Baseline Demographic and Clinical Features
A total of 570 patients had undergone ESD and 495 patients with 502 lesions were enrolled with the following criteria, among whom 265 patients with 271 lesions were enrolled in the AI group, 137 patients with 137 lesions in the EI group and 93 patients with 94 lesions in the BEI group. 5 patients were withdrawn during the follow-up. 450 patients were enrolled for the long-term outcome analysis in the end (Supplementary Figure 3).
As shown in Table 1, the average age was 65.5 (range, 44-87) and 82% were male. 291 patients (58.8%) had a smoking history, 263 patients (53.1%) had a drinking habit and 230 patients (46.5%) regularly ate pickled food. Nearly half of the patients (40.8%) had a family tumor history. The most common complication was hypertension (32.7%, 162/495). 14 patients (2.8%) suffered from gastroesophageal reflux disease.
Endoscopic and Pathological Characteristics
The average tumor size was 12 mm (range, 2-20) in the AI group, 23 mm (range, 6-66) in the EI group, and 27 mm (range, 5-65) in the BEI group (P<0.001) (Table 2). Most lesions were smaller than 20 mm (73.7%, 370/502). The most common site of EGCC was posterior curvature (51.0%, 256/502), followed by lesser curvature (42.2%, 212/502). The most common endoscopic infiltration growth pattern was INFa (66.3%, 333/502). The most common macroscopic pattern, in the descending order, was 0-IIc (45.2%, 227/502), 0-IIa+IIc (27.9%, 140/502), 0-IIa (14.9%, 75/502), 0-IIb (10.0%, 50/502), 0-III (1.2%, 6/502), and 0-I (0.8%, 4/502). All tumor lesions in the AI group were intramucosal and differentiated. Instead, in the BEI group, 5 patients (5.3%, 5/94) had undifferentiated carcinoma and the majority (94.6%, 89/94) had submucosal invasion (P<0.001). 24 patients (4.8%, 24/502) had ulceration over the three group and 13 patients (13.8%, 13/94) in the BEI group were positive. Focal distal esophageal involvement was detected in 12.0% (60/502) in all patients. Lymphovascular invasion was identified in 3.0% (15/502) in all patients. Gastritis cystica profunda was found in over 24% of lesions (122/502) and there were no differences among the three groups. Overall positive rate of helicobacter pylori was 56.8% and the difference among the three groups was non-significant. Atrophic gastric carditis was found in 89.4% (449/502).
Short-term Outcomes and Complications based on indication
As table 3 clarified, the overall en bloc resection rate was 99.8% (501/502), among which only one lesion in the BEI group was broken. Complete resection rate of all cases was 94.4% (474/502); 99.3% (269/271), 98.5% (135/137) and 74.5% (70/94), respectively (P<0.001). The rate of curative resection was 79.9% in all the patients: 98.5% (267/271), 97.8% (134/137) and 0% (0/94), respectively (P<0.001). Factors associated non-curative resection included lymphovascular infiltration (n=15), submucosal invasion (n=71), positive vertical margin (n=4), positive lateral margin (n=21), or positive lateral and vertical margin (n=3).
In the aspect of complications, 12 cases had significant bleeding, in which 8 cases were early delayed bleeding, 4 cases were late delayed bleeding. The early delayed bleeding rate was 0.4% (1/271), 1.5% (2/137), 5.3% (5/94) in the AI, EI and BEI group, respectively, which was a difference of statistics among the three groups (P=0.004). This could be due to the larger size of tumor lesions and submucosal invasion. Luckily, all the patients who suffered from bleeding were successfully managed by endoscopic hemostasis (n=11) and the use of hemostatic drugs such as octreotide acetate and somatostatin (n=1). Perforation was found in 1 AI patient and this patient's symptoms improved after conservative treatment. Stenosis was a common complication of patients who underwent ESD and was found in 3.6% (18/502) patients, among whom 4 patients (1.5%, 4/271) were in the AI group, 10 patients (7.3%, 10/137) belonged to the EI group and 4 patients (4.4%, 4/94) were members of the BEI group (P=0.002). After balloon dilation, the symptoms of stenosis were easily improved. The presence of stenosis was positively related to a circumferential extent of the mucosal defect of > 3/4 or longitudinal extent of > 5 cm in length according to the previous investigation[18]. The median time of ESD operations was 65.3 minutes (range 10-246), while the BEI patients required longer procedure time, which was 72.9 minutes (range 18-223) (P=0.007). The average hospitalization time was 6.6 days (range 2-19). Similarly, the EI and BEI patients stayed longer than the AI patients (P=0.012).
Long-term therapeutic outcomes according to the indication
During the median follow-up of 48.1 months (range 18-101), 45 patients (9.1%, 45/495) were lost during follow-up. Additional surgery was carried out for 7 AI patients (2.9%, 7/239), 10 EI patients (8.1%, 10/124) and 42 BEI patients (48.3%, 42/87) (Table. 3). Local recurrence was found in only one AI patient because of the residual tumor (P=1.000). Synchronous cancer was detected in 39 patients (8.7%, 39/450), and there was no statistically significant difference among the three groups. The similar conclusion was achieved for metachronous cancer. During the follow-up, 2 BEI patients (2.3%, 2/87) were found to have lymph node metastasis (P=0.037). These patients underwent the additional surgery and survived well during long-term follow-up. Distant metastasis was developed in one AI patient and two BEI patients (Table.4). They didn’t receive the additional surgery and passes away unfortunately. Disease-specific death occurred in 4 patients (0.9%, 4/450); 1 (0.4%, 1/239) in the AI group, no patient in the EI group and 3 (3.4%, 3/87) in the BEI group (P=0.030). The AI patient was dead because of metastasis whose lateral margin was positive and he refused the additional surgery. One BEI patient was dead due to the significant bleeding after the additional surgery. The other two patients passed away related to distant metastasis without additional surgery (Table 5). Five-year overall survival rates were 96.1% (AI), 98.3% (EI) and 89.1% (BEI), which didn’t have any statistic difference (P=0.180) (Fig.1 A). Five-year disease-specific survival rates of three groups were 99.6%, 100% and 96.6% in the AI, EI and BEI group, respectively, which was significantly different between each other (P=0.016) (Fig.1 B). Besides, we analyzed whether the additional surgery was necessary for the patients in the BEI group. After ESD surgery, 93 patients were enrolled into the BEI group based on postoperative pathology. 6 patients were lost during the follow-up. 42 of the remaining 87 BEI patients underwent additional surgery according to the doctor's advice. As we can see in Fig.1 C and D, the additional surgery didn’t have an influence on neither the overall survival nor the disease specific survival.
We wondered why the disease specific survival was no significantly difference between the two BEI group with or without the additional surgery. So, we applied the eCura system to evaluate whether there was any difference between these two group. According to the generally accepted method, we divided the BEI patients with or without additional surgery into three categories (Supplementary Table 1). The detailed scores had no difference between the two groups (Fig.1 E), which meant the LNM risk was resemble between the BEI patients with or without surgery, consistent with the survival analysis. In low risk category, disease specific survival was higher in the patient accepted the additional surgery (100% vs 97.1%), but was no significant difference for the two groups (p=0.389; Fig.1 F). In the intermediate risk category, we could achieve the similar conclusion: the patients received the additional surgery has higher five-year disease-free survival rate, but there was no significantly different (92.9% vs 88.9%, p=0.830; Fig.1 G). Since there was no disease-related death of the patients in the high-risk category and a small number of cases, we could not achieve a reliable survival analysis for the data (Fig.1 H). In summary, the patients in the BEI group may not benefit from the additional surgery after ESD in our research due to the small account of data.