Radical esophagectomy is one of the most stressful and invasive procedures due to the inherent risk of high mortality and morbidity rates. Recent large series have reported that minimally invasive esophagectomy (MIE) results in low incidences of pneumonia and mortality and has midterm, stage-specific survival rates similar to those of open esophagectomy [1–3]. Even with this new data, questions regarding the adequacy of lymph node clearance in MIE still are valid [4, 5].

For esophageal squamous carcinoma, extended or total mediastinal lymphadenectomy (ML) is recommended due to the high frequency of lymph node metastasis along the bilateral recurrent laryngeal nerves (RLNs) [6–8]. Lymphadenectomy along bilateral RLNs has remained a challenge in MIE due to the loss of depth perception on the two-dimensional monitor, the limitation of instrumental freedom, and the long training period [9–12].

After the first series of robot-assisted thoracoscopic esophagectomies (RATEs) had been introduced [13, 14], we reported the feasibility and safety of RATE with extended ML with the patient in the prone position in 2010 [15]. Based on our experience, we started performing RATE with total ML with the patient in the semi-prone position in October 2010. Because total ML includes a dissection of the left paratracheal, left RLN, and infraaortic nodes in addition to the extended ML [16], surgeons must be trained extensively before they are proficient with the minimally invasive approach. We hypothesized that the advantages of a robotic system (three-dimensional view, easier articulation of instruments, and tremor filtering) would facilitate the dissection of lymph nodes along the bilateral RLNs with acceptable mortality and morbidity rates.

This study aimed to investigate the feasibility and safety of a robot-assisted thoracoscopic lymphadenectomy along bilateral RLNs in radical esophagectomy for esophageal squamous cell carcinoma.

Materials and methods

This retrospective study was approved by the institutional review board of the hospital (IRB no. 4-2013-0180). Between October 2010 and July 2012, 40 consecutive patients who underwent RATE and total ML for esophageal squamous cell carcinoma were reviewed retrospectively. Preoperative endoscopic biopsy was performed for all the patients to confirm the histologic diagnosis of esophageal cancer. The staging workup included a thorough history and physical examination, esophagogastroduodenoscopy, endoscopic ultrasonography, and an integrated positron emission tomography/computed tomography scan. The surgical indications were similar to those for open esophagectomy.

For resectable T1–T2 tumors, surgery was followed by adjuvant chemotherapy or chemoradiation therapy according to the pathologic stage. For T3–T4a tumors below the carina, neoadjuvant concurrent chemoradiation therapy (CCRT) was used to ensure R0 resection at the time of surgery. When a tumor was located above the carina or when metastasis to both RLN chains was confirmed by frozen section diagnosis, bilateral cervical lymphadenectomy also was performed.

Lymph nodes along the dorsal side of the RLNs were removed in the initial 18 patients (group 1), and RLNs were skeletonized by dissection of all the lymph nodes around the nerve in the next 22 patients (group 2) (Fig. 1A, B).

Fig. 1
figure 1

Extent of lymphadenectomy along bilateral RLNs in group 1 (A, C) and group 2 (B, D). In group 2, the left RLN was skeletonized to remove all lymph nodes around the nerve. After the procedure, the left common carotid artery (asterisk) was identified at the base of the dissection field. SCA left subclavian artery, CCA left common carotid arterym N left recurrent laryngeal nerve, T trachea, AA aortic arch

Full size image

Port placement and setup

The patient was intubated with a Univent bronchial blocker tube (Fuji Systems Corp., Tokyo, Japan) and turned to the semi-prone position. After carbon dioxide insufflation at 8–10 mmHg, four trocars were placed with inspection of the pleural space as follows: a 12-mm trocar at the 8th intercostal space for a 30° angled 10-mm thoracoscope, an 8-mm trocar at the 6th intercostal space medial to the scapula at the posterior axillary line for a right robotic arm, an 8-mm trocar at the 10th intercostal space for a left robotic arm, and a 12-mm trocar at the 7th intercostal space along the midaxillary line for an accessory port (Fig. 2A). A da Vinci robotic cart (Intuitive Surgical, Mountain View, CA, USA) was introduced from the left cranial side of the patient (Fig. 2B).

Fig. 2
figure 2

Robotic port placement (A) and operating room setup (B). C camera port, A accessory port, OS operating surgeon, AS assistant surgeon, T trainee

Full size image

En bloc resection of the middle and lower esophagus

The arch of the azygos vein was isolated and cut after clips had been applied. The mediastinal pleura overlying the esophagus was incised, and the left side of the upper thoracic esophagus was mobilized from the trachea with care taken to avoid injury to the membranous portion of trachea. Dissection continued down to the left side of the middle thoracic esophagus, and the left vagus nerve was cut below the level of its pulmonary branch. The thoracic duct was clipped at the level of the diaphragm.

After dissection to the hiatus, the right side of the esophagus was dissected with periesophageal tissues. Once the right vagus nerve had been cut below its pulmonary branch, subcarinal lymph nodes were dissected. The thoracic duct, mediastinal pleura, and lymph nodes at paraesophageal, subcarinal, and peribronchial stations were dissected to remain en bloc with the esophagus.

Lymphadenectomy along bilateral RLNs

After en bloc resection of the middle and lower esophagus had been completed, dissection continued along the right vagus nerve, and the right RLN was identified at the lower margin of the right subclavian artery. The lymph nodes at the medial part of the right RLN were dissected in group 1 (Fig. 1A), and the lymph nodes with their associated fat pads around the right RLN were dissected completely in group 2 (Fig. 1B). The trachea was retracted with a grasper by an assistant, and the left RLN was identified below the aortic arch.

After dissection of infraaortic and left tracheobronchial nodes, the lymph nodes along the left RLN were dissected up to the thoracic inlet. In group 1, only the dorsal side of the left RLN was dissected (Fig. 1A, C), whereas in group 2, the lymph nodes with their associated fat pads were completely dissected along the nerve (Fig. 1B, D).

After completion of the total ML, a 28-Fr chest tube was placed, and the collapsed right lung was inflated. The patient then was turned to the supine position.

Gastric mobilization and cervical esophagogastrostomy

Gastric mobilization and upper abdominal lymph node dissection were performed using laparoscopy as previously reported [15]. At the end of the abdominal phase, left cervicotomy was performed, and the cervical esophagus was mobilized. A horizontal collar incision was chosen when bilateral cervical lymphadenectomy was indicated. The esophagogastric specimen was pulled out through the neck incision under laparoscopic observation. Esophagogastric anastomosis was performed with a side-to-side stapled anastomosis.

We recorded the operation time at each step of the procedure. The robot console time began at the start of dissection with the robot and ended at completion of the thoracic phase. The total operation time started at the skin incision for trocar placement in the thoracic phase and ended at skin closure of the neck and abdominal wounds.

Postoperative care

The patients were extubated in the operating room and then transferred to the intensive care unit for monitoring. Analgesia was achieved by intravenous patient-controlled analgesia for 2 days and then by fentanyl skin patch (25 mg/h) for 4–6 days. The nasogastric tube was removed, and esophagography was performed on postoperative day 7. Patients were discharged when they could tolerate a soft diet for 2–3 days without any significant gastrointestinal trouble.

Pulmonary complications were defined as cases of any pneumonic infiltrates shown on chest X-ray, acute lung injury, or adult respiratory distress syndrome (ARDS). At any sign of voice change or aspiration, RLN palsy was recorded, and this was evaluated regularly for 1 year.

Statistical analysis

Clinical and pathologic characteristics were described as mean ± standard deviation for continuous variables and frequency (%) for categorical variables. The differences between the two groups were tested using Student’s t test, the χ2 test, and Fisher’s exact test depending on the nature of the data. All p values were two-sided, and a p value lower than 0.05 was considered significant. All statistical procedures were performed using SPSS statistical software, version 20.0 (SPSS Inc., Chicago, IL, USA).

Results

Patient demographics

The demographic data of the patients are listed in Table 1. The two groups did not differ in terms of demographics. Of the 40 patients, 32 patients (80 %) had clinical stage 1 disease and 4 underwent neoadjuvant CCRT for a locally advanced cancer. Bilateral cervical lymphadenectomy was added for two patients with upper thoracic esophageal cancer.

Table 1 Patient demographics
Full size table

Operative outcomes

The operative outcomes are summarized in Table 2. All but one patient underwent a successful RATE. Thoracotomy conversion was performed for one patient due to bleeding from a superior segmental branch of the left inferior pulmonary vein. One patient with pulmonary emphysema died of ARDS after surgery subsequent to neoadjuvant CCRT. The total operation time and robot console time were prolonged in group 2, but the two groups did not differ in terms of blood loss, hospital length of stay, or complication profile.

Table 2 Postoperative outcomes
Full size table

Although the statistical difference was marginal, the incidence of RLN palsy was higher in group 2 (31.8 %), which seemed to be associated with more aggressive lymph node dissection along the RLNs. All palsies were left-sided according to a laryngoscopic examination, and none of the patients needed tracheostomy or feeding jejunostomy for this problem. All palsies resolved spontaneously within 1 year.

Number of harvested lymph nodes

The number of harvested lymph nodes was significantly greater in group 2 (Table 3). Although both groups had a number of abdominal nodes, group 2 had a greater number of mediastinal nodes (30.3 ± 7.9 vs 19.6 ± 8.2; p < 0.001). This difference resulted mainly from the increased number of dissected nodes along the RLN chains (13.5 ± 5.7 in group 2 vs 4.8 ± 3.6 in group 1; p < 0.001).

Table 3 Harvested lymph nodes and pathologic stage
Full size table

Pathologic staging

The postoperative pathologic stages of the patients are listed in Table 3. Although 33 patients (82.5 %) showed cN0 lesions in the preoperative workup, the pathologic reports from the specimens showed that only 19 patients (47.5 %) had pN0 disease. Of the 21 patients (52.5 %) who showed lymph node metastasis (N1 in 14, N2 in 5, and N3 in 2 patients), 7 (33.3 %) had metastatic nodes along one of the RLN chains. Two patients in group 1 and 3 patients in group 2 showed metastasis in the right RLN chain, whereas two patients in group 2 showed metastasis in the left RLN chain. Of the 19 patients with pT1b tumors, 12 (52.6 %) showed lymph node metastasis (right RLN chain in 2 patients, left RLN chain in 2 patients, other mediastinal nodes in 3 patients, and left gastric nodes in 5 patients).

Discussion

Since our 2010 report describing the feasibility and safety of RATE with extended ML [15], we have changed our policy so that RATE currently is to be performed with total ML because we believe that lymph node clearance along bilateral RLNs is crucial for both accurate staging and local control. We postulated that the advantages of the robotic system would enable surgeons to remove all lymph nodes around the RLNs with acceptable mortality and morbidity rates.

We dissected an average of 42.6 ± 14.1 lymph nodes: 9.6 ± 6.5 along the bilateral RLNs and 25.5 ± 9.6 from the mediastinum. With an increase in the number of dissected nodes, the number of nodes found to exhibit metastasis also increased [17, 18]. In our series, 33 patients had cN0 disease before surgery, but only 19 patients (47.5 %) had pN0 disease after pathologic examination. Among 21 patients with node metastasis, 7 (33.3 %) had metastatic nodes along one of the RLN chains.

A consensus conference of experts previously suggested that at least 15 nodes should be examined, [16] whereas other investigators have suggested that up to 30 nodes are necessary [19]. The total number of dissected nodes in our series fulfilled those requirements. Considering that the risk of metastasis in RLN chains is high [6–8] and that the N factor in the 7th tumor-node-metastasis (TNM) classification is determined by the number of metastatic nodes, dissection along the RLNs is critical for accurate staging.

Although a therapeutic benefit of dissecting more lymph nodes with potential metastatic disease has not been proven to date, surgeons should try to dissect more lymph nodes at the time of surgery because resection techniques linked to higher lymph node counts are associated with better survival [20, 21]. For maximal survival, Rizk et al. [19] recommended the dissection of at least 10 nodes for pT1, 20 nodes for pT2, and 30 or more nodes for pT3/T4. Schwarz and Smith [17] studied a cohort of 2,597 patients from the Surveillance, Epidemiology, and End Results database and projected that 5-year overall survival was improved by 10.7 % for every 10 extra lymph nodes examined in squamous cell carcinoma.

Our current approach to lymphadenectomy involves skeletonizing the RLNs so as to dissect all lymph nodes around the nerve. We believe that the advantages of RATE are of great value in lymph node dissection along the left RLN because it offers a three-dimensional view of the anatomy of the left tracheoesophageal groove and because a wrist-mimicking motion enables surgeons to remove all lymph nodes around the RLN in more comfortable way.

After skeletonization of the left RLN chains, we can always identify the left common carotid artery at the base of the dissection field (Fig. 1D). As a result, we retrieved 13.5 ± 5.7 nodes from the bilateral RLN chains, which was more than the 4.8 ± 3.6 nodes retrieved in group 1 (p < 0.001). Considering that the risk of node metastasis is high in this area, we believe that a skeletonization technique has the potential to provide better survival in addition to more accurate staging.

However, extensive nodal dissection around the trachea can interfere with lymphatic drainage and blood supply, which may contribute to the development of pneumonia. According to a systematic review, the incidence of pneumonia ranged from 4.7 to 49 % in open transthoracic esophagectomy (OTE) and from 5 to 36 % in MIE, with mortality rates ranging from 2 to 9.8 % in OTE and from 0 to 7.7 % in MIE [22].

A metaanalysis by Decker et al. [23] showed that the incidences of pneumonia and mortality in video-assisted thoracic surgery esophagectomy (VATS-E) were 16.7 and 2.4 %, respectively, whereas the number of dissected nodes was 16.5. Considering that our procedure included radical esophagectomy with extensive lymphadenectomy, the mortality rate (2.5 %) and the incidence of pulmonary complications (12.5 %) seem acceptable. Although we completely skeletonized the RLN chains in group 2, pulmonary complications did not increase significantly compared with group 1 (p = 0.064).

We believe that RATE has the potential to decrease the incidence of RLN palsy because high-definition, magnified images facilitate easier identification of the nerve, and tremor filtering minimizes the possibility of accidental injury to the nerve. However, we cannot definitely state that RATE decreases the incidence of RLN palsy because our study had no control arm (VATS-E or OTE).

Recently, Suda et al. [24] suggested a potential decrease in RLN palsy with RATE. They compared postoperative outcomes among 16 patients after RATE with those of 20 patients after VATS-E and reported that RATE significantly reduced the incidence of RLN palsy (p = 0.018) as well as ventilator time (p = 0.025).

In our series, eight patients (20 %) experienced RLN palsy, but no incidences of pneumonia or tracheostomy resulted from this problem. The affected patients had a transient voice change with or without minimal aspiration symptoms and still could be discharged with a soft diet. As expected, RLN palsy was more common in group 2 and in left RLNs because we skeletonized the left RLN from the infraaortic area to the thoracic inlet. Considering the risks and benefits, we think that a skeletonization technique is worthwhile. The benefits of its potential to offer accurate staging and better survival outweigh the risk of transient RLN palsy. Table 4 lists recent studies of VATS-E and RATE, all of which included a lymph node dissection along the RLNs. Our results are comparable with those of other Japanese series regarding the number of mediastinal lymph nodes and RLN nodes, as well as the operative mortality and morbidity rates. Thoracoscopy time and hospital length of stay were shorter according to all available data.

Table 4 Recent series of lymph node dissections along the RLNs by means of MIE
Full size table

In conclusion, robot-assisted thoracoscopic lymphadenectomy along bilateral RLNs was technically feasible and safe. Skeletonization of the RLNs yielded more lymph nodes, with acceptable mortality and morbidity. Nevertheless, the oncologic benefits of the procedure should be verified in future studies.