low (≤2.385 ng/mL) versus high (>2.385 ng/mL) expression groups (log-rank P = .417); (B) patients at risk; (C) survival distribution; (D) multivariate Cox regression results. CI = confidence interval, HR = hazard ratio.

3.8. Diagnostic performance

ROC analysis demonstrated that CLDN18.2 had excellent discriminatory properties for differentiating gastric adenocarcinoma from nonmalignant gastric tissue with an area under the curve of 0.956 (95% CI: 0.927–0.985, P < .001). The optimal cutoff point of 1.675 ng/mL yielded 92.4% sensitivity and 100% specificity. The evaluation of multiple threshold values showed that sensitivity increased when the cutoff value decreased but specificity remained 100% accurate at all tested values (Fig. 3).

Figure 3.

ROC curve analysis of Claudin 18.2 for gastric adenocarcinoma diagnosis. AUC: 0.956 (95% CI: 0.927–0.985, P < .001). Optimal cutoff: 1.675 ng/mL (sensitivity 92.4%, specificity 100%). AUC = area under curve, CI = confidence interval, NPV = negative predictive value, PPV = positive predictive value.

4. Discussion

The study examined how CLDN18.2 affects gastric adenocarcinoma tumor progression. Our findings revealed that this protein is significantly elevated in cancer tissue and plays an important role in the course of the disease. The research team discovered CLDN18.2 existed at high levels in tumor cells and showed stable expression compared to normal mucosa which directly correlated with aggressive disease features. The study shows CLDN18.2 serves as a diagnostic marker while simultaneously affecting the biological progression of gastric cancer.

The research findings showed that cancer tissue had higher CLDN18.2 concentrations than normal mucosa. The median CLDN18.2 level in cancer tissue was 2.385 ng/mL, while it was measured as 1.349 ng/mL in normal tissue, and this 2.19-fold increase was found to be statistically significant. Similarly, Jia and colleagues also reported significantly higher expression in the tumor region compared to the normal region.^[15] These findings support the potential use of CLDN18.2 as a biomarker for gastric cancer detection.

The study demonstrated that CLDN18.2 shows its most important prognostic value when analyzed as a continuous variable. The Kaplan–Meier analysis based on the median value of 2.385 ng/mL did not show significant difference between groups (log-rank P = .417), whereas multivariate Cox regression analysis established CLDN18.2 as an independent prognostic factor (hazard ratio: 1.28, 95% CI: 1.09–1.51, P = .003). This discrepancy reflects the well-known limitation of dichotomizing continuous variables, which reduces statistical power. Future research should focus on developing clinically validated thresholds through ROC curve analysis in larger populations.^[16-18]

The ROC analysis produced an area under the curve value of 0.956 which shows that the diagnostic test performed at an excellent level. The detection of 92.4% sensitivity and 100% specificity occurred at the optimal cutoff value of 1.675 ng/mL. These findings support Mohammed’s view that CLDN18.2 overexpression in gastric cancer could be used as a diagnostic biomarker.^[17] Although Qi and colleagues used a different evaluation method, their reporting of positivity rates ranging from 57.6% to 80.4% in immunohistochemical scoring confirms the diagnostic potential of CLDN18.2.^[18] The high diagnostic accuracy obtained with our quantitative ELISA method demonstrates that this protein can be used as an objective and reliable marker for the diagnosis of gastric adenocarcinoma.

We investigated the effect of H pylori infection on CLDN18.2 expression and obtained notable findings. The levels of CLDN18.2 in cancer tissue were significantly higher in H pylori-positive patients than in H pylori-negative patients (2.540 vs 1.950 ng/mL, P = .002). Moreover, the expression increase ratio from normal tissue to cancer tissue was higher in the H pylori-positive group (138.7% vs 96.2%, P = .034). These findings suggest that H pylori infection may increase CLDN18.2 expression in the gastric carcinogenesis process.

The research data revealed that CLDN18.2 expression associated to multiple clinicopathological features which were present in H pylori-positive patients and included tumor site and tumor grade and metastasis development. These findings are consistent with previous studies reporting associations between CLDN18.2 expression and tumor microenvironment components,^[19] poorly differentiated tumors and peritoneal metastasis,^[20-22] and overall survival.^[19] Our data notably demonstrate that H pylori presence strengthens these associations.

When we examined the relationship between CLDN18.2 expression and clinicopathological parameters, we found significantly higher levels in poorly differentiated tumors compared to well-differentiated tumors (3.560 vs 2.215 ng/mL, P < .001). This finding is consistent with the report by Weng et al^[23] and Son et al,^[24] who demonstrated that high CLDN18.2 positivity is associated with poor differentiation grade. These data suggest that CLDN18.2 expression increases in parallel with tumor differentiation loss.

The research findings showed that CLDN18.2 levels peaked at 3.780 ng/mL in patients with locally advanced disease yet patients with distant metastases had lower levels at 2.310 ng/mL. The observed pattern contradicts studies which indicate that peritoneal metastasis constitutes the principal link between CLDN18.2 and cancer.^[20-22,24] The 2 possible reasons for this difference could be that tumors at an advanced stage of local progression show their highest CLDN18.2 levels before they undergo epithelial-mesenchymal transition and that CLDN18.2 shows different expression patterns when cancer spreads through the hematogenous route rather than the peritoneum. The research should concentrate on examining different metastasis subtypes while researchers need to analyze CLDN18.2 concentration levels between primary tumors and their metastatic spread.^[20-22] The logistic regression model showed that each 1 ng/mL increase in the variable resulted in a 4.82 times higher risk of metastasis and poor differentiation led to a 4.31 times higher risk of metastasis.

The research findings from multivariate analysis showed that age acted as a protective factor which reduced the risk of metastasis by 42% for each 10-year age increment (P = .002) and males developed metastasis at a higher rate than females (odds ratio: 8.76, 95% CI: 3.12–24.58, P < .001). The wide CI shows that the estimates used for gender prediction produce highly uncertain results. The research findings could result from selection bias because our study at a single center only included elderly patients who were in good health before they underwent endoscopy procedures. The application of this method requires independent verification through prospective multi-center studies which must be conducted before medical use.^[25] The research by Pellino et al^[26] and Weng et al^[23] demonstrated that CLDN18.2 expression is associated with lymph node involvement and simultaneous metastasis development which makes it suitable for both diagnostic purposes and treatment approaches.

The research results demonstrated that tumor location and specific histological types of tumors determined the levels of CLDN18.2 expression. The highest expression levels were found in diffuse-type tumors compared to other localizations (5.735 ng/mL). This result is consistent with the findings of Kayikcioglu et al^[27] and Wong et al,^[28] who reported higher CLDN18.2 expression in diffuse adenocarcinomas compared to intestinal and mixed types. The research found that tumors located in the corpus region contained higher levels of CLDN18.2 than tumors found in the antrum and cardia regions (2.702 vs 2.245 vs 2.180 ng/mL, P = .016).^[27,28] The research findings indicate that CLDN18.2 plays an essential role in aggressive tumor types which makes it an attractive therapeutic option.

The strongest aspect of our study was the use of a quantitative measurement method to obtain objective results and the evaluation of both cancer and normal tissue in the same patient. The research contains multiple limitations which need to be acknowledged. We used tissue-based ELISA instead of immunohistochemistry (IHC), which would have shown protein distribution at the cellular level. The 2 pathologists confirmed that quantitative ELISA measurements remain reliable when tissue samples have equivalent cell density, and we excluded cases with major cellular density differences. The restricted access to CLDN18.2-specific IHC protocols which were cost-prohibitive during 2020 to 2023 led us to select this testing method.

The study faces 2 main limitations which are the dyspepsia control group originated from an area with elevated H pylori rates which could have influenced their initial CLDN18.2 gene expression levels. The study groups showed a statistically significant difference in age distribution which reached a P value of .039. The research had insufficient participants to perform separate analyses for different groups, and the study took place at a single location which reduces its ability to apply to other populations. The study results would become more robust through research conducted at multiple sites which includes bigger groups of participants who match in age to confirm CLDN18.2 serves as a diagnostic and prognostic tool for gastric adenocarcinoma. Future research should adopt prospective designs, perform molecular subtyping, and include parallel IHC validation to improve understanding of CLDN18.2 expression patterns.

5. Conclusion

CLDN18.2 was found to be a reliable diagnostic marker in gastric adenocarcinomas and closely associated with the clinical course of the disease. The molecule shows rising expression levels during tumor progression which makes it suitable for cancer diagnosis and prediction of patient outcomes. The research shows that H pylori infection is associated with altered CLDN18.2 expression levels, which may contribute to gastric cancer progression and metastatic potential.

Author contributions

Conceptualization: Enver Akbas, Gözde Ülfer.

Data curation: Enver Akbas.

Formal analysis: Enver Akbas.

Funding acquisition: Gözde Ülfer.

Investigation: Enver Akbas.

Methodology: Enver Akbas.

Project administration: Gözde Ülfer.

Resources: Enver Akbas.

Software: Enver Akbas.

Supervision: Enver Akbas, Gözde Ülfer.

Validation: Enver Akbas, Gözde Ülfer.

Visualization: Enver Akbas.

Writing – original draft: Enver Akbas.

Writing – review & editing: Enver Akbas, Gözde Ülfer.