Characteristics and Predictors of Long-Time Survivors in Non-Metastatic Gastric Signet Ring Cell Carcinoma: A Large Population-Based Study

Introduction

Gastric cancer is a lethal malignancy that accounts for approximately 8.27% of new cancer cases arising from the digestive system.¹ Histopathologically, most gastric tumors are adenocarcinomas.² Signet ring cell carcinoma (SRCC) is a relatively rare histological type with an extremely poor prognosis due to rapid tumor growth and diffuse infiltration into surrounding tissues;³ it accounts for 16.8% of all gastric cancer cases.⁴ SRCC is characterized by predominant intracytoplasmic mucin production in tumor cells (>50% size of cell) with the unique appearance of a signet ring.^5,6 A population-based analysis indicated that patients with gastric SRCC have significantly worse 5-year and 10-year survival rates than those with gastric adenocarcinoma (5-year survival: 19.2% vs 25.8%; 10-year survival: 16.0% vs 22.1%).⁴ The baseline demographics and clinical characteristics of gastric SRCC are also distinct from those of common gastric adenocarcinoma, including the mean age of onset, gender ratio, obesity prevalence, tumor grade, and advanced tumor stage.^4,7,8

Despite the overall dismal prognosis of gastric SRCC, the long-term survival (LTS) of patients with gastric SRCC has improved in recent decades owing to endoscopic treatment, surgery, and adjuvant and neoadjuvant treatments for early-stage patients, as well as systemic chemotherapy, hyperthermic intraperitoneal chemotherapy, and immunotherapy for metastatic patients.⁹ However, there are limited data on the optimal management of non-metastatic gastric SRCC possibly because of the rarity of this specific tumor type. Investigating the specific characteristics of patients with gastric SRCC who survive for a long time is thus critical.

In this study, long-term survivors (LTSs) were defined as patients with a cancer-specific survival (CSS) of >5 years.^10,11 Early death (ED) was defined as cancer-specific death within 5 years from the diagnosis of non-metastatic gastric SRCC. Eligible patients were identified from the Surveillance, Epidemiology and End Results (SEER) database, and the clinicopathological characteristics of non-LTSs and LTSs were compared. Independent predictors of LTS were identified and used to construct and validate a nomogram model to predict LTS in non-metastatic gastric SRCC.

Materials and Methods

Data Source and Patient Selection

Patients with non-metastatic gastric SRCC were collected from the SEER database between 2004 and 2015 using SEER*State (version 8.3.6). The SEER database includes population-based data from 18 registration centers, which covers approximately 30% of the US population.¹² The SEER database includes data on demographics, cancer incidence, tumor-related clinicopathological features, and cancer treatment, thereby providing accessible information for clinical studies of malignant tumors, especially for rare tumor entities.¹³ Ethical approval was obtained from the Ethics Committee of the Affiliated Hospital of the Medical School of Ningbo University. Written informed consent was not required in this SEER-based retrospective study.

First, 12,846 patients with gastric SRCC were identified from the SEER database between 2004 and 2015 according to the International Classification of Diseases in Oncology (ICD-O-3) (ICD-O-3: 8490). Patients were included if they met the following criteria: (1) patient age ≥18 years; (2) pathologically confirmed gastric SRCC; (3) single primary tumor; (4) active follow-up data; (5) survival time ≥1 month, which might be, otherwise, caused by perioperative complications; (6) clear cause of death; (7) non-metastatic patients with known TNM stage. As a result, 3906 eligible patients were enrolled in the study (Figure 1). Patients were divided into three groups according to age as follows: ≤50 years, 51–70 years, and >70 years.¹⁴

Figure 1 Flow chart of patient selection. Abbreviations: SRCC, signet ring cell carcinoma; CSS, cancer-specific survival; LTSs, long-time survivors.

LTSs were defined as patients with cancer-specific survival (CSS) >5 years. Patients with CSS <5 years were defined as non-LTSs.

For each patient, demographics, tumor features, treatment regimens, and patient survival data were retrieved from the SEER database.

Establishment and Validation of the Nomogram

Patients were randomly divided into the training group (N = 2759) and validation group (N = 1147) at a ratio of 7:3 by setting seed in R software. Independent predictors of LTS were identified by univariate and multivariate logistic regression analyses. Seven independent risk factors (age, race, sex, marital status, tumor size, T stage, and N stage) were used to construct a nomogram model to predict LTS.

The discrimination of the nomogram-based LTS prediction was assessed using a calibration plot in both cohorts. The C-index was also calculated. The predictive accuracy of the nomogram was evaluated using a receiver operating characteristic (ROC) curve and the area under the curve (AUC).

Statistical Analysis

The significance of differences in clinicopathological characteristics between LTSs and non-LTSs was evaluated using the chi square test. Outcomes of interest included overall survival (OS: the time interval from initial diagnosis to all-cause death) and CSS (the duration from initial diagnosis to death caused by gastric cancer). Survival curves were plotted using the Kaplan-Meier method, and survival differences were evaluated using the Log rank test. Potential predictors of LTS were identified by univariate logistic regression analysis, and variables with a P value <0.1 in the univariate analysis were included in the multivariate analysis. Results are expressed as odds ratio (OR) and 95% confidence interval (CI). SPSS statistics version 26.0 software (SPSS Inc, Chicago, United States) and R version 3.6.1 software (R Foundation for Statistical Computing, Vienna, Austria) were used for statistical analyses. A two-sided P value <0.05 indicated statistical significance.

Results

Baseline Characteristics of Patients

In total, 3906 patients with non-metastatic gastric SRCC were included according to the inclusion and exclusion criteria. The baseline characteristics of the patients are shown in Table 1. Specifically, 21.79% of patients were ≤50 years old (N = 851), 46.16% were 51–70 years old (N = 1803), and 32.05% were >70 years old (N = 1252). Most of the patients were white (N = 2642, 67.64%). There were no significant differences between the number of men (N = 2059, 52.71%) and women (N = 1847, 47.29%). Regarding marital status, 58.50% of the patients were married (N = 2285) and 37.15% were unmarried (N = 1451). Most patients showed poorly differentiated tumors (N = 3204, 82.03%). Patients were divided into different groups according to tumor size (Table 1). Regarding T stage, 30.18% (N = 1179), 42.01% (N = 1641), 23.02% (N = 899), and 4.79% (N = 187) of the patients had T1, T2, T3, and T4 tumors, respectively. Over half of the patients had no lymph node involvement (N = 2013, 51.54%), 32.51% had N1 stage (N = 1270), and 15.95% had N2 stage (N = 623). Most patients underwent surgery (N = 2896, 74.14%), 36.82% received radiation (N = 1438), and 57.76% received chemotherapy (N = 2256).

Table 1 Clinicopathological Characteristics of Included Patients

Differences in Clinicopathological Characteristics Between LTSs and Non-LTSs

The clinicopathological features of LTSs and non-TLSs were compared. Most variables were significantly different between the two groups (Table 2). There was a higher number of young patients (≤50 years) in the LTS group than in the non-LTS group. Regarding race, other race (except black and white) was more common among LTSs. The proportion of women (50.88%) and married patients (66.62%) was higher among LTSs.

Table 2 Differences of Clinicopathological Characteristics Between Non-LTSs and LTSs in Non-metastatic Gastric SRCC

Predictors of LTS in Non-Metastatic Gastric SRCC

Patients were randomly assigned to training and validation cohorts at a ratio of 7:3. Predictors of LTS were identified using univariate logistic regression analysis. Age, race, sex, marital status, tumor size, T stage, and N stage were associated with LTS (Table 3). These variables were incorporated into the multivariate logistic regression analysis, which showed that younger age, other race (not black or white population), female gender (OR = 1.21, P = 0.028), married status (OR = 1.59, P < 0.001), small tumor size, less advanced T stage, and less advanced N stage were significant independent predictors of LTS (Table 3).

Table 3 Logistic Regression Analysis to Identify Predictors of Long-Time Survival in Non-Metastatic Gastric SRCC in the Training Cohort

Establishment and Validation of a Nomogram

According to the results of the multivariate logistic regression analysis, age, race, sex, marital status, tumor size, T stage, and N stage were incorporated into a nomogram to predict the probability of LTS in a specific individual. As shown in Figure 2, T stage was the most important predictor of LTS with a maximal score of 100. Other variables had various effects on the probability of LTS.

Figure 2 Nomogram to predict the probability of long-time survival. For example, there was a 73-year old (0 point), white (0 point), married (30 points) man (0 point) with gastric SRCC. The tumor was > 6 cm (approximately 30 points), T1 stage (100 points), and negative lymph node involvement (30 points). The total score was 190 points. For this specific patient, the probability of long-time survival was approximately 0.2.

The nomogram showed good accuracy in predicting LTS in the training cohort, with a C-index of 0.745 (Figure 3A). The calibration plot showed good agreement between predicted and observed LTS (Figure 3A). Similarly, the C-index was 0.720 for the nomogram-based LTS prediction in the validation cohort (Figure 3B). The calibration curve showed good consistency between observed and predicted LTS probability. Finally, a ROC curve was generated to evaluate the predictive power of the nomogram-based prediction model for LTS. The AUC values were 0.745 and 0.720 in the training and validation cohorts, respectively (Figure 4).

Figure 3 Calibration curve of nomogram-predicted and actual probability of long-time survival in the training cohort (A, N = 2759) and the validation cohort (B, N = 1147).

Figure 4 Receiver operating characteristic (ROC) curve and area under the ROC curve (AUCs) in the training cohort (A, N = 2759) and the validation cohort (B, N = 1147). Abbreviations: ROC, receiver operating characteristic; AUC, area under the ROC curve.

Age Was an Independent Predictor of LTS in Non-Metastatic Gastric SRCC

Multivariate logistic regression analysis identified age as an independent predictor of LTS. To further examine the association between age and the probability of LTS, we analyzed patient survival stratified by age at diagnosis. As shown in Supplementary Figure 1, patients ≤50 years of age had the best survival, whereas those >70 years had the poorest prognosis.

Discussion

Gastric SRCC is a rare histological type of gastric cancer and is associated with poor survival compared with adenocarcinoma. The incidence of gastric SRCC is increasing, resulting in a considerable health burden to patients.¹⁵ Therefore, assessing the clinicopathological features of gastric SRCC is important. In addition, investigating the characteristics of LTSs may provide information essential for improving the clinical management of these patients. However, because of the rarity of gastric SRCC and the small sample size, demographic or survival analyses in a single institution are difficult. Here, we used the SEER database to examine the characteristics of LTSs among patients with non-metastatic gastric SRCC and identified predictors of LTS.

To the best of our knowledge, this is the first large population-based study investigating the characteristics of LTSs in non-metastatic gastric SRCC. Consistent with previous studies,^16,17 the incidence of gastric SRCC was higher among women, who accounted for 47.29% of patients in this study. Poorly differentiated or undifferentiated tumors accounted for the majority of tumors (85.18%), which was consistent with previous reports.¹⁸ Comparison of the clinicopathological features of LTSs and non-LTSs showed that LTSs included a higher proportion of young patients (≤50 years, 29.88%), other ethnicity (27.75%), female patients (50.88%), married patients (66.62%), those with a relatively smaller tumor size, and less advanced T and N stages (Table 2).

We further examined the independent predictors of LTS. Multivariate logistic regression analysis revealed that younger age, other race (not black or white population), female gender, married status, small tumor size, and less advanced T and N stages significantly predicted LTS. Age was a strong independent predictor of LTS. As shown in Table 3, patients ≤50 years and patients 51–70 years had a two-fold higher probability of LTS than those >70 years of age. The relatively better LTS in the young population was further confirmed by the survival curve (Supplementary Figure 1). Age at diagnosis is an independent predictor in several malignancies, such as colorectal cancer¹⁹ and hepatocellular carcinoma.²⁰ Saito et al analyzed 1473 gastric cancer patients who underwent curative gastrectomy and showed that age was an independent prognostic factor, although lymph node dissection was relatively limited and the proportion of patients receiving chemotherapy was lower among elderly patients.²¹ Consistently, a Japanese team demonstrated that more extensive lymph node dissection and fewer peri-operative complications might explain the better survival in young patients with gastric cancer.²² In addition, married patients were more likely to have LTS than unmarried patients. Spousal support is important for the active treatment and surveillance of patients after a cancer diagnosis,^23,24 whereas unmarried patients are more likely to experience undertreatment and lack of social support.²⁵

In addition to the demographic and tumor-specific factors analyzed in this study, several studies have demonstrated that molecular biomarkers are associated with patient prognosis in gastric cancer. Chen et al reported that the GALNT14-rs9679162 genotype was an effective predictor of prognosis in gastric SRCC.²⁶ These authors combined clinicopathological parameters with the GALNT14-rs9679162 genotype to develop a scoring system for stratifying advanced gastric SRCC patients into three distinguishable prognostic subgroups. However, most diagnostic and prognostic biomarkers of gastric SRCC are observed in the laboratory. For instance, hsa-miR-665 and hsa-miR-95 are downregulated in gastric SRCC and may be related to tumor invasion, metastasis, and chemoresistance.²⁷

The present study had several limitations. Firstly, selection bias is unavoidable due to the retrospective nature of the study. Although multivariate logistic regression analysis revealed that age was an independent predictor of LTS in non-metastatic gastric SRCC, the relatively poorer survival of elderly patients might also be associated with inadequate lymph node dissection and less intensive treatment. However, these data cannot be retrieved from the SEER database. Secondly, although the performance of the model was acceptable in both the training and validation cohorts, external validation is warranted to confirm the reliability and generalizability of the nomogram model. A prospective multicenter study with a large sample size is necessary to confirm the present findings. Finally, gastric SRCC with distant metastasis was excluded from the analysis because patients with metastatic disease are mainly treated with palliative systemic therapy and are likely to show a distinct prognostic pattern.²⁸ Thus, another study is necessary to examine this particular subgroup of patients.

In summary, we investigated the clinicopathological characteristics of non-metastatic gastric SRCC and identified predictors of LTS using the SEER database. Young age, other race (not black or white population), female gender, married status, small tumor size, lower tumor infiltration, and negative lymph node involvement were independent predictors of LTS in non-metastatic gastric SRCC. These seven parameters were used to construct a nomogram-based prediction model of LTS, which showed acceptable performance in both training and validation cohorts.