Full Text Article

Introduction

Gastric cancer (GC) is the fifth most common cancer, and the third leading cause of cancer death worldwide [1]. The GC incidence rates are high in Eastern/Southeastern and Central Asia, Eastern Europe, and parts of Central and Southern America, while the figures are relatively low in North America and Western Europe [2]. Since World War II, the incidence of GC has been declining globally, and now it is one of the least common cancers in North America, but it nevertheless remains a heavy burden of public health worldwide because the incidence of GC remains high in East Asia [1]. In 2020, an estimated 27,600 people will be diagnosed, and 11,010 people will eventually die of GC in the United States [3]. The prognosis of patients with GC is dismal since the 5- year relative survival rate was 30.1% for GC patients in the United States from 2005 to 2009 [4].

Most GCs include tumors of the noncardia and the subcardia (Siewert type III), with their center starting 2-5 cm below the esophagogastric junction, while cardia GC (CGC) is more proximal to the junction and is managed like esophageal and esophagogastric tumors [5-7]. In developed countries, the CGC incidence follows the distribution of esophageal cancer [8-10], while non-cardia GC shows marked geographic variation with countries in Eastern Asia and South America [11]. In 2012, non-cardia GC occurred more frequently than CGC, with an average ratio of 2:1, while in certain populations where non-cardia GC incidence rates are lower than the global average, CGC rates are similar to or higher than non-cardia GC rates [12].

Though both CGC and non-cardia GC are reported to be influenced by cigarette smoking [13-15] and high salt intake [16,17], and possibly by low intake of fruits and vegetables [18-20], the other risk factors differ between the two GC types. Indeed, the risk factors for CGC are similar to those for oesophageal adenocarcinoma, including obesity [21,22] and gastroesophageal reflux disease [23,24]. On the other hand, non- cardia GC is strongly associated with Helicobacter pylori infection [25,26]. Some evidence suggests that H. pylori infection might even be inversely associated with both esophageal adenocarcinoma and CGC [27,28], while studies in some populations have suggested a positive association between the H. pylori and CGC [29,30].

It is reported that CGC has more aggressive biological behavior than non-cardia GC. Data from Japanese [28] and Korean [30] studies have noted that patients with CGC are more likely to have advanced T and N stages at diagnosis. Data from a Chinese population revealed that compared with patients with non-cardia GC, patients with CGC tend to be diagnosed at a more advanced stage and have a worse prognosis after R0 resection [31]. On the other hand, data based on a USA population [32] suggested that the disease-free survival (DFS) and overall survival (OS) were similar between patients with CGC and those with non-cardia GC. Nevertheless, both studies are limited by small sample size, and whether non-cardia GC and CGC have different prognosis remains uncertain.

A large-scale study on the prognosis of GC patients and non-cardia GC is needed. The aim of this study is to compare the cancer-specific survival (CSS) of CGC with that of non-cardia GC in USA patients who had undergone gastrectomy and regional lymph node dissection.

Patients and Methods

Database

The patient population was based on the Surveillance, Epidemiology, and End Results (SEER) cancer registry (www.seer.cancer.gov). Sponsored by National Cancer Institute, the SEER program collects and publishes incidence, mortality, prevalence, survival, and lifetime risk statistics, which can be used to assess the impact of cancer in the general population. The current SEER database consists of 18 population-based registries, which cover approximately 26% of the United States population. It is the largest publicly available cancer database, including information on prevalence, incidence, age, sex, race and ethnicity, year of diagnosis, marital status, insurance, TNM stage, geographic region, and mortality.

Patient selection

The SEER-stat software (SEER*Stat 8.1.6) was used for data extraction and patient selection.

The inclusion criteria were 1) patients with a diagnosis of invasive adenocarcinoma of the stomach (International Classification of Disease [ICD]-O-3 code in the range of 8000-8152, 8154-8231, 8243-8245, 8250-8576, 8940-8950, and 8980-8990), 2) diagnosed from 2004 to 2010, 3) underwent gastrectomy, and 4) have recorded numbers of lymph nodes. The exclusion criteria were 1) multiple cancers, 2) unknown TNM stage, 3) M1 disease, 4) local tumor destruction, 5) local tumor excision; 6) unknown whether cancer-directed surgery was performed 7) no cancer-directed surgery of the primary tumor; or 8) primary site recorded as “stomach, NOS” and “overlapping lesion of the stomach”. Cause of death (COD) and follow-up were restricted to “Alive or dead due to cancer” and “Active follow-up”.

Variables

According to the SEER database, tumor location is described as cardia, fundus, body, greater curvature, lesser curvature, gastric antrum, pylorus, “stomach, NOS”, and overlapping lesion of the stomach. We divided the patients into two groups by tumor location: CGC and non-cardia GC (NCGC). The NCGC subgroup includes fundus, body, greater curvature, lesser curvature, gastric antrum, and pylorus. The race was classified as white, African-American, and others (including American Indian/AK Native, Asian/Pacific Islander), as determined by SEER. Marital status was identified as married, single, widowed, separated, divorced, and unknown. Single, widowed, separated, and divorced was collected as unmarried, so marital status was classified into three subgroups (married, unmarried, and unknown) in this study. According to the SEER database, cause-specific survival is a net survival measure representing survival to a specified cause of death in the absence of other causes of death. Estimates are calculated by specifying the cause of death. Individuals who died of causes other than the specified cause were censored. In this study, GC is the specified cause of death. The seventh American Joint Classification of Cancer (AJCC) TNM staging system was adopted [33].

Outcomes

The outcomes were OS (time from diagnosis to death from any cause) and CSS (time from diagnosis to GC-related death). Patients alive at last contact were censored at the last contact. For CSS, patients who died from non-GC causes were censored on the date of death.

Statistical analysis

Categorical variables are presented as n (%) and were analyzed using the chi-square test or Fisher’s exact test. Continuous data are presented as means ± standard deviations and were analyzed using Wilcoxon’s rank-sum test. Multivariable Cox regression analysis was used to determine the factors independently associated with CSS after adjustment for demographic and therapeutic factors. All P-values were two-sided, and P-values <0.05 were considered statistically significant. All confidence intervals (CIs) were stated at the 95% confidence level. All analyses were conducted using SPSS 22.0 (IBM, Armonk, NY, USA).

For the survival analysis, two models of the association between CSS and tumor location were constructed using a propensity score-matched (PSM) univariable Cox proportional hazards model and an unmatched univariable analysis based on the Kaplan-Meier estimator of inverse probability of treatment weight (IPTW). To construct the PSM model of OS, NCGC patients were matched 1:2 to CGC patients on propensity score by using a greedy, nearest neighbor matching algorithm, with maximum allowed differences of ±0.1% for propensity scores. Kaplan-Meier estimators were calculated for each group and were compared using the log-rank test. For the final model of OS, IPTW Kaplan-Meier estimators were calculated across all patients and compared between the two groups using the log-rank test. All calculations were performed with R software, version 3.6.3 (The R Project for Statistical Computing, www.r-project.org).

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. In the SEER database, all data are anonymized.

Results

Clinicopathological characteristics of the patients

From the SEER, 17,623 GC cases were extracted, but those patients included patients with stage I to IV cancer, and detailed medical records are not available for some patients. Therefore, to obtain more convincing results and reduce bias, only patients who had undergone surgery, with the type of surgery and number of examined lymph nodes being reported, were included. Finally, 7278 GC patients who had undergone gastrectomy and regional lymph nodes dissection were eligible for this study from January 2004 to December 2010 (Figure 1).

Among total of 7278 patients, there were 2366 (32.5%) and 4912 (67.5%) patients with CGC and NCGC, respectively. Their characteristics are listed in Table 1. Matching by propensity score achieved an adequate balance between the two groups for age, marital status, race, differentiated grade, and tumor stage (Appendix Table A1, online only). Patients diagnosed with cardia adenocarcinoma were younger (63.0 vs. 66.3 years, P<0.001), more frequently white compared to African-Americans/others, and had a higher percentage of males (79.2% vs. 54.9%, P<0.001). The CGC group also had larger proportion of well/moderate differentiated tumors (36.8% vs. 27.3%, P<0.001), higher lymph node metastasis rates (66.3% vs. 59.5%, P<0.001), relatively higher percentage of advanced T stage (T3/T4a/T4b, 68.6% vs. 64.4%, P<0.001), and more advanced TNM stage (stage II/III, 75.2% vs. 70.7%, P<0.001) than the NCGC group.

Prognostic factors for cancer-specific survival in GC patients

The median follow-up for the entire cohort was 72 months (interquartile range, 53 to 94 months). CSS was compared among the different subgroups (Table 2). Patients <60

years displayed better survival than those of 61-74 and the >75 years group (5-year CSS:

51% vs. 47% vs. 34%, P<0.001). Patients in marriage had longer survival than those who were unmarried (including single, widowed, separated, and divorced) (5-year CSS: 48% vs. 40%, P<0.001). White patients and African-American patients displayed worse survival than patients of other races (5-year CSS: 42% vs. 44% vs. 56%, P<0.001). Patients diagnosed in 2009-2010 had a better survival than those diagnosed in 2007- 2008 and 2004-2006 (5-year CSS: 53% vs. 45% vs. 41%, P<0.001). Patients with CGC showed significantly worse survival than NCGC patients (5-year CSS: 39% vs. 48%, P<0.001). Patients who had undergone partial/subtotal/hemi- gastrectomy showed better survival than those who had undergone gastrectomy with the removal of a portion of the esophagus and those who had undergone gastrectomy with en bloc resection of other organs (5-year CSS: 50% vs. 40% vs. 40%, P<0.001). From the multivariable Cox regression analysis, age, marital status, race, year of diagnosis, tumor location (cardia and non-cardia), differentiated grade, and tumor stage were independent prognostic factors for GC.

CGC patients had a worse prognosis than NCGC patients

Among the 7278 cases, the median CSS of CGC patients was 23 months shorter than in NCGC patients (34 vs. 57 months, P<0.001). The 5-year CSS rate was 39% for CGC and 48% for NCGC (Figure 2A, P<0.001). The result remained after PSM (Figure 2B, P<0.001). In the PSM Cox proportional hazard regression analysis, cardia adenocarcinoma was associated with a significantly worse CSS (hazard ratio [HR], 1.33; 95% CI: 1.228-1.447, P<0.001); the 5-year CSS was 37% in the CGC group versus 48% in the NCGC group (P<0.001). The IPTW analysis revealed similar results (Figure 2C; P<0.001). The survival rates of the cohort were stratified by TNM stage (Table 3), and the results showed that the survival disadvantage remained for CGC diagnosed at stage I/II (stage I, 68% vs. 83%, P<0.001; stage II, 37% vs. 48%, P<0.001), but for patients diagnosed at stage III, the survival rates between the two groups were similar (20% vs. 20%, P=0.520). The results remained consistent after PSM (Table 3).

Discussion

Whether patients suffering from CGC have a worse prognosis than those with NCGC remained controversial. Therefore, this study aimed to compare the CSS of CGC with that of NCGC in USA patients who had undergone gastrectomy and regional lymph node dissection. The results showed that CGC patients present a significantly worse prognosis than NCGC patients. CGC is an independent prognostic factor for GC patients, especially those at stages I and II.

This study included a large number of patients with GC. The results showed that CGC

patients were younger than NCGC patients at diagnosis. This finding is consistent with the findings of studies of populations from western countries, Japan, and Korea [30,32,34,35], but not those of populations from China and another population from Korea [29,36-38]. In addition, compared with NCGC, CGC was more common in men, as supported by previous studies from Korea, Japan, and China [34,36,38], while three studies from Korea and Japan reported no substantial differences in sex between the two groups [30,35,37]. The data displayed a higher percentage of married people in the CGC group than in the NCGC group. A previous study demonstrated that married patients with GC displayed better survival than those unmarried [39]. The prevalence of white people in the CGC cohort was significantly higher than in the NCGC cohort, keeping with the trend of a high incidence of CGC in western countries [40-44].

The pathological features of CGC patients were compared with those of NCGC patients. The results showed that CGC had a higher percentage of tumors with a well/moderately differentiated grade than NCGC. In a study with large sample size, the CGC patients had a significantly lower rate of poor-to-moderate tumor grade than the NCGC [31], supporting the present study. In this study, CGC patients had higher lymph node metastasis rates than NCGC patients. A previous study based on a US population demonstrated a similar lymph node metastasis rate between CGC and NCGC [32], while studies from Japan, Korea, and China reported higher rates of lymph node metastasis in CGC than that in NCGC [31,34,35,37].

The most significant finding from this study is that CGC is an independent prognostic factor for patients with GC. CGC patients had a worse prognosis than NCGC patients, and the results remained significant in stages I and II, even after being adjusted for clinicopathological characteristics and therapeutic management. A previous study based on a US population demonstrated a similar prognosis between CGC and NCGC, but the authors declared that long-term outcome was worse among patients with CGC and early-stage disease [32], which is partly consistent with the present study. Moreover, a study based on a Singapore population demonstrated that the R0 resection rates were similar, but the systemic recurrence rate was higher in CGC, and survival was poorer for CGC compared with NCGC. A study from a Korean population demonstrated that, regardless of curative probability, survival was worse for proximal GC than for distal GC. A study in Chinese patients revealed that CGC patients had a worse prognosis after R0 resection [31]. A study based on a single Japanese center reported that CGC patients had a worse survival than NCGC patients after curative resection [34]. In comparison to a previous US study [32], the present study confirmed that CGC patients had a worse prognosis than NCGC patients after gastrectomy and regional lymph node dissection in a US population.

CGC has a tendency to have different risk factors than NCGC. Similar to oesophageal adenocarcinoma, CGC is associated with obesity [21,22] and gastro-oesophageal reflux disease [23,24], while NCGC is strongly associated with Helicobacter pylori infection [25,26]. Moreover, the surgical approach, extent of resection, lymph node dissection, digestive tract reconstruction, and neoadjuvant therapy of CGC is still under debate [6,7]. Subtotal esophageal and proximal gastric resection with gastric pull-up or distal esophageal resection with total gastrectomy and esophagojejunostomy are competing procedures for advanced CGC, and gastrectomy maybe not adequate for tumors invading the lower esophagus [6,7,45,46]. Furthermore, the Japanese Gastric Cancer Association (JGCA) recommend that the dissection of the lymph nodes at stations 4, 5, and 6 is not necessary for AEG tumors [47,48] because the lymphatic metastases of CGC are found mainly in stations 1, 2, 3, and 7 [47,48]. In addition, lymph nodes at stations 19 and 20 are recommended to be dissected for T2-4 CGC. According to the NCCN guidelines [6,7], patients with CGC are recommended to be treated as described in the NCCN Guidelines for Esophageal and EGJ Cancers [7]. Neoadjuvant therapy is increasingly used in advanced GC. Since a survival difference was observed between the two groups in patients diagnosed at stages I and II, this could be explained, at least in part, by the omission of neoadjuvant therapy in stage I-II patients.

The multivariable Cox regression model in this study showed that age, marital status, race, year of diagnosis, cardia GC, differentiated grade, tumor stage, radiation therapy, number of examined lymph nodes, and number of positive lymph nodes were independent prognostic factors for GC. We found that married patients had a better prognosis than unmarried patients, which is consistent with another SEER-based study [39]. Patients classified as other races (which included a certain number of Asian patients) had a better prognosis than the white and African-American patients, as supported by another SEER-based study [45]. Patients diagnosed in 2009 and 2010 had better survival than those diagnosed in 2004-2008, which could be attributed to the improvement of medical treatments. Patients with poorly differentiated tumors had worse survival compared with those with well/moderately differentiated tumors. In this study, the number of examined lymph nodes and the number of positive lymph nodes were also prognostic factors for GC patients, which is in keeping with the AJCC staging system [33].

To the best of our knowledge, this report describes the first SEER-based study focusing on the differences in clinicopathological characteristics and cancer-specific survival between CGC and NCGC. Nevertheless, there are some limitations due to the retrospective nature of this study. First, the SEER database lacks information about body mass index, smoking, drinking, and H. pylori infection, which are important risk factors for CGC and NCGC. Second, the database has no record of whether the surgery is an R0 resection or not. Finally, further details regarding the use of systematic treatments and respective responses should be considered, but the SEER database lacks such information.

In conclusion, based on 7278 GC cases who had undergone gastrectomy and regional lymph node dissection, CGC is more likely to be T3-T4 lesions and has higher lymph node metastasis rates than NCGC tumors. Following gastrectomy, the CSS of CGC patients is significantly worse than that of NCGC. CGC is an independent prognostic factor for GC patients, especially for those diagnosed at stages I-II.