Surgery versus radical endotherapies for early cancer and high‐grade dysplasia in Barrett's oesophagus

Abstract

Background

Barrett's oesophagus is one of the most common pre‐malignant lesions in the world. Currently the mainstay of therapy is surgical management of advanced cancer but this has improved the five‐year survival very little since the 1980s. As a consequence, improved survival relies on early detection through endoscopic surveillance programmes. Success of this strategy relies on the fact that late‐stage pre‐malignant lesions or very early cancers can be cured by intervention. Currently there is considerable controversy over which method is best: that is conventional open surgery or endotherapy (techniques involving endoscopy).

Objectives

We used data from randomised controlled trials (RCTs) to examine the effectiveness of endotherapies compared with surgery in people with Barrett's oesophagus, those with early neoplasias (defined as high‐grade dysplasia (HGD) and those with early cancer (defined as carcinoma in‐situ, superficially invasive, early cancer or superficial cancer T‐1m (T1‐a) and T‐1sm (T1‐b)).

Search methods

We used the Cochrane highly sensitive search strategy to identify RCTs in MEDLINE, EMBASE, the Cochrane Central Register of Controlled Trials (CENTRAL), ISI Web of Science, EBMR, Controlled Trials mRCT and ISRCTN, and LILACS, in July and August 2008. The searches were updated in 2009 and again in April 2012.

Selection criteria

Types of studies: RCTs comparing endotherapies with surgery in the treatment of high‐grade dysplasia or early cancer. All cellular types of cancer were included (i.e. adenocarcinomas, squamous cell carcinomas and more unusual types) but will be discussed separately.

Types of participants: patients of any age and either gender with a histologically confirmed diagnosis of early neoplasia (HGD and early cancer) in Barrett's or squamous lined oesophagus.

Types of interventions; endotherapies (the intervention) compared with surgery (the control), all with curative intent.

Data collection and analysis

Reports of studies that meet the inclusion criteria for this review would have been analysed using the methods detailed in Appendix 9.

Main results

We did not identify any studies that met the inclusion criteria. In total we excluded 13 studies that were not RCTs but that compared surgery and endotherapies.

Authors' conclusions

This Cochrane review has indicated that there are no RCTs to compare management options in this vital area, therefore trials should be undertaken as a matter of urgency. The problems with such randomised methods are standardising surgery and endotherapies in all sites, standardising histopathology in all centres, assessing which patients are fit or unfit for surgery and making sure there are relevant outcomes for the study (i.e. long‐term survival (over five or more years)) and no progression of HGD.

Plain language summary

Surgery versus endoscopic therapies for early oesophageal cancer in Barrett's oesophagus

This Cochrane review has indicated that there are no randomised controlled trials to compare management options in this vital area, therefore trials should be undertaken as a matter of urgency. Current use of endotherapies in the care of patients with early cancer or high‐grade dysplasia of Barrett's oesophagus should be at the recommendation of the multi‐disciplinary team involved in individual care. Properly conducted randomised controlled trials comparing surgery with endotherapies should be conducted before any conclusions can be drawn.

Background

Description of the condition

Barrett's oesophagus is an acquired condition resulting from persistent and severe mucosal injury to the oesophagus. It develops as a result of chronic reflux of duodenogastric contents (Caldwell 1995; Vaezi 1995). Exposure to acid, bile and other constituents of refluxate causes cellular and deoxyribonucleic acid (DNA) damage as well as altering the differentiation of proliferating cells. The normal squamous epithelium of the distal oesophagus is replaced with the columnar type of epithelium normally found in gastric cardia, the gastric fundus and the upper intestine resulting in a macroscopically visible change, obvious through a normal endoscope. The British Society of Gastroenterology guidelines define Barrett's oesophagus as "an endoscopically apparent area above the gastro‐oesophageal junction that is suggestive of Barrett's which is supported by the findings of columnar lined oesophagus on histology" (Playford 2006).

Barrett's is asymptomatic and common, being detected in 0.5% of post‐mortem studies (Cameron 1990), 15% to 20% of those undergoing upper gastrointestinal endoscopy for symptoms of chronic gastro‐oesophageal reflux (Winters 1987) and 1.6% of the general Swedish population (Ronkainen 2005). The major concern is the potential of progression to malignancy. The cancers that occur in areas of Barrett's oesophagus are almost exclusively adenocarcinomas (Byrne 2002).

It is thought that adenocarcinoma in Barrett's epithelium does not arise as a single event but rather follows a stepwise process from metaplasia through low‐grade (LGD) and high‐grade dysplasia (HGD) to early carcinoma and eventually to invasive cancer (Jankowski 1995). Progression from non‐dysplastic Barrett's oesophagus to malignancy is rare but a diagnosis of dysplasia within Barrett's section confers an increased risk, most significant with HGD (Buttar 2001; Reid 2000).

Every year in the UK around 7000 cases of oesophageal cancer are diagnosed. The vast majority of these (> 85%) are either adenocarcinomas (Byrne 2002) or squamous cell carcinomas. Worldwide, the incidence rate of squamous cell carcinoma of the oesophagus is fairly stable, but there is an increasing trend of incidence for oesophageal adenocarcinomas in some European countries and the US (Vizcaino 2002). In the UK and US (in white men), adenocarcinoma is now the dominant histology (CRUK 2012).

These two histological types differ in their risk factors and pathobiology. Adenocarcinomas tend to develop in white males with centripetal obesity and Barrett's oesophagus (Corley 2007; Edelstein 2007; Hampel 2005). Squamous cell carcinomas are more strongly associated with smoking (Chen 1997), alcohol (Thun 1997), and diets rich in salted meat, salted fish, pickled vegetables and lacking in fresh fruit, fresh vegetables and vitamins A and C (Brown 1998; Hung 2004). Variations in alcohol metabolism may contribute to the risk of developing squamous cell cancers (Engel 2003; Morita 1994). Chak 2002 demonstrated in a small case control study that individuals with Barrett's oesophagus and adenocarcinoma had a positive family history of these disorders more frequently than control cases who did not have a family history of these conditions. There is probably a genetic susceptibility to gastro‐oesophageal reflux disease (GORD) (de Vries 2009) and this may be a component in the development of Barrett's oesophagus and oesophageal cancer (Lagergren 1999; Lagergren 2000; Romero 2002).

Both types of cancer can present with a non‐specific early symptomatology, and are often detected late, leading to a poor prognosis (Green 2011). An incidence of 12 per 100,000 and a five‐year survival rate of 9% together make oesophageal cancer the third most common cause of cancer death (Hellier 2006). Endoscopic developments have enabled diagnosis of these lesions at progressively earlier stages but there is currently no clear guidance for optimal management of early disease such as HGD or mucosal/submucosal cancer with no metastases (T1m/sm N0 M0) once detected.

Description of the intervention

Description of the interventions and how the interventions are thought to work

Surgery

For many years, open total oesophagectomy has been the treatment of choice for early tumours. There are three common techniques ‐ transhiatal, Ivor‐Lewis trans‐thoracic and three stage (McKeown) with or without neo‐adjuvant chemotherapy and/or radiotherapy. No one technique has shown overall superiority and the method selected will depend on the tumour and the preference/experience of the surgeon. All carry a degree of risk in terms of morbidity (Holscher 1997; Thomas 1996) and mortality. A range of post‐surgery mortality rates is reported from up to 20% post‐operatively and 85% at five years (Urba 2001), to as low as 2% (Reed 2005); 4% (Hulscher 2002) and 4.9% (at 30 days) (Jamieson 2004); results are improved at high‐volume centres using careful case selection (Birkmeyer 2002; Chang 2006; Rouvelas 2007; Whooley 2001). The third annual report of the UK national oesophago‐gastric audit, which included an estimated 84% of all operations for upper gastrointestinal in England and Wales in 2010, reported an overall inpatient mortality of 4.5% (90‐day mortality 5.7%) following oesophagectomy with curative intent. About 29.8% of patients suffered one or more complications with 8.3% suffering from anastomotic leak. The rate of re‐operation was 10.2% (National Oesophago‐gastric Audit 2010). There is some evidence that oesophagectomy performed for HGD or intramucosal cancer (IMC) is associated with a slightly lower morbidity and mortality as patients with early disease tend to be younger with fewer co‐morbidities.

New surgical techniques, such as jejunal interposition (Stein 2007) and minimally invasive laparoscopic and thoracoscopic techniques, have been developed, in addition to other less radical surgical interventions including limited lymphadenectomy and vagal sparing oesophagectomy (Akiyama 1982; Banki 2002). The latter techniques may represent an appropriate alternative to traditional surgery or radical endotherapy for patients with early neoplastic disease (HGD, IMC) and have demonstrated low morbidity and mortality in tertiary centres (DeMeester 2008; Peyre 2007). Many clinicians choose to avoid surgery and instead manage selected patients with surveillance alone or with one of several endoscopic therapies to resect or ablate the tumour.

Endoscopic treatments

Ablation therapies

Ablation therapies aim to destroy the affected tissue, allowing replacement with normal squamous tissue (Sampliner 1996).

Methods include:

• argon plasma coagulation (APC), which uses ionised argon gas to conduct a monopolar current to the tissue, thereby causing necrosis through electrocoagulation. Different depths of necrosis can be achieved by varying the power of the generator, the distance of the probe from the tissue, the amount of gas flow and the duration of application;

• multipolar electrocautery (MPEC), in which the targeted tissue completes an electrical circuit between two electrodes on the probe tip, thereby causing thermal necrosis;

• laser therapy, which can be used at various wavelengths to destroy cells at different depths;

• cryotherapy, which involves the application of extreme cold to destroy the target cells;

• radiofrequency ablation (RFA) is a relatively recent development. The Barrx system delivers thermal energy through radiofrequency electrodes either through a focal device or coiled around a balloon that inflates to come in contact with the oesophageal wall; and

• photodynamic therapy (PDT) uses a phototoxic reaction to induce cell death. A photosensitising agent is administered that is absorbed into normal and abnormal cells but retained for a longer period in the abnormal cells. After a suitable time period, an endoscopy is performed to expose the oesophagus to light of an appropriate wavelength. The Barrett's and cancerous cells that have retained the photosensitiser then undergo a delayed cell death.

National Institute for Health and Clinical Evidence (NICE) guidance (UK) currently recommends the use of RFA or PDT in the treatment of flat HGD, combined where appropriate with endoscopic resection (ER) (see below) but it does not prioritise one therapy over another. The use of APC, laser ablation or MPEC alone are not recommended unless as part of a clinical trial but these ablative techniques may be combined with ER (NICE 2010).

Endoscopic resection

ER encompasses a range of endoscopic techniques to remove the mucosal and submucosal layers (down to the muscularis propria) of the oesophagus. The conventional technique involves either a simple snare, or a double channel gastroscope with grasping forceps down one of the channels and a snare down the other (strip biopsy technique). Techniques involve suction of the lesion into a plastic cap with or without submucosal injection, using a cap and snare technique or a ligation device.

Endoscopic submucosal dissection (ESD) uses a variety of endoscopic knives to dissect out lesions. If the lesion is larger than 2 cm then conventional endoscopic mucosal resection (EMR) techniques will result in multi‐piece resection. Multi‐piece resection can prevent the pathologist from accurately staging the tumour and could lead to a higher rate of local recurrence. The goal and a primary principle of surgical oncology is to remove all of the tumour in one piece and demonstrate that the resected tumour has margins of cancer‐free tissue. In the case of larger lesions, ESD meets this oncological principle, permitting en‐bloc resection of lesions of any size in one piece and avoiding the problems associated with multi‐piece resection.

Why it is important to do this review

The occurrence of HGD and early mucosal cancer in Barrett's oesophagus is an important watershed in the progression of carcinogenesis. Eradication of HGD and early mucosal cancer leads to a very high cure rate whereas missing the consequential invasive cancer leads to an unacceptable progression to incurable disease. In addition, there are several new technologies that have been specifically shown to deal with this condition.

The risk of invasive oesophageal adenocarcinoma (invasion into or beyond the submucosa) in patients with biopsy‐detected HGD has formerly been over‐estimated. The majority of cancers detected in this group are T1 cancers and so potentially amenable to endoscopic treatment. The widespread use of ER has dramatically improved disease staging in patients with early lesions and one +

meta‐analysis concluded that the incidence of submucosally invasive cancer is only 12.7% in patients with visible HGD in Barrett's and 3% if the HGD does not produce a visible abnormality (Konda 2008). In addition, rates of lymphatic metastases appear to be low or absent in T1m (and possibly T1sm1) cancers demonstrating potential for curative endotherapy in these patients without the need for surgical lymphadenectomy (Alvarez Herrero 2010; Lui 2005). Improvements in endoscopic imaging are also increasingly enabling diagnosis of oesophageal neoplasia at an earlier, potentially curative, stage.

Increasingly clinicians are choosing to avoid surgery as first‐line therapy and instead manage selected patients with endoscopic therapy to resect and ablate the tumour. Surgery is now often reserved as a second‐line intervention following failure of endoscopic therapy, particularly in patients with focal flat HGD; however, this strategy is not fully evidence based.

It is vital that clear guidance is available on the optimal treatment of early Barrett's neoplasia. Currently its management remains controversial and clinical practice varies significantly both nationally and internationally.

The morbidity and mortality of endoscopic therapies is probably considerably less than surgery. It seems likely (but unconfirmed) that complete ablation of the Barrett's segment, in addition to removal of the early cancer/dysplastic area, may be necessary to prevent the development of metachronous lesions. The long‐term outcomes of these relatively recent developments are becoming clearer, with five‐year survival in one large single centre study being similar to that of an age‐ and country‐matched population (Pech 2008). The lack of any cohesive policy is emphasised by a report suggesting that patients with HGD or IMC within Barrett's are far more likely (86% versus 12%) to undergo oesophagectomy if first seen by a surgeon rather than a gastroenterologist (Yachimski 2008).

Objectives

The objective of this review was to use data from randomised controlled trials (RCTs) to examine the effectiveness of endotherapies (the intervention) compared with surgery (the control), in two groups of people with Barrett's oesophagus; those with early neoplasias (defined as HGD, and those with early cancer (defined as carcinoma in‐situ, superficially invasive, early cancer or superficial cancer T‐1m (T1‐a) and T‐1sm (T1‐b)).

To investigate the following:

1) In patients with either HGD or early cancer, who received either endotherapies or surgery, what are the overall survival rates at one, five or more years?

We are interested in progression of existing disease, that is, for people who have HGD, the effect of the therapies on rates of progression to cancer and for people having early cancer, progression to more invasive cancer, defined as cancer having greater depth of invasion into the surrounding tissues or spread to distant sites. We have examined the effects of the therapies on both progression and cancer‐free survival as follows:

2a) In patients with HGD, are endoscopic therapies effective in terms of reducing rates of progression to cancer?

2b) In patients with early cancer, do endoscopic therapies have an effect on rates of progression to more invasive cancer?

2c) Are there any differences in the rates of cancer‐free survival for participants with either HGD or early cancer who received endotherapies compared with trial participants who received surgery?

In both groups of patients (either HGD or early cancer), we hope to be able to obtain information to address the following questions:

4) Do endoscopic therapies prove effective in preventing recurrence of the original lesion?

5) Do endoscopic therapies have an effect on rates of development of metachronous oesophageal neoplasia?

6) Are there any differences between the intervention and the control therapies in the rates of procedure related mortality (30 day), and early and late complications?

7) Are there differences in the quality of life (QoL) experienced after surgery or endoscopic therapy?

It was anticipated that there would be few appropriate RCTs and that evidence from other non‐comparative or open trials would form the basis for discussion. We also searched for, and included in the discussion, RCT studies which compare endotherapies with each other.

Methods

Criteria for considering studies for this review

Types of studies

RCTs in which one of the endotherapies is compared with surgery in the treatment of early neoplasia (defined as HGD, or early cancer defined as carcinoma in‐situ, superficially invasive, early cancer or superficial cancer T‐1m and T‐1sm, also commonly referred to as T1a and T1b) on a background of Barrett's oesophagus. All cellular types of cancer were eligible for inclusion (i.e. adenocarcinomas, squamous cell carcinomas and more unusual types) but would be discussed separately owing to their differing evolution and prognosis.

Types of participants

Patients of any age and either gender found to have a histologically confirmed diagnosis of early neoplasia (HGD and early cancer) on a background of Barrett's or squamous lined oesophagus.

Types of interventions

Endotherapies (the intervention) compared with surgery (the control), all with curative intent.

Types of outcome measures

Primary outcomes

In patients with HGD:

1. Progression to cancer at one or more years.

In patients with early cancer:

2. Overall survival at one or more years. All reports of survival will be considered, such as hazard ratios, log rank tests, percentages.

Secondary outcomes

In patients with either HGD or early cancer:

1. Cancer‐free survival at one, five and, if data are available, 10 years;
2. Progression‐free survival at one, five and, if data are available, 10 years.

If data are available for patients with either HGD or early cancer, we shall collect data for the following outcomes:

3. Recurrence of original lesion;
4. Development of a metachronous oesophageal cancer;
5. Procedure‐related mortality (30 day);
6. Early complications such as perforation, bleeding, infection;
7. Late complications such as stricture development;
8. QoL.
9. Persistence of Barrett's oesophagus.
Economic data will not be analysed but will be addressed in the discussion.

Search methods for identification of studies

Electronic searches

Search strategy to locate RCTs

We used the Cochrane Highly Sensitive Search Strategy for identifying RCTs in MEDLINE (Appendix 1; Higgins 2011), combined with the search terms in Appendix 2 to identify RCTs in MEDLINE. The MEDLINE search strategy was adapted for use in the other databases searched (see the corresponding appendix for the search terms used).

We searched the following databases in 2008 (Green 2008):

• Cochrane Upper Gastrointestinal and Pancreatic Diseases (UGPD) Group Trials Register;

• MEDLINE (January 1996 to August 2008) (Appendix 2), searched 11 August 2008;

• EMBASE (January 1996 to week 3 2008) (Appendix 3), searched using the RCT filter on July 23 2008 and again, in order to widen the search, without the RCT filter on August 11 2008;

• the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2008, Issue 3) (Appendix 4), searched 24 July 2008;

• ISI Web of Science (January 2000 to July 2008) (Appendix 5), searched 23 July 2008;

• EBMR (January 2000 to August 2008) (Appendix 6), searched 6 August 2008;

• Controlled Trials mRCT and ISRCTN (January 2000 to July 2008) (Appendix 7), searched 23 July 2008; and

• LILACS (January 2000 to July 2008) (Appendix 8), searched 23 July 2008.

The searches were updated in November 2009 in EMBASE, MEDLINE and CENTRAL and Controlled Trials mRCT and ISRCTN. The search for LILACS was updated and re‐run from 1 January 2000 to November 2009 (Green 2008). We revised the LILACS search strategy to use DeCS (Descriptors in Health Sciences) as it is trilingual (English, Portuguese and Spanish) vocabulary coding. We also simplified the strategy in the hope of finding more relevant citations. Because this was a change from the search strategy published in the protocol of this review, we re‐searched LILACS from January 1 2000 to November 2009.

We updated the searches in 2012 as follows:

• Ovid MEDLINE(R) (1946 to March week 4 2012);

• EMBASE (1980 to 2012 week 14);

• EBMR, CENTRAL (April 2012);

• Web of Knowledge (v.5.5) (searched 11 April 2012);

• LILACS (searched 10 April 2012);

• www.controlled‐trials.com/ (searched 17 May 2012).

Adverse outcomes

We did not do a separate search for adverse events.

Language

We did not confine our search to English language publications and translations were sought where necessary.

Searching other resources

Reference lists from trials selected by electronic searching were scanned to identify further relevant trials. Published abstracts from conference proceedings from the United European Gastroenterology Week (published in Gut) and Digestive Disease Week (published in Gastroenterology) were handsearched. We searched for relevant published trials in The Cochrane UGPD Group, in CENTRAL in The Cochrane Library (November 2009).

Grey Literature

We attempted to find unpublished studies through searching grey literature sources (Auger 1998).

Handsearching

The Cochrane UGPD Group's trials register, which contains the results of a comprehensive programme of ongoing handsearching of gastroenterology journals and conference proceedings, was searched.

Reference lists

We scanned the bibliographies of published studies and reviews for possible references to RCTs.

Correspondence

We tried to identify unpublished or on‐going trials, by writing to trial authors who published reports of trials relevant to this review up to 10 years ago, regardless of the trial's inclusion or exclusion in this review.

Data collection and analysis

Selection of studies

Three review authors (SG, CB and JJ) independently reviewed titles and abstracts of references retrieved from the searches and selected all potentially relevant studies. Copies of these articles were obtained, and reviewed independently by the same authors against the inclusion criteria of the study. Two review authors (SG and CB) independently reviewed the results of the updated searches.

Authors were not blinded to the names of the trial authors, institutions or journal of publication. There was no disagreement between the authors and they did not approach or appeal to the editorial base of the Cochrane UGPD Group for consensus. We did not identify any studies that met the inclusion criteria. If updated searches retrieved any reports of studies that meet the inclusion criteria for this review, they would have been analysed using the methods detailed in Appendix 9.

Results

Description of studies

Results of the search

We searched electronic databases in 2008 and received 2568 reports of potential RCTs. The review authors identified an additional 13 citations. Two hundred and sixty‐five reports were duplicates and were discarded (Figure 1). No ongoing trials that compared surgery to endotherapy in Barrett's oesophagus were found.

1.

Study flow diagram.

When the searches were updated in November 2009, a further 571 studies were identified. None of these was an RCT; we formally excluded one of these studies (Yachimski 2008).

In April 2012 in total we found 626 new citations from electronic searches (MEDLINE 160, EMBASE 544, CENTRAL 29, LILACS 0, ISI 14). CB and MA screened the abstracts and titles of all the hits and identified 11 potential RCTs, which we obtained in full text. None was an RCT, we discarded five of these newly identified studies and the remaining six were formally excluded (Farrell 2011; Pech 2011; Rosmolen 2010; Schembre 2010; Tian 2011; Zehetner 2011).

A bibliography search in 2012 identified Thomas 2005, which we obtained in full text and formally excluded as it was not an RCT, and Comay 2007, a cost analysis that we refer to elsewhere in the review (Discussion).

We searched for ongoing clinical trials and identified 153 registered trials. None of these was a comparison of surgery versus endotherapies. We received details of one trial that has not commenced from the principal investigator (BRIDE 2012), which we have placed in the Studies awaiting classification list until more details are available for a full appraisal.

Included studies

No studies met the inclusion criteria.

Excluded studies

None of the studies identified in the searches were suitable for inclusion. There were five trials that compared endoscopic with surgical treatment of HGD or early cancer, or both, in Barrett's oesophagus (Das 2008; Pacifico 2003; Prasad 2007; Reed 2005; Schembre 2008) but all were retrospective. One further excluded study was identified (Yachimski 2008) during the updated searches carried out in November 2009.

In 2012 we excluded six studies that we identified from electronic database searches (Farrell 2011; Pech 2011; Rosmolen 2010; Schembre 2010; Tian 2011; Zehetner 2011) and a further study (Thomas 2005) from a bibliography search of Menon 2010.

In total there are 13 excluded studies (see Characteristics of excluded studies).

Risk of bias in included studies

There were no included studies.

Effects of interventions

We did not perform a meta‐analysis as there were no included studies. We summarised data from the excluded studies in Table 1, Table 2 and Table 3 and the results from these in the context of treatment for HGD/early cancer lesion within a Barrett's oesophagus are discussed.

1. Demographic data.

Study	Type	Treatment modalities investigated	No. of patients	Ratio endoscopic: surgical therapy	Male (%)	Mean age years (/surgery)	Histological type	Follow‐up time
Pacifico 2003	Retrospective single institution	Combined PDT and EMR vs. oesophagectomy	88	24:64	90	67 endotherapy group 68 surgery group	Adenocarcinoma T1m / T1sm (stage 1)	17 months
Prasad 2007	Retrospective single institution	PDT vs. oesophagectomy	199	129:70	91	65 endotherapy group 60 surgery group	HGD in Barrett's	5 years
Das 2008	Retrospective from SEER database	ER, thermal ablation, PDT or a combination vs. oesophagectomy	742	99:643	82	72 endotherapy group 68 surgery group	All histological types TisN0M0 (stage 0) or T1N0M0 (stage 1)	Mean follow‐up time in SEER database (23.9 (SE 1.8) for endoscopic treatment; 25.3 (SE 0.7) months for surgery
Reed 2005	Retrospective single institution	PDT or EMR vs. oesophagectomy	96	47:49	81	70 endotherapy group 59 surgery group	HGD in Barrett's	The median follow‐up for 48 patients who were operative survivors after resection of HGD was 56 months (range 1‐250 months). No follow‐up data were given for the endoscopy group
Schembre 2008	Retrospective single institution	EMR, APC, PDT or a combination vs. oesophagectomy	94	62:32	73% endotherapy group 93%  oesophagectomy group	70 endotherapy group 64 surgery group	HGD or IMC in Barrett's	Mean follow‐up for the endoscopic therapy was a median 20 months' with a range of 6 to 84 months. For the surgery group, the median follow‐up was 48 months with a range of 6 to 88 months
Yachimski 2008	Retrospective single institution	Endoscopic ablation vs. oesophagectomy	159	80:70	86	69.7 endotherapy group 63.1 surgery group	Barrett's oesophagus patients who had undergone oesophagectomy or endoscopic ablation. Prior to treatment, 70.4% of patients were diagnosed as HGD/IMC or T1sm, but 29.6% of participants were stage T2 or greater). Post‐oesophagectomy pathology was only reported in the participants with HGD or IMC. Disease was limited to the mucosa in 80% (24 of 30 participants). Efficacy was not reported	No long‐term outcomes reported (not the primary focus of the study)
Farrell 2011	Retrospective single institution	EMR + RFA (therapeutic and salvage) vs. oesophagectomy	96 in total, 73 reported in detail.	33:40	85 (ratio 5.4:1)	Median age 65 years (range 29 to 86 years)	Barrett's oesophagus with early malignant lesions, defined as HGD, IMC and submucosal T1 adenocarcinoma. No further details given in study report, as 1 patient had T1N1 disease, the study may have included patients with more or less advanced disease	Described as a 10‐year audit but length of follow‐up not clear (very short report)
Pech 2011	Retrospective data from 2 high‐volume centres	ER + APC vs. en bloc transthoracic oesophageal resection with 2‐field lymphadenectomy	114	76:38	96 endotherapy group 97 surgery group	Median age 62.2 endotherapy group 62.7 surgery group	Matched cases with respect to age, gender, infiltration depth (pT1m1‐3), differentiation grade (G1/2 vs. 3) and follow‐up period	4.1 years endotherapy group 3.7 years surgery group
Rosmolen 2010	Retrospective cross‐sectional review by questionnaire	Endotherapy vs. oesophagectomy	114	81:33 (66 endoscopy and 29 surgery responded)	83 endotherapy group 82 surgery group	64.1 (SD 8.2) endotherapy group 63 (SD 9.5) surgery group	HGD, early cancer (cancer limited to the mucosa [T1m] or superficial submucosa [T1sm1]), or both	12 to 60 months after treatment
Schembre 2010	Retrospective single institution questionnaire	Endotherapy (various) vs. oesophagectomy	77 (40 responded)	27:13	85 endotherapy group 93 surgery group	70 endotherapy group 63 surgery group	Barrett's oesophagus with HGD or IMC	Unclear, study examines cases from 1998 to 2007
Tian 2011	Retrospective single institution	EMR vs. oesophagectomy (2/3 transhiatal; 1/3 transthoracic)	68 (31 had oesophagectomy without CRT; 15 had EMR without CRT)	15:31	82.8 endotherapy group 92.3 surgery group	69.9 endotherapy group 67.1 surgery group	Submucosal oesophageal adenocarcinoma (T1b oesophageal adenocarcinomas)	5 years
Thomas 2005	Retrospective review of pathology database	APC vs. oesophagectomy (vs. surveillance)	15 (13 having APC or surgery)	5:8	100 endotherapy group 87.5 surgery group	70 endotherapy group 58 surgery group	HGD and medically unfit or chose not to have oesophagectomy in APC group. Persistent HGD on repeat biopsies done after a mean interval of 2 months in surgery group. Results reported for HGD groups (not adenocarcinoma) only	12 months APC, up to 60 months oesophagectomy
Zehetner 2011	Retrospective single institution	EMR vs. oesophagectomy (transthoracic en bloc, transhiatal, minimally invasive, or vagus sparing)	101	40:61	100 endotherapy group 70 surgery group	66 endotherapy group 68 surgery group	HGD or IMC	3 and 5 years

APC: argon plasma coagulation; CRT: chemoradiotherapy; EMR: endoscopic mucosal resection; ER: endoscopic resection; HGD: high‐grade dysplasia; IMC: intramucosal cancer; PDT: photodynamic therapy; SD: standard deviation; SE: standard error; SEER: Surveillance, Epidemiology, and End Results.

2. Primary outcome: survival.

Study	Survival measure	Survival in endoscopic treatment group	Survival in surgical group	Significantly different?
Pacifico 2003	Overall survival at end of follow‐up	92% at 12 months	97% at 19 months	No
Prasad 2007	Overall survival at 5 years	91% at 5 years	91.5% at 5 years	No
Das 2008	Median cancer‐free survival	56 months	59 months	No difference between the 2 groups (P = 0.41)
Reed 2005	Overall survival at 5 and 10 years	Not reported	83% at 5 years (disease specific 93%), 64% at 10 years	NA
Schembre 2008	Adjusted 4‐year survival	89% (0 due to oesophageal cancer)	93% (none due to oesophageal cancer)	No
Yachimski 2008	Not reported	Not reported	Not reported	"Vital status" was reported in Table 1 of this study, for the aggregated group of all patients, regardless of cancer stage (54% of patients were ≥ stage T2). There were no significant differences (P = 0.60) between the numbers of deceased participants in the oesophagectomy (10%) or ablation (13%) groups. The interval between treatment and death and cause of death were not reported
Farrell 2011	Not reported	Not reported	Not reported	Not reported
Pech 2011	Overall survival 1, 3 and 5 years	99% at 1 year 96% at 3 years 89% at 5 years	97% at 1 year 93% at 3 years 93% at 5 years	No differences between the 2 groups
Rosmolen 2010	Not reported	Not reported	Not reported	Not reported, not the focus of the study (QoL focus)
Schembre 2010	Not reported	Not reported	Not reported	Not reported
Tian 2011	Median overall survival duration	48.9 months	38.4 months	No (note: survival rates are for the whole group including those who had CRT in addition to EMR or surgery). In multivariate Cox modelling, age, co‐morbidities sex and oesophagectomy were not associated with survival duration.
Thomas 2005	Overall survival	Unclear (not the focus of the study)	62% at mean 18.5 months	Unclear, not reported
Zehetner 2011	Overall survival at 3 and 5 years	94% at 3 years	94% at 3 years	No

CRT: chemoradiotherapy; EMR: endoscopic mucosal resection; QoL: quality of life.

3. Secondary outcomes.

Study	Cancer‐free survival		Treatment failures		Procedure‐related mortality (30 day)		Complications		QoL
Treatment group	Endoscopic treatment	Surgical treatment	Endoscopic treatment	Surgical treatment	Endoscopic treatment	Surgical treatment	Endoscopic treatment	Surgical treatment	Endoscopic treatment	Surgical treatment
Pacifico 2003	83% (20/24) remained free of cancer at 12 ± 2 and 19 ± 3 months	All patients remained free of cancer at 12 ± 2 and 19 ± 3 months	Persistent cancer at first follow‐up 17% (4/24 patients). No later recurrence	None	None	2%	Stricture: 8% Photo‐ sensitivity: 8%	Stricture: 16% Anastomotic leak: 8% Wound infection: 8% Dumping syndrome: 5% Empyema (pus in the pleural cavity): 3% Blood transfusion: 3% Atrial fibrillation: 3% Aspiration: 2% Chylothorax (lymphatic effusion in the pleural cavity): 2%	Not reported
Prasad 2007	No significant difference between groups but tendency towards lower cancer‐free survival in endoscopic therapy group	HGD eliminated: at 1 year: 88%; at 3 years: 86% HGD detected within 12 months: 26% (33/129) i.e. failed initial therapy and retreated, HGD eradicated in 70% of these. Recurrence of HGD (detected after 12 months free of HGD): 7.8% (10/129) Cancer detected during follow‐up (up to 18 months after PDT): 6.2% (8/129)	None	None	1.4%	No early complications Stricture formation: 27% Photosensitivity: 60%	38% Postoperative morbidity Stricture formation: 12.6%	No QoL data, performance score only given ECOG PS?1 in 73% at end of FU* Proportions unchanged from pre‐therapy	No QoL data, performance score only given ECOG PS?1 in 94% at end of FU* Proportions unchanged from pre‐therapy
Das 2008	(Only oesophageal cancer‐specific cause of death considered) 56 months (50 to 61 months)	59 months (57 to 67 months) (no significant difference)	Not reported	Not reported	Not reported	Not reported
Reed 2005	Not reported	5/42 PDT and 2/5 EMR patients had recurrent HGD/cancer after mean 13.5 months. 1 developed metastatic spread but 0 died during follow‐up	3/49 patients died from recurrent disease	Not reported	2% (cerebrovascular accident in postoperative period)	Not reported	Anastomotic leaks: 4%	Not reported
Rosmolen 2010	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Significant bleeding: 3%, perforation: 2%, oesophageal stenosis: 31%	Symptomatic anastomotic leakage: 15 %, pulmonary problems: 30 %, serious infections: 15 % and vocal cord paralysis: 15 %. Anastomotic stenosis in 37% during follow‐up	QoL measured (SF‐36, EORTC‐QLQ‐C30, EORTC‐QLQ‐OES18 questionnaires). HADS, WOCS. Endoscopy patients reported more fear of recurrence on the WOCS than surgery patients (P = 0.003). No significant differences were found between the 2 groups on the other outcomes	Surgery patients had significantly more eating problems (OR 18.3; P < 0.001) and reflux symptoms (OR 3.4; P = 0.05) on the EORTC‐OES18 questionnaire
Schembre 2008	Not reported	Residual Barrett's segment in 44%, persistent dysplasia in 13%, progression to cancer in 6%	Residual Barrett's segment in 3%, persistent dysplasia in 3%, no progression to cancer	2%: 1 patient died from diverticulitis	None	Major: 8% Minor: 32%	Major: 13% Minor: 66% (significantly different)	Not reported	‐	‐
Yachimski 2008	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported
Farrell 2011	Not reported	Not reported	In patients treated by EMR + RFA, there was 100% disease control; 72% had complete eradication of Barrett's disease, and 1 patient represented with low‐grade dysplasia	4.8% surgical patients relapsed with HGD or cancer and 1 patient with T1N1 disease died of recurrent disease	Unclear from study report	Unclear from study report	2% mortality and 32% morbidity	2% morbidity from EMR + RFA (statistically significant compared with surgery P=0.001)	Not reported	Not reported
Pech 2011	98.7% at 4.1 years	100% at 3.7 years	6.6% (repeat endoscopic treatment was successful in all patients who experienced disease recurrence	No deaths from tumour‐related disease during follow‐up	None	Mortality at 40 days 2.6%	Minor complication in 17% (bleeding managed by clipping or injection). No major complications (severe bleeding, perforation stenosis) observed	32% major complications. including anastomotic leakage, pneumonia, multiple organ failure following sepsis, cardiac problems	Not reported	Not reported
Schembre 2010	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	SF‐36: no significant differences compared with surgery GIQLI: higher (better) scores among ASA Class 1 and 2 endotherapy patients compared with surgery (not statistically significant) and higher among young endotherapy patients than young surgery patients	SF‐36: no significant differences other than superior physical functioning in patients 65 years of age and older GIQLI: no significant differences in scores of older patients for surgery and endotherapy or SAS 3 patients. Negative QoL impact appears to be greater for younger patients undergoing surgery. (overall QoL scores: "QOL scores among EG and ET groups were not significantly different than sex age‐matched controls”
Tian 2011	Not reported	Not reported	In group who had EMR as only therapy, 21.4% had cancer recurrence at 45 months follow‐up (remaining 11 had mean cancer‐free periods of 21 months)	Not reported	None	1 death at 34 days owing to anastomotic leak (with 'do not resuscitate' order)	Not reported	Not reported	Not reported	Not reported
Thomas 2005	Not reported	Not reported	2/5 went on to develop metastatic adenocarcinoma at 9 and 12 months follow‐up	2/8 patients developed recurrent adenocarcinoma at 22 and 60 months after surgery, 1 of whom died	None	None	Postoperative morbidity in 4 patients (pneumonia, adult respiratory distress syndrome, pneumothorax) 2/8 developed postoperative strictures at 3 to 6 months' postsurgery requiring ongoing dilation	Not reported	Not reported	Not reported
Zehetner 2011	100% at 3 years; 100% at 5 years	100% at 3 years; 88% at 5 years	20% had new or metachronous cancer	None	None	None	None reported	39% had complications including long‐term complications of anastomotic strictures, ventral hernias, reflux, postprandial dumping or diarrhoea	Not reported	Not reported

*ECOG PS 0 = fully active, able to carry on all pre‐disease performance without restriction.

ECOG PS 1 = restricted in physically strenuous activity but ambulatory and able to carry out work of a light or sedentary nature, e.g., light house work, office work.

ASA: American Society of Anesthesia; EMR: endoscopic mucosal resection; EORTC‐QLQ: European Organization for Research and Treatment of Cancer Quality of Life Questionnaire; FU: fluorouracil; GIQLI: Gastrointestinal Quality of Life Index; HADS: Hospital Anxiety and Depression Scale; OR: odds ratio; PDT: photodynamic therapy; QoL: quality of life; RFA: radiofrequency ablation; SF‐36: 36‐item Short Form; WOCS: Worry Of Cancer Scale.

Discussion

There are currently no RCTs comparing surgery with endoscopic therapy for early cancer and HGD in Barrett's oesophagus. This discussion is based on available evidence derived from reports of non‐randomised studies. It is difficult to draw definitive conclusions from the data obtained from non‐RCTs owing to selection bias.

Updated searches will be carried out on a regular basis and if RCTs or potential RCTs become available they will be assessed and, where appropriate, incorporated into an updated review. If further details are required to assess the trials, for example method of randomisation, the study investigators will be contacted to obtain further details.

Summary of main results

We excluded 13 published studies that compared surgery with endotherapy in early cancer and HGD in Barrett's oesophagus (Das 2008; Farrell 2011; Pacifico 2003; Pech 2011; Prasad 2007; Reed 2005; Rosmolen 2010; Schembre 2008; Schembre 2010; Thomas 2005; Tian 2011; Yachimski 2008; Zehetner 2011) as all were retrospective and non‐randomised. All of these excluded studies are likely to be biased as endoscopic therapy has, until very recently, been offered largely to those considered unfit for surgery owing to age and frailty, hence the decision to treat by endoscopic methods was generally not based on the stage of the disease. In all of the excluded studies, patients in the endotherapy groups were older than in the surgical groups, and had shorter Barrett's segments or less invasive disease, or both. In Schembre 2008, the study investigators comment on this, stating that in their study, participants in the endotherapy group were older but "not necessarily sicker", and that "endotherapy was initially reserved for poor surgical candidates and only recently have younger healthier patients been referred for consideration of non‐surgical treatment". Other factors may influence the treatment pathway. In a review of cases between 2003 and 2007, Yachimski 2008 concluded that patients with Barrett's HGD or IMC, who were aged 60 years or under, and who had a cancer stage of T1sm or greater were more likely to be treated by oesophagectomy. Even when the age and cancer stage were controlled for, patients were more likely to be treated by oesophagectomy if the initial consultation was performed by a surgeon.

For healthcare providers such as the National Health Service (NHS) in the UK, there are considerations in terms of the organisation of care for patients who have high‐grade oesophageal dysplasia such as implementation of new technologies and the need for repeat observations and treatments. In 2012 the BAD CAT (Barrett's and Dysplasia Cancer Action Taskforce) consensus group created evidence‐based consensus statements for the management of patients with Barrett's oesophagus and early‐stage oesophageal adenocarcinoma. This approach identified important clinical features of HGD in Barrett's oesophagus and areas for future studies where management could be made more consistent or improved (Bennett 2012).

In this review, with the limited data available and lack of published RCTs comparing endoscopic treatment with oesophagectomy, it is impossible to reach firm comparative conclusions.

Overall completeness and applicability of evidence

We were unable to find any RCTs of direct comparisons of surgery versus endotherapies for the treatment of early cancer and HGD in Barrett's oesophagus. Therefore the studies that we identified and excluded are not sufficient to address all of the objectives of the review. However the results of other studies that address risk of progression, effects of treatment in terms of cure, progression and complications are available from non‐randomised studies and we have attempted to summarise some of these data here, with reference to the studies that we excluded.

Risk of progression from Barrett's oesophagus to cancer

Treating all patients with Barrett's oesophagus is currently unfeasible as it occurs in at least 0.5% of the UK population (and is not supported by number‐needed‐to‐treat analyses) (Cameron 1990). Progression from non‐dysplastic Barrett's oesophagus to malignancy is rare. Large studies from Northern Ireland and Denmark suggest a malignant progression rate of 0.12% to 0.39% of patients per year (Bhat 2011; Hvid‐Jensen 2011; Yousef 2008), although these may be underestimates as the studies defined Barrett's histologically rather than endoscopically and are therefore likely to have included a number of patients without true Barrett's. The most common type of metaplastic epithelium, specialised intestinal metaplasia (intestinalisation), is considered to have a major potential to develop into adenocarcinoma. Some studies have suggested that genetic changes play a key role in this progression to cancer. Adenocarcinoma can develop in Barrett's oesophagus with these genetic changes even in the absence of intestinalisation (Kelty 2007; Liu 2009). It is hoped that research in progress will identify appropriate strategies for risk assessment, treatment and surveillance in this population (Konda 2012).

There is marked variation in the reported malignant potential of LGD. Skacel 2000 followed 25 patients with LGD for 26 months and found that 8% had persistent LGD, 60% regressed and 28% went on to develop HGD or adenocarcinoma (4% were indefinite for dysplasia at follow‐up). Sharma 2006 followed 156 patients with LGD for a mean of 4.1 years and reported stable LGD in 21%, regression in 66% and progression to HGD or cancer in 13%. These data support an earlier audit of a Barrett's epithelium surveillance database, where one in 10 cases of LGD progressed to cancer over an observation period of five years (Basu 2004). However, other studies have shown progression rates below 5% (Wani 2009). In one prospective cohort study of 713 patients with Barrett's oesophagus, including 111 with LGD, which reported that, compared to non‐dysplastic disease, LGD was a significant risk factor for progression to HGD or adenocarcinoma (risk ratio (RR) 9.7; 95% confidence interval (CI) 4.4 to 21.5) (Sikkema 2010).

This variation in the reported malignant potential of LGD partly reflects difficulties in establishing an accurate histological diagnosis. It has been demonstrated that non‐dysplastic Barrett's oesophagus is frequently over‐diagnosed as LGD by non‐specialist pathologists (Curvers 2010). The study authors concluded that the true significance of LGD may be underestimated by the majority of studies that have relied on single pathologist reporting.

In this review, we chose not to address LGD in Barrett's oesophagus. We included both HGD and IMC (mucosal neoplasia) in this review, owing to the suboptimal techniques available to differentiate between them. In addition, their biological behaviour is similar, as both have a low risk of nodal spread.

Literature suggests that, in those with biopsy‐confirmed HGD, the rate of submucosally invasive cancer is 12.7% in those with endoscopically visible HGD and 3% for those with no detectable lesion (Konda 2008). There is variability in the reported risk of lymph node metastases in patients with intramucosal and submucosal (T1) oesophageal adenocarcinoma, which may be accounted for by differences in depth of invasion, size, presence of lympho‐vascular invasion and tumour grade. One retrospective review of 126 T1a and T1b tumours reported that although lymph node metastases were rare (1.3%) with intramucosal tumours, submucosal invasion of any extent carried a significant risk of lymph node metastases (and is thus unsuitable for endoscopic treatment) (Leers 2011). This has not been the experience of others who have concluded that superficial submucosal invasion is low risk for metastases (Alvarez Herrero 2010). Careful histopathology of ER specimens should influence the decision whether to treat endoscopically or by oesophagectomy (Peters 2008). Mortality rates for oesophagectomy may exceed 2%, hence a small risk of lymph node metastases may be acceptable after endoscopic therapy.

Surgery or endotherapy?

If visible lesions suspicious of dysplasia are found in the setting of Barrett's oesophagus, ER should be performed in order to obtain an accurate histological diagnosis. A variety of options including resection or ablative therapies, followed by adequate surveillance, are available to treat the epithelium.

Surgical therapies, although effective, are costly and carry appreciable patient mortality and morbidity. Ablative endotherapies allow restoration of the normal squamous lining and thereby remove both existing dysplasia and the potential for development of dysplasia (Sampliner 1996). In patients with HGD, endotherapies may offer an alternative to oesophagectomy with comparable survival, lower mortality and morbidity (Pacifico 2003, Schembre 2008), and less disruption of QoL (Rosmolen 2010; Schembre 2010), especially since patients with Barrett's oesophagus have considerable co‐morbidity (Moayyedi 2008).

Post‐oesophagectomy mortality and morbidity are important issues; DeMeester 2008 suggested that quoted high postoperative mortality rates were not supported by more recent studies of patients with HGD or early cancer who were treated by oesophagectomy, although there is a volume to performance relationship (Chang 2006). In one retrospective case series, vagal‐sparing oesophagectomy was associated with less peri‐operative morbidity, a shorter hospital stay and fewer late complications such as weight loss, diarrhoea and dumping, than other surgical techniques (Peyre 2007). In that study, postoperative mortality was similar for all types of surgical resection (30‐day mortality percentage; vagal‐sparing 2%, transhiatal 5%; en‐bloc 0%) and survival was reported to be excellent (percentage survival over 60 months, in 85 patients diagnosed with cancer, all‐cause mortality: vagal sparing 89%; transhiatal 65%; en bloc 95%). While oesophagectomy can potentially offer the chance of complete resection of diseased tissue in patients with HGD or early (intramucosal) adenocarcinoma, recurrent Barrett's metaplasia or neoplasia may still appear after resection with curative intent (Wolfson 2004).

While different surgical techniques used for oesophagectomy offer broadly similar results, the endotherapy arm of many of the reports reviewed here were found to be much more heterogeneous. Published studies do consistently report endoscopic therapy to be safe and effective. For example in Shaheen 2009, a multicentre, sham‐controlled randomised trial, participants received either RFA or a sham procedure. In patients with dysplastic Barrett's oesophagus, RFA was associated with a high rate of complete eradication of both dysplasia and intestinal metaplasia and a reduced risk of disease progression, and few serious adverse effects. Medium‐term follow‐up data have demonstrated that these effects are durable at three years (Shaheen 2011).

While we found no RCTs that directly compared the results of surgery versus endotherapies, Attwood 2003 followed 29 patients with Barrett's‐related HGD who were deemed unfit for or declined surgery, instead undergoing argon beam photocoagulation. He showed that survival over seven years was not different to the general age‐ and sex‐matched population. This would suggest that, in those patients unable or unwilling to undergo oesophagectomy, endoscopic therapy offers a significant survival benefit over no therapy at all. Greenstein 2008 used the Surveillance, Epidemiology, and End Results (SEER) cancer registry data to compare survival in patients having either endoscopic therapy or no therapy. There was a significantly better oesophageal cancer‐related survival in the endoscopically treated group. It cannot be assumed that all of this group were deemed unfit for surgery, but non‐cancer‐related survival curves were similar between the groups suggesting similar co‐morbidities.

Schembre 2008 concluded that "both endotherapy and esophagectomy can effectively treat high‐grade dysplasia and intramucosal carcinoma associated with Barrett's esophagus". Endotherapy is associated with a higher risk of tumor progression, although this is uncommon. Esophagectomy incurs higher initial costs and results in more frequent minor complications but is usually curative. In this study, exclusion of the 'unfit for surgery' participants resulted in only 3% developing cancer (5% HGD) at the latest follow up (with no major complications and 27% minor complications), significantly fewer compared with the surgical group. One retrospective matched cohort comparison of oesophagectomy versus ER and APC in IMC reported effective treatment outcomes for both strategies after four years of follow‐up (Pech 2011). Complete remission of Barrett's was achieved in all patients in the surgical group and 98.7% in the ER group. 90 day mortality was 2.6% following surgery and 0% following endotherapy and major complications were only observed in the surgical group (32%). No tumour‐related mortality was observed in either group demonstrating that both strategies provided effective treatment in the medium term.

Adverse effects and treatment failure

Complications may arise as a result of endotherapy and there is the possibility of treatment failure as local endoscopic therapies can leave much of the oesophagus intact and often untreated.

A review of 12 publications on ER for early Barrett's neoplasia (Pech 2007) showed recurrence rates of 0% to 21%, minor complications in 0% to 46% and few major complications. Ell 2007 published single centre results of ER in 100 consecutive patients with early Barrett's adenocarcinoma (total 144 resections) showing no major complications, 11 minor complications (all bleeds without decrease in haemoglobin (Hb) > 2 g/dL) and local remission in 99% of patients after 1.9 months with a five‐year survival rate of 98% (both of the patient deaths were from unrelated causes).

One retrospective matched cohort comparison of oesophagectomy versus ER and APC in T1 cancer reported cancer‐free remission for both strategies after four years of follow‐up (Pech 2011). Complete remission of Barrett's was achieved in all patients in the surgical group and 98.7% in the ER group. Ninety‐day mortality was 2.6% following surgery and 0% following endotherapy and major complications were only observed in the surgical group (32%). No tumour‐related mortality was observed in either group demonstrating that both strategies provided effective treatment in the medium term.

When focused local therapy is used to remove an HGD/early cancer lesion within a Barrett's segment, there often remains a significant amount of residual Barrett's. After endotherapy, the remaining Barrett's has a high malignant potential and small patches of early dysplasia can easily be missed at initial treatments. In patients who have already developed an area of HGD/early cancer, it is likely that the residual Barrett's has a greater malignant potential (field change effect) than in those who have never developed a dysplastic lesion. Many studies have therefore used a combination of local therapies, for example EMR to remove the HGD/early cancer lesions and wide‐field ablation modalities (most commonly PDT or RFA) to treat the residual Barrett's in an attempt to reduce the rate of metachronous lesions found at follow‐up.

One randomised study of HGD in Barrett's by Overholt 2007 compared PDT plus omeprazole to omeprazole alone and demonstrated that at five years those in the PDT group were less likely to have detectable HGD (77% versus 39%) or cancer (29% versus 15%). Photosensitivity reactions and oesophageal strictures were reported as adverse events, but no long‐term effects of oesophageal strictures or photosensitivity reactions were seen by the end of the study. There is some concern that PDT may simply 'bury' the pre‐malignant tissue under neo‐squamous epithelium (Mino‐Kenudson 2007), although these areas seem to have a low malignant potential (Biddlestone 1998; Hornick 2007) and may simply atrophy over time (Hornick 2005). PDT also has a relatively high rate of complications (e.g. strictures and photosensitisation). One retrospective review of 50 patients treated by PDT for HGD or early oesophageal cancer showed there was no procedure‐related mortality, but 42% of the patients developed oesophageal stricture (Keeley 2007). Bronner 2009 reported that squamous overgrowth following PDT did not obscure the most advanced neoplasia in any patient.

Recurrence or metachronous lesions occur relatively frequently but repeat local therapy is usually possible and successful. Schnell 2001 demonstrated that the risk of cancer developing from HGD after the first year was 16% in a group of 75 people followed up for a mean of seven years (managing these patients with a strategy of frequent endoscopic surveillance with biopsies). May 2002 showed that in a group of 115 patients with dysplastic areas, most of whom had early cancer (96 early cancer, 19 HGD), 30% developed a metachronous cancer after undergoing local therapy, usually EMR alone (EMR alone in 70, PDT alone in 32, EMR and PDT in 10 and APC alone in three) over a mean follow‐up period of 34 months. A study by Pech 2006 including 304 patients undergoing ER, PDT, or both, followed for a mean of 69.5 months reported tumour‐related deaths in two inoperable patients. Complications in this study included bleeding in 11% and stricture formation in 3.3%, all of which were managed endoscopically. Pech 2008 reported long‐term results from a cohort of patients with Barrett's neoplasia, who were treated endoscopically. Three hundred and forty‐nine patients were followed up for a mean of 63.6 months. A complete response was seen in 96.6% and surgery was necessary in 3.7% after endoscopic therapy failed. Metachronous lesions developed during the follow‐up in 74 patients (21.5%); 56 died of concomitant disease, but none died of Barrett's neoplasia. The calculated five‐year survival rate was 84%. The reason for the high rate of recurrence seemed to be a result of the undetected neoplasia in the Barrett's segment after endoscopic therapy. Moss 2010, in Australia, demonstrated the effects of ER on histological grade and stage of adenocarcinoma, in a prospective study at two centres (N = 75) using ER for HGD or oesophageal adenocarcinoma, there was no recurrence at resected sites, 11% developed metachronous lesions and five developed submucosal invasion treated by oesophagectomy. There were no deaths due to adenocarcinoma. The study authors concluded that ER altered histological grade or local T stage and resulted in reduced oesophagectomy rates.

There are promising results from using either circumferential ER (Giovannini 2004; Peters 2006) or RFA (Gondrie 2008a; Gondrie 2008b; Sharma 2008) to reduce the future malignant potential of the residual Barrett's safely. In terms of eradication of Barrett's oesophagus tissue, RFA seems to be able to remove benign Barrett's oesophagus completely without any major complications (Sharma 2007) but the results of further studies, ideally RCTs, are awaited (Rösch 2008; Sharma 2009; Singh 2009). After endotherapy, patients must continue to attend for surveillance endoscopies and of note 13% of the endotherapy group in Schembre 2008 failed to attend the follow‐up appointments upon which success may depend.

Quality of life

The assessment of QoL in clinical trials does not have a universally accepted definition, although Efficace 2003 suggested that a number of standard criteria could be employed. Ideally the measures should be multi‐dimensional and self‐reported (Blazeby 2007). Both treatment modalities appeared to have a similar impact on QoL one year or more after treatment compared with age‐matched controls and younger patients who had oesophagectomy had a greater impairment in QoL than older patients who had surgery (Schembre 2010). Prasad 2007 made some attempt to look at QoL retrospectively by noting the Eastern Cooperative Oncology Group performance scores (ECOG PS) pre‐treatment and at the end of follow‐up. The proportion of patients with an ECOG PS of one or less 1 (equivalent to being symptomatic but completely ambulatory or being asymptomatic) in the endoscopic therapy/surgical groups was similar before and after treatment.

Rosmolen 2010 explored QoL and fear of cancer recurrence in 81 patients over a follow‐up period of 12 to 60 months after endoscopic or surgical treatment for early Barrett's neoplasia using a non‐randomised cross‐sectional design. There were more eating problems and reflux symptoms in the surgery group but those treated endoscopically were reported to have more fear of recurrence surgery patients (statistically significant). The study investigators commented that strategies to educate and address fear of cancer recurrence may be useful. Future studies would ideally report on QoL as well as survival and incidence of recurrent or new dysplasia and cancer.

Cost

Schembre 2008 reported that hospital charges were USD40,079 in the endotherapy group as compared with USD66,060 in the surgical group over a mean follow‐up period of 20 and 48 months, respectively. However, the surgical group was unlikely to generate further costs whereas the endoscopic therapy group would require long‐term surveillance and possibly further treatments. A study by Hur 2003 showed that, for these reasons, PDT‐treated patients would generate higher costs than those undergoing oesophagectomy in the long term. A subsequent detailed cost‐effectiveness analysis was undertaken in Shaheen 2004 using decision analysis modelling to compare endoscopic surveillance, endoscopic ablation therapies and oesophagectomy in HGD in Barrett's oesophagus. They found endoscopic ablation therapy to be both cheaper and more effective (in terms of discounted quality‐adjusted life years) than oesophagectomy, even when the model was weighted towards oesophagectomy by assuming all surgery was undertaken in high volume centres and all endoscopic ablation patients required costly long‐term endoscopic surveillance post‐treatment. Endoscopic surveillance was the least expensive option but resulted in almost 20% of the patients developing cancer. Oesophagectomy provided the lowest risk of developing cancer but only when yearly rates of progression of HGD to cancer exceeded an unrealistic 30% did it become the most cost‐effective option. Vij 2004 used a similar decision analysis model to compare costs of PDT and oesophagectomy and concluded that PDT followed by endoscopic surveillance incurred greater life‐time cost but was more cost‐effective in terms of quality‐adjusted life years. In a hypothetical cohort of male patients with Barrett's oesophagus and HGD, PDT and oesophagectomy or surveillance were examined. Surveillance was the least costly approach, but the least effective. The report concluded that PDT and oesophagectomy were cost‐effective alternatives to surveillance only (Comay 2007). This is further supported by a decision analysis model by Pohl 2009. This mathematical modelling study concluded that although only limited long‐term outcomes of endotherapies are available, endoscopic therapy for early Barrett's oesophagus adenocarcinoma is more effective and costs less than surgical therapy.

A cost‐utility analysis using a decision analysis model based on a hypothetical cohort of 50‐year old patients with a diagnosis of Barrett's oesophagus, endoscopic ablation of HGD in subjects with Barrett's oesophagus was found to be "probably cost effective" in that it was estimated to be less costly and more effective than oesophagectomy or surveillance alone. Ablation of HGD was found to be capable of extending life expectancy by more than three quality‐adjusted life years at an incremental cost of less than USD6000 compared with no intervention. While the costs of different ablation types varied, they seem to be similar in terms of effectiveness. The authors noted that "as further post ablation data become available, the optimal strategy to be implemented will be clarified" (Inadomi 2009).

One UK‐based cost‐utility analysis of RFA or oesophagectomy concluded that a policy of RFA with oesophagectomy for HGD recurrence was a cost‐effective strategy when compared to immediate oesophagectomy (Boger 2010). RFA generated an additional 0.4 quality‐adjusted life years at a cost saving of GBP1902 compared to oesophagectomy. Boger 2010 concluded that only with an RFA failure rate (progression or persistence of HGD) of more than 44% or an annual risk of HGD recurrence of more than 15% following ablation would immediate oesophagectomy become the more cost‐effective option in the management of HGD. Similarly, one systematic review of the cost‐effectiveness of PDT in HGD identified three out of four studies where PDT ablation was found to be cost‐effective compared to oesophagectomy or endoscopic monitoring, although reported cost‐effectiveness ratios varied widely between each study (Sanchez 2010).

Potential biases in the review process

None known.

Agreements and disagreements with other studies or reviews

With the limited data available and lack of published RCTs comparing endoscopic treatment with oesophagectomy, it is impossible to reach firm comparative conclusions. This observation is consistent with that of a systematic review comparing safety and effectiveness of endotherapy compared to oesophagectomy in the treatment of HGD and early cancer (Menon 2010). Endotherapy was shown to be safer in terms of mortality (0.04% versus 1.2%) although high stricture rates were identified as a significant problem following PDT ablation (18.5% with PDT compared to 5.3% with oesophagectomy). The absence of randomised trial evidence meant that selection and publication bias were significant. In addition, the studies included were predominantly conducted prior to the widespread uptake of RFA.

Authors' conclusions

Implications for practice.

No data were available to permit an analysis of the relative benefits or harms of endotherapies compared with surgery (oesophagectomy). Current use of endotherapies in the care of patients with early cancer or HGD of Barrett's oesophagus should be at the recommendation of the multi‐disciplinary team involved in individual patient care. Properly conducted, RCTs comparing surgery with endotherapies should be carried out before any conclusions can be drawn.

Implications for research.

Collaborative work between patients who make informed decisions to participate in RCTs, carers, clinicians and researchers in order to produce large clinically relevant data is paramount, to ensure outcomes are relevant and participation is maximised. Future RCTs, comparing oesophagectomy with endotherapies are required. Attention should be paid to:

• collection of detailed demographic/baseline data, fitness for surgery, histologically confirmed stage of cancer or HGD of oesophagus (good internal validity);

• employing randomised allocation to treatment groups;

• a comparison of surgery versus endotherapies in comparable participant groups;

• adequate power (employing perhaps a multicentre design);

• relevant outcomes, including long‐term follow‐up, adverse effects (researchers should also collect and report reasons for drop‐out); reporting long‐term outcomes compared in an RCT for surgery versus endotherapies for treatment of oesophageal HGD. These data would add to the evidence base to assist in the planning and implementation of cancer care services by healthcare providers;

• cost analyses; and

• endoscopic and surgical interventions standardised throughout sites (internal validity).

The results of such studies should be incorporated into evidence‐based practice guidelines.

What's new

Date	Event	Description
11 May 2020	Amended	Typo in abstract corrected, from "treatment of or early cancer" to “treatment of high‐grade dysplasia or early cancer”, results and conclusions not changed
11 May 2020	New citation required but conclusions have not changed	Typo corrected in abstract, conclusions not changed.

History

Protocol first published: Issue 3, 2008
Review first published: Issue 2, 2009

Date	Event	Description
30 June 2012	New citation required but conclusions have not changed	Updated, new author, no change to conclusions.
17 May 2012	New search has been performed	Updated, no new included studies found. New excluded studies found. New author added and order of authors revised.
5 October 2010	Amended	Contact details updated.
17 March 2010	New citation required but conclusions have not changed	Author line changed, no new randomised controlled trials identified, conclusions not changed.
18 January 2010	New search has been performed	Searches updated in November 2009 including LILACS search which was revised.
8 July 2008	Amended	Typographical error fixed (protocol)

Notes

We revised the LILACS search strategy to use DeCS (Descriptors in Health Sciences) as it is trilingual (English, Portuguese and Spanish) vocabulary coding. We also simplified the LILACS search strategy in hope of identifying more relevant citations. Because this was a change from the search strategy published in the protocol of this review, in 2009, we re‐searched LILACS from January 1 2000 to November 2009. No new RCTs were found. The updated searches in 2012 also used DeCS for LILACS.

Appendices

Appendix 1. Cochrane Highly Sensitive Search Strategy

Cochrane Highly Sensitive Search Strategy, sensitivity maximising version; Ovid format.

1. randomized controlled trial.pt.
2 controlled clinical trial.pt.
3 randomized.ab.
4 placebo.ab.
5 drug therapy.fs.
6 randomly.ab.
7 trial.ab.
8 groups.ab.
9 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8
10 humans.sh.
11 9 and 10

Appendix 2. MEDLINE search strategy

Ovid MEDLINE(R) 1996 to November week 1 2009
1. randomized controlled trial.pt.
2. controlled clinical trial.pt.
3. randomized.ab.
4. placebo.ab.
5. drug therapy.fs.
6. randomly.ab.
7. trial.ab.
8. groups.ab.
9. or/1‐8
10. (animals not (humans and animals)).sh.
11. 9 not 10
12. exp barrett esophagus/
13. (barret$ adj3 esophagus).tw.
14. (barret$ adj3 oesophagus).tw.
15. (metaplas$ adj5 epitheli$).tw.
16. (columnar adj5 line$).tw.
17. (columnar adj5 metaplas$).tw.
18. (intest$ adj5 metaplas$).tw.
19. (intest$ adj5 dysplas$).tw.
20. or/12‐19
21. exp esophagectomy/
22. esophagectomy.tw.
23. (esophag$ adj5 neoplas$).tw.
24. (oesophag$ adj5 neoplas$).tw.
25. (esophag$ adj5 cancer$).tw.
26. (oesophag$ adj5 cancer$).tw.
27. (esophag$ adj5 carcin$).tw.
28. (oesophag$ adj5 carcin$).tw.
29. (esophag$ adj5 tumo$).tw.
30. (oesophag$ adj5 tumo$).tw.
31. (oesophag$ adj5 metasta$).tw.
32. (esophag$ adj5 metasta$).tw.
33. (esophag$ adj5 malig$).tw.
34. (oesophag$ adj5 malig$).tw.
35. (adenocarcinoma$ adj5 esophag$).tw.
36. (transthoracic adj10 esophag$).tw.
37. (transthoracic adj10 oesophag$).tw.
38. (thoracic adj10 resect$).tw.
39. (ivor adj lewis).tw.
40. (tanner adj25 esophag$).tw.
41. (tanner adj25 oesophag$).tw.
42. (santy adj25 esophag$).tw.
43. (santy adj25 oesophag$).tw.
44. (transhiatal adj10 esophag$).tw.
45. (orringer adj25 esophag$).tw.
46. (orringer adj25 oesophag$).tw.
47. surgery.tw.
48. exp surgery/
49. (surgical adj treatment).tw.
50. operative.tw.
51. merendino.tw.
52. (limited adj resection$).tw.
53. (vagal sparing adj oesophagectomy).tw.
54. (vagal sparing adj esophagectomy).tw.
55. exp cryotherapy/
56. cryotherapy.tw.
57. (cryothera$ adj5 endoscop$).tw.
58. exp lasers/
59. exp laser coagulation/
60. exp light coagulation/
61. exp catheter ablation/
62. (argon adj5 plasma adj5 coagulat$).tw.
63. APC.tw.
64. exp sclerotherapy/
65. sclerotherap$.tw.
66. exp electrocoagulation/
67. electrocoagulat$.tw.
68. (multipolar adj5 coagulat$).tw.
69. (therm$ adj5 coagulat$).tw.
70. (heater adj5 probe).tw.
71. (argon$ adj5 laser$).tw.
72. (YAG adj5 laser$).tw.
73. (yag adj5 nd).tw.
74. (yag adj5 ktp).tw.
75. (monopolar adj5 coagulat$).tw.
76. (bipolar adj5 coagulat$).tw.
77. (multipolar adj5 coagulat$).tw.
78. mpec.tw.
79. exp photochemotherapy/
80. (photodynamic adj5 therap$).tw.
81. PDT.tw.
82. (ala adj5 pdt).tw.
83. (aminolaevulin$ adj5 acid).tw.
84. (photophrin adj5 pdt).tw.
85. (endoscopic adj5 mucosal adj5 resect$).tw.
86. (ultrasonic adj3 surgical adj3 aspirat$).tw.
87. exp fundoplication/
88. fundoplicat$.tw.
89. (antireflux adj5 surgery).tw.
90. (nissen or rossetti).tw.
91. (toupet or lind or watson or besley).tw.
92. (partial$ adj5 fundoplication$).tw.
93. (laparoscop$ adj5 fundoplication$).tw.
94. exp catheter ablation/
95. (monopolar adj5 coagulat$).tw.
96. (endoscopic adj5 therapy).tw.
97. (endoscopic adj5 submuscosal adj5 dissect$).tw.
98. (endoscopic adj5 resect$).tw.
99. ER.tw.
100. (endoscopic adj5 removal).tw.
101. (endoscopic adj5 ablation).tw.
102. (multipolar adj5 electrocaut$).tw.
103. (laser adj5 therapy).tw.
104. (radio adj3 frequency adj3 ablation).tw.
105. RFA.tw.
106. ablation.tw.
107. (barrx adj3 system).tw.
108. (early adj cancer).tw.
109. (early adj neoplas$).tw.
110. (high adj grade adj dysplas$).tw.
111. HGD.tw.
112. precancer$.tw.
113. (pre adj cancer$).tw.
114. premalignant.tw.
115. (pre adj malignant).tw.
116. (invasive adj high adj grade adj neoplas$).tw.
117. (high adj grade adj intraepithelial adj metaplas$).tw.
118. (vienna adj2 classification).tw.
119. (indefinite adj2 neoplas$).tw.
120. (indefinite adj2 dysplas$).tw.
121. (non adj2 invasive adj2 high adj grade adj2 neoplas$).tw.
122. (non adj2 invasive adj2 low adj grade adj2 neoplas$).tw.
123. or/21‐122
124. 11 and 20 and 123
125. limit 124 to ed=20080101‐20091108

Appendix 3. EMBASE

EMBASE 1996 to 2009 week 45
1. exp randomized controlled trial/
2. randomized controlled trial$.tw.
3. exp randomization/
4. exp single blind method/
5. exp double blind method/
6. or/1‐5
7. animal.hw.
8. human.hw.
9. 7 not (7 and 8)
10. 6 not 9
11. exp clinical trial/
12. (clin$ adj3 (stud$ or trial$)).ti,ab,tw.
13. (clin$ adj3 trial$).ti,ab,tw.
14. ((singl$ or doubl$ or treb$ or tripl$) adj3 (blind$ or mask$)).ti,ab,tw.
15. exp placebo/
16. placebo$.ti,ab,tw.
17. random.ti,ab,tw.
18. (crossover$ or cross‐over$).ti,ab,tw.
19. or/11‐18
20. 19 not 9
21. 20 not 10
22. exp comparative study/
23. exp evaluation studies/
24. exp prospective studies/
25. exp controlled study/
26. (control$ or prospective$ or volunteer$).ti,ab,tw.
27. or/22‐26
28. 27 not 9
29. exp barrett esophagus/
30. (barret$ adj3 esophagus).tw.
31. (barret$ adj3 oesophagus).tw.
32. (metaplas$ adj5 epitheli$).tw.
33. (columnar adj5 line$).tw.
34. (columnar adj5 metaplas$).tw.
35. (intest$ adj5 metaplas$).tw.
36. (intest$ adj5 dysplas$).tw.
37. or/29‐36
38. exp esophagus resection/
39. (esophag$ adj5 resect$).tw.
40. (oesophag$ adj5 resect$).tw.
41. esophagectomy.tw.
42. (esophag$ adj5 neoplas$).tw.
43. (oesophag$ adj5 neoplas$).tw.
44. (esophag$ adj5 cancer$).tw.
45. (oesophag$ adj5 cancer$).tw.
46. (esophag$ adj5 carcin$).tw.
47. (oesophag$ adj5 carcin$).tw.
48. (esophag$ adj5 tumo$).tw.
49. (oesophag$ adj5 tumo$).tw.
50. (oesophag$ adj5 metasta$).tw.
51. (esophag$ adj5 metasta$).tw.
52. (esophag$ adj5 malig$).tw.
53. (oesophag$ adj5 malig$).tw.
54. (adenocarcinoma$ adj5 esophag$).tw.
55. (transthoracic adj10 esophag$).tw.
56. (transthoracic adj10 oesophag$).tw.
57. (thoracic adj10 resect$).tw.
58. (ivor adj lewis).tw.
59. (tanner adj25 esophag$).tw.
60. (tanner adj25 oesophag$).tw.
61. (santy adj25 esophag$).tw.
62. (santy adj25 oesophag$).tw.
63. (transhiatal adj10 esophag$).tw.
64. (orringer adj25 esophag$).tw.
65. (orringer adj25 oesophag$).tw.
66. surgery.tw.
67. exp surgery/
68. (surgical adj treatment).tw.
69. operative.tw.
70. merendino.tw.
71. (limited adj resection$).tw.
72. (vagal sparing adj oesophagectomy).tw.
73. (vagal sparing adj esophagectomy).tw.
74. exp cryotherapy/
75. cryotherapy.tw.
76. (cryothera$ adj5 endoscop$).tw.
77. exp laser/
78. exp laser coagulation/
79. exp light coagulation/
80. exp catheter ablation/
81. (argon adj5 plasma adj5 coagulat$).tw.
82. APC.tw.
83. exp sclerotherapy/
84. sclerotherap$.tw.
85. exp electrocoagulation/
86. electrocoagulat$.tw.
87. (multipolar adj5 coagulat$).tw.
88. (therm$ adj5 coagulat$).tw.
89. (heater adj5 probe).tw.
90. (argon$ adj5 laser$).tw.
91. (YAG adj5 laser$).tw.
92. (yag adj5 nd).tw.
93. (yag adj5 ktp).tw.
94. (monopolar adj5 coagulat$).tw.
95. (bipolar adj5 coagulat$).tw.
96. (multipolar adj5 coagulat$).tw.
97. mpec.tw.
98. exp photochemotherapy/
99. (photodynamic adj5 therap$).tw.
100. PDT.tw.
101. (ala adj5 pdt).tw.
102. (aminolaevulin$ adj5 acid).tw.
103. (photophrin adj5 pdt).tw.
104. (endoscopic adj5 mucosal adj5 resect$).tw.
105. (ultrasonic adj3 surgical adj3 aspirat$).tw.
106. exp fundoplication/
107. fundoplicat$.tw.
108. (antireflux adj5 surgery).tw.
109. (nissen or rossetti).tw.
110. (toupet or lind or watson or besley).tw.
111. (partial$ adj5 fundoplication$).tw.
112. (laparoscop$ adj5 fundoplication$).tw.
113. exp catheter ablation/
114. (monopolar adj5 coagulat$).tw.
115. (endoscopic adj5 therapy).tw.
116. (endoscopic adj5 submuscosal adj5 dissect$).tw.
117. (endoscopic adj5 resect$).tw.
118. ER.tw.
119. (endoscopic adj5 removal).tw.
120. (endoscopic adj5 ablation).tw.
121. (multipolar adj5 electrocaut$).tw.
122. (laser adj5 therapy).tw.
123. (radio adj3 frequency adj3 ablation).tw.
124. RFA.tw.
125. ablation.tw.
126. (barrx adj3 system).tw.
127. (early adj cancer).tw.
128. (early adj neoplas$).tw.
129. (high adj grade adj dysplas$).tw.
130. HGD.tw.
131. precancer$.tw.
132. (pre adj cancer$).tw.
133. (pre adj malignant).tw.
134. premalignant.tw.
135. (invasive adj high adj grade adj neoplas$).tw.
136. (high adj grade adj intraepithelial adj metaplas$).tw.
137. (vienna adj2 classification).tw.
138. (indefinite adj2 neoplas$).tw.
139. (indefinite adj2 dysplas$).tw.
140. (non adj2 invasive adj2 low adj grade adj2 neoplas$).tw.
141. or/38‐140
142. 28 and 37 and 141
143. limit 142 to em=200800‐200945

Appendix 4. CENTRAL search strategy

Search date November 12 2009

#1barrett* *esophagus in Clinical Trials
#2MeSH descriptor Barrett Esophagus explode all trees
#3(#1 AND #2)
#4(#3), from 2008 to 2009
•Search string limited to Barrett Esophagus owing to limited number of results when used more focused search strategy.

Appendix 5. ISI Web of Science

Search date: 23 July 2008

Search limited 2000 to 2008

Set
Results
Combine Sets
AND OR
Delete sets
# 931
#6 and #7 and #8
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
# 8135
#1 and #4 and #5
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
# 7105
#1 and #3 and #5
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
# 6113
#1 and #2 and #5
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
# 5>100,000
TS=((random* or RCT*))
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
# 476,852
TS=((ER or "endoscopic removal" or "endoscopic ablation*" or "multipolar electrocaut*" or "laser therapy" or "radio frequency ablation" or RFA or ablation or "barrx system" or "early cancer" or "early neoplas*" or "high grade dysplas*" or HGD or precancer* or "pre cancer*" or "pre malignant" or premalignant or "invasive high grade neoplas*" or "high grade interaepithelial metaplas*" or "vienna classification" or "indefinite neoplas*" or "indefinite dysplas*" or "non invasive low grade neoplas*"))
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
# 3>100,000
TS=((cryotherapy or "cryothera* endoscop*" or laser* or "laser coagulation" or "light coagulation" or "catherer ablation" or "argon plasma coagulat*" or APC or sclerotherap* or electrocoagulat* or "multipolar coagulat*" or "therm* coagulat*" or "heater probe" or "argon laser*" or "YAG laser*" or "yag nd" or "yag ktp" or "monoploar coagulat*" or "bipolar coagulat*" or "multipolar coagulat*" or mpec* or photochemotherapy or "photodynamic therap*" or PDT or "ala pdt" or "endoscopic mucosal resect*" or "ultrasonic surgical aspirat*" or "stomach fundoplication" or fundoplicat* or "antireflux surgery" or nissen or rossetti or troupet or lind or watson or besley or "partial* fundoplication*" or "catherer ablation" or "monopolar coagulat*" or "endoscopic therapy" or "endoscopic submuscosal dissect*" or "endoscopic resect*"))
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
# 2>100,000
TS=(("esophag* resect*" or "oesophag* resect*" or esophagectomy or "esophag* neoplas*" or "oesophag* neoplas*" or "oesophag* neoplas*" or "esophag* cancer*" or "oesophag* cancer*" or "esophag* carcin*" or "oesophag* carcin*" or "oesphag* tumo*" or "esophag* tumo*" or "oesophag* metasta*" or "esophag* metasta*" or "esophag* malig*" or "oesophag* malig*" or "adenocarcinoma* esophag*" or "transthoracic esophag*" or "transthoracic oesophag*" or "thoracic resect*" or "ivor lewis" or "tanner esophag*" or "tanner oesophag*" or "santy esophag*" or "santy oesophag*" or "transhiatal esophag*" or "orringer esophag*" or "orringer oesophag*" or surgery or "surgical treatment" or operative or merendino or "limited resection*" or "vagal sparing oesophagectomy" or "vagal sparing esophagectomy"))
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
# 14,230
TS=(("barrett* oesophagus" or "barrett* oesophagus" or "metaplas* epitheli*" or "columnar line*" or "intest* metaplas*" or "intest* dysplas*"))
Databases=SCI‐EXPANDED, SSCI, A&HCI Timespan=2000‐2008
AND OR

Appendix 6. Evidence‐Based Medicine Reviews (EBMR)

Search date 6 August 2008

Database: All EBM Reviews ‐ Cochrane DSR, ACP Journal Club, DARE, CCTR, CMR, HTA, and NHSEED
Search Strategy:

1 randomised controlled trial.pt. (246,310)
2 controlled clinical trial.pt. or clinical control trial.sd. (145246)
3 randomised controlled trials.sh. or randomised control trial.sd. (137,383)
4 random allocation.sh. (20,277)
5 double blind method.sh. (81,451)
6 single‐blind method.sh. (7713)
7 or/1‐6 (524,480)
8 (animals not human).sh. (5857)
9 7 not 8 (518,760)
10 clinical trial.pt. (273,478)
11 exp clinical trials/ (1542)
12 (clin$ adj25 trial$).ti,ab. (49,617)
13 ((singl$ or doubl$ or trebl$ or tripl$) adj25 (blind$ or mask$)).ti,ab. (108,246)
14 placebos.sh. (18,293)
15 placebo$.ti,ab. (101,806)
16 random$.ti,ab. (246,559)
17 research design.sh. (2607)
18 or/10‐17 (402,427)
19 18 not 8 (396,897)
20 19 not 9 (18,561)
21 comparative study.sh. (4907)
22 exp evaluation studies/ (17,823)
23 follow up studies.sh. (28,686)
24 prospective studies.sh. (49,126)
25 (control$ or prospectiv$ or volunteer$).ti,ab. (236,985)
26 or/21‐25 (271,606)
27 26 not 8 (268,361)
28 27 not (9 or 20) (9693)
29 9 or 20 or 28 (547,014)
30 exp barrett oesophagus/ (116)
31 (barret$ adj3 oesophagus).tw. (120)
32 (barret$ adj3 oesophagus).tw. (86)
33 (metaplas$ adj5 epitheli$).tw. (42)
34 (columnar adj5 line$).tw. (12)
35 (columnar adj5 metaplas$).tw. (11)
36 (intest$ adj5 metaplas$).tw. (116)
37 (intest$ adj5 dysplas$).tw. (28)
38 or/30‐37 (321)
39 exp esophagectomy/ (157)
40 esophagectomy.tw. (184)
41 (esophag$ adj5 neoplas$).tw. (144)
42 (oesophag$ adj5 neoplas$).tw. (14)
43 (esophag$ adj5 cancer$).tw. (463)
44 (oesophag$ adj5 cancer$).tw. (211)
45 (esophag$ adj5 carcin$).tw. (399)
46 (oesophag$ adj5 carcin$).tw. (123)
47 (esophag$ adj5 tumo$).tw. (54)
48 (oesophag$ adj5 tumo$).tw. (35)
49 (oesophag$ adj5 metasta$).tw. (13)
50 (esophag$ adj5 metasta$).tw. (18)
51 (esophag$ adj5 malig$).tw. (66)
52 (oesophag$ adj5 malig$).tw. (32)
53 (adenocarcinoma$ adj5 esophag$).tw. (53)
54 (transthoracic adj10 esophag$).tw. (33)
55 (transthoracic adj10 oesophag$).tw. (9)
56 (thoracic adj10 resect$).tw. (91)
57 (ivor adj lewis).tw. (4)
58 (tanner adj25 esophag$).tw. (4)
59 (tanner adj25 oesophag$).tw. (3)
60 (santy adj25 esophag$).tw. (0)
61 (santy adj25 oesophag$).tw. (0)
62 (transhiatal adj10 esophag$).tw. (28)
63 (orringer adj25 esophag$).tw. (0)
64 (orringer adj25 oesophag$).tw. (0)
65 surgery.tw. (44,cf499)
66 exp surgery/ (221)
67 (surgical adj treatment).tw. (2258)
68 operative.tw. (10259)
69 merendino.tw. (0)
70 (limited adj resection$).tw. (14)
71 (vagal sparing adj oesophagectomy).tw. (0)
72 (vagal sparing adj esophagectomy).tw. (0)
73 exp cryotherapy/ (282)
74 cryotherapy.tw. (522)
75 (cryothera$ adj5 endoscop$).tw. (1)
76 exp lasers/ (439)
77 exp laser coagulation/ (259)
78 exp light coagulation/ (407)
79 exp catheter ablation/ (529)
80 (argon adj5 plasma adj5 coagulat$).tw. (50)
81 APC.tw. (175)
82 exp sclerotherapy/ (360)
83 sclerotherap$.tw. (897)
84 exp electrocoagulation/ (971)
85 electrocoagulat$.tw. (180)
86 (multipolar adj5 coagulat$).tw. (5)
87 (therm$ adj5 coagulat$).tw. (27)
88 (heater adj5 probe).tw. (48)
89 (argon$ adj5 laser$).tw. (416)
90 (YAG adj5 laser$).tw. (246)
91 (yag adj5 nd).tw. (124)
92 (yag adj5 ktp).tw. (2)
93 (monopolar adj5 coagulat$).tw. (12)
94 (bipolar adj5 coagulat$).tw. (35)
95 (multipolar adj5 coagulat$).tw. (5)
96 mpec.tw. (5)
97 exp photochemotherapy/ (475)
98 (photodynamic adj5 therap$).tw. (455)
99 PDT.tw. (310)
100 (ala adj5 pdt).tw. (65)
101 (aminolaevulin$ adj5 acid).tw. (47)
102 (photophrin adj5 pdt).tw. (0)
103 (endoscopic adj5 mucosal adj5 resect$).tw. (26)
104 (ultrasonic adj3 surgical adj3 aspirat$).tw. (11)
105 exp fundoplication/ (171)
106 fundoplicat$.tw. (260)
107 (antireflux adj5 surgery).tw. (98)
108 (nissen or rossetti).tw. (191)
109 (toupet or lind or watson or besley).tw. (148)
110 (partial$ adj5 fundoplication$).tw. (41)
111 (laparoscop$ adj5 fundoplication$).tw. (152)
112 exp catheter ablation/ (529)
113 (monopolar adj5 coagulat$).tw. (12)
114 (endoscopic adj5 therapy).tw. (352)
115 (endoscopic adj5 submuscosal adj5 dissect$).tw. (0)
116 (endoscopic adj5 resect$).tw. (98)
117 ER.tw. (1238)
118 (endoscopic adj5 removal).tw. (61)
119 (endoscopic adj5 ablation).tw. (23)
120 (multipolar adj5 electrocaut$).tw. (1)
121 (laser adj5 therapy).tw. (832)
122 (radio adj3 frequency adj3 ablation).tw. (14)
123 RFA.tw. (69)
124 ablation.tw. (1550)
125 (barrx adj3 system).tw. (0)
126 (early adj cancer).tw. (51)
127 (early adj neoplas$).tw. (8)
128 (high adj grade adj dysplas$).tw. (74)
129 HGD.tw. (28)
130 precancer$.tw. (203)
131 (pre adj cancer$).tw. (40)
132 premalignant.tw. (107)
133 (pre adj malignant).tw. (26)
134 (invasive adj high adj grade adj neoplas$).tw. (0)
135 (high adj grade adj intraepithelial adj metaplas$).tw. (0)
136 (vienna adj2 classification).tw. (0)
137 (indefinite adj2 neoplas$).tw. (0)
138 (indefinite adj2 dysplas$).tw. (6)
139 (non adj2 invasive adj2 high adj grade adj2 neoplas$).tw. (0)
140 (non adj2 invasive adj2 low adj grade adj2 neoplas$).tw. (0)
141 or/39‐140 (58,222)
142 29 and 38 and 141 (141)
143 limit 142 to yr="2000 ‐ 2008" [Limit not valid in: DARE; records were retained] (102)
144 from 143 keep 1‐102 (102)
•Deleted results not part of Cochrane CENTRAL Trials
•Deduplicated records that occurred in search of Cochrane Library (Wiley Interface)

Appendix 7. Search strategy for ongoing clinical trials (Controlled Trials mRCT and ISRCTN)

ISRCTN Register (International) ‐ copy of ISRCTN Register

 

Action Medical Research (UK) ‐ subset from ISRCTN Register

  

NIH ClinicalTrials.gov Register (International) ‐ subset of randomised trial records

The Wellcome Trust (UK) ‐ subset from ISRCTN Register

UK trials (UK) ‐ subset from ISRCTN Register, UK trials only

Search date 17 May 2012

Search string "Barretts and (esophagus or oesophagus)"

• Search not limited by date

• Search not deduplicated

Appendix 8. LILACS

January 1 2000 to November 11 2009
( C06.405.117 OR C04.588.274.476 OR C06.198.102) [DeCS Category] and (E04.210.346 OR VS3.003.001.006.002 OR E01.370.372 OR E01.370.388.250.250 OR VS3.003.001.006 OR E02.594) [DeCS Category]

References found 42

Appendix 9. Methods to be used in future updates of the review

Data extraction and management

Two review authors (CB and SG) will independently extract the data using a specially designed data extraction form. A third team member (JJ) will resolve any differences in opinion. Two review authors (CB and SG) will independently check and enter the data into RevMan 5 (RevMan 2011).

Assessment of risk of bias in included studies

We shall describe each study in a 'Risk of bias' table, and address the following issues, which may be associated with biased estimates of treatment effect, that is, sequence generation, allocation sequence concealment, blinding, incomplete outcome data, selective outcome reporting and other potential sources of bias (Higgins 2011). We shall report specifically on:

(a) the method of generation of the randomisation sequence;
(b) the method of allocation concealment ‐ it will be considered 'adequate' if the assignment could not be foreseen;
(c) who was blinded and not blinded (participants, clinicians, outcome assessors) if this is appropriate;
(d) how many participants were lost to follow‐up in each arm, and whether reasons for losses were adequately reported;
(e) whether all participants were analysed in the groups to which they were originally randomised (intention‐to‐treat principle).

In addition we may report on:
(f) the baseline assessment of the participants for age, sex, duration, location and stage of cancer;
(h) whether outcome measures were described and their assessment was standardised;
(j) the use and appropriateness of statistical analyses, where tabulated data cannot be extracted from the original publication.

We shall record information on all these components in a 'Risk of bias' table. We shall also summarise the general quality of all the studies in the section on 'Risk of bias in the included studies'.

Measures of treatment effect

For studies of a similar type of radical endoscopic therapy (that is, endoscopic therapy with curative intent) versus surgery, if sufficient trials are available and their populations are clinically similar, meta‐analyses of primary and secondary end‐points will be carried out.

For the time‐to‐event outcomes (survival), these outcomes may be reported as hazard ratio (HR) or log rank test and where these measures are provided in a trial report, they will be used.

For meta‐analyses of dichotomous outcomes, RRs or odds ratios (OR) will be calculated with 95% CIs and combined for meta‐analysis with RevMan 5 software (RevMan 2011). Data will be combined for the same duration of follow‐up rounded to the nearest six months.

Continuous data will be combined for meta‐analysis. We will use mean and standard deviations to derive a weighted mean difference (WMD) with 95% CIs using a fixed‐effect model.

Unit of analysis issues

We will seek input from the Cochrane UGPD Group editorial base for analysis issues involving any included trials with multiple treatment groups, and cluster randomised designs. We do not expect to identify any cross‐over trials as this design is inappropriate for this type of intervention, given that surgery and endoscopic therapies are intended to have a curative rather than temporary effect.

Dealing with missing data

We shall contact trial authors or sponsors of studies less than 10 years old to provide missing data. Where there is uncertainty about the specifics of a trial, we shall contact the trial authors for clarification.

Assessment of heterogeneity

Heterogeneity will be assessed with the I² statistic, which provides a measure of the degree of inconsistence in the studies' results with 95% uncertainty intervals (Higgins 2011). A value of 0% indicates no heterogeneity and a value greater than 50% is considered to indicate substantial heterogeneity.

Assessment of reporting biases

We shall test publication bias by the use of a funnel plot when adequate data are available for similar types of endoscopic therapies. If our analysis contains sufficient trials to make visual inspection of the plot meaningful (there is no standard for this, but we will assume that five trials is the minimum), and if the presence of asymmetry in the inverted funnel suggests a systematic difference between large and small trials in their estimates of treatment effect, we may comment on this in the Discussion section.

Data synthesis

HGD and early cancer will be evaluated separately where possible. As a general rule, a fixed‐effect model will be used for calculations of summary estimates and their 95% CIs unless there is significant heterogeneity, in which case results will be confirmed using a random‐effects statistical model (DerSimonian 1986). We will compare relative risk of surviving in the endotherapy compared with surgery at one or more years depending on the data available.

Subgroup analysis and investigation of heterogeneity

Subgroup analysis

We plan to perform subgroup analyses where adequate information is given in reports of studies. The groups will be different types of disease (HGD, mucosal cancer, submucosal cancer), histological type (adenocarcinoma or squamous cell carcinoma) and the age of participants as described by the trial authors. We may also consider subgroup analysis for types of endoscopic therapies or surgical techniques and on the basis of study design and reporting (adequate versus unclear allocation concealment for RCTs).

Investigation of heterogeneity

If substantial heterogeneity (I² > 50%) exists between studies for the primary outcome, we shall explore the reasons for heterogeneity; such as age, cancer stage, histological type and treatment type. If it is inappropriate to pool the data because of clinical or statistical heterogeneity a systematic review without meta‐analysis will be performed or a meta‐analysis excluding outlying studies will be performed.

Sensitivity analysis

We plan to conduct sensitivity analyses to examine the effects of excluding studies defined as those with a moderate or high risk of bias as described in the Cochrane Handbook of Systematic Reviews of Interventions (Higgins 2011).

Adverse outcomes

We shall record:

1. the appropriateness of the methods used to detect adverse events;

2. the adequacy of reporting.

We shall describe the information qualitatively.

Characteristics of studies

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Das 2008	Retrospective review of cases from SEER database. Not an RCT
Farrell 2011	Retrospective review of patients treated with either endoscopic therapies or surgery, review of cases from a single institution. Not an RCT
Pacifico 2003	Retrospective review of patients treated with either endoscopic therapies or surgery. Not an RCT
Pech 2011	Retrospective review of patients treated with either endoscopic therapies or surgery, review of cases from 2 high‐volume centres. Not an RCT
Prasad 2007	Retrospective review of patients treated with either endoscopic therapies or surgery, review of cases from a single institution. Not an RCT
Reed 2005	Retrospective review of patients treated with either endoscopic therapies or surgery. Not an RCT
Rosmolen 2010	Retrospective review, cross‐sectional design. Quality of life and fear of cancer recurrence outcomes. Not an RCT
Schembre 2008	Retrospective review of patients treated with either endoscopic therapies or surgery, review of cases from a single institution. Not an RCT
Schembre 2010	Quality of life survey (retrospective review) of patients treated with either endoscopic therapies or surgery, review of cases from a single institution. Not an RCT
Thomas 2005	Review of clinical practice and outcomes from a pathology database. Not an RCT
Tian 2011	Retrospective review of patients with submucosal oesophageal adenocarcinoma treated with either endoscopic therapies or surgery, review of cases from a single institution. Not an RCT
Yachimski 2008	Retrospective case series of participants identified through a patient registry (single institution). Not an RCT
Zehetner 2011	Retrospective review of patients treated with either endoscopic therapies or surgery, review of cases from a single institution. Not an RCT

RCT: randomised controlled trial; SEER: Surveillance, Epidemiology, and End Results.

Characteristics of studies awaiting classification [ordered by study ID]

BRIDE 2012.

Methods	RCT (feasibility study) NIHR funded
Participants	Up to 100 patients with Barrett's HGD/T1m cancer, from 6 UK centres
Interventions	ER + RFA vs ER + APC; qualitative interviews (patients and surgeons) to ascertain willingness to participate in potential future trial comparing endotherapy to surgery
Outcomes	Feasibility of recruitment target to inform fully powered trial of 2 forms of endotherapy; feasibility of RCT comparing surgery to endotherapy
Notes	The study is not powered to detect a difference between ER + RFA vs. ER + APC

APC: argon plasma coagulation; ER: endoscopic resection; HGD: high‐grade dysplasia; NIHR: National Institute for Health Research; RCT: randomised controlled trial; RFA: radiofrequency ablation.

Differences between protocol and review

We modified the text of the Background following advice from the authors. We added persistence of Barrett's oesophagus as a secondary outcome.

Contributions of authors

All review authors contributed equally to the writing of the protocol. CB guided the review process. CB, SG and JJ screened the results of the database searches. CB and MA screened the updated searches in 2012. AT provided additional information for the first published version of the review. All authors provided feedback and contributions to the text of the review. SG (with CB and JJ) wrote the text of the Discussion. JJ wrote the Abstract. CB, HB, PB, J DeC and KR provided editorial advice and amended the review text in the first published version and subsequent updates. CB, MA, RS, SG and AT checked the data in the additional tables. SG, CB and JJ edited the text of the full review in response to the peer referee's comments. CB and SG updated the review in March 2010 and MA contributed to screening of search results and review content in April 2012.

Declarations of interest

JJ is a consultant to AstraZeneca.

CB is an independent researcher and the proprietor of Systematic Research Ltd, a company that provides research support in evidence‐based medicine.

Edited (no change to conclusions)