Early gastric cancer (EGC) is defined as gastric cancer that is confined to the mucosa or submucosa (T1 cancer), irrespective of the presence of regional lymph node metastases.1 Currently, almost 10 000 cases of EGC are being detected every year in Japan, corresponding to 40–50% of all gastric cancers.2 Endoscopic mucosal resection (EMR) is widely accepted as a standard treatment for EGC with nominal risk of lymph node metastasis, as it is minimally invasive, safe and convenient.3 4 At present, the standard guideline criteria for EMR, which were established by the Japanese Gastric Cancer Association, have been generally accepted, and they state that: (1) elevated EGCs <2 cm in diameter and (2) small (⩽1 cm) depressed EGCs without ulceration are absolutely indicated for EMR.5 However, the conventional snaring procedure is not reliable for lesions >20 mm in diameter or lesions with ulcer findings.6 7 Conventional EMR is associated with a high risk of local recurrence (range, 2–35%) in such cases, especially when resections are not accomplished en bloc or the margins are not clear.8

Endoscopic submucosal dissection (ESD) has been developed to dissect directly along the submucosal layer using specialised devices, including an insulation-tipped diathermy knife (IT knife).9 10 Preliminary studies have been published showing the advantage of ESD over conventional EMR for removing larger or ulcerated EGC lesions in an en bloc manner.6 11 12 Thus, ESD allows precise histological assessment of the resected specimens, and it may prevent residual disease and local recurrence.4 7 8 Although the early results of ESD are promising, except for a higher risk of procedure-related complications,8 11^–13 the long-term outcomes of EGC patients treated by ESD remain unknown. We sought to evaluate the feasibility and efficacy of ESD for treatment of EGC and to conduct a follow-up study on clinical outcomes, including tumour recurrence and disease-specific and overall survival, in a large consecutive series.

PATIENTS AND METHODS

Patients

A total of 713 EGCs in 661 consecutive patients were treated by ESD at the hospitals of Nagasaki University School of Medicine from January 2001 to December 2007. The patients were enrolled based on the expanded criteria proposed by Gotoda et al,4 14 wherein differentiated gastric cancers (well and moderately differentiated tubular adenocarcinoma and papillary adenocarcinoma) with no lymphatic–vascular involvement correlating with a nominal risk of lymph node metastasis were defined as: mucosal cancer without ulcer findings irrespective of tumour size; mucosal cancer with ulcer findings ⩽3 cm in diameter; and minute (<500 μm from the muscularis mucosae) submucosal invasive cancer ⩽3 cm in size. EGCs which did not fall into one of the categories were excluded from the study and the patientts were urged to receive a gastrectomy with removal of lymph nodes. Written informed consent was obtained from all patients before ESD.

ESD

EGCs were first identified and demarcated using white-light endoscopy and chromoendoscopy with indigo–carmine solution, and then marking around the lesions was carried out with spotty cautery using a needle-knife. Glycerol (10% glycerol and 5% fructose; Chugai Pharmaceutical, Tokyo, Japan) was then injected into the submucosal layer to lift the mucosa. A circumferential mucosal incision was made around the lesion using the IT knife (Olympus Optical, Tokyo, Japan). Submucosal dissection was performed for complete removal of the lesion using the IT knife and a Hook knife (Olympus). High-frequency generators (ICC200 or VIO 300D; ERBE Elektromedizin, Tübingen, Germany) were used during marking, incision of the gastric mucosa and exfoliation of the gastric submucosa. ESD was performed by experienced gastrointestinal endoscopists (HI, SS, NY, KI, HN and KO), and there were no statistically significant differences in their outcomes of ESD.

All patients were sedated by intravenous injection of 5–7.5 mg of diazepam (Cercine; Takeda Pharmaceutical, Osaka, Japan) and 15 mg of pentazocine (Pentazin; Daiichi-Sankyo Pharmaceutical, Tokyo, Japan), and 2.5 mg of diazepam was additionally given for conscious sedation as needed throughout the procedure.

Procedure-related bleeding after ESD was defined as bleeding that required transfusion or surgical intervention, or bleeding that caused the haemoglobin level to fall by 2 g/dl.13 Perforation was diagnosed endoscopically or by the presence of free air on an abdominal plain radiograph or CT scan.

Histopathological evaluation

The location of EGC was classified into the upper, middle and lower thirds of the stomach. The macroscopic type of EGC was divided into the elevated type and the flat/depressed type. The excised specimens were sectioned perpendicularly at 2 mm intervals. Histology was classified into differentiated adenocarcinoma (well or moderately differentiated adenocarcinoma or papillary adenocarcinoma) or undifferentiated adenocarcinoma (poorly differentiated adenocarcinoma or signet-ring-cell carcinoma). The tumour size, depth of invasion, presence of ulcerative changes, lymphatic and vascular involvement, and tumour involvement to the lateral and vertical margins were assessed.

En bloc resection refers to a resection in one piece.13 When the lesion had to be removed in multiple segments, the piecemeal-resected specimens were reconstructed as completely as possible. The curability of ESD was classified as either curative or non-curative.13 15 Resections were deemed curative when the removal was achieved with tumour-free lateral and vertical margins, and there was no submucosal invasion deeper than 500 μm from the muscularis mucosae and no lymphatic and vascular involvement. Non-curative resection was defined as one that did not meet the curative criteria or when compartments of undifferentiated carcinoma were found.

Follow-up

Endoscopic examinations were scheduled at 1, 3, 6 and 12 months after ESD and annually thereafter. Biopsy specimens during each follow-up endoscopy were taken from the treatment-related scar or any other suspicious abnormalities to assess the presence of local recurrent tumour or metachronous cancer of the stomach. To detect lymph node and distant metastases, contrast-enhanced CT and ultrasound sonography of the abdomen, and chest x rays were performed annually.

Procedure-related mortality was defined as any death within 30 days after ESD.13 The cumulative disease-specific and overall survivals were estimated.

Statistical analysis

The significance of differences in patients’ characteristics and clinicopathological features was determined using Fishers exact test, the χ2 test, the Mann–Whitney U test, or Student t test, as appropriate. Factors associated with resectability (en bloc or piecemeal) and curability of ESD were analysed using logistic regression analysis. Odds ratios (ORs), together with 95% CIs, were calculated to estimate the relative risk of piecemeal resection or non-curative resection and their associations with various parameters. Data for the long-term outcomes were calculated using the Kaplan–Meier method and analysed by the log rank test. p Values <0.05 were considered statistically significant.

RESULTS

Clinicopathological characteristics

Of the 661 patients with 713 EGCs treated by ESD, 20 patients were diagnosed as having undifferentiated cancer based on the resected specimens and were excluded from this study. An additional 59 patients were excluded because of detection of lymphatic and/or vascular invasion. Twenty-four patients with massive submucosal invasion, 10 patients with minute submucosal invasion >30 mm in size and 11 patients with intramucosal cancer and ulcer findings >30 mm in size were also excluded. Thus, 551 patients, with a total of 589 EGC lesions, were enrolled in the study and received endoscopic follow-up for 6–89 months (median, 30 months). There were 76 patients with multiple simultaneous EGCs at the time of ESD. The patients’ median age was 72 years (mean, 71 years; range 38–92 years), and the male/female ratio was 2.6:1 (396:155).

Clinicopathological characteristics of the enrolled 589 EGCs are summarized in table 1. The median tumour size was 17 mm, and the mean size was 20 mm, ranging from 2 to 75 mm.

Resectability and curability of ESD

On the whole, en bloc resection was achieved in 559 of 589 EGCs (94.9%). Table 2 shows the association of clinicopathological characteristics of the EGC lesions and resectability. Based on the multivariate logistic analysis, tumour size was significantly associated with resectability of ESD (p<0.05); larger EGCs were at higher risk of piecemeal resection.

Of the 589 lesions, 8 could not be evaluated for curability due to difficulties in histopathological assessment; these were attributable to the burn effect or insufficient reconstruction of the piecemeal fragments. Thus, 550 (94.7%) of 581 lesions were defined as curative. Table 3 shows the association of various factors with curability of ESD. On univariate analysis, resectability and the presence of ulcer findings had a significant impact on ESD curability (p<0.001 and 0.05, respectively). On a multivariate basis, however, piecemeal resection was the sole significant contributor to non-curative resection (p<0.001). Of the patients with non-curative resections, eight underwent gastrectomy with removal of lymph nodes. Of those, four had local recurrence, while the remainder had no residual or recurrent tumour and no lymphatic involvement. Four patients with non-curative resections were successfully treated by argon plasma coagulation for margins with cancerous glands, and two underwent repeat ESD with one-piece curative resection.

Complications

Procedure-related bleeding was seen in 10 patients (1.8%). All haemorrhagic episodes were successfully treated by endoscopic clipping or coagulation. Perforations related to ESD occurred in 25 patients (4.5%) and could also be managed by conservative medical treatment after endoscopic closure with clipping. Procedure-related bleeding was not associated with any clinicopathological characteristics, including age, gender, tumour size, tumour location and macroscopic appearance of EGC. On the other hand, the upper location and tumour size had impact on ESD-related perforation based on the univariate analysis, while on the multivariate analysis these factors were not significantly associated with ESD-related perforation. There were no treatment-related deaths (procedure-related mortality rate, 0%).

Clinical outcomes of ESD

Of the 551 EGC patients treated by ESD, 75 with a follow-up period of <1 year were excluded from the analyses of clinical outcomes including local recurrence, metachronous gastric cancer development, and disease-specific and overall survival. Thus, 510 lesions in 476 patients were eligible for the analyses, including 481 lesions with curative resection and 29 with non-curative resection. Local recurrence of the tumours occurred in four lesions from four EGC patients; one (0.2%) was from a curative but piecemeal resection, and the other three (10.3%) were from a non-curative resection. There was a significant difference in the recurrence rate between the curative and non-curative groups (p<0.001), though the follow-up duration for each was not statistically different (median follow-up period, 29 and 32 months, respectively). The recurrent tumours were situated in the middle (3) or lower (1) third of the stomach, and the sizes of the tumours varied from 16 to 28 mm. Tumour recurrence was seen from 13 to 24 months after ESD, and each patient underwent gastrectomy with D2 lymph node dissection. All recurrent tumours were limited to the mucosal layer without lymphatic and vascular involvement, and no lymph node metastases were observed in any of the cases.

Metachronous gastric cancers that were not local recurrences developed in 13 patients with curative resection and one with non-curative resection after ESD, with a median follow-up period of 14 months (range, 12–42 months). No other patients had metastases to either the lymph nodes or distant organs such as the liver and lungs during the study period.

Among the 476 patients enrolled for the analyses of clinical outcomes, two patients whose vital status was unknown and six patients who underwent gastrectomy were excluded from the disease-specific and overall survival analysis, and thus 468 patients treated by ESD were eligible for the survival analyses. The 3- and 5-year overall survival rates were 98.4% and 97.1%, respectively (fig 1A). The causes of death, which were confirmed by death certificates, included two hepatocellular carcinomas, two oesophageal cancers, one lung cancer, one urinary bladder cancer and two heart diseases. None of the patients died of gastric cancer during the present study, and the 3- and 5-year disease-specific survival rates were 100%. Next, we compared the prognosis of the patients with EGCs that did not meet the guideline criteria but fulfilled the expanded inclusion criteria (expanded group) with that of the patients with the lesions that fulfilled the standard guideline criteria (standard group). Both the en bloc resection and curative resection rates were significantly lower for EGCs in the expanded group than for the standard group lesions (p<0.05 for each, table 4). The expanded group lesions were at significantly higher risk of ESD-associated perforation (p<0.05, table 4). However, the incidences of both local tumour recurrence and metachronous gastric cancer development were comparable between the standard and expanded group (table 4). Of note, overall survival was adequate irrespective of the indication criteria (fig 1B). The 3-year overall survival rates were excellent: 99.3% in the standard group and 97.2% in the expanded group. The 5-year overall survival rates were 97.1% and 97.2% in the standard and expanded group, respectively; the difference between the groups was not significant. The disease-specific survival rates were 100% in both the groups.

DISCUSSION

Despite the increasing use of ESD for EGC, the clinical outcomes have not been fully evaluated. In the present larger consecutive series, the 3-year overall survival after ESD was excellent, with the rate being nearly 99%. In a multicentre study of endoscopic resection for EGC, Oda et al reported a comparable 3-year overall survival between the EMR and ESD groups (99.7% and 98.5%, respectively).16 The 5-year survival rate after ESD reached 97.1% in our patients, which was equivalent to those after EMR documented in previous reports.17 18 Of note, both the 3- and 5-year disease-specific survival rates after ESD were 100%, similar to those after EMR in 12 major Japanese institutions.17 Moreover, the prognosis for patients in the expanded group was similarly excellent compared with that for patients in the standard group (fig 1B), albeit that the resectability and curability of ESD were significantly lower in the expanded group lesions. This is the first study to clarify the prognosis of EGC patients after ESD, although the follow-up periods are still limited compared with those following surgical treatment. EGC has excellent clinical outcomes, with 10- and 20-year survival rates after gastrectomy with removal of lymph nodes as high as 95%.17 Confirmation of whether ESD can equal surgery will require further long-term prospective studies.

The present results confirmed that local recurrence of gastric cancer after ESD was nominal when curative resection was achieved, whereas about 10% of patients with non-curative resection had local recurrence. This implies that EGC patients with non-curative ESD require close follow-up surveillance for cancer recurrence for at least 2 years, as the recurrent tumours developed 13–24 months after ESD. On the other hand, patients with curative ESD may also require periodic endoscopic surveys for new EGC lesions. In the present study, 14 metachronous gastric cancers (13 in the curative and 1 in the non-curative resection groups) developed 12–42 months after ESD. Similarly, another study of the clinical outcomes of ESD for EGC documented 14 (6.2%) metachronous lesions among 225 EGCs at unknown stages following ESD.7 The incidence of metachronous gastric cancer varies from 1.8% to 8.1% after EMR,17 19 and, thus, the necessity for continued surveillance is an intrinsic drawback of endoscopic therapies, irrespective of curability.

Takenaka et al analysed the clinicopathological factors associated with local recurrence in 225 EGCs after ESD, and tumour size (>30 mm) and tumour location (upper) were the significant predictors for local recurrence.7 On the other hand, in the present study, no recurrent tumours were located in the upper third of the stomach and exceeded 30 mm in size. The reason for this discrepancy is unclear, but the larger size of EGCs and upper third location may substantially affect the curability of ESD in their series, resulting in a higher incidence of local recurrence in such cases. Gastrectomy with removal of lymph nodes has been recommended in cases of recurrent gastric cancer after ESD,8 and we managed them by additional surgery. All recurrent tumours in the present study were, however, mucosal cancers without lymph node metastasis, suggesting the potential for additional ESD or other local treatment. In fact, ESD allows en bloc resection for locally recurrent EGC even after prior EMR, with rates from 89.1% to 93.1%.6 15

En bloc resection of ESD provides much higher curative resection rates than piecemeal resection. Using logistic regression analysis, we assessed the impact of various factors on the curability of ESD. On a univariate basis, piecemeal resection and ulcer findings interfered with curative resection. Oka et al also showed that ulceration prevented complete removal of EGCs, notably in lesions larger than 21 mm.6 A similar tendency was seen in EGC patients treated by ESD for recurrent EGC after previous EMR.15 Nevertheless, multivariate analysis revealed that en bloc resection was the sole significant contributor to curative ESD in the present study. En bloc resection permits precise histopathological examination of curability in order to guide further management and to stratify a patient’s risk of developing metastases, offering successful outcomes following ESD.4 8

Oda et al13 reported that upper and middle location, tumour size >21 mm and positive ulcer findings were associated with piecemeal resection.13 Oka et al reported a marked decrease in en bloc resection rates from 92.9% for EGC without ulcer findings to 19.2% for tumours with ulceration.11 Only 9.1% of the ulcerative lesions >21 mm in size were resected en bloc.11 In the present series, on the multivariate analysis, there was a significant association of tumour size with lesion resectability. Collectively, larger and/or ulcerative EGCs could be at higher risk of piecemeal resection; therefore, their treatment requires a high level of expertise and experience. Understanding the characteristics of lesions with an increased risk of piecemeal resection may help in the complete reconstruction of the multiple segments using endoscopic pictures and clue markings prior to ESD.

ESD still has relatively high complication rates, with 1.8% and 4.5% for procedure-related bleeding and perforation, respectively, in the present study. Oda et al reported that post-ESD bleeding was seen in 6% of cases, and it has a significant association with upper third location,13 though procedure-related bleeding was not associated with any clinicopathological characteristics in the present series. Since the lesions in the upper third tend to be more tangential to endoscopes, it may be rather difficult to recognise the lesions en face during treatment and to access them anatomically in certain parts of the upper stomach, leading to more frequent bleeding during and after the procedures. In addition to these factors, positive ulcer findings could be related to a higher risk of perforation.13 On the other hand, no significant factors related to ESD-related perforation were found in the present multivariate analyses.

In conclusion, ESD has an advantage over EMR for treatment of EGC by facilitating en bloc resection. It is feasible to assess the histopathological curability of the resected specimens precisely, reducing local recurrence. The relatively long-term outcomes may be excellent irrespective of the inclusion criteria, but continued surveillance is necessary for recurrent tumours in cases of non-curative resection and for metachronous cancers even after curative ESD.