Will robot-assisted minimally invasive esophagectomy improve patient outcomes compared to conventional minimally invasive esophagectomy?

INTRODUCTION

Although preoperative chemotherapy or chemoradiotherapy (CRT) followed by esophagectomy with radical lymph node dissection has a major role in the standard treatment of resectable esophageal cancer, radical esophagectomy is a highly-invasive procedure with high morbidity and mortality rates^[1-3]. Therefore, the global demand for minimally invasive esophageal cancer surgery has resulted in the rapid spread of thoracoscopic esophagectomy [minimally invasive esophagectomy (MIE)] as a minimally invasive surgical treatment in recent years^[4]. Recently, robot-assisted MIE (RAMIE) has been introduced and rapidly standardized for esophageal cancer surgery at various institutions in Japan. RAMIE has the advantages of recognizing microanatomy based on 3D high-definition images, multi-joint function, and vibration filtering, thereby improving operability, especially for superior mediastinal dissection, and enabling precise surgical manipulation^[5-7]. However, whether the introduction of RAMIE improves patient outcomes compared to conventional MIE is controversial.

The aim of this study was to evaluate the surgical outcomes of RAMIE compared to standardized MIE and to determine if RAMIE improves short-term patient outcomes.

METHODS

RAMIE and MIE operative procedures

MIE and RAMIE preparation

The patients were intubated with a single-lumen spiral tube with bilateral lung ventilation in preparation for MIE and RAMIE. An artificial pneumothorax was established with 8 mmHg carbon dioxide gas using the Airseal® Intelligent Flow System (ConMed, Utica, NY, USA) to secure the visual field.

A pneumothorax was initiated during MIE by puncturing the Airseal® 12-mm port using the optical view technique at the extension of the 9th intercostal space (ICS) at the subscapular angle. This port was used as a camera port with a 3D high-definition flexible scope. Next, five ports were inserted as follows: a 12-mm port was inserted in the posterior axillary line of the 5th and 7th ICSs; a 5-mm port was inserted in the mid-axillary line of the 3rd and 8th ICSs; and a 5-mm port was inserted in the 6th ICS slightly midline of the subscapular angle [Figure 1A]. The energy devices used are mainly electrocautery or vessel sealing systems.

Figure 1. (A) Port placement of MIE. The Airseal® 12-mm port, used as the camera port, was inserted at the extension of the 9th intercostal space (ICS) at the subscapular angle. Next, five ports were inserted as follows: a 12-mm port was inserted in the posterior axillary line of the 5th and 7th ICSs; a 5-mm port was inserted in the mid-axillary line of the 3rd and 8th ICSs; and a 5-mm port was inserted in the 6th ICS slightly midline of the subscapular angle; (B) Port placement of RAMIE. The da Vinci Xi System ports for arms 1, 2, 3, and 4 were used for the 9th ICS at the subscapularis line, the 7th ICS behind the axillary line, the 5th ICS behind the axillary line, and the 3rd ICS at the mid-axillary line, respectively. In addition, an assistant port was placed at the 6th ICS slightly anterior to the mid-axillary line. An Airseal® 5-mm port was inserted on the dorsal aspect of the 8th ICS for use with the Airseal® Intelligent Flow System.

RAMIE was performed using a robotic system (da Vinci Xi System; Intuitive Surgical Inc., Sunnyvale, CA, USA). The da Vinci Xi System ports for arms 1, 2, 3, and 4 were used for the 9th ICS at the subscapularis line, the 7th ICS behind the axillary line, the 5th ICS behind the axillary line, and the 3rd ICS at the mid-axillary line, respectively. In addition, an assistant port was placed at the 6th ICS slightly anterior to the mid-axillary line. Because an artificial pneumothorax with an 8-mmHg carbon dioxide gas was used with the Airseal® Intelligent Flow System, we used an Airseal® 5-mm port on the dorsal aspect of the 8th ICS [Figure 1B]. During RAMIE, we primarily used a 30^o-angle thoracoscope. The primary instruments used by the surgeon during RAMIE were sharp monopolar cutting scissors, which were used to dissect, coagulate, and cut tissue during the procedure.

The surgical procedures for MIE and RAMIE were similar, although the port placement and the surgical devices used were different Supplementary Video 1A and B.

Surgical procedure

Middle and lower mediastinal dissection

The right pulmonary ligament was dissected, the pericardial surface was exposed, the pericardial side of the tracheal bifurcation lymph node was dissected, and the membranous part of the left main bronchus was identified. The lower pulmonary branch of the right vagal nerve was sacrificed; only the right upper pulmonary branch was preserved. During dissection of the tracheal bifurcation, it is important to be aware of the tracheal sheath [Figure 2A]. The azygos arch was also divided using a linear stapler. Next, a descending peri-aortic dissection was performed. At this time, we were aware of the ligament interpleural de Morosow^[8] [Figure 2B]. Patients with cT3 or more disease in our department were treated with combined resection of the thoracic duct, while patients with cT1-2 disease were treated with preservation of the thoracic duct. In cases involving thoracic duct preservation, the ligament interpleural de Morosow was preserved, but in cases involving combined resection of thoracic ducts, the ligament interpleural de Morosow was incised, and the thoracic duct and surrounding lymph nodes were excised [Figure 2B].

Figure 2. (A) Lymph node dissection of the tracheal bifurcation. The yellow arrow indicates the tracheal sheath; (B) Peri-aortic dissection of the descending aorta. Yellow arrows indicate the ligament interpleural de Morosow; (C) Lymph node dissection around the right recurrent laryngeal nerve. The right recurrent laryngeal nerve was identified at the recurrent part (yellow arrow); (D) Lymph node dissection around the right recurrent laryngeal nerve. The red arrow indicates the tracheoesophageal sheath; (E) Intraoperative view after lymph node dissection around the right recurrent laryngeal nerve; (F) Lymph node dissection around the left recurrent laryngeal nerve. The dissected tissues were temporarily gathered on the esophageal side along the inner surface of the tracheoesophageal sheath; (G) Lymph node dissection around the left recurrent laryngeal nerve. The ventrally-dropped lymphatic chain was explored to the pre-tracheal border, then clipped and dissected; (H) Intraoperative view after lymph node dissection around the left recurrent laryngeal nerve. AZ: Azygos arch; E: esophagus; Lt. RLN: left recurrent laryngeal nerve; Rt. RLN: right recurrent laryngeal nerve; TD: thoracic duct; Tr: trachea.

Dissection of the right laryngeal nerve lymph nodes

It is important to be aware of the tracheoesophageal sheath and right tracheoesophageal artery (TEA) during right recurrent laryngeal nerve (RLN) lymph node dissection^[9]. The mediastinal pleura was incised up to the right subclavian artery (SCA) while preserving the epineurium of the right vagal nerve. The level of the vagal nerve epineurium, the tracheoesophageal sheath, and the right RLN were identified at the recurrent part [Figure 2C]. The upper thoracic esophagus was mobilized from the trachea and vertebral side, and the dissected right RLN lymph nodes were detached from the tracheal sheath at the right wall of the trachea. Next, three to five esophageal branches of the right RLN were dissected, and then the right RLN was dropped ventrally, continuing the dissection of the tissue to the pre-tracheal region as far ventrally as possible [Figure 2D]. The right RLN lymph nodes were dissected as far as possible into the neck. In some cases, it was possible to identify the pulsation of the right inferior thyroid artery and the lower pole of the thyroid gland [Figure 2E].

Dissection of the left RLN nodes

It is also important to be aware of the microanatomy of the tracheoesophageal sheath and left TEA, as well as the left RLN lymph node dissection^[9] [Figure 2F]. After taping the upper thoracic esophagus, the trachea was expanded ventrally. From the left aspect of the trachea, the dissected tissues of the left RLN lymph nodes were temporarily gathered on the esophageal side [Figure 2F]. Next, dissecting forceps were inserted into the anterior surface of the left RLN along the epineurium, and the tracheoesophageal arteriovenous branch and the tracheal branch of the left RLN were dissected. Thereafter, separating the lymphatic chain from the left RLN on the ventral side allowed subsequent manipulation without any traction to the left RLN. The ventrally-dropped lymphatic chain was explored to the pre-tracheal border and then clipped and dissected [Figure 2G]. Next, the esophageal branch of the left RLN was cut from the caudal-to-cephalic side, and the recurrent nerve was freed ventrally, then the tracheoesophageal sheath was incised, the esophageal branch of the TEA from the dorsal tracheoesophageal sheath to the esophagus was dissected, and the remaining dissected tissue was gathered to the esophagus and removed with the esophagus [Figure 2H].

Abdominal procedure and reconstruction

Gastric mobilization and upper abdominal lymph node dissection were performed using a hand-assisted laparoscopic approach, with the exception of cases with massive metastases of abdominal lymph nodes or a history of a laparotomy. Bowel continuity was principally reconstructed using a gastric conduit via the retrosternal route. Cervical esophagogastric anastomosis was performed using the modified Collard technique^[10] or a hand-sewn anastomosis. Both MIE and RAMIE abdominal manipulations were performed in the same way.

RESULTS

MIE was introduced in 2012, and 535 cases were performed until August 2022. RAMIE was started in 2019, and 84 cases were performed by August 2022. Of the 535 MIE cases, 208 cases of MIE were performed from January 2019 to August 2022.

To analyze the number of cases required for RAMIE standardization, the console time was calculated using CUSUM analysis. The calculated CUSUM learning curve was identified graphically to consist of three phases: phase I (28 initial cases), phase II (14 mid-term cases), and phase III (42 final cases; Figure 3). The slope of the CUSUM for phase I was positive, indicating insufficient procedural proficiency. The slope for phase II was variable but generally plateaued, suggesting that the console surgeon had achieved the learning point. In contrast, the slope of CUSUM in phase III tended to decrease, indicating that the surgeon has acquired technical proficiency.

Figure 3. The CUSUM plot of console time. A plateau was reached after the 29th case, followed by a further decrease after 43 cases. Therefore, we defined the period up to the 28th case as phase I, the period from the 29th-42nd cases as phase II, and the period after the 43rd case as phase III.

DISCUSSION

In the present study, it was found that RAMIE required 29 cases to achieve surgical proficiency. Moreover, the console time tended to decrease further after 43 cases. In addition, although the incidence of RLN palsy increased in phase I immediately after the introduction of RAMIE, the incidence of RLN palsy decreased significantly after proficiency was acquired.

Several reports have focused on learning curves for robotic-assisted esophagectomy. Hernandez et al.^[13] reported that the learning curve for robotic Ivor-Lewis surgery required a minimum of 20 cases to become proficient with respect to operative time and incidence of postoperative complications. In addition, de la Fuente et al.^[14] observed a trend toward lower complication rates after the first 29 surgeries using the robotic Ivor-Lewis system. In contrast, Park et al.^[15] reported that serving as an assistant surgeon in 50 cases prior to performing robotic surgery did not require a learning curve with respect to operative time or the number of lymph nodes dissected, whereas the postoperative results were similar to previous studies involving robotic esophagectomy outcomes^[15]. Although we implemented RAMIE without any experience in robotic surgery, the learning curve and postoperative outcomes were similar to reports by other authors after 29 cases. Moreover, we found an improvement in the incidence of RLN palsy and the duration of hospital stay, while other outcome measures were similar to reports by other authors. We conclude that these results could have been obtained by extrapolating the standardized MIE technique to robotic surgery. Moreover, the console operative time was shortened after 43 cases in our department. One reason for the robotic surgery biphasic learning curve in our hands, which differs from previous reports^[13-15], is that the proportion of cT3 cases was increased in phase II and the number of cases with combined thoracic duct resection increased. In addition, the indication for salvage surgery was expanded to robot-assisted esophagectomy, which may have resulted in a longer console operative time.

Several single-center studies have reported that robotic surgery tends to increase thoracic operative time compared to MIE^[16,17]; however, a recent RAMIE study showed that robotic-assisted surgery significantly reduces the operative time^[7]. These conflicting results may reflect the fact that most of the previous single-center studies tended to show longer operative times for robot-assisted surgery because of learning curves in the early phases after introduction. Therefore, we believe that if RAMIE is standardized at each institution, the console operative time will improve, as we have reported in our department.

The efficacy of robotic-assisted surgery compared to MIE has not been fully validated. Several systematic reviews have shown reduced blood loss, an increased number of lymph nodes dissected, and shorter hospital stays, but most of the reports were single-center studies with a small number of patients; large multicenter studies are lacking^[18]. The results of ongoing multicenter prospective studies, such as the RAMIE^[15] and REVATE trials^[19], are warranted.

Our results showed that the incidence of RLN palsy was reduced in the cases in which standardized RAMIE was performed. While there have been recent reports that RAMIE reduces the incidence of RLN palsy compared to MIE^[7,16,20], several studies have demonstrated conflicting results^[21-24].

Therefore, whether RAMIE is superior to MIE in reducing the incidence of RLN palsy is debatable; however, because the microanatomy detail provided by the multi-joint capabilities and 3D high-definition images in robotic surgery contributes to improving the accuracy of superior mediastinal lymph node dissection, we are of the opinion that standardized robotic surgery may contribute to a reduction in the incidence of RLN palsy. The results of the ongoing REVATE study, in which the primary endpoints are successful lymph node harvesting around the left RLN and the incidence of left RLN palsy, are very interesting and eagerly awaited.

Because RLN palsy is a risk factor for postoperative pneumonia^[20,25,26], a reduction in the incidence of RLN palsy has the potential to decrease the risk of postoperative pneumonia and thereby improve the prognosis. Therefore, it is important to determine whether robotic-assisted surgery reduces the incidence of RLN palsy. Our results showed a trend toward a decrease in the incidence of RLN palsy with standardization of RAMIE but no decrease in the incidence of postoperative pneumonia. These results are similar to previous studies, and few studies have reported that RAMIE reduces the incidence of pneumonia compared to MIE^[7,16,20]. Because risk factors for postoperative pneumonia vary widely, including intraoperative position^[27], postoperative antibiotic management^[28], and the presence of sarcopenia^[29], it may be difficult to demonstrate a reduction in the incidence of pneumonia simply by a reduction in the incidence of RLN palsy alone.

The incidence of chylothorax tended to be higher in the RAMIE group, at 21%, compared to 13% in the MIE group (P = 0.081). The reason for the slightly higher incidence of chylothorax compared to other centers may be attributed to the use of early enteral feeding, which started four hours postoperatively, for postoperative management at our institution. In addition, the difference in surgical devices used between MIE and RAMIE may have also affected the incidence of chylothorax: in MIE, the vessel sealing system was used during periaortic lymph node dissection, whereas in RAMIE, only a monopolar electrocautery scalpel was initially used. This may have resulted in incomplete closure of lymphatic branches during the periaortic dissection in the RAMIE group. Therefore, the use of bipolar coagulation method during periaortic lymph node dissection resulted in a reduced incidence of chylothorax in the RAMIE group (MIE vs. phases II and III, P = 0.618). Moreover, all patients who developed chylothorax were healed conservatively by discontinuing enteral nutrition, with no significant impact on the postoperative course.

The incidence of RLN palsy was reduced in the RAMIE group, and the number of retrieved mediastinal lymph nodes and estimated intraoperative blood loss did not differ between groups. However, it is the surgeon, not the robot, who performs better lymph node dissection. We should not confuse the surgeon operating the robot with the robot^[30]. We believe that the fact that this study was performed by two surgeons experienced in both esophageal cancer surgery and robotic surgery contributed to these favorable results.

This study had several limitations. First, this was a retrospective study. Therefore, additional multicenter prospective studies are needed to fully clarify the superiority of RAMIE compared to MIE. Second, this study was ≤ 4 years in duration, and although the surgical technique and perioperative patient management were consistent, some of the surgical instruments differed. Third, although console operating time is one of the most important factors for achieving the learning curve, there are other important factors, such as the amount of blood loss, the number of retrieved mediastinal lymph nodes, and the time density of the surgical volume and operation frequency. This study did not consider all these factors in the learning curve. Regarding the amount of blood loss and the number of retrieved mediastinal lymph nodes, no difference was observed between the two groups in this study, and similar reports are available in the literature^[7,31]. However, we believe that one of the limitations of this study is the fact that the time density of surgical volume and operation frequency were not considered in the study.

In conclusion, RAMIE required 29 cases to achieve surgical proficiency. Our results suggest that the introduction and standardization of prone robot-assisted esophagectomy may decrease the incidence of RLN palsy in patients compared to MIE. Further analysis of more cases in future multicenter studies will clarify the potential impact of RAMIE in esophageal cancer.