Combined inflammation/nutrition-based prognostic score for overall survival in gastric adenocarcinoma: A retrospective cohort study

Abstract

BACKGROUND

Gastric cancer is the fifth most common malignancy and the fourth leading cause of cancer-related death. The tumor-node-metastasis (TNM) staging system, while fundamental, does not account for systemic inflammation and nutritional status, which impact survival.

AIM

To assess the prognostic value of a novel inflammation/nutrition-based combined score (INCS) for predicting overall survival in patients undergoing curative gastrectomy for gastric adenocarcinoma.

METHODS

This retrospective cohort study included 154 consecutive patients (2019-2022). INCS integrated neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, prognostic nutritional index, and prognostic index, categorized as low (0-3) or high (4-6). Overall survival was analyzed using Kaplan-Meier, log-rank, and multivariable Cox regression.

RESULTS

Median age was 68 years (interquartile range: 61-76), 62.3% were male. During follow up (median 25.1 months), 61.0% died. High INCS was associated with higher mortality (87.9% vs 53.7%, P < 0.001) and advanced TNM stage (P < 0.001). In multivariable analysis without TNM stage, INCS independently predicted mortality (hazard ratio = 1.20, 95%CI: 1.04-1.39, P = 0.013). The TNM + INCS model showed a higher area under the curve than TNM alone (0.805 vs 0.787), with no statistically significant difference.

CONCLUSION

INCS is a simple, and potentially useful prognostic tool for gastric cancer patients undergoing curative gastrectomy. Combined with TNM staging, it may enhance risk stratification and guide postoperative management.

Key Words: Adenocarcinoma; Inflammation; Nutritional status; Prognosis; Stomach neoplasms; Survival analysis

Core Tip: This retrospective cohort study found that the inflammation/nutrition-based combined score was associated with overall survival in patients undergoing curative gastrectomy for gastric adenocarcinoma. When added to tumor-node-metastasis staging, inflammation/nutrition-based combined score may improve prognostic accuracy. The score reflects systemic inflammation and nutritional status, offering complementary information to tumor anatomy. While these findings are promising, they require validation in larger, multicenter studies before clinical application.

INTRODUCTION

Gastric cancer (GC) continues to pose one of the most serious threats to the world’s health community because it ranks the fifth most common malignancies globally and the fourth leading death-related cancers[1]. Despite the progressive improvement in the methods used for its detection and therapy, the outcome for its patients also appears dismal in the advanced stages of the malignancy[2].

Presently, the tumor-node-metastasis (TNM) system of staging, which primarily reflects the anatomical extent of disease, is the most common approach being used to predict the prognosis of GC. Nevertheless, it fails to take individual variations in prognosis within the same stage into consideration. This indicates the existence of other host-related biological parameters influencing the outcome of the condition[3].

Systemic inflammation also contributes importantly to the tumor process. Inflammatory reactions may favor the tumor through the production of cytokines, angiogenesis, and the suppression of anti-tumoral immunity[4,5]. Thus, some inflammation-based markers like the neutrophil-to-lymphocyte ratio (NLR), the platelet-to-lymphocyte ratio (PLR), the prognostic nutritional index (PNI), and the prognostic index (PI) had all been recognized as prognostic predictors for various malignancies, including GC[6]. Nevertheless, the prognostic ability of each index used individually was not ideal. Thus, the most effective way of combining those markers in predicting the outcome in patients with GC is yet unclear.

In the given context, the use of an integrated PI combining various inflammation-related parameters along with nutrition markers could prove to be an effective approach[5,7]. Although the prognostic significance of the integrated index in the context of various malignancies like solid cancers has been studied previously, it is still in its initial stages in the case of gastric malignancies. Various other scoring systems for the prognosis of malignancies are also being developed in attempts to improve prognosis prediction[8-11].

Thus, the aim of the current retrospective cohort study was to determine the ability of the inflammation/nutrition-based combined score (INCS) in predicting the overall survival in gastric adenocarcinoma patients undergoing curative gastrectomy. Unlike single-parameter indices such as systemic immune-inflammation index (SII) or PNI, the INCS integrates multiple inflammatory and nutritional pathways, potentially providing a broader representation of the host systemic response to tumor progression. Our hypothesis was that patients who received higher scores on the combination prognostic system would also exhibit higher levels of clinicopathological parameters along with poor overall survival. By addressing the aforementioned challenge in the field, we aim to pave the way for the future utilization of a simple prognostic system.

MATERIALS AND METHODS

Study design and participants

It was a retrospective cohort study involving 181 patients who had curative surgery for gastric adenocarcinoma at a tertiary institute in March 2019 through to October 2022. However, the study excluded patients who had benign histopathology results, patients for whom crucial data was incomplete, and those who had the surgery during an emergency. Thus, the study used 154 patients. Data was obtained from the patients’ electronic records. Preoperative values for the parameters involved for each patient involved in the study had been measured in venous blood samples one week prior to the operations.

Ethical considerations and power analysis

This study was evaluated for ethics by the Ethics Committee of Ankara City Hospital (No. E1/3303/2023). Initially, an a priori power calculation was undertaken to determine the minimum patient population needed to assess the difference in the rate of overall survival between low (0-3) and high (4-6) INCS use. Using an α level of 0.05 in the two-tailed test, 80% power, along with the probability of event rate of 60%, the log rank test (Schoenfeld’s method) showed that 150 patients would provide adequate results. Using this estimation, the period studied was defined to encompass the time frame during which at least the same number of consecutive patients could be retrieved from the institutional database. In order to minimize bias, consecutive patients were selected, and preoperative lab results were measured using consistent criteria. Moreover, multivariate analysis was conducted to account for any possible confounders.

Clinical and pathological assessments

Staging of all cancers was performed according to the 8^th American Joint Committee on Cancer/Union for International Cancer Control TNM system in compliance with the American Joint Committee on Cancer/Union for International Cancer Control Cancer Staging Manual based on the final pathological stage[12]. Participants in the study comprised patients aged 18-90 years old diagnosed histopathologically for gastric adenocarcinoma undergoing curative surgery. Additionally, in the study, the administration of neoadjuvant therapy was noted for all patients. Also in the study, survival was evaluated by follow-up documentation until May 1, 2025. Since the mortality rates in the study were confirmed through official death certificates, no patients were lost to follow-up.

Definition of the INCS

The INCS was developed to integrate systemic inflammatory response and nutritional status into a single composite index. The included parameters were selected based on their established roles in cancer-related inflammation and immunonutritional status.

NLR and PLR were used to reflect the systemic inflammatory response and immune balance. PI, incorporating C-reactive protein and leukocyte counts, represents acute-phase inflammation. PNI, calculated using albumin and lymphocyte counts, reflects both nutritional and immune status.

Calculations for the NLR, PLR, PI, and PNI allowed us to assess the inflammation levels and nutritional condition of the patients, and a PI integrating both these inflammation and nutrition markers was developed and referred as to INCS. The score was analyzed both as an ordinal variable (0-6) and as a dichotomized variable (low: 0-3 vs high: 4-6). While the ordinal analysis demonstrated a stepwise association with survival, the dichotomized model was preferred for its greater clinical interpretability and ease of application in routine practice (0-3 = low INCS group, 4-6 = high INCS group). Distribution of points assigned to each variable for the INCS are given in the Table 1. An additive model with equal weighting was used to maintain simplicity and clinical applicability, and to reduce the risk of overfitting in this retrospective cohort. The cutoff values for NLR, PLR, and PNI were selected based on findings from previously published meta-analyses[13-15]. To better reflect the importance of nutritional status within the combined score, an additional point was assigned to the PNI component, thereby preventing it from being underweighted compared with inflammatory markers.

Table 1 Scoring criteria for inflammation/nutrition-based combined score.

Marker	Cut-off definition	Assigned score
NLR	≥ 2.75	1
PLR	≥ 178	1
PNI	40–45	1
	< 40	2
PI	CRP > 1.0 mg/dL or WBC > 11000/mm3	1
	CRP > 1.0 mg/dL and WBC > 11000/mm3	2

CRP: C-reactive protein; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; PNI: Prognostic nutritional index; PI: Prognostic index; WBC: White blood cell count.

Statistical analysis

Patients were stratified by stage for subgroup analysis. TNM stages were categorized as follows: (1) Stage I (early disease); (2) Stage II (local disease); (3) Stage III (locally advanced disease); and (4) Stage IV (metastatic disease). Multivariable Cox regression models were constructed to assess independent prognostic factors affecting mortality. Two models were created after univariate evaluations: (1) The first included the INCS, age, perineural invasion, neoadjuvant treatment, and surgical type; and (2) In the second model, TNM stage was added as an ordinal variable (1-4). Multicollinearity was checked using variance inflation factors (VIF), with VIF > 2.5 considered indicative of problematic collinearity.

Statistical analyses were performed using SPSS version 20.0 (IBM Corp., Armonk, NY, United States) and R software (version 4.3.3). Overall survival was defined as the primary outcome variable and calculated from the date of surgery to death or the last follow-up. Categorical variables (e.g., sex, stage group, neural invasion, differentiation, treatment status) were expressed as frequencies and percentages, and compared using the χ² test. Fisher’s exact test was applied when the expected cell count was less than 5. Continuous variables were assessed for normality using the Shapiro-Wilk test. As most variables were not normally distributed, they were expressed as medians [interquartile range (IQR)] and compared using the Mann-Whitney U test for two-group comparisons and the Kruskal-Wallis test for comparisons involving more than two groups. Survival analyses were performed using the Kaplan-Meier method, and differences between groups were evaluated using the log-rank test. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the ability of each marker to predict mortality. Area under the curve (AUC) values were calculated, and pairwise comparisons between AUCs were conducted using DeLong test. ROC curves and pairwise comparisons are presented in Supplementary Figure 1 and Supplementary Table 1.

A multivariable Cox proportional hazards regression model was used to identify independent predictors of overall survival. Variables included in the model were selected based on clinical relevance and the results of univariate analyses. Hazard ratios (HRs) with 95%CI were reported. Decision curve analysis (DCA) was performed using the rmda package in R to evaluate the net clinical net benefit of different prognostic models. The proportional hazards assumption was assessed using Schoenfeld residuals and no significant violation was detected.

All analyses were performed using complete-case data, and any observation with a missing value was discarded. The significance level was set at α = 0.05 for all statistical analyses.

RESULTS

Patients’ characteristics and follow-up

Of 181 initially screened patients who underwent curative-intent gastrectomy for gastric adenocarcinoma, 154 patients formed the final analysis after exclusions for benign histopathology (n = 14), emergency surgery (n = 3), and incomplete crucial data (n = 10, Figure 1).

Figure 1 Patient selection flowchart. Of 181 patients screened, 154 were included in the final analysis after excluding patients with benign histopathology (n = 14), emergency surgery (n = 3), and incomplete data (n = 10). IQR: Interquartile range.

The median age was 68 years (IQR: 61-76), and 96 patients (62.3%) were male. Neoadjuvant chemotherapy was administered to 29 patients (18.8%). During the follow-up period, 94 deaths (61.0%) were recorded. The median follow-up duration for the entire cohort was 25.1 months (IQR: 5.3-41.8). Among survivors, the median follow-up reached 42.7 months (IQR: 36.8-56.9). The median survival time among patients who died during follow-up was 9.5 months (IQR: 1.3-20.9).

Baseline demographic and clinical characteristics of the cohort are presented in Table 2. After stratification of patients into low (0-3) and high (4-6) INCS groups, corresponding laboratory characteristics are shown in Table 3.

Table 2 Baseline demographic and clinical characteristics stratified by inflammation/nutrition-based combined score, median (interquartile range)/n (%).

Variable	Total (n = 154)	Low score (0-3, n = 121)	High score (4-6, n = 33)	P value
Age (years)	68 (61-76)	67 (60-74)	75 (70-83)	< 0.001a
Sex (male)	96 (62.3)	78 (64.5)	18 (54.5)	0.401
Type of surgery				0.050
Subtotal	59 (38.3)	41 (33.9)	18 (54.5)
Total	95 (61.7)	80 (66.1)	15 (45.5)
Neoadjuvant chemotherapy	29 (18.8)	25 (20.7)	4 (12.1)	0.389
Overall mortality	94 (61.0)	65 (53.7)	29 (87.9)	< 0.001a
TNM stage	n = 151			0.016a
Stage I	19 (12.6)	18 (14.9)	1 (3.0)
Stage II	45 (29.8)	39 (32.2)	6 (18.2)
Stage III	65 (43.0)	48 (39.7)	17 (51.5)
Stage IV	22 (14.6)	13 (10.7)	9 (27.3)
Perineural invasion	107/153 (69.9)	79 (65.3)	28 (84.8)	0.058
Differentiation (poor)	n = 108			0.136
Poor	48 (44.4)	36 (29.8)	12 (36.4)
Moderate	43 (39.8)	30 (24.8)	13 (39.4)
Good	17 (15.7)	16 (13.2)	1 (3.0)

^aP < 0.05.

Mann Whitney U test was used for age, and χ² was used for the categoric variables. The total number (n) of available cases for variables with missing data is indicated separately. TNM: Tumor-node-metastasis.

Table 3 Laboratory characteristics stratified by inflammation/nutrition-based combined score, median (interquartile range).

Variable	Low score (0-3)	High score (4-6)	P value
WBC (× 109/L)	6.55 (5.44-7.67)	7.29 (5.56-9.85)	0.058
Neutrophil (× 109/L)	3.88 (2.96-4.95)	5.78 (3.61-7.79)	< 0.001c
Lymphocyte (× 109/L)	1.79 (1.41-2.21)	1.02 (0.79-1.41)	< 0.001c
Platelet (× 109/L)	259 (218-315)	303 (241-373)	0.002b
CRP (mg/L)	8.1 (3.4-18.0)	29.2 (13.6-54.4)	< 0.001c
NLR	2.22 (1.68-3.01)	5.29 (4.07-8.19)	< 0.001c
PLR	150 (112.5-202.6)	284.96 (197.6-379.8)	< 0.001c
PNI	46.9 (43.45-51.15)	36.25 (31.7-38.9)	< 0.001c

^bP < 0.01.

^cP < 0.001.

Mann-Whitney U test was used for all variables. CRP: C-reactive protein; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; PNI: Prognostic nutritional index; WBC: White blood cell count.

Patients in the high INCS group were significantly older (median 75 years vs 67 years, P < 0.001) and had more advanced TNM stages (P = 0.016) compared to the low INCS group. Overall mortality was significantly higher in the high INCS group (87.9% vs 53.7%, P < 0.001). No significant differences were observed between groups regarding sex, type of surgery, neoadjuvant chemotherapy administration, perineural invasion, or tumor differentiation (Table 2).

Laboratory analysis revealed that patients with high INCS had significantly higher neutrophil counts, platelet counts, CRP levels, NLR, and PLR, while lymphocyte counts and PNI were significantly lower compared to the low INCS group (all P < 0.05; Table 3).

Survival analysis

The Kaplan-Meier analysis revealed a significant association between increasing INCS (0-6) and progressively poor survival. Further, when the score was stratified into low (0-3) and high (4-6) groups, survival remained significantly different, reflecting notably poorer outcomes for the high-score group (Figure 2). This prognostic separation persisted even after the exclusion of metastatic patients, a finding that confirmed the robustness of the score in non-metastatic disease (log-rank χ² = 18.035, P = 0.006).

Figure 2 Kaplan-Meier survival curves by inflammation/nutrition-based combined score. Left panel shows all seven score categories (0-6), and right panel compares low (0-3) vs high (4-6) score groups. Log-rank tests indicated significant survival differences in both analyses.

Because neoadjuvant therapy may influence systemic inflammatory and nutritional parameters, an additional analysis was performed exclusively in patients who did not receive neoadjuvant treatment (n = 125). In this subgroup, the INCS was associated with mortality in univariable analysis (P < 0.001). However, Kaplan-Meier analysis comparing low (0-3) and high (4-6) score groups revealed no statistically significant difference in overall survival (log-rank χ² = 2.20, P = 0.138), indicating that the prognostic value of the INCS may be primarily driven by patients who received neoadjuvant therapy. Although a trend toward poorer survival was observed in patients with higher scores, the lack of statistical significance in this subgroup raises questions about the score’s generalizability to patients undergoing upfront surgery.

Prognostic accuracy of INCS

We compared the prognostic accuracy of the INCS with other established markers individually, including the PNI, NLR, PLR, and PI. ROC analysis demonstrated that TNM stage exhibited the highest discriminative performance (AUC = 0.787, 95%CI: 0.717-0.850), followed by PNI (AUC = 0.666, 95%CI: 0.580-0.752), INCS (AUC = 0.623, 95%CI: 0.536-0.711), NLR (AUC = 0.598, 95%CI: 0.502-0.688), PI (AUC = 0.542, 95%CI: 0.471-0.607), and PLR (AUC = 0.539, 95%CI: 0.443-0.636). When evaluated individually, TNM stage had higher AUC values than the inflammation-based markers. INCS showed similar performance to several established markers and had a higher AUC than PLR (P = 0.015), while differences between INCS and PNI (P = 0.136) and NLR (P = 0.404) were not statistically significant (Supplementary Figure 1 and Supplementary Table 1). All ROC analyses were based on observed mortality status.

The predictive performance of the INCS and TNM was also assessed for when they are used together, and the combined model showed a higher AUC compared to TNM stage alone (P = 0.223, AUC = 0.805, 95%CI: 0.731-0.870). This finding is supported by the DCA, which indicates TNM resulted in the highest net benefit for a wide range of threshold probabilities (Supplementary Figure 2). Although at lower thresholds the model of the PNI showed higher net benefit, the INCS exhibited equal or better performance at intermediate and higher thresholds.

Subgroup analysis

Subgroup analyses supported the prognostic relevance of the INCS. Higher INCS was significantly associated with poorer survival in both subtotal and total gastrectomy patients. Survival rates did not differ between the two types of surgery (P = 0.406). Regarding the early, local, and locally advanced disease stages, the INCS was consistently able to separate survival curves (P < 0.001). Its prognostic effect disappeared in metastatic disease as the stage dominated the outcome. Among patients receiving neoadjuvant treatment, higher INCS was associated with poorer survival (P = 0.014).

Multivariable Cox regression analysis

To assess the independent prognostic value of the INCS, multivariable Cox regression analysis was carried out. The variables included in the model were INCS as an ordinal variable, age, perineural invasion, neoadjuvant treatment status, and type of surgery. In this initial model, INCS remained an independent predictor of mortality; every one-unit increase in the INCS represented a 20% increased risk of mortality (HR = 1.20, 95%CI: 1.04-1.39, P = 0.013). Perineural invasion (HR = 2.58, 95%CI: 1.45-4.57, P = 0.001) and neoadjuvant therapy (HR = 2.26, 95%CI: 1.31-3.88, P = 0.003) were also significant predictors, whereas age (P = 0.250) and type of surgery (P = 0.845) did not show independent associations with overall survival. When we used neoadjuvant therapy as a stratification variable in this model, the association between INCS and mortality weakened and did not retain statistical significance (P = 0.067), whereas perineural invasion remained significant (P < 0.001).

A second Cox regression analysis was then performed incorporating TNM stage (ordinal, 1-4). In this expanded model, TNM stage was identified as the strongest prognostic factor (HR = 2.79), while neoadjuvant therapy remained significant and age demonstrated a modest but significant effect. The INCS showed a borderline association with mortality (HR = 1.14, P = 0.070), indicating that part of its prognostic effect overlaps with tumor stage. Perineural invasion lost statistical significance consistent with its moderate correlation with stage (Table 4). No multicollinearity or significant correlation between the variables was observed (all VIF < 2, Supplementary Figure 3).

Table 4 Multivariable Cox regression analysis including tumor-node-metastasis stage.

Variable	HR	95%CI (lower)	95%CI (upper)	P value
INCS	1.14	0.99	1.31	0.070
Age	1.02	1.00	1.05	0.030
Perineural invasion	0.96	0.51	1.83	0.910
Neoadjuvant therapy	2.72	1.56	4.74	< 0.005
TNM stage (1-4)	2.79	2.01	3.86	< 0.005
Surgical type	1.02	0.66	1.59	0.930

Inflammation/nutrition-based combined score (0-6) and tumor-node-metastasis stage (1-4) were treated as ordinal variables. Multivariable Cox proportional hazards regression analysis was used to calculate hazard ratios and 95%CI (concordance index: 0.73). HR: Hazard ratio; INCS: Inflammation/nutrition-based combined score; TNM: Tumor-node-metastasis.

DISCUSSION

The present study indicates a potential association between the INCS and overall survival in patients with gastric adenocarcinoma who underwent curative surgery. Higher scores were associated with poorer results and this trend remained clear even upon exclusion of metastatic patients.

Unlike single-parameter indices, INCS combines inflammatory and nutritional components within a single framework, allowing for a more comprehensive assessment of tumor-related inflammation and patient nutritional status. By incorporating four parameters, INCS seeks to give an improved estimate of both the systemic inflammatory response and the nutritional reserve. Previous reports have also indicated that the combination of inflammation-based markers is superior in prognostic performance to single markers. The findings by Kudou et al[7] support this principle, and our results are consistent with their observations. The modest AUC of our score in isolation and its better performance when combined with TNM staging indicate that multi-marker approaches capture biological information that single markers do not offer.

The elements of our score capture various aspects of the inflammatory and nutritional response, which modulate tumor progression. High levels of neutrophil and platelet count reflect active inflammation, whereas low lymphocytes and hypoalbuminemia suggest a poor immune response and nutrition reserve, respectively[5,16]. These combined signals would appear to explain the relationship between this score and aggressive tumor features.

Prognostic models, including those using machine learning–based systems, have also identified the advantage of integrating multiple biological and clinical markers[17]. Sekiguchi et al[11] recently demonstrated how integrated scoring systems could identify prognostic subgroups not achievable with conventional staging alone. Indeed, a number of studies incorporating imaging[18], immune profiling[19], or molecular signatures[20] have also identified multi-parameter models that enhance prognostic accuracy in GC. These methods are more complex than our blood-based score but support the general principle that combining disparate biological signals yields a clearer perspective on patient risk. Our findings are in agreement with this principle, and may offer a simpler and more accessible tool for routine practice.

Other inflammation-based indices, such as the modified Glasgow prognostic score and SII, have also been associated with prognosis in patients with GC[14,21,22]. Although these systems were not directly compared to our score, the conceptual similarity implies that they capture related biological pathways. Our score represents a practical alternative, combining multiple markers of inflammation and nutrition into a simple format that can be calculated preoperatively.

Importantly, our study shows that the model that combines the INCS with the TNM stage yielded a higher AUC (0.805) than TNM alone (AUC = 0.787). Therefore, our score captures valuable biological information that is complementary to the anatomical information provided by the TNM system. The clinical implication of this synergy may be the potential for enhanced risk stratification. For example, a patient who has early-stage disease according to TNM but a high INCS can be distinguished as being at greater risk and might be a candidate for closer surveillance.

Among the individual elements of our score, the PNI performed best, consistent with previous evidence supporting the role of nutritional status as a prognostic factor in GC[6,23]. The comparable performance of the INCS implies that the addition of other inflammation-based markers to the PNI offers a more global index of host response. DCA further supported the clinical utility of both the PNI and the INCS over a wide range of threshold probabilities.

The INCS demonstrated prognostic value in most subgroups, including different surgical types and disease stages. However, in patients who did not receive neoadjuvant therapy, it failed to reach statistical significance (log-rank P = 0.136). This implies that the prognostic significance of the INCS, as demonstrated in the current study, may be due to the patient population that underwent neoadjuvant treatment. This result has several implications. First, the result underscores the fact that neoadjuvant treatment may improve the prognostic significance of the inflammatory and nutritional parameters included in the INCS. Second, the result raises a concern over the discriminative ability of the INCS for the patient population that underwent upfront surgery, for which the score was not found to demonstrate prognostic significance. Similar observations have been reported in studies evaluating inflammation-based markers after neoadjuvant therapy[24]. Likewise, our findings did not show any difference in survival between subtotal and total gastrectomy. Furthermore, previous studies suggest that, in patients undergoing total gastrectomy, either a laparoscopic or an open approach does not meaningfully change long-term survival as long as an adequate D2 lymphadenectomy is performed[25].

To evaluate the independent predictive value of the INCS, two multivariable Cox regression models were developed. When the INCS was included without the TNM stage, the INCS retained its predictive value for mortality (HR = 1.20, 95%CI: 1.04-1.39, P = 0.013). This suggests that the INCS provides independent prognostic information beyond some of the selected clinicopathological variables. However, when the TNM stage was included in the model, the INCS failed to achieve statistical significance at the α = 0.05 level (HR = 1.14, 95%CI: 0.99-1.31, P = 0.070) and thus may be considered to be indirectly related to the outcome. This suggests that the prognostic information provided by the INCS might be indirectly related to the outcome via its relationship with the TNM stage. The loss of significance of the INCS when the TNM stage is included does not diminish the overall value of the INCS. When the INCS and the TNM stage are used together to predict the outcome, the INCS showed a higher AUC when combined with TNM stage (0.805 vs 0.787). Patients with the same stage by TNM classification but different INCS may have different outcomes, and the score may help identify high-risk patients within each stage group that may warrant closer surveillance or adjuvant therapy.

Age in our multivariable model showed only a modest association with mortality. This pattern is consistent with literature that has documented that, while age is a contributor to prognosis, generally its effect is smaller than tumor stage and biological factors[26]. Further, several analyses have also demonstrated variable age-related trends in survival, especially in different models of populations[27].

In our multivariable model, neoadjuvant therapy was found in association with increased mortality. This finding should be interpreted cautiously because the study was not designed to assess the effect of neoadjuvant treatment. Pathological treatment may have underestimated the true burden of disease, and those selected for neoadjuvant therapy tend to have more aggressive tumor features.

From a clinical point of view, the combination of inflammation-based scores and TNM staging can be useful in a more holistic approach to prognosis. While TNM staging gives an idea about the extent of the tumor in terms of anatomy, inflammation-based and nutrition-based scoring systems can give an idea about biological factors related to the hosts that cannot be explained by TNM staging. In this regard, the combination of TNM and INCS can help in a better prognostic assessment, especially in patients who have similar TNM stages but different inflammation-based scoring. Such prognostic information can help clinicians in managing such patients. The use of a dichotomized version of the score may further enhance its clinical applicability and facilitate its use in research settings, allowing for simplified risk stratification and more straightforward statistical modeling.

This study has several important limitations, including its retrospective design and reliance on data from a single center. External validation in larger and multi-center cohorts will be required before routine clinical use. Inflammatory and nutritional markers were measured at a single preoperative time point, reflecting only a snapshot of systemic inflammation without capturing temporal fluctuations. Details of chemotherapy regimens were not assessed, and this heterogeneity may have influenced survival patterns. Pathological staging after neoadjuvant treatment may also have underestimated true disease burden in some patients. Additionally, the small number of patients who received neoadjuvant therapy limits the reliability of subgroup analyses in this population. We did not compare our score against other composite indices such as the modified Glasgow prognostic score or SII, which could provide further context regarding its relative performance. Although the median follow-up duration may appear relatively limited, the mortality rate observed in our cohort likely reflects the advanced disease stages and biological aggressiveness of gastric adenocarcinoma. Furthermore, mortality data were obtained from official death registry records, minimizing the risk of follow-up bias.

CONCLUSION

In conclusion, this study demonstrates that a simple blood-based combined inflammation/nutrition score is associated with overall survival in gastric adenocarcinoma and provides prognostic information complementary to TNM staging. Improvement in performance when the score was integrated with stage suggested that systemic inflammation and tumor burden reflect different aspects of disease biology. Validation in larger and multi-center cohorts will clarify its clinical value and potential role in risk stratification.