Development and validation of nomograms for predicting overall survival and cancer-specific survival in female gastric signet-ring cell carcinoma: a SEER database analysis

Introduction

Gastric cancer is a common but aggressive cancer, ranking 5^th among the most prevalent tumors and 4^th among the deadliest tumors worldwide (1,2). Gastric signet-ring cell carcinoma (GSRC) is a specific subtype of gastric cancer that is pathologically characterized by an abundance of distinctive intracytoplasmic mucin components and compression of the surrounding nuclei (3). Several studies have demonstrated that compared with other types of gastric cancer, it is characterized by low differentiation, high invasiveness, and poor prognosis (4-6). In addition, the majority of GSRC patients usually present with locally progressive disease at the time of initial diagnosis, usually showing an advanced disease state involving lymph node and neighboring organ invasion (7). In recent years, it has been found that patients with signet-ring cell histology were more likely to be young women (8,9), and their clinical features and molecular mechanisms are significantly different from those of males (10,11). Although the overall incidence of gastric cancer is higher in men than in women, GSRC has a relatively higher incidence in women and is more common in younger women. This may be because female reproductive factors (such as menstrual status) may have a protective effect on signet-ring cell carcinoma of the stomach, and the risk of death in pre-menopausal women is significantly lower than that in men and postmenopausal women (10). For example, one study showed that women accounted for 57.1% of newly diagnosed GSRC patients in a given population, compared to 45% in the SEER database (12). Furthermore, an epidemiological study observed that the decline in the incidence rate of GSRC among males was more significant than that among females (11). The development of this histological type might be affected by female sex hormones, since estrogen receptor (ER) expression is frequently observed in poorly differentiated gastric carcinomas, suggesting a potential association between hormonal milieu and gastric tumorigenesis (13). Previous studies have shown that young female patients with gastric adenoma (GA) are prone to develop GSRC. Over 80% of GSRC cells can secrete mucus and express ERs, and they are more likely to metastasize to the ovary, suggesting that GSRC has a high affinity for estrogen and can promote tumor growth and invasion (14,15). A study has shown that the estrogen signaling pathway is abnormally activated in signet-ring cell carcinoma, suggesting that blocking the estrogen signaling pathway is a potential treatment method for GSRC (16), confirming that estrogen plays a key role in the development of GSRC (17).

Currently, many studies have developed similar prognostic models, Wan et al. developed the prognosis of GSRC patients with radiotherapy (18), Guo et al. constructed and verified GSRC overall survival (OS) and cancer-specific survival (CSS) prognosis curve (19), Jiang et al. developed web-based dynamic nomogram (20) for GSRC, Wang et al. established nomogram to predict the survival of GSRC in nonelderly adults (21), Wu et al. developed a nomogram for predicting OS after radical gastrectomy (22), Huang et al. developed a nomogram for predicting OS at 3, 6, and 12 months for patients with metastatic GSRC (23), these models can help doctors estimate the prognostic status of GSRC patients. However, these models also have the following drawbacks: firstly, these models do not conduct gender stratification analysis, ignoring the unique clinical and pathological features of women. Secondly, some models do not simultaneously predict OS and CSS in outcome indicators. Thirdly, the establishment of the nomogram does not consider the gender variables to be included. In addition, there are some models that only make predictions regarding the short-term survival time. This limitation is particularly important in clinical practice, as the unique factors of women may significantly affect treatment response and survival outcomes. There are no studies on the prognostic model of females with GSRC. Therefore, it is necessary to develop a specific prediction model for female GSRC patients in order to achieve more precise and personalized treatment.

The Surveillance, Epidemiology, and End Results (SEER) is a large-scale and high-quality cancer registry system, with its variables (such as tumor size) conforming to international standards and serving as a platform for developing basic models. Furthermore, the SEER database includes various racial subgroups (such as Asian Americans), which can partially reflect Asian characteristics. The prognosis of cancer patients is commonly evaluated using the American Joint Committee on Cancer (AJCC) staging framework; however, several clinicopathological variables—including age, marital status, and therapeutic interventions such as chemotherapy and surgery—that may influence the survival of patients with GSRC are not incorporated into this system. Consequently, the AJCC alone may be insufficient to provide individualized prognostic assessment for these patients. A large number of studies have been conducted to investigate the epidemiological characteristics, prognostic factors, and survival analysis of GSRC based on the SEER database and external databases, and partial risk scores or multifactorial prediction models have been established (18-20). However, almost none of the studies considered the survival of GSRC in females. Therefore, it is necessary to analyze the prognosis of this population in order to facilitate individualized clinical treatment and management.

A nomogram serves as a visualized statistical model for individualized risk estimation (24). It facilitates personalized treatment planning and prognostic prediction, thereby enhancing patient management, and has been increasingly applied in gastric cancer research. In the present study, a cohort of female GSRC cases was extracted from the SEER database to evaluate incidence patterns, prognostic factors, and survival outcomes. Based on retrospective data analysis, prediction models for 1-, 3-, and 5-year OS and CSS were developed and internally validated. These findings provide a reference for tailored clinical management and support evidence-based decision-making. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2193/rc).

Methods

Patients selection

The SEER database is a publicly accessible population-based cancer registry that collects data on cancer incidence, treatment modalities, and survival outcomes, representing roughly 30% of the United States (U.S.) population and including multiple regions and multiple ethnic groups (25). The SEER database only collected information on organ transfers since 2010, so we use the SEER*Stat software (http://seer.cancer.gov/seerstat/) included in 2010 to 2015 in the diagnosis of females with GSRC patients. Cancer was selected using the International Classification of Diseases for Oncology, 3^rd edition (ICD-O-3) criteria based on the primary site of the stomach, histological type of signet-ring cell carcinoma [8490], and gender being female. Data for related variables were downloaded from the database “Incidence-SEER Research Data. 17 Registries, Nov 2023 Sub” (e.g., age at diagnosis, race, gender, marital status, differentiation grade, AJCC stage, SEER stage, AJCC tumor stage (T stage), AJCC node stage (N stage), AJCC metastasis stage (M stage), human epidermal growth factor receptor 2 (HER2), cancer antigen-125 (CA-125), carcinoembryonic antigen (CEA), surgical status, radiotherapy status, chemotherapy status, bone metastasis, brain metastasis, liver metastasis, lung metastasis, cancer-specific cause of death, survival months, and survival status). This ensures the wide representation of female GSRC patients. Quality control involved excluding records with incomplete key data. Patients who met any of the following criteria were excluded from the analysis: those diagnosed based on autopsy or death certificate, those under 18 years old, those in T0/TX, NX stage, those with unknown marital status, and those with missing relevant clinical information. Variables such as lymphovascular invasion (LVI) and perineurial invasion (PNI) were not available in the SEER data during this study period and, therefore, were not included in the analysis. Finally, 976 females with GSRC were enrolled and randomly divided into two cohorts (7:3) based on survival time per 15 months: the training set (n=687) and the validation set (n=289). Figure 1 summarizes the flow chart of the patient selection process. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Figure 1 Detailed flow diagram of study population selection. AJCC, American Joint Committee on Cancer; CA-125, cancer antigen-125; CEA, carcinoembryonic antigen; HER2, human epidermal growth factor receptor 2; ICD-O-3, International Classification of Diseases for Oncology, 3^rd edition; M, metastasis; N, node; SEER, Surveillance, Epidemiology, and End Results; T, tumor; WHO, World Health Organization.

Development and validation of prediction models

Numbers and percentages are used to describe the classified data, and Chi-squared tests are used to compare the differences between the training and validation sets. Nineteen independent variables were considered candidate predictors, including age at diagnosis, race, tumor grade, AJCC N stage, AJCC M stage, AJCC stage, radiotherapy, chemotherapy, marital status, tumor size, bone metastasis, brain metastasis, liver metastasis, lung metastasis, tumor number, surgical site, SEER stage, and primary site. The assessment of ‘treatment response’ was achieved indirectly by comparing the survival hazard ratios (HRs) of patients who received specific treatments (such as chemotherapy) with those who did not receive any treatment. OS is defined as the time from diagnosis to either death from any cause or the last follow-up date. CSS is defined as the time from the date of GSRC diagnosis to either death due to GSRC or the last follow-up date.

Statistical analysis

The R (4.4.1) software was used for the statistical analysis and figure drawing in this paper. The least absolute shrinkage and selection operator (LASSO) regression (26) analysis was first used by to select the appropriate clinical variables from these 19 clinical variables. LASSO regression selects the optimal λ value (λmin) through 10-fold cross-validation. To ensure that the sample size is sufficient to support model construction without causing overfitting, we calculated the event per variable (EPV). Among all the samples (n=976), the number of events for OS was 679, and the number of events for CSS was 591. For the 19 candidate predictors, the EPV was 35.7 events/var for OS and 31.1 events/var for CSS, both of which were higher than the recommended threshold (EPV ≥10) (27). This indicates that the sample size is sufficient and can effectively avoid the risk of overfitting. Next, we used multivariate Cox analysis to identify independent predictors and then fed predictors of P<0.05 into the Cox proportional hazards model to construct the predictive model for females with GSRC patients and represented by a nomogram. The predicted results were 1-, 3-, and 5-year OS and CSS probability, respectively. The recognition ability of the model is evaluated by the concordance index (C-index). The C-index range is from 0.5 to 1. Here, 0.5 represents complete randomness, and the model has no predictive effect; the range of 0.5 to 0.7 indicates low accuracy; the range of 0.71 to 0.90 represents medium accuracy; the range above 0.90 indicates high accuracy; and 1 represents complete consistency. The receiver operating characteristic (ROC) curve over time evaluated the accuracy of the model, and the area under the curve (AUC) was greater than 0.7, indicating that the model had good predictive power. Calibration curves measure how close the predicted risk is to the actual risk, with closer curves indicating better predictions. Decision curve analysis (DCA) assessed the clinical utility of the model and quantified the net benefit at different threshold probabilities. Patients were divided into high- and low-risk groups based on the median risk score predicted by the model. Kaplan-Meier (K-M) analysis was used to draw survival curves of high- and low-risk groups to predict OS and CSS, and the log-rank test was used for survival analysis. The HR was calculated with a 95% confidence interval (CI). When the P value <0.05, the difference was significant.

Results

Baseline characteristics

Our study included 976 female patients with GSRC between 2010 and 2015, of whom 687 were in the training set and 289 were in the validation set. The Chi-squared test for each variable of the training set and validation set showed no significant difference between all the variables of the training set and validation set (P>0.05). Among all patients, 57.8% were older than 60 years old, and 8.3% were younger than 40 years old. Tumor size was less than 5 cm in 59.9% of patients, and in tumor-node-metastasis (TNM) staging, T4 (36.6%), N0 (43.1%), and M0 (81.4%) stages were the most common among patients. Of all 976 patients, 583 (59.7%) received chemotherapy, 270 (27.7%) received radiotherapy, and 798 (81.8%) received surgery. In addition, the median duration for the entire cohort was 25 months [interquartile range (IQR), 9–80 months], with an OS rate of 30.4% and a CSS rate of 39.4% by the end of follow-up. More detailed baseline characteristics of these two cohorts are shown in Table 1.

Development of prediction models

The LASSO regression model included 19 independent candidate variables (Table 1, Figure 2A,2B). When the partial likelihood deviance reached its minimum, 15 and 14 covariates were recognized as potential prognostic indicators for OS and CSS, respectively. To construct a more parsimonious and interpretable model, log(λ) parameters corresponding to one standard error above the minimum criterion were applied, and predictors with non-zero coefficients were retained. Ultimately, we selected seven variables (age, AJCC N stage, AJCC T stage, AJCC stage, tumor size, surgery, and chemotherapy) for multifactorial regression analysis of OS, and five variables (AJCC N stage, AJCC T stage, AJCC stage, tumor size, and surgery) were selected for multivariable regression analysis of CSS (Figure 2C,2D). A variable was defined as a risk factor for mortality when its regression coefficient >0 or HR>1, and was regarded as a protective factor otherwise. Based on multivariable Cox regression for OS, age (>60 years), AJCC T stage (T3–T4), AJCC N stage (N3), chemotherapy (no/unknown), AJCC stage (stages 2–4), and tumor size (>5 cm) were considered risk factors for the OS training group, and surgery (yes) was considered a protective factor for the OS training group. Based on multivariable Cox regression of CSS, AJCC stage T (T4), AJCC stage N (N3), AJCC stage (stages 2–4), and tumor size (>5 cm) were considered risk factors for the CSS training group, and surgery (yes) was considered a protective factor for the CSS training group. Table 2 shows the results of the multivariate Cox analysis performed on the training set. As shown in the column line diagrams, the selected variables were incorporated into the final 1-, 3-, and 5-year OS and CSS models (Figure 3A,3B).

Figure 2 Plot of LASSO coefficient profiles of the 19 candidate predictors for OS (A) and CSS (C) and partial likelihood deviance for OS (B) and CSS (D). CSS, cancer‑specific survival; LASSO, least absolute shrinkage and selection operator; OS, overall survival.

Figure 3 Nomogram for OS (A) and CSS (B). AJCC, American Joint Committee on Cancer; CSS, cancer‑specific survival; N, node; OS, overall survival; T, tumor.

Validation of prediction models

The C-index of the OS prediction model was 0.755 (95% CI: 0.733–0.777) in the training set and 0.766 (95% CI: 0.734–0.799) in the validation set. Meanwhile, the C-index of the CSS prediction model was 0.753 (95% CI: 0.728–0.777) in the training set and 0.753 (95% CI: 0.716–0.791) in the validation set. The calibration curves showed (Figures 4,5) that the predictions were consistent with the observations. Time-dependent ROC curves were used to compare the predictive performance of each prognostic factor and the prediction model, which showed that the predictability of the two models was high. In the training set, the AUCs of the OS prediction model for 1-, 3-, and 5-year were 0.783 (95% CI: 0.746–0.819), 0.855 (95% CI: 0.827–0.883), and 0.867 (95% CI: 0.839–0.895) (Figure 6A), and in the validation set, the 1-, 3-, and 5-year AUCs of the OS prediction model were 0.828 (95% CI: 0.778–0.878), 0.871 (95% CI: 0.830–0.911), and 0.887 (95% CI: 0.848–0.927) (Figure 6B). Similarly, in the training set, the 1-, 3-, and 5-year AUCs of the CSS prediction model were 0.758 (95% CI: 0.719–0.798), 0.863 (95% CI: 0.835–0.891), and 0.882 (95% CI: 0.856–0.909) (Figure 7A), and in the validation set, the 1-, 3-, and 5-year AUCs of the CSS prediction model were 0.777 (95% CI: 0.719–0.836), 0.847 (95% CI: 0.801–0.893), and 0.871 (95% CI: 0.825–0.916) (Figure 7B). In addition, the results of DCA also showed that the clinical applicability of both prediction models was superior to any single risk factor (Figures 8,9).

Figure 4 Calibration curves for 1-, 3-, and 5-year OS in the training set (A-C) and validation set (D-F). OS, overall survival.

Figure 5 Calibration curves for 1-, 3-, and 5-year CSS in the training set (A-C) and validation set (D-F). CSS, cancer‑specific survival.

Figure 6 Time-dependent ROC curves comparing prognostic accuracy between the OS prediction model in the training set (A) and validation set (B). AUC, area under the curve; CI, confidence interval; OS, overall survival; ROC, receiver operating characteristic.

Figure 7 Time-dependent ROC curves comparing prognostic accuracy between the CSS prediction model in the training set (A) and validation set (B). AUC, area under the curve; CI, confidence interval; CSS, cancer‑specific survival; ROC, receiver operating characteristic.

Figure 8 DCA for 1-, 3-, and 5-year OS prediction models in the training (A-C) and validation set (D-F). AJCC, American Joint Committee on Cancer; DCA, decision curve analysis; N, node; OS, overall survival; T, tumor.

Figure 9 DCA for 1-, 3-, and 5-year CSS prediction models in the training set (A-C) and validation set (D-F). AJCC, American Joint Committee on Cancer; CSS, cancer‑specific survival; DCA, decision curve analysis; N, node; T, tumor.

Survival analysis

Risk scores for all patients were generated using the established prediction model, and participants were subsequently stratified into low- and high-risk groups according to the median score. The K-M curves (Figure 10) showed that the prognosis of patients in the OS low-risk group was significantly better than that of the high-risk group (P<0.001, high vs. low HR =4.52), and the same result was also seen in the CSS group (P<0.001, high vs. low HR =5.29). In addition, we noted that the entire cohort responded well to surgery (OS, HR =0.29, 95% CI: 0.24–0.35; CSS, HR =0.28, 95% CI: 0.23–0.33; Figure 11). This suggests that our model can help physicians identify patients who may benefit from surgery, thus allowing for personalized treatment planning.

Figure 10 JenyK-M survival curves for OS (A) and CSS (B) in low-risk and high-risk female GSRC patients. CI, confidence interval; CSS, cancer‑specific survival; GSRC, gastric signet-ring cell carcinoma; HR, hazard ratio; K-M, Kaplan-Meier; OS, overall survival.

Figure 11 JenyK-M survival curves between surgery and no surgery groups for OS (A) and CSS (B). CI, confidence interval; CSS, cancer‑specific survival; HR, hazard ratio; K-M, Kaplan-Meier; OS, overall survival.

Discussion

Previous studies have shown that in early and late stages, the incidence of GSRC in young female patients is higher than that of gastric adenocarcinoma, and the incidence of GSRC has been rising in recent years (28). Mechanistic studies of GSRC have revealed that female reproductive factors induce diffuse gastric cancer through estrogenic activity (29). Additionally, studies have shown that the mortality risk of women with GSRC during menstruation is significantly lower than that of male patients (HR =0.58, 95% CI: 0.42–0.82) (10). This is consistent with previous studies on the effects of female reproductive factors on other tumors, such as in a study involving 758 patients, it was proposed that female reproductive hormones may be a potential protective factor for intestinal gastric cancer, and the incidence of intestinal gastric cancer in postmenopausal women is comparable to that in men (29). A study in Japan showed that women during their menstrual period had a lower risk of developing gastric cancer (HR =0.33, 95% CI: 0.23–0.49). A protective effect was observed in differentiated histological types (HR =0.25, 95% CI: 0.11–0.55) and undifferentiated histological types (HR =0.39, 95% CI: 0.23–0.63) (30). The study on the ER mechanism of gender differences in GSRC reveals that there are several types of ERs, including ERα, ERβ, and ERγ. The biological effects of estrogen are mediated by two ERs, ERα and ERβ, which belong to the nuclear receptor superfamily (31). It has been reported that both ERα and ERβ are expressed in poorly differentiated adenocarcinoma, with specific expression in GSRC adenocarcinoma with sex hormone-dependent features (32,33). However, there are few studies on the regulation of GSRC occurrence by estrogen or ERs, especially in the female population, suggesting that further research is urgently needed in the future.

In this study, we developed a predictive model for females with GSRC 1-, 3-, and 5-year OS and CSS outcomes based on a large number of clinical samples in the SEER database, and patients were divided into high-risk and low-risk groups based on the median point of the nomogram (OS: 182.95 points; CSS: 160.15 points; Figure 12). Seven and five parameters that were significantly associated with the patient’s OS and CSS, respectively, were included as independent prognostic factors by LASSO regression and multivariate Cox analysis. Previous studies have shown that young age can improve the survival of GSRC patients (34), and our study also showed that the younger patients, the better prognosis and the better prognosis of OS. Grinlinton et al. showed that patients with signet-ring cell carcinoma of the upper digestive tract seem to have a reasonable mid-term survival rate after surgery (35), and providing surgery may improve the prognosis analysis of this disease, which is similar to our results. Compared with patients with tumors without surgery, the total death risk of tumor patients undergoing surgery is reduced by 75% (HR =0.25), and the risk of cancer-specific death was reduced by 74% (HR =0.26). In addition, previous studies have shown that the prognosis of patients with early GSRC is better than that of patients with non-GSRC, while the prognosis of patients with advanced GSRC is worse (36). Similarly, according to our analysis, females with GSRC patients with higher AJCC stage have progressively worse prognosis and higher risk, among which the risk of patients with AJCC4 stage is 45% times that of those with stage 1 (HR =1.45). Wu et al. showed that distant lymph nodes were the most common metastatic sites in gastric GSRC patients (37), and previous studies showed that the higher the N stage, the worse the prognosis of the patients (18). Our study showed that the risk of N3 was 75% higher than N0 (HR =1.75), which is consistent with previous studies. Previous studies have shown that chemotherapy can provide GSRC patients with long-term disease-free survival and is the preferred therapy for metastatic cancer (38,39). The study by Pernot et al. also shows that first-line treatment of docetaxel-5FU-oxaliplatin (TEFOX) for advanced GSRC appears to be effective (40). According to our analysis, chemotherapy is also considered to increase OS in females with GSRC patients. However, studies have shown that chemotherapy drugs can cause drug resistance in the treatment of GSRC patients (41,42), resulting in many adverse reactions to patients. Therefore, it is important to consider the specific clinical facts of each patient when deciding whether to use adjuvant therapy. At the same time, it has been reported that GSRC patients can benefit from radiotherapy treatment (43,44), but in OS model construction, based on LASSO regression analysis, “radiotherapy” is excluded, possibly because it may have collinearity with “chemotherapy”. A number of past studies have shown that tumor size has a significant impact on the prognosis of cancer, including GSRC (45-47). Similarly, our study also showed that the larger the tumor size, the worse the prognosis of patients. When the patient’s tumor is larger than 10 cm, the risk is increased by nearly 50% (HR =1.46). However, individual organ metastases such as bone, brain, lung, and liver metastases were not identified as risk factors, which may be due to the low proportion of patients with individual organ metastases in the current cohort. However, it is necessary to screen these sites in the clinic.

Figure 12 JenyK-M survival curves stratified by the median point of the nomogram: OS (A) and CSS (B). CSS, cancer‑specific survival; K-M, Kaplan-Meier; OS, overall survival.

From the model validation results, it can be seen that the C-index of all models is greater than 0.7, which is 0.755 (95% CI: 0.733–0.777) and 0.766 (95% CI: 0.734–0.799) in the training set and validation set of the OS prediction model, respectively. In the training set and validation set of the CSS prediction model, the values were 0.753 (95% CI: 0.728–0.777) and 0.753 (95% CI: 0.716–0.791), respectively. This indicates that the models provide moderately accurate estimations. We also applied ROC curve, calibration curve, and DCA to prove the validity of the model. The results show that the AUC values of these models are relatively high, all of which are >0.7, which indicates that our prediction model has high accuracy. The calibration curve is close to the diagonal line, which indicates that the model has good predictability and accuracy. DCA results showed that the nomogram prediction model had better clinical outcomes than any single factor, including AJCC stage, and the prediction model produced higher clinical benefit than any single risk factor. These findings demonstrate the accuracy, clinical utility, and universality of the column diagram. On the risk score calculated by the predictive model, we divided the entire cohort into low- and high-risk groups. Across the cohort, patients in the high-risk group had higher all-cause mortality and cancer-specific mortality than those in the low-risk group, respectively (P<0.001). In addition, the K-M survival analysis of all cases with or without surgery showed a better prognosis in the surgery group (P<0.001), suggesting that the model made a significant difference for patients at different risk for 1-, 3-, and 5-year outcomes and that surgery had an important role in improving patient outcomes.

The developed nomogram was employed to visualize and implement the prediction model, offering an intuitive approach for estimating multiple clinical outcomes and supporting personalized clinical decision-making in female GSRC patients. Moreover, all prognostic indicators included in this study are easily obtainable in routine clinical settings, facilitating practical use and integration into clinical workflows. Our model was evaluated for unique females with GSRC patients, and at the same time, we randomly divided the data from the SEER database into a development set and a validation set using stratified randomization, which prevents an imbalance of known factors influencing prognosis between the two groups. Besides this, we also downloaded the data from different datasets of the SEER database as an extended validation of the model, thus allowing for more efficient testing of the performance of the model. Extended validation also showed our model with good validity and accuracy.

Although the established nomogram demonstrated good predictive performance for OS and CSS in female GSRC, several limitations should be acknowledged. First, cases with incomplete or ambiguous records were excluded from the analysis, which may have introduced potential selection bias. GSRC was found to be mostly advanced and accompanied by organ metastasis (7). Molecular heterogeneity is an important determinant of tumor invasiveness and prognosis. For instance, in glioblastoma, tumor cells exist in multiple states (such as neuroprogenocytoid, oligodendrocyte progenocytoid, astrogliocytoid, and mesenchymoid), and these states can switch between each other, increasing the complexity of treatment (48). However, in this study, relevant tumor molecular indicators such as HER2, CA-125, CEA, and related markers were not recorded. This might be because SEER mainly collects clinicopathological variables (such as TNM stage, histology) rather than relevant molecules, which is a limitation of population-based cancer registry. In the future, through multi-regional sampling and transcriptomic analysis, molecular heterogeneity within tumors can be better identified, and its impact on prognosis can be evaluated. Furthermore, almost all patients (>98%) did not have bone, liver, lung, or brain metastasis, and the sample size of patients with organ metastasis was too small for us to evaluate the relationship between survival and it. This may be because the data is excluded in the preprocessing. Secondly, one significant limitation of the SEER database is the low distant transfer rate of its records, a view that many studies have shown (49,50). Furthermore, the SEER database does not record certain key information, such as LVI, PNI (51), the specific chemotherapy plan, and the dosage and plan of radiotherapy (52). These factors are closely related to the increase in distant metastasis rates, which limits the comprehensive analysis of factors related to distant metastasis. In the future, it is recommended to conduct prospective studies to reduce the selection bias and information deficiency problems in retrospective studies, and cases of clinical distant organ metastases from respective hospitals can be included to enhance the statistical testing power and the reliability of the results. In addition, Helicobacter pylori (Hp) infection is one of the risk factors for gastric cancer (53), and the lack of specific information on whether patients have been infected with Hp in the SEER database will not be conducive to all-around analysis. Multi-center studies may help overcome the limitations of single-center studies and provide more representative results. Third, with the advancement of precision medicine, relying solely on basic clinical and pathological characteristics may be insufficient for accurate assessment of tumor prognosis. Combining multiple gastric cancer biomarkers such as CEA and CA19-9 with clinical features of targeted therapeutic sites, such as epidermal growth factor receptor (EGFR), HER2, and vascular endothelial growth factor (VEGF)/VEGF receptor (VEGFR) (54) and may provide more important prognostic value. Fourthly, this study is a retrospective study rather than a prospective cohort study, which relies on historical records and is prone to selection bias and information bias, and it is difficult to control the interference of confounding factors. Fifth, the cohort analyzed in this study consisted of patients diagnosed from 2010 to 2015. With the progress of medicine, many new therapies such as immunotherapy, targeted therapy, and neoadjuvant therapy have been applied to cancer patients, so the patients included in this study have a lag and are limited in clinical guidance. Sixth, this study included only the data from the SEER database, excluding other publicly available data from female with GSRC patients, introducing an inherent bias. However, there are differences in the epidemiology and clinical manifestations of GSRC between Western and Asian populations. This is reflected in the screening methods, surgical standards (such as the D2 lymph node dissection rate in Asian regions), and the biological characteristics of the tumors, among other aspects. Even though the SEER database includes subpopulations of demographics (such as Asian Americans), this may partially bridge the gap, and the variables of the model (such as AJCC staging, tumor size) are generally applicable. However, future external validation in Asian cohorts, such as using clinical data from East Asia, is crucial for confirming its generalizability. Future research with larger and more diverse datasets will be required to develop more robust and generalizable prediction models.

Conclusions

Using clinical variables from the SEER database combined with clinicopathological features identified through LASSO regression and multivariate Cox analysis, prediction models for 1-, 3-, and 5-year OS and CSS were developed and internally validated in female GSRC. The established models demonstrated superior predictive accuracy compared with the AJCC staging system and effectively estimated individual survival outcomes. Furthermore, the proposed nomogram can assist in risk stratification, support clinical prognostic evaluation, and help clinicians design personalized therapeutic strategies.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2193/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2193/prf

Funding: This study was funded by the Natural Science Foundation of Fujian Province (No. 2022J05062), the Natural Science Foundation of Xiamen (No. 3502Z20227041), and the Scientific Research Funds of Huaqiao University (No. 21BS126).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2193/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
3. Chu PG, Weiss LM. Immunohistochemical characterization of signet-ring cell carcinomas of the stomach, breast, and colon. Am J Clin Pathol 2004;121:884-92. [Crossref] [PubMed]
4. Hass HG, Smith U, Jäger C, et al. Signet ring cell carcinoma of the stomach is significantly associated with poor prognosis and diffuse gastric cancer (Lauren's): single-center experience of 160 cases. Onkologie 2011;34:682-6. [Crossref] [PubMed]
5. Hu B, El Hajj N, Sittler S, et al. Gastric cancer: Classification, histology and application of molecular pathology. J Gastrointest Oncol 2012;3:251-61. [PubMed]
6. Zu H, Wang H, Li C, et al. Clinicopathologic characteristics and prognostic value of various histological types in advanced gastric cancer. Int J Clin Exp Pathol 2014;7:5692-700. [PubMed]
7. Voron T, Messager M, Duhamel A, et al. Is signet-ring cell carcinoma a specific entity among gastric cancers? Gastric Cancer 2016;19:1027-40. [Crossref] [PubMed]
8. Theuer CP, Nastanski F, Brewster WR, et al. Signet ring cell histology is associated with unique clinical features but does not affect gastric cancer survival. Am Surg 1999;65:915-21. [Crossref] [PubMed]
9. Bamboat ZM, Tang LH, Vinuela E, et al. Stage-stratified prognosis of signet ring cell histology in patients undergoing curative resection for gastric adenocarcinoma. Ann Surg Oncol 2014;21:1678-85. [Crossref] [PubMed]
10. Li Y, Zhong YX, Xu Q, et al. Protective effects of female reproductive factors on gastric signet-ring cell carcinoma. World J Clin Cases 2022;10:5217-29. [Crossref] [PubMed]
11. Yu ZH, Zhang LM, Dai ZQ, et al. Epidemiology and prognostic nomogram for locally advanced gastric signet ring cell carcinoma: A population-based study. World J Gastrointest Oncol 2024;16:2610-30. [Crossref] [PubMed]
12. Fung BM, Patel M, Patel N, et al. Signet Ring Cell Gastric Carcinoma: Clinical Epidemiology and Outcomes in a Predominantly Latino County Hospital Population. Dig Dis Sci 2021;66:1240-8. [Crossref] [PubMed]
13. Matsui M, Kojima O, Kawakami S, et al. The prognosis of patients with gastric cancer possessing sex hormone receptors. Surg Today 1992;22:421-5. [Crossref] [PubMed]
14. Jin X, Wu W, Zhao J, et al. Clinical Features and Risk Factors for Lymph Node Metastasis in Early Signet Ring Cell Gastric Cancer. Front Oncol 2021;11:630675. [Crossref] [PubMed]
15. Yang XF, Yang L, Mao XY, et al. Pathobiological behavior and molecular mechanism of signet ring cell carcinoma and mucinous adenocarcinoma of the stomach: a comparative study. World J Gastroenterol 2004;10:750-4. [Crossref] [PubMed]
16. Zhao W, Jia Y, Sun G, et al. Single-cell analysis of gastric signet ring cell carcinoma reveals cytological and immune microenvironment features. Nat Commun 2023;14:2985. [Crossref] [PubMed]
17. Tang W, Liu R, Yan Y, et al. Expression of estrogen receptors and androgen receptor and their clinical significance in gastric cancer. Oncotarget 2017;8:40765-77. [Crossref] [PubMed]
18. Wan G, Wang Q, Li Y, et al. Development and validation of a nomogram for predicting survival in gastric signet ring cell carcinoma patients treated with radiotherapy. Sci Rep 2024;14:29963. [Crossref] [PubMed]
19. Guo Q, Wang Y, An J, et al. A Prognostic Model for Patients With Gastric Signet Ring Cell Carcinoma. Technol Cancer Res Treat 2021;20:15330338211027912. [Crossref] [PubMed]
20. Jiang Y, Hu H, Shao X, et al. A novel web-based dynamic prognostic nomogram for gastric signet ring cell carcinoma: a multicenter population-based study. Front Immunol 2024;15:1365834. [Crossref] [PubMed]
21. Wang H, Peng Y, Huang Q, et al. Prognostic Nomograms for Nonelderly Adults with Gastric Signet Ring Cell Carcinoma. Biomed Res Int 2021;2021:1274527. [Crossref] [PubMed]
22. Wu J, Wang J, Chen N, et al. A prognostic nomogram for predicting overall survival in gastric signet ring cell carcinoma patients: a SEER database and Chinese registry analysis. Front Mol Biosci 2025;12:1704157. [Crossref] [PubMed]
23. Huang D, Li X, Huang Y, et al. Individualized prediction tool for patients with metastatic gastric signet cell carcinoma. Sci Rep 2025;15:33163. [Crossref] [PubMed]
24. Bonnett LJ, Snell KIE, Collins GS, et al. Guide to presenting clinical prediction models for use in clinical settings. BMJ 2019;365:l737. [Crossref] [PubMed]
25. Cronin KA, Ries LA, Edwards BK. The Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute. Cancer 2014;120:3755-7. [Crossref] [PubMed]
26. Tibshirani R. The lasso method for variable selection in the Cox model. Stat Med 1997;16:385-95. [Crossref] [PubMed]
27. Peduzzi P, Concato J, Feinstein AR, et al. Importance of events per independent variable in proportional hazards regression analysis. II. Accuracy and precision of regression estimates. J Clin Epidemiol 1995;48:1503-10. [Crossref] [PubMed]
28. Kao YC, Fang WL, Wang RF, et al. Clinicopathological differences in signet ring cell adenocarcinoma between early and advanced gastric cancer. Gastric Cancer 2019;22:255-63. [Crossref] [PubMed]
29. Kim SM, Min BH, Lee J, et al. Protective Effects of Female Reproductive Factors on Lauren Intestinal-Type Gastric Adenocarcinoma. Yonsei Med J 2018;59:28-34. [Crossref] [PubMed]
30. Persson C, Inoue M, Sasazuki S, et al. Female reproductive factors and the risk of gastric cancer in a large-scale population-based cohort study in Japan (JPHC study). Eur J Cancer Prev 2008;17:345-53. [Crossref] [PubMed]
31. Matsuyama S, Ohkura Y, Eguchi H, et al. Estrogen receptor beta is expressed in human stomach adenocarcinoma. J Cancer Res Clin Oncol 2002;128:319-24. [Crossref] [PubMed]
32. Kang MH, Choi H, Oshima M, et al. Estrogen-related receptor gamma functions as a tumor suppressor in gastric cancer. Nat Commun 2018;9:1920. [Crossref] [PubMed]
33. Zhao XH, Gu SZ, Liu SX, et al. Expression of estrogen receptor and estrogen receptor messenger RNA in gastric carcinoma tissues. World J Gastroenterol 2003;9:665-9. [Crossref] [PubMed]
34. Ren J, Niu G, Wang X, et al. Effect of Age on Prognosis of Gastric Signet-Ring Cell Carcinoma: A SEER Database Analysis. Med Sci Monit 2018;24:8524-32. [Crossref] [PubMed]
35. Grinlinton M, Furkert C, Maurice A, et al. Gastroesophageal signet ring cell carcinoma morbidity and mortality: A retrospective review. World J Gastrointest Surg 2024;16:1629-36. [Crossref] [PubMed]
36. Nie RC, Yuan SQ, Li YF, et al. Clinicopathological Characteristics and Prognostic Value of Signet Ring Cells in Gastric Carcinoma: A Meta-Analysis. J Cancer 2017;8:3396-404. [Crossref] [PubMed]
37. Wu J, Fang D, Man D, et al. Clinical Correlates and Prognostic Value of Different Metastatic Sites in Gastric and Colorectal Signet Ring Cell Carcinoma. Engineering 2020;6:1028-34. [Crossref]
38. Yan GJ, Ji ZH, Liu G, et al. CRS + HIPEC combined with IP + IV chemotherapy for gastric signet-ring cell carcinoma: Case report of long-term survival. Medicine (Baltimore) 2020;99:e22647. [Crossref] [PubMed]
39. Ji J, Zhang X, Yuan S, et al. Survival impact of gastrectomy and chemotherapy on gastric signet ring-cell carcinoma with different metastatic lesions: A population-based study. Asian J Surg 2024;47:1769-75. [Crossref] [PubMed]
40. Pernot S, Dubreuil O, Aparicio T, et al. Efficacy of a docetaxel-5FU-oxaliplatin regimen (TEFOX) in first-line treatment of advanced gastric signet ring cell carcinoma: an AGEO multicentre study. Br J Cancer 2018;119:424-8. [Crossref] [PubMed]
41. Shu Y, Zhang W, Hou Q, et al. Prognostic significance of frequent CLDN18-ARHGAP26/6 fusion in gastric signet-ring cell cancer. Nat Commun 2018;9:2447. [Crossref] [PubMed]
42. Wang SY, Wang JH, Chen RK, et al. Mapping the landscape of gastric signet ring cell carcinoma: Overcoming hurdles and charting new paths for advancement. World J Clin Oncol 2025;16:98983. [Crossref] [PubMed]
43. Sasako M, Sakuramoto S, Katai H, et al. Five-year outcomes of a randomized phase III trial comparing adjuvant chemotherapy with S-1 versus surgery alone in stage II or III gastric cancer. J Clin Oncol 2011;29:4387-93. [Crossref] [PubMed]
44. Wei F, Lyu H, Wang S, et al. Postoperative Radiotherapy Improves Survival in Gastric Signet-Ring Cell Carcinoma: a SEER Database Analysis. J Gastric Cancer 2019;19:393-407. [Crossref] [PubMed]
45. Hui X, Zhou G, Zheng Y, et al. Development and validation of a tumor size-stratified prognostic nomogram for patients with gastric signet ring cell carcinoma. Updates Surg 2024;76:2813-24. [Crossref] [PubMed]
46. Fukui T, Fukumoto K, Okasaka T, et al. Prognostic impact of tumour size in completely resected thymic epithelial tumours. Eur J Cardiothorac Surg 2016;50:1068-74. [Crossref] [PubMed]
47. Liu Y, He M, Zuo WJ, et al. Tumor Size Still Impacts Prognosis in Breast Cancer With Extensive Nodal Involvement. Front Oncol 2021;11:585613. [Crossref] [PubMed]
48. Nussinov R, Yavuz BR, Jang H. Molecular principles underlying aggressive cancers. Signal Transduct Target Ther 2025;10:42. [Crossref] [PubMed]
49. Lu T, Xu H, Dong X, et al. Epidemiology and survival of patients with central nervous system solitary fibrous tumors: A population-based analysis. Front Oncol 2022;12:977629. [Crossref] [PubMed]
50. Ma Z, Yang S, Yang Y, et al. Development and validation of prediction models for the prognosis of colon cancer with lung metastases: a population-based cohort study. Front Endocrinol (Lausanne) 2023;14:1073360. [Crossref] [PubMed]
51. Cai YL, Lin YX, Jiang LS, et al. A Novel Nomogram Predicting Distant Metastasis in T1 and T2 Gallbladder Cancer: A SEER-based Study. Int J Med Sci 2020;17:1704-12. [Crossref] [PubMed]
52. Xie T, Qiu BM, Luo J, et al. Distant metastasis patterns among lung cancer subtypes and impact of primary tumor resection on survival in metastatic lung cancer using SEER database. Sci Rep 2024;14:22445. [Crossref] [PubMed]
53. Tan Y, Matsuzaki J, Saito Y, et al. Environmental factors in gastric carcinogenesis and preventive intervention strategies. Genes Environ 2025;47:5. [Crossref] [PubMed]
54. Luo D, Liu Y, Lu Z, et al. Targeted therapy and immunotherapy for gastric cancer: rational strategies, novel advancements, challenges, and future perspectives. Mol Med 2025;31:52. [Crossref] [PubMed]