Characteristics of Metachronous Remnant Gastric Cancer After Proximal Gastrectomy: A Retrospective Analysis

Abstract

Purpose

Despite annual endoscopy, patients with metachronous remnant gastric cancer (MRGC) following proximal gastrectomy (PG) are at times ineligible for endoscopic resection (ER). This study aimed to clarify the clinical risk factors for ER inapplicability.

Materials and Methods

We reviewed the records of 203 patients who underwent PG for cT1 gastric cancer between 2006 and 2015. The remnant stomach was categorized as a pseudofornix, corpus, or antrum.

Results

Thirty-two MRGCs were identified in the 29 patients. Twenty MRGCs were classified as ER (ER group, 62.5%), whereas 12 were not (non-ER group, 37.5%). MRGCs were located in the pseudo-fornix in 1, corpus in 5, and antrum in 14 in the ER group, and in the pseudo-fornix in 6, corpus in 4, and antrum in 2 in the non-ER group (P=0.019). Multivariate analysis revealed that the pseudo-fornix was an independent risk factor for non-ER (P=0.014). In the non-ER group, MRGCs at the pseudo-fornix (n=6) had more frequent undifferentiated-type histology (4/6 vs. 0/6), deeper (≥pT1b2; 6/6 vs. 2/6) and nodal metastasis (3/6 vs. 0/6) than non-pseudo-fornix lesions (n=6). We examined the visibility of the region developing MRGC on an annual follow-up endoscopy one year before MRGC detection. In seven lesions at the pseudofornix, visibility was only secured in two (28.6%) because of food residues. Of the 25 lesions in the non-pseudo-fornix, visibility was secured in 21 lesions (84%; P=0.010).

Conclusions

Endoscopic visibility increases the chances of ER applicability. Special preparation is required to ensure the complete clearance of food residues in the pseudo-fornix.

INTRODUCTION

Proximal gastrectomy (PG) is a function-preserving surgery for early gastric cancer of the upper stomach. The primary benefit of PG is organ preservation, which results in an increase in oral intake, a decrease in body weight loss, and a better quality of life than compared to the outcomes observed in total gastrectomy [1,2,3,4,5].

However, PG has some disadvantages, such as reflux esophagitis, anastomotic stricture, and metachronous remnant gastric cancer (MRGC) [3,6,7,8,9,10,11]. The incidence of MRGC after PG is reported to be 1.7%–9.1%, which is much higher than that observed after distal gastrectomy [6,8,10,12,13]. When MRGC without distant metastasis is detected, complete gastrectomy with nodal dissection is necessary to cure the cancer, except in cases of mucosal tumors with a low possibility of nodal metastasis. However, surgery for MRGC is technically difficult because of severe adhesions after primary surgery, which can cause severe morbidity. Furthermore, completion gastrectomy severely impairs food intake and causes weight loss, which decreases the quality of life. Therefore, it is critical to determine whether endoscopic resection (ER) or surgery is applicable for MRGC. The Japanese Gastric Cancer Association (JGCA) guidelines recommend annual endoscopic surveillance for five years [14,15], which is useful for the early detection of MRGC.

Despite strict annual postoperative endoscopic examinations, patients with MRGC who are not candidates for ER are occasionally found, highlighting the need for a more precise understanding of MRGC after PG. Therefore, the present study investigated the characteristics of MRGCs for which ER after PG was not applicable to improve clinical practice in detecting and managing MRGC in post-PG patients.

MATERIALS AND METHODS

Patients

We retrospectively reviewed the clinical records of 203 patients diagnosed with cT1 gastric cancer located in the upper third of the stomach or esophagogastric junction who underwent R0 PG with lymph node dissection between January 2006 and December 2015 at the National Cancer Center Hospital of Japan.

Surgical procedures

The surgical procedure has been previously described in detail [16,17]. Briefly, the celiac, hepatic, and pyloric branches of the vagus nerve were preserved routinely. Reconstruction was performed in several ways, including jejunal interposition (JI) reconstruction, esophagogastrostomy (EG) with the double-flap technique (DFT), gastric tube reconstruction, double-tract reconstruction (DTR), and EG using Toupet’s method [7,18]. The JI was the first choice during the study period. The DFT was temporarily implemented from 2008 to 2009 [19]. Other methods were selected when the JI was not applicable.

Reconstruction method of JI

After transection of the abdominal esophagus and lymph node dissection, the stomach was divided using a linear stapler (Fig. 1). The transected ends are reinforced with seromuscular sutures. Next, a segment of the jejunum was transected–20–30 cm from the ligament of Treitz. The proximal jejunum was anastomosed to the esophagus using a circular stapler. The jejunal limb was dissected to a length of 6–8 cm. An 8–10 cm jejunal segment was sacrificed to remove the mesenteric tension of the jejunal limb. During jejunogastrostomy, a 2 cm circular hole was made in the anterior wall at the midpoint and 2 cm on the anal side of the gastric transection line. The distal end of the jejunal limb and the remaining stomach were anastomosed using Gambee sutures. This created a pseudofornix. Fig. 1B shows an endoscopic view of the pseudofornix.

Fig. 1. The definition of the location of the remnant stomach. (A) A remnant stomach was divided into three parts: pseudo fornix, corpus, and antrum. (B) Endoscopic image showed pseudo-fornix region (pale blue area).

Reconstruction method of DFT

After PG, an H-shaped seromuscular flap (2.5×3.5 cm) was created on the anterior wall of the gastric remnant, 3–4 cm from the top. The gastric mucosa was opened for anastomosis–8–10 mm from the lower edge of the flap. The posterior wall was anastomosed with a single layer of continuous sutures between the esophagus and stomach. The anterior wall was sutured in layers, starting with a continuous suture between the mucosa, followed by interrupted sutures between the esophageal muscle and the gastric seromuscular layer. Finally, the double flap was closed in a Y-shape with interrupted sutures to cover the anastomosis.

Surveillance and secondary intervention

Annual follow-up surveillance was performed using computed tomography, tumor markers (CEA and CA19-9), and endoscopy. All endoscopies were performed by a highly skilled staff with at least 10 years of experience. When remnant gastric cancers were detected, a secondary intervention was performed according to the latest JGCA gastric cancer guidelines for each period [14,15].

Protocol of preparation before endoscopy

Patients were requested to fast from 14:00 hours after lunch the day before endoscopy, except for water. Just before the endoscopy, the patients ingested 100 mL of water containing 2 mL of simethicone (Gascon®; Kissei Pharmaceutical Co., Tokyo, Japan) and 20,000 units of Pronase (Sigma-Aldrich, St. Louis, MI, USA). All endoscopic examinations were performed between 9:00 and 12:00 hours.

Clinicopathological features of MRGC

The incidence, clinicopathological findings, and treatment outcomes of MRGC were recorded and assessed retrospectively. The MRGC location was divided into three parts: the pseudofornix, corpus, and antrum (Fig. 1A). MRGCs were then divided into two groups depending on the applicability of ER: the ER group and non-ER group. Curative resection was defined in accordance with the JGGA gastric cancer guidelines. Curative resection was achieved for lesions in the ER group achieved curative resection [14,15], while those in the non-ER group did not undergo ER.

The evaluation of the second-to-last endoscopic images

In cases of MRGC, we examined second-to-last endoscopic images and evaluated the visibility of the region where the MRGC was detected during the most recent endoscopic evaluation. The reasons for not observing the area were classified into two categories. One was due to food residue and the other was due to insufficient gastric mucosal extension (Fig. 2).

Fig. 2. The 2 categories for the poor visibility of the region where metachronous remnant gastric cancer was detected at the last endoscopic evaluation. (A) Food residues obscured the observation of mucosal structures at the pseudo-fornix, thus preventing clear visualization. (B) Inadequate distension of the gastric mucosa hindered the complete observation of the mucosal surface at the pseudo-fornix.

Statistical analyses

Categorical variables within each group were analyzed using the chi-square test. However, when more than 20% of the cells displayed an expected frequency of less than 5, the Fisher’s exact test was used. Cumulative incidences were calculated using the Aalen-Johansen method [20]. Differences were considered statistically signicant at P<0.05. Logistic regression analysis was performed to compare the ER and non-ER groups in terms of sex, age, tumor location, circumference, and interval from onset. An age of 65 years and an interval to onset of 60 months were used to divide the two groups for the analysis. Statistical analyses were performed using EZR v1.61 (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria) [21].

Ethical considerations

This study was approved by the National Cancer Center Institutional Review Board (No. 2017–077). All research procedures were performed in accordance with the Declaration of Helsinki and its amendments. Informed consent was not obtained from individual participants included in this study due to the opt-out bias of our institute.

RESULTS

Patient characteristics

The patient backgrounds are presented in Table 1. Males predominated (84%), and the median age at onset of primary gastric cancer was 67 (37–84) years. Most cases (n=190, 94%) were reconstructed by JI. The DFT was performed on 10 patients. Gastric tube reconstruction, DTR, and EG with Toupet’s method were performed in one case each. Most lesions were located in the upper gastric body (n=155, 76%), while others were located at the esophagogastric junction (n=41, 20%). Histologically, intestinal-type gastric cancer was the predominant cancer type (n=143, 70%). pT2 or deeper was included in 10% (pT2, n=12; pT3, n=8) of the cases, and lymph node metastasis was found in 9% (pN1, n=14; pN2, n=4; pN3a, n=1).

Table 1. Clinicopathological features of the patients who received proximal gastrectomy for cT1 gastric cancer from 2006 to 2015 (n=203).

Variables			Values (n=203)
Age at the primary surgery (yr, median (range))			67 (37–84)
Sex
	Male		170
	Female		33
Operation reconstruction
	JI		190
	EG		12
	DTR		1
Pathological features of primary gastric cancer
	Macroscopic type
		0-I	25
		0-IIa	56
		0-IIb	1
		0-IIc	119
		1	2
	Tumor size (mm, median (range))		25 (4–103)
Histological type
	Differentiated		158
	Undifferentiated		40
	Others		5
	Location
		EGJ	41
		U	155
		M	7
	pT
		1a	46
		1b	137
		2	12
		3	8
	pN
		0	184
		1	14
		2	4
		3a	1
	Esophageal invasion		37/166

JI = jejunal interposition; EG = esophagogastrostomy; DTR = double-tract reconstruction; EGJ = esophagogastric junction; U = upper gastric body; M = middle gastric body.

Incidence and characteristics of MRGC

Among the 203 patients, the median observation time after the primary surgery was 72 months (range, 1–168 months). MRGC were detected in 32 lesions in 29 patients. The overall incidence of MRGC was 15.8%. The cumulative incidence was 8.1% (confidence interval [CI], 4.7%–12.7%) at 5 years, and 24.3% (95% CI, 15.5%–34.2%) at 10 years. One patient died due to peritoneal metastasis of the primary cancer and eight died from other causes. Twenty (62.5%) lesions in 18 patients were eligible for ER (ER group), whereas 12 (37.5%) lesions in 11 patients were not eligible (non-ER group). Among the 11 patients in the non-ER group, 4 had received ER at first but resulted in non-curative resection and proceeded to surgery with nodal dissection.

Risk factor for non-ER

The differences in the clinical and histological features of MRGC between the ER and non-ER groups are shown in Table 2. When MRGC developed at the pseudofornix, only one of seven lesions (14.3%) was indicated for ER. In contrast, ER was indicated in the corpus in 5 of 9 (55.6%) patients and in the antrum in 14 of 16 (87.5%) patients. Thus, the dominant location of MRGC differed significantly between the ER and non-ER groups (P=0.019). On multivariate analysis, only tumor location at the pseudo-fornix remained significant (odds ratio [OR], 27.88; 95% CI, 1.958–396.9; P=0.014).

Table 2. Clinical characteristics of the MRGCs after PG that were eligible for ER and not eligible (n=32).

Variables		Non-ER group (n=12)	ER group (n=20)	P-values	Multivariate analysis
					OR (95% CI)	P-values
Tumor location				0.006*	27.88 (1.958–396.9)	0.014
	Pseudo-fornix	6	1
	Corpus	4	5
	Antrum	2	14
Cross-sectional circumference				0.703†	0.531 (0.058–4.844)	0.575
	Posterior wall	5	8
	Lesser curvature	4	5
	Anterior wall	2	4
	Greater curvature	1	3
Sex				0.619	0.353 (0.026–4.733)	0.432
	Male	10	18
	Female	2	2
Age of MRGC (yr, median (range))		69 (45–79)	72 (62–86)	0.058	0.691 (0.108–4.365)	0.691
	<65	4	5
	≥65	8	15
Reconstruction methods				1
	JI	12	19
	EG	0	1
Interval time (mo, median (range))		72 (24–111)	60 (11–132)	0.642	2.103 (0.331–13.35)	0.431
	<60	5	9
	≥60	7	11

Italicized P-values indicate statistically significant results (<0.05).

MRGC = metachronous remnant gastric cancer; PG = proximal gastrectomy; ER = endoscopic resection; OR = odds ratio; CI = confidence interval; JI = jejunal interposition; EG = esophagogastrostomy.

^*Pseudo-fornix vs. corpus and antrum, ^†Posterior wall vs. anterior wall, lesser curvature, and greater curvature.

Characteristics of the tumor located at the pseudo-fornix

As the pseudo-fornix was a risk factor for non-ER, we examined the characteristics of the tumors that developed at the pseudo-fornix compared with the non-pseudo-fornix. The 12 MRGCs in the non-ER group showed clearly different characteristics from the tumors in the pseudo-fornix and non-pseudo-fornix subgroups (Table 3). Six MRGCs in the non-pseudo-fornix were pure differentiated-type early gastric cancers. The histologic non-eligible factors for ER were pT1b2 with lymphovascular invasion (LVI) in 2 lesions (33%) and only LVI in the other 2 MRGCs (33%). Among the two MRGCs for which histology indicated that the ER was suitable, one was surgically resected along with other organs, whereas the other was selected for observational management based on the patient's condition. In contrast, MRGCs in the pseudofornix have been identified as more advanced cancers. Two MRGCs (33%) were unresectable owing to multiple liver metastases. The remaining four MRGCs (67%) were surgically resected and classified as pT1b2N0, pT2N0, pT4aN3a, and pT4bN0. The median interval time tended to be shorter in the pseudo-fornix group than in the non-pseudo-fornix group (60 vs. 78 months). Furthermore, MRGC were not detected in the pseudofornix during the first and second years of surveillance.

Table 3. Comparison of clinicopathological features between pseudo-fornix located MRGCs and non-pseudo-fornix located MRGCs of the non-ER group (n=12).

Variables		Pseudo-fornix (n=6)	Non-pseudo fornix (n=6)	P-values
Interval time (mo)		60.0 (31–96)	78.0 (24–111)	1
Macroscopic type				0.182
	Type 0-IIa/Type 0-IIc	3 (50.0)	6 (100.0)
	Type 2/Type 3	3 (50.0)	0 (0.0)
Tumor size (mm)		36.0 (7–57)	14.0 (9–24)	0.060
Histological type				0.060
	Differentiated type	2 (33.3)	6 (100.0)
	Undifferentiated type	4 (66.7)	0 (0.0)
Tumor depth				0.060
	≤pT1b1(SM1)	0 (0.0)	4 (66.7)
	≥pT1b2(SM2)	6 (100.0)	2 (33.3)
Lymph node metastasis				0.182
	Positive	3 (50.0)	0 (0.0)
	Negative	3 (50.0)	6 (100.0)
Distal metastasis				0.456
	Positive	2 (33.3)	0 (0.0)
	Negative	4 (66.7)	6 (100.0)
Lymphovascular invasion				0.567
	Positive	2 (33.3)	4 (66.7)
	Negative or not evaluated	4 (66.7)	2 (33.3)
Treatment				1
	Surgery	4 (66.7)	5 (83.3)
	Chemotherapy or best supportive care	2 (33.3)	1 (16.7)

Values are presented as number (%) or median (range).

MRGC = metachronous remnant gastric cancer; ER = endoscopic resection.

Endoscopic findings during annual surveillance

Regular endoscopic surveillance was conducted at least once a year for all patients who developed MRGC. Table 4 shows the findings of the second-to-last endoscopic examinations, focusing on the visibility of the region where MRGC was subsequently detected. In cases of MRGCs in the pseudofornix, the endoscopic findings of the region were significantly less visible and contained more food residue than those in the non-pseudofornix.

Table 4. Visibility of the region where MRGC was identified in the latest endoscopy before its detection, along with the influencing factors (n=32).

Variables		Pseudo-fornix (n=7)	Non-pseudo fornix (n=25)	P-values
Visibility of the region				0.010
	Yes	2 (28.6)	21 (84)
	No	5 (71.4)	4 (16)
Food residue				0.047
	Yes	4 (57.1)	4 (16)
	No	3 (42.9)	21 (84)
Gastric mucosal distention for observation				0.001
	Adequate	3 (42.9)	25 (100)
	Inadequate	4 (57.1)	0

Values are presented as number (%).

MRGC = metachronous remnant gastric cancer.

Fig. 3 shows the endoscopic images of the three patients who had tumors at the pseudofornix and had no food residue at the latest endoscopy. In case 1, there was no food residue, and visibility was good at the latest endoscopy. Only the gastric ulcer scar was observed (Fig. 3A). On endoscopy, the tumor was detected, and the patient underwent ER (Fig. 3B). In case 2, there was no food residue, and visibility was good at the latest endoscopy. No tumors were observed (Fig. 3C). The tumor was detected on endoscopy and diagnosed as T2 (Fig. 3D). Both MRGCs in the pseudofornix with good visibility were located near the border of the corpus. In Case 3, there was no food residue, but behind the His angle of the anastomotic site, it was hidden owing to poor distension at the latest endoscopy (Fig. 3E). On endoscopy, the tumor was detected at the site and was diagnosed as T4a (Fig. 3F).

Fig. 3. The endoscopic images of three cases in the pseudo-fornix group were not detected despite the absence food residue. (A) In case 1, there was no food residue and good visibility on the latest endoscopy. Only a gastric ulcer scar was found. (B) At the present endoscopy, a 0-IIa lesion was detected and the patient received endoscopic resection (white arrowhead). (C) In case 2, no food residue was found and the visibility was good on the latest endoscopy. (D) A 0-I+IIc lesion was detected and was diagnosed with T2 at the present endoscopy. (E) In case 3, there was no food residue but behind the His angle of the anastomotic site was hidden possible due to poor distension at the latest endoscopy. (F) In the present endoscopy, a Type 2 tumor was detected at that site and was diagnosed with T4a.

DISCUSSION

In this study, we retrospectively analyzed the characteristics of MRGCs treated with PG for cT1 gastric cancer. Our study clarified that most MRGCs that developed at the pseudofornix had no indication for endoscopic treatment for relatively progressive disease, possibly because of poor visibility on annual follow-up endoscopy after primary surgery, in contrast to those that developed at the corpus and antrum. These results suggest that optimal endoscopic visibility based on complete clearance of the stomach can contribute to the early detection of MRGCs, independent of location.

The incidence of MRGC in our study was 15.8%, which was the highest reported in similar studies [6,8,10,11,12]. This high incidence may be attributed to the fact that endoscopic examinations were performed annually over a five-year observation period [22]. In our study, 60% of MRGC cases were eligible for ER, which is consistent with previous literature reporting rates ranging from 50% to 83.3% [8,9,10,11].

To analyze the factors associated with MRGC in the non-ER group, we focused on the location of MRGC. The remnant stomach was classified into three compartments: the pseudo-fornix, corpus, and antrum. A total of 22% of the MRGCs occur in the pseudofornix. All but one MRGC in the PF region was not eligible for curative ER treatment. Furthermore, multivariate analysis revealed that the pseudo-fornix was an independent factor for non-ER. Ishida et al. [10] divided the stomach into the body and antrum, with 68% of MRGCs observed in the body. Among MRGCs in the gastric body, 29% showed tumor invasion deeper than the submucosal layer. Our results are comparable because the pseudofornix was included in their study.

In the present study, MRGCs in the pseudofornix were larger, showed deeper invasion, and had a higher proportion of undifferentiated histology than those in the non-pseudofornix. These differences can be attributed to two potential reasons: distinct tumor development sites and variations in detection intervals. The pseudofornix is initially created from the gastric corpus [16]. The gastric corpus and antrum are conventional hotspots of gastric cancer, with no marked differences in pathogenesis [23,24,25]. Therefore, another factor (the different timing of detection) could have led to the different outcomes of MRGCs, as the time interval to detect MRGC was slightly longer in the pseudo-fornix than in the non-pseudo-fornix. Furthermore, no lesions were found within two years in the pseudo-fornix group, whereas two lesions (2/25, 8%) were found in the non-pseudo-fornix group.

Kato et al. reported that the missing lesion rate accounted for 10% of synchronous lesions after ER [22]. These data support the hypothesis that MRGC in the pseudofornix cannot be detected until it develops into an advanced cancer. The rate of advanced MRGC was significantly higher with the DTR and long JI methods than with the EG and short JI methods, because it was difficult to reach and survey the remnant stomach [6,26]. In the present study, five of seven tumors detected in the PF region were not observed in the corresponding area during the previous year's endoscopy. Suzuki et al. conducted annual surveillance after ER and examined cases in which curative ER was deemed inappropriate [27]. They found that the area where the cancer was located could not be observed in 35% of the cases, suggesting that a lack of observation, not oversight, was primarily responsible for ER ineligibility.

In our study, the observation of the gastric mucosa was difficult in some cases because of the presence of food residue, which occurred with all reconstruction methods. The incidence in cases reconstructed using JI was reported to be 8.5%–31.8%, while those with EG and DTR were 21.8% and 45.8%, respectively [17,18,28]. Several strategies have been reported to reduce food residue and improve delayed gastric emptying, such as vagus nerve preservation, creation of a short jejunal limb, pyloroplasty, and ensuring an optimal remnant gastric volume [13,16].

In the present cohort, we selected a fasting protocol that excluded dinner on the previous day, in addition to breakfast on that day. The fasting period in this protocol is longer than that required for screening endoscopy. However, food residue was still present in some patients in this cohort. A longer fasting period has been previously reported to decrease food residue levels [29]. To completely clear the food residue, other methods have been used, such as drinking 1 L of water at night on the day before examination, taking peristalsis-promoting drugs on the day before examination, or erythromycin infusion just before examination [30,31,32]. However, these studies predominantly examined patients who underwent distal gastrectomy. Further research is required to tailor preparation protocols for patients following PG. Additionally, gastric mucosal distension occurs due to inefficient air accumulation in the remnant stomach. Reflux into the jejunal limb and the strict gastric suture line that forms the pseudo-fornix may restrict mucosal visualization. It is crucial to scrutinize the pseudofornix region and prevent missing new MRGC, including changing the patient's position and attempting to empty the stomach. Additionally, the examiners should not hesitate to reschedule examinations by severely restricting the food intake.

The present study has several limitations. First, most cases involved reconstruction using an open JI. In recent years, the number of laparoscopic and robotic PG has increased, with EG becoming an increasingly common reconstruction method. However, our findings remain applicable as the creation of the pseudofornix and the proportion of food residue are comparable across different reconstruction methods. Second, the study was conducted at a single institution. Multicenter studies have high generalizability; however, their results may vary in testing accuracy among centers [8,10,22]. In contrast, our single-institution approach with dedicated endoscopists experienced in cancer treatment conducting annual surveillance ensured the high accuracy of our findings. Despite our meticulous surveillance protocol, the presence of MRGCs that are ineligible for ER indicates that annual endoscopy alone is inadequate. Third, our sample size was small. Therefore, the present results must be confirmed in a multicenter study with a larger cohort.

In conclusion, most MRGCs that developed at the pseudofornix had no indication for endoscopic treatment for relatively progressive disease, possibly because of poor visibility on annual follow-up endoscopy after primary surgery, in contrast to the findings for lesions in the corpus and antrum. Annual endoscopy in post-PG patients should be performed using a special preparation protocol that involves elimination of the food residue.