Comparing open and minimally invasive surgical procedures for Comparing open and minimally invasive surgical procedures for oesophagectomy in the treatment of cancer: the ROMIO (Randomised Oesophagectomy: Minimally Invasive or Open) feasibility study and pilot trial

The French MIRO trial

The oesphagectoMIe pour cancer paR voie conventionnelle ou coeliO-assistée (MIRO) trial of patients with oesophageal cancer, excluding patients with type II and III tumours involving the gastro-oesophageal junction, compared open two-phase surgery (abdomen and right chest) with two-phase laparoscopically assisted oesophagectomy (LAO; minimal access for the abdomen and open right chest incision) [see http://clinicaltrials.gov/show/NCT00937456 (accessed 21 January 2016)]. The primary end point was 30-day morbidity (a composite of all complications from grade II to grade IV on the Clavien-Dindo system12) and the trial was powered to test the hypothesis that minimal-access surgery leads to a reduced rate of complications (45% vs. 25%) at 30 days. The MIRO trial recruited 207 patients from 12 centres. The results, to date reported only at the American Society of Clinical Oncology Gastrointestinal Cancers Symposium 2015,13 found that 67 (64.4%) patients in the open surgery group had major morbidity compared with 37 (35.9%) in the minimal-access group [odds ratio 0.31, 95% confidence interval (CI) 0.18 to 0.55; p = 0.0001]. There are, however, a number of methodological weaknesses. Although the randomisation sequence was computer generated, allocation concealment was unclear, using sealed envelopes. The outcome assessors were not blinded to the intervention type and methods to quality assure surgical procedures are not described. 14

The Dutch TIME trial

The Traditional Invasive versus Minimally invasive Esophagectomy (TIME) trial included patients with oesophageal cancer, excluding patients with type II and III tumours involving the gastro-oesophageal junction. 15 It compared open two- or three-phase oesophagectomy with totally MIO (both abdomen and chest access performed with minimal-access approaches in the prone position). Pulmonary complications, strictly defined and graded, were the primary measure of outcome. The criteria for surgeon involvement in this trial were evidence of prior completion of 10 minimally invasive procedures and production of one video showing surgical competence. This trial recruited 120 patients from seven surgical centres in four countries (the Netherlands, Spain, India and Italy). In total, 115 patients were recruited, with 16 out of 56 (29%) patients undergoing open surgery having a pulmonary infection in the first 2 weeks compared with five out of 59 (8%) in the minimally invasive surgery group (relative risk 0.30, 95% confidence interval 0.12 to 0.76; p = 0.005). The trial includes a comprehensive assessment of HRQoL with the Short Form questionnaire-36 items (SF-36)16 and the European Organization for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire Oesophageal Cancer Module (QLQ-OES18),17 but there are no cost analyses and no monitoring of surgical procedures.

Rationale for a UK trial

Although the two previously conducted trials (MIRO13 and TIME15) provide some evidence to inform practice, both have methodological flaws that preclude firm conclusions being drawn from their results and neither will be applicable to the NHS and UK surgeons. In particular, the sample sizes are small and only a modest contribution is made to the evidence of equivalent survival benefits with the different surgical approaches. The primary end points reflect surgical interest and do not incorporate meaningful benefit for minimal-access surgery from the patients’ perspective. Both trials are at risk of biased outcome assessment without blinding of assessors and the French trial used sealed envelopes for randomisation with the concealment of allocation consequently being questionable. 14 In addition, the interventions in the Dutch trial15 (totally minimally invasive surgery) are still being developed in the UK and, as this is an evolving procedure, few UK surgeons and anaesthetists are comfortable with oesophagectomy in the prone position.

Rationale for a feasibility study and pilot trial

There are many challenges to conducting high-quality randomised trials of non-pharmaceutical interventions; the particular hurdles that may affect a full trial of open oesophagectomy and MIO are (1) strong preferences held by surgeons about the two procedures, which may mean that it is difficult to recruit centres to the main trial, and (2) preferences of patients that will not permit randomisation. Addressing the first issue, we planned to secure provisional agreement from clinical centres to take part in the main trial. The recruitment of patients to the study was addressed by the qualitative research integrated within this feasibility study. The feasibility of the main trial would also be in doubt if (3) the number of eligible patients was lower than a mean of 2.5 patients per month per centre. Further potential challenges to the quality of the main trial are (4) an inability to specify the surgical procedures being evaluated and to demonstrate that the specified procedures are being adhered to; (5) bias in outcome measures because of patients and assessors not being blinded to treatment allocation; (6) the lack of a widely accepted battery of outcome measures on which to compare different surgical procedures; and (7) poor standards of pathology with consequently a weak measure of oncological success. These were all addressed by the feasibility study and an unexpected failure to find a practical solution to any of them would compromise the main trial.

Open surgery

The oesophagectomy was carried out in two or three phases according to the surgeons’ judgement with regard to the patient and the tumour.

• Abdominal phase. After initial inspection of the abdomen to exclude inoperability, complete gastric mobilisation was performed with the surgeons’ usual practice based on the right gastroepiploic and right gastric arteries. Pyloroplasty, pyloromyotomy or no drainage were at the surgeons’ discretion. Lymphadenectomies along the common hepatic artery and left gastric and splenic arteries either en bloc or separately were performed and removal of sufficient crural fibres and a cuff of diaphragm was performed if required for tumour clearance. The pericardial fat pad and strips of pleura were removed. Transection of the lesser curve was undertaken during the abdominal phase or left to the thoracic phase of the operation. Placement of a feeding jejunostomy or nasojejunal tube was optional at the surgeons’ discretion as was placement of intra-abdominal and intra-thoracic drains. Methods to close the abdomen were at the surgeons’ discretion.

• Thoracic phase. After initial inspection of the chest to exclude inoperability, the mediastinal pleura overlying the oesophagus was excised in continuity with the oesophagus. The posterior limit of the dissection was normally the anterolateral wall of the aorta, so that the thoracic duct was mobilised en bloc or separately to the oesophagus and peri-oesophageal tissues. The thoracic duct was ligated and divided at the level of the diaphragm. The oesophagus was mobilised to the level of at least the aortic arch. Para-oesophageal and diaphragmatic nodes were removed in continuity or separately with the oesophagus (at the surgeons’ discretion). Lymph nodes at the tracheal bifurcation and along the right and left main bronchi to the pulmonary hilus were removed en bloc or separately at the surgeons’ discretion. The anastomotic technique and use of chest drainage was at the surgeons’ discretion. Methods to close the chest were at the surgeons’ discretion.

• Cervical phase. A left neck incision was made if a three-phase operation was undertaken. The oesophagus was mobilised preserving the recurrent laryngeal nerve and the anastomosis was performed using the surgeons’ preferred methods. Use of a drain was optional.

Laparoscopically assisted oesophagectomy

The oesophagectomy was performed as described in the previous section and access to the abdominal cavity was achieved using between four and six small incisions, which were placed any way in the abdominal wall at the surgeons’ discretion. Placement of a feeding jejunostomy was at the surgeons’ discretion and was performed laparoscopically or by extending a port site to an 8-cm abdominal incision. The thoracic part of the operation was performed as described in the previous section with a standard open incision. A cervical incision and a neck anastomosis was normally undertaken if required for tumour clearance.

Totally minimally invasive oesophagectomy

This consisted of performing the steps of the abdominal and chest phases of the operation as described in Open surgery but using laparoscopic and thoracoscopic techniques for access to each body cavity for each phase, respectively. A two- or three-phase minimally invasive operation was permissable. In the two-phase procedure the anastomosis was performed in the chest if necessary creating a 15-cm incision in addition to the minimal-access ports. In the three-phase operation the anastomosis was performed with a left cervical incision.

Setting

Two centres, University Hospitals Bristol NHS Foundation Trust and Plymouth Hospitals NHS Trust, recruited patients and carried out procedures within the pilot trial. Both centres have teams of upper gastrointestinal cancer surgeons (six in Bristol and five in Plymouth). These two centres also undertook the associated feasibility work. Methodological support for the pilot trial was predominantly based in Bristol, with the development of quality assurance protocols for surgical procedures and pathology being based at Imperial College London.

Allocation to treatment arm

Allocation of patients to surgical procedure was at random, the allocation being conducted separately for the two centres and further stratified by whether or not patients had undergone neoadjuvant treatment. Patients were randomly allocated to one of the three procedures at Bristol and to one of the open or two-phase laparoscopically assisted procedures at Plymouth. Randomisation within blocks of varying size prevented large imbalances in the number of patients in each treatment arm. Allocation was concealed through centralised randomisation, with patients being logged into the study before their allocation was revealed to the surgical team.

Sample size

At the two lead centres, recruitment to the pilot RCT was planned for a 12-month period, with 72 potentially eligible patients being expected during that time. This would allow a true 50% recruitment rate to be estimated with a 95% CI of approximately 38% to 62% (CI calculation based on the binomial distribution – the Clopper–Pearson approach19). If 11 patients were randomly allocated to each surgical procedure, this would allow a true difference of 1.25 standard deviations between two procedures on a continuous measure of early outcome to be detected with 80% power at the 5% significance level. Hence, the pilot RCT was planned to provide an acceptably precise estimate of the recruitment rate to inform plans for the main trial and perhaps to provide evidence suggestive of an intervention having promise for a beneficial impact on short-term outcomes.

Statistical methods

Summary statistics that inform plans for the main trial are presented, including the number of potentially eligible patients per month per centre, the percentage of these patients confirmed as eligible, the percentage of patients agreeing to be randomly allocated to a study procedure in the pilot trial and the percentage of randomised patients completing outcome measurements. Mean scores on short-term outcome measures are presented for each study arm, with p-values and 95% CIs presented for treatment comparisons when at least 10 patients have been randomised to each study arm. Additional summary statistics arise from the feasibility work, for example mean scores on the blinding scale achieved by different blinding procedures.

Planned follow-up

Participating patients completed baseline measurements prior to random allocation. On the second day post surgery, patients completed assessments of pain and blinding; this was followed by completion of the full assessment 6 days, 6 weeks and 3 and 6 months after surgery. During the first year of the feasibility study, permission was obtained to continue follow-up for 3 years post randomisation. Patient-reported outcomes were subsequently administered at 9, 12, 18, 24 and 36 months post randomisation (Table 1).

Primary outcome measure

The primary outcome measure for the main trial was to be determined during the feasibility study based on expert input from the Trial Steering Committee and the consensus meetings held to inform the core outcome set described in Development of a core outcome set to assess outcome for oesophageal cancer surgery. At the outset of the feasibility study, the primary outcome for the main trial was planned to be a patient report of fatigue using the Multidimensional Fatigue Inventory (MFI-20). 22 This validated questionnaire has been used in trials of minimal-access surgery for nephrectomy and it is the primary outcome of an open trial of minimal-access compared with open surgery within an enhanced recovery programme [the Enhanced Recovery Open versus Laparoscopic (EnROL) trial]. Consideration was given to using a dual primary end point including a self-reported measure of fatigue and an assessment of morbidity.

Secondary outcome measures

The secondary outcome measures for the main trial were to be determined during the feasibility study. During the pilot trial we measured surgical morbidity using the Accordian [see www.accordionclassification.wustl.edu/ (accessed 21 January 2016)] and Clavien–Dindo23 classifications, which include assessment of in-hospital mortality and need for reoperation. Survival time was recorded for any individual dying during the follow-up period, as was the time until the onset of palliative care/diagnosis of recurrent disease. Procedural outcome measures included the lymph node count, duration of operation and blood loss. Generic and disease-specific measures of patient-reported health were taken, including the EORTC Quality of Life Questionnaire Core 30 (QLQ-C30)21 and QLQ-OES1817 plus the European Quality of Life-5 Dimensions five-level version (EQ-5D-5L). 24 Length of hospital stay, defined as time from day of operation to discharge home, was noted for each trial participant. Further measures of resource use were considered in the health economics feasibility work.

Further measures

Presurgical pathological staging used evidence from a multidisciplinary team review of imaging (computerised tomography, positron emission tomography, endoscopic ultrasound and laparoscopy) and was recorded according to the tumour, node, metastasis (TNM) system (Table 2). Instances in which the primary tumour (T×), nodes (N×) or metastases (M×) could not be assessed, in which there was no evidence of a primary tumour (T0) and in which carcinoma in situ (Tis) was diagnosed were also recorded using the indicated additional codes. The TNM categories were combined into numerical stages for tabulation, these being stage 1A (cancer in the lining of the oesophagus only), stage 1B (cancer beginning to spread to the muscle wall), stage 2A (cancer has grown through the wall of the oesophagus), stage 2B (cancer has spread to one to two lymph nodes) and stages 3A, 3B or 3C (cancer is beginning to spread beyond the oesophagus into surrounding tissues, lymph nodes and organs).

Understanding recruitment issues

The first aspect was addressed through in-depth interviews, undertaken by an experienced qualitative researcher, with a maximum variation sample of between 10 and 15 patients eligible for the trial (selected to include the range of age, disease severity, socioeconomic status and centre). These interviews explored patient perspectives of oesophageal cancer and its treatment, views about surgery and study information and the acceptability of randomisation between procedures, including potential barriers to randomisation. In-depth interviews with clinical investigators and staff undertaking recruitment to the trial explored their ability to summarise the details of the trial design and protocol, their views about the evidence on which it is based, their perceptions of levels of uncertainty/equipoise in relation to the trial arms, their views about how the arms are or can be delivered in their clinical centre, methods for identifying eligible patients and examples of actual recruitment successes and difficulties. Numbers of interviews were guided by the concept of data saturation: the need to continue sampling until findings become repetitious. Interview data were analysed thematically and using case study approaches, looking for shared or disparate views among investigators, clinicians, recruiters and potential participants and within or between centres. 32,33

Digital audio recorders were provided to the recruitment staff so that they could audio record appointments at which they provided information for patients and attempted to recruit them to the ROMIO trial. These recordings were analysed thematically and also using a version of conversation analysis pioneered in previous studies33–35 to identify aspects of information provision that were unclear, disrupted and hinder recruitment and that could thus be targeted for improvement.

The qualitative researcher worked with the two study nurses to map the patient eligibility and recruitment pathways that are established, including when patients receive information about the trial and which members of the clinical team they meet and when. Logs of eligible and recruited patients were assembled using simple flow charts and counts to display numbers and percentages of patients at each stage of the eligibility and recruitment process and allow comparison with the original protocol. Previous research29–31 has shown that logistical and other local issues sometimes prevent full implementation of the protocol and can consequently lead to less efficient recruitment pathways.

Development of a recruitment strategy

The second aspect was the development of a recruitment strategy. The findings from these qualitative investigations and their potential impact on recruitment were summarised and presented to the ROMIO investigators. When necessary the recruitment process was modified in light of the specific qualitative research findings, the changes including a mix of bespoke and generic measures. Actions specific to the ROMIO trial were grounded in the findings of the integrated qualitative research and could include, for example, changes to the descriptions of surgical procedures in study information. Training of recruitment staff is an example of a generic approach that has been found to be useful in previous trials,29,32 with topics including how to elicit and address patient preferences and how to manage personal (clinical) preferences.

Once the recruitment plan was agreed it was implemented during the latter part of the feasibility study recruitment period. Recruiters were strongly encouraged to continue recording their appointments with potential participants so that these could be analysed iteratively to check that the new process was being implemented and to identify any emergent difficulties. Individual feedback to recruiters was provided confidentially by the qualitative researcher and anonymised findings from the recordings were used in a recruitment training programme for the main trial. Logs of eligible patients and randomisation rates continued to be measured.

Semistructured interviews and structured observations

Semistructured interviews and structured observations,36 performed with peer-identified expert surgeons at specialist centres for oesophagectomy in the UK, the USA and Japan, provided experience of variations in clinical practice worldwide. Interviews were digitally audio recorded (with the participants’ verbal consent) before being transcribed, checked for accuracy and qualitatively analysed using thematic analysis. 37 By means of contingency, shorthand written records were also kept at the time of interview in case of digital data loss. Structured observations were written in a research diary kept by the researcher. A second researcher was present for some of the procedures in the UK and all of those in Japan. A debrief was held at the end of each observed operation, which permitted comparison of notes between researchers. A second researcher was not present in the USA, although digital video recordings were obtained, with the appropriate approvals in place, for analysis in the UK.

Hierarchical task analysis for oesophagectomy

Findings from the published literature and online digital media were combined with thematic analysis of the semistructured interviews and structured observations to create a hierarchical task analysis for two-field oesophagectomy. 38

Several iterations of the hierarchical task analysis were written and revised. The accuracy of each hierarchical task analysis was tested against a real-time or recorded oesophagectomy until no additional changes were identified. A final version of the hierarchical task analysis was then constructed and tested against a series of subsequent operations, for which the researcher was present as an observer.

Delphi consensus process for the oesophagectomy hierarchical task analysis

Ten peer-identified expert surgeons, each with significant personal experience of performing oesophagectomy, were electronically invited to participate in a Delphi consensus process. 39 This method was selected to ensure the underlying face and content validity of the final assessment tool as well as surgeon acceptance of the defined procedure. Expert surgeons who had been involved during the development of the hierarchical task analysis were excluded. The 10 invited surgeons were Mr Barham (ROMIO), Mr Berrisford (ROMIO) and Mr Hardwick from the UK; Professor Van Lanschot, Professor Hoelscher, Professor Pera and Professor Zaninotto from Europe; Dr Donald Low from North America; and Professor Simon Law from Hong Kong.

In the first Delphi round, each surgeon was provided with the final hierarchical task analysis and a questionnaire on which they were requested to rate the necessity of each step identified in the analysis as mandatory, optional or prohibited, rather than using the traditional Likert-scale rating system. This was in line with ROMIO trial requirements. Additional free space was available for comments and suggestions on each step of the hierarchical task analysis as well as in general. Completed questionnaires were returned to Professor Hanna electronically and analysed by the clinical research fellow. An arbitrary consensus agreement of 75% was sought for each step.

This process was repeated in the second Delphi round, during which the nine respondents from the first Delphi round were electronically sent the percentage agreement and anonymised comments for each step of the hierarchical task analysis (rather than the traditional mean and standard deviation). The original analysis was resent, along with a fresh Delphi questionnaire. If respondents’ answers remained outside of the majority agreement, they were asked to provide reasons for this in the comments section. For steps in which consensus agreement could not be reached by the end of the second Delphi round, the researcher consulted Professor Hanna at Imperial College London. It was decided that a majority opinion could be upheld if it reflected the findings of the evidence-based hierarchical task analysis.

Development of an operation manual for oesophagectomy

A full operation manual was constructed for circulation to surgeons entering the ROMIO trial, based on the Delphi consensus approved hierarchical task analysis and its original evidence base (i.e. the published literature, semistructured interviews and structured observations). The operative steps for both open oesophagectomy and MIO were identical. As such, each step of the operation, both mandatory and optional, was described in detail and diagrams illustrated the required en-bloc lymphadenectomy.

Development of an operation note for oesophagectomy

An operation note was constructed by way of an iterative process. It was designed to satisfy both the clinical requirement of providing a legal record of the operation performed and the research requirement of quality assurance. Its structure was based on the standard operation note currently in use at St Mary’s Hospital, London, and combined its generic content (namely patient identifiers, operating surgeon and assistants, procedure performed, operation date, incision, findings, procedure, closure, and postoperative instructions) with the evidence-based operation manual specific to the ROMIO trial.

The body of the operation note included a tick-box version of the operation manual, permitting surgeons to rapidly provide a detailed outline of the procedure performed. White-space boxes were incorporated for patient identifiers; operative findings; specific details of a task, for example the anastomosis; deviations from the trial protocol; errors and their recovery; and postoperative instructions. Two upper gastrointestinal cancer surgeons at St Mary’s Hospital, London, piloted the operation note over the course of 3 months and provided informal verbal constructive feedback to the primary researcher. Surgeons at the two centres involved in the ROMIO feasibility study also piloted the operation note.

Development and examination of the reliability of a video assessment tool for oesophagectomy

Results obtained from the semistructured interviews and structured observations performed previously confirmed the importance of the technical performance (i.e. process) and oncological quality (i.e. immediate outcome) of the operation, in accordance with the Systems Engineering Initiative for Patient Safety (SEIPS) model using a structure–process–outcome approach. 40 Various techniques were considered that would permit independent remote blind evaluation of (1) the process (safety and efficiency) and (2) the quality of the end product. It was decided that only a video assessment tool could address all of these aspects.

Elements relevant to the safety and efficiency of the operative process, as well as the oncological quality of the end product, were identified from thematic analysis of the semistructured interviews and structured observations such that clear definitions for each of the terms used were composed. An existing validated consultant-level surgical assessment tool41 was deconstructed and its underlying principles adapted during the structural development of this video assessment tool.

Several different video assessment tools were written and piloted at St Mary’s Hospital, London, over the course of 3 months. Each iteration placed a different emphasis on rating the element being assessed, with the intention of balancing the technical safety with which each task was performed with the surgeons’ efficiency and the oncological quality of their dissection. Eight videos, randomly selected from the two centres involved in the study, were rated by two independent blind assessors (senior consultant surgeon with 20 years’ experience in upper gastrointestinal surgery and a post Certificate of Completion of Specialist Training clinical fellow at consultant grade) to determine inter-rater reliability, which was calculated using Cohen’s kappa.

Development of a photographic assessment tool for oesophagectomy

While in Japan, researchers from Imperial College London had witnessed how photographs taken intraoperatively could be used for quality assurance purposes. Therefore, the outcome component of the final video assessment tool was isolated for use as a photographic assessment tool, focusing on the completeness of the lymphadenectomy and exposure of the relevant anatomical landmarks.

Recruitment

When planning the feasibility study we stated that:

We anticipate a smooth progression to the main trial if a randomisation rate of at least 50% of eligible patients can be achieved in the pilot trial, . . . , and if at least a mean of 2.5 eligible patients per month can be identified in each pilot trial centre.

The feasibility work has shown that it is possible to recruit to a trial in which patients are randomly allocated to the different approaches to oesophagectomy. Figure 1 shows the accumulation of participants in the ROMIO randomised trial until the end of December 2014. In the 21 months of recruitment 263 patients have been assessed for eligibility, of whom 135 (51%) were found to be eligible and 104 (77%) were randomised across the two centres. These figures exceeded our targets for eligible patients (2.5 patients × two centres × 21 months = 105 eligible patients) and randomised patients (105 eligible patients × 50% = 52.5 patients randomised) by the end of December 2014. Of the 104 randomised patients, 46 were at Plymouth (open surgery, n = 22; LAO, n = 24) and 58 were at Bristol (open surgery, n = 19; LAO, n = 20; MIO, n = 19).

FIGURE 1.

Actual and target recruitment up to December 2014. Recruitment to the study is continuing.

In the light of the excellent rate of recruitment to the ROMIO pilot trial, we continued beyond the planned closure date of 31 March 2014 and successfully applied for a contract variation to allow recruitment until the start of the main trial (planned start of the main trial is mid-2016). As recruitment is ongoing the database is still live and so, for the purposes of the remainder of this report, we focus on those patients randomised on or before 31 August 2014. By this date there were 48 participants at Bristol and 33 participants at Plymouth. Furthermore, as we now plan to include the pilot trial data in the main trial analysis we no longer plan an unblinded analysis during the feasibility stage and so the only post-randomisation variables that we present by study arm are actual surgery undergone, the assessment of blinding during the first week post surgery and follow-up completion rates.

Balance across treatment arms

Table 4 presents summary statistics for the baseline measures of several important variables by centre and trial arm. Approximately equal numbers of patients were allocated to each of the study arms available within each centre. As expected, by far the majority of participants were male and the mean age at randomisation was approximately 66 years for each treatment group, with a standard deviation of approximately 7 years. A lower proportion of patients had undergone neoadjuvant treatment before surgery in Plymouth than in Bristol, although the proportion in each treatment arm was well balanced within each centre.

The majority of participants had a history of smoking, with perhaps a slightly lower prevalence in Plymouth. At both centres smoking was well balanced across the study arms. About one-third of men had consumed alcohol on ≥ 3 days per week over the 12 months preceding diagnosis at both centres, with this variable being well balanced between the treatment arms.

For the feasibility study, baseline spirometry measurements were taken at Plymouth only (the Bristol group has recently bought a bedside spirometer for use in the study and intend to carry out these measurements in the planned definitive trial). The mean forced expiratory volume in 1 second (FEV₁) and forced vital capacity (FVC) were markedly lower than expected for men of this average age44 and both measures were balanced across the two treatment arms at Plymouth.

Table 5 presents the TNM categories and numerical disease staging by intervention arm for each centre. Most participants had a tumour that was growing through the wall of the oesophagus but involving one to two lymph nodes at most (stages 2a and 2b) or a cancer that had just started to spread to the tissues surrounding the oesophagus (stage 3a). There was no convincing evidence of an imbalance between the randomised intervention arms in terms of disease staging. Summarising the three numerical stages, it is apparent that the vast majority of patients had a stage-2 or stage-3 tumour.

Adherence to allocated treatment

Tables 6 and 7 present the surgical approach actually undertaken for patients in each arm of the study for the Bristol and Plymouth centres, respectively. For patients recruited up to 31 August 2014 at the Bristol centre, the allocated surgical approach was followed for 79% (38/48) of patients.

The three patients for whom the allocation to open surgery could not be followed were found to be inoperable. Of the patients allocated to LAO, one was converted to open surgery to stop a bleed, one requested MIO having spoken to a patient who had undergone this procedure previously and one underwent a gastrectomy for more extensive disease. Four patients allocated to MIO underwent LAO, two because of the position of the tumour and two because a surgeon was not available to undertake the MIO procedure.

For patients recruited up to 31 August 2014 at the Plymouth centre, the allocated approach to surgery was followed in every case (see Table 7). Note that one man at the Plymouth Centre withdrew from the study the day after being allocated to LAO, owing to being very deaf and becoming confused concerning the trial procedures.

Postsurgical measures

Table 8 presents summary statistics for a number of postsurgical measures for 80 patients recruited on or before 31 August 2014. The median length of postsurgical stay across the two centres was 10 days, with stays typically lasting between 8 and 16 days. Two deaths occurred within 30 days of surgery, with a further death occurring on the 31st postsurgical day. Three patients experienced anastomotic leaks, all of which resulted in a return to theatre. A further nine patients returned to theatre one or more times. Eight patients returned to the intensive care unit, six because of respiratory problems and two because of renal failure. Nine patients were readmitted within 30 days of surgery.

Completeness of follow-up

Table 9 enumerates the numbers of patients who should have been invited to complete the visual analogue scale of pain at days 2 and 6 post surgery, the numbers actually completing the scale and the reasons for non-completion.

Completion of the pain scale at day 2 was low (40/80, 50%), the reasons for non-completion being illness or insufficient recovery (9/80, 11%), lack of cover to administer the scale at the weekend (16/80, 20%) and administrative errors (9/80, 11%), which included erroneously skipping the pain scale if the patient was unblinded. Completion of the pain scale was considerably better at day 6 (65/80, 81%).

Table 10 indicates the completion of the QLQ-C30 at baseline, 6 days post surgery and 6 weeks and 6 months post randomisation.

The physical function subscale of the QLQ-C30 assessed at 6 weeks will be the primary outcome measure for the planned definitive trial. This assessment was available for 75% (60/80) of participants. Nine of these participants were missing a baseline assessment of this measure, likely because of the short interval between assessment appointment and surgery at one centre. Those participants not having an assessment of the QLQ-C30 were equally split between those who did provide later assessments and those who provided no subsequent information. The nine patients in the latter group included two patients who had been found to be inoperable during surgery.

Semistructured interviews and structured observations

In total, eight separate semistructured interviews were performed with six surgeons from the UK and two from the USA (Table 11). Themes arising from the qualitative analysis of these interviews are documented in Table 12. In addition, > 50 operations, performed by 16 surgeons from the UK (n = 9), the USA (n = 6) and Japan (n = 1), were observed in seven different hospitals. Structured observation notes were combined with findings from the ‘operative procedure’ theme identified from the semistructured interviews and incorporated into the hierarchical task analysis.

Hierarchical task analysis

The result of the hierarchical task analysis is presented in Appendix 1.

Operation manual and essential tasks

Appendix 2 presents the operation manual detailing mandatory (essential) and optional tasks. Given the length of the full manual, a separate summary document describing 10 essential steps for each of the abdominal and thoracic phases of the operation was also produced and is presented in Appendix 3.

Operation note

The operation note is presented in Appendix 4. This note allows recording of performance of or variation from steps in the manual, errors incurred and recovery mechanisms. This will allow root cause analysis of adverse events.

Video and photographic assessment tools

To allow a standardised assessment of video and photographic recording of surgical procedures, assessment tools were produced and are presented in Appendices 5 and 6, respectively.

Implementation in the feasibility study

The manuals, notes and assessment tools for quality assurance were approved by the ROMIO Trial Steering Committee on 15 November 2013. The approved documents were piloted during the ROMIO feasibility phase. An online encrypted file transfer process was set up by the team at Imperial College London. The Imperial College London Clinical Governance Officer arranged local access to the server and a nominee from each site, for example a nurse specialist, was identified to upload the files. The system was tested in both centres and all surgeons submitted video records from their surgery. Twenty-five operative videos were transferred using the online system. Blind independent assessors evaluated the operative performance using the video assessment tool, indicating that an objective video-based assessment system is feasible and can be applied in the main trial. Sets of intraoperative photographs of the operative field were obtained by the surgeons and transferred using the same online system. Surgeons used the operation notes to record the steps of the procedures.

Initial inpatient stay incorporating the oesophagectomy

Initial meetings with finance personnel at one of the sites and subsequent approaches to finance departments within all of the sites established how resource use was costed within the different hospitals. This will enable resource use information to be collected in the correct way in terms of measurement (e.g. different types of medical staff contacts can be measured using the number of visits to the patient or the actual time spent visiting the patient), which will then ensure that costs from the finance departments can be applied.

Follow-up inpatient stays and outpatient visits

The follow-up Clinical Report Form (CRF) was designed to enable HRGs to be created. The level of detail needed for this to be carried out correctly can be time-consuming. Different approaches to obtaining the HRGs for any subsequent inpatient stays have been explored.

Coders within the hospitals were initially approached as they allocate codes to diagnoses, procedures, complications and comorbidities. These codes are then submitted into a HRG ‘grouper’, a software package available for each financial year from the Health and Social Care Information Centre (www.hscic.gov.uk), which produces the HRG code for an individual’s inpatient stay. Time constraints may, however, limit their ability to obtain HRG codes for each individual within the trial.

Clinical commissioning groups were approached as they use individual patient HRGs to reimburse hospitals. However, the groups obtain patient HRGs only in a pseudo-anonymised format.

Finally, the use of individual patient records as the source to obtain data on resource use was explored. Finance and costing departments for Bristol, Plymouth and Bath were contacted. Bristol and Plymouth have so far confirmed that individual patient records are available for ROMIO patients. If individual patient records are available for all other sites in the main ROMIO trial then this would be the preferred option. However, although information on operating times would be available through the individual patient records, it is unlikely that differences in equipment use for the main surgical intervention would be accurately captured as needed. This emphasises the need to use microcosting techniques to cost the initial operation, even if individual patient records are used for the remaining aspects of the initial inpatient stay.

Resource use logs (diaries)

These were designed to be used by the patient as an ‘aide memoire’ to prospectively record health and social service contacts and other expenses incurred, with the format of the logs mirroring that of the questionnaires. Resource use logs have the potential to reduce missing data46 and therefore lead to more accurate estimates of costs. To allow piloting of the resource use diaries and the telephone-administered questionnaires, a substantial amendment to the ethical approval was sought and this was granted in April 2014. Patients were given the 0–3 months resource use diary by a research nurse at hospital discharge following their oesophagectomy. If they were too unwell to record events a relative or friend could record this information for them. At 3 months following their discharge, a nurse telephoned those patients given the resource use diary to ask them questions about their resource use during that period of time. To not overburden the patients, nurses were encouraged to call back or speak with a relative who could answer the questions on a patient’s behalf if the patient did not feel able to answer the questions at that time.

To assess the feasibility of the 6-month telephone administration of the resource use questionnaire the following process was established. Participants who had been discharged prior to ethical approval being granted but who had not yet reached 3 months post discharge were posted the 4–6 months resource use diary at 3 months post discharge and asked to complete it for the following 3 months. At 6 months following their discharge from hospital a research nurse telephoned those patients who had received a 4–6 months resource use diary and asked questions about their resource use.

Initial findings on this process

The organisation of the study, especially for patients who undergo surgery in one site and are followed up in another site (as in the case of Bath patients), means that, if resource use diaries are to be used in the main trial, site-specific resource use diaries and resource use questionnaires will have to be created. Some patients did not want to complete the questionnaire over the telephone; one patient wanted only to read out what he had written in the resource use diary rather than be asked questions. Another patient wanted to return the resource use diary rather than complete the telephone questionnaire. These findings have implications for how resource use is obtained for the main trial, as these two actions resulted in missing data.

To reduce the burden on patients, one suggestion is to use a much narrower NHS perspective or NHS and Personal Social Services (PSS) perspective for the main trial and include an exploration of other resource use within the qualitative component of the trial.

A shortened resource use instrument will still be required as not all secondary care visits and inpatient stays take place within the main treating hospitals where the initial surgery took place, and therefore medical records cannot provide this information. In some instances, when a substantial group of patients is followed up at another site, for example as in Bath in the feasibility study, it makes sense for the correct approvals to be gained so that access to medical records can be granted. In addition, community-based NHS resource use still needs to be obtained.

One solution would be to use the patient diary as the main patient source of information, using telephone interviews for those patients who do not return the diaries. Alternatively, given that patients did seem to be able to complete the resource use diaries, a shortened resource use questionnaire may be acceptable.

Establishing cost differences between the arms

One potential area in which cost differences will exist between the arms of the trial is in relation to the actual operation. A CRF was created that outlined potential equipment and consumables that could be used in all of the operations in terms of brand and quantity and the operation staff potentially involved in terms of their role and time present during the operation. Currently, this CRF is being piloted in Plymouth on the two different types of oesophagectomy. Once the main differences in theatre resource have been identified in this pilot, theatre records will be obtained from Bristol to ascertain whether or not the main differences in resource use can be captured easily within the Bristol centre.

Qualitative recruitment intervention

Recruitment was successful in no small part because, as our qualitative investigation has demonstrated, surgeons were able to present the equipoise that exists between the different approaches in consultations. The qualitative recruitment intervention has identified the key ways to balance the presentation of interventions and to explain to patients the need for a trial. 47,48 We will continue the qualitative recruitment intervention in the proposed definitive trial to ensure that the five centres joining the main trial will meet with the same success so that the definitive ROMIO trial as a whole recruits on time and to target.

Sample size for the planned definitive trial

In this section we illustrate the thinking behind the sample size calculation, as included in the current proposal for a definitive ROMIO trial. For simplicity, and to indicate the minimum statistical power that will be achieved for the comparison of recovery, we consider just the 6-week assessment of physical function. The planned analysis, with the baseline assessment of physical function as a covariate, is likely to have greater power than indicated here. We are assuming that having adjusted analyses for centre, there will no further need to accommodate clustering of outcomes by surgeon. In fact, as a team of surgeons is involved in each case (in decision-making and hospital care and often in theatre), it would be difficult to do this in practice.

A recent review of patient-reported outcomes has indicated that the minimum clinically important difference on the QLQ-C30 physical function subscale is 0.4 standard deviations. 49 Allowing for 5% of patients allocated to LAO actually undergoing open surgery and 10% of patients in each arm being found during surgery to have more extensive disease can be achieved by reducing the effect size to be detected to 0.34 standard deviations. In this situation 182 patients in each arm (364 patients in total) will allow a true treatment effect (LAO vs. open oesophagectomy) of 0.4 standard deviations to be detected with 90% power at the 5% significance level when up to 15% of patients are not able to follow their allocated procedure. Allowing for up to 10% of missing primary outcome data, for example because of patients being unwell, increases the target sample size to 364/0.9 = 406 patients in total. Hence, our sample size target for the definitive ROMIO trial is 203 patients allocated to LAO and 203 patients allocated to open oesophagectomy.

This sample size will also give adequate statistical power to detect a clinically important reduction in postsurgical length of stay. The mean length of stay in the pilot trial (all arms combined) was 13 days (standard deviation 8 days). Allowing for the skewed distribution,50 182 patients per group will allow a ratio of means of 0.84 to be detected with 80% power at the 5% significance level. This ratio of means corresponds to a 2.25-day reduction, from 14 to 11.75 days. Meta-analysis with data from the feasibility study will allow us to detect smaller differences in the average length of stay.

We are discussing carrying out an individual patient data meta-analysis of survival data from the pilot (n ≈ 150) and main (n = 406) ROMIO trials, the TIME trial15 (n = 115) and the MIRO trial14 (n = 200), giving a combined data set of > 800 patients. A data set of this size will rule out an absolute difference in mortality exceeding 6% (e.g. 7.5% with open surgery and 13.5% with minimally invasive surgery at 1 year) with 85% power and a 95% one-sided CI if true survival is equivalent.

Our experience with the pilot trial indicates that it is reasonable to expect an average of 22 patients to be recruited per centre per year. Seven centres recruiting for 2.5 years will enrol 2.5 × 7 × 22 = 385 patients. The two established centres will continue recruiting up to the definitive trial period and hence will also recruit during the 6 months preceding the main recruitment phase (0.5 × 2 × 22 = 22 patients), hence meeting the sample size target of 406 patients.

Task 1: Abdominal access

1.1 Opening the abdomen

1.1.1 Obtain safe access to the abdominal cavity.

1.1.2 Confirm the absence of metastatic disease and the appropriateness of the planned procedure.

Task 2: Diaphragmatic hiatus

2.1 Diaphragmatic hiatus

2.1.1 Mobilise the oesophagus from the diaphragmatic hiatus to the gastro-oesophageal junction, resecting a cuff of diaphragm and right and left paracardial lymph (LN) tissue (LN stations 1 and 2).

2.1.2 Dissect along the pericardial adventitia to remove the pericardial fat.

2.1.3 Resect the right pleura to expose the right lung.

2.1.4 Resect the left pleura to expose the left lung.

2.1.5 Dissect along the pre-aortic fascia.

Task 3: Gastric mobilisation

3.1 Gastric mobilisation

3.1.1 Identify the right gastroepiploic artery, which will provide the blood supply to the gastric tube.

3.1.2 Divide the greater omentum, ensuring that the gastroepiploic artery is preserved.

3.1.3 Enter the lesser sac and continue the dissection along the greater curvature of the stomach towards the spleen, dividing the short gastric and left gastroepiploic vessels and resecting the associated LN tissue (LN stations 4sa and 4sb).

Task 4: Coeliac axis

4.1 Portal vein and coeliac axis

4.1.1 Retract the stomach and dissect LN tissue along the superior border of the pancreas, to expose the portal vein.

4.1.2 Sling the common hepatic artery.

4.1.3 Dissect LN tissue along the proper hepatic artery, common hepatic artery, coeliac artery, left gastric artery and proximal splenic artery (LN stations 7, 8a, 9, 11p and 12a).

4.1.4 Ligate and divide the left gastric vein at the portal vein.

4.1.5 Ligate and divide the left gastric artery at its origin from the coeliac artery.

4.1.6 Dissect LN tissue from the left side of the coeliac artery to the left crus at the oesophageal hiatus and left side of Gerota’s fascia.

4.2 Splenic artery

4.2.1 Continue the dissection along the anterior surface of the proximal splenic artery towards the splenic hilum.

4.2.2 Ligate the posterior gastric vessels at their origin from the splenic artery.

4.2.3 Dissect the remaining LN tissue along the distal splenic artery, clearing to the splenic vein inferiorly and the abdominal wall posteriorly, until the splenic hilum is reached (LN station 11d).

4.2.4 Clear the splenic hilum of LN tissue (LN station 10).

Note: This dissection should marry up with that performed in Task 3.

Task 5: Gastric tube

5.1 Gastric tube

5.1.1 Perform a lymphadenectomy along the lesser curvature of the stomach until the expected lower resection margin is reached (LN stations 3a and 3b).

5.1.2 Create the gastric tube.

5.1.3 Pyloroplasty, pyloromyotomy, other (e.g. botulinum toxin, Botox^®, Allergan) or no action may be performed.

Task 6: Insertion of surgical adjuncts

6.1 Feeding jejunostomy

6.1.1 A feeding jejunostomy may be placed.

6.1.2 If the abdominal phase of the operation has been performed minimally invasively, a port site may be extended to an 8-cm incision to facilitate the placement of a feeding jejunostomy.

6.2 Abdominal drains

6.2.1 Abdominal drain(s) may be placed.

Task 7: Abdominal closure

7.1 Abdominal closure

7.1.1 Perform abdominal lavage.

7.1.2 Confirm haemostasis.

7.1.3 Close the abdomen.

7.1.4 Dress the wound.

Abdominal lymph node stations

Japanese Gastric Cancer Association. Japanese classification of gastric carcinoma: 3^rd English Edition. Gastric Cancer 2011;14:101–12.

See table 5 in the above paper.

Task 1: Thoracic access

1.1 Thoracic access

1.1.1 Obtain safe access to the patient’s right chest.

Task 2: Thoracic lymphadenectomy

2.1 Thoracic lymphadenectomy

2.1.1 Mobilise the lower lobe of the right lung.

2.1.2 Ligate and divide the azygos vein at the azygos arch.

2.1.3 Dissect along the pericardium until the left lung is reached including the pleura of the left lung in the radial excision margin.

2.1.4 Perform a subcarinal lymphadenectomy (LN station 107).

2.1.5 Clear both bronchi of LN tissue until the hilum of each lung is reached (LN station 109).

2.1.6 Dissect along the right pulmonary veins, continuing posteriorly until the left pulmonary veins are reached.

2.1.7 Dissect the mediastinal pleura at the anterolateral border of the thoracic aorta.

2.1.8 Dissect along the pre-aortic fascia from the proximal resection margin towards the diaphragm (LN station 112).

2.1.9 Dissect LN tissue along the aorto-pulmonary window, clearing the arch of the aorta, pulmonary artery and recurrent laryngeal nerve as it hooks around the arch of the aorta.

2.1.10 Identify and ligate the thoracic duct at the proximal resection margin and at the level of the diaphragm such that it is resected with the specimen.

Task 3: Specimen excision

3.1 Specimen excision

3.1.1 Ensure that the thoracic part of the specimen is circumferentially free, from the previously completed diaphragmatic mobilisation (performed during the abdominal phase) to at least the level of the aortic arch (LN stations 108, 110 and 111).

3.1.2 Deliver the stomach into the right chest cavity, ensuring that the gastric tube can reach the site of anastomosis without tension or torsion.

3.1.3 Excise the specimen with a suitable distal resection margin.

3.1.4 Send the specimen to pathology as per protocol.

Task 4: Anastomosis

4.1 Oesophago-gastrostomy

4.1.1 Perform an oesophago-gastrostomy.

4.1.2 If performing a two-phase minimally invasive procedure, an incision of up to 5 cm may be made in addition to the existing ports.

4.1.3 If performing a three-phase minimally invasive procedure, a left cervical incision is permitted for the anastomosis to be made.

Task 5: Insertion of surgical adjuncts

5.1 Nasogastric/nasojejunal tube

5.1.1 A nasogastric or nasojejunal tube may be placed.

5.2 Thoracic drains

5.2.1 Thoracic drains should be placed prior to the closure of the thoracic incision.

Task 6: Thoracic closure

6.1 Thoracic closure

6.1.1 Perform lavage.

6.1.2 Confirm haemostasis.

6.1.3 Close the chest.

6.1.4 Dress the wound.

Thoracic lymph node stations

Japan Esophageal Society. Japanese classification of esophageal cancer, tenth edition: part 1. Esophagus 2009;6:1–25.

See page 12 in the above paper.

Task 1: Abdominal access

1.1 Abdominal access

1.1.1 Obtain safe access to the abdominal cavity.

1.1.2 Confirm the absence of metastatic disease and the appropriateness of the planned procedure.

Task 2: Diaphragmatic hiatus

2.1 Diaphragmatic hiatus

2.1.1 Mobilise the oesophagus from the diaphragmatic hiatus to the gastro-oesophageal junction, resecting right and left paracardial lymph node (LN) tissue (LN stations 1 and 2).

2.1.2 A cuff of diaphragm should be resected in advanced disease to achieve a clear circumferential margin if advanced disease.

2.1.3 Dissect along the pericardial adventitia to remove the pericardial fat.

2.1.4 Resect the right pleura to expose the right lung in advanced disease.

2.1.5 Resect the left pleura to expose the left lung in advanced disease.

2.1.6 Dissect along the pre-aortic fascia.

Task 3: Gastric mobilisation

3.1 Gastric mobilisation

3.1.1 Identify the right gastroepiploic artery, which will provide the blood supply to the gastric tube.

3.1.2 Divide the greater omentum to enter the lesser sac, ensuring that the gastroepiploic vessels are preserved.

3.1.3 Continue the dissection along the greater curvature of the stomach towards the spleen, dividing the short gastric and left gastroepiploic vessels.

3.1.4 Resect the associated LN tissue (O) (LN stations 4sa and 4sb).

Task 4: Coeliac axis

4.1 Portal vein and coeliac axis

4.1.1 Retract the stomach and dissect along the superior border of the pancreas to expose the portal vein (O).

4.1.2 Sling the common hepatic artery (O).

4.1.3 Dissect LN tissue along the proper hepatic artery (O) (LN station 12a).

4.1.4 Dissect LN tissue along the common hepatic artery, coeliac artery, left gastric artery and proximal splenic artery (LN stations 7, 8a, 9 and 11p).

4.1.5 Ligate and divide the left gastric vein close to the portal vein.

4.1.6 Ligate and divide the left gastric artery at its origin from the coeliac artery.

4.1.7 Dissect LN tissue from the left side of the coeliac artery to the left crus at the oesophageal hiatus and left side of Gerota’s fascia.

4.2 Splenic artery

4.2.1 Continue the dissection along the anterior surface of the proximal splenic artery towards the splenic hilum.

4.2.2 Ligate the posterior gastric vessels at their origin from the splenic artery.

4.2.3 Dissect the remaining LN tissue along the distal splenic artery, clearing to the splenic vein inferiorly and the abdominal wall posteriorly, until the splenic hilum is reached (O) (LN station 11d).

4.2.4 Clear the splenic hilum of LN tissue (O) (LN station 10).

Task 5: Gastric tube

5.1 Gastric tube

5.1.1 The lesser curvature of the stomach is cleared of LN tissue at the appropriate level until the expected distal resection margin is reached (LN stations 3a and 3b).

5.1.2 Create the gastric tube. This may be performed in the chest.

5.1.3 Oversew the staple line (O).

5.1.4 Pyloroplasty, pyloromyotomy or other may be performed (O).

Task 6: Insertion of surgical adjuncts

6.1 Feeding jejunostomy

6.1.1 A feeding jejunostomy may be placed (O).

6.1.2 If the abdominal phase of the operation has been performed minimally invasively, a port site may be extended to facilitate the placement of a feeding jejunostomy (O).

6.2 Abdominal drains

6.2.1 Abdominal drain(s) may be placed (O).

Task 7: Abdominal closure

7.1 Abdominal closure

7.1.1 Perform abdominal lavage (O).

7.1.2 Confirm haemostasis.

7.1.3 Close the abdomen.

7.1.4 Nursing staff dress the wound as per ROMIO trial guidelines.

Abdominal lymph node stations

Japanese Gastric Cancer Association. Japanese classification of gastric carcinoma: 3^rd English Edition. Gastric Cancer 2011;14:101–12.

See table 5 in the above paper.

Note: The abdominal lymphadenectomy is identical to that for a D2 gastrectomy except LN stations 4d (right gastroepiploic artery), 5 (suprapyloric LN) and 6 (infrapyloric LN) are not included.

Task 1: Thoracic access

1.1 Thoracic access

1.1.1 Obtain safe access to the patient’s right chest.

1.1.2 Confirm the absence of metastatic disease in the chest.

Task 2: Thoracic lymphadenectomy

2.1 Thoracic lymphadenectomy

2.1.1 Divide the inferior pulmonary ligament.

2.1.2 Ligate and divide the azygos arch.

2.1.3 Dissect along the pericardium until the left lung is reached, including resection of the left pleura in advanced disease to achieve a clear circumferential margin.

2.1.4 Perform a subcarinal lymphadenectomy (LN station 107).

2.1.5 Clear both bronchi of LN tissue until the hilum of each lung is reached (LN station 109).

2.1.6 Dissect along the right pulmonary veins, continuing posteriorly until the left pulmonary veins are reached (O).

2.1.7 Dissect the mediastinal pleura at the anterolateral border of the thoracic aorta.

2.1.8 Dissect along the pre-aortic fascia, from the proximal resection margin towards the diaphragm (LN station 112).

2.1.9 Dissect LN tissue along the aorto-pulmonary window, clearing the arch of the aorta, pulmonary artery and recurrent laryngeal nerve as it hooks around the arch of the aorta (O).

2.1.10 Identify and ligate the thoracic duct at the proximal resection margin and above the level of the diaphragm such that it is resected with the specimen.

Task 3: Specimen excision

3.1 Specimen excision

3.1.1 Ensure that the thoracic part of the specimen is circumferentially free, from the previously completed diaphragmatic mobilisation (performed during the abdominal phase) to at least the level of the aortic arch (LN stations 108, 110 and 111).

3.1.2 Deliver the stomach into the right chest cavity ensuring that the gastric tube can reach the site of anastomosis without tension or torsion.

3.1.3 Excise the specimen with suitable proximal and distal resection margins.

3.1.4 Send the specimen to pathology as per the ROMIO trial protocol.

Task 4: Anastomosis

4.1 Oesophago-gastrostomy

4.1.1 Perform an oesophago-gastrostomy using your preferred method.

4.1.2 If performing a two-phase minimally invasive procedure, a further incision may be made in addition to the existing port sites (O).

4.1.3 If performing a three-phase procedure, a left cervical incision is permitted for the anastomosis to be made (O).

Task 5: Insertion of surgical adjuncts

5.1 Nasogastric/nasojejunal tube

5.1.1 A nasogastric or nasojejunal tube may be placed (O).

5.2 Thoracic drains

5.2.1 Thoracic drain(s) should be placed prior to the closure of the thoracic incision.

Task 6: Thoracic closure

6.1 Thoracic closure

6.1.1 Perform lavage (O).

6.1.2 Confirm haemostasis.

6.1.3 Reinflate the lung under direct vision.

6.1.4 Close the chest.

6.1.5 Nursing staff dress the wound as per ROMIO trial guidelines.

Thoracic lymph node stations

Japan Esophageal Society. Japanese classification of esophageal cancer, tenth edition: part 1. Esophagus 2009;6:1–25.

See page 12 in the above paper.

Abdominal access

1. Confirm the absence of metastatic disease.

Diaphragmatic hiatus

1. Mobilise the gastro-oesophageal junction, resecting right and left paracardial lymphatic (LN) tissue (LN stations 1 and 2).

2. Resect a cuff of diaphragm and pleura to achieve a clear circumferential margin in advanced disease.

3. Dissect along the pre-aortic fascia.

Gastric mobilisation

1. Mobilise the stomach based on the right gastroepiploic vessels.

Coeliac axis

1. Dissect LN tissue along the common hepatic artery, coeliac artery, left gastric artery and proximal splenic artery (LN stations 7, 8a, 9 and 11p).

2. Ligate and divide the left gastric vein close to the portal vein and the left gastric artery at the coeliac artery.

3. Dissect LN tissue from the left side of the coeliac artery to the left crus at the oesophageal hiatus and the left side of Gerota’s fascia.

4. Continue the dissection along the anterior surface of the proximal splenic artery towards the splenic hilum and ligate the posterior gastric vessels at their origin from the splenic artery.

Gastric tube

1. Create the gastric tube, removing tissue along the lesser curvature of the stomach (LN stations 3a and 3b). This step may be carried out in the chest.

Thoracic access

1. Exclude metastatic disease in the chest.

Thoracic lymphadenectomy

1. Divide the inferior pulmonary ligament and ligate and divide the azygos arch.

2. Dissect along the pericardium until the left pulmonary vein is reached, including the left pleura in advanced disease.

3. Perform a subcarinal lymphadenectomy (LN station 107) and clear both bronchi of LN tissue (LN station 109).

4. Dissect the mediastinal pleura at the anterolateral border of the thoracic aorta and dissect along the pre-aortic fascia from the proximal resection margin towards the diaphragm (LN station 112).

5. Identify and ligate the thoracic duct at the proximal resection margin and above the diaphragm.

Specimen excision

1. Ensure that the thoracic part of the specimen is circumferentially free, from the previously completed diaphragmatic mobilisation (performed during the abdominal phase) to at least the level of the aortic arch (LN stations 108, 110 and 111).

2. Deliver the stomach into the right chest cavity ensuring that the gastric tube can reach the site of anastomosis without tension or torsion.

3. Excise the specimen with suitable proximal and distal resection margins and send it to pathology as per the ROMIO trial protocol.

Anastomosis

1. Perform an oesophago-gastrostomy using the preferred technique.