Barrett's Esophagus

Comparison of Benign and Malignant Cases

DAVID B. SKINNER, M.D., BRUNO C. WALTHER, M.D., ROBERT H. RIDDELL, M.D., HELMUT SCHMIDT, M.D.,CLEMENT IASCONE, M.D., THOMAS R. DEMEESTER, M.D.

Using strict criteria for diagnosis, 23 patients having benignBarrett's esophagus, and 20 patients with adenocarcinoma aris-ing in this epithelium have been analyzed. Evidence supportssevere gastroesophageal reflux as a cause of Barrett's esoph-agus. Successful antireflux surgery leads to stabilization andpossibly regression of the dysplasia in Barrett's epithelium, andcan be followed by squamous epithelial regeneration in some.Antireflux surgery is advocated in all patients with Barrett'sesophagus demonstrated to have abnormal reflux regardless ofsymptoms. The malignant potential of the columnar epitheliumis higher in men who smoke, in patients with intestinal-typemetaplasia who continue to have severe reflux, and in patientswho develop dysplasia. In those with high grade dysplasia, theprobability of carcinoma is high and esophagectomy should beseriously considered in the hopes that the pathological stage ofthe neoplasm is still favorable.

IN 1950, BARRETT DESCRIBED the condition of pepticulceration in the esophagus arising in a zone of gas-

tric-type epithelium.' Allison and Johnstone, in 1953,published additional cases in which the esophagus waslined with gastric-like mucosa and suggested the term"esophagus lined with columnar epithelium."2 Barrettaccepted this terminology and provided a more com-plete description of the disease in 1957.3 Controversyhas persisted to the present concerning the etiology, def-inition and classification, complications, and clinicalmanagement of Barrett's esophagus.4The etiology of the columnar-lined esophagus has

Presented at the 103rd meeting of The American Surgical Associ-ation, Boca Raton, Florida, May 11-14, 1983.

Supported in part by The Triad Trust, The Chicago CommunityTrust, and University of Chicago Clinical Research Center, USPHSGrant No. RR00055.

Reprint requests: Dr. David B. Skinner, Department of Surgery,The University of Chicago Medical Center, 950 E. 59th Street, Chi-cago, IL 60637.

Submitted for publication: May 16, 1983.

From the Departments of Surgery and Pathology, TheUniversity of Chicago Pritzker School of Medicine,

Chicago, Illinois

been debated extensively. Among the original descrip-tions of the condition, Barrett proposed a congenitalorigin, whereas Allison and Johnstone indicated the pos-sibility that the condition might be acquired. Descrip-tions of a cephalad migration of peptic esophagitis andstricture above an ascending boundary of a columnar-lined esophagus as in the cases of Goldman and Beck-man,5 and Mossberg6 provided strong evidence that thecondition could be progressive and result from persistingsevere gastroesophageal reflux and complicating esoph-agitis. The concept of erosive reflux esophagitis healingby upward migration of adjacent columnar epitheliumwas advanced by Hayward in 196 .7 The possibility ofrepair by extension from esophageal glands followingreflux esophagitis was proposed by Adler in 1963.8 It isnow widely accepted that a substantial proportion ofpatients with Barrett's esophagus do have severe gas-troesophageal reflux.

Almost simultaneously with the reporting of the be-nign version of this condition, cases were described inwhich adenocarcinoma of the esophagus developed inthe aberrant gastric-type mucosa, the first being de-scribed by Morson and Belcher in 1953,9 and in thiscountry by McCorkle and Blades in 1955.10 Althoughthe potential for the columnar-lined esophagus to un-dergo malignant degeneration has become well known,the interrelationships between the benign and malignantform of the disease, and potential for progression fromone to the other, have been poorly understood. The roleof persisting reflux in causing progression of the con-dition towards malignant degeneration has been uncer-

0003-4932/83/1000/0554 $01.40 C) J. B. Lippincott Company

554

BARRETT'S ESOPHAGUS 555tain, and the effect of correcting the reflux on the fateof the columnar-lined epithelium has remained unclear.The purposes of this study are to compare cases with

the benign or malignant type of columnar-lined esoph-agus to detect differences between the two groups whichmight suggest etiological or prognostic factors. The fateof the epithelium in benign cases after successful anti-reflux repair is examined to detect evidence ofregressionor change in the type or degree of dysplasia in the ep-ithelium. Based upon results achieved in the benign andmalignant cases, suggestions are made concerning treat-ment.

DefinitionsFor comparisons among patients with Barrett's esoph-

agus to be meaningful, precise definitions of the abnor-mality prove necessary. For the purposes of this study,Barrett's esophagus is the condition in which 3 or morecm of the distal tubular esophagus are lined by colum-nar-type epithelium. This definition is deliberately cho-sen, recognizing that the squamo-columnar junctionmay be irregular, and that the distal 1-2 cm of normaltubular esophagus are lined with cardia-type columnarepithelium. Equivocal cases or patients with only smalltongues of columnar epithelium are thereby excluded.The definition requires that the 3 or more cm of colum-nar epithelium occur within the tubular esophagus, re-gardless of the presence or absence of a hiatal hernia(Fig. 1). The tubular esophagus in cases of hiatal herniais defined as that segment of foregut which participatesin esophageal peristaltic contractions, as are seen duringendoscopy or barium swallow, and recorded duringmanometric studies.

The malignant version of Barrett's esophagus is de-fined as the condition in which adenocarcinoma occursabove the junction of the tubular esophagus with stom-ach, and in which mucosa adjacent to the carcinomademonstrates columnar epithelium, meeting the abovedefinition of Barrett's esophagus.

For the purposes ofthis study, two cases ofcarcinomaarising in ectopic gastric-type epithelium occurring inthe cervical esophagus above a zone of confluent squa-mous epithelium are not considered in the analysis. Theetiology in such cases appears to be congenital," andnot necessarily related to the apparent reflux-inducedcause of Barrett's epithelium in continuity with the car-dia and gastric mucosa.

Patient PopulationThe records of the Esophageal Laboratory in the De-

partment of Surgery at the University of Chicago Med-ical Center between the yrs 1974-82 were reviewed toidentify patients with either benign or malignant typesof Barrett's esophagus. Based upon radiographic, en-doscopic, or biopsy material, 29 patients had been seenwho were initially identified as having benign Barrett'sesophagus. Among these, 23 were accepted as meetingthe stringent definition described above and having thecondition proved by biopsies of columnar epitheliummore than 3 cm above the junction oftubular esophaguswith stomach. Six cases were excluded because of un-certainty as to the location of biopsies or endoscopicfindings. Additional cases in which columnar epitheliumextended into the esophagus after esophagectomy andesophagogastrostomy were also excluded.

During the same time, 29 patients were seen with

Barrett 'sEsophagus

and Hiatal hernia

Barrett 'sEsophagus

and adenocarcinoma

FIG. 1. Diagrams illustratingthe definitions ofbenign andmalignant epithelium. Re-gardless of the presence ofan hiatal hernia, the colum-nar lining of the tubular >3cmesophagus must extend atleast 3 cm above the junc-tion with the stomach.

Barrett 'sEsophagus

Vol. 198 * No. 4

SKINNER AND OTHERS

adenocarcinoma of the esophagus in whom the cancerwas suspected after surgery to arise from Barrett's epi-thelium. Among these, 20 cases were accepted as meet-ing the definition ofadenocarcinoma arising in Barrett'sepithelium following detailed analysis of the patholog-ical specimens. Using these precise definitions, the clin-ical records of the 23 patients having benign Barrett'sesophagus were compared with the clinical data among20 patients having adenocarcinoma arising in an esoph-agus which had benign Barrett's epithelium elsewherein the specimen.

Statistical MethodFor comparison of measurements between different

groups, a normal probability plot was performed. Whenthe data was not normally distributed, the Mann-Whit-ney nonparametric test was used, otherwise Student'st-test for unpaired data was applied. In each evaluation,both the t-test and Mann-Whitney were performed, andin every evaluation they gave the same results concern-ing significance. For nominal data, the chi square testwas used when the total of the table was more than 40.For smaller numbers, Fisher's exact probability test hasbeen applied.'2

Comparisons ofBenign and Malignant CasesAge, sex, race, and habits. Among the 23 patients with

benign columnar lining in the esophagus, the mean agewas 52 yrs (range 13-78) and the median age was 55.Among the malignant group, the average of60 yrs (range45-78) was older (p = .052 by Student's t-test), but themedian age of 59.5 yrs was not significantly differentfrom the benign group (Mann-Whitney nonparametrictest).

Barrett's esophagus was observed in 9 women and 34men. The incidence of malignancy was high amongmen, as 18 of the 20 cases of cancer occurred in mencompared to 16 of 23 benign cases being in men(p = 0.1).

Barrett's esophagus was a disease occurring over-whelmingly in white patients. All 20 cases of definitemalignant Barrett's esophagus, and all 9 of the addi-tional cases considered as possible malignant Barrett'sesophagus but excluded based on pathological analysis,occurred in Caucasians. Among the benign cases, 21occurred in whites and 2 in blacks; all 6 patients ex-cluded as not meeting the stringent definition were alsowhite. This predominance of the disease in white pa-tients was in contrast to the racial mixture of our overallsurgical patient population which has been approxi-mately one-third black, and our patients with squamous

cell carcinoma of the esophagus, 30% of whom wereblack.

In reviewing social habits, 10 of 23 patients with be-nign Barrett's esophagus had a smoking history. Amongthe malignant group, the incidence of smoking wasgreater, in that 16 out of 20 smoked two or more packsper day (p < .025). Among the male patients, the cor-relation of smoking with malignancy was even morestriking in that 16 of 18 men with cancer smoked, com-pared to 8 of 16 with benign Barrett's (p < .025).

Alcoholic beverages were ingested "socially" or moreregularly by ten people having benign Barrett's esoph-agus, and by 13 in the malignant group. The differencewas not significant.

Symptoms. The important role of gastroesophagealreflux in Barrett's esophagus was indicated by the ob-servation that all but 2 of the 23 patients with benignBarrett's complained ofeither heartburn or regurgitationaggravated by postural change. In addition, 18 of thebenign cases experienced dysphagia, 8 had upper gas-trointestinal bleeding, and 4 had odynophagia. Amongthe malignant group, 13 out of 20 complained of heart-burn and/or regurgitation, but the frequency of thesesymptoms was significantly less (p < .05) than in thebenign group (Fig. 2). In the cancer group, 15 com-plained of dysphagia, 3 bleeding, and 5 odynophagia.The length of the history of esophageal complaints var-ied from 5 mnths to 29 yrs in the benign group (mean12, median 10 yrs). This did not differ significantly fromthe duration of history ofesophageal disorder in patientswith malignant disease, which varied from 0 to 40 yrswith a mean of 8.7 yrs and median of 2 yrs.

Radiology. A hiatal hernia was identified radiograph-ically in 20 of the 23 patients with benign disease andin 16 of 20 with malignant disease. In the latter group,assessment of the hiatus could not be made in two be-cause ofthe obstruction caused by the tumor. A stricturewas identified by the radiologist during barium swallowin 14 of 23 patients with benign disease and in 6 of 20with malignant disease. A mass or narrowing, suggestingcarcinoma, was seen in 11 patients. Three patients hav-ing adenocarcinoma had neither stricture nor mass seenradiographically. Among the benign cases, an esopha-geal ulcer was observed in 9 patients. Except for thepresence ofa bulky tumor mass in some ofthe malignantcases, radiologists could not differentiate with certaintybetween patients with Barrett's esophagus having benignor malignant epithelium. In patients without either ulceror stricture, the diagnosis was frequently not madeat all.

Esophagealfunction tests. Twenty-two of the 23 pa-tients with benign Barrett's epithelium were studied

556 Ann. Surg. * October 1983

BARRETT'S ESOPHAGUS 557

FIG. 2. The frequencyof esophageal symptomsamong the 23 patients withbenign and 20 patients withmalignant Barrett's esopha-gus are shown. Differencesin heartburn and regurgita-tion are significant by thechi square test.

20

0*.- 1 8E 1 6

.

1 4.0 12Ez 10

8642

#Benign n=23

Heartburn Regurgitation Dysphag

#Malignant n=20

jia Bleeding Odynophogia

p< 0.025 p

Ann. Surg. * October 1983SKINNER AND OTHERS

FIG. 4. Cardia type of epithelium in Barrett's esophagus (right) contrastedwith intestinal mucosa (left). Note the pyloric glands in the lower partof the figure (X O00).

FIG. 5. Intestinal type metaplasia in Barrett's esophagus. Note the mucindepletion in the crypt on the right, the nuclei of which are enlarged,hyperchromatic, and may represent the earliest stages of neoplastictransformation. However, as this falls short of unequivocal dysplasia,it is categorized as indefinite for dysplasia, unknown (X300).

TABLE 1. Types ofEpithelium in Barrett's Esophagus

Benign Malignant

IT* 7 12CTt 4 0FTf 0 0IT + CT 4 6IT + CT 2 0CT + FT 2 0IT + CT + FT 4 2

*IT = intestinal type. tCT = cardia type. tFT = fundic type.

found in association with IT in two patients, CT in twopatients, and all three cell types were present in fourpatients.Among the malignant cases in which more thorough

pathological examination could be made in the resectedspecimens, IT epithelium was identified in all 20 spec-imens, CT in the tubular esophagus 3 or more cm abovethe junction in eight specimens, and FT in two. In 12of the specimens, IT was the only type epithelium iden-tified, IT and CT were found together in six, and allthree types were identified in two specimens. There wereno significant differences in the patterns of epitheliumidentified between the benign and malignant cases. Itappeared that the specialized intestinal type epithelium,featuring goblet cells, was the hallmark of Barrett'sesophagus, particularly in patients at risk to developcarcinoma.

Epithelial dysplasia. All biopsies and resected speci-mens were carefully reviewed to identify dysplasia in theBarrett's epithelium. Dysplasia was defined as being "anunequivocal neoplastic alteration of the glandular (co-lumnar-lined) esophagus." It should be stressed thatsuch dysplastic epithelium might not only be a markeror precursor ofcarcinoma, but might itselfbe malignantand associated with direct invasion into the underlyingtissue. Dysplasia was graded (Table 2) using the criteriaestablished for inflammatory bowel disease.'7 Mucosawas scored as negative for dysplasia if the nuclearchanges of the metaplastic epithelium were similar tothose found in the same epithelium in its normal ana-

TABLE 2. Classification ofDysplasia in the Mucosaof "Barrett's" Epithelium

Negative for dysplasiaIndefinite for dysplasiaa) Probably negativeb) Unknownc) Probably positivePositive for dysplasiaa) Low gradeb) High grade

558

BARRETT'S ESOPHAGUS

tomic location (e.g., stomach, intestine). Actively regen-erating epithelium was considered to be negative fordysplasia. If dysplastic nuclei were largely confined tothe basal parts of the cells, it was arbitrarily called lowgrade (Fig. 6). More severe changes with nuclei regularlyapproaching the upper pole of the cells and all changes,up to and including carcinoma-in situ, were called highgrade (Fig. 7). All epithelia not falling into either un-equivocally positive or unequivocally negative weregraded as indefinite for dysplasia. Within this category,changes thought most likely to represent the results ofactive inflammation were termed 'indefinite for dyspla-sia, probably negative,' if most likely but not unequiv-ocally to represent a neoplastic process they were termed'indefinite for dysplasia, probably positive.' Remainingbiopsies were categorized as 'indefinite for dysplasia,unknown.' When a series of changes, or multiple biop-sies were encountered showing more than one of thesecategories, the most severe was utilized.Among the biopsies from the 23 benign cases, definite

low grade dysplasia was present in two with IT epithe-lium. In one patient with a stricture, this was treated byesophagectomy and colon interposition. The secondpatient refused esophagectomy and underwent a BelseyMark IV anti-reflux repair. Biopsies obtained 7 yrs latershowed only changes indefinite for dysplasia based uponinspection of 8 biopsies. Five other benign cases hadbiopsies showing indefinite, probably negative criteriafor dysplasia, four in IT, and one in FT epithelium.

In the malignant cases, zones of benign Barrett's ep-ithelium were found in all resected specimens. Dysplasiaof high grade was found in IT epithelium in seven spec-imens, low grade dysplasia in IT of nine specimens, andno dysplasia in the Barrett's epithelium was noted ad-jacent to the adenocarcinoma in four specimens. Thehigh incidence of dysplasia in the IT epithelium amongthe malignant cases compared to the benign cases un-derscored the seriousness with which dysplasia shouldbe taken as a predictor of malignant degeneration inBarrett's epithelium.Among the 18 patients in whom definite low grade

or high grade dysplasia was seen, 16 had adenocarci-noma. In three patients undergoing esophagectomy, dys-plasia, but not frank carcinoma, was found in the pre-operative biopsies. Two of these patients had invasivecarcinoma in the resected specimen, and one of thesehad a positive esophageal cytology. In one additional caseof ectopic gastric epithelium in the cervical esophagus,not in this series, a preoperative biopsy showed only dys-plasia in IT epithelium, but the resected specimen showedmicroinvasive carcinoma. Dysplasia was the most seriousindicator ofpotential malignant degeneration in Barrett'sepithelium, and was particularly associated with the in-testinalized type of epithelium.

559

FIG. 6. Low grade dysplasia in Barrett's epithelium. Compared with thecrypt in Fig. 5, the nuclei are uniformly enlarged, hyperchromatic andthere is marked mucin depletion. Nuclei are largely confined to the cellbases, hence the categorization of low grade dysplasia (X80).

L* Its 1 _

FIG. 7. High grade dysplasia in Barrett's epithelium. Compared to Fig.6, the neoplastic nuclei regularly reach the upper part of the cells (X80).

Vol. 198 * No. 4

560 SKINNERTreatment: Benign Cases

Anti-reflux repair. From the 23 benign cases, 13 pa-tients were selected for treatment by anti-reflux repair,nine of whom had the Belsey Mark IV procedure'8 andfour who had the Nissen fundoplication operation."9 Inall but one, the operation was performed through a leftthoracotomy so that the entire esophagus could be in-spected, and lymph nodes sampled as a further precautionagainst missing malignant disease. Three ofthese patientshad undergone previous unsuccessful antireflux surgery.Eight had a stricture requiring preoperative dilatations,and five had an ulcer in the Barrett's epithelium. Ofthese13 patients, 12 have been followed to date and ten haverecently been restudied by esophageal function tests, ra-diography, and esophagoscopy with multiple biopsies.These evaluations were done from 2-7 yrs after surgery(mean 4 yrs). One patient moved away and has been lostto follow-up, one was asymptomatic and deferred testing,and one has been evaluated symptomatically to have arecurrence, but has not yet agreed to restudy. Sympto-matically, the results of antireflux surgery are shown in

Fig. 8. Eleven out of twelve are asymptomatic or haveonly minimal heartburn or regurgitation.

Compared to preoperative values, the follow-up func-tion tests showed a significant increase in distal esoph-ageal segment (DES) pressure from a mean of4.4 mmHgto 12.4 mmHg (p < .001). Before operation, no patienthad DES pressure above 9 mmHg, and after antirefluxsurgery, nine of ten patients had a DES pressure greaterthan 9 mmHg. Length of the DES high pressure zonedid not change significantly after operation. Reflux mea-sured by both the Standard Acid Reflux Test and 24 hrpH monitoring demonstrated that the frequency andduration of reflux was reduced to normal in eight oftenpatients after surgery (Fig. 9). Two patients were com-pletely asymptomatic, but had abnormal frequency andduration of reflux 4 and 7 yrs after Belsey repairs. TheSART was positive in each. In each, a stricture and ulcerseen after surgery had healed completely.By visual endoscopic inspection and biopsies, the lo-

cation of the junction of squamous and Barrett's epi-thelium has remained unchanged in eight patients. Intwo, the junction was found to be 4 and 5 cm distal to

Heartburn

pre11/12

post2/12

Regurgitation

pre9/12

post1/12

Dysphogia

pre9/12

FIG. 8. Comparison of symptoms before and 2-7 yrs after antireflux surgery in 12 patients. The one highly symptomatic patient has thus far refusedconfirmatory esophageal function tests.

3k-

2k

E04-

EI.13S0

0

000h..CD

0F-

post3/12

AND OTHERS Ann. Surg. * October 1983

I

I

BARRETT'S ESOPHAGUS

the preoperative level when endoscopy was repeatedwithin 1 yr of operation. These changes confirmed bybiopsy were thought due to surgical relocation of theesophagus rather than regression ofthe Barrett's process.On a second postoperative endoscopy 2 yrs after oper-ation, no Barrett's epithelium was identified by biopsyin one of these patients.

Extension of patchy zones of squamous epitheliuminto regions ofpreviously columnar-lined esophagus wasseen in two patients. By visual endoscopic examination,this regenerating squamous epithelium appeared as dis-tinct thin patches of opaque epithelium overlying thereddened residual columnar epithelium. This was con-firmed by multiple biopsies which showed superficialnormal squamous epithelium with residual glandularepithelium beneath (Fig. 10). To our knowledge, thishas not been described previously. In both patients, re-flux had been totally corrected by Belsey antireflux re-pairs. One patient was a nonsmoker, the other hadstopped smoking at the time of his antireflux procedure.Among five patients with no dysplasia on preoperative

biopsies, four still showed no dysplasia, but one asymp-tomatic patient with positive reflux tests progressed fromno dysplasia to an indefinite, probably negative gradeof dysplasia. Of the four patients with indefinite, prob-ably negative dysplasia before surgery, three have hadtheir reflux corrected and two continued to have indef-inite dysplasia in postoperative biopsies, while one nowhas no dysplasia. The fourth who has persistent asymp-tomatic reflux and continues to smoke has progressedto low grade dysplasia 4 yrs after surgery. The one pa-tient with low grade dysplasia before a successful anti-reflux repair has shown an indefinite, probably positivestatus 7 yrs later.

Esophagectomy. Four patients were selected foresophagectomy and isoperistaltic left colon interposi-tion.20 In one, the indication was a deep penetratingBarrett's ulcer into the mediastinum, and in three theindication was stricture, complicated by dysplasia in oneand by two previous antireflux repairs in a second. Allfour patients had an excellent functional result with fol-low-ups of 2, 2, 4, and 5 yrs. One patient died recentlyof a cerebral vascular accident 5 yrs after colon inter-position but was eating well until his stroke. One patientdeveloped an ulceration at the gastro-colonic anasto-mosis requiring revision 3 yrs after colon interposition,but recovered fully and resumed eating normally.No operation. Six patients did not undergo surgery.

In three, the decision was made because of other ill-nesses, including lung cancer, recent stroke, and acutepsychosis. None had dysplasia on biopsy. One patientwas referred after a successful Belsey antireflux repair

56124 Hour Esophageal pH Monitoring

40F

cr

E

LO

C

0

.-

0-0

C

0

Ca0

0

'C0

cr

0

30k

20_

1o0

MedianRange

Mann - Whitney

.* 100E10

* a.0 8

, 60* t:

In 400

C

0

20,

pre post18.50 00-37 0-14

p< 0.005

_

)_

)_ :

pre post43.35 0.305-108.7 0-90

p

SKINNER AND OTHERS

FIG. 11. High grade dysplasia that evolved over a 3 yr period in apatient who smoked heavily but had only minimal reflux symptoms(X500).

tinued smoking, and operation has been strongly ad-vised. The other patient did not have dysplasia and hasremained asymptomatic after 4 yrs, but has not been re-evaluated.

Treatment: Malignant Cases

All 20 patients with malignant Barrett's esophagusunderwent resection. In 16, a radical en bloc resectionof the mediastinum was performed,2' and in four, pal-liative operations were undertaken because of concom-itant cardiovascular disease in two, obvious extensionof disease in one, and the surgeon's choice in one case.Among the 16 patients undergoing radical resection,there were two deaths in hospital within 30 days, andfive additional patients died within the first 6 mnths,two from pneumonia, one from a late subphrenic ab-scess, and two from metastatic cancer. Four patientsremained living and well without evident disease at 9mnths, 2 yrs, 2'/2 yrs, and 8 yrs after radical esophagec-tomy. The survival curve by actuarial analysis in the

patients undergoing radical en bloc esophagectomy foradenocarcinoma arising in Barrett's esophagus has notvaried significantly from the survival curve in 78 otherpatients undergoing the same operation for other typesof esophageal carcinoma. Among the total group of pa-tients undergoing en bloc resection for esophageal neo-plasms, the 5 yr survival determined by actuarial tableanalysis was 22%. Among the four patients undergoingpalliative resections, three died of cardiovascular com-plications within the first 30 days after operation, andone died 14 months after surgery from metastases.

Previous analysis of esophagectomy specimens con-taining carcinoma has demonstrated that spread tolymph nodes or full thickness wall penetration were thetwo most significant factors determining prognosis.22Among the 16 patients undergoing radical en bloc re-section, five had no positive lymph nodes, and threehave survived over 2 yrs without evident disease. Onedied of an early myocardial infarction, and the seconddied of pneumonia and respiratory failure 6 wks aftersurgery. Five patients had no esophageal wall penetra-tion, including the three long-term survivors, and thepatient with negative lymph nodes who died from pneu-monia. The fifth patient in this group had five positivelymph nodes and died of widely disseminated cancerwithin 3 months. The combination of both negativelymph nodes and lack of wall penetration appeared tobe significant in predicting long-term survival after suc-cessful operations for cancer arising in Barrett's esoph-agus, as in other types of esophageal carcinoma. Thedismal results in patients with more advanced diseasestressed the importance of close follow-up in patientshaving benign Barrett's esophagus, and the necessity oftreating patients with dysplasia aggressively.

Discussion

Barrett's esophagus is no longer a medical rarity orcuriosity. To permit comparison of data among variousseries, more precise descriptions of the condition aredesirable. Previous reports do not specify an exact def-inition. The definitions proposed and used in this reportacknowledge that the squamo-columnar junction maybe irregular, and that columnar epithelium of the cardiatype extends upwards into the tubular esophagus for 1-2 cm. The squamo-columnarjunction is situated within,or just proximal to the competent distal esophagealsphincter mechanism seen endoscopically when theesophagus is dilated by insufflation of air. Observationof the squamo-columnar junction at this level is notabnormal nor indicative of Barrett's esophagus. A 3 cmzone of columnar epithelium is utilized to ensure that

562 Ann. Surg. * October 1983

BARRETT'S ESOPHAGUS

no equivocal cases are included. Measurement of morethan 3 cm of columnar epithelium lining the tubularesophagus is a more precise definition than the mea-surement of the squamo-columnar junction in centi-meters from the incisor teeth. When the latter deter-mination is used, a large hiatal hernia may lead to prox-imal displacement of the squamo-columnar junction,even in the absence of Barrett's epithelium. The use ofa precise definition may reduce some of the confusionin reports of this condition.The high incidence of severe gastroesophageal reflux

documented in the benign cases in this series confirmsmany other reports which show a high incidence of hia-tal hernia and reflux in Barrett's esophagus. Our pre-vious report of greater severity of reflux in Barrett'sesophagus cases, compared to results of reflux testing inpatients with ulcerative esophagitis but without Barrett'sepithelium indicates that the Barrett's cases have a moreadvanced abnormality ofthe cardia and acid clearance.'5This data adds further evidence to the likelihood thatpathological reflux is a cause of Barrett's esophagus. Theexperimental data of Bremner, Lynch, and Ellis pointsto reflux causing columnar transformation of injuredmucosa.23 The report of Hamilton and Yardley of Bar-rett's epithelium developing proximal to esophagogastricanastomoses after esophagectomy in patients with post-operative reflux further suggests this etiology.24 Althoughnot reported in this series, similar findings after esoph-agogastrostomy are seen in some of our cases.The absence of symptomatic heartburn and regurgi-

tation does not exclude pathological reflux in patientswith Barrett's esophagus. It is well known that there aremarked individual differences in sensitivity to reflux, andthat patterns of reflux vary in different patients.25 Per-sistent severe reflux in two ofour postoperative cases whoare asymptomatic suggests that Barrett's epithelium maynot be sensitive to acid after ulceration or strictures heal.As compared to patients with reflux esophagitis withoutBarrett's epithelium in our earlier report, patients with'benign Barrett's have less severe symptoms of heartburnand regurgitation in spite of significantly greater durationof acid reflux in the distal esophagus.'5 This lack of cor-relation between symptoms and reflux emphasizes theimportance of objective measurements of reflux in Bar-rett's esophagus cases by a method such as one of theesophageal pH reflux tests.

These data suggest that the malignant degenerationof Barrett's esophagus occurs in the presence of con-tinuing reflux and persistent smoking. All three patientswith invasive carcinoma without a stricture or large tu-mor mass studied by pH testing show severe reflux. Thir-teen ofthe 20 malignant cases complained of heartburn

and regurgitation, and the incidence of hiatal hernia issimilar to that in the benign cases. Of particular im-portance are the follow-up studies which demonstrateprogression to dysplasia in three patients who have con-tinued reflux in spite of surgical or medical therapy andwho continued to smoke; two of these patients areasymptomatic and one has minimal symptoms. Thissuggests that persistent reflux can lead to developmentof dysplasia in the absence of symptoms of heartburnand regurgitation, a change that seems to be potentiatedby smoking. It suggests an explanation for the lack ofreflux symptoms before the onset of dysphagia in someof the patients developing cancer. A corollary is foundin one patient having low grade dysplasia after surgerywho had a successful antireflux operation and gave upsmoking. Multiple biopsies have failed to confirm hisdysplasia. These findings suggest that the combinationof a successful antireflux procedure and cessation ofsmoking may be effective in preventing the subsequentdevelopment of dysplasia and carcinoma.The importance of reflux is underscored by the ob-

servation that dysplasia does not progress or regresses inall eight patients in whom reflux is documented to becorrected by pH studies after successful antireflux sur-gery. Since the two patients with continuing reflux andprogressive dysplasia after antireflux surgery are asasymptomatic as those with successful repairs, the needfor follow-up after surgery by quantitative measure-ments of reflux by pH studies is apparent. The lack ofsuch studies may explain the occasional case reports inwhich carcinoma develops at a later time, following anapparently successful antireflux repair.The 46% incidence of malignancy in this series is the

highest reported. Although there are few papers pub-lished which provide a basis for judging incidence ofcarcinoma in Barrett's esophagus, the most commonestimates are approximately a 10% risk of malignantdegeneration.26 In a large series, Naef, et al. describe 12cases of adenocarcinoma arising among 140 patientswith documented Barrett's esophagus, an incidence of8.6%.27 However, two other reports describe a higherincidence comparable to that which we observe. Radi-gan, et al. report 5 of 19 (26%) patients with Barrett'sesophagus have adenocarcinoma.28 Dees, Van Blanken-stein, and Frendel describe 13 cases of adenocarcinomaarising in Barrett's esophagus, and approximately 32cases of benign Barrett's esophagus among 5,300 en-doscopies carried out in Rotterdam.29 This is an inci-dence of approximately 30% malignancy in Barrett'sesophagus cases. The incidence of cancer in our labo-ratory is probably higher than the overall incidence inthe population of Barrett's esophagus, since ours is a

563Vol. 198 * No. 4

564 SKINNER AND OTHERS Ann. Surg October 1983surgical laboratory and it can be expected that only themore symptomatic patients are those referred. In a re-ferral practice from a broad and ill-defined base, thetotal number of Barrett's esophagus patients in the pop-ulation from which the adenocarcinomas are referredcannot be stated. The high incidence of stricture andulceration in our benign Barrett's cases points to theselected nature of these cases.

Several other risk factors for development of carci-noma are identified in this series. The higher proportionof men than women developing adenocarcinoma inBarrett's esophagus is noted by Berardi and Devaiah,who find that approximately two-thirds of all cases re-ported in the literature occur in males.4 Haggitt, et al.observe 8 out of 12 cases of Barrett's adenocarcinomain men.2' The median age of 59.5 yrs in our cases cor-responds exactly with the average ages of 58 and 60 inother reports.4'30 Our patients have an average durationof symptoms of 8.7 yrs compared to an average of 7.6yrs in Berardi and Devaiah's review of the literature.4The association with smoking is not reported previously,to our knowledge.The type of epithelium found in Barrett's esophagus

and in cases with malignant degeneration will vary de-pending upon the number of biopsies. Intestinal meta-plasia is present in all of our cases which develop ma-lignancy. These findings strongly suggest that the spe-cialized type of intestinal metaplasia is the type mostlikely to undergo malignant degeneration. However, weobserved dysplasia in all types of epithelium (IT, CT,FT). Others do not always report intestinal metaplasiain cases of adenocarcinoma arising in Barrett's esoph-agus.' No dysplasia adjacent to the carcinoma is notedin 4 of our 20 patients, possibly the result ofdestructionof dysplastic epithelium by the carcinoma.Of special interest is the finding in two patients that

squamous epithelium regenerates in zones of previousBarrett's epithelium. In the previous report by Brand,et al., the technique of demonstrating regeneration ofsquamous epithelium can be criticized because the biop-sies are obtained with suction capsules without knowl-edge of the precise location of the biopsy and its rela-tionship to the squamo-columnar junction. The findingin two of our cases that the squamo-columnar junctioncan be reset distally, shortly following antireflux repair,suggests that this can be a mechanical sequel to the sur-gery and need not represent true regeneration ofmucosa.However, the visual and biopsy findings in our two pa-tients are convincing and suggest a mechanism for squa-mous cell regeneration actually replacing areas of co-lumnar epithelium by superficial overgrowth, with sub-mergence and diminution in the glandular epithelium.

Whether the glandular epithelium will persist deep tothe squamous layers, gradually atrophy and disappearcompletely, or even persist and become malignant isunknown.

Based upon the findings in this study and availablepublished information, all patients identified to haveBarrett's esophagus by the strict definition given shouldundergo quantitative testing for gastroesophageal reflux.If abnormal reflux, intestinal type metaplasia, and anydegree of progression towards dysplasia is observed, anantireflux repair should be performed regardless ofsymptoms. Patients must remain under long-term fol-low-up and have correction of the reflux demonstrated,as well as rebiopsy of the mucosa and/or cytologicalstudies at intervals. Patients with symptomatic refluxdemonstrating zones of ulcerative esophagitis, ulcera-tion in Barrett's mucosa, or stricture should be treatedsurgically to correct the reflux. In most instances, thestricture is dilatable prior to operation and an antirefluxrepair can be expected to be successful. Patients withhigh grade dysplasia, which includes carcinoma-in situ,should be considered at high risk of having carcinomasomewhere in the Barrett's epithelium.

It should be stressed that carcinoma-in situ is the endpoint of noninvasive intraepithelial neoplasia, and thatall grades ofdysplasia, as defined here, have the potentialto give rise directly to invasive carcinoma. Further,many do so without going through the morphologicalstage of carcinoma-in situ. Both patients in this studywith high grade dysplasia that were resected had invasivecarcinoma, one of which invaded through the esopha-geal wall and had numerous lymph node metastases, theother which had only invasion into the submucosa andhas been previously reported.3' Such patients should bestrongly advised to undergo esophagectomy and recon-struction rather than run the risk ofdeveloping clinicallyinvasive carcinoma which carries a grave prognosis.

References1. Barrett NR. Chronic peptic ulcer of the oesophagus and oeso-

phagitis. Br J Surg 1950; 38:175-182.2. Allison PR, Johnstone AS. The oesophagus lined with gastric

mucous membrane. Thorax 1953; 8:87-101.3. Barrett NR. The lower esophagus lined by columnar epithelium.

Surgery 1957; 41:881-894.4. Berardi RS, Devaiah KA. Barrett's esophagus. Surg Gynecol &

Obstet 1983; 156:521-538.5. Goldman MC, Beckman RC. Barrett syndrome; case report with

discussion about concepts of pathogenesis. Gastroenterology1960; 39:104-110.

6. Mossberg SM. The columnar-lined esophagus (Barrett syndrome);an acquired condition? Gastroenterology 1966; 50:671-676.

7. Hayward J. The treatment of fibrous stricture of the oesophagusassociated with hiatal hernia. Thorax 1961; 16:45-55.

Vol. 198 * No. 4 BARRETTS ESOPHAGUS 5658. Adler RH. The lower esophagus lined by columnar epithelium;

its association with hiatal hernia, ulcer, stricture and tumor.J Thorac Cardiovasc Surg 1963; 45:13-32.

9. Morson BC, Belcher JR. Adenocarcinoma of the oesophagus andectopic gastric mucosa. Br J Cancer 1953; 6:127-130.

10. McCorkle RG, Blades B. Adenocarcinoma of the esophagus aris-ing in aberrant gastric mucosa. Am Surg 1955; 21:781-785.

11. Rector LE, Connerley ML. Aberrant mucosa in the esophagus ininfants and in children. Arch Pathol 1941; 31:285-294.

12. Hill AB. Principles ofMedical Statistics. 9th Edition. Oxford NewYork: University Press, 1971.

13. Johnson LF, DeMeester TR. Twenty-four hour pH monitoringof the distal esophagus: a quantitative measure of gastroesoph-ageal reflux. Am J Gastroenterol 1974; 62:325.

14. Skinner DB, Booth DJ. Assessment of distal esophageal functionin patients with hiatal hernia and/or gastroesophageal reflux.Ann Surg 1970; 172:627.

15. Iascone C, DeMeesterTR, Little AG, Skinner DB. Barrett's esoph-agus: functional assessment, proposed pathogenesis, and sur-gical therapy. Arch Surg 1983; 118:543-549.

16. Paull A, Trier JS, Dalton MD, et al. The histologic spectrum ofBarrett's esophagus. N Engl J Med 1976; 295:476-480.

17. Riddell RH, Goldman H, Ransohoff DF, et al. Dysplasia in in-flammatory bowel disease: standardized classification with pro-visional clinical applications. Human Pathology (in press).

18. Skinner DB, Belsey RHR. Surgical management of esophagealreflux and hiatus hernia: long-term results with 1030 patients..J Thorac Cardiovasc Surg 1967; 53:33.

19. Nissen R. Gastropexy and "fundoplication" in surgical treatmentof hiatal hernia. Amer J Dig Dis 1961; 6:954.

DIscuSSION

E. R. WOODWARD, M.D.: This is most certainly a timely presentation.Our profession generally is not adequately aware of the malignant po-tential in Barrett's mucosa. Would the authors recommend that allpatients with symptomatic gastroesophageal reflux have a thorough en-doscopy as a part oftheir initial work-up and before instituting therapy?I fear that many such patients receive, at the most, an upper G.I. x-rayexamination before treatment is started. In addition, would the authorsconsider that the presence of Barrett's mucosa is in itself sufficient in-dication for antireflux? Would this be a reasonable move for cancerprevention?

In a recent paper, the authors have pointed out a strong correlationbetween the severity ofgastroesophageal reflux and the extent of Barrett'smucosa. This certainly supports the theory that the columnar epitheliumrepresents a metaplastic change. One would expect, therefore, that surgicalcorrection of reflux would be followed by reversion to squamous mucosa.We have four patients followed 5 to 15 years wherein most of thethoracic esophagus is columnar-lined. Transabdominal Nissen fundo-plication provided symptomatic relief and healing of the ulcers andstrictures at the squamo-columnar junction. However, annual endoscopyhas shown no gross change in the Barrett's mucosa. Do you supposethere are occasional congenital cases?

DR. LAWRENCE DENBESTEN (Los Angeles, California): Dave, in ad-enocarcinoma where do you stop the resection? Frequently the Barrett'swill extend very much proximally, with significant dysplasia, and I amalways troubled as to how far one chases it.

Possibly, this is another of these diseases where multi-institutionalsurveillance might be in order. At UCLA, we have a study wherein allpatients are being evaluated in a method similar to that described, andfollowed linearly for an indefinite period. We certainly invite otherinstitutions and groups to participate and pool data with us, to get some

20. Belsey R. Reconstruction of esophagus with left colon. J ThoracCardiovasc Surg 1965; 49:33.

21. Skinner DB. En bloc resection for neoplasms of the esophagusand cardia. J Thor & Cardiovasc Surg 1983; 85:59-70.

22. Skinner DB, Dowlatshahi KD, DeMeester TR. Potentially curablecancer of the esophagus. Cancer 1982; 50(Suppl.):2571-2575.

23. Bremner CG, Lynch VP, Ellis FH Jr. Barrett's esophagus; con-genital or acquired? An experimental study of esophageal mu-cosal regeneration in the dog. Surgery 1970; 68:209-216.

24. Hamilton SR, Yardley JH. Regeneration of cardiac type mucosaand acquisition of Barrett mucosa after esophagogastrostomy.Gastroenterology 1977; 72:669-675.

25. DeMeester TR, Johnson LF, Joseph GJ, et al. Patterns of gas-troesophageal reflux in health and disease. Ann Surg 1976;184:459-470.

26. Sjogren RW Jr, Johnson LF. Barrett's esophagus: A review. AmerJ Med 1983; 74:313-321.

27. NaefAP, Savary M, Ozzello L, Pearson FG. Columnar-lined loweresophagus. Surgery 1975; 70:826-834.

28. Radigan LR, Glover JL, Shipley FE, Shoemaker RE. Barrettesophagus. Arch Surg 1977; 112:486-491.

29. Dees J, Van Blankenstein M, Frendel M. Adenocarcinoma inBarrett's esophagus; a report of 13 cases. Gastroenterology1978; 74:1119.

30. Haggitt RC, Tryzelaar J, Ellis FG, Colcher H. Adenocarcinomacomplicating columnar epithelium-lined (Barrett's) esophagus.Am J Clin Pathol 1978; 70:1-5.

31. Berenson MM, Riddell RH, Skinner DB, Freston JW. Malignanttransformation of esophageal columnar epithelium. Cancer1978; 41:554-561.

handle on the effect on regression from antireflux procedures, and alsothe effect on dysplasia, or the potential for carcinoma, when successfulantireflux procedures have been carried out.

DR. JOHN S. SPRArr (Louisville, Kentucky): In that there is a tendencyfor patients who are on potassium-losing diuretics to have earlier de-velopment of cancer from villous adenomas of the rectum, at a smallersize, I would wonder whether any of these esophagus cases were on anypotassium-losing diuretics.

DR. DAVID B. SKINNER (Closing discussion): In response to Dr.DenBesten, we believe that all of the Barrett's epithelium should beresected in the malignant cases. We treat this very much the same aswe do carcinoma of the esophagus, aim for 10 cm proximal and distalmargins when possible, and do an extensive mediastinal resection inthe favorable cases.

Dr. Spratt, we have not made the correlation with potassium-losingdiuretics. I am not aware that any of our patients have been treatedwith that, but we will certainly go home and double-check that obser-vation.

Dr. Woodward asked about endoscopy in the initial work-up of apatient with symptomatic gastroesophageal reflux. We believe patientswith reflux should be endoscoped before a decision is made about treat-ment, since the degree of ulcerative esophagitis cannot be predicted bythe type of symptoms, and patients with ulcerative esophagitis must befollowed much more closely and tightly, and probably should undergoantireflux repair early in the course of their disease. Early endoscopywill then clearly detect Barrett's ifyou are looking for it, although someof our patients have been endoscoped several times by other endoscopistsbefore we saw them, and the Barrett's was not always identified. So ithas to be looked for to be seen.

Dr. Woodward also commented about our policy for antireflux surgery