Predictive value of neutrophil-to-lymphocyte ratio combined with carcinoembryonic antigen in postoperative adjuvant chemotherapy after gastric cancer surgery

Abstract

BACKGROUND

Gastric cancer (GC) is an aggressive malignancy of the gastrointestinal tract characterized by high recurrence rates following radical gastrectomy. To enhance treatment efficacy, reduce recurrence, and improve patient survival, adjuvant chemotherapy is commonly administered based on established postoperative guidelines. Despite advancements in chemotherapy delivery, the overall response rate remains below 50%, primarily due to the lack of targeted therapies tailored to specific patient populations.

AIM

To explore sensitive biomarkers to assess the efficacy of postoperative adjuvant chemotherapy in appropriate patient subgroups.

METHODS

This study retrospectively analyzed 1628 patients who underwent radical gastrectomy for GC at our hospital in 2017 and 2018, with a subsequent five-year follow-up. Patients were divided based on whether they received postoperative adjuvant chemotherapy. The study aimed to determine optimal cutoff values for various biomarkers-neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio, carcinoembryonic antigen (CEA), carbohydrate antigen (CA) 199, CA724, and CA242-using receiver operating characteristic (ROC) curves. Based on the optimal ROC cut-off, a novel combined metric, NLR-CEA, was developed to assess the efficacy of adjuvant chemotherapy following GC surgery.

RESULTS

Cox subgroup analysis demonstrated that postoperative adjuvant chemotherapy significantly improved overall survival in the NLR-CEA_Low group with a rate of 0.41 (0.26-0.63). In the NLR-CEA_Middle group, the protective effect was observed at 0.69 (0.54-0.87), while in the NLR-CEA_High group, it was 0.73 (0.53-0.99). Notably, a 32% difference in the efficacy of chemotherapy was observed between the NLR-CEA_Low and NLR-CEA_High groups.

CONCLUSION

NLR-CEA is an effective biomarker for evaluating the efficacy of postoperative adjuvant chemotherapy in GC. Patients with NLR-CEA_Low exhibit significantly better responses to chemotherapy compared to those with NLR-CEA_High.

Key Words: Neutrophil-to-lymphocyte ratio; Carcinoembryonic antigen; Gastric cancer; Adjuvant chemotherapy; Tumor microenvironment

Core Tip: This investigation conducted a retrospective analysis on a cohort of gastric cancer patients, revealing a 32% disparity in therapeutic efficacy of postoperative adjuvant chemotherapy across distinct neutrophil-to-lymphocyte ratio (NLR)-carcinoembryonic antigen (CEA) stratification groups. Utilizing a stratified approach based on adjuvant chemotherapy administration status, we performed covariate-adjusted Cox proportional hazards regression analyses to evaluate comparative therapeutic outcomes among NLR-based, CEA-based, and combined NLR-CEA stratification models. The findings demonstrate significantly greater differential therapeutic benefits in the combined NLR-CEA stratification group compared to isolated NLR or CEA parameter stratification, suggesting synergistic prognostic value of this composite biomarker approach.

INTRODUCTION

Gastric cancer (GC), originating from the epithelial cells of the gastric mucosa[1], ranks among the most invasive malignancies of the digestive system, contributing substantially to global cancer incidence and mortality[2]. Representing nearly 10% of all cancer-related deaths, GC constitutes a significant public health burden[3-7]. Surgical resection, typically involving total or subtotal gastrectomy with lymphadenectomy, remains the primary curative modality. Nonetheless, postoperative complication rates remain significantly high, with reported incidences reaching 46%[8]. To improve prognosis, adjuvant chemotherapy is routinely administered to eligible patients based on established clinical guidelines[9]. However, despite notable improvements in chemotherapeutic regimens, response rates remain suboptimal-persistently below 50%[10-12]-due primarily to the lack of precise clinical indicators and individualized therapeutic planning. The identification of non-invasive, sensitive, specific, and durable biomarkers is therefore urgently required to optimize postoperative chemotherapy evaluation. Such biomarkers would enhance the selection of patients most likely to obtain benefit from chemotherapy and support the development of alternative strategies for those demonstrating resistance, thereby advancing precision oncology in GC management.

An increasing volume of evidence implicates the tumor microenvironment (TME) in driving tumor progression, metastasis, and resistance to chemotherapy[13-17]. Chemotherapeutic interventions within the TME enhance inflammatory responses in a bidirectional manner, thereby contributing to these oncogenic processes[18]. Inflammation associated with malignancy and systemic inflammatory activity in the TME strongly correlate with enhanced tumor growth and reduced therapeutic efficacy[19,20]. GC, in particular, is commonly linked to persistent inflammatory states[21]. In inflammation-driven tumors, immune cell dynamics substantially influence disease advancement, especially at sites of sustained inflammation[22]. Neutrophils, monocytes, platelets, and lymphocytes-key hematological mediators of inflammation-are quantifiable through standard serological assays[23]. These cellular elements orchestrate pro- and anti-inflammatory signaling, modulate immune surveillance, and participate in coagulative pathways, collectively shaping tumor development and clinical trajectories across solid malignancies[22]. In GC, unfavorable outcomes are frequently associated with elevated platelet counts, increased circulation of immature platelet forms, and disruptions in neutrophil-to-lymphocyte ratios (NLR)[24,25].

Several pre-treatment inflammatory and tumor biomarkers-such as NLR, platelet-to-lymphocyte ratio (PLR), carbohydrate antigen (CA) 199, and carcinoembryonic antigen (CEA)-have been established as prognostic indicators for tumor progression[26]. Among these, NLR and PLR exhibit greater sensitivity and provide a more accurate depiction of the TME and immune balance than traditional serum inflammatory indices[27]. Advances in elucidating the interplay between inflammation and malignancy have prompted increasing attention toward the integrative use of inflammatory and tumor markers. This synergistic strategy demonstrates enhanced prognostic efficacy for survival prediction compared to the isolated evaluation of either marker category[28]. In light of this, the integrated assessment of inflammatory and tumor markers emerges as a more robust indicator of the TME and a potentially more informative tool for assessing chemotherapeutic response. Therefore, patients with a high NLR exhibit a certain degree of resistance to chemotherapy, while CEA serves as a crucial biomarker for the early detection of GC[29]. Based on extensive clinical experience, we have observed that CEA and NLR levels remain more stable within the same patient. Preliminary statistical analyses indicate that the differences in the efficacy of postoperative adjuvant chemotherapy for GC patients are more pronounced in subgroup of NLR and CEA. Furthermore, there is a notable scarcity of studies exploring the combined prognostic value of inflammatory and tumor markers in the context of postoperative chemotherapy for GC. This study, therefore, aims to evaluate the role of NLR in conjunction with CEA in determining the efficacy of postoperative chemotherapy for GC, with the objective of identifying the patient population most likely to benefit from such treatment.

MATERIALS AND METHODS

Data acquisition

Numerous investigations into postoperative adjuvant chemotherapy have employed overall survival (OS) as the primary endpoint[30-32]. In the present analysis, the focus shifts to a composite metric-NLR-CEA-as the principal variable, with five-year OS serving as the definitive outcome measure. The objective was to evaluate differential responses to postoperative adjuvant chemotherapy across subgroups delineated by the NLR-CEA index. A retrospective cohort was constructed using a comprehensive array of clinicopathological and demographic variables extracted from the institutional database. This included patient age, sex, tumor anatomical site, histopathological classification, and clinical tumor, node, metastasis (TNM) stage, categorized according to the seventh edition of the American Joint Committee on Cancer TNM classification published in 2010[33]. Five-year survival data were obtained via the hospital’s follow-up system. All variables were anonymized prior to data processing and were compiled into Microsoft Excel 2013 (Microsoft Corporation, Redmond, WA, United States) for statistical analysis and prognostic index derivation.

Study population

This single-center retrospective cohort study evaluated data from patients undergoing curative-intent gastrectomy between June 2017 and June 2018 at Shanxi Provincial Cancer Hospital (Chinese Academy of Medical Sciences Cancer Hospital Shanxi Hospital) in Taiyuan, China. Eligibility criteria included patients aged over 18 years, with histologically confirmed gastric adenocarcinoma, and having received radical gastrectomy for GC. Exclusion criteria comprised any prior history of malignancies or multiple primary tumors, medical conditions influencing peripheral blood cell counts or CEA levels, early loss to follow-up resulting in incomplete NLR-CEA and chemotherapy records, and diagnoses of GC subtypes other than adenocarcinoma. A total of 1512 patients satisfied these criteria (Figure 1). Peripheral blood samples were collected at the time of admission. Despite the retrospective nature of the study rendering it non-interventional, approval was granted by the Ethics Committee of Shanxi Cancer Hospital (Approval No: KY-2024131). Additional informed consent beyond routine admission protocols was not required. The study adhered to the principles outlined in the Declaration of Helsinki (1975, revised 2008).

Figure 1  Flowchart depicting the inclusion and exclusion criteria for the study population.

NLR, PLR, CEA, CA242, CA724, and CA199

All hematological evaluations adhered to standardized protocols within a single institution, maintaining uniformity in instrumentation, reference thresholds, and analytical integrity. Routine analyses included quantification of neutrophils, lymphocytes, monocytes, platelets, as well as serum levels of CEA and CA199. Inflammatory and oncological biomarkers were systematically obtained upon patient admission.

NLR and PLR were derived by calculating the ratios of absolute neutrophil to lymphocyte counts and absolute platelet to lymphocyte counts, respectively. To categorize NLR, PLR, CEA, CA242, CA724, and CA199 effectively, receiver operating characteristic (ROC) curves were applied to determine optimal cutoff values, which were established using the maximal Youden index. The identified cut-off points were as follows: NLR (2.18), PLR (128.71), CEA (2.72 ng/mL), CA242 (10.15), CA724 (2.56), and CA199 (31.65) (Figure 2). Based on these thresholds, the biomarkers were classified into two groups for further analysis: NLR: < 2.18 or ≥ 2.18, PLR: < 128.71 or ≥ 128.71, CEA: < 2.72 or ≥ 2.72 ng/mL, CA242: < 10.15 or ≥ 10.15, CA724: < 2.56 or ≥ 2.56, and CA199: < 31.65 or ≥ 31.65. Furthermore, GCs were classified into three groups according to the NLR-CEA scoring system: NLR-CEA = 0: NLR < 2.18 and CEA ≥ 2.72 ng/mL, NLR-CEA = 1: NLR < 2.18 and CEA < 2.72 ng/mL or NLR≥ 2.18 and CEA ≥ 2.72 ng/mL and NLR-CEA = 2: NLR ≥ 2.18 and CEA < 2.72 ng/mL.

Figure 2 Receiver operating characteristic curves of receiver operating characteristic, platelet-to-lymphocyte ratio, carcinoembryonic antigen, carbohydrate antigen 199, carbohydrate antigen 242, and carbohydrate antigen 724. NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; CEA: Carcinoembryonic antigen; CA: Carbohydrate antigen.

Statistical analysis

Patients were stratified according to the administration of postoperative adjuvant chemotherapy. Categorical variables were summarized as frequencies or percentages. Comparative analyses between the “chemotherapy” and “no chemotherapy” cohorts employed the χ² test for categorical variables, normality testing for continuous data, and the Mann-Whitney U test for non-normally distributed variables (Table 1).

Table 1 Baseline data table for research variables grouped by chemotherapy, n (%).

Variables	Total (n = 1512)	No chemotherapy (n = 841)	Chemotherapy (n = 671)	P value
Gender				0.279
Female	282 (18.7)	165 (19.6)	117 (17.4)
Male	1230 (81.3)	676 (80.4)	554 (82.6)
Old, mean ± SD	59.8 ± 9.9	61.6 ± 9.8	57.5 ± 9.4	< 0.001
Smoke				0.459
No smoking	816 (54.0)	461 (54.8)	355 (52.9)
Long term smoking	696 (46.0)	380 (45.2)	316 (47.1)
Drink				0.421
No drinking	1268 (83.9)	711 (84.5)	557 (83)
Long term drinking	244 (16.1)	130 (15.5)	114 (17)
Glycuresis				0.349
No glycuresis	1407 (93.1)	778 (92.5)	629 (93.7)
Glycuresis	105 (6.9)	63 (7.5)	42 (6.3)
Hypertension				0.206
No hypertension	1199 (79.3)	657 (78.1)	542 (80.8)
Hypertension	313 (20.7)	184 (21.9)	129 (19.2)
Diseases				0.828
No diseases	1397 (92.8)	774 (92.7)	623 (93)
Cardiovascular and cerebrovascular diseases	108 (7.2)	61 (7.3)	47 (7)
Location				0.263
Proximal	826 (55.5)	444 (53.8)	382 (57.7)
Middle	282 (19.0)	159 (19.2)	123 (18.6)
Distal	380 (25.5)	223 (27)	157 (23.7)
Lauren				0.148
Intestinal	252 (32.1)	139 (34.9)	113 (29.2)
Diffuse	258 (32.9)	120 (30.2)	138 (35.7)
Mixed	275 (35.0)	139 (34.9)	136 (35.1)
Histological grade				0.005
Moderately	384 (28.6)	232 (32.2)	152 (24.4)
Moderately to highly	385 (28.6)	200 (27.8)	185 (29.6)
Highly	574 (42.7)	287 (39.9)	287 (46)
AJCC stage				< 0.001
I	259 (18.4)	230 (30.4)	29 (4.4)
II	325 (23.0)	161 (21.3)	164 (25)
III	778 (55.1)	347 (45.9)	431 (65.8)
IV	48 (3.4)	17 (2.2)	31 (4.7)
Invasion depth T				< 0.001
I	212 (15.0)	188 (24.7)	24 (3.7)
II	120 (8.5)	82 (10.8)	38 (5.8)
III	424 (30.0)	200 (26.3)	224 (34.3)
IV	658 (46.5)	290 (38.2)	368 (56.3)
Vascular invasion				< 0.001
No invasion	772 (54.2)	488 (63.2)	284 (43.5)
Invasion	653 (45.8)	284 (36.8)	369 (56.5)
Perineural invasion				< 0.001
No invasion	827 (58.0)	502 (65)	325 (49.7)
Invasion	599 (42.0)	270 (35)	329 (50.3)
Surgical type				< 0.001
Laparoscope surgery	532 (35.2)	343 (40.8)	189 (28.2)
Traditional surgery	937 (62.1)	464 (55.2)	473 (70.6)
Non radical surgery	41 (2.7)	33 (3.9)	8 (1.2)
Neutrophils, mean ± SD	3.9 ± 1.7	3.9 ± 1.7	4.0 ± 1.7	0.207
NLR	2.3 ± 1.6	2.3 ± 1.6	2.3 ± 1.5	0.651
PLR	153.2 ± 73.8	148.7 ± 71.8	158.8 ± 75.8	0.008
CEA	6.7 ± 17.9	5.7 ± 16.0	8.0 ± 19.9	0.017
CA199	41.3 ± 88.2	38.2 ± 88.1	45.0 ± 88.1	0.154
CA50	24.9 ± 109.2	28.3 ± 129.0	21.1 ± 82.1	0.284
CA724	13.9 ± 36.7	13.7 ± 37.1	14.1 ± 36.2	0.808
CA242	24.3 ± 63.0	20.7 ± 57.7	28.4 ± 68.5	0.025
TPS	126.3 ± 153.7	127.7 ± 145.9	124.8 ± 162.2	0.762
AFP	90.5 ± 1774.9	14.8 ± 163.4	178.4 ± 2602.0	0.09
State				0.011
Survival	770 (50.9)	453 (53.9)	317 (47.2)
Death	742 (49.1)	388 (46.1)	354 (52.8)

NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; CEA: Carcinoembryonic antigen; CA: Carbohydrate antigen; TPS: Tissue polypeptide specific antigen; AFP: Alpha-fetoprotein.

Baseline characteristics (Table 1, P < 0.05), Cox univariate regression outcomes (Table 2, OR > 1.1 or < 0.9), and clinical assessment collectively informed the identification of confounders, including historical grade, AJCC TNM stage, vascular invasion, perineural invasion, and depth of tumor infiltration (T). These covariates were incorporated into a stratified Cox regression model (Table 3, Figure 3) to control for potential biases and enhance the reliability of the association estimates.

Figure 3 Forest plot illustrating Cox subgroup analysis results. NLR: Neutrophil-to-lymphocyte ratio; CEA: Carcinoembryonic antigen.

Table 2 Cox single factor regression analysis.

Variables	HR (95%CI)	P value
Gender: Male vs female	1.05 (0.84, 1.31)	0.668
Old	1.02 (1.01, 1.03)	< 0.001
Smoke: No smoking vs long term smoking	1.0006 (0.8662, 1.1559)	0.994
Drink: No drinking vs long term drinking	0.92 (0.76, 1.13)	0.438
Glycuresis: Glycuresis vs No Glycuresis	0.95 (0.72, 1.27)	0.736
Hypertension: Hypertension vs No Hypertension	0.96 (0.8, 1.15)	0.65
Diseases: Cardiovascular and cerebrovascular diseases vs no diseases	0.95 (0.72, 1.27)	0.748
Location: Ref		0.408
Proximal
Middle	1.01 (0.83, 1.23)	0.913
Distal	0.89 (0.75, 1.07)	0.214
Lauren: Ref		< 0.001
Intestinal
Diffuse	1.84 (1.43, 2.36)	< 0.001
Mixed	1.57 (1.22, 2.03)	< 0.001
Histological grade: Ref
Moderately		< 0.001
Moderately to highly	1.71 (1.38, 2.12)	< 0.001
Highly	1.9 (1.55, 2.31)	< 0.001
AJCC stage: Ref
I		< 0.001
II	2.41 (1.65, 3.5)	< 0.001
III	7.35 (5.27, 10.27)	< 0.001
IV	12.07 (7.69, 18.96)	< 0.001
Invasion depth T: Ref		< 0.001
I
II	1.85 (1.11, 3.1)	0.019
III	4.88 (3.31, 7.19)	< 0.001
IV	7 (4.8, 10.21)	< 0.001
Vascular invasion: Invasion vs no invasion	3.03 (2.59, 3.54)	< 0.001
Perineural invasion: Invasion vs no invasion	2.5 (2.15, 2.91)	< 0.001
Surgical type: Ref
Laparoscope surgery		< 0.001
Traditional surgery	1.21 (1.03, 1.42)	0.018
Non radical surgery	8.1 (5.76, 11.38)	< 0.001
Postoperative chemotherapy: Chemotherapy vs no chemotherapy	1.13 (0.98, 1.3)	0.103
CEA	1.01 (1.01, 1.01)	< 0.001
CA199	1.0029 (1.0022, 1.0036)	< 0.001
CA50	1.0021 (1.0015, 1.0026)	< 0.001
CA724	1.0075 (1.006, 1.009)	< 0.001
CA242	1.0034 (1.0025, 1.0044)	< 0.001
TPS	1.001 (1.0006, 1.0014)	< 0.001
AFP	1 (1, 1.0001)	0.001
NLR	1.1 (1.06, 1.13)	< 0.001
PLR	1.003 (1.0022, 1.0038)	< 0.001

NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; CEA: Carcinoembryonic antigen; CA: Carbohydrate antigen; TPS: Tissue polypeptide specific antigen; AFP: Alpha-fetoprotein.

Table 3 Cox subgroup analysis of postoperative neoadjuvant chemotherapy among neutrophil-to-lymphocyte ratio, carcinoembryonic antigen and neutrophil-to-lymphocyte ratio-carcinoembryonic antigen combined indicators.

	Variable	Total	Event, n (%)	Follow up (time)	HR (95%CI)	P value
NLR_Low
	No chemotherapy	441	165 (37.4)	20024	1 (Ref)
	Chemotherapy	335	156 (46.6)	14443	0.64 (0.51-0.82)	0
NLR_High
	No chemotherapy	349	191 (54.7)	12575	1 (Ref)
	Chemotherapy	304	180 (59.2)	11268	0.71 (0.56-0.89)	0.004
CEA_Low
	No chemotherapy	252	149 (59.1)	8847	1(Ref)
	Chemotherapy	210	125 (59.5)	7660	0.64 (0.49-0.83)	0.001
CEA_High
	No chemotherapy	538	207 (38.5)	23752	1 (Ref)
	Chemotherapy	429	211 (49.2)	18051	0.72 (0.59-0.9)	0.003
NLR_CEA_ Low
	No chemotherapy	83	52 (62.7)	2832	1 (Ref)
	Chemotherapy	97	47 (48.5)	3872	0.41 (0.26-0.63)	0
NLR_CEA_ Middle
	No chemotherapy	387	154 (39.8)	16780	1 (Ref)
	Chemotherapy	318	167 (52.5)	13109	0.69 (0.54-0.87)	0.002
NLR_CEA_ High
	No chemotherapy	231	111 (48.1)	9089	1 (Ref)
	Chemotherapy	199	105 (52.8)	7844	0.73 (0.53-0.99)	0.042

NLR: Neutrophil-to-lymphocyte ratio; CEA: Carcinoembryonic antigen.

ROC analysis was applied to establish optimal thresholds for NLR, PLR, CEA, CA199, CA724, and CA242 (Figure 2, Table 4). Each point on the ROC curve corresponds to a specific threshold, visually representing the area under the curve. This curve balances sensitivity and specificity, while maximizing the Youden index to identify the optimal cutoff values for each indicator. Based on these values, a composite indicator, “NLR-CEA,” was generated.

Table 4 Area under the curve.

Variable	AUC (95%CI) (%)	Specificity	Sensitivity
CEA	59.7769 (56.7646-62.7892)	0.7885	0.3729
CA199	60.6168 (57.5979- 63.6356)	0.8878	0.3098
NLR	58.9996 (55.9899-62.0094)	0.6561	0.4872
PLR	58.89 (55.8704-61.9096)	0.5223	0.6256
CA242	59.5793 (56.5495-62.609)	0.836	0.3549
CA724	63.0845 (60.1236-66.0455)	0.659	0.5714

AUC: Area under the curve; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; CEA: Carcinoembryonic antigen; CA: Carbohydrate antigen.

Covariate-adjusted Cox stratified models were subsequently employed to compare the therapeutic impact of postoperative adjuvant chemotherapy across stratified NLR, CEA, and NLR-CEA groups (Table 3, Figure 3).

Statistical computations were conducted using R software (version 3.3.2, The R Foundation, http://www.R-project.org) and Free Statistics software (version 1.1.8). A two-sided test was utilized, with statistical significance defined as P < 0.05. Continuous variables were expressed as mean ± SD, and categorical data were reported as counts and proportions. A HR < 1 accompanied by a 95% confidence interval excluding 1 indicated a prognostic benefit of postoperative adjuvant chemotherapy for GC.

RESULTS

A total of 1512 participants were included in the final analysis (Figure 1). Baseline characteristics, stratified by chemotherapy status, were presented in Table 1. The mean age was 59.8 ± 9.9 years, with males comprising approximately 81.3% of the cohort (Table 1). Variables exhibiting statistical significance (P < 0.05) included age, histological grade, AJCC TNM stage, vascular invasion, perineural invasion, invasion depth, and surgical type (Table 1). Additionally, covariates demonstrating an OR > 1.1 or < 0.9 comprised histological grade, AJCC TNM stage, vascular invasion, perineural invasion, invasion depth, and surgical type (Table 2). Integrating statistical results with clinical rationale, histological grade, AJCC TNM stage, vascular invasion, perineural invasion, and invasion depth were designated as confounders for subsequent adjustment.

ROC curve analysis identified optimal threshold values for NLR, PLR, CEA, CA242, CA724, and CA199 as 2.18, 128.71, 2.71, 10.14, 2.5, and 31.65, respectively. A composite metric, NLR-CEA, was constructed using NLR and CEA cutoffs: NLR-CEA = 0 (NLR < 2.18 and CEA ≥ 2.72 ng/mL), NLR-CEA = 1 (either NLR < 2.18 with CEA < 2.72 ng/mL or NLR ≥ 2.18 with CEA ≥ 2.72 ng/mL), and NLR-CEA = 2 (NLR ≥ 2.18 and CEA < 2.72 ng/mL) (Figure 2, Table 4).

Cox subgroup analysis, adjusted for relevant covariates, revealed that postoperative adjuvant chemotherapy conferred a survival benefit across all NLR-CEA stratifications. HRs for OS were 0.41 (0.26-0.63) in the NLR-CEA_Low group, 0.69 (0.54-0.87) in the NLR-CEA_Middle group, and 0.73 (0.53-0.99) in the NLR-CEA_High group. A 32% differential in chemotherapy efficacy emerged between the NLR-CEA_Low and NLR-CEA_High groups-markedly exceeding the variations noted within the NLR (7%) and CEA (8%) subgroups (Table 3, Figure 3).

DISCUSSION

To date, no investigations have assessed chemotherapeutic efficacy based on the integrated evaluation of systemic inflammatory and tumor markers. The observed 32% disparity in postoperative chemotherapy effectiveness between the NLR-CEA_High and NLR-CEA_Low cohorts demonstrates its clinical applicability. Moreover, neutrophil, lymphocyte, and CEA levels are routinely obtained during baseline assessments. Given this accessibility, NLR-CEA may serve as a dependable biomarker for informing adjuvant chemotherapy decisions following GC resection, thereby contributing to the refinement of individualized postoperative treatment planning.

The TME constitutes a specialized niche integral to cancer cell survival and proliferation[33]. Its influence on clinical outcomes has been substantiated across multiple studies[34]. Both tumor-associated and systemic inflammatory responses are implicated in promoting tumor progression, facilitating metastasis, and inducing resistance to chemotherapy, thereby negatively affecting OS[13-17,19]. In clinical serological evaluations, neutrophils and lymphocytes serve as accessible biomarkers of inflammation. Circulating neutrophil levels represent the systemic inflammatory state, whereas lymphocytes, as central mediators of immune surveillance, are involved in direct tumor eradication. The NLR, a composite index reflecting the relative proportions of these two cell types, offers an indirect estimation of immune competence within oncologic and inflammatory contexts. This ratio captures the dynamic interplay between pro-tumor inflammatory processes and anti-tumor immune activity, enabling prognostic stratification. Elevated NLR indicates a lymphocyte-to-neutrophil disequilibrium suggestive of TME disruption[27]. The current analysis corroborates this association, indicating diminished sensitivity to postoperative adjuvant chemotherapy among patients with higher NLR values compared to those with lower NLR (Table 2). This trend supports the role of NLR as a potential indicator of chemotherapeutic response and disease trajectory.

CEA serves as an established biomarker for early GC detection[35] and provides partial insight into TME dynamics. Increasing evidence has evaluated the prognostic utility of combining systemic inflammatory markers with tumor markers in GC, indicating that composite indices deliver enhanced sensitivity and specificity over isolated parameters[28]. This integrative strategy is increasingly applied in clinical assessment frameworks. Integrating NLR, a pretreatment inflammatory indicator, with CEA more accurately captures TME status than using either metric independently. The present findings reinforce this premise. In subgroup analyses based on NLR-CEA classification, the benefit of postoperative adjuvant chemotherapy in the NLR-CEA_Low cohort surpassed that in the NLR-CEA_High cohort by 32% (Table 4), a markedly greater margin compared to those observed in the NLR (7%) and CEA (8%) subgroups (Table 3, Figure 3).

Patients with postoperative GC exhibiting low NLR-CEA levels consistently demonstrate favorable responses to standard adjuvant chemotherapy, reflected in superior therapeutic efficacy and prolonged survival outcomes[9]. Conversely, elevated NLR-CEA levels-associated with a more immunologically active TME-correlate with increased resistance to chemotherapy, diminishing its effectiveness in this subgroup. For such patients, therapeutic strategies beyond conventional adjuvant regimens warrant consideration. Recent studies have also suggested that assuming Anti-inflammatory and targeted therapies that modulate inflammatory pathways-such as WNT-β-catenin and others-could offer promising options to maximize OS[36-39]. Stratifying therapeutic approaches based on TME profiles may enhance treatment specificity and improve prognostic outcomes in GC management.

Limitations of this study

The current study has certain limitations: (1) This study exclusively examined GCs, excluding other tumor subtypes such as gastric neuroendocrine carcinoma and gastric squamous cell carcinoma; (2) The research primarily focused on evaluating the role of NLR-CEA in the context of adjuvant chemotherapy following GC surgery, aiming to identify the most suitable patient population for postoperative adjuvant chemotherapy. While patients with elevated NLR-CEA levels generally show poorer responses to adjuvant chemotherapy, alternative treatment strategies that may be more effective for this subgroup remain undetermined. Further prospective studies are needed to explore and validate these alternatives; (3) As a retrospective observational study with a large sample size (n = 1628), variations in adverse reactions, drug resistance, economic factors, and drug sensitivity influenced chemotherapy regimen selection. Treatment plans, including drug choice and duration, were tailored based on clinical guidelines. Consequently, while this study provides insights into the overall efficacy of adjuvant chemotherapy after GC surgery, it does not offer a detailed evaluation of the efficacy of individual chemotherapy drugs or regimens; (4) Due to the inherent limitations of retrospective studies, this research may be subject to biases, including those related to chemotherapy selection and the variability in follow-up quality; and (5) Due to the use of Cox subgroup regression analysis as the statistical method in this study, further investigation into the interactions between the results and each subgroup is not feasible.

CONCLUSION

NLR-CEA can serve as a clinically informative biomarker for assessing postoperative chemotherapy responsiveness in GC patients. Individuals exhibiting lower NLR-CEA levels tend to derive greater therapeutic benefit from adjuvant chemotherapy after surgical resection compared to those with elevated levels.

ACKNOWLEDGEMENTS

We thank Jie Liu, PhD (Department of Vascular and Endovascular Surgery, Chinese PLA General Hospital) for his helpful review and comments regarding the manuscript.