Introduction

Esophagectomy with lymphatic dissection is a highly invasive procedure with high mortality and morbidity, although it plays a major role in esophageal cancer treatment [1]. Thoracoscopic esophagectomy (TE) has become widespread as a minimally invasive surgical treatment. TE is superior to open esophagectomy (OE) regarding short-term outcomes and postoperative quality of life and has comparable survival benefit [2,3,4,5,6]. However, TE is generally not routinely applied as a standard approach for advanced esophageal cancer because of its high level of technical complexity. Recently, robot-assisted thoracoscopic esophagectomy (RATE) was introduced to overcome the technical difficulty of TE [7]. The robotic platform provides improved visualization with a magnified three-dimensional view, articulation of instruments with seven degrees of freedom, and tremor filtering, all of which enable precise manipulation [8, 9].

We have performed TE in the lateral decubitus position (LDP) in every case of resectable esophageal cancer to achieve adequate mediastinal dissection while maintaining safety in emergency thoracotomy. In our TE in the LDP, the surgeon uses an endoscopic view similar to the surgical view in OE [10]. We introduced RATE in the LDP with new refinements of the robotic system application to enhance our previous surgical experience obtained through performing more than 300 cases of TE in the LDP. In the present study, we report our methods used to optimize RATE in the LDP, and assess its usefulness by comparing the surgical results of RATE and TE.

Materials and methods

Patients

At Kanazawa University, all patients with resectable thoracic esophageal cancer underwent TE. We started to perform RATE in July 2018. RATE was initially limited to T1 tumors without lymph node metastasis. After the initial two cases of RATE, we enlarged the indication for RATE to include all resectable tumors. We performed 30 RATE procedures between July 2018 and April 2020 (initial RATE group). We changed the port arrangement in May 2020, and performed ten RATE procedures between May 2020 and August 2020 (recent RATE group). All RATE procedures were performed by one of the authors (I.N.). The same operator (I.N.) performed TE in 51 consecutive patients between April 2014 and May 2019. The patients underwent TE included more advanced stage tumor rather than initial RATE group. To minimize the cancer stage difference between the TE and initial RATE groups, TE cases were propensity score matched to initial RATE cases according to the clinical stage. We imposed a caliper of 0.30 of the standard deviation of the logit of the propensity score. Propensity score-matched analysis produced 30 patients underwent TE (TE group).

Surgical procedure

RATE in LDP was performed using the da Vinci Xi Surgical System (Intuitive Surgical Inc., Sunnyvale, CA) after intubation with a left-side double-lumen tube. Ports setting and operating view are presented in Figs. 1 and 2, respectively. Camera image for the operator was vertically and horizontally inverted by camera rotation to create an operative view similar to that achieved under OE. We used a forward-oblique viewing endoscope with a 30° down-facing orientation. To obtain appropriate hand–eye coordination we changed hand control assignment manually. In initial RATE, the right and left hands were assigned to arm 1 and arm 3, respectively. Arm 4 was used as an assistant. In recent RATE, the right and left hands were assigned to arm 1 and arm 4, arm 2 and arm 4, or arm 1 and arm 2, respectively, depending on the situation. The remaining arm was used as an assistant. From the third case onward, a valveless insufflation system (AirSeal; ConMed, Utica, NY) was used to achieve artificial pneumothorax with CO2 insufflation at a pressure of 8 mmHg. The operator used monopolar curved scissors, Vessel Sealer or Vessel Sealer Extend (Intuitive Surgical Inc.), a medium–large clip applier, and a large needle driver in the right hand, and Maryland bipolar forceps or a large bipolar grasper in the left hand. Organ retraction was done using Cadiere forceps held in the assistant arm. The assistant performed tracheal rotation using a narrow tracheal retractor and dried the surgical field using suction. The image for the assistants was inverted horizontally and vertically. TE was performed as described previously [11].

Fig. 1
figure 1

Positions of the patient, assistants, patient cart, vision cart, and the trocar sites in the a initial and b recent robot-assisted thoracoscopic esophagectomy. Roman numerals show the rib numbers. Encircled numerals show the port sites. Port no. 1–4 are the 8-mm da Vinci ports (Intuitive Surgical Inc.) connected to the corresponding robotic arms. Port no. 5 and 6 are the 12-mm assistant ports. No. 6 port is the AirSeal (ConMed) trocar

Full size image
Fig. 2
figure 2

Operative view by recent procedure in the a right upper mediastinum, b left upper mediastinum, c sub-aortic arch area, d sub-carinal area, e left pulmonary hilum area, and f lower mediastinum. The left and upper sides of the monitor image are cranial and ventral, respectively. Encircled numerals show the port number used for instruments insertion. Port no. 1–4 connected to the corresponding robotic arms. AA aortic arch, AV azygos vein, DA descending aorta, E esophagus, LMB left main bronchus, LPV left pulmonary vein, LRLN left recurrent laryngeal nerve, P pericardium, PA pulmonary artery, RMB right main bronchus, RPV right pulmonary vein, RRLN right recurrent laryngeal nerve, S suction, SCA subclavian artery, T trachea, TR trachea retractor, VN vagal nerve

Full size image

Data analysis

Tumors were staged in accordance with the TNM classification of the American Joint Committee on Cancer and the Union Internationale Contre le Cancer (8th edition). Robot setup time was defined as the time from the end of port insertion to the start of the console procedure. Surgical morbidities were evaluated using the Clavien–Dindo classification [12]. The Chi-squared test and Fisher’s exact test were used to compare categorical variables. The Mann–Whitney U test was applied to compare the continuous variables between the groups. A p value of < 0.05 was considered statistically significant. All analyses were done with IBM SPSS (IBM Statistics version 25; IBM, NY).

Results

The demographics and clinical characteristics of the 70 included patients are shown in Table 1. There were no significant differences between groups regarding patient demographics, except that the tumor location differed between the TE and initial RATE groups.

Table 1 Patient demographics
Full size table

The thoracic surgical outcomes are summarized in Table 2. There was no conversion to thoracotomy in all three groups. The duration of the thoracic procedure was significantly longer in the initial RATE group than in the TE group (p = 0.04) and recent RATE group (p = 0.011). The amount of thoracic blood loss was significantly less in the initial RATE group than in the TE group (p = 0.003). The console time was significantly shorter in the recent RATE group than in the initial RATE group (p = 0.011). Transition of the duration of thoracic procedure and console time in RATE was shown in the supplemental figure (Online Resource).

Table 2 Surgical outcomes of the thoracic procedure
Full size table

There was no surgical mortality in all groups. The incidence of surgical morbidity did not differ between the three groups (Table 3). There were no specific complications related to the use of robotic surgery.

Table 3 Surgical mortality and morbidities related to the thoracic procedure
Full size table

Discussion

Camera rotation and manual hand control assignment in RATE in the LDP reproduced the surgical view and manipulation achieved in OE and TE in the LDP. Our ingenuity aided in the introduction of RATE with maximum enhancement of our surgical experience and anatomical knowledge obtained through more than 300 cases of TE in the LDP. Although RATE necessitated more time than TE, optimization of the port arrangement minimized the console and operation times. As a result, we successfully and safely introduced RATE in the LDP.

The two patient positions used in TE and RATE are the LDP and the prone position (PP). The ideal patient position in TE is controversial [13]. A recent study suggested that the PP is superior to the LDP regarding pulmonary complications, blood loss, and mediastinal lymph node harvest [14]. Furthermore, LDP necessitates much assistant’s effort to create adequate surgical view rather than PP. However, the LDP has advantage to perform emergent thoracotomy. Lymphatic dissection around the left RLN is easily performed in TE in LDP, because surgeons can introduce dissection device from the dorsal and caudal directions [15].

Many studies have evaluated RATE in the LDP [16,17,18,19]; the videoscope images and hand control manipulation in these studies are the same as those in RATE in the PP, where the arms for the right and left hands are introduced into thoracic cavity from the cranial and caudal directions, respectively. We have performed TE in the LDP under the image rotated by 180°, where the left side was cranial and the upper side was ventral; this image is similar to the view under OE [20]. We successfully reproduced the same endoscopic view and hand manipulation in RATE in the LDP as in our TE via inversion of the image with camera rotation and manual hand control assignment. In our RATE, artificial pneumothorax with the AirSeal insufflation system (ConMed) was used to obtain a good view with minimizing assistant’s efforts to compress the lung. We could perform RATE by drawing on the experience and anatomical knowledge obtained through the performance of more than 300 TE procedures, leading to a safe and smooth introduction of robotic esophagectomy.

The present data showed that RATE resulted in less blood loss and a longer operative time than TE. Meticulous manipulation by articulated forceps under the three-dimensional view might have contributed to the control of unnecessary bleeding. The most time-consuming problem in RATE was robotic arm collision. However, robotic arm collisions could be minimized by the optimization of the port settings. The port setting was rearranged to enable every two-arm combination among arm 1, arm 2, and arm 4 to be co-axial to the camera (arm 3). Hand control with a 3:1 motion scale might have been another cause of the elongated operation time. Thus, we changed the motion scale to 2:1 in the recent RATE group, which successfully shorten the console time. The learning effect might also contribute the time shortening. Previous report showed that learning phase without proctoring necessitated 70 procedures [16]. The continuous data plot of our operation time did not show the apparent learning effect. Forty procedures in our institute may be still within learning phase.

The application of RATE for superior mediastinum lymph node metastasis is reportedly associated with increased mortality and morbidity [21]. We experienced frequent RLN palsy. Most palsy was mild and occurred in the left RLN. Although the paralysis was low-grade and transient, more careful dissection avoiding RLN retraction and heat injury is desirable to avoid RLN palsy [22, 23]. The main cause of RLN palsy may be nerve retraction via esophageal branch during the lymphatic dissection around left cervical-mediastinal border area. To avoid RLN retraction, we transect esophageal branch to expose upper esophageal wall before esophageal retraction in most recent operation.

In conclusion, the robotic platform is beneficial in achieving meticulous dissection and reducing blood loss during TE in the LDP. Our ingenuity and optimization of the port arrangement during RATE in the LDP contributed to the safe introduction of the procedure and minimization of the operative duration. Reproduction of the surgical view and manipulation achieved in OE by camera rotation and manual hand control assignment in RATE in LDP is also applicable to RATE in PP and is beneficial to perform upper mediastinal dissection in RATE regardless of the patient position. Based on the advantages of meticulous robotic manipulation, we consider that optimized RATE can take the place of TE.