Gut 1997; 40: 710-715

Short segment Barrett's oesophagus: prevalence,diagnosis and associations

S Nandurkar, N J Talley, C J Martin, T H K Ng, S Adams

Department ofMedicine, UniversityofSydney, AustraliaS NandurkarN J Talley

Department ofSurgery, University ofSydney, AustraliaC J Martin

Department ofPathology, NepeanHospitalT H K NgS AdamsCorrespondence to:N J Talley,Professor of Medicine,Clinical Sciences Building,PO Box 63, Penrith,NSW 2751, Australia.

Accepted for publication23 December 1996

AbstractBackground-Prevalence of short seg-ment Barrett's (SSB) oesophagus, definedas the absence of macroscopic Barrett'sbut histologically identifiable intestinalmetaplasia, has been reported to be 18%based on haematoxylin and eosin (H&E)staining.Aims-To define the prevalence of SSBoesophagus using H&E and alcian bluestaining and to determine whether SSBoesophagus is associated with inflam-mation at the gastro-oesophageal junction(GOJ).Subjects-Consecutive patients (n=158)presenting for endoscopy completed astructured interview.Methods-Two biopsy specimens takenfrom the GOJ were stained with H&E,alcian blue and Giemsa. A third specimenwas obtained from the distal oesophagus.Intestinal metaplasia was diagnosed ifgoblet cells were definitely identified bytwo independent observers.Results-SSB oesophagus was present in46 (prevalence 36%, 95% confidence in-terval (CI) 28.5-43.5) using alcian bluestaining. IfH&E had been the sole stainingmethod used, 50% cases of intestinalmetaplasia would have been overlooked.There were no cases of intestinal meta-plasia identified by H&E but missed byalcian blue stainling. Logistic regressionanalysis identified age (odds ratio (OR)per decade 103, 95% CI 1-01-1-06), his-tological oesophagitis (OR 3*2, 95% CI1.4-7.2) and inflammation at the gastro-oesophageal junction (OR 5*9, 95% CI2-2-15-6) as independent risk factors forSSB oesophagus.Conclusion-Unrecognised SSB oesopha-gus is highly prevalent in patients pre-senting for diagnostic upper endoscopy ifalcian blue staining is applied.(Gut 1997; 40: 710-715)

Keywords: Barrett's oesophagus, gastro-oesophagealreflux, histology, intestinal metaplasia.

Traditionally, Barrett's oesophagus is definedas the replacement of the distal oesophageallining by three or more centimetres of circum-ferential columnar epithelium in continuitywith the gastric mucosa.' However, the lengthof three centimetres, as required by the abovedefinition, is purely arbitrary. The standard

definition excludes tongues of columnarepithelium that are often seen at the gastro-oesophageal junction.' Some investigators haverecommended that the definition should bemore appropriately based on the histologicaltype rather than length.2 Cardia or junctionaltype epithelium in the distal oesophagus isnot diagnostic of Barrett's oesophagus,3 butdetection of specialised intestinal epithelium(as indicated by the presence of goblet cells) isabnormal. In patients who fail to fulfil thetraditional definition, the presence of special-ised intestinal epithelium has been referred toas short segment Barrett's oesophagus.' 3

Barrett's oesophagus is a strong risk factorfor adenocarcinoma of the oesophagus.9 Theneoplastic risk seems to be related to thepresence of specialised intestinal type epi-thelium, and not the junctional or cardiatypes.'0 "1 There is increasing evidence thatthese adenocarcinomas can arise from veryshort segments of metaplastic epithelium.6 12Short segments of Barrett's oesophagus mayoften be overlooked during routine endoscopybecause the gastro-oesophageal junction ap-pears to be normal and biopsy specimens arenot obtained. In a recent study, the prevalenceof short segment Barrett's oesophagus wasreported to be 18%.13 Another preliminaryreport found a prevalence of 24%./14 However,the true prevalence in the community is stillunclear as these studies used haematoxylin andeosin (H&E) staining alone to detect thespecialised intestinal type epithelium. Alcianblue has been shown to be superior to H&E indetection of intestinal metaplasia,4 '5 but thismethod of diagnosis has not been evaluated inpatients without standard Barrett's oeso-phagus. We hypothesised that the prevalence ofshort segment Barrett's oesophagus has beenunderestimated because of the methods used.

Barrett's oesophagus has been causallylinked to chronic gastro-oesophageal refluxdisease in humans'6-18 and in experimentalmodels of Barrett's oesophagus,'9-22 but inthe only published paper on short segmentBarrett's oesophagus, symptoms of gastro-oesophageal reflux were not associated withthe condition.'3 This study did not evaluatethe association of short segment Barrett'soesophagus with inflammation of thegastro-oesophageal junction. We postulatedthat such a link would be detected. Hence, weaimed to estimate the prevalence, comparedifferent methods of diagnosis and assess forassociations of short segment Barrett'soesophagus in consecutive patients presentingfor endoscopy.

710

on Septem

ber 1, 2020 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.40.6.710 on 1 June 1997. Dow

nloaded from

Short segment Barrett's oesophagus

Methods

SUBJECTS

Consecutive outpatients booked for routinegastroscopy at Nepean Hospital (irrespectiveof their indication) were invited to participatein this study. The study was approved by theWentworth Area Health Service Research andEthics Committee. Patients were excludedfrom the study if they had known Barrett'soesophagus, were unwilling to participate orhad conditions that did not permit a safebiopsy (for example, coagulopathy, oeso-phageal varices).

SYMPTOMSAll patients completed a structured interviewthat measured symptoms of gastro-oeso-phageal reflux disease (that is, heartburn, acidregurgitation and dysphagia). All subjects wereinterviewed individually by one ofus (SN) andall the terms were specified to minimisemeasurement bias. The frequency of symp-toms was graded as follows: 1, less than oncea month; 2, once a month; 3, once a week; 4,several times a week; or 5, daily. The severitywas graded as follows: 1, mild - can be ignoredif I don't think about it; 2, moderate - cannotbe ignored but does not affect my lifestyle; 3,severe - affects my lifestyle; or 4, very severe- noticeably affects my lifestyle.The frequency and severity scores were

added to compute combined scores for eachsymptom (ranges: heartburn score 0-9; acidregurgitation score 0-9; and dysphagia score0-9). Relevant current medication history (useof antacids, H2-receptor antagonists or protonpump inhibitors) was also obtained.

ENDOSCOPY

At endoscopy, macroscopic oesophagitis, ifpresent, was noted and graded according tothe Hetzel grading system23 as follows: grade0, normal appearing mucosa; grade 1, mu-cosal oedema, hyperaemia, and/or friabilityof mucosa; grade 2, superficial erosions invol-ving <10% of mucosal surface of the distal 5cm of oesophageal squamous mucosa; grade3, superficial erosions/ulcerations involving10-50% of distal oesophagus; grade 4, deeppeptic ulceration anywhere in the oesophagusor confluent erosion of >50% of the distaloesophageal squamous mucosa. The endos-copist was unaware of the symptom scores andhistology results.

HISTOLOGYIn the absence of clinically apparent Barrett'soesophagus, two biopsy specimens of thegastro-oesophageal junction were taken in eachpatient. The gastro-oesophageal junction wasidentified by noting the change in colourbetween the squamous and columnar epi-thelium. The biopsy sample was carefully takenfrom a piece of the columnar epitheliumimmediately distal to the z-line. These sampleswere fixed with 10% buffered formalin and

then embedded in paraffin wax. Serial sectionswere cut at three levels from each specimen.Sections were stained with haematoxylin andeosin (H&E), alcian blue and Giemsa.

All sections were analysed independently forthe presence of intestinal metaplasia, Helico-bacter pylori and inflammation. The pathol-ogists were unaware of the endoscopy andclinical data. On H&E staining, intestinal met-aplasia was considered to be present if gobletcells were identified based on their goblet cuplike shape and characteristic staining (basalnucleus and clear spherical apical cytoplasm).'3

Goblet cells contain acid mucin and stainintensely blue with alcian blue at pH 2.5.24 25Columnar cells which stain less intensely withalcian blue are not diagnostic of Barrett'soesophagus as they can also be found in themucus neck region of gastric glands. Positiveidentification of goblet cells with alcian blue inthis study required the characteristic gobletshape and intense blue staining.4 Absence ofeither of the two characteristics negated thediagnosis of intestinal metaplasia in order toavoid overdiagnosis. A section of the smallintestine was also stained simultaneously as acontrol.H pylo?i was identified by using the Giemsa

stain using the criteria of its characteristicmorphology (small curved basophilus bacil-lus), position and distribution (closely appliedto epithelial cell surface, present within the pitsand in overlying mucus).26

Inflammation at the gastro-oesophagealjunction was diagnosed if increased numbers oflymphocytes and polymorphonuclear leuco-cytes were present in the lamina propriabeneath the columnar epithelium.27 TheSydney System was used for the diagnosis ofacute and chronic gastritis.28 Patients in whomboth of the gastro-oesophageal junction biopsyspecimens showed only stratified squamousepithelium but failed to include any columnarepithelium, were labelled as sampling error andwere not included in the final analyses.

If endoscopically obvious Barrett's oeso-phagus was present (circumferential columnarlined oesophagus >3 cm in length), patientsdid not undergo biopsy according to theprotocol. Instead, biopsy specimens were takenas a part of a Barrett's oesophagus screeningprogramme. These patients were not includedin further analyses.An additional biopsy specimen was also

taken from the oesophagus, 2 cm above thegastro-oesophageal junction. This was pro-cessed in identical manner (see earlier), butstained with H&E only and graded indepen-dently. Oesophagitis was identified if there was(1) basal cell hyperplasia >15% of the totalepithelium, (2) increased papillary length>66% of the squamous epithelium, and (3)infiltration by polymorphonuclear leucocytesor eosinophils, or both.29 At least two of theabove findings had to be present for a positivediagnosis of microscopic oesophagitis.30 Aseparate analysis was carried out where allthree criteria were required for diagnosis.

All biopsy specimens were scrutinised by twoexperienced pathologists. Any disagreement

711

on Septem

ber 1, 2020 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.40.6.710 on 1 June 1997. Dow

nloaded from

Nandurkar, Talley, Martin, Ng, Adams

(</5%) between the two investigators wasresolved by consensus. Based on the presenceor absence of intestinal metaplasia, patientswere divided into those with and without shortsegment Barrett's oesophagus.

PATIENT SELECTIONFrom 1 April to 1 August 1995, 182 con-secutive patients attending Nepean Hospitalfor upper endoscopy consented to enter thestudy. Only one patient declined for personalreasons. Of the 182 patients, four patients(2-2%) in whom the diagnosis of typicalBarrett's oesophagus was made for the firsttime, during endoscopy, were excluded from

~~~~~~~tpo

..4

4.t

S'S

FO~~~~~~~~~'

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~4

Figure 1: Biopsy samplefrom the gastro-oesophageal junction stained with alcian blue(pH 2-5) showing goblet cells with the characteristic features of intense blue staining and thetypical spherical shape.

Figure 2: Biopsy samplefrom the gastro-oesophagealjunction stained with haematoxylinand eosin showing goblet cells with the characteristic features ofa basal nucleus and a clearspherical apical cytoplasm.

TABLE I Comparison of alcian blue and haematoxylin andeosin staining in the detection of intestinal metaplasia

Haematoxylin and eosin

Positive (n) Negative (n)

Alcian blue:Positive 23 23Negative 0 112

further analysis. In a further 20 patients (11%),the biopsy specimens contained only stratifiedsquamous epithelium, indicating that thegastro-oesophageal junction was missed.Because of this sampling error, these specimenscould not be evaluated for metaplasia and sothe patients were excluded.

STATISTICAL ANALYSIS

Univariate and multiple logistic regressionanalyses were used to assess the relation be-tween the prevalence of short segmentBarrett's and age, sex, symptoms of gastro-oesophageal reflux disease, medication, micro-scopic oesophagitis, inflammation of thegastro-oesophageal junction, and H pyloristatus. Unadjusted and age adjusted oddsratios (OR) (with 95% conficence intervals(CI)) are reported for each potential risk factor.Exact two-tailed p values, based on a Waldstatistic, are reported which relate to thestatistical Null hypothesis of a population oddsratio of unity. The independence of multiplerisk factors was examined using forward step-wise selection of logistic regression models.

Results

PREVALENCEIn total, 158 patients were included in thefinal analysis (mean age 50x8 years; 66%women). Intestinal metaplasia was detected in46 patients (prevalence rate 36%, 95% CI28x5-43.5%) by alcian blue staining. Meta-plasia at the gastro-oesophageal junction wasdetected in both biopsy samples in 15 (33%)patients and in one sample in 31 (67%)patients.

COMPARISON OF H&E AND ALCIAN BLUE

STAINING

Figure 1 shows intestinal metaplasia as detec-ted by staining with alcian blue. On H&Estaining alone (Fig 2), intestinal metaplasia wasdetected in only 23 (15%) patients. There wereno cases of intestinal metaplasia that weredetected by H&E but missed by alcian bluestaining. The prevalence of intestinal meta-plasia detected by alcian blue and H&Estaining is shown in Table I.

ASSOCIATIONSTable II summarises the clinical features of thepatients with and without short segmentBarrett's oesophagus, and the univariate andage adjusted analyses. The patients withintestinal metaplasia as a group were older thanthose without (p=0 009). However, there wasno difference in the sex ratios between the twogroups.Symptom scores constructed from the reflux

questionnaire did not differ between the twogroups. Erosive oesophagitis (Hetzel grade > 1)was more prevalent in the group with intestinalmetaplasia (24% v 14%). However, adjustingfor age, this difference did not reach statistical

* e * -0

~~~~~~,.~~~~~~~~~~~~.P."

712

on Septem

ber 1, 2020 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.40.6.710 on 1 June 1997. Dow

nloaded from

Short segment Barrett's oesophagus

TABLE II Characteistics ofpatients (n=158) with and without short segment Barrett'soesophagus

Short segment Barrett'soesophagus p Value

Positive Negative Adjusted(n=46) (n=112) Univariate for age

Mean age (years) 55-7 47-9 0 009 0 009Women (O/o) 65-2 68-3 0-92 0-89Mean heartburn score (range: 0-9) 7 6-8 0-69 0-84Mean acid regurgitation score 5-2 5-7 0-46 0 55

(range: 0-9)Mean dysphagia score (range: 0-9) 5-1 4-8 0-89 0-88Treatment (%):No treatment 40-2 37 0-71 0-87H2-receptor antagonists (H2RA) 45-5 50 0-61 0 49Proton pump inhibitors (PPI) 8-9 8-7 0-96 0-86H2RA and PPI 5-4 4-3 0-79 0-54

Macroscopic oesophagitis (%) - - 0 07 0-13No oesophagitis 60-9 69-6 -

Oesophagitis:Grade 1 15-2 161 -

Grade 2 8-7 8-9 -

Grade 3 10 9 3-6 -

Grade 4 4-3 18 -

Basal cell hyperplasia (%/o) 54 21 <0-001 <0*001Increased papillary length (%) 41 20 0-008 0-004Infiltration with neutrophils/ 32 14 0-01 0-006

eosinophils (O/o)Microscopic oesophagitis (/) 43 19 0-001 0-001

(.2 criteria present)Microscopic oesophagitis (/) 28 13 0-02 0-01

(all 3 criteria present)Gastro-oesophageal junction: 87 50 <0-001 <0-001

inflammation (%)Gastro-oesophageal junction: 13 17 0-54 0-56

Hpylori (%/)

*p Values were obtained from a logistic regression analysis adjusted for age.

significance (OR 1-4, 95°/o CI 0-9-2-3,p=0-13). Histological oesophagitis was asso-ciated with short segment Barrett's oesophagus(OR 3.9, 95°/o CI 1-8-8&7, p=0-001). Whendata were re-analysed using the presence of allthree criteria for the diagnosis of histologicaloesophagitis, the association was strengthenedwhen adjusted for age (OR 3-25, 95% CI1V3-7-9, p=001). Short segment Barrett'soesophagus was also strongly associated withinflammation at the gastro-oesophageal junc-tion when adjusted for age (OR 6-7, 95% CI2-6-17-3, p<0001). Short segment Barrett'soesophagus was not associated with medi-cation use or the presence of H pylon at thegastro-oesophageal junction.

Logistic regression analysis confirmed thatage (OR per decade 1-03, 95% CI 1-01-1V06,p=0-005), histological oesophagitis (OR 3-2,95% CI 1-47-2, p=0 006), and inflammationat the gastro-oesophageal junction (OR 5 9,95% CI 2.2-15-6, p<0-001) were indepen-dent risk factors for short segment Barrett'soesophagus.

DiscussionIntestinal metaplasia characterised by thepresence of goblet cells is the sine qua non forthe diagnosis of short segment Barrett'soesophagus. Other types of epithelium that arefound in standard Barrett's oesophagus such asjunctional or fumdic mucosa are per se notdiagnostic of Barrett's oesophagus as they canoccur in the distal oesophagus of peoplewithout the condition.3 In our study, we foundthat one in three patients presenting forendoscopy had intestinal metaplasia. Thisprevalence is higher than that published in the

USA,3 14 in which H&E was the sole stainingmethod used to detect intestinal metaplasia.This discrepancy is explained by our use ofalcian blue which seemed to be superior tostaining with H&E. In our study, the preva-lence of intestinal metaplasia using H&Ealone was 15%, which is comparable with theprevalence described by Spechler et al.'3 Wefound no cases of intestinal metaplasia thatwere detected by H&E but missed by alcianblue. We do not believe that the additionalcases detected by alcian blue were likely to befalse positives because we adhered to strictcriteria for diagnosis.Our results are consistent with the appli-

cation of alcian blue staining in other settings.Cooper et al obtained biopsy samples from 11children (mean age 5 9 years) with standardBarrett's oesophagus.3' They identified fiveadditional cases of specialised intestinal epi-thelium on staining with alcian blue that weremissed with H&E staining. Gottfried et al, too,found alcian blue to be more sensitive thanH&E in the diagnosis of intestinal metaplasia.4When strict criteria were used (that is, rejectionof patients showing only diffuse acid mucinin the columnar cells with alcian blue), theyconcluded that goblet cell metaplasia detectedby alcian blue is highly specific for the histo-logical diagnosis of columnar lined (Barrett's)oesophagus. Because we used strict criteria inour study for the positive identification ofgoblet cells, we may have actually under-estimated the prevalence of short segmentBarrett's oesophagus, but we believe that anysuch error would be small. We recommendthat alcian blue be used as a routine stain inthe diagnosis of short segment Barrett'soesophagus.We found that the prevalence of short

segment Barrett's oesophagus was higher inolder subjects. Epidemiological studies byCameron et al32 and GOSPE33 have shown thatstandard Barrett's oesophagus is also agedependent. This supports the view that boththese conditions are acquired. We adjusted forage in our analyses as this may have otherwiseconfounded our results.A causal relation between gastro-oeso-

phageal reflux disease and standard Barrett'soesophagus is believed to exist in humans,'1-8and refluxate has been used to induce Bar-rett's mucosa in experimental models.'9-22Nakano et al subjected Wistar rats to gas-trectomy and an end to side anastomosisbetween the oesophagus and jejunum, re-sulting in free jejuno-oesophageal reflux.22They showed the appearance of columnarepithelium in the oesophagus replete withabsorptive and goblet cells. These goblet cellswere positively identified on staining with highiron diamine/alcian blue and concanavalin Atype III. Martin et al performed gastroenter-ostomy and cardioplasty in dogs to promotefree pancreaticoduodenal reflux into the oeso-phagus.2' Gastric acid secretion was inhibitedwith an oral proton pump inhibitor (omepra-zole -40 ,umol/kg/day). The oesophagealmucosa was resected and allowed to heal in thismodified environment. They noted healing

713

on Septem

ber 1, 2020 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.40.6.710 on 1 June 1997. Dow

nloaded from

Nandurkar, TaUley, Martin, Ng, Adams

with squamous epithelium at the proximalborder whereas distal healing occurred witha columnar epithelium (with goblet cells).Barrett's epithelium has also been reported asa consequence of acid reflux after oesophagealmyotomy in patients with achalasia.34

It has been hypothesised that destruction ofthe stratified squamous epithelium and thesubsequent repair process leads to the develop-ment of islands of intestinal metaplasia in theoesophagus. We identified a trend (24% v

14%) for erosive oesophagitis to be associatedwith short segment Barrett's oesophagus. Thelack of statistical significance may reflect a typeII error; we calculate, based on our results, thata population of 390 patients would be neededto detect a difference of 20% with a power of80%. Histological changes in the oesophagussuch as basal cell hyperplasia, increasedpapillary length and infiltration by neutrophilsor eosinophils, or both, have been reported tobe useful predictors of pathological acid reflux,although this is controversial.35 In a study byIsmail-Beigi et al these changes were presentin 85% of patients with symptomatic reflux,based on history and a short pH study, andin 10% of normal controls.30 In our study,histological oesophagitis was a more sensitivemarker of reflux disease than macroscopicchanges. We believe that these histologicalchanges suggest that gastro-oesophageal refluxdisease is linked to short segment Barrett'soesophagus, but further studies are neededusing 24 hour oesophageal pH monitoring. Wefound, however, that the symptoms of gastro-oesophageal reflux disease were not associatedwith the presence of short segment Barrett'soesophagus, consistent with an earlier study.'3This may reflect the lack of sensitivity ofsymptoms in general in identifying thepresence of Barrett's oesophagus.33We found that inflammation in the columnar

epithelium at the gastro-oesophageal junctionhad the strongest association with intestinalmetaplasia. An increased prevalence of inflam-mation at the gastro-oesophageal junction inpatients with short segment Barrett's oeso-phagus was also noted by Clark et al.'4 In theirpreliminary study, ambulatory 24 hour pHtesting showed increased exposure to oeso-

phageal acid in a cohort of patients with shortsegment Barrett's oesophagus. In contrast tothe present study, when inflammation at thegastro-oesophageal junction and histologicaloesophagitis were analysed together, onlyinflammation at the gastro-oesophageal junc-tion remained statistically significant. Wehypothesise that inflammation at the gastro-oesophageal junction acts as a nidus formetaplastic change. The exact cell of origin ofBarrett's oesophagus is unknown but undiffer-entiated pluripotent stem cells, basal cells ofsquamous epithelium and epithelial cells, andcongenital gastric rests are all candidates.Epidermal growth factor receptor (EGFr) isconcentrated in the basal and prickle cell layersof oesophageal epithelium as well as inBarrett's mucosa.36 3' Acid stimulates EGFrwhich in turn seems to be involved in cellularproliferation and maturation, stimulating

wound healing and maintaining tissue in-tegrity.36 3 Hence, it can be postulated thatinflammation at the gastro-oesophagealjunction secondary to reflux of acid andpossibly duodenal juice may activate EGFr(and possibly other cytokines) which in turninduces cellular dedifferentiation leading tointestinal metaplasia.We did not find any difference in the prev-

alence of H pylori at the gastro-oesophagealjunction in the patients with and withoutshort segment Barrett's oesophagus. We hadpostulated that a higher local prevalence ofHpylori in patients with short segment Barrett'smay account for increased inflammatorychanges at the gastro-oesophageal junction. Asthe antrum and the corpus were not sampled,it is likely that we have underestimated theprevalence of H pylori distally. However,increased detection ofHpylori from the antrumwould not necessarily negate reflux as a causalpossibility as H pylori can coexist in patientswith gastro-oesophageal reflux disease withoutbeing linked causally to the latter. We suspectthat gastro-oesophageal reflux disease ratherthan H pyloni is the underlying pathophysio-logical mechanism causing inflammation at thegastro-oesophageal junction in patients withintestinal metaplasia.Although there was an age related increase

in prevalence in standard Barrett's oesophagusin one large study, no significant increase inlength was noticed.32 This report suggestedthat standard Barrett's oesophagus develops toits full length all at once, usually many yearsbefore it is diagnosed, with no significantfurther increment in its length with time.Whether short segment Barrett's oesophagusis a completely separate entity or a precursorlesion for typical Barrett's oesophagus isunknown, but the former seems more likelybased on the data available.A significant increase in the incidence of

oesophageal adenocarcinoma has been notedin the past few decades,38 39 and these tumoursare almost always associated with underlyingspecialised intestinal type epithelium ratherthan junctional or fundic type.'0 11 Anecdotalevidence based on case reports supports theview that adenocarcinoma is associated withshort segment Barrett's oesophagus.6 12 Thevery high prevalence of short segment Barrett'soesophagus may therefore be of clinicalimportance. There are no studies which havedetermined the cumulative risk of developmentof adenocarcinoma in patients with shortsegment Barrett's oesophagus. Moreover, thenatural history of short segment Barrett's oeso-phagus has not been characterised carefully.Further studies are required before it will beknown whether guidelines for surveillance inpatients with short segment Barrett's oeso-phagus will need to be developed.We thank Dr M Jones PhD and Dr J Bai MPH for theirassistance with the statistical analyses. We also thank DrsL Brooks, A Keegan and M Cox for their help in providing uswith patients for the study.

1 Stein HJ, Siewart JR. Barrett's esophagus: pathogenesis,epidemiology, functional abnormalities, malignantdegeneration, and surgical management. Dysphagia 1993;8: 276-88.

714

on Septem

ber 1, 2020 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.40.6.710 on 1 June 1997. Dow

nloaded from

Short segment Barren's oesophagus

2 Spechler SJ, Goyal RK. Barrett's esophagus. N Engl J Med1986; 315: 362-71.

3 Haggitt RC. Barrett's esophagus, dysplasia, and adeno-carcinoma. Hum Pathol 1994; 25: 982-93.

4 Gottfried MR, McClave SA, Boyce HW. Incompleteintestinal metaplasia in the diagnosis of columnar linedesophagus (Barrett's esophagus). Am J Clin Pathol 1989;92: 741-6.

5 Pera M, Cameron AJ, Trastek VF, Carpenter HA,Zinsmeister AR. Increasing incidence of adenocarcinomaof the esophagus and esophagogastric junction. Gastro-enterology 1993; 104: 510-3.

6 Schnell TG, Sontag SJ, Chejfec G. Adenocarcinomasarising in tongues of short segments of Barrett'sesophagus. Dig Dis Sci 1992; 37: 137-43.

7 Hamilton SR, Smith RR, Cameron JL. Prevalence andcharacteristics of Barrett's esophagus in patients withadenocarcinoma of the esophagus or esophagogastricjunction. Hum Pathol 1988; 19: 942-8.

8 Clark GWB, Smyrk TC, Burdiles P, Hoeft SF, Peters JH,Kiyabu M, et al. Is Barrett's metaplasia the source ofadenocarcinomas of the cardia? Arch Surg 1994; 129:609-14.

9 Thompson nJ, Zinsser KR, Enterline HT. Barrett'smetaplasia and adenocarcinoma of the esophagus andgastroesophageal junction. Hum Pathol 1983; 14: 42-61.

10 Skinner DB, Walther BC, Riddell RH, Schmidt H,Lascone C, DeMeester TR. Barrett's esophagus: com-parison of benign and malignant cases. Ann Surg 1983;198: 554-66.

11 Hamilton ST, Smith RR. The relationship betweencolumnar epithelial dysplasia and invasive adenocar-cinoma arising in Barrett's esophagus. Am J Clin Pathol1987; 87: 301-12.

12 Cameron AJ, Lomboy CT, Pera M, Carpenter HA.Adenocarcinoma of the esophagogastric junction andBarrett's esophagus. Gastroenterology 1995; 109: 1541-6.

13 Spechler SJ, Zeroogian JM, Antonioli DA, Wang HH,Goyal RK. Prevalence of metaplasia at the gastro-oesophageal junction. Lancet 1994; 344: 1533-6.

14 Clark GWB, Ireland AP, Chandrasoma P, DeMeester TR,Peters JH, Bremner CG. Inflammation and metaplasia inthe transitional epithelium of the gastroesophagealjunction: a new marker for gastroesophageal reflux disease[abstract]. Gastroenterology 1994; 106: A63.

15 Cooper JE, Spitz L, Wikins BM. Barrett's oesophagus inchildren. A histologic and histochemical study of 11 cases.J Pediatr Surg 1987; 3: 191-6.

16 DeMeester TR, Attwood SEA, Smyrk RC, Therkildsen DH,Hinder RA. Surgical therapy in Barrett's esophagus. AnnSurg 1990; 212: 528-40.

17 Bozymski EM, Herlihy KJ, Orlando RC. Barrett's eso-phagus. Ann Intern Med 1982; 97: 103-7.

18 Wmters C, Spurling TJ, Chobanian S, Curtis DJ,Esposito RI, Hacker JF, et al. Barrett's esophagus: aprevalent, occult complication of gastroesophageal refluxdisease. Gastroenteroogy 1987; 92: 118-24.

19 Fink MA, Martin CJ, Ewing HP, Machet D, Rode J. Alongitudinal study of the development of experimentalBarrett's oesophagus in the dog [abstract]. Gastro-enterology 1991; 100: A825.

20 Poliness CM, Martin CJ, Ewing HP, Thompson ME,Machet D, Rode J. Bile affects the pattern of metaplasiain experimental columnar lined oesophagus [abstract].Gastroenterology 1993; 104: A173.

21 Martin CJ, Shaw ME, Ewing HP, Machet D. Pancreaticreflux produces intestinal metaplasia in experimentalcolumnar lined oesophagus [abstract]. Aust NZ Jf Surg1995; 65: A422.

22 Nakano K, Sato H, Kimura T, Noda M, Kodama T,Kashima K, et al. An experimental model of Barrett'sesophagus by jejunoesophageal reflux [abstract]. Gastro-enterology 1996; 110: A206.

23 Hetzel DJ, Dent J, Reed WD, Narielwala FM,Mackinnon M, McCarthy JH, et al. Healing and relapseof severe peptic esophagitis after treatment withomeprazole. Gastroenterology 1988; 95: 903-12.

24 Rothery GA, Patterson JE, Stoddard CJ, Day DW.Histological and histochemical changes in the columnarlined (Barrett's) oesophagus. Gut 1986; 27: 1062-8.

25 Peuchmaur M, Potet F, Goldfain D. Mucin histochemistryof the colurnrar epithelium of the esophagus (Barrett'sesophagus): a prospective biopsy study. 7 Clin Pathol1984; 37: 607-10.

26 Wyatt JI, Gray SF. Detection of Helicobacter pylori byhistology. In: Rathbone BJ, Heatley RV, eds. Helicobacterpylori and gastroduodenal disease. Oxford: BlaxwellScientific Publications, 1992: 51-7.

27 Robert ME, Weinstein WM. Helicobacter pylori associatedgastric pathology. Gastroenterol Clin North Am 1993; 22:59-72.

28 Misiewicz Jn, Tytgat GNJ, Price AB, Goodwin CS,Strickland RG, Sipponen P. The Sydney System: A newclassification of gastritis. Jf Gastroenterol Hepatol 1991; 6:207-8.

29 Frierson HF. Histology in the diagnosis of reflux eso-phagitis. Gastroenterol Clin NorthAm 1990; 19: 631-44.

30 Ismail-Beigi F, Horton PF, Pope CE. Histological con-sequences of gastro-esophageal reflux in man. Gastro-enterology 1970; 58: 163-74.

31 Cooper JE, Spitz L, Wilkins BM. Barrett's oesophagus inchildren. A histologic and histochemical study of 11 cases.J7 Pediatr Surg 1987; 3: 191-6.

32 Cameron AJ, Lomboy CT. Barrett's esophagus: Age,prevalence and extent of columnar epithelium. Gastro-enterology 1992; 103: 1241-5.

33 Gruppo Operativo Per Lo Studio Della Precancerosi Dell'Esofago (GOSPE). Barrett's esophagus: Epidemiologicaland clinical results of a multicentric survey. Int Jf Cancer1991; 48: 364-8.

34 Agha FP. Barrett's esophagus complicating achalasia afteresophagomyotomy. A clinical, radiologic and pathologicstudy of 70 patients with achalasia and related motordisorders.GCIin Gastroenterol 1987; 9: 232-7.

35 Schindlbeck NE, Wiebecke B, Klauser AG, Voderholzer WA,Muiler-Lissner SA. Diagnostic value of histology in non-erosive gastro-oesophageal reflux disease. Gut 1996; 39:151-4.

36 Jankowski J, Murphy S, Coghill G, Grant A, Wormsley KG,Sanders DSA, et al. Epidermal growth factor receptors inthe oesophagus. Gut 1992; 33: 439-43.

37 Jankowski J, Coghill G, Tregakis B, Hopewood D,Wormsley KG. Epidermal growth factor in theoesophagus. Gut 1992; 33: 1448-53.

38 Hansson L-E, Sparen P, Nyren 0. Increasing incidence ofcarcinoma of the gastric cardia in Sweden from 1970 to1985. BrJSurg 1993; 80: 374-7.

39 Blot WJ, Devesa SS, Kneller RW, Fraumeni JF. Risingincidence ofadenocarcinoma of the esophagus and gastriccardia.JAMA 1991; 265: 1287-9.

715

on Septem

ber 1, 2020 by guest. Protected by copyright.

http://gut.bmj.com

/G

ut: first published as 10.1136/gut.40.6.710 on 1 June 1997. Dow

nloaded from