ORIGINAL CONTRIBUTIONS

Progression or Regression of Barrett’s Esophagus—Is It All in the Eye of the Beholder?Roy Dekel, M.D.,Donald E. Wakelin, M.D., Chris Wendel, M.S., Colleen Green, M.S.,Richard E. Sampliner, M.D.,Harinder S. Garewal, M.D., Ph.D., Patricia Martinez, R.N., andRonnie Fass, M.D.Section of Gastroenterology, Department of Medicine, Southern Arizona Veterans Affairs Health Care Systemand University of Arizona Health Science Center, Tucson, Arizona

OBJECTIVES: Accurate measurements of Barrett’s esopha-gus length are important in clinical follow-up as well as instudies of therapeutic intervention in Barrett’s esophagus.Our aim was to evaluate both the inter- and intraobserverreliability in measuring Barrett’s length during two consec-utive endoscopies by either the same or different experiencedendoscopists. The relationship between Barrett’s mucosalength and the absolute change in Barrett’s length measure-ments on a follow-up endoscopy was also evaluated.

METHODS: A total of 96 Barrett’s patients underwent twoconsecutive endoscopies. The diagnosis of Barrett’s esoph-agus was confirmed by the presence of intestinal metaplasiaon biopsy. The Barrett’s esophagus length was carefullymeasured and recorded during the two endoscopies. Proce-dures were performed by only two experienced endosco-pists, who were not aware of previous endoscopic measure-ments. Only patients with long-segment (�3 cm) Barrett’sesophagus were included in this study.

RESULTS: The 55 patients who had their consecutive endos-copies performed by the same endoscopist had a mean1.6-cm difference between the two measurements as com-pared to 1.4 cm in the 41 patients who had their endoscopiesperformed by different endoscopists (p � 0.3). The agree-ment between the two Barrett’s length measurements washigh in both groups, although it was slightly higher forendoscopies performed by the same endoscopist (r � 0.79vs r � 0.67). Linear regression analysis of the absolutechange in Barrett’s length between the two endoscopicmeasurements and Barrett’s mucosa length demonstrated asignificant relationship (r � 0.28,p � 0.005). For every 1-cmincrease in the mean length of Barrett’s mucosa, a 0.15-cmincrease in the absolute difference between two consecutiveendoscopic measurements of Barrett’s length was observed.

CONCLUSIONS: Consecutive measurements of Barrett’slength performed by different experienced endoscopists orby the same experienced endoscopist demonstrated a highdegree of agreement. A range of variability in Barrett’slength measurement was determined (�1.4–1.6 cm). Trueregression or progression of Barrett’s mucosa should beconsidered only if the change is greater than the range of

variability. In addition, endoscopists should be well awarethat the longer the Barrett’s mucosa the greater the absolutedifference in Barrett’s length measurement on follow-upendoscopy. (Am J Gastroenterol 2003;98:2612-2615. ©2003 by Am. Coll. of Gastroenterology)

INTRODUCTION

Barrett’s esophagus is defined by the replacement of normalesophageal squamous epithelium by specialized intestinalepithelium. This metaplastic change has been long consid-ered to be the result of gastroesophageal reflux (1). Theimmense interest in Barrett’s esophagus stems from itsmalignant potential. Studies have demonstrated that Bar-rett’s mucosa is a precancerous lesion for esophageal ade-nocarcinoma, which is currently one of the fastest risingcancers in Western countries (2, 3).

Diagnosis of Barrett’s mucosa depends on identifying theabnormal columnar epithelium extending into the distal esoph-agus. Multiple biopsies are necessary to confirm the presenceof intestinal metaplasia, the hallmark of Barrett’s esophagus.

Once Barrett’s esophagus is suspected, it is essential todescribe its length. When measuring the length of the Bar-rett’s mucosa, two anatomic landmarks need to be accu-rately identified. The first one is the esophago-gastric junc-tion, which is defined by the proximal margin of the gastricfolds. The second one is the squamo-columnar junction (Zline), which represents the transition point from squamousto columnar epithelium. Normally these two anatomic land-marks coincide, but in Barrett’s esophagus patients thesquamo-columnar junction is proximally displaced. Thelength of Barrett’s esophagus consists of the distance be-tween these two important endoscopic landmarks (4).

Endoscopic evaluation of Barrett’s esophagus length re-quires skill and is influenced by various intraluminal factors.The squamo-columnar junction may be obscured by peri-stalsis, respiratory movements, tortuous esophagus, or ero-sive esophagitis. Presence of hiatal hernia may interfere indefining the exact location of the esophago-gastric junction.Over- or under-insufflation of air, faulty measurements(such as from the lips instead of from the incisors), andinappropriate biopsy technique all may compromise proper

THE AMERICAN JOURNAL OF GASTROENTEROLOGY Vol. 98, No. 12, 2003© 2003 by Am. Coll. of Gastroenterology ISSN 0002-9270/03/$30.00Published by Elsevier Inc. doi:10.1016/S0002-9270(03)00665-8

length measurement (5). More experienced endoscopistsdemonstrate a higher diagnostic accuracy (6).

At present, most authorities believe that Barrett’s esophagusevolves rapidly to its maximal length with little change after-ward (7). To date, there is no proven medical therapy that caneliminate Barrett’s epithelium. However, recognizing gastro-esophageal reflux disease as the underlying etiology for Bar-rett’s esophagus has led to the suggestion that aggressive acidsuppression, mainly through the use of high dose proton pumpinhibitors (PPI), may result in regression or even completeelimination of Barrett’s esophagus. Several studies have at-tempted to evaluate the effect of PPI treatment on Barrett’sesophagus length. These studies differed in study population,type, dosage of PPI, and treatment duration. Results variedfrom no change to a significant change in Barrett’s esophaguslength in the range of 0.6–2.6 cm (8–14).

As previously mentioned, measuring Barrett’s esophaguslength is an integral part of routine clinical follow-up. Ac-curate measurements are essential for learning about thenatural history of the lesion and for developing better diag-nostic techniques as well as new therapeutic options. Inac-curate measurements may lead to possible clinical pitfalls,such as false impressions of disease regression or progres-sion as well as overlooking Barrett’s esophagus or evenadenocarcinoma of the esophagus during biopsy.

Our study had dual aims. The first was to evaluate boththe inter- and intraobserver reliability of Barrett’s esophaguslength measurements by experienced endoscopists. The sec-ond was to determine the relationship between Barrett’smucosa length and the absolute change in Barrett’s esoph-agus length on subsequent consecutive endoscopy.

MATERIALS AND METHODS

Patient Population and Data CollectionA total of 96 consecutive Barrett’s esophagus patient whopresented to the Southern Arizona Veterans Affairs HealthCare System GI Laboratory were included in this study.

Only patients with long segment Barrett’s esophagus (de-fined by length �3 cm) who underwent two endoscopies wereincluded in this study. Diagnosis of Barrett’s esophagus wasconfirmed by the presence of intestinal metaplasia on multiplebiopsies. The lengths of Barrett’s mucosa at the last and pre-ceding endoscopies were recorded. In addition, the dates of theendoscopies and the identity of the physicians who performedthe examinations were documented. Endoscopists were notaware of previous endoscopic measurements.

Exclusion criteria included lack of two consecutive endos-copies, presence of short segment Barrett’s esophagus (�3cm), presence of esophageal cancer, history of any form ofendoscopic ablative therapy, and history of esophageal surgery.

Upper EndoscopyAfter an overnight fast, patients were placed in the leftlateral position. Sedation was achieved with a combinationof midazolam (Roche, Nutley, NJ) and meperidine (SanofiWinthrop, New York, NY). The endoscope (GIF 160, GIF

140, and GIFQ160, Olympus, Japan) was inserted throughthe mouth and into the esophagus. The distal portion of theesophagus was carefully evaluated to determine the pres-ence of mucosal abnormality. The stomach and duodenumwere also routinely inspected.

The distal portion of the esophagus was examined todetermine the presence of Barrett’s esophagus, which wasindicated by red color and velvet-like texture extending intothe esophagus. Barrett’s esophagus was defined as the pres-ence of intestinal metaplasia on biopsy from mucosa thatwas abnormal in appearance. Measurement of Barrett’sesophagus length was performed from the proximal marginof continuous Barrett’s epithelium to the end of the tubularesophagus or the proximal margin of the hiatal hernia folds(15). In patients with suspected long segment Barrett’sesophagus, four quadrant biopsy samples, every 2 cm start-ing at the distal end, were obtained to identify intestinalmetaplasia and to exclude dysplasia.

All specimens were placed immediately in a 10% buff-ered formalin solution, embedded in paraffin, and stainedwith hematoxylin and eosin combined with Alcian blue atpH 2.5. A designated pathologist reviewed the specimens todetermine whether intestinal metaplasia was present.

Statistical AnalysisResults are shown as mean � SD. Characteristics of the twogroups including age, sex, and follow-up duration were com-pared by t test and �2 test where appropriate. Linear regressionwas used to evaluate relationship between difference in lengthand mean length after inspection of difference by length cate-gories for linearity. We analyzed inter-rater reliability to char-acterize the consistency and repeatability of the Barrett’s lengthmeasurement between two experienced endoscopists. Becausethe measurement of length was continuous, inter-rater reliabil-ity was evaluated with a Pearson correlation coefficient.

RESULTSA total of 55 patients had two endoscopies performed by thesame endoscopist (one operator only), and 40 patients hadtwo endoscopies performed by two different endoscopists.All endoscopists were experienced, well trained, and famil-iar with the technique of Barrett’s measurements as previ-ously detailed. The study population was composed pre-dominantly of men, and both groups were similar for age atfirst endoscopy as well as sex. In addition, both groups hadsimilar mean duration of endoscopic follow-up, which was18.6 months for the group whose procedures were carriedout by the same endoscopist and 15.7 months for that withdifferent endoscopists (p � 0.41) (Table 1). The reliabilityfor the entire cohort was also high (r � 0.75).

The absolute difference in measurements of Barrett’sesophagus length from the first to the second endoscopy (�� 1o measurement of Barrett’s esophagus length –2o mea-surement of Barrett’s length) did not differ between the twogroups. Both follow-ups (i.e., with the same endoscopist ordifferent endoscopists) showed high levels of reliability,although it was somewhat higher in the group whose pro-

2613AJG – December, 2003 Regression or Progression of Barrett’s Esophagus

cedures were carried out by the same endoscopist (r � 0.79vs r � 0.67) (Table 2).

The absolute mean difference between the first and secondendoscopies was 1.6 cm for measurements made by the sameendoscopist and 1.4 cm for measurements made by differentendoscopists (p � 0.3). It is apparent that regardless of themethod of follow-up, in both situations a Barrett’s esophaguslength change in the range of �1.4–1.6 cm is expected.

Linear regression analysis between the absolute value ofmeasurement change (between the first and second endos-copy) and Barrett’s esophagus length showed a significantrelationship (r � 0.28, p � 0.005). For each 1-cm increasein mean length there was a 0.15-cm increase in the differ-ence between the first and second measurement (Fig. 1).Thus, the longer the segment of Barrett’s mucosa, thegreater the absolute difference in Barrett’s length measure-ments between two consecutive endoscopies.

Interestingly, summation of the positive and negativedifferences (increase vs decrease in Barrett’s length on theconsecutive endoscopy) in Barrett’s length showed that thetotal length of Barrett’s mucosa remained unchanged (pos-itive change of 70.5 cm vs negative change of 70.5 cm).

DISCUSSION

The length of Barrett’s mucosa must be addressed either inclinical practice or while conducting clinical research. De-termining Barrett’s esophagus length is important not just asa descriptive tool but also for classifying it in clinical terms,

studying its natural history, and evaluating regression orprogression during interventional studies using antirefluxmedications or endoscopic ablative therapy.

Our study evaluated the impact of interobserver and in-traobserver reliability on Barrett’s length measurement. Themean difference in length between two consecutive endos-copies was 1.6 cm for patients who were followed by thesame endoscopist and 1.4 cm for patients followed by dif-ferent endoscopists. Although the mean difference in lengthbetween these two groups was not statistically significant,these findings demonstrate a consistent range of variabilityin both groups. As previously mentioned, different studiesthat evaluated acid suppressive therapy for Barrett’s esoph-agus regression demonstrated a decrease in Barrett’s esoph-agus length ranging from 0 to 2.6 cm (8–14). In a recentstudy, for example, a regression of 1 to 2 cm in Barrett’sesophagus length was demonstrated in six (67%) of ninepatients undergoing treatment with omeprazole or lansopra-zole twice daily plus ranitidine at bedtime for a mean of 16months (14). In a different study, 13 (68%) of 19 patientsthat demonstrated a change in Barrett’s mucosa length hada regression of �2 cm (12). Our findings suggest that thesetypes of reports in studies that assessed regression of Bar-rett’s mucosa may still be within the range of inter- orintraobserver variability. Thus, this study points out for thefirst time to a regression/progression threshold for futurestudies. Claims of presumed progression or regression ofBarrett’s esophagus should demonstrate a change that is atleast �1.6 cm (inter- and intraobserver range of variability).

Another important finding of our study is that there wasa significant correlation between Barrett’s segment lengthand the absolute mean change in length measurements onconsecutive endoscopies. The longer the Barrett’s mucosa,the greater the absolute change in Barrett’s esophagus mea-surement observed. For example, in the previously men-tioned studies, a high percentage of patients with a relativelylong segment of Barrett’s tissue (�8 cm, 44.4% in one studyand 35% in the other) were included (12, 14).

Table 1. Demographic Characteristics of Patients

SameEndoscopist

(N � 55)

DifferentEndoscopists

(N � 40) p Value

Age, yr* 66.6 � 11.2 66.6 � 12.4 0.97Range, yr* 33–86 27–86

Sex (male/female) 52/3 38/2 0.922Follow-up (mo)* 18.6 � 18.9 15.71 � 14.9 0.41

* Data are shown as mean � SD.

Table 2. Barrett’s Esophagus Measurements in Two ConsecutiveEndoscopies

SameEndoscopist*

(N � 55)

DifferentEndoscopists*

(N � 40) p Value

Barrett’s segmentlength at firstendoscopy

7.6 � 2.7 6.8 � 2.6 0.17

Barrett’s segmentlength at secondendoscopy

7.5 � 2.9 6.9 � 2.8 0.3

Difference betweenfirst and secondmeasurements(cm)

1.6 � 1.4 1.4 � 1.2 0.3

Reliability* 0.79 0.67

Data are shown as mean � SD.* Reliability of consecutive measurements (level of correlation between first and

second measurements) was evaluated using the Pearson correlation coefficient.

Figure 1. Regression analysis was performed between the absolutedifference in Barrett’s esophagus length measurement (betweenthe first and second endoscopy) and the mean length of Barrett’smucosa. Significant relationship was found between these twoparameters (r � 0.28, p � 0.005). For every 1-cm increase in meanBarrett’s esophagus length there is a 0.15-cm change in measure-ment of Barrett’s esophagus length on follow-up endoscopy.

2614 Dekel et al. AJG – Vol. 98, No. 12, 2003

Regression analysis of the difference between the first andsecond endoscopic measurements of Barrett’s length showedthat for every 1-cm increase in the mean length of Barrett’sesophagus, one can expect a 0.15-cm increase in the changebetween two consecutive measurements. For example, whenmeasuring a 4-cm vs 10-cm Barrett’s esophagus, one wouldexpect a 0.9-cm increase in absolute difference between twoconsecutive measurements of the latter length as compared tothat of the former. The increase in absolute difference is relatedsolely to the length of the measured Barrett’s mucosa. Futurestudies evaluating changes in Barrett’s esophagus lengthshould be aware of this phenomenon. In addition, cliniciansand researchers should specify Barrett’s esophagus length notonly as short vs long segment but also according to the absolutelength of Barrett’s mucosa.

Both measurements made by the same endoscopist anddifferent endoscopists showed high reliability. Measure-ments made by the same endoscopist revealed a somewhathigher level of reliability. This finding suggests that toprovide a reliable follow-up as well as Barrett’s esophaguslength assessment, there is no need for all endoscopies to beperformed by the same endoscopist provided that all endos-copists are trained in careful assessment of Barrett’s esoph-agus length by criteria mentioned.

Summation of the positive and negative differences inBarrett’s esophagus length showed that the total length ofBarrett’s mucosa remained unchanged. This finding sug-gests that during Barrett’s esophagus follow-up, some mea-surements might show an increase whereas others show adecrease in Barrett’s length. However, all these changesseem to be within the range of measurement variability, andthe actual length of Barrett’s mucosa remains unchanged.

Our study population included mainly veterans in theUnited States, which explains the male predominance. Wedo not believe that the very small number of women in ourstudy compromises its generalizability. In general, Barrett’sesophagus measurements do not seem to be influenced bygender. A potentially better study design for evaluatingintraobserver reliability would be to have two differentendoscopists assess the same patient at the same time. How-ever, this study design eliminates the follow-up perspectiveand creates an “artificial” situation.

In summary, we found no significant difference inmeasurements of Barrett’ s esophagus length between twoconsecutive endoscopies performed by a single experi-enced endoscopist vs different endoscopists similarly ex-perienced. There is a “ range of variability” of up to 1.6cm when measuring Barrett’ s esophagus length. Conse-quently, only absolute changes in Barrett’ s esophaguslength that are greater than this range of variabilityshould be regarded as a true regression or progression ofBarrett’ s mucosa. Finally, there is a close relationshipbetween the length of Barrett’ s esophagus that is beingmeasured and the absolute difference observed in mea-surement of Barrett’ s esophagus length on a consecutiveupper endoscopy. This finding suggests that in future

studies assessing regression or progression of Barrett’ smucosa, investigators should expect a greater range ofvariability in Barrett’ s esophagus length measurement inpatients with longer Barrett’ s epithelium.

ACKNOWLEDGMENTThis study was supported in part by a fellowship grant fromThe American Physicians Fellowship for Medicine in Israel(to R.D).

Reprint requests and correspondence: Ronnie Fass, M.D.,F.A.C.P., F.A.C.G., Southern Arizona VA Health Care System,3601 S. 6th Avenue (1-111G-1), Tucson, AZ 85723.

Received Mar. 17, 2003; accepted June 12, 2003.

REFERENCES

1. Sampliner RE, The Practice Parameters Committee of theAmerican College of Gastroenterology. Updated guidelinesfor the diagnosis, surveillance, and therapy of Barrett’s esoph-agus. Am J Gastroenterol 2002;97:1888–95.

2. Blot WJ, Devesa SS, Kneller RW, Fraumeni JF Jr. Risingincidence of adenocarcinoma of the esophagus and gastriccardia. JAMA 1991;265:1287–9.

3. Devesa SS, Blot WJ, Fraumeni JF, Jr. Changing patterns in theincidence of esophageal and gastric carcinoma in the UnitedStates. Cancer 1998;83:2049–53.

4. Spechler SJ, Goyal RK. Barrett’s esophagus. N Engl J Med1986;3185:362–71.

5. Fass R, Sampliner RE. Gastroesophageal reflux disease andBarrett’s esophagus. In: DiMarino AJ Jr, Benjamin SB, eds.Gastrointestinal disease, an endoscopic approach. 2nd ed.Thorofare, NJ: Slack, 2002:191–214.

6. Padda S, Ramirez FC. Accuracy in the diagnosis of short-segment Barrett’s esophagus: The role of endoscopic experi-ence. Gastrointest Endosc 2001;54:605–8.

7. Cameron AJ, Lomboy CT. Barrett’s esophagus: Age, preva-lence, and extent of columnar epithelium. Gastroenterology1992;103:1241–5.

8. Neumann CS, Iqbal TH, Cooper BT. Long term continuousomeprazole treatment of patients with Barrett’s oesophagus.Aliment Pharmacol Ther 1995;9:451–4.

9. Sharma P, Sampliner RE, Camargo E. Normalization ofesophageal pH with high-dose proton pump inhibitor therapydoes not result in regression of Barrett’s esophagus. Am JGastroenterol 1997;92:582–5.

10. Malesci A, Savarino V, Zentilin P, et al. Partial regression ofBarrett’s esophagus by long-term therapy with high-dose ome-prazole. Gastrointest Endosc 1996;44:700–5.

11. Wilkinson SP, Biddlestone L, Gore S, Shepherd NA. Regres-sion of columnar-lined (Barrett’s) oesophagus with omepra-zole 40 mg daily: Results of 5 years of continuous therapy.Aliment Pharmacol Ther 1999;13:1205–9.

12. Gore S, Healey CJ, Sutton R, et al. Regression of columnarlined (Barrett’s) oesophagus with continuous omeprazole ther-apy. Aliment Pharmacol Ther 1993;7:623–8.

13. Peters FT, Ganesh S, Kuipers EJ, et al. Endoscopic regressionof Barrett’s oesophagus during omeprazole treatment: A ran-domised double blind study. Gut 2000;45:489–94.

14. Srinivasan R, Katz PO, Ramakrishnan A, et al. Maximal acidreflux control for Barrett’s oesophagus: Feasible and effective.Aliment Pharmacol Ther 2001;15:519–24.

15. Sampliner RE, Fennerty B, Garewal HS. Reversal of Barrett’sesophagus with acid suppression and multipolar electrocoagula-tion: Preliminary results. Gastrointest Endosc 1996;44: 532–5.

2615AJG – December, 2003 Regression or Progression of Barrett’s Esophagus