Introduction

Gastric cardiac carcinoma arises in the gastric cardia below the gastroesophageal junction [1,2,3]. Because of the paucity of clinical symptoms, most patients with tumors diagnosed at advanced stages have a poor prognosis. The 5-year survival rate was reported to be lower than 10% [1]. At present, the best strategy proven to improve patient survival is early diagnosis and resection as gastric cardiac carcinomas at early (pT1) stage have significantly improved 5-year survival rates of over 90% [2]. With advances in endoscopic therapy, early gastric carcinoma can be effectively resected by endoscopic mucosal resection or endoscopic submucosal dissection with fewer complications, better preservation of gastric function, and lower cost compared with surgical resection [4,5,6]. However, the clinical decision making for choosing endoscopic vs. surgical resection for early gastric carcinoma is based on the clinicopathologic finding of negligible risk of lymph node metastasis. Unfortunately, risk factors for nodal metastasis in early gastric cardiac carcinoma remain elusive.

Gastric cardiac carcinoma is uncommon in Western countries as well as some East Asian countries such as Japan and Korea [7]. There are a handful of single-center studies with small sample sizes that show a lower risk of lymph node metastasis in early gastric cardiac carcinoma [8,9,10] compared with early gastric non-cardiac carcinoma [2, 10,11,12]. These limited single-center experiences on assessment of risk of lymph node metastasis in early gastric cardiac carcinoma require in-depth investigation with large samples. Herein, we conducted a multicenter clinicopathologic study of risk factors associated with lymph node metastasis in 495 consecutive early gastric cardiac carcinoma radical resections in the Jiangsu Province in China where gastric cancer remains one of leading malignancies in terms of incidence and mortality. Notably, the most important confounding factors for investigation of early gastric cardiac carcinoma, such as Barrett’s esophagus and esophageal adenocarcinoma are rare in this study patient population [13,14,15,16].

Materials and methods

Study design, patient selection, and groups

In this multicenter retrospective study, we searched the electronic pathology database stored in the Departments of Pathology of four participating tertiary medical centers for the final pathologic diagnosis of gastric cancer in radical gastrectomy specimens over the 11-year period from 2005 through 2016 for The Nanjing Drum Tower Hospital, from 2011 to 2015 for The Jiangsu Provincial Hospital of Traditional Chinese Medicine, and from 2011 to 2016 for The Changzhou Second Hospital and The Affiliated First Hospital of Soochow University. Each pathology report was scrutinized to exclude the cases with tumor invasion into the muscularis propria and beyond. The resection cases with tumor invasion limited to the mucosa and submucosa (pT1) were selected as early gastric carcinoma for the study. All histology slides, including special stains, of each case were retrieved and investigated. Excluded were esophageal adenocarcinoma, stump carcinoma, lymphoma, synchronous carcinomas, high-grade dysplasia, tumor invasion into the muscularis propria, and cases with a history of neoadjuvant therapy before radical resection. In addition, cases without histology slides or paraffin blocks for recuts were also eliminated. Pathology reports of all selected cases were reviewed for patient demographic information, tumor epicenter location, size, shape, surface characteristics, and relationship to the gastroesophageal junction. In cases with equivocal information on tumor location, size, and shape, the original upper endoscopic reports and images along with the surgical resection notes were reviewed for accurate information. All identifying patient information was deleted from the data sheet to protect patient privacy. The pathology accession number of each case was used as the reference for all communication and data analyses. The study protocol was approved by the Medical Ethics Committee of all 4 participating medical centers.

Pathologic study

All study pathologists had over 5-year clinical practice experience with strong interest in gastrointestinal pathology and applied the World Health Organization criteria to guide early gastric carcinoma diagnosis in daily practice [17]. Early gastric cardiac carcinoma was defined as tumors with epicenters located in a narrow region of about 3 cm below the gastroesophageal junction line, as previously described [2, 3]. Early gastric non-cardiac carcinoma referred to early gastric carcinomas with epicenters in other regions of the stomach. Tumor gross characteristics were sub-grouped into five patterns as follows: (1) polypoid/protruding (type 0–I), (2) slightly elevated with rough surface (type 0–IIa), (3) flat (type 0–IIb), (4) slightly depressed with erosion (type 0–IIc), and (5) excavated with ulcer involving the muscularis mucosae and beyond (type 0–III) [17]. Tumor size was measured after routine overnight formalin fixation. Depth of tumor invasion was determined microscopically from the muscularis mucosae to the lowest point of invasion and divided into four categories: (1) M2, tumors limited to the lamina propria without the involvement of the muscularis mucosae; (2) M3, tumors involving the muscularis mucosae; (3) SM1, tumor invasion into the superficial submucosa (< 0.5 mm from the muscularis mucosae); and (4) SM2, tumor involvement of the deep submucosa ( > 0.5 mm from the muscularis mucosae) [2]. With the World Health Organization criteria on gastric carcinoma classification and differentiation [17], gastric adenocarcinoma was divided into five main types: tubular, papillary, mucinous, poorly cohesive, and mixed. Also tabulated and analyzed were other rare variants such as adenosquamous and neuroendocrine carcinomas, carcinoma with lymphoid stroma, mixed mucinous and tubular/papillary adenocarcinoma, and mixed poorly cohesive carcinoma with tubular/papillary adenocarcinoma.

Appropriate immunohistochemical and in situ hybridization tests were employed to confirm the diagnosis of neuroendocrine carcinoma and carcinoma with lymphoid stroma, respectively, on the basis of the initial impression on routine hematoxylin–eosin stained tumor sections [2]. The World Health Organization tumor differentiation grading system on gastric carcinoma was adopted and applied primarily to tubular and papillary adenocarcinomas, in which well differentiated tumors were defined as having well-formed neoplastic glands in over 95% of the estimated tumor volume, while poorly differentiated tumors were defined as having well-formed glands in < 49% of the tumor, and thus moderately differentiated tumors were defined as having a proportion of discernable glands that was in between [17]. Poorly cohesive carcinoma, including signet-ring cell carcinoma, was thus classified as poorly differentiated. In each case, lymphovascular invasion, and perineural invasion were also recorded and analyzed. The tumor pN staging was carried out, according to the American Joint Committee on Cancer staging manual 7th edition [12].

As a routine quality control procedure for this four-center study, cases with equivocal diagnoses were discussed between the study pathologists via teleconferences. The senior study pathologist and organizer (QH) often traveled to each center to help out, periodically audited the preliminary study results, and sometimes requested further investigations of unusual findings reported by study pathologists. Frequent discussions via emails or WeChatTM among study pathologists were regularly conducted to ensure a smooth study process.

Statistical analysis

Differences between the groups with and without lymph node metastasis were statistically analyzed and compared with respect to patient age, gender, tumor gross pattern, size, invasion depth, histology type, differentiation, lymphovascular, and perineural invasion. The χ2, Fisher’s exact, or Kruskal–Wallis H test was utilized, where appropriate. A logistic regression analysis model was used for univariate and multivariate analyses on risk factors of lymph node metastasis. Statistical Package for Social Sciences (SPSS, version 17, Chicago, USA) was employed for all statistical analyses. P values < 0.05 were defined as statistically significant.

Results

Over the study period, a total of 12,050 radical gastrectomies for gastric cancer were reviewed. After exclusion of advanced gastric cancer (N = 9737) and disqualified (N = 281) cases, 2101 consecutive early gastric carcinoma cases were eligible for the study, in which 495 were classified as early gastric cardiac carcinoma (23.6%, 259/2101) and 1606 (76.4%, 1606/2101) were as early gastric non-cardiac carcinoma.

Lymph node metastasis

The average number of lymph nodes retrieved in each case was 17.2 (range: 2–61) for the early gastric cardiac carcinoma group and 18.1 (range: 2–65) for the early gastric non-cardiac carcinoma group. Lymph node metastasis was identified in 308 (14.7%, 308/2101) cases and was significantly more frequent in early gastric non-cardiac carcinomas (17.1%, 275/1606) than in early gastric cardiac carcinomas (6.7%, 33/495) (p < 0.0001). In 33 early gastric cardiac carcinoma cases with nodal metastasis, 26 (78.8%, 26/33), 6 (18.2%, 6/33), and 1 (3%, 1/33) cases were staged as pN1, pN2, and pN3, respectively. As shown in Table 1, there were no significant differences in patient age, gender, and tumor macroscopic growth pattern between early gastric cardiac carcinomas with lymph node metastasis and early gastric cardiac carcinomas without. For tumors with a size of 0.9 cm or smaller, the risk of lymph node metastasis was significantly lower (p < 0.05). In contrast, for tumors measuring 3 cm or greater, the risk of lymph node metastasis was significantly higher (p < 0.0001). There was no significant difference in the risk of lymph node metastasis for tubular adenocarcinoma, papillary adenocarcinoma, mucinous adenocarcinomas, and poorly cohesive carcinoma. Although most early gastric cardiac carcinoma cases with lymph node metastasis (78.8%, 26/33) were tubular (54.5%, 18/33) or papillary (24.4%, 8/33) adenocarcinomas (Fig. 1), mixed poorly cohesive carcinoma with tubular/papillary adenocarcinomas (18.2%, 6/33) were substantial. One case (3.0%, 1/33) with lymph node metastasis was mucinous adenocarcinoma. No lymph node metastasis was found in uncommon types of early gastric cardiac carcinoma, such as neuroendocrine (Fig. 2) and adenosquamous (Fig. 3) carcinomas, carcinoma with lymphoid stroma (Fig. 4), and mixed mucinous and tubular/papillary adenocarcinomas. However, the risk of nodal metastasis was significantly increased in early gastric cardiac carcinomas with mixed poorly cohesive carcinoma and tubular/papillary adenocarcinomas (p < 0.05).

Table 1 Risk factors of lymph node metastasis in early (pT1) gastric cardiac carcinoma
Full size table
Fig. 1
figure 1

Microscopic features of early tubular and papillary cardiac adenocarcinomas. a A poorly differentiated tubular adenocarcinoma exhibited solid and cribriform growth patterns. b The tumor showed papillary morphology with well-defined fibrovascular cores lined by dysplastic columnar cells. c lymphovascular invasion (black arrows) was identified in the submucosa

Full size image
Fig. 2
figure 2

Early gastric cardiac neuroendocrine carcinoma showing submucosal invasion with a sharp invasion front in the submucosal space (a) and demonstrating salt-pepper nuclear chromatin features (b) and immunoreactivity to synaptophysin (c)

Full size image
Fig. 3
figure 3

Early gastric cardiac adenosquamous carcinoma exhibiting a sharp tumor invasion front in the submucosa (a). The squared area in (a) was enlarged in (b) to demonstrate both adenocarcinoma (glands) and squamous cell carcinoma components. Note the comedo necrosis in the squamous cell carcinoma component

Full size image
Fig. 4
figure 4

Early gastric cardiac carcinoma with lymphoid stroma featuring an ulcerated surface, a pushing submucosal invasion border (a), and heavy infiltration of small lymphocytes into poorly differentiated tubular adenocarcinoma (b). Nuclei of poorly differentiated tubular adenocarcinoma were decorated by a positive nuclear in situ hybridization stain for the Epstein–Barr viral antigen (c)

Full size image

As shown in Table 1, well differentiated early gastric cardiac carcinomas exhibited a significantly lower risk of lymph node metastasis (p < 0.0001), whereas poorly differentiated early gastric cardiac carcinomas showed a significantly higher risk of lymph node metastasis (p < 0.0001). The risk of lymph node metastasis was also significantly increased in early gastric cardiac carcinomas with lymphovascular invasion (p < 0.00001), but not with perineural invasion. By a multivariate logistic regression analysis (Table 2), poor differentiation and lymphovascular invasion were identified as significant independent risk factors of lymph node metastasis in early gastric cardiac carcinomas with odds ratios of 6.0 (95% confidence interval: 1.41–25.88, p < 0.05) and 7.6 (95% confidence interval: 2.75–20.24, p < 0.0001), respectively. Among 33 early gastric cardiac carcinoma cases with nodal metastasis, 16 (48.5%, 16/33) showed lymphovascular invasion, 16 (48.5%, 16/33) revealed poor differentiation, 6 (18.2%, 5/33) exhibited both lymphovascular invasion and poor differentiation. Only 7 (21.2%, 7/33) cases did not demonstrate either. Thus, most early gastric cardiac carcinoma cases with nodal metastasis (78.8%, 26/33) displayed either lymphovascular invasion or poor differentiation or both.

Table 2 Independent risk factors of lymph node metastasis in early (pT1) gastric cardiac carcinoma
Full size table

Depth of invasion and lymph node metastasis

In this cohort, the risk of lymph node metastasis was 6.3% (62/982) for intramucosal carcinoma and significantly increased to 22.0% (246/1119) for submucosal carcinoma (p < 0.0001). As shown in Table 3, no lymph node metastasis occurred in intramucosal (M2 and M3 invasion depths) early gastric cardiac carcinomas. Despite more frequent submucosal invasion of early gastric cardiac carcinomas into SM1 (23.2%) and SM2 (37.8%), compared with early gastric non-cardiac carcinomas (19.7% and 31.2%, respectively), the frequency of lymph node metastasis was significantly lower for these 33 early gastric cardiac carcinoma cases with 10/115 (8.7%) for SM1 and 23/187 (12.3%) for SM2, compared with early gastric non-cardiac carcinomas (26.1%, 213/817) with 58/316 (18.4%) for SM1 (p < 0.05) and 155/501 (30.9%) for SM2 (p < 0.0001), respectively.

Table 3 Comparison of relationship of invasion depth and nodal metastasis between early gastric cardiac (EGCC) and non-cardiac (EGNCC) carcinomas
Full size table

Discussion

In this multicenter study of 495 early gastric cardiac carcinoma cases, we showed a lack of lymph node metastasis in intramucosal carcinoma and uncommon types of carcinomas, such as neuroendocrine and adenosquamous carcinomas, carcinoma with lymphoid stroma, and mixed mucinous and tubular/papillary adenocarcinomas, regardless of the tumor gross growth pattern, size, and invasion depth. In contrast, lymphovascular invasion and poor tumor differentiation in early gastric cardiac carcinomas were unveiled as two significant independent risk factors of lymph node metastasis. On the other hand, the tumor gross ulcerative pattern, tumor size > 3 cm, and submucosal invasion were not identified as significant independent risk factors for lymph node metastasis in early gastric cardiac carcinomas. Overall, our data suggest a very low risk of lymph node metastasis in early gastric cardiac carcinomas, as reported previously in single-center studies [2, 8, 18, 19].

Uncommon histologic types of early gastric carcinomas, such as adenosquamous and neuroendocrine carcinomas as well as carcinoma with lymphoid stroma arise primarily in the gastric cardia and share some common pathologic characteristics, such as a large size, submucosal invasion, and a pushing border. Our findings of the lack of lymph node metastasis in those uncommon types of early gastric carcinomas in the gastric cardia parallel to those of previous reports [2, 20, 21], especially for carcinomas with lymphoid stroma that occur mainly in the proximal stomach with Epstein–Barr viral infection and frequent surface ulceration and submucosal invasion [2, 3, 21].

Most gastric cardiac carcinomas correspond to Siewert type II and type III adenocarcinomas of the gastroesophageal junction [13, 22]. Lymph node metastasis in early gastroesophageal junction carcinomas occurs predominantly in the upper abdomen and very infrequently in the lower mediastinum [23,24,25]. Matsuda et al. investigated lymph node metastasis in the lower mediastinum in 15 patients with early gastric cardiac carcinoma (cT1) and reported the absence of lymph nodal metastasis in all sentinel lymph nodes identified in the lower mediastinum (5 patients staged as pT1a, 10 as pT1b). No tumor recurrence occurred in all patients without sentinel nodal metastasis during the 38-month follow-up period [26]. Those results, taken along with our findings of the absence of nodal metastasis in the abdominal lymph nodes in intramucosal and rare types of early gastric cardiac carcinomas lend support to the role of endoscopic therapy, especially by endoscopic submucosal dissection, in early gastric cardiac carcinoma resections as a safe alternative to surgical resection of this early cancer [4, 5].

In early gastric carcinoma, including early gastric cardiac carcinoma, lymphovascular invasion and poor tumor differentiation have been repeatedly demonstrated as the most important risk factors for lymph node metastasis [11, 18, 19, 27,28,29] as confirmed in this study. These features have been incorporated into many practice guidelines as contra-indications of endoscopic therapy [30,31,32]. From the pathologist’s point of view, these two histology features may be diagnosed with a minimal inter-observer variation. Therefore, in a biopsy specimen from the gastric cardia, it is essential for the pathologists to document tumor histology type, size, differentiation, and the presence or absence of lymphovascular invasion, among other parameters, to guide clinical management of patients with early gastric cardiac carcinoma.

The high risk of lymph node metastasis in poorly cohesive carcinoma is well known [11, 27,28,29] but not validated in this study on early gastric cardiac carcinoma. The discrepancy may be related to the small sample size (2.2%, 11/495) for this carcinoma in the current study since poorly cohesive carcinoma arises primarily in the distal stomach and rarely in the cardia [2, 33]. Although we observed only two cases with lymph node metastasis, poorly cohesive carcinoma mixed with tubular/papillary adenocarcinoma components demonstrated a significantly increased risk of lymph node metastasis [34]. Although additional risk factors such as submucosal invasion certainly contribute to this high risk, the underlying complex multidirectional differentiation in mixed poorly differentiated carcinomas may hold the key for tumor spread and lymph node metastasis. This speculation requires further investigation.

Early gastric cardiac carcinoma has a high propensity for submucosal invasion [2, 8, 26, 33]. In an early study in Japan on early gastric cardiac carcinoma, Tanaka et al. reported the presence of submucosal invasion in 88.9% (8/9) cases [8], which is a bit higher than that (61%, 302/495) in our cohort. Despite deeper submucosal invasion, lymph node metastasis was absent in all eight cases in that Japanese study and detected in only 33 (10.9%) cases in this cohort, which was a significantly lower proportion than that for early gastric non-cardiac carcinoma (26.1%, 213/817). Although the reason for the significantly lower risk of lymph node metastasis in submucosal early gastric cardiac carcinoma is unknown, marked fibromuscular hyperplasia in the gastric cardiac region with the frequent presence of “the double muscularis mucosae” is well known and may be one of the plausible defense mechanisms against lymph node metastasis in early gastric cardiac carcinoma [35].

The risk of lymph node metastasis in early gastric carcinoma has been reported not to be gastric region-dependent in many Japanese and Korean studies [11, 27, 36,37,38], which differs from our findings. This difference appears to result from various disease classification systems used in different countries. In Japan and Korea, gastric cardiac cancer is rare [39, 40] and is classified along with upper third gastric cancers [30], which includes cancers in the fundus and proximal corpus as well as in the gastric cardia. From the pathologic point of view, early gastric cardiac carcinoma differs from early gastric non-cardiac carcinoma in the gastric fundus and proximal corpus in multiple ways. First, the cellular origins for early gastric cardiac carcinoma are much more complex and tumors may arise in cardiac, mixed mucoxyntic mucosa, and residual embryonic stem cells, among others [41]. In contrast, the cells of origin for early gastric non-cardiac carcinoma of the fundus and proximal corpus may be related to oxyntic mucosa only. Second, most early gastric cardiac carcinoma tumors are Lauren intestinal-type adenocarcinomas [2, 3, 42], whereas carcinomas in the gastric fundus show considerable proportions of undifferentiated poorly cohesive carcinoma [43]. Third, hyperplastic fibromuscular tissue is present in the gastric cardia but absent from the fundus and corpus. This may explain a lower rate of lymph node metastasis in early gastric cardiac carcinoma compared to early gastric non-cardiac carcinoma, as also reported previously in other studies [2, 8, 18, 19].

There are several strengths of this study: (1) a large number of consecutive early gastric cardiac carcinoma surgical resection cases (N = 495) specifically for investigation of risk factors of lymph node metastasis, which is unprecedented; (2) minimal confounding factors in the study patient population, such as Barret’s adenocarcinoma, which was intentionally excluded; and (3) uniform execution with a stringent investigation protocol by specialized, experienced gastrointestinal pathologists guided with the World Health Organization diagnostic criteria on early gastric carcinoma. An important limitation of our analysis is not unique to our study but commonly associated with any retrospective multicenter histopathology study on lymph node metastasis in gastric cancer: we could not control for the surgical methods of lymphadenectomy, which is often inconsistent among surgeons, surgical centers, and countries. As such, variation in the number of lymph nodes retrieved was unavoidably present. However, the average number of lymph nodes studied in the cohort was high. We are confident that any analysis on the risk of lymph node metastasis in early gastric cardiac carcinoma would have similar results.

In summary, we demonstrated, in this multicenter study, the absence of lymph node metastasis in intramucosal and uncommon types of early gastric cardiac carcinomas, irrespective of tumor gross patterns, size, and invasion depth. Even in early gastric cardiac carcinomas with submucosal invasion, the risk of lymph node metastasis was still significantly lower than that in early gastric non-cardiac carcinoma, especially for early gastric cardiac carcinoma without lymphovascular invasion and poor tumor differentiation.