Robotic-assisted Esophagectomy vs Video-Assisted Thoracoscopic Esophagectomy (REVATE): study protocol for a randomized controlled trial

A study flowchart is shown in Fig. 1. This is an investigator-initiated, investigator-driven multicenter, open-label, randomized controlled trial (termed REVATE) enrolling patients with ESCC scheduled to undergo LND along the RLN. Patients will be randomly assigned (1:1 ratio) to undergo minimally invasive esophagectomy (MIE) with either RE or VATE. All participating surgeons will be required to have competence with both techniques. The study will take place in two high-volume surgical centers located in Taiwan and China, and will be conducted with adherence to the principles of the World Medical Association’s Declaration of Helsinki.

Our study population will consist of adult patients with ESCC requiring MIE. Inclusion criteria will be as follows: age between 18 and 80 years; histology-proven primary intrathoracic ESCC; scheduled treatment with MIE (McKeown procedure) and bilateral RLN LND; European Clinical Oncology Group performance status 0–1; presence of surgically resectable disease (cT1−4a, N0–3, M0) defined according to the American Joint Committee on Cancer (AJCC) Staging Manual, eighth edition [21]; normal bilateral vocal cord function confirmed by preoperative laryngoscopic examination; and willingness to provide written informed consent. If the candidates meet any of the following criteria, they will not be eligible for the study: history of previous major thoracic surgery that renders MIE unfeasible; malignancies different from ESCC; and inability to undergo curative resection and/or to comply with the follow-up schedule.

Patients who do not require preoperative neoadjuvant chemoradiotherapy (nCRT) will be informed about potential inclusion in the first 3 weeks of the study before the scheduled operation. When nCRT will be necessary, candidates will be approached only upon completion of the nCRT course.

A computerized randomization tool will be used to randomly allocate patients to either RE or VATE (1:1 ratio). In order to minimize the potential confounding effect of local treatment variables (e.g., postoperative care), patients will be stratified in a 1:1 fashion according to the enrolling hospital. A permuted-block randomization with varying block size will be implemented. Allocation and block size will be concealed from all investigators. Blinding of surgeons is unfeasible because of the obvious technical differences between RE and VATE. However, the occurrence of the study endpoints will be determined from medical records by an independent assessor who will be blinded to the surgical technique used for each case. Patients will have the option to withdraw from the study at any time.

Regardless of the treatment arm (i.e., RE versus VATE), all patients will undergo a total two-field lymphadenectomy according to the consensus proposed in 1994 by the International Society of Disease of Esophagus (ISDE) [22]. The operating surgeon will be exempted from performing RLN LND in the presence of at least one of the following conditions: low surgical curability as evidenced by tumor ingrowth into adjacent organs (i.e., T4b status) or unexpected detection of distant metastases (i.e., M1 status) during surgery; and/or occurrence of previously unforeseen perioperative adverse events and/or findings which will require discontinuation of surgery (all of these events will be detailed on an individual basis). Other surgical variables—including the use of laparotomy versus laparoscopy, the technique used for anastomosis, and the route of reconstruction—are not expected to influence the results of RLN LND and will be left to the surgeon’s discretion.

The location of bilateral RLN lymph nodes will be defined according to the criteria proposed by the Japan Esophageal Society (JES) [23]. Right (station 106recR) and left (station 106recL) RLN lymph nodes correspond, respectively, to stations termed 2R and 2L in the AJCC Staging Manual, eighth edition (Table 1).

At inclusion, an unequivocal identification code will be generated for each participant, with its access being restricted to the principal investigator (PI) and the study coordinators. A good clinical practice (GCP)-compliant digital case record form (CRF) will be used for data collection, for which the study coordinators and/or research nurses will be in charge. Nonelectronic data will be secured in locked cabinets located at data coordinating centers, being accessible to the PI, as well as research nurses and physicians. Each participating center will be allowed to request information from the principal database, but direct access will be permitted only for locally generated data. Upon termination of the study, data access requests may be forwarded to the PI. The completed CRFs will be cross-checked with the original sources to ensure that data regarding the primary and secondary outcome measures will be accurate and reliable. All clinical records will be stored anonymously to protect privacy and all investigators will adhere to local confidentiality regulations.

The primary outcome measure will be the rate of unsuccessful LND along the left RLN, which will be defined as: failure to remove lymph nodes along the left RLN (i.e., no identifiable lymph nodes confirmed by pathology reports); or occurrence of permanent (duration > 6 months) left RLN palsy following LND. Regardless of the presence of hoarseness, vocal cord function will be assessed by an experienced otolaryngologist using a flexible laryngoscope within 1 week of surgery. RLN palsy will be classified according to the following variables: site (unilateral versus bilateral); duration (temporary (i.e., recovering within 6 months) versus permanent (i.e., not recovering within 6 months)); and type of treatment required (type I, no therapy required; type II, injury requiring an elective surgical procedure; type III, injury requiring an urgent surgical procedure) [24]. All patients diagnosed with postoperative RLN palsy will undergo additional laryngoscopic examinations at 1, 3, and 6 postoperative months to confirm the occurrence of permanent RLN palsy (defined as its persistence at 6 months after surgery).

Secondary endpoints will include: the number of nodes removed along the right and left RLN; the incidence of pneumonia, defined according to the Revised Uniform Pneumonia Score which includes temperature, leukocyte count, and pulmonary radiography (Table 2) [25]; the incidence of other postoperative complications (e.g., anastomotic leakage, chylothorax), defined according to the Esophagectomy Complications Consensus Group system; the total percentage of surgery-related major postoperative complications (i.e., grade IIIa or higher according to the Clavien–Dindo criteria) [26]; in-hospital, 30-day, and 90-day mortality rates, defined as any death occurring during the same hospitalization and within 30 or 90 days after surgery, respectively; R0 resection rates, defined as microscopically negative proximal/distal and circumferential margins; surgery-related parameters (including thoracic, abdominal, and total surgical time (expressed in minutes); unexpected events and complications occurring during surgery (e.g., massive hemorrhage, perforation of other organs); blood loss during surgery (expressed in milliliters per phase); and number of patients requiring conversion to thoracotomy and related reasons); postoperative recovery parameters (including length of mechanical ventilator use after surgery (expressed in minutes), length of ICU stay (expressed in hours), length of hospital stay (expressed in days), and need for reintubation or readmission to the ICU); quality of life and psychometric measures (including the EORTC QLQ-C30, OES-18, and Hospital Anxiety and Depression Scale (HADS) questionnaires, administered at the following time points: 5 days before surgery; 4 weeks, 3 months, and 6 months after surgery; and on a yearly basis up to 5 years thereafter); and oncological outcomes (including 2-year, 3-year, and 5-year disease-free and overall survival rates).

In order to avoid a learning curve bias, only surgeons with a proven track of at least 50 previously performed RE and VATE procedures will be involved in the trial. Every operation performed within the trial will be video-recorded. All of the events of interest (related to the study outcomes) and surgical videos will be regularly reviewed by a local independent committee consisting of surgeons with proven expertise in MIE and involved in the conduct of the trial. All disagreements will be handled by a senior member of the Upper GI International Robotic Association (http://​ugira.​org/​)—whose decision will be final.

Randomization will be considered the start of the study for each participant. All patients will be followed up according to the schedule shown in Fig. 2. The primary study endpoint will be assessed at 1 week (in the entire study cohort) and 6 months (in the subgroup of patients with confirmed postoperative left side RLN palsy at 1 postsurgical week). Data on postoperative complications, readmissions, and deaths occurring within 30 and 90 postoperative days will be collected. The study visits are planned at the following time points: 2 weeks before surgery; 4 weeks, 3 months, and 6 months after surgery; and every 6 months thereafter (until 5 postoperative years).

The study started on November 12, 2018, with a projected 3-year inclusion period. Analysis of short-term results and long-term oncological outcomes will be conducted at 6 months and 5 years after discharge of the last randomized patient, respectively. The expected total duration of the trial is 8 years, including prearrangement and data analysis. The investigators will release interim reports at the end of each preplanned follow-up period.

Based on a preliminary study conducted by our group, the projected rates of successful left RLN LND (according to the definition proposed in the current trial) are 70–75% and 85–90% for VATE and RE, respectively [20]. The purpose of our study is to demonstrate that the rate of successful left RLN LND will be at least 15% higher for RE compared with VATE (based on a conservative 75% rate for VATE). Because our null hypothesis is that the difference between the two arms is ≤15%, we will adopt one-sided testing. In a one-sided test, the “extreme” portion of distribution is decided beforehand as meaning “sufficiently large” (i.e., 15% in our case) [20]. Assuming an alpha error of 0.05 and an 80% power using one-sided two-sample comparisons of proportions under a balanced trial design, a total of 95 patients per treatment arm will be required. Assuming a dropout rate of 10% in the entire study cohort, we plan to enroll a total of 212 patients (i.e., 106 in each arm).

A formal statistical analysis plan will be prepared by both an independent statistician and the principle investigator ahead of the final data collection (both blinded to treatment allocation and the study results). The final study findings will be released by the study statistician. Data analysis will be performed in patients who will receive RLN LND according to the procedure assigned at randomization and who will have complete follow-up data on nerve palsy recovery status at the 6-month assessment (for confirmation of permanent nerve palsy). A small percentage of patients (estimated 5% maximum) are expected not to undergo the procedure assigned at randomization either because of self-withdraw from the study or because of occurrence of unexpected intraoperative findings (as described in the “Surgical approach” section). Another 5% of patients may potentially be lost to follow-up, resulting in a maximum expected total dropout rate of 10%. We will perform both an in-depth analysis of the reasons for dropout and a comparison of dropouts and completers. The primary outcome will be analyzed using a two-sample proportion test (single-sided). Categorical secondary outcomes will be expressed as frequencies and compared with the Fisher’s exact test. Continuous secondary outcomes will be given as means ± standard deviations (for normally distributed variables) or medians and interquartile ranges (for skewed parameters). Comparisons will be performed using Student’s t test and the nonparametric Mann–Whitney U test, respectively. The Cochran–Mantel–Haenszel test will be used to examine the differences in categorical outcomes between the two treatment groups after adjusting for study center and use of nCRT. All calculations will be conducted with SAS (version 9.3; SAS Institute Inc., Cary, NC, USA) and SPSS (version 20.0; SPSS Inc., Chicago, IL, USA) statistical packages. P < 0.05 will be considered statistically significant.