Esophagectomy

CRITICAL ELEMENTS

Resection of the Primary Tumor to Negative Margins
Lymph Node Dissection
Gastrointestinal Reconstruction
- Construction of Gastric Conduit
- Performance of a Gastric-Emptying Procedure
- Alternate Conduits
- Cervical Versus Intrathoracic Anastomosis
- Hand-Sewn Versus Stapled Anastomosis
- End-to-End Versus End-to-Side Anastomosis

1. RESECTION OF THE PRIMARY TUMOR TO NEGATIVE MARGINS

Recommendation: To achieve an optimal oncologic outcome, the proximal, distal, and circumferential esophageal margins should be free of cancer both macroscopically and microscopically. Although most surgeons attempt to divide the esophagus 5 to 8 cm proximal and 5 cm distal to the gross tumor to maximize the likelihood of a microscopically negative longitudinal margin, a firm recommendation in this regard cannot be made based on existing data. To ensure a microscopically negative proximal margin, intraoperative pathologic evaluation of the margin should be considered unless resection of additional proximal esophagus would be impossible. Intraoperative assessment of the gastric margin should also be considered when the microscopic status of the margin is of concern. Skeletonization of the esophagus should be avoided to maximize the likelihood of a microscopically negative circumferential margin as the status of this margin cannot be effectively assessed intraoperatively. In all cases, the goal of negative margins should be balanced with the ability to complete the operation safely and to reestablish gastrointestinal continuity successfully.

Type of Data: Retrospective data.

Grade of Recommendation: Strong recommendation, low-quality evidence.

Rationale

Multiple factors may affect the survival of patients who undergo esophagectomy. Among them, the status of the proximal, distal, and circumferential surgical margins is recognized as particularly important.¹ Because the esophagus lacks a serosal layer, cancer of the esophagus can infiltrate into important adjacent structures such as the pericardium, heart, great vessels, trachea, and lung. Cancer of the esophagus is also characterized by microscopic submucosal extension, which may not be apparent grossly at the resection margins. To achieve an optimal oncologic outcome, proximal, distal, and circumferential margins must be free of cancer both macroscopically and microscopically (R0). However, no scientific evidence or published guideline supports the acquisition of any length of tissue between cancer cells and the inked margins beyond an R0 resection. Likewise, there is little evidence to support one surgical technique over another to ensure macroscopically and microscopically negative margins.

Proximal Margin

Few articles report on the prognostic value of the status of the proximal resection margin. Mirnezami et al¹ showed that the incidence of microscopically positive proximal/distal resection margins was 7% and was a robust predictor of survival. Barbour et al² retrospectively examined reports of 352 patients with Siewert I, II, or III adenocarcinoma in an effort to determine whether length of esophageal resection or operative approach influenced outcome. Patients had undergone complete gross resection without neoadjuvant therapy. Proximal margin lengths were recorded ex vivo as the distance from the gross tumor edge to the esophageal transection line. Their results indicated that a 3.8-cm (approximately 5 cm in situ) clear proximal margin was an independent prognostic factor. Mariette et al³ showed that, of 94 patients who underwent esophagectomy, 8 had a positive proximal resection margin with a median survival of 11.1 months versus a survival of 36.3 months in the group that had a negative proximal margin (P = 0.02). There was infiltration of tumor cells in two patients with proximal margin >6 cm, but no infiltration was observed in patients with margins >7 cm. Tam et al⁴ reported a prospective study on 95 esophagectomies and found a 20% anastomotic recurrence when the clear margin was <5 cm, an 8% recurrence rate when the clear margin was between 5 and 10 cm, and 0% recurrence when the clear margin was >10 cm. Hennessy and O’Connell⁵ reported a series of 62 esophagectomies in which a clearance of 5 cm was attempted and found that the macroscopically negative proximal resection margin was microscopically positive in 15% of cases. Migliore et al⁶ performed a systematic review that included many of the studies described above. Their analysis determined the optimal proximal margin should be at least 8.5 cm from the apex of the tumor.⁶

Based on these data, most surgeons generally attempt to achieve a gross proximal margin of 5 to 8 cm in an attempt to maximize the likelihood of a microscopically negative margin (Fig. 7-1). However, in all cases, the surgeon must balance the goal of
a negative margin with the ability to complete the operation safely and to reestablish GI continuity. In this regard, intraoperative pathologic evaluation of the proximal margin should be considered unless resection of additional proximal esophagus would be ill-advised.

FIGURE 7-1 A proximal distance of 5 to 8 cm and distal length of 5 cm from the respective tumor edge maximizes likelihood of a microscopically negative margin.

Distal (Gastric) Margin

Even fewer authors have considered the significance of a positive distal resection margin. In 2007, DiMusto and Orringer⁷ demonstrated that 1.8% of 1,044 patients undergoing transhiatal esophagectomy had a positive distal resection margin. They showed that transection of the esophagus 4 to 6 cm distal to the palpable esophageal mass led to a histopathologically negative margin in 98% of cases. Casson et al⁸ argued that to consistently achieve a negative distal resection margin, at least 5 cm of macroscopically normal foregut below the distal margin of the primary tumor is necessary (Fig. 7-1). They concluded that there was a trend toward reduced postoperative survival for patients with a histologically positive distal resection margin, in particular for patients with involvement of the cardia (median, 5.7 vs 15.4 months if negative; P = 0.0001).

Intraoperative frozen-section evaluation of the distal gastric margin should be considered when the status of the margin is of concern based on intraoperative findings or tumor location. In such cases, the surgeon should be prepared to perform a proximal or total gastrectomy and have a strategy for an alternate conduit (see “Alternate Conduits”).

Circumferential (Radial) Margin

The College of American Pathologists (CAP) defines a positive circumferential margin as the presence of esophageal cancer directly at the inked margin. However, the United Kingdom Royal College of Pathologists (RCP) defines a positive circumferential margin as the presence of esophageal cancer at or within 1 mm of the inked margin.⁹,¹⁰ A meta-analysis of 14 cohort studies, including 3,566 patients with resectable esophageal cancer, found the overall 5-year mortality rates were higher for patients with a positive circumferential margin compared with rates of those with a negative circumferential margin.¹¹ Rates of margin positivity were 15.3% and 36.5%, according to the CAP and RCP criteria, respectively. Overall, 5-year mortality rates were significantly higher in patients with a positive circumferential margin compared with those who had a negative circumferential margin according to both (odds ratio [OR], 4.02; P <0.001) and RCP (OR, 2.52; P <0.001) criteria. Wu et al¹² performed a systematic review and meta-analysis to assess the prognostic significance of a positive circumferential margin on overall survival in patients with esophageal cancer and found that a positive circumferential margin was associated with a lower survival rate in patients with stage T3 disease and in patients receiving neoadjuvant therapy.

Because a negative circumferential margin has an impact on survival of patients with esophageal cancer, one would reason that surgery should be conducted by using a wider resection that includes the healthy tissue surrounding the lesion. An en bloc esophagectomy, as classically defined by Altorki et al,¹³ was developed on this principle. In this procedure, the tumor is excised within a wide envelope of surrounding tissue, including both pleural surfaces laterally and the thoracic duct posteriorly. For tumors that traverse the hiatus, the envelope of tissue includes a 2.54-cm (1-inch) circumferential cuff of diaphragm as well.¹³ Lymph node dissection includes all lymphatic and areolar tissue between the tracheal bifurcation and the celiac axis, including the retroperitoneal, common hepatic, and splenic nodal tissue.¹³ The GI tract is divided with 8- to 10-cm proximal and distal margins.¹³

Altorki et al¹³ and Skinner et al¹⁴ compared results of en bloc esophagectomy versus standard esophagectomy in patients with stage III disease. These studies showed that a standard esophagectomy had a narrower circumferential margin, a less radical lymphadenectomy, and a lower median survival (standard, 12 months; en bloc, 27 months; P <0.001).¹³,¹⁴,¹⁵,¹⁶ Some studies have reported reduced local recurrence with en bloc (1%,¹⁷ 8%,¹⁸ 5%¹⁹) in comparison with standard esophagectomy (21%,²⁰ 25%,²¹ 35%,²² and 38% with surgery alone when performed as described in the Chemoradiotherapy for Oesophageal Cancer Followed by Surgery Study [CROSS]²³). However, there is significant selection bias at play and potential stage migration. This remains a controversial topic, and insufficient data exist to support a specific technical approach. At present, attention should be paid to excision of a generous circumferential margin by avoiding skeletonization of the esophagus, especially at the level of the tumor.

Technical Aspects

During mobilization, the gastrohepatic ligament is incised, taking a wide breadth to maximize the circumferential margin. The peritoneum overlying the hiatus is incised and the crural pillars and esophagogastric junction identified. Some surgeons advocate taking a few millimeters or more of the diaphragmatic hiatus, at least including the peritoneal lining of the crura, again for a wide circumferential margin. The esophageal mass is palpated and assessed for resectability and mobility to ensure that it is not fixed to the prevertebral fascia, aorta, or surrounding mediastinal structures. The stomach is further mobilized (Fig. 7-2A and B; also see Construction of Gastric Conduit).

A lymphadenectomy at the celiac axis is performed (see Lymph Node Dissection). Retractors can be placed into the hiatus to allow ligation of the periesophageal tissues to the level of the carina under direct vision. At least 10 cm of the distal esophagus can be mobilized under direct vision. This is also a critical step in preserving the circumferential margin. The maneuver is facilitated by retracting the esophagus from one side to the other in the lower mediastinum to create tension on the tissues.

Here exists a critical point in the operation. Local invasion into the diaphragm or nonvital mediastinal structures (pericardium, thoracic duct) should prompt surgical resection of these structures to achieve a negative circumferential margin. If the tumor is immobile and/or fixed to surrounding structures (aorta, airway, prevertebral fascia), then resection should be aborted. The esophagus is divided 5 to 8 cm from the apex of the tumor, either sharply or by firing a stapler (Fig. 7-1), and the margins are inspected grossly. Areas for concern on margins may be sent for immediate histopathologic evaluation at the time of operation to ensure an R0 resection, as detailed previously.

2. LYMPH NODE DISSECTION

Recommendation: Radical dissection of all the lymphatic tissue in the periesophagogastric, mediastinal, and celiac axis distributions should be performed regardless of current stage, use of induction therapy, or surgical approach.

Type of Data: One randomized controlled study, multiple retrospective cohort studies.

Grade of Recommendation: Strong recommendation, moderate-quality evidence.

Rationale

Lymph node involvement in the setting of resectable esophageal cancer has been shown to be an indicator of poor prognosis. To date, the majority of data on extent of lymph node dissection and number of lymph nodes removed for esophagectomy for esophageal cancer have been generated from large retrospective case series by highvolume centers. The only randomized data available are a study by Hulscher et al,²⁴ who compared transthoracic esophagectomy to transthoracic esophagectomy with extended en bloc lymphadenectomy in a study of 220 randomly assigned patients. This trial demonstrated that even with an en bloc lymphadenectomy the increased number of nodes dissected did not result in a statistically significant improvement in disease-free or overall survival. Other large, single-center experiences have been able

to consistently demonstrate 5-year survival rates between 25% and 30%.²⁵,²⁶,²⁷ These series mostly involve two-field (thoracic and abdominal) lymph node dissections or sampling through either a transhiatal or transthoracic approach. Altorki et al²⁸ demonstrated that a three-field approach with associated three-field lymph node dissection (cervical, thoracic, and abdominal) can be safely performed with low mortality and 5-year survival rates approaching 50%. These outstanding results have not been replicated in other centers, limiting the applicability of these data.

FIGURE 7-2 Circumferential dissection plane of the esophagus during esophagectomy for malignancy. A: Axial view of computed tomography scan (top) and illustration (bottom) show tumor in the esophagus and the suggested resection plane (black dotted line) to facilitate negative radial margin.

FIGURE 7-2 (continued) B: Illustration shows suggested resection plane to include mediastinal and periesophageal tissue but to avoid trauma to the adjacent structure, such as the airway, to obtain negative radial margins.

The number of nodes needed to resect for optimal prognostic value has also been intensely debated, and recently Rizk et al²⁹ attempted to quantify the number of nodes that should be resected by reviewing the Worldwide Esophageal Cancer Collaboration database, which contains patient data on esophagectomy for cancer from a consortium of institutions. Based on their review of data from more than 4,000 patients, they recommended that the number of nodes resected be based on the T stage of the cancer: 10 nodes for pT1, 20 nodes for pT2, and >30 nodes for pT3/4.²⁹ This study was limited in that there was no standardized protocol for extent of lymphadenectomy and it was highly variable. There was also no standardized protocol for pathology review, so the true number of lymph nodes resected may not have been known for many cases.

Given the heterogeneity in technique of lymphadenectomy, in addition to the optimal number of lymph nodes resected and the specific lymph node stations dissected,
consensus regarding the optimal approach to and extent of lymphadenectomy has not yet been achieved.

Despite the suggested “minimum” number of lymph nodes, the current recommendation by the National Comprehensive Cancer Network (NCCN) for resected esophageal adenocarcinoma with positive lymph node involvement in patients who have been treated with neoadjuvant chemotherapy and radiation is surveillance, regardless of the number of resected lymph nodes. For those patients who have not received preoperative therapy, the recommended treatment is postoperative chemotherapy and radiation. Patients with resected squamous cell cancer who have positive lymph nodes at time of resection are treated with surveillance, regardless of whether the patient has received neoadjuvant therapy.

Technical Aspects

There are three primary nodal fields from which lymph nodes may be excised during esophagectomy: the abdominal field, the mediastinal/thoracic field, and the cervical field. The lymph node fields correspond to the course of the esophagus, distally from the abdomen, through the mediastinum, and into the neck. The extent of dissection and the number of fields examined are very much dependent on the operative approach. In the following section we describe the anatomic considerations for each of the lymph node fields (Fig. 7-3).

FIGURE 7-3 Regional lymph node stations for staging esophageal cancer. Cervical (blue), mediastinal/thoracic(yellow) and abdominal (green) lymph nodes. Two field lymphadenectomy includes the mediastinal/thoracic and abdominal lymph nodes while three field lymphadenectomy includes the cervical, mediastinal/thoracic and abdominal lymph nodes. The eighth edition of the AJCC uses the number of metastatic lymph nodes from any of the fields as the basis for N staging. A = left sided, B = right sided, C = anterior

Abdominal Field

The abdominal lymph node field consists of a broad range of nodes that extend along the lesser curve of the stomach, which include nodes along the celiac artery, left gastric artery, splenic artery, and common hepatic artery. During gastric mobilization, nodes along the retroperitoneum and the diaphragm should be included in the dissection.

Mediastinal/Thoracic Field

The mediastinal/thoracic field includes the nodes along the length of the thoracic esophagus and within the mediastinum. Included within this dissection are the subcarinal nodes as well as nodes associated with the thoracic duct. Paratracheal and paraesophageal lymph nodes above the carina may also be resected from the right thoracotomy approach.

Cervical Field

The addition of this “third field” was first suggested by Japanese surgeons who discovered that up to one-third of patients had recurrences in this lymph node station after a two-field lymphadenectomy. The cervical field includes the nodes along the upper esophagus and nodes that course along the recurrent laryngeal nerve as it moves into the neck. Nodes along the jugular chain and supraclavicular region are also included in this dissection.

3. GASTROINTESTINAL RECONSTRUCTION

3A. CONSTRUCTION OF GASTRIC CONDUIT

Recommendation: Existing evidence does not support gastrointestinal reconstruction using any specific width of gastric conduit. Irrespective of the specific technique chosen, the submucosal collateral blood supply to the gastric conduit should be preserved to minimize the risk for anastomotic leak.

Type of Data: Retrospective studies.

Grade of Recommendation: Weak recommendation, low-quality evidence.

Rationale

The stomach is the most common and reliable conduit for reconstruction after esophageal resection. Attention to the technical aspects of conduit formation may minimize the complications of anastomotic leak and conduit necrosis. Several critical steps in gastric conduit preparation can be used to minimize ischemia, anastomotic leak, and conduit necrosis as well as to obtain optimal function of the conduit.

Vascular Anatomy of the Gastric Conduit

To achieve an adequate lymphadenectomy when performing an esophagectomy for an esophageal or gastroesophageal junction malignancy, the surgeon must divide the left gastric artery at its origin from the celiac trunk. Division of the short gastric arteries and the left gastroepiploic vessel is also required to facilitate mobilization
of the stomach into the chest or neck for anastomosis with the proximal esophagus. Therefore, with two of the four major vessels supplying the stomach divided, the most proximal end of the mobilized stomach is always relatively ischemic. Previous anatomic studies investigating the vasculature of cadaveric adult stomachs with angiography have demonstrated that the arterial supply to the gastric conduit is derived entirely from the caudal end of the conduit.³⁰,³¹,³²,³³,³⁴,³⁵ The right gastroepiploic arcade provides the vast majority, if not all, of the arterial inflow to the conduit. Both the celiac axis through the gastroduodenal artery and the superior mesenteric artery through the pancreaticoduodenal arteries supply the gastroepiploic artery. The right gastric artery also provides inflow to the conduit from the celiac axis through the hepatic artery. The left gastroepiploic arcade and short gastric vessels arise from the splenic artery. Retrograde flow from intramural and highly variable anastomoses between the right and left gastroepiploic arcades may play a role in perfusion of the cranial aspect of the gastric conduit.³²,³⁵

The optimal morphology of the gastric conduit remains a topic of debate. The entire stomach may be used as an esophageal substitute, or the stomach can be fashioned into a “wide” or “narrow” gastric tube. Anatomic cadaveric studies indicate that whole-stomach conduits have a more robust vascular supply because of preservation of intramural collateral flow. The definitions of wide and narrow in terms of conduit diameter or width remain ill-defined and vary in the literature. When a gastric tube is created (Fig. 7-4), the width of the conduit is typically measured from the greater curve to the line of division of the stomach on the lesser curve. A tube that measures <4 cm in width may disrupt intramural gastric collateral circulation in some patients sufficiently to compromise flow to the cranial third of the conduit. Resection of a portion of the lesser curve that maintains a conduit width of at least 4 to 5 cm preserves reliable collateral flow that conduit perfusion is maintained. Therefore, a narrow conduit may be best defined as a width <4 cm, whereas a wide conduit would measure at least 4 to 5 cm from the greater curve.

Leakage from the esophagogastric anastomosis remains one of the most common and potentially devastating major complications reported after esophagectomy.
Technical errors and poor perfusion of the gastric conduit at the anastomotic site likely account for the majority of these leaks. Based on the previously mentioned anatomic studies, inadequate perfusion of the gastric conduit and resulting anastomotic leaks may be related to the manner in which the conduit is fashioned.³⁰,³¹,³²,³³,³⁴,³⁵ However, only a limited number of studies have assessed the potential association of gastric conduit width and anastomotic leak.

FIGURE 7-4 A conduit width that is <4 cm in diameter increases the preservation of submucosal blood supply and viability.

Two prospective randomized studies assessed the association of gastric conduit width and anastomotic leak, and both failed to note a significant difference between wide and narrow conduits. Tabira et al³⁶ carried out a prospective randomized trial comparing use of a subtotal stomach (n = 22) with a narrow gastric conduit (n = 22) for reconstruction after esophagectomy for the management of esophageal cancer. A McKeown approach with retrosternal passage of the conduit and a onelayer, hand-sewn cervical anastomosis was performed in all patients. The subtotal stomach conduit was created by resecting the proximal half of the lesser curvature and cardia, which likely results in a wide gastric conduit. The narrow conduit was 3 cm. Minor leaks, all of which were treated conservatively, occurred in six patients (13.6%). No significant difference was noted between cohorts in the anastomotic leak rate (subtotal, 4.5%; narrow, 22.7%; P = 0.19). In a second prospective randomized study, Zhang et al³⁷ compared outcomes after esophagectomy and reconstruction with a whole-stomach conduit (n = 52) and narrow gastric tube conduit (n = 52). An open McKeown approach was used for all proximal tumors with a hand-sewn cervical anastomosis (whole, 15.4%; narrow, 9.6%). An open Ivor Lewis approach was used for distal tumors (whole, 84.6%; narrow, 90.4%), and a circular stapler was used for the intrathoracic anastomosis. All conduits were brought up through the posterior mediastinum. There was no significant difference between the two groups in approach or anastomotic technique. The line of division for the whole-stomach cohort was not clearly defined. The narrow gastric conduit was 3 to 4 cm in width. No difference in the anastomotic leak rate was found (8% of anastomoses leaking in each group).

Four additional, larger retrospective series were identified, reporting conflicting results. Shu et al³⁸ reported the results of 850 patients who underwent esophagectomy with an open McKeown approach and a two-layer, hand-sewn cervical anastomosis. In this large retrospective study, the whole stomach was used in 397 patients and a tubular stomach measuring 4 to 6 cm in diameter in 453 patients. The groups were comparable in terms of tumor stage and location. Anastomotic leak rates were significantly higher in the whole-stomach cohort (9.3% vs. 5.5%; P = 0.033). In another retrospective study from China, Shen et al³⁹ reported the outcomes of 259 patients undergoing minimally invasive McKeown esophagectomy using a narrow (n =126) or wide gastric conduit (n = 133). Anastomotic leaks occurred more frequently when a wide conduit was used (17.3% vs. 8.7%; P = 0.041). Collard et al⁴⁰ performed a retrospective review of 212 patients undergoing esophagectomy for both benign and malignant indications using a variety of approaches and conduit routes—posterior mediastinal, retrosternal, or intrapleural. The authors compared the results of a whole-stomach conduit to those with a tubular gastric conduit. However, the dimensions of the gastric tube were not reported. Contrary to the results from Shu et al,³⁸ the incidence of anastomotic leak was significantly less when the
whole stomach was used as a conduit (1% vs. 7.9%; P = 0.021). Subgroup analysis by route of the conduit revealed consistent results favoring a whole-stomach conduit. Finley⁴¹ retrospectively reviewed single-institution outcomes of 295 patients treated with esophagectomy for management of malignant disease using a transhiatal approach with a 4-cm gastric tube, using a McKeown approach with a whole-stomach conduit and a thoracoabdominal approach with anastomosis to the “distal stomach.” Cervical anastomoses were performed with the transhiatal and McKeown approaches, whereas an intrathoracic anastomosis was used with the thoracoabdominal approach. The majority of patients had a gastric-emptying procedure with either a pyloroplasty or pyloromyotomy. Anastomotic leak rates were similar irrespective of operative approach or conduit width (tubular/transhiatal, 16%; whole stomach/McKeown, 22%; distal stomach/thoracoabdominal, 14%; P = not significant [NS]).

Strong evidence is lacking to support a recommendation of use of either a wholestomach gastric conduit or a wide 4- to 5-cm tubularized conduit. There is limited evidence supporting the use of a tubularized narrow conduit <4 cm in size. Regardless of the technique chosen, the submucosal collateral blood supply should be preserved to minimize anastomotic leak.

3B. PERFORMANCE OF A GASTRIC-EMPTYING PROCEDURE

Recommendation: A gastric-emptying procedure should be performed routinely at esophagectomy. Existing evidence does not support the exclusive use of any single procedure.

Type of Data: Retrospective.

Grade of Recommendation: Weak recommendation, low-quality evidence.

Rationale

Delayed gastric emptying (DGE) is a relatively common cause of morbidity after esophagectomy and is strongly associated with postoperative respiratory complications, primarily aspiration pneumonia.⁴² Typically both proximal vagus nerves are divided during standard esophagectomy, significantly altering the gastric motor activity and causing both impaired relaxation and contraction of the conduit. However some intrinsic motor activity of the stomach appears to recover at least partially over time.⁴³,⁴⁴ Emptying of the stomach in its normal anatomic position is thought to be facilitated by the positive pressure of the abdomen. However, transposition of the gastric conduit into the chest results in a shift to a negative-pressure environment, which may also alter transit through the stomach.⁴⁵ Gastric emptying may be facilitated by a transient increase in intraluminal pressure resulting from passage of a food bolus through the conduit. In general, however, the neo-esophagus is thought to be an inert conduit that allows passage of oral intake primarily by the forces of gravity.

Several aspects of conduit preparation may have an effect on emptying and have been studied, particularly conduit width and employment of a drainage procedure. Some authors have proposed that a narrower conduit may result in improved gastric-conduit emptying by generating more intraluminal pressure.⁴⁶,⁴⁷ Additionally,
gastric-emptying procedures have been proposed to decrease the resistance at the level of the pylorus, thereby facilitating emptying of the denervated conduit. Despite the literature on these topics, definitive answers have not emerged.

Few studies have investigated the association of gastric conduit width and DGE. Variability in operative technique, definition, and manner in which DGE was assessed and the interval between esophagectomy and assessment of gastric emptying make it difficult to draw any sound conclusions. Three of the previously mentioned studies also reported the incidence of DGE they encountered. Zhang et al³⁷ observed no difference in the incidence of DGE when a narrow 3- to 4-cm conduit (6%) or wholestomach conduit (6%) was used in their prospective randomized study with routine pyloroplasty. Shen et al³⁹ reported only one case of DGE in their narrow conduit cohort (conduit, 3 cm; n = 126), whereas none of the patients in the wide-conduit cohort (conduit, 5 cm; n = 133) developed symptoms of delayed emptying after minimally invasive McKeown esophagectomy without a pyloric emptying procedure. Although Finley et al⁴¹ did not observe a significant difference in anastomotic leaks when comparing approach and conduit width in their single-institution experience (n = 295), they did find a higher incidence of DGE in six patients (11%) after Ivor Lewis esophagectomy using a whole-stomach conduit, nine patients (4%) after a transhiatal approach with a 4-cm conduit, and in one patient (5%) after a thoracoabdominal approach with anastomosis to the distal stomach (P <0.05). None of these studies provided a clear definition of DGE or the manner in which this outcome was assessed.

Lee et al⁴⁸ retrospectively reviewed their outcomes for 56 patients who had gastric emptying assessed with a solid-food radionuclide study after esophagectomy for malignant disease using an Ivor Lewis (n = 51), McKeown (n = 2), or transhiatal (n = 3) approach. Emptying could not be measured in six patients because of rapid passage of the meal into the duodenum. Of the 50 patients analyzed, a whole-stomach conduit was used in 34 and a tubular gastric conduit in 16. The width of tubular conduit was not reported. All patients had some form of pyloric emptying procedure (finger fracture, 46; pyloroplasty, 4). DGE was defined as 50% emptying time (T50) >180 minutes. No difference in the incidence of DGE was noted when comparing the two types of conduits (16 had a whole conduit, 47.1%; 5 had tubular conduits, 31.3%; P = 0.365). Interestingly, the incidence of DGE was significantly lower in patients studied 6 months or later after esophagectomy than in those studied earlier than 6 months postsurgery (2 [14%] had DGE at >6 months, but 19 [52.9%] had DGE at <6 months; P = 0.02). However, the proportion of patients with tubular or wholestomach conduits was not clarified in terms of early or late assessment. Bemelman et al⁴⁹ performed a retrospective study of 140 patients who underwent esophagectomy with reconstruction using a whole-stomach conduit (n = 40), distal two-thirds or wide gastric conduit (n = 65), or narrow gastric conduit measuring 2 to 3 cm in diameter (n = 35). Cervical anastomoses were performed in those with wholestomach and narrow conduits, whereas intrathoracic anastomoses were created for those with a wide gastric conduit. None of the patients with a narrow conduit had a pyloric emptying procedure, whereas pyloroplasty was performed in 22.5% (n = 9) of those with a whole-stomach conduit and in 30.8% (n = 20) of those with a wide conduit. In this study, DGE was defined as an inability to resume a diet of solid
foods within 1 week after normal findings on an esophagram routinely performed on postoperative day 7. DGE was significantly higher when the whole stomach was used (n = 15; 37.5%) than with wide (n = 9; 13.8%; P = 0.05) and narrow (n = 1; 2.8%; P = 0.007) gastric conduits. No difference in DGE was noted comparing wide with narrow conduits (P = 0.02), nor was there a difference with or without a pyloroplasty in the whole-stomach and wide-conduit cohorts.

Multiple retrospective studies have been performed to determine if patients developed DGE less frequently if a pyloric drainage procedure was performed at the time of esophagectomy. Six retrospective studies were identified over the time period of 1990 to 2016 that directly compared outcomes of esophagectomy patients with or without a drainage procedure. A reduction in DGE was demonstrated by Cerfolio et al (2009),⁵⁰ who used botulinum toxin A (Botox) (n = 221; P = 0.002), and Deng et al (2010),⁵¹ who used finger fracture (n = 78; P =0.05); however, no statistically significant difference in DGE was noted by Lanuti et al (2011)⁵² (n = 242; P = 0.078); Palmes et al (2007),⁵³ who used pyloroplasty or myotomy (n = 198; P = not reported [NR]); Nguyen et al (2010),⁵⁴ who used pyloroplasty (n = 140; P = NR); or Mehran et al (2011),⁵⁵ who used pyloroplasty (n = 88, P = 0.219). It is not possible to determine from these retrospective studies if pyloric drainage is necessary at the time of esophagectomy to reduce DGE.

Two randomized controlled trials (RCTs) that employed pyloroplasty were also performed during this time period and published in English. Mannell et al (1990)⁵⁶ (n = 40; P = 0.1) found no difference in DGE on long-term follow-up, whereas Fok et al (1991)⁵⁷ (n = 200; P = 0.001) did demonstrate a significant difference in emptying. Two meta-analyses examined the question of pyloric drainage, and both concluded there was a nonsignificant trend toward improved DGE with drainage.⁵⁸,⁵⁹

Several confounding factors were then considered while reviewing the above studies. First, tubularization of the conduit has been shown to be associated with a reduced incidence of DGE.⁵⁰,⁵⁴ The methods in the aforementioned studies were not consistently using whole stomach or tubularized stomach; therefore, it is difficult to compare the studies directly. Secondly, the definition of DGE is not standardized. Some studies used water soluble contrast swallows from postoperative day 4 (early DGE), patient questionnaires, and nuclear medicine gastric-emptying studies. Given that DGE can improve with time, the data suggest that the most clinically relevant evaluation is validated patient questionnaires combined with objective testing done >7 days from the index procedure (late DGE). Finally, a well-designed randomized study is needed to remove some of the bias inherent in retrospective studies and give greater clarity to the discussion.

Part of the rationale for performing a pyloric drainage procedure is that providing protection for DGE early in the postoperative period may protect against aspiration and pulmonary complications. Some studies in the early 1990s showed an increase in pulmonary complications without drainage: Fok et al⁵⁷ found that 4% had complications without drainage, whereas Mannell et al⁵⁶ reported 15% experienced complications. No patients in either study who had drainage experienced complications. However, more recent studies have not demonstrated an increase in pulmonary complications related to a lack of drainage: those with pulmonary complications along
the drainage-no drainage divide included studies from Mehran et al⁵⁵ (34% vs. 32%), Cerfolio et al⁵⁰ (17% vs. 22%), and Palmes et al⁵³ (16% vs. 13%). These studies reported perioperative mortality rates of 1% to 3%, which included patients who died of pulmonary complications.

Because the incidence of DGE in this setting is estimated to be 15% to 25% and the complications can be devastating, a gastric-emptying procedure should be performed routinely at the time of esophagectomy. However, existing evidence does not support the exclusive use of any single procedure.

Technical Aspects

Botox Injection of the Pylorus

At the outset, 100 units of Botox are dissolved in 4 mL of saline, and 1 mL of this is primed into an endoscopic needle. A standard adult endoscope combined with an endoscopic 25-gauge needle is then used to inject 1 mL of Botox into each of the four quadrants of the pylorus. A 1-mL saline “chaser,” or follow-up injection, can be used to deliver the last milliliter. Botox treatments usually last 3 months, and patients may require more treatments in the future.

Pyloromyotomy

During the initial operation, the serosa at the gastroduodenal junction is scored transversely. The pylorus muscle is then divided. If there is any breach of the mucosa, then conversion to pyloroplasty is recommended. Overlaying an adjacent patch of omentum on the suture line to buttress the area of the myotomy is simple to do.

Pyloroplasty

Pyloroplasty can be accomplished laparoscopically or by open approach. After an esophagectomy and cervical esophagogastric anastomosis, the pylorus typically comes to rest 3 to 4 cm below the level of the hiatus. If done at a second operation, the pylorus is exposed by applying a liver retractor and performing lysis of adhesions until the hiatus is exposed. Great care is required to prevent injury to the right gastroepiploic vascular arcade. A 3- to 4-cm longitudinal transmural incision is made across the pylorus. A Heineke-Mikulicz pyloroplasty is then performed to close the gastrotomy in a transverse fashion.

SUMMARY

There is much debate over preparation of gastric conduit width with little supporting evidence in the literature. Expert consensus supports optimizing the procedure according to well-known key principles. The submucosal collaterals are essential for preserving the blood flow to the gastric conduit to minimize anastomotic leaks. The gastric tip is the area most at risk for ischemia. Preserving the “crow’s feet” along the lesser curve of the gastric conduit is a crucial step to not compromising collateral blood supply. Most US surgeons avoid a gastric conduit narrower than 4 to 5 cm, and if needed for adequate cancer margin, an alternate conduit is recommended (discussed subsequently). Regarding gastric emptying, there does not appear to be a difference in
function between the use of the whole stomach or a 4- to 5-cm conduit, but we recommend a gastric-emptying procedure to reduce the inherent 15% to 20% risk of DGE.

3C. ALTERNATE CONDUITS

Recommendation: A gastric conduit should generally be used for gastrointestinal reconstruction. Although alternatives such as the colon and small bowel can be used safely and effectively in selected cases, assistance from surgeons with expertise in microvascular surgery may be necessary to perform a vascular anastomosis in the neck to supercharge a colon or a jejunal interposition graft.

Type of Data: Retrospective data, technical manuscripts.

Grade of Recommendation: Weak recommendation, low-quality evidence.

Rationale

Full-length esophageal reconstruction in patients who lack sufficient stomach to serve as a viable replacement have two alternative conduits for reconstruction: colon or jejunum. A segment of colon is the most common option other than stomach to reestablish GI continuity after esophageal resection (Fig. 7-5). Patients with colon grafts have been shown to have good short-term and long-term satisfaction.⁶⁰,⁶¹ Colon interposition is generally regarded as inferior to stomach as an esophageal substitute because of multiple concerns. The colonic interposition is a more complex and technically demanding operation than gastric replacement. Three anastomoses are required to reestablish GI continuity versus a single anastomosis when the stomach is used. Long term, colon grafts also have a tendency to dilate and form redundant loops.⁶² The advantages of colon interposition include adequate conduit length and consistent blood supply. The jejunum is uniquely suited for reconstruction because it is abundant, is disease free, is similar in size to the esophagus, has intrinsic peristalsis like an esophagus, and generally does not experience senescent lengthening to the extent a colonic interposition may.³⁰,³¹ Interest in and experience with using the jejunum as a conduit is growing.

Preoperative technical considerations include determining a patient’s current anatomy and studying conduit options. Patients will present at different time points in their treatment for consideration of complex esophageal reconstruction. Some will present without having had an intervention, and the surgeon must decide whether to perform a single operation with esophagectomy and formation of conduit or to perform a staged procedure (diversion followed later by reconstruction). Other patients will present having already undergone an esophagectomy with failed reconstruction or diversion. It is essential that with these patients, prior to proceeding with reconstruction, surgeons have a clear understanding of their previous interventions and current anatomy. Some patients may have residual stomach, which may be used as conduit or as a reservoir for distal conduit anastomosis.³¹,⁴⁴

Computed tomography (CT) is helpful in many aspects of preprocedural planning. It images residual gastric volume to determine whether there is an adequate remnant for the reestablishment of GI continuity with stomach itself or whether an alternative option should be prepared. The CT also helps delineate the potential
routes of reconstruction (e.g., posterior mediastinum or retrosternal; see Fig. 7-6). The posterior mediastinal route in the esophageal bed is ideal because it is the shortest and has an excellent functional result for patients undergoing immediate reconstruction. Unfortunately, significant scarring from prior esophagectomy makes this route impossible in those patients who have undergone esophagectomy. In these cases, a retrosternal approach is recommended.⁴⁴ If the retrosternal route is not an option (e.g., previous cardiac surgery), the subcutaneous route is also an option although less optimal because of the longer length of conduit needed. Finally, mesenteric angiography interrogates the superior mesenteric artery to rule out calcification or disease, provides anatomic information in patients who have had previous abdominal surgery, and importantly reveals the configuration of the jejunal arcades. Colonoscopy or barium enema allows the entire colonic mucosa to be examined, and biopsies of suspicious areas may be performed if found.

FIGURE 7-5 Stomach (A), colon (B), and jejunum (C) are the most commonly used conduits for esophageal replacement.

FIGURE 7-6 The posterior mediastinal route (A) requires the shortest length of conduit for reconstruction. When the posterior mediastinal space is obliterated or unavailable, the retrosternal (B) and rarely parahilar (intrapleural) (C) or subcutaneous (D) routes are used.

Technical Aspects: Colon

Left Colon

The isoperistaltic left colon interposition is based on the ascending branch of the left colic artery via the marginal artery communicating with the divided middle colic artery. The length of colon necessary for esophageal replacement is assessed by using topographic landmarks.

The marginal artery proximal and distal to the marked areas of the left colon is exposed. If the proximal line of division is to the right of the middle colic artery, the base of the artery must be exposed proximal to the marginal artery. Atraumatic vascular clamps are then placed on the proximal and distal segments of the marginal artery and, if necessary, the middle colic artery. Adequacy of arterial flow to the graft can be assessed by inspection, palpation, and Doppler ultrasound. Spectroscopy (fluorescein angiography or indocyanine green spectroscopy) is sometimes useful in visualizing blood flow along the length of the graft. Once adequate vascularity of the left colon segment has been verified, the middle colic artery is ligated and divided proximal to its branching into the marginal artery. The proximal and distal portion of the marginal artery are also ligated and divided. The colon is transected proximally and distally with a GI linear stapler. The remaining portions of the transverse mesocolon are divided to facilitate straightening of the graft. A retrosternal route is created by detachment of the diaphragm from the sternum below the xiphoid process. The superior entry is prepared by division of the deep cervical fascia and entry into the retrosternal
midline space behind the manubrium. The retrosternal tunnel is then completed by blunt distal dissection from the superior and inferior entry sites. The neck incision is created, and the left clavicular head, first rib, and the left half of the manubrium are resected to create space for the jejunum or colon to enter the neck without constriction. This also prevents venous congestion of the conduit from compression by the thoracic inlet. This holds true for any retrosternal conduit—gastric, jejunal, or colonic (Fig. 7-7).³¹,⁴⁴,⁶³,⁶⁴

FIGURE 7-7 The cervical incision (A) and resection of a portion of the manubrium, first rib, and clavicular head (B) to accommodate the conduit without kinking or impingement. A small portion of a jejunal graft is externalized as a “monitoring flap” (C) of graft perfusion.

Right Colon

The isoperistaltic right colon graft is based on the middle colic vessels. The approach is similar to that used for the left colon. After mobilization of the greater omentum, the entire right colon and terminal ileum are mobilized from the retroperitoneum. The length of the colon segment is measured in a fashion similar to that described previously.

The ileocolic, right colic, and ileal (if necessary) arteries are then test clamped with atraumatic vascular clamps. Once the adequacy of the vascular supply has been confirmed, the vessels are divided. The proximal anastomosis is constructed first and then any redundant colon is pulled back into the abdomen. The colon is tacked to the edges of the diaphragm with three stitches to prevent paraconduit herniation of bowel contents into the chest. Any redundant colon is excised, and the cologastric or colojejunal anastomosis is then constructed.

Technical Aspects: Jejunum

The jejunum is mobilized from the ligament of Treitz, and the jejunal vascular arcade is transilluminated.³¹,⁶⁵ The first branch of the superior mesenteric artery (SMA) to the jejunum is spared, and the jejunal flap is created by severing the second and third jejunal vessels while preserving the vascular arcades to the small bowel. The blood supply to the jejunum will be provided from the fourth jejunal branch. The superior-most vessel (second jejunal branch) should be divided at the base of the arch, and small clips or atraumatic vascular bulldog clamps placed, leaving the artery intact for possible microvascular augmentation in the neck.³¹,⁶³,⁶⁵ The jejunum is transected proximally 3 to 5 cm from the ligament of Treitz. Once the second and third vascular pedicles are transected, the small bowel begins to unfurl and yields approximately 40 to 50 cm of jejunum for reconstruction.²⁵ In this case, the distal small bowel conduit will maintain vascular supply from the SMA, and vascular augmentation is mandatory to maintain blood flow to the proximal third of the graft above the line of mesenteric transection.³¹,⁶³,⁶⁵ The jejunal graft is then evaluated for appropriate length to reach the neck and for redundancy (Fig. 7-8). Ideally the jejunal graft should unfurl completely to facilitate reach and to improve emptying.⁶⁵,⁶⁶ Adequacy of arterial flow may be assessed by bowel inspection, palpation, Doppler, or spectroscopy as performed for the colon graft.

When performing the neck incision, careful attention is paid to avoid injury to the left internal mammary artery, which is used in vascular augmentation and reconstruction. If it is to be used for reconstruction, the left internal mammary artery is carefully skeletonized, clipped, and transected distally.⁶³ The jejunal conduit may follow a retrocolic or antecolic route. In most cases, jejunal interposition is performed in delayed fashion, and therefore the best position for the conduit will be a retrosternal one.⁶³,⁶⁴ The jejunal graft is passed retrosternally, taking great care to avoid tearing the
vascular arcades or creating an intimal flap within the delicate vessels. Microvascular anastomoses are performed between the most proximal jejunal arcade (most commonly the second jejunal artery and vein) and either the internal mammary vessels (artery and vein) or the external jugular vein and branch of the external carotid artery. Anastomosis is performed under the microscope. In some cases, there is inadequate length, and the saphenous vein must be harvested to bridge the gap.³¹,⁶⁵

FIGURE 7-8 Jejunal conduit is draped over the sternal region to evaluate for adequate length of a viable portion of the bowel.

A monitoring flap is helpful for postoperative evaluation of the health of the jejunal flap. This can be accomplished by leaving a small portion of the incision open, allowing direct visualization of the conduit. A terminal 2 to 3 cm of the jejunal flap can be separated from the conduit before reconstruction and externalized, or Doppler monitors may be left around the venous and arterial anastomoses to monitor flow.³¹,⁶⁵,⁶⁶ No drains should be left directly against the visceral anastomosis nor around the microvascular anastomosis.³¹

In the abdomen, a Roux-en-Y reconstruction is performed. A side-to-side functional end-to-end jejunojejunostomy is performed. A jejunostomy feeding tube is placed distal to this anastomosis. It is important to remember to close Petersen’s and any mesenteric defects to avoid the creation of internal hernias that can result in bowel necrosis.³¹,⁴³ Additionally, the hiatal opening should be narrowed to admit the graft along with food bolus only, and the conduit is tacked to the hiatal opening.

3D. CERVICAL VERSUS INTRATHORACIC ANASTOMOSIS

Recommendation: Cervical and intrathoracic anastomoses appear to be equally safe with regard to perioperative mortality, pulmonary complications, anastomotic stricture, and quality of life. In general, the location of the anastomosis should be dictated by the location of the tumor. For distal esophageal or gastroesophageal junction tumors, existing evidence does not support the assertion that construction of a cervical anastomosis reduces recurrence or improves survival over an intrathoracic anastomosis, as long as microscopically negative margins are achieved.

Type of Data: Randomized controlled trials and meta-analysis.

Grade of Recommendation: Strong recommendation, moderate-quality evidence.

Rationale

A systematic search yielded four RCTs⁶⁷,⁶⁸,⁶⁹,⁷⁰ and two meta-analyses that compared the thoracic and the cervical anastomotic approach (Grade 2b).⁷¹,⁷² Table 7-1 summarizes the RCTs referenced in this section. In one of the RCTs, the cervical approach was associated with a significant increase in anastomotic leakage rates when compared with leakage rates associated with the intrathoracic approach.⁶⁷ To the contrary, three other RCTs did not show a significant difference with regard to anastomotic integrity. However, a nonsignificant trend toward more leakage in the cervical group was suggested by two trials.⁶⁹,⁷⁰ This trend was confirmed by two meta-analyses that included all four RCTs, comprising 298 patients, and found that anastomotic leaks were seen significantly more often in the cervical group than in the intrathoracic group (cervical: OR, 4.7; 95% confidence interval [CI], 1.6-13.9; P = 0.005; intrathoracic: OR, 3.43;

95% CI, 1.1-10.8; P = 0.03).⁷¹,⁷² It is hypothesized that this difference may be explained by the finding of less tissue ischemia with an intrathoracic anastomosis because of the shorter length the blood supply must travel and therefore less tension on the intrathoracic anastomosis.⁷¹,⁷²,⁷³

TABLE 7-1 Characteristics and Results of Relevant Prospective Randomized Trials Regarding Anastomotic Location and Technique

Author, Year	N	Comparison, Outcomes, and Results	Grade of Evidence	Risk of Bias	Conclusion of Paper	Comments
Saluja et al, 2012⁸¹	174	Hand-sewn double layer interrupted (n = 87) vs. stapled side-to-side cervical anastomosis (n = 87) In-hospital mortality: 6/87 (6.9%) vs. 5/87 (5.7%) (P = 1.00) Anastomotic leakage: 16/87 (18.4%) vs. 14/87 (16.1%) (P = 0.33) Operating time (min): 245 ± 62 vs. 252 ± 58 (P = 0.45) Hospital stay (d): 12.8 ± 8 vs. 11.9 ± 6 (P = 0.32) Blood loss (mL): 570 ± 444 vs. 566 ± 448 (P = 0.61) Anastomotic stricture: 12/69 (17.3%) vs. 6/68 (8.8%) (P = 0.20)	High	High	Anastomotic leakage and postoperative outcomes are comparable among patients undergoing esophagectomy with stapled or handsewn anastomoses.	Data available on patientand treatment-related characteristics are limited. Although anastomotic technique was standardized throughout the study period, different surgical approaches were used (transhiatal [n = 145] and transthoracic [n = 29]).
Nederlof et al, 2011⁸⁷	128	End-to-side (n = 64) vs. end-to-end (n = 64) hand-sewn single continuous layer cervical anastomosis In-hospital mortality: 2/64 (3%) vs. 7/64 (11%) (P = 0.16) Anastomotic leakage: 10/64 (16%) vs. 21/64 (33%) (P = 0.04) Anastomotic stricture: 25/64 (40%) vs. 10/64 (18%) (P <0.01) Pneumonia: 11/64 (17%) vs. 28/64 (44%) (P = 0.002) Median (range) hospital stay (d): 15 (9-125) vs. 22 (8-281) (P = 0.02)	High	Low	End-to-side anastomosis is associated with lower anastomotic stricture rates than is an end-to-end anastomosis at the cost of higher anastomotic leakage rates and longer hospital stay.	The surgeons who performed the end-to-side anastomoses in the RCT were acquainted with the construction of an end-to-end anastomosis and were possibly still learning the creation of the end-to-side anastomosis.
Zhang et al, 2010⁸³	516	Hand-sewn two-layered continuous and interrupted (n = 244) vs. circular stapled (n = 272) intrathoracic anastomoses 30-day mortality: 3/244 (1.2%) vs. 7/272 (2.6%) (P = NS) Anastomotic leakage: 1/244 (0.4%) vs. 6/272 (2.2%) (P = NS) Operating time (min): 240 ± 31 vs. 220 ± 26 (P = NS) Anastomotic stricture: 2/236 (0.8%) vs. 13/261 (5.0) (P = 0.016)	High	Low	Compared with a stapled anastomosis, a layered manual anastomosis reduces the incidence of anastomotic strictures.	Data on patient- and treatment-related characteristics that may affect anastomotic outcomes are limited. Data on other outcomes are also limited. No in-hospital mortality significant differences were noted.
Luechakiettisak et al, 2008⁸⁴	117	Hand-sewn (n = 59) continuous single-layer vs. circular stapled (n = 58) intrathoracic anastomoses 30-day mortality: 7/59 (11.8%) vs. 6/58 (10.3%) (P = 0.97) Anastomotic leakage: 4/59 (6.7%) vs. 2/58 (3.4%) (P = 0.69) Pulmonary complications: 8/59 (13.5%) vs. 10/58 (17.2%) (P = 0.77) Operating time (min): 218.1 (range, 182-278) vs. 7 (range, 182-228) (P <0.001) Blood loss (mL): 864.4 (range, 600-1,300) vs. 803.4 (range, 500-1,100) (P = 0.02) Anastomotic stricture: 10/52 (19.2%) vs. 19/52 (36.5%) (P = 0.08)	High		Both hand-sewn and stapled esophagogastric anastomoses are safe. The stapled method has a higher incidence of anastomotic stricture.	Data available on patientand treatment-related characteristics are limited. The postoperative mortality rate was high.
Okuyama et al, 2007⁶⁹	32	Hand-sewn two-layer interrupted cervical anastomosis (transhiatal procedure) vs. circular stapled intrathoracic (transthoracic procedure) anastomosis In-hospital mortality: 0 (0%) vs. 0 (0%) (P = NS) Anastomotic leakage: 3/18 (16.7%) vs. 1/14 (7.1%) (P = NS) Recurrent laryngeal nerve palsy: 8/18 (38.8) vs. 1/14 (7.1%) (P <0.05) Pneumonia: 2/18 (11.1%) vs. 5/14 (35.7%) (P = NS) Blood loss (mL): 537 ± 281 vs. 702 ± 252 (P = NS) Operating time (min): 547 ± 95 vs. 593 ± 57 (P = NS) Anastomotic stricture: 0/18 (0%) vs. 2/14 (14.2%) (P = NS) QOL: No difference in QOL 6 mo after surgery (heartburn, regurgitation, stenotic sense, abdominal fullness, cough, and pain) Tumor recurrence: 4 (22%) vs. 2 (14%) (P = NS) Survival (5-y): 72.2% vs. 85.7% (P = 0.41)	High	Low	Both methods of anastomosis yielded similar anastomotic outcomes. The incidence of recurrent laryngeal nerve injury was higher after hand-sewn cervical anastomosis	Small sample size made comparisons among different anastomotic techniques at different sites while using different surgical techniques interesting but yielded few actionable results.
Walther et al, 2003⁶⁸	83	Hand-sewn cervical (n = 41) vs. circular stapled intrathoracic (n = 42) anastomoses In-hospital mortality: 1/41 (2.4%) vs. 1/42 (2.4%) (P = NS) Anastomotic leakage: 1/41 (2.4%) vs. 0/42 (0%) (P = NS) Hospital stay (d): 14 (range, 6-68) vs. 14 (range, 0-83) (P = NS) Operating time (min): 555 (382-850) vs. 553 (290-750) (P ± NS) Blood loss (mL): 950 (250-3,000) vs. 950 (200-4,000) (P ± NS) Pulmonary complications: 2/41 (4.9%) vs. 4/42 (9.5%) (P = NS) Anastomotic stricture: 8/41 (19.5%) vs. 12/42 (28.6%) (P = 0.433) Recurrent nerve palsy: 1/41 (2.4%) vs. 0/42 (0.0%) (P = NS)	High	Low	Cervical and intrathoracic anastomoses are equally safe.	There was variation in the use of stapled and sutured anastomosis at each site. The incidence of anastomotic leaks was low, making correlation to site and severity difficult.
Hsu et al, 2004⁷⁸	63	Hand-sewn (n = 32) vs. stapled (n = 31) cervical anastomosis In-hospital mortality: 4/31 (13%) vs. 3/31 (10%) (P = NS) Anastomotic leakage: 7/32 (22%) vs. 8/31 (26%) (P = NS) Anastomotic stricture: 4/28 (14%) vs. 5/28 (18%) (P = NS) Operating time (min): 524 vs. 447 (P <0.001) Pulmonary complications: 9/32 (28%) vs. 11/31 (36%) (P = NS) Mean hospital stay (d): 43 vs. 38 (P = NS) Survival after 3 y: comparable	High	Low	A stapled anastomosis results in a shorter operating time. Other outcomes were comparable.	This study with a relatively small sample size had high mortality and morbidity rates.
Laterza et al, 1999⁷⁹	43	Hand-sewn (n = 21) vs. stapled (n = 20) cervical anastomosis In-hospital mortality: 2/21 (9.5%) vs. 1/20 (5%) (P = NS) Anastomotic leakage: 1/21 (4.8%) vs. 4/20 (20%) (P = NS) Anastomotic stricture: 2/21 (9.5%) vs. 3/20 (15%) (P = NS)	High	Low	Esophagogastric anastomoses should be hand-sewn rather than stapled.	A small sample size made correlation of the type of anastomosis with different outcomes difficult.
Law et al, 1997²¹	122	Hand-sewn single layer continuous (n = 61) vs. circular stapled (n = 61) intrathoracic anastomosis In-hospital mortality: 4/61 (6.6%) vs. 6/61 (9.8%) (P = NS) Anastomotic leakage: 1/61 (1.6%) vs. 3/61 (4.9%) (P = NS) Pulmonary complications: 6/61 (9.8%) vs. 11/61 (18%) Operating time (min): 214 ± vs. 217 ± 3.4 (P = NS) Blood loss (mL): 721 ± 31 vs. 708 ± 33 (P = NS) Anastomotic stricture: 5/55 (9.1%) vs. 20/50 (40%) (P <0.001)	High	Low	A stapled anastomosis results in more anastomotic stricture formation.	Differences were found in the size of stapler used throughout the study. Moreover, some patients were treated with adjuvant radiotherapy.
Valverde et al, 1996⁸²	154	Hand-sewn, single- or two-layer interrupted or continuous anastomosis (n = 74) vs. circular stapled (n = 78) cervical or intrathoracic anastomosis 90-day mortality: 7/74 (9.5%) vs. 3/78 (3.8%) (P = NS) Anastomotic leakage: 12/74 (16%) vs. 12/78 (15%) (P = NS) Pulmonary complications: 37/74 (50%) vs. 41/78 (52.3%) (P = NS) Operating time (min): 401 ± 130 vs. 390 ± 120 Anastomotic stricture: 8/63 (13%) vs. 7/53 (11%) (P = NS)	High	High	The disadvantages of mechanical staples (technical mishaps and higher costs) are not counterbalanced by a gain of time or a decrease in rate or severity of anastomotic leakage.	High variation in surgical techniques and site of anastomosis was noted.
Craig et al, 1996⁷⁷	100	Hand-sewn, single-layer interrupted (n = 50) vs. circular stapled (n = 50) intrathoracic anastomosis In-hospital mortality: 1/50 (2%) vs. 4/50 (8%) (P = NS) Anastomotic leakage: 3/50 (6%) vs. 4/50 (8%) (P = NS) Operating time (min): 145 vs. 126 (P <0.05) Blood loss (mL): 381 vs. 344 (P <0.05 Anastomotic stricture: 13/49 (27%) vs. 13/46 (28%) (P = NS)	High	Low	A stapled anastomosis is quick to perform, allowing a shorter operating time and lower blood loss. In this comparison with hand-sewn anastomoses, it does not appear to be associated with a previously noted increased incidence of benign anastomotic stricture formation.	Data available on patientand treatment-related characteristics are limited.
Ribet et al, 1992⁷⁰	60	Hand-sewn cervical three-stage approach to anastomosis (n = 30) vs. two-stage approach with stapled intrathoracic anastomosis (n = 28) In-hospital mortality: 5/30 (17%) vs. 4/28 (13%) (P = NS) Anastomotic leakage: 9/30 (30%) vs. 3 (10%) (P = NS) Pulmonary complications: 21/30 (70%) vs. 11/28 (37%) (P = 0.001) Recurrent nerve trauma: 6/30 (20%) vs. 1/28 (3%) (P = NS) Operating time (min): 405 vs. 375 Hospital stay (d): 24.2 vs. 16.6 Anastomotic stricture: 4/30 (13%) vs. 1/28 (3%) (P = NS) Positive resection margin: 3/30 (10%) vs. 10/28 (33%) (P <0.05) Median survival (mo): 9 vs. 12 (P = NS)	High	Low	Respiratory complications and recurrent laryngeal nerve trauma were more frequent after cervical anastomosis.	Sample was small. Data available on patientand treatment-related characteristics were limited. Postoperative mortality rate was high. Patients were not matched for neoadjuvant treatment.
Chasseray et al, 1989⁶⁷	123	Cervical transhiatal or three-stage approach to anastomosis (n = 43) vs. intrathoracic anastomosis with two-stage approach (n = 49) 30-d mortality: 4/43 (9.3%) vs. 7/49 (14.2%) (P = NS) Anastomotic leakage: 11/43 (9.3%) vs. 2/49 (4%) (P = 0.02) Pulmonary complications: 7/43 (16%) vs. 15/49 (29%) (P = NS) Operating time (hr): 7 (range, 4.5-10) vs. 7 (range, 3.5-10) (P = NS) Length/hospital stay (d): 19.5 (range, 3-71) vs. 18 (range, 2-122) (P = NS) Anastomotic stricture: 23/43 (53%) vs. 14 (29%) (P = NS) Positive resection margin: 1/43 (2%) vs. 1/49 (2%) (P = NS) Median survival (mo): 23 vs. 20 (P = NS)	High	High	Cervical anastomosis is associated with a significantly increased risk of anastomotic leakage.	Surgical approach and operative technique were not standardized.
RCT, randomized controlled trial; NS, not significant; QOL, quality of life.

Although leakage rates from the intrathoracic approach may be lower, some surgeons prefer the cervical approach because leakage after this approach can be easily diverted by opening the neck wound. However, cervical anastomotic leakages can also cause potentially life-threatening intrathoracic complications, including pleural empyema and mediastinal abscess, with reported incidence rates of up to 44% to 62% after anastomotic leakage.⁷⁴,⁷⁵,⁷⁶ To this regard, the four previously mentioned RCTs found no significant difference in in-hospital mortality between the two groups. This finding was confirmed by a meta-analysis that found no difference in postoperative mortality between the cervical and intrathoracic anastomosis (OR, 1.24; 95% CI, 0.35-4.41; P = 0.74).⁷¹

The functional results (i.e., anastomotic stricture, recurrent nerve palsy) and quality of life (QOL) were compared after cervical or intrathoracic anastomosis after esophagectomy. None of the four RCTs that compared the thoracic and the cervical anastomotic approaches reported a significant difference between them in stricture formation after surgery.⁶⁷,⁶⁸,⁶⁹,⁷⁰ This finding was supported by a meta-analysis that combined the results of these four trials and concluded that benign stricture formation requiring dilatation was not associated with the location of the anastomosis.⁷¹ Of note is the absence of a consensus definition and diagnostic criteria and threshold of treatment for anastomotic strictures, which limits comparison between these trials. Performing a cervical anastomosis has been associated with a significantly higher incidence of recurrent nerve trauma by a meta-analysis of four RCTs (OR, 7.14; 95% CI, 1.75-29.14; P = 0.006), although none of the four RCTs had comparable definitions of how postoperative recurrent nerve trauma was diagnosed.⁷¹ This is likely related to the close anatomic relation of the recurrent nerve with the esophagus in the cervical region.

One trial evaluated QOL 6 months after surgery on the basis of questions put to the patients about experiencing heartburn, regurgitation, stenotic sense, abdominal fullness, cough, or wound pain.⁶⁹ After 6 months there was no difference in reported QOL between the two groups.⁶⁹ When looking at oncologic outcomes, one trial found a significantly higher invasion of the resection specimen in patients undergoing a thoracic anastomosis than in those undergoing cervical anastomosis (33% versus 10%, respectively; P <0.05).⁷⁰ Two other trials that also reported on the resection margin could not confirm this finding.⁶⁷,⁶⁸ With regard to tumor recurrence, none of the four RCTs reported a significant difference between the cervical and intrathoracic groups,⁶⁷,⁶⁸,⁶⁹ which was confirmed by a meta-analysis that found no difference with regard to tumor recurrence between the groups.⁷¹

Methodologic concerns that may have influenced the results of the various studies include a relatively small sample size and great heterogeneity in surgical approach among the included studies.⁶⁸,⁶⁹,⁷⁰ Moreover, in most studies the anastomotic technique was not standardized, resulting in a variation in the use of stapled and sutured anastomoses at each site.⁶⁷,⁶⁸,⁶⁹ As the quality of a meta-analysis reflects that of the constituent studies, the aforementioned meta-analyses are inherently limited by the drawbacks
of combining the heterogeneous data of these RCTs. Therefore, robust, adequately powered studies with a standardized surgical technique are warranted to determine potential differences between intrathoracic and cervical anastomoses.

3E. HAND-SEWN VERSUS STAPLED ANASTOMOSIS

Recommendation: Based on existing evidence, construction of the anastomosis using a hand-sewn or stapled technique are both acceptable. Rates of anastomotic leak, early mortality, and postoperative complications associated with hand-sewn anastomoses are not different than those associated with stapled anastomoses. However, stapled anastomoses may be associated with an increased rate of anastomotic stricture when compared with hand-sewn anastomoses.

Type of Data: Randomized controlled trials.

Grade of Recommendation: Strong recommendation, moderate-quality evidence.

Rationale

Esophagogastric anastomosis can be constructed by hand or with a mechanical stapling device (Figs. 7-9, 7-10, and 7-11). Currently, the choice of a hand-sewn or stapled approach depends on surgeon preferences. A systematic search yielded 10 RCTs²¹,⁶⁸,⁶⁹,⁷⁷,⁷⁸,⁷⁹,⁸¹,⁸²,⁸³,⁸⁴ and two meta-analyses⁸⁵,⁸⁶ that compared the hand-sewing technique with the stapling technique. No RCT or meta-analysis showed a statistically significant difference in either anastomotic leakage rates or early mortality between
techniques. However, in one subgroup analysis of 302 patients, it was suggested that stapling might have a benefit over single-layer, hand-sewn esophagogastric construction in reducing the risk of anastomotic leakage (relative risk [RR] = 0.37; 95% CI, 018-0.76; P = <0.01).⁸⁵ Except for the single-layer hand-sewn technique, the current evidence suggests that it is appropriate to consider both anastomotic techniques (handsewn vs. stapled) as equivalent options for esophagogastric anastomotic construction.

FIGURE 7-9 The hand sewn end to side anastomosis approximates the end of the esophagus to a gastrotomy made in the side of the gastric conduit (A). It can be performed in an interrupted or running (B) fashion. The gastrotomy should be positioned to avoid proximity to the gastric staple line along the lesser curvature and the gastroepiploic/short gastric vessels on the greater curvature when the anastomosis is completed (C).

FIGURE 7-10 The esophagus is aligned along the gastric conduit (A) with enough overlap to accommodate the anvil and cartridge of a linear stapler (B) through the open end of the esophagus and a gastrotomy. Firing the stapler creates a common channel to form to back wall of the anastomosis. The remaining opening is closed with suturing (shown) or stapler (not depicted) (C). The gastrotomy should be positioned equidistant from the gastric staple line and greater curvature to decrease the incidence anastomotic ischemia.

There were no differences in incidence of pulmonary complications between groups.⁶⁸,⁶⁹,⁷⁸,⁸²,⁸⁴ A stapled esophagogastric anastomosis was significantly faster to construct during surgery than a hand-sewn anastomosis in some studies.⁸⁴ Also blood loss was significantly reduced in some studies, although the absolute reduction in blood loss was very minimal (range, 37 to 61 mL).⁷⁷,⁸⁴ Although some RCTs report statistical differences between these outcomes, one can challenge the clinical relevance of these findings.

Two of the ten RCTs that compared the hand-sewing technique with the stapling technique reported a significantly higher incidence of anastomotic strictures after stapled anastomotic construction,²¹,⁸³ and a third reported a nonsignificant trend toward more strictures in the stapled group.⁸⁴ The other seven RCTs did not show a significant difference with regard to anastomotic strictures.⁶⁸,⁶⁹,⁷⁷,⁷⁸,⁷⁹,⁸¹,⁸² A meta-analysis demonstrated that a stapled anastomosis is associated with an increased risk of postoperative anastomotic stricture formation (OR, 1.76; 95% CI, 1.09-2.86; P = 0.02).⁸⁶ It is hypothesized that this finding is caused by the compression of tissue by the stapling device, which results in excessive fibrosis or, alternatively, ischemia of the tissue between
staples and further fibrosis. Of note is the lack of a consensus definition on diagnostic criteria for anastomotic strictures, which is an important limitation when comparison between trials is attempted.

FIGURE 7-11 One method of performing an end to end stapled anastomosis requires inserting the anvil into the proximal stump of the esophagus and securing the open end around the anvil stem with a purse string suture (A). The cartridge is inserted through a gastrotomy in the proximal stomach and a puncture is made in the gastric wall where the anastomosis is planned to join the cartridge to the stem (B). After the stapler is fired, the portion of the stomach with the gastrotomy and GE junction is amputated with a linear stapler (C).

In one of the RCTs QOL was evaluated 6 months after surgery on the basis of questions asking the patients if they suffered from heartburn, regurgitation, stenotic sense, abdominal fullness, cough, or wound pain.⁶⁹ After 6 months there was no difference in reported QOL between the two groups.⁶⁹ Methodologic concerns that may have influenced the results of the various studies include a large degree of heterogeneity with regard to surgical approach, anatomic location of anastomosis, type of stapler (e.g., circular or linear), type of suture materials, and sewing technique (single-layer or double-layer stitching).

3F. END-TO-END VERSUS END-TO-SIDE ANASTOMOSIS

Recommendation: Based on existing evidence, construction of the anastomosis using an end-to-end or end-to-side configuration are both acceptable. End-to-side anastomoses are weakly associated with an increased risk of anastomotic leak when compared with end-to-end anastomoses. However, end-to-end anastomoses are associated with an increased risk of stricture.

Type of Data: Randomized controlled trials.

Grade of Recommendation: Weak recommendation, high-quality evidence.

Rationale

A systematic search yielded one RCT of 128 patients that compared the end-to-end with the end-to-side technique.⁸⁷ In this study by Nederlof et al,⁸⁷ the end-to-side anastomosis was associated with a higher incidence of anastomotic leakage (16% vs. 33%; P = 0.04), pneumonia (17% vs. 44%; P = 0.002), and a longer hospital stay (15 vs. 22 days; P = 0.02). More strictures were observed after the end-to-end approach than after the end-to-side approach (40% vs. 18%; P <0.001).⁸⁷ One clinical retrospective study that compared end-to-end anastomosis with end-to-side anastomosis could not confirm a difference in leak rates, but it did suggest that patients with an end-to-end anastomosis require dilatation at a statistically significantly higher rate than do those whose surgeons employed the end-to-side anastomotic technique.⁸⁸ Methodologic concerns that may have influenced the results include that the surgeons who performed the end-to-side anastomoses in the included RCT were primarily acquainted with the construction of an end-to-end anastomosis and were possibly during the trial still within their learning curve for the creation of end-to-side anastomoses.⁸⁹

Technical Aspects

The gastric conduit is laid next to the proximal esophagus. The site of the anastomosis is identified, and a 1.5-cm vertical gastrostomy is made in the anterior wall of the stomach. The stapled end of the cervical esophagus is amputated. The posterior wall of the esophagus and the anterior wall of the stomach are aligned by using either sutures or clamps. A side-to-side anastomosis between the cervical esophagus and gastric fundus is performed by using an Endo GIA 30 stapler with 3.5-mm staples (Medtronic). The traction sutures or clamps are gently retracted downward as the stapler is advanced, ensuring that the posterior wall of the esophagus and anterior wall of the stomach are well aligned. The stapler is then fired, creating a 3-cm side-to-side anastomosis. A nasogastric tube is carefully guided across the anastomosis. The anterior opening of the anastomosis is then closed in one or two layers. For a two-layer closure, the inner mucosal layer is closed with running 4-0 PDS (polydioxanone sutures; Ethicon), while the outer layer is performed with interrupted 4-0 PDS sutures.⁹⁰

Key Question: Esophagectomy

What constitutes optimal management of T1N0 esophageal adenocarcinoma?

INTRODUCTION

Esophageal adenocarcinoma remains the fastest increasing cancer in the United States and develops as a consequence of gastroesophageal reflux disease and Barrett’s esophagus. Increasingly, patients with early stage disease are being identified, either fortuitously related perhaps to more liberal use of upper endoscopy or as part of a Barrett’s surveillance program. In contrast to combination therapy with chemotherapy or concurrent chemoradiotherapy followed by esophagectomy, which has been demonstrated to improve survival in local and regionally advanced tumors,⁹¹,⁹² the optimal treatment strategy for early stage disease is still being defined. A great majority of those with early stage esophageal cancer will be cured of their disease. The introduction of endoscopic resection and ablation techniques has revolutionized the therapy for superficial esophageal cancer and allowed organ preservation rather than esophagectomy in appropriate patients.⁹³ Therefore, a critically important concept guiding therapy is that long-term QOL and alimentary satisfaction are necessary considerations in these patients.

From an oncologic standpoint, endoscopic eradication therapy can be applied safely only when patients are at low risk of lymph node metastases.⁹⁴,⁹⁵,⁹⁶,⁹⁷,⁹⁸,⁹⁹,¹⁰⁰,¹⁰¹,¹⁰²,¹⁰³,¹⁰⁴,¹⁰⁵ Unfortunately, typical staging studies, such as endoscopic ultrasound (EUS) and CT with positron-emission therapy (CT-PET), are of limited utility in patients with superficial tumors¹⁰⁶,¹⁰⁷,¹⁰⁸,¹⁰⁹; therefore, the risk of nodal metastases must be evaluated on the basis of tumor characteristics including size, depth of invasion, degree of differentiation, and presence of lymphovascular invasion. The decision for endotherapy or esophagectomy also should depend on patient factors, including the length of the Barrett’s esophagus segment, severity of gastroesophageal reflux, presence of comorbid conditions, and patient compliance and willingness to undergo multiple procedures.⁹³,⁹⁴,⁹⁶,¹¹⁰ The following sections will review the evaluation, treatment, and outcomes in patients with T1 esophageal adenocarcinoma and serve to guide clinicians in the selection of the most appropriate therapy for each patient.

METHODOLOGY

The English-language literature was queried via PubMed to include articles published between 1990 and 2016. Search terms were generated from the key statements: “Endoscopic eradication therapy results in equivalent overall survival when compared to esophagectomy for clinical T1aN0 esophageal adenocarcinoma,” and “Esophagectomy provides better overall survival than endoscopic eradication therapy for cT1b esophageal adenocarcinoma.” Specific search terms included adenocarcinoma, Barrett esophagus, esophageal neoplasms, ablation techniques, catheter ablation, cryosurgery, esophagoscopy, photochemotherapy, and esophagectomy (see Fig. 7-12). No RCTs comparing endoscopic eradication therapy with surgery for clinical T1N0 esophageal adenocarcinoma or subgroups (T1a or T1b) were found with this strategy. Filters were
then applied to include nonrandomized clinical trials including retrospective case series and comparative studies. Abstracts were reviewed and included if they met criteria of adenocarcinoma in all or most of the cohort, had >30 patients, provided a comparison between endoscopic eradication therapy and surgery, and described long-term outcomes after treatment with either endoscopic eradication therapy or surgery. Studies were excluded if they were case reports or reviews, contained cohorts of either all or mostly squamous cell carcinoma, did not contain patients with invasive adenocarcinoma, or did not report either long-term overall survival or disease-related survival. Full-text review of the selected articles applied the same inclusion and exclusion criteria. Articles selected were assessed according to the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) system, and Table 7-2 was created. 111 Recommendations were made based on the strength of available evidence.

FIGURE 7-12 Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram: T1N0 esophageal adenocarcinoma.

TABLE 7-2 Results and Characteristics of Observational studies of T1N0 Esophageal Adenocarcinoma

Author, Year (Study Years)	Patients	Study Type	Procedure(s)	Outcomes	Results	Conclusions	Comments
Ancona et al, 2008⁹⁴ (1980-2006)	Total: 98; T1a: 27; T1b: 71; SM1: 36; SM2: 7; SM3:28; SCC: 67; ADC: 31	Retrospective	Ivor Lewis and McKeown	R0 resection Lymph node metastasis Angiolymphatic invasion Lymphocytic infiltration 5-yr overall survival (%) 10-yr overall survival (%)	95 (97%) T1a: 0; T1b: 20 (28%) Absent: 53 (54%); present: 45 (45.9%) High grade: 34 (34.7%); low grade: 64 (65.3%) T1: 56.7; T1a: 77.7; T1b: 53.3; T1a/SM1: 67.3; SM2/3: 44.2; SCC: 43.3; ADC: 100 SCC: 37.3; ADC: 90.9	Independent predictive factors for lymph node metastasis were angiolymphatic invasion, depth of invasion, low-grade lymphocytic invasion, and neural invasion.	Survival reporting by depth of invasion was of a mixed population of SCC and ADC. Survival reported according to histologic subtype did not further segregate by depth of invasion.
Badreddine et al, 2010¹⁴³ (1997-2007)	Total: 80; T1b: 80; sm1:31; sm2: 23; sm3: 26; ADC: 100%	Retrospective	Ivor Lewis: 56; THE: 24	Mean follow-up (mo) In-hospital mortality 30-d mortality Lymph node metastasis LVI Overall survival (mo) Disease-free survival (mo)	40.5 ± 4 5 (6 %) 5 (6%) T1b: 14 (17.5%); SM1: 4 (12.9%); SM2/3: 10 (20.4%) T1b: 27 (34%); SM1:10 (32%); SM2/3: 17 (35%) T1b: 46.5; SM1: 52; SM2/3: 38 T1b: 30; SM1: 35.3; SM2/3: 27.8	Superficial submucosal (sm1) invasion is a risk factor for lymph node metastasis (12.9%). Overall and cancer-free survival were not different between superficial submucosal (sm1) and deeper submucosal (sm2/3) invasion.	All cases were adenocarcinoma. Adjuvant chemoradiation was administered to five patients with metastatic lymphadenopathy.
Barbour, 2010¹¹³ (1991-2008)	Tot: 85; T1a: 35; T1b: 49; Unknown: 1; ADC: 100%	Retrospective	Ivor Lewis: 6; MIE: 73; total gastrectomy: 5; proximal gastrectomy: 1	Median follow-up (mo) In-hospital mortality R0 resection Lymph node metastasis LVI 5-yr overall survival (%) Disease-free survival (%)	59 0 85 (100%) T1a: 0; T1b: 8; Unknown: 1 No: 71 (83.5%); Yes: 14 (16.5%); T1a: 1; T1b: 13 T1a: 97; T1b: 65; SM1: 70 (NS); SM2: 60; SM3:71 T1a: 100; T1b: 70; SM1: 82 (NS); SM2: 60; SM3: 82	Four risk groups for lymph node metastases—I: T1a (0%); II: T1b, well/moderate differentiation/no LVI (4%); III: T1b, poor differentiation/no LVI (22%); IV: T1b with LVI (46%)	Overall and diseasefree survival reported for the T1b cohort included eight patients with nodal metastasis (N1).
Bogoevski et al, 2011⁹⁵ (1992-2007)	Total: 113; HGIEN: 16; T1a: 41 T1b: 56 SCC: 56 ADC: 57	Retrospective	THE (LR): 51; LR of the esophagogastric junction: 21; thoracoabdominal (ER): 41	Median follow-up (mo) In-hospital mortality R0 resection Lymph node metastasis Overall survival (all) Overall survival T1b	39 3 (2.7%) 113 (100%) T1a: 0; T1b: 15 (27%) LR: 203 mo, 5-yr 67.7%; ER: 88 mo, 5-yr 57.8% LR: 63 mo, 5-yr 51.5%; ER: 85 mo, 5-yr 53.3%	Patients with T1b submucosal invasion have a high (27%) rate of lymph node metastasis making them unsuitable for endoscopic treatment. LR esophagectomy is a valuable alternative to extended esophagectomy for T1b tumors with no difference in overall survival.	The survival outcome reported mixed ADC and SCC histologies. Survival was not reported for T1a tumors and overall survival included 16 patients with HGIEN.
Bolton et al, 2009⁹⁶ (1979-2007)	Total: 133; T1a: 64; T1b: 69; ADC: 133	Retrospective	Ivor Lewis: 57; THE: 63; 3-field: 7; MIE: 6	Median follow-up (mo) R0 resection Lymph node metastasis LVI 3-yr overall survival (%) 5-yr overall survival (%)	42 131 (98.5%) T1a: 5%; T1b:23% No: 102 (77%); Yes: 23 (17%); Missing: 8 (6%); T1a: 7%; T1b: 29% T1a: 100; T1b: 90 T1a: 90; T1b: 69	In combination with submucosal invasion, tumor length >3 cm identifies a group of patients with T1 esophageal adenocarcinoma at high risk of lymph node involvement and decreased survival.	All adenocarcinoma. Survival reported according to tumor depth included patients with nodal disease.
Estrella et al, 2011⁹⁸ (1997-2010)	Total: 99; T1a: 69; T1b: 30; ADC: 99	Retrospective	THE: 43; Ivor Lewis: 41; MIE: 10; 3-field: 5	Lymph node metastasis LVI Mean RFS (mo)	T1a: 1 (1.4%); T1b: 10 (33%) All: 23 (23%); T1b: 18 (60%) T1a: 108; T1b: 124
Kaneshiro et al, 2011¹¹⁴ (1983-2008)	Total: 185; T1a: 150; T1b: 35 (sm1 only)	Retrospective	Esophagectomy	Mean follow-up (mo) Lymph node metastasis Overall survival	T1a: 7.7±4.6; T1b: 7.1±4 T1a: 1 (0.7%); T1b: 3 (8.6%) No difference between groups		No survival difference seen according to depth of invasion. T1a survival was reported according to the histologic sublayer and not in total.
Kauppi et al, 2013¹¹⁵ (1984-2001)	Total: 85; T1a: 35; T1b: 44; Unknown: 6; ADC: 85	Retrospective	En bloc transthoracic: 36; THE: 39; vagalsparing esophagectomy: 5; EMR: 5	Median follow-up (yr) In-hospital mortality R0 resection (%) Lymph node metastasis 5-yr overall survival (%) 10-yr overall survival (%) 5-yr disease-specific survival (%) 10-yr disease-specific survival (%) 5-yr RFS (%) 10-yr RFS (%)	5 4 (5%) Surgery: 100; EMR: 100 (number of resections: median 2, range 2-7) T1a: 1/35 (3%); T1b: 6/44 (14%) T1a: 73; T1b: 62 T1a: 58; T1b: 40 T1a: 92; T1b: 72 T1a: 86 T1b: 68 T1a: 89; T1b: 70 T1a: 89; T1b: 70	Recurrence of superficial ADC was the most important determinant of 5-year survival. Risk of recurrence of intramucosal tumors may be higher than reported. Beyond 5 years age-related disease is most predictive of mortality.	All adenocarcinoma. Survival reported according to tumor depth included patients with nodal disease.
Leers et al, 2011⁹⁹ (1985-2008)	Total: 126; T1a: 75; T1b: 51; ADC: 126	Retrospective	En bloc esophagectomy: 64; THE: 56; transthoracic esophagectomy: 6	Median follow up (mo) LVI present Lymph node metastasis 5-yr overall survival (%) 5-yr disease-specific survival (%)	50 T1a: 6 (8%); T1b: 22 (43%) T1a: 1 (1.3%); T1b: 11 (22%) (P = 0.0003); SM1: 21%; SM3: 26% T1: 78; T1a: 82; T1b:71 (P = 0.25) T1a: 98; T1a: 78 (P = 0.001); N0: 92; N1: 70 (P = 0.001)	No significant difference in overall survival between T1a and T1b tumors. No difference in lymph node metastases between SM1 and SM3.	All ADC. Main outcome measure was prevalence of lymph node metastases. Patients undergoing esophagectomy where < six lymph nodes were harvested were excluded.
Lorenz et al, 2014¹⁰⁰ (2000-2012)	Total: 168; T1a: 42; m1-2: 10; m3-4: 32; T1b: 126; sm1: 37; sm2: 33; sm3: 56; ADC: 168	Retrospective	En bloc esophagectomy: 130; limited esophagectomy: 30; THE: 2; MIE: 6	Median follow-up (mo) Lymphatic invasion present Vascular invasion present Lymph node metastasis Hospital mortality (%) 5-yr overall survival (%) Tumor-specific survival (%)	64 T1a: 13 (31%); m1-2: 0; m3-4: 13 (31%); T1b: 50; sm1: 12 (24%); sm2: 12 (24%); sm3: 26 (52%) T1a: 0; m1-2: 0; m3-4: 0; T1b: 9; sm1: 3; sm2: 1; sm3: 5 T1a: 4; m 4: 4; T1b: 26; sm1: 3; sm2: 9; sm3: 14 2.97 All: 79; m1-2: 100; m3-4: 89.2; sm1: 83.5; sm2: 67.8; sm3: 70.4 (NS) All: 79; m1-2: 100; m3-4: 96.4; sm1: 92.3; m2: 89.2 sm3: 81 (NS)	The presence of lymph node metastasis was the most significant factor for 5-yr survival: 87.1 vs 56% (P <0.001). Lymph node infiltration was the only prognostic factor for overall survival, tumor-specific survival, and tumor recurrence in multivariate analysis. On univariate analysis, deeper submucosal infiltration (sm2,3) correlated with overall survival.	All ADC.
Manner et al, 2013¹³⁰ (1996-2010)	Total: 66; T1b sm1: 66; ADC: 66	Retrospective	ER + APC: 66	Mean follow-up (mo) CER Number of ERs to obtain CER Long-term remission Metachronous neoplasias Mean period from CER to diagnosis of metachronous neoplasias (mo) Lymph node metastasis 5-yr overall survival (%)	47±29.1 Overall: 53/61 (86.9%); lesions <2 cm: 30/31 (97%); lesions ≥2 cm: 23/30 (77%) 2.6±2.9 Overall: 51/61 (83.6%); lesions <2 cm: 28/31 (90%); lesions ≥2 cm: 23/30 (77%) 10/53 (19%) 22 1/53 (1.9%) 84	Risk of developing lymph node metastases after EMR is lower than the risk of surgery. No tumorassociated deaths were observed.	Main outcome measure was CER in T1b ADC. After EMR, thermal ablation of the residual Barrett’s segment was done for prophylaxis of metachronous lesions in 79% of patients. All MN treated endoscopically.
Manner, et al, 2017²²⁹	Total: 62; T1b sm2: 23; T1b sm3:39; ADC: 62	Retrospective	T1b sm2—hisLR: endoscopic resection: 3; esophageal resection: 9; hisHR: endoscopic resection: 1; esophageal resection: 10. T1b sm3—hisLR: Endoscopic resection: 0; Esophageal resection: 7; hisHR: endoscopic resection: 0; esophageal resection: 7. Note: A total of 58 patients underwent esophageal resection for whom pT1b sm2 vs. sm3 status was not clarified.	Lymph node metastasis Lymph node metastasis per histologic risk 30-d mortality of surgery Mean follow-up with EUS after endoscopic resection (mo)	T1b sm2: 5/23 (21.7%); T1b sm3:14/39 (35.9%) T1b sm2—hisLR: 8.3% (1/12); hisHR: 36.3% (4/11); combLR: 0% (0/5); combHR: 27.8% (5/18). T1b sm3—hisLR: 28.6% (2/7); hisHR: 37.5% (12/32); combLR: 25% (1/4); combHR: 37.1% (13/35) 1.7% (1/58) 41-43	In ADC with pT1b sm2/3 invasion, the frequency of lymph node metastasis depends on macroscopic and histologic risk patterns. The rate of lymph node metastasis appears to be higher than the mortality risk of surgery. Whether a highly selected group of pT1b sm2 patients with a favorable risk pattern may be candidates for endoscopic therapy cannot be determined until the results of larger case volumes are available.	There was no treatmentrelated mortality of ET. Rate of lymph node metastasis was analyzed depending on risk patterns—hisLR): G1-2, L0, V0; hisHR: ≥1 criterion not fulfilled; macLR: gross tumor type I-II, tumor size ≤2 cm; macHR: ≥1 criterion not fulfilled; combLR: hisLR1macLR; combHR: at least 1 risk factor.
Ngamruengphong et al, 2013¹¹⁶ (1998-2009)	Total: 1,618; Endoscopic—Tis: 55; T1a: 174; T1b: 39; T1, NOS: 38; Surgery—Tis: 110; T1a: 561; T1b: 523; T1, NOS: 118	Retrospective from SEER database	Endoscopic therapy: 306 (19%); ER: 149 (48.7%); ER NOS: 61 (19.9%); ER + PDT: 16 (5.2%); ER + electrocautery: 16 (5.2%); ER + laser: 15 (4.9%); ER + cryosurgery 3 (1.0%); PDT: 29 (9.5%); laser treatment: 10 (3.3%); cryosurgery: 5 (1.6%); electrocautery: 1 (0.3%); ablation, NOS: 1 (0.3%); surgery: 1,312 (81%); Partial esophagectomy + partial gastrectomy: 608 (46.3%); partial esophagectomy + gastrectomy, NOS: 210 (16.0%); partial esophagectomy: 237 (18.1%); total esophagectomy: 79 (6.0%); total esophagectomy + gastrectomy, NOS: 51 (3.9%); esophagectomy, NOS: 127 (9.7%)	Median follow-up (mo) Overall survival for T1b vs Tis/T1a (HR) ECSS T1b vs Tis/T1a (HR) Overall survival for surgery vs. ET (HR) ECSS surgery vs. ET (HR) 5-yr overall survival Tis/T1 (%) 5-yr overall survival Tis/T1a (%) 5-yr overall survival T1b (%) 5-yr ECSS Tis/T1 (%) 5-yr ECSS Tis/T1 (%) 5-yr ECSS Tis/T1a (%) 5-yr ECSS T1b (%) Deep margin on EMR	36 1.37 (95% CI, 1.12-1.69) 1.65 (95% CI, 1.28-2.16) Tis/T1: 1.21; Tis/T1a: 1.29; T1b: 0.97 (NS) Tis/T1: 0.74; Tis/T1a: 0.69; T1b: 0.46 (NS) ET: 58; surgery: 70 (P = 0.003) ET: 60; surgery: 76 (P = 0.002) Surgery: 66; ET: 64 (P = 0.26) Surgery: 81; ET: 77 (P = 0.26) Surgery: 81; ET: 77 (P = 0.26) Surgery: 83; ET: 84 (P = 0.64) Surgery: 89; ET: 71 (P = 0.31) Negative: 53; positive: 26; not reported: 2	Older age and radiation therapy were associated with worse overall survival and ECSS. Submucosal invasion was significantly associated with worse overall survival and ECSS. Outcomes were not different for surgery compared with ET when stratified by depth of invasion.	All ADC. 15% received radiation therapy.
Nurkin et al, 2014¹¹⁷ (2001-2012)	Total: 81; Tis: 37; T1a: 10; at least T1a: 13; T1b: 6; at least T1b: 15	Retrospective	EMR: 77; ESD: 4; esophagectomy: 14	Median follow-up (yr) 5-yr overall survival (%) 5-yr disease-free survival (%)	3.25 T1a: 83 T1a: 100; T1b: 100	ER can result in a cure for T1a disease. Patients with T1b disease, low-risk features, and negative deep margins can be observed. Obtaining a negative margin on EMR was a predictor of improved 5-yr disease-free and overall survival.	Of T1b patients, only two were observed, two had esophagectomy, and two chemoradiotherapy. Series is too small to support the conclusion for T1b.
Oh et al, 2006¹¹⁸ (1987-2006)	Total: 78; T1a: 78	Retrospective	En bloc esophagectomy: 23; THE: 31; vagal-sparing esophagectomy: 20; transthoracic esophagectomy: 4	5-yr overall survival (%) Lymph node metastasis (in patients undergoing en bloc resection) Overall mortality rate	88% 1/23 (4%) 2.6%	Vagal-sparing technique had less morbidity than other forms of resection and no mortality. Survival after all types of resection was similar.
Pech et al, 2005¹²³ (1996-2002)	Total: 66; HGD: 35; T1a: 31	Prospective	PDT-ALA	Median follow-up (mo) Complete remission Recurrence 5-yr overall survival (%) 5-yr disease-specific survival (%)	HGD: 37; T1a: 36 HGD: 34/35 (97%); T1a: 31/31 (100%) HGD: 6/34 (18%); T1a: 9/31 (29%) HGD: 97; T1a: 80 HGD: 89; T1a: 68	Local recurrence in 10 patients in the T1a group.
Pech et al, 2011¹¹⁹ (1996-2009)	Total: 114; T1a: 114	Retrospective	Transthoracic esophagectomy: 38; ER + APC: 76	Median follow-up (yr) Complete remission Major complications (%) 90-d mortality rate Overall rate of recurrence/metachronous neoplasia (%) Long-term remission rates allowing for repeat endoscopic therapy (%) Disease-free survival rate Lymph node metastasis	Esophagectomy: 3.7; ER + APC: 4.1 Esophagectomy: 38/38 (100%); ER + APC: 75/76 (98.7%) Esophagectomy: 32; ER + APC: 0 Esophagectomy: 1/38 (2.6%); ER + APC: 0/76 (0%) ER + APC: 6.6 Esophagectomy: 100; ER + APC: 98.7 Esophagectomy: 100%; ER + APC: 75/76 (98.7%) Esophagectomy: 0/38 (0%); ER + APC: 0/76 (0%) on follow-up EUS	Esophagectomy was associated with a higher morbidity rate and a mortality rate of 2.6% vs. 0% mortality in the EMR cohort (difference between cohorts not significant). Recurrence rate was higher in the EMR cohort (difference not significant). Age was a prognostic factor for survival (HR, 1.13; 95% CI, 1.04-1.23 for each year, with P = 0.005).	Baseline characteristics were comparable between the two cohorts. Only significant difference between cohorts was Barrett’s length with shorter Barrett segments in the EMR cohort.
Pech et al, 2014⁹³ (1996-2010)	Total: 1,000; T1a: 1,000	Prospective	ER	Mean follow-up (mo) Complete remission of neoplasia Long-term complete remission rate Recurrence of neoplasia (HGD or adenocarcinoma)/metachronous lesions 5-yr overall survival (%) 10-yr overall survival (%) 5-yr disease-specific survival (%)	56.6 963 (96.3%) 938 (93.8%) 140/963 (14.5%) 91.5 75 87.1	No mortality and major complications were <2% (bleeding, perforation). Only factor negatively associated with complete remission was presence of long segment BE (P = 0.0001). Poorly differentiated mucosal ADC (G3) was associated with significantly higher risk of recurrence and failure of endoscopic treatment than patients with well-differentiated or moderately differentiated mucosal ADCs.	Repeat endoscopic treatment was successful in 82.1%. Surgery was necessary in 3.7% after failure of endoscopic therapy.
Peyre et al, 2007¹²⁰	Total: 109; HGD: 36; T1a: 73	Retrospective	Vagal-sparing esophagectomy: 49; THE: 39; en bloc esophagectomy: 21	30-d mortality (%) Duration of follow-up (mo)	Vagal-sparing esophagectomy: 2; THE: 5; en bloc esophagectomy: 0 Vagal-sparing esophagectomy: 39; THE: 25; en bloc esophagectomy: 60	Length of hospital stay and the incidence of major complications was significantly reduced with vagalsparing esophagectomy compared with transhiatal or en bloc resection. Recurrent cancer has developed in only one patient.
Prasad et al, 2009¹¹⁰ (1998-2007)	Total: 178; T1a: 178	Retrospective	Endoscopic: 132; EMR alone 75 (57%); EMR+PDT 57 (43%); surgery: 46; THE: 20 (43%); transthoracic: 29 (57%)	Mean follow-up (mo) Median number of endoscopic treatments Endoscopic remission rate Lymph node metastasis Follow-up (person-yr) Cumulative mortality Overall mortality, incidence rate (person-yr) Recurrent cancers Recurrence, incidence rate (person-yr) 5-yr overall survival (%) 5-yr disease-specific survival (%)	Endo: 43; Surgery: 64 1 (IQR, 1-2) 124/132 (94%) Surgery: 4/46 (8.6%) Endoscopic: 464.6; surgery: 244 Endoscopic: 23/132 (17%); surgery: 9/46 (20%) Endoscopic: 4.9/100; surgery: 3.7/100 Endoscopic: 16 (12%); surgery: 1 (2%) Endoscopic: 5.5/100; surgery: 0.56/100 Endoscopic: 83; surgery: 95 Endoscopic: 80; surgery: 97	Cumulative mortality in the endoscopic group (17%) was comparable with the surgery group (20%) (P = 0.75). Possible predictors of recurrence: presence of residual BE (without dysplasia), increasing length of BE segment at baseline, and presence of a carcinoma detected after 6 months of surveillance for HGD.
Rice et al, 2001¹⁰¹ (1985-1999)	Total: 122; HGD: 38; T1a N0: 51; T1a N1: 2; T1b N0: 25; T1b N1: 6	Retrospective	Esophagectomy—THE: 75; thoracotomy/abdominal incision: 46; laparotomy: 1	Follow-up (mo) Operative mortality Lymph node metastasis Overall survival (%) Overall survival:T1a Overall survival: T1b (%) Number of deaths (n)	Mean: 47±41; median: 38 2.5% (3/122) T1a: 2/122 (2%); T1b: 6/122 (5%) 5-yr: 77; 10-yr: 68 5-yr: 77%; 10-yr: 65% 5-yr: 61; 10-yr: 40 (P = 0.009) Cancer related: 13; noncancer related: 6	Depth of invasion predicted survival on univariate analysis but on multivariate analysis, lymph node metastasis was most predictive of poor long-term survival.
Saha et al, 2009¹²¹ (2000-2009)	Total: 55; T1: 44; T1b: 11	Retrospective	Esophagectomy—Ivor Lewis: 24; THE: 20	Median followup (mo) Operative mortality Complications Lymph node metastasis 1-yr disease-free survival (%) 3-yr survival (%) Recurrence rate (%)	44 2 (5%) 9 (20%) 2 (4%, both T1b) 95 93 7	Differentiation appears to be a marker of poor prognosis.
Sepesi et al, 2010¹⁰² (2000-2008)	Total: 54; T1a: 25; T1b: 29; SM1: 14; SM2: 11; SM3: 6	Retrospective	En bloc esophagectomy: 10; THE: 39; Ivor Lewis: 2; vagal sparing esophagectomy: 2; McKeown: 1	Median follow-up (mo) LVI Lymph node metastasis 5-yr overall survival, estimated (%) Died with recurrence	42 7 (13%) Total: 9 (17%); T1a: 0 SM1: 3/14 (21%); SM2: 4/11 (36%); SM3: 2/4 (50%) T1a: 85; T1b: 60 (P = 0.08) 5/15 (33%)	Nodal metastasis rate was significantly different for T1a and T1b tumors but not different by depth of submucosal invasion. Nodal metastasis in submucosal tumors was high.
Sgourakis et al, 2013¹²⁴ (1997-2011)	Total: 4,241; endoscopic: 2,092; LGD: 4%; HGD/CIS: 33.6%; T1a: 54%; T1b: 16%; surgery: 2,149 included ADC and SCC	Systematic review	Endoscopic therapy—EMR and/or ESD (42 studies); APC (3 studies); PDT (2 studies); RFA (2 studies); surgery—various (38 studies)	LVI Microvascular invasion Lymph node metastasis (11 studies) Mean follow up (mo) EMR resection margins (18 studies) Endoscopic local recurrence (30 studies) Development of metachronous lesions after endoscopic management (10 studies)	EMR and/or ESD: 0%-30% EMR and/or ESD: 0-33% Endoscopic—31/371 (8%); ADC: 5%; SCC: 1%; surgery—888/2,149 (41%); T1b: ADC: 26%; SCC: 45% Endoscopic: 12-62 Positive: 294/880 (3%); ADC: 9%; SCC: 7% 0%-17%; ADC: 0.8%; SCC: 1% 2%-14%; ADC: 6%; SCC: 1%	In endoscopically managed patients, local tumor recurrence was predicted by poor differentiation and piecemeal resection; lymph node positivity by LVI. In surgically resected patients for ADC, the best predictor for lymph node positivity was LVI.	Includes both ADC and SCC. A significantly greater number of SCC patients were submitted to surgery than ADC patients. For surgically resected patients, there were significant differences in patients with positive lymph nodes between ADC and SCC.
Stein et al, 2000¹⁰⁴ (1982-1999)	Total: 94; T1a: 38; T1b: 56	Retrospective	Esophagectomy—subtotal: 71; limited resection/jejunal interposition: 24	Median follow-up (mo) 30- and 90-d mortality (%) Complications Lymph node metastasis 5-yr overall survival (%)	15 Subtotal: 4.2; limited: 0 Subtotal: 31/71 (43.7%); limited: 5/24 (20.8%) T1a: 0; T1b: 10/56 (18%) T1a: 85; T1b: 78	No survival difference T1aN0 and T1bN0. Lymph node metastasis had a significant effect on survival.
Stein et al, 2005¹⁰³ (1990-2004)	Total: 290; ADC: 157; SCC:133; HGD: 14; T1a: 82; ADC: 57; SCC: 25; T1b: 194; ADC: 87; SCC: 107	Retrospective	Esophagectomy—abdominothoracic: 146; radical transhiatal: 144	Median follow-up (mo) Postoperative mortality Lymph node metastasis ADC 5-yr overall survival (%)	66 5/290 (1.7%) T1a: 0; T1b: 18/87 (20.7%) ADC: 83.4; SSC: 62.9	Presence of lymph node metastasis and histologic tumor type were the only significant factors affecting survival favoring node negative disease and adenocarcinoma.
Tian et al, 2011¹⁴⁴ (1995-2010)	Total: 68; all T1b	Retrospective	EMR: 68; esophagectomy: 39	In-hospital mortality Lymph node metastasis at esophagectomy Median survival (mo) Overall mortality (%) 5-yr survival (%) Risk factors for overall survival, univariate (HR, 95% CI)	1/39 (2.6%) 13 (33%) All: 39.5 (IQR, 23.9-70.3); EMR: 34.8 surgery: 48.9 (P = 0.09) EMR: 37.9; surgery: 25.6 (P = 0.28) All: 69.5 Surgery: 0.4, 0.2-1.1 (P = 0.09); lymph node metastasis: 6.8, 1.4-33.8 (P = 0.02)	Patients who underwent esophagectomy after EMR lived longer but not significantly so. Most of the EMR-only mortality was in patients deemed not fit for esophagectomy.	29 patients did not undergo esophagectomy and were treated with CRT (5), continued endoscopic therapy (15), both (5), or neither (4).
Westerterp et al, 2005¹⁰⁵	Total: 120; HGD: 13; T1: 107	Retrospective	THE	Median follow-up (mo) Lymph node metastasis 5-yr recurrence free (%) 5-yr RFS (%) Recurrence-free period risk for T1N0 5-yr disease-free survival (%)	44 T1a/T1b-sm1: 1/79 (1.3%); T1-sm2/3: 18/41 (44%) T1a/T1b-sm1: 97; T1-sm2/3: 57 (P <0.0001) T1a/T1b-sm1: 80; T1-sm2/3: 42 (significant) T1sm2-3 HR, 7.5, (95% CI, 2.0-2.7 vs. T1m1-3/sm1) 68 (all)	N-stage was the only independent predictor of recurrence at 5 years but when excluded, depth of invasion was significant.
SCC, squamous cell carcinoma; ADC, adenocarcinoma; THE, transhiatal esophagectomy; LVI, lymphovascular invasion; MIE, minimally invasive esophagectomy; HGIEN, high-grade intraepithelial neoplasia; LR, limited resection; ER, extended resection; EMR, endoscopic mucosal resection; RFS, relapse free survival; APC, argon plasma coagulation; CER, complete endoluminal remission; MN, metachronous neoplasias; hisLR, histologically low risk; hisHR, histologically high risk; combLR, combined low risk; combHR, combined high risk; macLR macroscopically low risk; macHR, macroscopically high risk; NOS, not otherwise specified; PDT, photodynamic therapy; HR, hazard ratio; ECSS, esophageal cancer specific survival; NS, not significant; ESD, endoscopic submucosal dissection; HGD, high grade dysplasia; ET, endoscopic therapy; ALA, 5-delta-aminolevulinic acid; CI, confidence interval; IQR, interquartile range; BE, Barrett’s esophagus; LGD, low-grade dysplasia; CIS, carcinoma in situ; RFA, radiofrequency ablation; CRT, chemoradiotherapy.