Treatment of early oesophageal and gastric cancer

Definition of early oesophageal and early gastric cancers

This chapter deals with the management of early upper gastrointestinal neoplasms, which is described in two parts: early oesophageal cancer (part A) and EGC (part B).

Risk and development of early oesophageal cancer

Oesophageal cancer is the seventh most common cancer globally, with approximately 572 000 new cases and 509 000 deaths in 2018. The two main histologic subtypes are squamous cell carcinoma and adenocarcinoma. Worldwide, squamous cell carcinoma is most prevalent. Squamous cell carcinoma can develop in any part of the oesophagus but is mostly located in the middle third and develops from the stratified squamous epithelium of the oesophagus. In high-income countries, major risk factors are heavy drinking and smoking. Risk factors for high-incidence countries in parts of Asia and Africa have not been clarified yet and are thought to include environmental and dietary factors, drinking very hot beverages, and genetic factors. Adenocarcinomas account for the majority of cases in Western countries and the incidence continues to increase, whereas the incidence of squamous cell carcinoma is expected to decrease. Major risk factors for developing adenocarcinoma are obesity and gastro-oesophageal reflux disease (GORD), which are also risk factors for the development of metaplastic columnar epithelium (Barrett’s oesophagus). Metaplasia, in Barrett’s oesophagus or in the intestinal columnar mucosa in the stomach or gastro-oesophageal junction, may progress to increasing grades of dysplasia and eventually to invasive carcinoma. Rarely, adenocarcinomas occur in the proximal or mid oesophagus in the absence of Barrett’s oesophagus, and are thought to develop from (sub)mucosal glands or ectopic columnar epithelium (especially the cervical inlet patch).

Classification of early oesophageal cancer

Early oesophageal cancer is confined to the mucosa (T1a) or extends into the submucosa (T1b) and is considered eligible for local endoscopic removal ( Fig. 7.1 ). Early carcinomas comprise a spectrum from very subtle carcinomas to deep submucosal tumours. While subtle carcinomas can be difficult to detect, deep submucosal carcinomas are easier to identify, but assessment of endoscopic resectability can sometimes be difficult. Detection and early assessment of oesophageal lesions are performed with endoscopy using white light and (virtual) chromoendoscopy such as narrow-band imaging (NBI). Endosonography can be performed on indication to determine lymph node status. Histopathologic evaluation from endoscopically resected specimens is considered as the optimal staging step to determine the indication for subsequent therapy.

Endoscopic appearance of early lesions

Superficial neoplastic lesions in the gastrointestinal tract are classified by their endoscopic appearance according to the Paris classification ( Fig. 7.2 ). Superficial lesions are classified as protruding (Paris 0–I), elevated (0–IIa), depressed (0–IIc), or excavated and often ulcerated lesions (0–III).

The Paris classification has been developed to unify the description of superficial lesions within the oesophagus, stomach, and colon. ^, The Japanese Society for Esophageal Diseases originally developed morphological criteria for cancers of the digestive tract and type 0 lesions were defined as endoscopically superficial cancers.

The Paris classification distinguishes three main subtypes within the type 0 category: protruding lesions (0–I), non-protruding non-excavated lesions (0–II), and excavated lesions (0–III) or ‘ulcers’ (see Fig. 7.2 ). The protruding types protrude above the mucosal surface and may be pedunculated (0–Ip) or sessile (0–Is). Pedunculated polyps are narrow at the base, whereas sessile polyps have a width at the base equal to that of the top. Non-protruding lesions are classified as slightly elevated (0–IIa), flat (0–IIb), slightly depressed (0–IIc), or excavated (0–III). Depending on the depth of the depressed lesion relative to the adjacent mucosa, the lesion is classified as type 0–IIc or type 0–III. The cut-off is at 1.2 mm in the columnar epithelium of the oesophagus and in the stomach, and at 0.5 mm in the stratified epithelium of the oesophagus. Lesions that have both elevated and depressed components are classified into two groups: depressed lesions, in which most of the surface is depressed and there is elevation in a portion of the peripheral ring are classified as 0–IIc + IIa, while elevated lesions with a central depression encircled by the elevated ring at the periphery are called 0–IIa + IIc. The combined patterns of excavation and depression are called 0–III + IIc or 0–IIc + III, depending on the relative surface area of the ulcer and of the depressed area. The classification helps to predict the extent of invasion into the submucosal layer and thus the choice between endoscopic or surgical treatment, which is applicable to all neoplastic lesions throughout the gastrointestinal tract ( Table 7.1 ). Type 0–I and 0–III are more at risk of submucosal invasion and thus not good candidates for endoscopic resection, in contrast to types 0–IIa, 0–IIb, and 0–IIc.

Endoscopic staging before endoscopic resection does provide useful information on the prediction of invasion into the proper muscle layer. Lesions that are confined to the mucosa with or without the submucosa tend to move over the peristaltic waves, whereas peristaltic waves appear to curve around tumours that have invaded the proper muscle layer. The latter provides a strong argument against endoscopic resection. In most endoscopic resection techniques, submucosal injection of fluid is used to lift the early cancer from the proper muscle layer. This method has three major benefits: 1) it provides information on invasion depth and thus endoscopic resectability; 2) it facilitates endoscopic resection by increasing the visibility of and access to the submucosal resection plane; and 3) it provides a safety fluid cushion for resecting the superficial lesion without damaging the deeper layers when using snares, knives, or electrocautery. The amount of lifting provides information on the invasion depth of the lesion. Mucosal or superficial submucosal lesions usually demonstrate complete lifting (m–sm1), whereas the lesions that infiltrate into the deeper submucosal layers often lift incompletely (sm2–sm3). A non-lifting sign most often represents invasion deeper than sm3.

Other features that predict deeper invasion are the size of the lesion, poor tumour differentiation grade, and lymphovascular invasion (LVI). The histological features can be obtained through simple biopsy.

Chromoendoscopy

Chromoendoscopy is a technique that is used to highlight neoplastic changes in the mucosa that are less apparent when white light is used. This is traditionally done by using dye spray to stain the mucosa during endoscopy.

Chromoendoscopy with Lugol’s iodine solution is traditionally used to outline neoplastic areas in the squamous epithelium that can be recognised as Lugol-voiding lesions. The Japanese 2020 guideline advises to use iodine solution at a low concentration of ≤ 1% to prevent damage to the epithelium. For Barrett’s oesophagus, acetic acid can be used as stain, leaving dysplastic or cancerous areas more reddish than the surroundings. Virtual chromoendoscopy has become available as an alternative option for the use of stains. Examples are NBI, i-Scan, and Fujinon Intelligent Chromoendoscopy (FICE). These techniques employ a narrow-band optical filter using a spectrum corresponding with the absorption of haemoglobin to visualise the microvascular mucosal structure, which facilitates identification of abnormal microvessels.

In a per-patient analysis, the sensitivity and specificity of Lugol chromoendoscopy were 92% and 82%, respectively, and the sensitivity and specificity of NBI were both 88%. NBI is generally preferred since it has a shorter examination time and has no side-effects. In patients with many multiform Lugol-voiding lesions, however, NBI has a higher chance of misdiagnosis of squamous cell carcinoma than with Lugol chromoendoscopy, as was reported in one randomised non-inferiority trial.

Endosonography

Routine staging with endoscopic ultrasonography (EUS) for early lesions is not recommended, since EUS is not sufficiently accurate to distinguish T1a from T1b or T1 from T2. A meta-analysis assessing the diagnostic accuracy of EUS to differentiate T1a from T1b adenocarcinomas and squamous cell carcinomas reported a large degree of heterogeneity due to the site of the lesion, the histology, the frequency of the ultrasound probe, the method of use (radial versus linear or both), and the experience of the operator. For T1a staging the pooled sensitivity was 85% and pooled specificity was 87%. The pooled sensitivity and specificity for T1b staging were both 86%.

In a setting of patients with Barrett’s oesophagus who were referred for high-grade dysplasia or adenocarcinoma, a meta-analysis showed that performance of T staging with EUS to identify ≥ T1sm1 was suboptimal, with a sensitivity of 56%, specificity of 89%, positive predictive value of 63% and negative predictive value of 85%. In a subsequent meta-analysis of the same research group, EUS was found to over-stage 9% of patients with Barrett’s oesophagus and suspected early carcinoma and to under-stage another 9%. Over-staging of pT1a carcinomas has been reported to occur in 20–40%, as reported in two cohort studies including squamous cell carcinoma patients. ^, In recent patient series including two prospective cohort studies, cT2-staged tumours by EUS have been reported to be overstaged in 30–40%. To prevent over-staging and consequent overtreatment in cT2-staged lesions by EUS, endoscopic reassessment by an expert interventional endoscopist is recommended for cT2N0 adenocarcinomas.

EUS with fine-needle aspiration (FNA) is advised to detect lymph node metastases when submucosal disease is suspected or when pT1b is histopathologically confirmed after staging endoscopic resection. ^{, ,} EUS outperforms positron emission tomography (PET) and computed tomography (CT) for clinical T staging and has similar accuracy compared to PET and CT for clinical N staging. EUS has a pooled sensitivity for detecting N+ of 71% (95% confidence interval (CI) 49–87%) and a pooled specificity of 94% (95% CI 89–97%) in patients with high-grade dysplasia or adenocarcinoma in Barrett’s oesophagus. Patients with invasive submucosal or T2 disease being considered for oesophagectomy should undergo complete staging with EUS and diagnostic CT and PET-CT.

Revised Vienna classification

Large discrepancies between Western and Japanese pathologists in the diagnostic criteria for adenoma, dysplasia, and carcinoma in the gastrointestinal tract have led to considerable problems in the comparison between Western and Japanese data. This led to a consensus meeting in 1998 in Vienna, ultimately resulting in the Vienna classification of gastrointestinal epithelial neoplasia and the revised Vienna classification in 2002. This classification allows not only for a more universal nomenclature of gastrointestinal epithelial neoplasia, but also corresponds more properly with clinical management. The revised Vienna classification is shown in Table 7.2 . Histology from biopsies and resection specimens are nowadays classified according to the revised Vienna classification.

Endoscopic treatment versus surgery

The decision for endoscopic treatment or surgery depends on the extent of the disease, the risk of the individual patient to develop lymph node metastases, and the operative risk of the individual patient. In early cancers, endoscopic treatment is considered the therapy of choice for patients in whom the risk of lymph node metastases is below the mortality risk of oesophagectomy, taking patient preferences and patient performance status into consideration.

Risk of nodal involvement

For adenocarcinoma and squamous cell carcinoma, differences exist in the risk of developing nodal metastases ( Table 7.3 ). The risk of lymph node metastases increases with the depth of tumour invasion in the oesophagus. T1a carcinoma is intramucosal carcinoma and can be subclassified according to the level of invasion into the epithelium (m1), the lamina propria (m2), or the muscularis mucosae (m3) (see Fig. 7.1 ). For assessment of the extent of submucosal invasion in T1b carcinomas, a pragmatic approach is to divide the submucosa into three layers of equivalent thickness: superficial (sm1), middle (sm2), and deep (sm3). However, after endoscopic resection, the submucosal divisions cannot be obtained from the resected specimen since the full submucosa is not available for examination. The depth of tumour invasion is obtained by measuring from the lower limit of the muscularis mucosae in microns (μm). Pathologists should be aware of the finding of a duplication of the muscularis mucosae in Barrett’s oesophagus to avoid overestimation of the invasion depth. The cut-off between sm1 and sm2/3 is defined at 200 μm for squamous neoplasia and at 500 μm in Barrett’s neoplasia. Most international guidelines advise oesophagectomy at an invasion depth more than m3 in squamous carcinoma and more than m3 or sm1 in adenocarcinoma.

Specimens of endoscopic resection should be assessed according to standardised protocols, reporting on adverse prognostic features that could change patient management, including the depth of invasion, tumour size, tumour differentiation grade, margin status and presence of LVI.

Risk of lymph node metastases in T1a

Many studies have been conducted to estimate the risk of lymph node metastases related to the extent of tumour invasion depth. It is likely that the rates of risk of lymph node involvement as reported based on surgical series have been subject to overestimation. In these time periods, the estimation of exact tumour invasion in surgical specimens did not change patient management. Also, histopathologic analysis of surgical specimens is usually performed at slices of 5 mm, whereas endoscopically resected specimens are assessed at 2-mm intervals. As a consequence, reported estimates of lymph node metastases in surgical series might actually correspond to more deeply invading tumours and thus the true risk of lymph node metastases may be lower.

High-grade dysplasia or tumours not infiltrating deeper than the epithelium (m1) in Barrett’s oesophagus or in the squamous epithelium have not been associated with the development of nodal disease. ^, The occurrence of lymph node metastases with m2 squamous cell carcinoma or adenocarcinoma is rare, ^, but has been reported up to 5.6% in a large, single-centre surgical series. With squamous cell carcinoma invading up to m3, the node-positive rate has been reported to be 8.8% in a meta-analysis. For mucosal adenocarcinomas the overall rate of lymph node metastases ranges between 0% and 3%. ^{, , ,}

Risk of lymph node metastases in T1b

The risk of lymph node metastases in early oesophageal cancer increases with the depth of tumour invasion ( Table 7.3 ) but is also dependent on the presence of additional risk factors. Several studies highlight the reasons for the difference in recommendations based on the presence of (histopathological) features when it comes to management of the disease. ^,

A recent meta-analysis evaluated 20 Asian studies including 3 983 patients who underwent surgery for early squamous cell carcinoma. The positive lymph node rate was 23% for tumours in sm1, 29% for sm2, and 48% for sm3. Risk factors for lymph node metastases were tumour size > 20 mm, Paris type 0–I and 0–III tumours, poor tumour differentiation, deeper submucosal invasion depth, and LVI. Age, sex, and tumour location were not statistically significant risk factors for the development of lymph node metastases. The association between LVI and lymph node metastases was also reported in a meta-analysis including 4 749 patients with squamous cell carcinoma or adenocarcinoma from 23 studies in the USA, Europe, and Asia from 2000 to 2018 (odds ratio for lymph node metastases of 5.72 for patients with LVI compared to patients without LVI).

In a large retrospective cohort study of 1 283 patients with T1a or T1b adenocarcinomas, a scoring system for the presence of lymph node metastases was developed based on tumour invasion depth, tumour size, and LVI. A subgroup of T1a adenocarcinomas could be defined with a high risk (> 15%) of nodal metastases, occurring in approximately 5% of the T1a tumours. Also a subgroup of T1b adenocarcinomas with low risk (< 5%) of nodal metastases could be defined, occurring in approximately 10% of the T1b tumours. In another internally validated prediction model for T1b adenocarcinomas, the risk to develop lymph node metastases was associated with the presence of LVI, every 500-μm increase in invasion depth and every 10-mm increase in tumour size. Depending on the combination of these factors, lymph node risks ranged between 5.9% and 70%.

Several series that included patients who underwent endoscopic resection have reported lower lymph node metastasis risks for submucosal adenocarcinoma than previously reported in surgical series. In low-risk T1b sm1 adenocarcinomas, defined as tumours with R0 resection margins after endoscopic resection, with well or moderate tumour differentiation and without LVI, the risk for lymph node metastases was less than 2%, ^, whereas high-risk sm1 disease is seen with a rate of 9% lymph node metastasis. In T1b sm2–3 adenocarcinomas, the overall rate of lymph node metastasis is 0–22% for sm2 and 0–36% for sm3. ^, In subgroups of low-risk sm2 and low-risk sm3, the positive lymph node rates were 0–8% and 25–29%, respectively. In subgroups of high-risk sm2 and high-risk sm3, the lymph node rates were 28–36% and 37–38%, respectively. In these patients the risk of lymph node metastases clearly exceeded the mortality rate of oesophagectomy, which is estimated to be 1.7–3%. ^,

For patients with an T1bN0M0 adenocarcinoma, strict follow-up after endoscopic resection might be a feasible option. In these patients, surgery might be safely postponed until lymph node metastases have been detected during follow-up examinations. This strategy has been investigated in a multicentre, multinational prospective study and the results are pending at the time of writing (PREFER study, NCT03222635).

Endoscopic resection

There are two commonly used methods to remove gastrointestinal neoplastic lesions. The first is endoscopic mucosal resection (EMR) and is based on the principle of creating a ‘pseudopolyp’ with the use of suction. This ‘polyp’ can be resected with a snare instrument. The second method is endoscopic submucosal dissection (ESD), in which the mucosa can be carefully dissected with the use of an endoscopic knife. EMR or ESD allow for thorough histopathologic assessment of the resected early cancer with the possibility of avoiding subsequent surgery when the removed lesion is found to be within accepted criteria.

Endoscopic mucosal resection

The principle of EMR is the removal of a mucosal lesion by resecting it from its deeper layers using a snare instrument ( Fig. 7.3 ). This method does not allow for lesions larger than 2 cm to be removed en bloc. Larger lesions can be removed by EMR, but only in a piecemeal fashion. This technique is fundamentally different from ESD, where the submucosal layer is carefully dissected in a stepwise manner. Using EMR, early neoplasia is often lifted from the proper muscle layer before resection by using different solutions of saline or viscous fluids for submucosal injection. The lesions can then be sucked into a cap that is placed at the tip of the endoscope with a snare preloaded into the rim of the cap. After sucking the lesion into the cap, the snare is pulled. The content of the snare is then resected using a high-frequency current. An alternative resection method that is frequently used is the so-called ‘band-and-cut’ method (video reference: https://youtu.be/WltDtBfKCUk ). A lesion is sucked into a modified multiband ligator, and by ligating the mucosa bearing the lesion, a pseudopolyp is created that can be resected using a snare. This multiple-band mucosectomy allows for larger segments of neoplasia to be completely resected. This method is mostly used in early Barrett’s-related cancer but is also applicable in the cardia and antrum of the stomach or the rectum.

Endoscopic submucosal dissection

ESD was originally developed in Japan for the local treatment of superficial EGC limited to the mucosal layer or with minimal invasion into the submucosal layer. The main goal of submucosal dissection is to retrieve the lesion en bloc for histopathological staging and to minimise the risk of local recurrence. ESD is performed in several steps ( Fig. 7.4 ). First, the lesion is delineated by placing circumferential dots using electrocautery around the lesion, with a few millimetres of free margin. The lesion is then lifted from the proper muscle layer by submucosal injection in the same fashion as in EMR. The solutions used are stained with either indigo carmine or methylene blue. This dye will colour the submucosal layer and facilitate recognition of the separate layers and blood vessels. After lifting the lesion, a full circumferential incision of the mucosal layer is made, placing the markers just inside the circumferential cut. From this point onwards the submucosal layer is carefully dissected from the muscle layer, often with additional submucosal injections. Different approaches have been described using a tunnelled approach underneath the lesion in order to leave the lesion in its anatomical position until the final stage of the resection. This allows for better orientation and overview in the narrow lumen of the oesophagus. After markers have been placed, a distal incision is created first, followed by a proximal incision, and a submucosal tunnel dissection is performed towards the distal incision (video reference: https://youtu.be/OIM3iJ33LR0 ). The lateral circumference of the lesion is then used as a tether to keep the lesion in place and is cut in the final stage after complete submucosal dissection.

Different speciality knives have been developed for ESD. A breakthrough in ESD was the development of the insulated tip (IT) knife in 1996. An insulated small ceramic sphere is mounted on the top of a high-frequency needle knife, allowing for safe and easy incision and separation of the mucosal and submucosal layers. Subsequently, the design of the original IT knife has been further adapted, leading, for example, to the IT-2 knife, the hook knife, the flex knife, the triangular tip (TT) knife, and the hybrid knife. This hybrid technique combines electrocautery with a distance-dependent water-jet application, allowing for easier, faster, and safer submucosal lifting and cutting in ESD without having to change instruments. ^,

ESD is recognised as a challenging procedure reserved for centres with experience in the technique. ESD needs more operator skills, has a much longer learning curve, takes more time, and is more expensive than EMR. Common difficulties for ESD in the oesophagus include working within the tubular structure of the oesophagus and the delineation of neoplastic lesions, the circular cutting, and the stepwise and partial lifting of the lesion in an area with movement from the heartbeat and breathing. Over the last few years, more and more experience has been obtained with ESD in American and European countries. Results regarding the (oncological) safety of ESD have been reported to comply with quality benchmarks.