1472 WANG ET AL GASTROENTEROLOGY Vol. 128, No. 5( 3 cm) Barrett’s esophagus.5 A multicenter study of cancer should be performed at the time of diagnosis of536 subjects found that long-segment Barrett’s esopha- the head-and-neck cancer. Extensive alcohol consump-gus was only present in 0.36% of subjects, although tion alone has been found to be an important criterion forshort-segment Barrett’s esophagus was found in 5.6%.6 screening in Asian patient populations.25 Partial gastrec-This study did ﬁnd that patients with heartburn had a tomies have been associated with an increased incidencesigniﬁcantly higher prevalence of long-segment Barrett’s of squamous cell cancers of the esophagus; these reportsesophagus than those who did not. This is signiﬁcant stem from areas of the world where squamous cell cancerbecause past studies have shown that the degree of neo- is commonly found, and there is no convincing evidenceplasia found in Barrett’s esophagus appears to be corre- to support the screening of patients in the Unitedlated with the length of Barrett’s esophagus.7 Other States.26 The occurrence of squamous cell cancer is notgroups that may be at risk would include those with decreasing in minority populations in the United Statesfamilial occurrence of adenocarcinoma, but this has and remains the most frequent form of esophageal canceronly been described in a limited number of families. in black and Hispanic populations.27,28 Other conditionsThe cost-efﬁcacy of screening family members because such as Plummer–Vinson webs have also been identiﬁedof a history of Barrett’s esophagus has not been as being associated with an increased risk for squamousdemonstrated.8 –12 cell cancers, but the decreasing frequency of these webs Screening methods for detection of Barrett’s esophagus makes this almost a historical footnote in Westernhave included standard endoscopy, unsedated endoscopy countries.29with ultrathin endoscopes, catheter-based cytology, and Screening methods such as cytologic balloons orballoon cytology.13–16 Although preliminary studies in- sponges for screening for squamous cell cancers have beendicate some promise with these technologies in terms of extensively tested in Asia. Balloons appear to be betterscreening for adenocarcinoma, there have not been any than sponges, with an increased sensitivity of 44% com-deﬁnitive trials to allow recommendation of these pared with 18% for sponges.30 This technology has beentechniques. used as the primary screening tool in high-risk areas of Squamous cell cancer. Screening for squamous China.31cell cancer in the general population of the United States Summary of Evidence—There is level III evidence thatcannot be justiﬁed because of the low incidence of this endoscopic screening of male white patients older thanform of cancer. However, speciﬁc subgroups may be 50 years of age with symptoms of gastroesophageal reﬂuxidentiﬁed that warrant screening endoscopy for squa- may be cost-effective. Patients with tylosis, lye-inducedmous cell cancer. The most likely to beneﬁt from screen- strictures, or Fanconi’s anemia would beneﬁt froming would be those who have tylosis, which is a genetic screening endoscopy for squamous cell cancers (level IIIdefect in the 17q25 region that is found in patients with evidence). In addition, patients with long-term tobaccothickened palms and soles.17,18 This group is likely to and alcohol use, achalasia, or prior head-and-neck cancersdevelop cancer by the age of 65 years and should undergo could be considered for screening, depending on theirscreening for squamous cell cancer. Patients with lye- other risk factors (level III evidence). The interval forinduced or caustic strictures develop cancers approxi- surveillance of these patients has not been established,mately 46 years after ingestion and have been described but yearly investigations would seem to be reasonableto have an 8% incidence of cancer.19,20 The developmentof cancer in patients with long-standing achalasia is (level IV evidence).infrequent, especially in women, and it is unclear if Surveillance for Esophageal Cancersurveillance is warranted in these patients.21 Most pa-tients with achalasia are found to have prevalent cancers Signiﬁcance of surveillance. The importance ofbecause they are usually diagnosed when obstruction and Barrett’s esophagus derives from its association with esoph-esophageal dilation are found.22 Fanconi’s anemia has ageal adenocarcinoma.32,33 Approximately 5%–10% of pa-also been associated with the development of esophageal tients diagnosed with Barrett’s esophagus may developcancer and may become more common as patients survive esophageal adenocarcinoma based on older studies.32,34,35longer after bone marrow transplantation.23 Patients The incidence of cancer will most likely decrease withwith existing aerodigestive tumors, especially those of the increasing detection of Barrett’s esophagus throughthe oral cavity, who have had long-term extensive expo- screening programs. Surveillance of Barrett’s esophagussure to alcohol and tobacco might beneﬁt from screening can detect dysplasia in Barrett’s esophagus, which canfor cancers of the esophagus, although the rationale for predict the risk of future diagnosis of esophageal cancerthis is based on case series.24 Screening for esophageal in these patients.36,37
May 2005 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 1473 The incidence of cancer in patients with Barrett’s carcinoma), intraobserver agreement was near perfectesophagus is low in general, which makes surveillance (mean 0.82 and 0.80) and interobserver agreementcost-ineffective unless at-risk populations can be identi- was substantial ( 0.66 – 0.70). When the analysis wasﬁed.38,39 At the current time, dysplasia is used as the performed using the 4 separate classiﬁers (Barrett’s with-primary means to discriminate at-risk patients. A sys- out dysplasia, indeﬁnite for dysplasia and low-grade dys-tematic endoscopic biopsy protocol, generally accepted plasia, high-grade dysplasia, and carcinoma), the meanto be 4 quadrant biopsy specimens taken every 2 cm of intraobserver was once again substantial (0.64 – 0.68)Barrett’s mucosa, can provide tissue for histologic diag- but the mean interobserver was only 0.43– 0.46 (mod-nosis of dysplasia. The grading system used to classify erate agreement). Because of this difﬁculty in determin-dysplasia in Barrett’s esophagus is based on the system ing the presence of dysplasia, using recent recommenda-developed for ulcerative colitis.40 – 42 Although molecular tions to screen and perform surveillance only if dysplasiamarkers have shown promise, dysplasia is currently the is found may be ineffective if histologic interpretation isbest indicator for the risk of cancer. Increasing grade and not accurate.53extent of dysplasia are associated with increasing risk of Despite the difﬁculties with surveillance, the detec-cancer.37,43– 48 Information from several prospective stud- tion of high-grade dysplasia or early cancer has beenies and a Barrett’s esophagus registry provide a total shown to have the potential to improve survival inof 783 patients with a follow-up of 2.7–7.3 patients with Barrett’s esophagus.54,55 Cancers conﬁnedyears.37,43,45,47,49,50 These combined series indicate that to the esophagus are associated with a 5-year survival rate2% of patients without dysplasia progressed to cancer. of 70% compared with a survival rate of 20% for patientsProgression from low-grade dysplasia to cancer was noted with more invasive cancers.56 Nodal involvement, whichin 7% of patients and from high-grade dysplasia to helps determine prognosis, is far less likely to occur incancer in 22% of patients. Unfortunately, it is difﬁcult to patients who are found to have cancer on surveillanceobtain true cancer incidence per year of observation based endoscopies compared with those patients who are noton the data provided in these series. It is also apparent undergoing surveillance.1,3,56,57 This leads to signiﬁ-that low-grade dysplasia had a wide range of rates of cantly improved survival in surveyed patients comparedprogression between the series. This may be because of with those patients who underwent surgery for symp-the variability in interpretation of low-grade dysplasia.46 tomatic disease. Retrospective studies of patients in sur- Other ﬁndings on endoscopy that have been shown to veillance programs suggest that the patients with inci-predict the development of esophageal cancer include the dental cancers detected during surveillance are likely topresence of any mucosal abnormalities noted on endos- have earlier staged disease than those who present be-copy and the extent of high-grade dysplasia noted on cause of symptoms.58,59 Although these case series seemhistology.48 The presence of mucosal nodularity has been convincing, there are several concerns. Many of the ret-associated with a 2.5-fold increased risk of cancer devel- rospective studies compared patients in surveillanceopment in patients with high-grade dysplasia (P .01). groups with historical controls. Patients of advanced ageFocal high-grade dysplasia, deﬁned as involvement of 5 or with signiﬁcant comorbidity are less likely to becrypts in one biopsy specimen from the entire set of enrolled in a surveillance program. On the other hand,biopsy specimens, has been found to be associated with subjects enrolled in surveillance programs might havesigniﬁcantly less risk of cancer development than those healthier lifestyles, might have health-seeking behaviors,with greater amounts of high-grade dysplasia (P or were selected because of overall better health status..02).48 Survival differences between these 2 groups may well be Unfortunately, the classiﬁcation of dysplasia is subjec- due to selection bias. In addition, earlier detection oftive and does have signiﬁcant observer variation. Inter- cancer can lead to a false increase in survival time becauseobserver pathologic agreement can improve to 85%– of lead-time bias, which results from an early diagnosis of87% when the degree of dysplasia is placed in a 2-tier preclinical cancer rather than any true effect of detectionsystem. This is done by combining high-grade dysplasia on subsequent mortality.and intramucosal carcinoma in one category and low- In at least 2 small surveillance cohorts, esophagealgrade dysplasia, indeﬁnite for dysplasia, and no dysplasia cancer has been reported to be an uncommon cause ofin another.51 This was substantiated by a recent study death in patients with Barrett’s esophagus.60,61 There-using a group of expert pathologists.52 In this study, fore, the morbidity and mortality from surveillance andwhen analysis was performed using 2 broad diagnostic subsequent treatment might overshadow any advantagecategories (Barrett’s esophagus without dysplasia, indef- that surveillance provides by early detection of cancer.inite and low-grade dysplasia vs high-grade dysplasia and However, a retrospective population-based cohort study
1474 WANG ET AL GASTROENTEROLOGY Vol. 128, No. 5of endoscopic surveillance in 23 patients with Barrett’s abnormality, followed by rigorous sampling from each ofesophagus among 589 patients with adenocarcinoma the 4 quadrants of the esophagus every 1–2 cm startingfound an improved survival beneﬁt.62 Eleven of 15 pa- at the gastroesophageal junction and stopping at thetients who were diagnosed with cancer while in surveil- squamocolumnar junction using standard or jumbo bi-lance programs were alive compared with none of 8 opsy forceps. The goal is to determine if the patient haspatients not under surveillance. None of the deaths in the high-grade dysplasia or early cancer so that appropriatesurveillance group were from cancer compared with 4 management can be initiated. The effectiveness of takingdeaths in the nonsurveillance group that were directly 4-quadrant biopsy specimens every 1 cm along therelated to cancer. Three patients in the surveillance length of the Barrett’s segment with jumbo biopsy for-group and one patient in the nonsurveillance group died ceps instead of the traditional standard biopsy forceps hasas a result of complications of surgery. been assessed.42 In a retrospective analysis of the Seattle Allocation of resources based on risk stratiﬁcation has protocol, it was shown that taking 4-quadrant biopsythe potential to make surveillance more cost-effective. A specimens every 2 cm would have missed 50% of therecent decision analysis model examined the cost-utility cancers that were found by taking 4-quadrant biopsyratio for different surveillance strategies.38 A cohort of specimens every 1 cm.65 Jumbo biopsy forceps acquire50-year-old patients with Barrett’s esophagus without more tissue in a single biopsy than regular biopsy forcepsdysplasia was evaluated for surveillance strategies every and have not been associated with increased complica-1–5 years and no surveillance, with esophagectomy per- tions.66 The use of jumbo biopsy forceps has not beenformed if high-grade dysplasia was diagnosed. This study shown to decrease the number of samples necessary forsuggested that, with an annual cancer risk of 0.4%, surveillance, and they require the use of a therapeuticsurveillance every 5 years was the only cost-effective endoscope. The need for use of jumbo biopsy forceps hasstrategy. More frequent surveillance was associated with been questioned, and one study found no statistical dif-more cost and yielded a lower life expectancy. The in- ference in the rate of unsuspected cancers found at esoph-cremental cost-utility ratio for surveillance every 5 years agectomy when patients with Barrett’s esophagus withwas $98,000 per quality-adjusted life year (QALY) high-grade dysplasia underwent preoperative endoscopygained, comparable to the incremental cost-effectiveness using jumbo biopsy forceps versus standard biopsyratios of accepted practices. forceps.67 Another detailed cost-utility analysis assessed the It is generally accepted that patients with mucosalstrategy of surveillance of Barrett’s esophagus detected abnormalities in Barrett’s esophagus should undergo en-during screening.53 The strategy of surveillance of only doscopic mucosal resection to increase detection of neo-patients with dysplasia appears to be cost-effective rela- plasia. In one small series, endoscopic mucosal resectiontive to other currently accepted medical practices. The of suspicious lesions diagnosed superﬁcial adenocarci-strategy yielded an incremental cost-effectiveness ratio of noma in 13 patients (52%) and high-grade dysplasia in$10,440 per QALY saved compared with no screening. A 4 (16%) of 25 patients with Barrett’s esophagus.68strategy of no screening would incur direct medical costs Methods of surveillance. Several methods haveof caring for those developing cancer of $16 million. been studied to perform surveillance. These includeScreening and surveillance of patients with dysplasia brush cytology, chromoendoscopy, magniﬁcation endos-would prevent 2580 deaths caused by esophageal adeno- copy, and optical imaging techniques. All of these tech-carcinoma at a cost of $262 million. Fifty-nine patients niques have certain potential advantages over routinewould need to be screened and surveyed to prevent one biopsy.death from cancer. Although this model was thorough, it Brush cytology. Brush cytology has been advocatedis limited by several assumptions, such as that surveil- in the surveillance of patients with Barrett’s esophaguslance for patients with high-grade dysplasia can cease because it is faster to perform and it can sample moreafter 2 years and that there are no adverse effects on the than the 1% of the total Barrett’s mucosal area typicallyquality of life of patients undergoing surveillance, which achieved with standard biopsy. The diagnostic value ofother studies have shown to be decreased.63 cytology in patients with Barrett’s esophagus was as- Biopsy method. Patients who are undergoing sessed using 66 pairs of endoscopic biopsy specimens andsurveillance should follow a systematic endoscopic biopsy cytology specimens.69 Cytologic analysis was used toprotocol after they are free from esophagitis.64 The evi- correctly identify 7 of 8 esophageal cancers, with onedence for this is based on expert opinion, limited data, false-negative sample described on analysis as highlyand the outcome of cohorts of patients managed empir- suspicious for cancer. In other studies, cytologic brushingically. Biopsy specimens are ﬁrst taken from any mucosal was found to be a complementary technique in detecting
May 2005 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 1475serious Barrett’s lesions when combined with the histo- doscopy with Lugol’s solution in Western populationslogic examination of esophageal biopsy specimens.70,71 are sparse. A prospective analysis of a French cohort ofHowever, there was only 72% (47/65) concordance be- 158 alcoholic and/or smoking patients showed no signif-tween the 2 diagnostic techniques. In the 18 cases that icant advantage of staining with Lugol’s solution relativewere discordant, cytologic analysis diagnosed 13 at a to the visual diagnostic accuracy achieved by videoen-higher level of abnormality than the biopsy diagnosis and doscopy alone.81 Lugol’s-assisted endoscopy may be use-underdiagnosed 5. There is potential to further improve ful in populations living in highly endemic regions forthe efﬁcacy of cytology by using objective assessment esophageal squamous cell cancer.tools such as digital image analysis, which involves a Methylene blue stains absorptive cells (intestinal-typecomputerized analysis of the digitized microscopic im- epithelium), including specialized intestinal metaplasia,age.72,73 This has been demonstrated in China, where but does not stain normal squamous or gastric epithe-quantitative DNA ﬂuorescence has been found to en- lium. Methylene blue may be particularly useful in thehance detection of squamous cell cancers when applied to diagnosis of short segments of columnar-lined esopha-cytology taken in a high-risk population.74 As men- gus.82 Other uses include identifying dysplastic focitioned previously, balloon cytology for the surveillance of within Barrett’s epithelium and identifying residual focisquamous cell cancers has become an accepted part of of Barrett’s esophagus following mucosal ablative ther-practice in high incidence areas, although such devices apy. Chromoendoscopy with methylene blue requireshave not been shown to be beneﬁcial in Western pretreatment of the mucosa with a mucolytic agent, apopulations.30,75 dwell time for the dye to take effect, and a rinse phase to Chromoendoscopy. Chromoendoscopy involves the remove excess dye. In general, methylene blue stainingtopical application of dyes during endoscopy in an effort adds 5–10 minutes to the procedure time, but the ac-to enhance the detection of mucosal patterns or lesions on cessories required are inexpensive.83 Specialized intesti-the basis of their staining characteristics.76 Stains are nal metaplasia typically stains blue, while a “lighter”classiﬁed as vital (absorptive), contrast, or reactive, ac- intensity and increased heterogeneity in the stainingcording to their mode of action. Vital stains (e.g., Lugol’s pattern predict high-grade dysplasia and/or cancer.84 Fo-solution and methylene blue) enter speciﬁc cell types by cal unstained areas may represent dysplastic foci within apreferential absorption or diffusion across the cell mem- background of heterogeneous methylene blue staining.85brane. Contrast stains (e.g., indigo carmine and acetic These ﬁndings, however, are not consistently replicatedacid) seep into mucosal interstices to highlight mucosal in other studies, and the reproducibility of stainingtopography. Reactive stains chemically react with cellu- patterns among Barrett’s tissue types remains debat-lar constituents. able.86 This technique has been used to allow the gas- The body of the literature regarding chromoendoscopy troenterologist to perform fewer biopsies to ﬁnd signif-in the esophagus consists primarily of staining with icantly more dysplasia or cancer in a biopsy specimenLugol’s solution for detection of squamous dysplasia and (12% vs 6%) compared with the random biopsy tech-superﬁcial carcinoma and staining with methylene blue nique.85 However, 2 prospective trials comparing thefor assessment of Barrett’s esophagus. Lugol’s solution is methylene blue– guided biopsy technique with the ran-an iodine-based vital dye that has an afﬁnity for glycogen dom biopsy technique found no signiﬁcant differences incontained in nonkeratinized squamous epithelium. Chro- the detection of dysplasia or early cancer between the 2moendoscopy with Lugol’s solution has been used pri- techniques.87,88 Methylene blue has also recently beenmarily to screen individuals at high risk for squamous shown to induce oxidative damage of DNA because ofcell cancer.77– 80 The application of Lugol’s solution dur- photoactivation of the dye by the white light of theing endoscopy is not standardized, with nonuniform endoscope, which certainly increases the risk of its use inconcentrations (1%–3%) and volumes (10 –50 mL) as premalignant tissue.89well as variable deﬁnitions of “unstained” areas that Magniﬁcation endoscopy. Most new electronicmerit biopsy.77– 80 Despite these differences, these studies videoendoscopes are equipped with charge-coupled de-in aggregate show unstained lesions to have a high vice chips of high pixel density (400K), enabling high-sensitivity (91%–100%) and moderate speciﬁcity (40%– image resolution (ability to discriminate 2 closely95%) for squamous dysplasia and carcinoma.77 Lesions approximated points). Megapixel density (850K) endo-are found to be more clearly deﬁned and can be signiﬁ- scopes have recently been introduced, which may furthercantly larger after staining,81 providing critical informa- enhance detection. Magniﬁcation endoscopy enlarges thetion, particularly if local therapy, such as endoscopic image, which enhances mucosal detailing and is gener-mucosal resection, is contemplated. Studies of chromoen- ally used in combination with chromoscopy. Magniﬁca-
1476 WANG ET AL GASTROENTEROLOGY Vol. 128, No. 5tion chromoendoscopy has been used primarily to char- to harbor dysplastic or early cancerous changes. Recentacterize Barrett’s esophagus.90 –93 Ridged or villous studies of ﬂuorescence endoscopy, however, have showedmucosal patterns observed under magniﬁcation endos- limited diagnostic utility in Barrett’s esophagus.88,105 Incopy using a variety of stains have been correlated with a small prospective study of 35 patients, the sensitivityintestinal metaplasia or dysplasia in case series.91–93 and speciﬁcity of autoﬂuorescence imaging for the diag-Whether magniﬁcation endoscopy will decrease the need nosis of dysplasia or cancer versus Barrett’s mucosa with-for endoscopic biopsy or signiﬁcantly enhance the diag- out dysplasia were 21% and 91%, respectively.88 Anostic yield over conventional techniques remains to be prospective, randomized, crossover study of 50 patientsdetermined in large controlled trials. with Barrett’s esophagus, published in abstract form, Optical biopsy. Optical biopsy is a term that en- showed no signiﬁcant beneﬁt of ﬂuorescence-guided bi-compasses a variety of techniques that use light to en- opsies over a standard surveillance biopsy protocol for thehance detection of dysplastic lesions. Included in this detection of high-grade dysplasia and early adenocarci-category are spectroscopic techniques and low coherence noma.105 Fluorescence endoscopy is an evolving technol-interferometry (optical coherence tomography [OCT]). ogy. Optical biopsy and imaging techniques are not yetSpectroscopy is a light-based diagnostic technique that optimized for clinical use.exploits properties of light-tissue interactions, such as OCT is analogous to ultrasonography but uses lightﬂuorescence, elastic scattering, and Raman scattering, for instead of sound waves to produce high-resolution, cross-tissue characterization. Collected in the form of spectra, sectional imaging of tissue. The spatial resolution ofthese optical (light) signals are sensitive to microstruc- 10 –25 m achieved by endoscopic OCT systems istural and/or molecular changes accompanying various 10-fold better than that of high-frequency ultrasonogra-stages of tissue pathology, and spectral differences can phy but at the expense of a limited sampling depth ( 2then be correlated to speciﬁc histopathologic diagnoses. mm).106 –109 Current endoscopic OCT devices are cathe- Assessment of optical spectroscopic techniques in the ter based (2–2.7 mm in diameter), allowing scanning ofesophagus has been limited to small-scale, proof-of-prin- the esophagus in a linear,107 transverse,106 or radial108ciple studies.94 –102 Most studies deal with ﬂuorescence fashion. Using predetermined OCT criteria for special-spectroscopy, which shows moderate to high sensitivity ized intestinal metaplasia, the linear scanning OCT sys-(75%–100%) and speciﬁcity (65%–95%) in differenti- tem was found to be 97% sensitive and 92% speciﬁc forating high-risk (high-grade dysplasia/adenocarcinoma) identifying specialized intestinal metaplasia.110 A quan-from low-risk (nondysplastic to low-grade dysplasia) titative analysis of the OCT signal seemed to identifysamples of Barrett’s esophagus.94 –98 False positives occur high-grade dysplasia with high sensitivity (100%) andin the presence of inﬂammatory or reactive changes.94 speciﬁcity (85%) in a small retrospective study, whichThe range of diagnostic accuracies primarily relates to will require prospective validation on a larger samplemethodological differences, because the technique has size.110 Similar to spectroscopy, OCT is a technique inyet to be optimized for Barrett’s esophagus. It is also evolution, and further reﬁnements are anticipated thatunclear whether ﬂuorescence spectroscopy, aided by an may improve its diagnostic accuracy for the detection ofexogenously administered ﬂuorescent agent with afﬁnity preneoplastic lesions.for neoplastic tissues (e.g., porﬁmer sodium or 5-amin- Frequency of surveillance. The frequency of en-olevulinic acid), improves discrimination of Barrett’s ep- doscopic surveillance in patients with Barrett’s esopha-ithelia over naturally occurring tissue ﬂuorescence gus has been debated without the beneﬁt of well-con-(autoﬂuorescence).99 –101,103 Future comparative studies structed studies. Surveillance intervals recommended inare required to address this issue. Other emerging spec- this section are supported by limited evidence, which istroscopic techniques, including elastic scattering spec- based on our understanding of the natural history oftroscopy,102 Raman spectroscopy,103 and multimodal op- development of esophageal adenocarcinoma dependingtical spectroscopy,104 have shown initial promise in on the grade and extent of dysplasia in Barrett’s esoph-differentiating Barrett’s epithelia in feasibility studies. agus. The goal of surveillance is to identify patients at Fluorescence endoscopy is an extension of its spectro- risk and assess them in a timely fashion to prevent orscopic counterpart into an imaging technique, which detect cancer at an earlier treatable stage. Summaries ofconsists of sensitive cameras capturing the emitted ﬂu- the studies that have investigated the incidence of cancerorescence emanating from a mucosal ﬁeld of view ap- in Barrett’s esophagus appear in Tables 1– 4. Patientsproaching that of a conventional endoscope. The obvious with nondysplastic Barrett’s esophagus should undergoadvantage over point spectroscopy is the ability to per- repeat endoscopy 1 year after the initial endoscopy. Pa-form widespread assessment of the mucosa for areas likely tients with persistent nondysplastic Barrett’s esophagus
May 2005Table 1. Diagnosis of Prevalent or Incidental Cancer on Follow-up of Patients With No Dysplasia Mean orInstitute or geographic Study No. of median Total Incidental area Study type duration patients age (y) Duration of follow-up cancers (%) cancers (%) Comments ReferenceUniversity of Chile, LGD developed in 15 Csendes et al,115 Santiago, Chile Prospective 1978–1991 152 52 100 mo (mean) 4 (2.6) 4 (2.6) patients on follow-up 1998University of Otago Medical School, Dunedin, New Zealand Prospective 1980–1986 52 63 16.4 mo (mean) 0 0 Cooper et al,182 1987Deutsche Klinik fur Diagnostik, Wiesbaden, 5 patients developed Germany Prospective 1980–1999 60 61 10 y (mean) 2 (3.3) 2 (3.3) LGD Eckardt et al,340 2001University Hospital, El Palmar, Murcia, Spain Prospective 1982–1993 59 37 287 patient-years 2 (3.3) 2 (3.3) Ortiz et al,341 1996 One patient developed HGD, 3 had persistent LGD, 18 had transient LGD, and 56 remained nondysplastic. AMERICAN GASTROENTEROLOGICAL ASSOCIATION Patients with prevalent cancerVA Medical Center diagnosed at entry or and University of within the ﬁrst 6 Kansas Medical 40 mo (mean) or months were Center Prospective — 80 61 362 patient-years 2 (2.5) 2 (2.5) excluded Weston et al,45 1999 5-year cumulative incidence of cancer was reported to beFred Hutchinson 2.4 y (median) or 1.7% and 8-year Cancer Research 3.9 y (mean) for cumulative incidence Center, University of Prospective those who did not was reported to be Washington cohort 1983–1998 129 62 develop cancer 5 (3.9) 5 (3.9) 8.4% Reid et al,43 2000LGD, low-grade dysplasia; HGD, high-grade dysplasia. 1477
1478 WANG ET ALTable 2. Diagnosis of Prevalent or Incidental Cancer on Follow-up of Patients in Which the Degree of Dysplasia Was Not Deﬁned or There Were 10 Patients in a Single Dysplasia Category Mean or Institute or geographic Study No. of median Duration of Total Incidental area Study type duration patients age (y) follow-up cancers (%) cancers (%) Comments ReferenceMayo Clinic, Rochester, Retrospective 1960–1983 122 58 8.5 y (mean) 20 (16) 2 (1.9) Cameron et al,342 1985 MNBoston VA Medical Center, Retrospective 1962–1983 115 58 3.3 y (mean) 11 (9.5) 2 (1.9) 4 patients had LGD at baseline; one of Spechler et al,34 1984 Boston University them progressed to HGD and ﬁnally School of Medicine, and to cancer. HGD developed in 7 Tufts University School patients of MedicineVA Outcomes Group, Retrospective 1970–1994 102 63 4.8 y (median) 3 (3) 3 (3) 8% of nondysplastic patients Katz et al,343 1998 Dartmouth-Hitchcock developed dysplasia. In 2 patients, Medical Center HGD developed before cancerCastle Hill Hospital, Retrospective 1971–1991 26 62 11.5 y (mean) 4 (15.4) 4 (15.4) Moghissi et al,344 1993 Cottingham, United KingdomLahey Clinic, Burlington, Retrospective 1973–1989 176 56 3 y (median) 5 (2.8) 5 (2.8) 11% of patients developed dysplasia Williamson et al,33 1991 MA on follow-upUniversity of Rotterdam, Retrospective 1973–1994 155 42 9.3 y (mean) 8 (5.3) 8 (5.3) Van der burgh et al,60 Dutch cohort cohort or 1440 1996 patient- yearsHelsinki University, Central Retrospective 1975–1985 32 59 166 patient- 3 (9.3) 3 (9.3) Ovaska et al,345 1989 Hospital, Finland yearsMayo Clinic, Rochester, Retrospective 1975–1994 113 68 6.5 y (median) 3 (2.6) 3 (2.6) Patients in this study underwent McDonald et al,346 1996 MN fundoplication; one patient developed HGDUniversity Hospital, Prospective 1976–1990 102 63 54 mo (mean) 4 (3.9) 4 (3.9) 2 patients had dysplasia at baseline Iftikhar et al,347 1992 Nottingham, England and a total of 12 had dysplasia on GASTROENTEROLOGY Vol. 128, No. 5 follow-upNinewells Hospital and Retrospective 1976–1986 44 58 9.5 y (mean) 2 (4.5) 2 (4.5) Patients were not in formal Rana et al,348 2000 Medical School, Dundee surveillance programCleveland Clinic Registry 1979–1995 136 58 4.2 y (mean) 2 (1.4) 2 (1.4) One patient without dysplasia and one O’Connor et al,50 1999 with LGD developed cancer. Five patients developed HGD, and 24 developed LGD.
May 2005Table 2 (continued). Diagnosis of Prevalent or Incidental Cancer on Follow-up of Patients in Which the Degree of Dysplasia Was Not Deﬁned or There Were 10 Patients in a Single Dysplasia Category Mean or Institute or geographic Study No. of median Duration of Total Incidental area Study type duration patients age (y) follow-up cancers (%) cancers (%) Comments ReferenceAcademic Medical Center, Prospective — 50 59 5.2 y (mean) 5 (10) 5 (10) 6 patients had LGD and one had HGD Hameeteman et al,349 University of Amsterdam at baseline. At the end of the study, 1989 10 had LGD and 3 had HGDUniversity of Minnesota, Retrospective 1980–1995 149 — 510 patient- 20 (13.4) 7 (5) 13 prevalent cancers were detected at Streitz et al,350 1998 Duluth Clinic years esophagectomyPrincess Alexandra Prospective 1981–1988 81 63 3.6 y (mean) 3 (3.7) 3 (3.7) 2 patients had HGD at baseline, and Miros et al,37 1991 Hospital, Brisbane, both progressed to adenocarcinoma. Australia Ten patients had LGD, and one progressed to adenocarcinoma.VA Medical Center, Arizona Prospective 1982–1995 177 62 4.8 y (mean) 11 (6.2) 4 (2.3) All patients who progressed to cancer Drewitz et al,351 1997 Health Sciences Center cohort or 834 developed dysplasia before the patient- development of cancer years AMERICAN GASTROENTEROLOGICAL ASSOCIATIONUniversity of Leeds, Retrospective 1984–1995 307 58 72 mo (mean) 12 (3.9) 12 (3.9) Patients were excluded if the cancer Bani-Hani et al,59 2000 Leeds, England was present at baseline or within the ﬁrst 6 monthsVA Cooperative Study Retrospective 1986–1999 108 58 1037 patient- 4 (3.7) 4 (3.7) Patients were not in formal Spechler et al,125 2001 Group years surveillance programGOSPE, Torino, Italy Prospective 1987–1995 187 — 36 mo (mean) 3 (1.6) 3 (1.6) 3 patients had LGD at baseline. 5 Ferraris et al,352 1997 cohort patients developed LGD, and 2 developed HGDLGD, low-grade dysplasia; HGD, high-grade dysplasia. 1479
1480 WANG ET ALTable 3. Diagnosis of Prevalent or Incidental Cancer on Follow-up of Patients With Low-Grade Dysplasia Mean orInstitute or geographic Study No. of median Duration of area Study type duration patients age (y) follow-up Total cancers (%) Incidental cancers (%) Comments ReferenceVA Outcomes Group, 3 patients had HGD Dartmouth- or cancer. Cancer Hitchcock Medical and HGD were Katz et al,343 Center Retrospective 1970–1994 24 63 4.8 y (median) * * pooled together 1998 2.4 y (median) or 3.9 y (mean) for 5- and 8-yearFred Hutchinson those who cumulative Cancer Research did not incidence of cancer Center, University of Prospective develop was reported to be Reid et al,43 Washington cohort 1983–1998 43 62 cancer 3 (7) 3 (7) 12% 2000 An additional 5 patients developed HGD. When pathologists agreed on the diagnosis of LGD, 7 of 17 11 mo progressed to HGD Skacel etCleveland Clinic Retrospective 1986–1997 25 67 (median) 2 (8) 2 (8) or cancer al,46 2000 4 patients progressed to extensive HGD. 12 patients had persistent LGD. 31 patients regressed 4 (8); one patient 4 (8); one patient to no dysplasia GASTROENTEROLOGY Vol. 128, No. 5VA Medical Center had deﬁnite and 3 had deﬁnite and 3 Patients with and University of had suspected had suspected prevalent cancer Kansas Medical intramucosal intramucosal diagnosed at entry Weston et Center Prospective — 48 62 41 mo (mean) cancer cancer were excluded al,353 2001*Not deﬁned in manuscript, 24 patients developed LGD anytine during follow up, the high grade dysplasia or cancers were pooled; HGD, high-grade dysplasia; LGD, low-grade dysplasia.
May 2005Table 4. Diagnosis of Prevalent or Incidental Cancer on Follow-up of Patients With High-Grade Dysplasia Mean orInstitute or geographic Study No. of median Duration of Total area Study type duration patients age (y) follow-up cancers (%) Incidental cancers (%) Comments ReferenceVA Hospital, Hines, IL Prospective 1979–1996 75 62 7.3 y (mean) 12 (16) 12 (16) 4 patients in whom cancer was Schnell et al,47 cohort detected in the ﬁrst year of 2001 surveillance were excluded and considered to have prevalent cancersJohns Hopkins Retrospective 1982–1994 30 61 — 13 (43) — Esophagectomy. No Heitmiller et surveillance al,354 1996Fred Hutchinson Prospective 1983–1998 76 62 2.4 33 (43.4) 33 (43.4) 5-year cumulative incidence of Reid et al,43 Cancer Research cohort y (median) cancer was reported to be 2000 Center, University of or 3.9 59% Washington y (mean) for those who did not develop cancerUniversity of Retrospective 1985–1995 11 61 — 8 (72.7) — Esophagectomy. No Edwards et Louisville, surveillance al,355 1996 Louisville, KYVA Medical Center Prospective — 15 61 37 7 (470; 4 7 (470; 4 patients had Patients had unifocal HGD. In Weston et and University of mo (mean) patients deﬁnitive cancer and 3 had 4 patients, the extent of al,356 2000 Kansas Medical had possible intramucosal dysplasia increased. 5 Center deﬁnitive cancer) patients became AMERICAN GASTROENTEROLOGICAL ASSOCIATION cancer nondysplastic, and 2 had and 3 had LGD. Patients with prevalent possible cancer diagnosed at entry intramucosal were excluded cancer)Mayo Clinic, Retrospective 1991–1999 100 67.5 30 32 (32%) 13 (16) after excluding Patients with cancer at entry Buttar et al, Rochester, MN cohort mo (mean) cancers that were detected were excluded. 4 patients 200148 for focal in the ﬁrst 6 months with focal and 28 patients HGD and with diffuse HGD were 15 diagnosed with cancer mo (mean) for diffuse HGDHGD, high-grade dysplasia; LGD, low-grade dysplasia. 1481
1482 WANG ET AL GASTROENTEROLOGY Vol. 128, No. 5should have periodic surveillance at 5-year intervals, rationale for repeating surveillance every 3 months fordepending on the patient’s health. The rationale of re- the ﬁrst 2 years is to capture any prevalent cancers thatpeating surveillance twice in the ﬁrst year is to diagnose were not detected initially. This surveillance intervalany prevalent dysplasia that may have been missed dur- should be used even if ablative procedures are performeding initial endoscopy. Most initial procedures that diag- for high-grade dysplasia or cancer because of the risk ofnose Barrett’s esophagus do not involve complete sur- recurrence of the disease.112,113 The reason to increase theveillance biopsies because the diagnosis has not been surveillance interval to 6 months after the ﬁrst 2 years isestablished. The reason to increase the surveillance in- that the majority of cancers detected in patients withterval to 5 years is based on decision analysis models high-grade dysplasia are within the ﬁrst year and the riskshowing that surveillance in this group of patients is not of incidental cancer after the ﬁrst 2 years of surveillancecost-effective. The caveat is that this recommendation is is low.47,48 After ablative therapy, surveillance should bebased on set assumptions made in the cost-utility model, continued as per the preablative recommendation be-and any major change in these assumptions in the future cause the long-term consequences of ablation are as yetwill require modiﬁcation of the surveillance interval in undetermined and because of reports of recurrent neo-patients with nondysplastic Barrett’s esophagus.38 This is plasia after ablation.different from prior gastrointestinal societal guidelines, Summary of Evidence—The use of surveillance in pa-which suggested that surveillance should be performed at tients with Barrett’s esophagus is supported by level IImore frequent intervals, but there has not been sufﬁcient evidence, although it seems that the most cost-effectiveevidence to support prior recommendations.64 Therefore, approach is to target patients at higher risk for develop-we recommend ending surveillance when the anticipated ment of cancer. Surveillance should be performed usinglife expectancy is limited ( 1 year) or the patient is 4-quadrant biopsy specimens taken every 1–2 cm ofunable to tolerate any therapeutic measures. Even Barrett’s mucosa, depending on the degree of dysplasiayounger patients may not warrant surveillance if therapy (based on level III evidence). The use of jumbo biopsyfor cancer cannot be performed. forceps in routine surveillance is not recommended based Patients with low-grade dysplasia should undergo re- on existing evidence. The use of jumbo biopsy forceps inpeat endoscopy twice in the ﬁrst year, and those patients patients with high-grade dysplasia has not been provenwho have persistent low-grade dysplasia in Barrett’s to provide increased beneﬁt. There are as yet insufﬁcientesophagus should have periodic surveillance at 1- to data to recommend chromoendoscopy with Lugol’s solu-2-year intervals until the age of 80 years, depending on tion over conventional videoendoscopy in Western co-health status. The risk of cancer development in patients horts at risk for esophageal squamous cell cancer (basedwith low-grade dysplasia is intermediate.46 Patients with on level III evidence). Methylene blue–assisted chro-low-grade dysplasia whose diagnosis is agreed on by 2 moendoscopy is not clearly beneﬁcial and should not bepathologists carry a relatively higher risk of cancer diag- used in the routine assessment of Barrett’s esophagusnosis; therefore, we recommend a surveillance interval of (based on level II evidence). Magniﬁcation, optical spec-1 year for this group of patients. Patients whose diagno- troscopy, and imaging techniques are still being devel-sis of low-grade dysplasia is not agreed on by 2 pathol- oped and should be considered research tools at theogists but who do not have any evidence of high-grade current time (based on level III evidence). Surveillance ofdysplasia can be surveyed at 2-year intervals. nondysplastic Barrett’s mucosa established by 2 endos- Patients with high-grade dysplasia in Barrett’s esoph- copies should be at 5-year intervals (based on level IIIagus that has been conﬁrmed by 2 experienced patholo- evidence). Surveillance of patients with low-grade dys-gists should be recommended to proceed with deﬁnitive plasia established by 2 endoscopies should be at 1- tosurgical or endoscopic management, especially if the 2-year intervals, depending on whether or not the diag-high-grade dysplasia is diffuse or mucosal abnormalities nosis is conﬁrmed by 2 pathologists. If both pathologistsare found on endoscopy. If the patient does not wish to concur that low-grade dysplasia is present, 1-year sur-proceed with deﬁnitive treatment until the diagnosis of veillance intervals should be used; if agreement cannot becancer is made, surveillance should be considered. Sur- reached (dysplasia is not believed to be present by at leastveillance endoscopy should be repeated every 3 months one pathologist), 2-year intervals should be sufﬁcientin the ﬁrst 2 years, followed by 6-month intervals indef- (based on level IV evidence). High-grade dysplasiainitely. The rationale for recommending immediate sur- should be conﬁrmed in a second endoscopy and reviewedgical or endoscopic therapy is that many patients may by expert pathologists. If this is conﬁrmed, the patientalready have coexisting esophageal cancer that was not can be offered surveillance at 3-month intervals for 2detected due to the limitations of surveillance.111 The years and then every 6 months (based on level III evi-
May 2005 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 1483dence). Any mucosal abnormality should be carefully Ablation Therapyinvestigated and removed for diagnosis. Surgical resec- Ablation therapy in conjunction with acid controltion should be recommended and mucosal ablation with has been promoted for the treatment of Barrett’s esoph-photodynamic therapy (PDT) can be offered for patients agus to decrease the risk for esophageal cancer.127,128 Thiswith conﬁrmed high-grade dysplasia (based on level III is based on the observation that elimination of the meta-evidence). plastic epithelium can result in squamous mucosa regen- eration.129 Multiple ablation therapies have been pro- Chemoprevention for Esophageal posed and involve the use of multipolar coagulation, Cancer argon plasma coagulation, neodymium:yttrium-alumi- Acid Inhibition num-garnet (Nd:YAG) laser coagulation, and PDT. Multipolar coagulation has been performed in a prospec- Because esophageal cancer is uncommon and in- tive multicenter study with 58 patients that eliminatedterventions are usually only warranted in high-risk histologic evidence of Barrett’s mucosa in 78% of treatedgroups, strategies for chemoprevention of this condition patients.130 Argon plasma coagulation has been used inare being advocated. These include the use of acid sup- one randomized, nonblinded, prospective trial in 40pression, nonsteroidal anti-inﬂammatory drugs patients after fundoplication for acid control.131 Barrett’s(NSAIDs), and lifestyle modiﬁcations. Antireﬂux surgery mucosa was initially found underneath squamous mucosaor prolonged acid suppression using high doses of proton in this study in more than 35% of patients 1 month afterpump inhibitors result in the appearance of squamous treatment but decreased to 5% at 1 year. Patients whoislands in Barrett’s esophageal segment but have not did not achieve complete ablation had 95% regression inconsistently been shown to regress metaplastic epithe- the amount of Barrett’s mucosa. Multiple case series oflium or prevent esophageal adenocarcinoma.114 –117 Al- argon plasma coagulation have been reported with sig-though the rationale for intensive antireﬂux therapy niﬁcant reduction in Barrett’s mucosa and decrease inseems reasonable, there is not good clinical evidence to dysplasia.132–139 Many ablation trials emphasize that re-support this approach. The primary rationale for these duction in the length or area of Barrett’s mucosa shouldproposals is that proton pump inhibitors incompletelysuppress gastric acid and fail to completely reduce duo- correspond with a decreased risk for cancer, although thisdenal-esophageal bile reﬂux.118,119 Acid and bile salts can has not been shown in a prospective study. PDT has beenactivate several key cellular pathways in Barrett’s esoph- approved by the Food and Drug Administration (FDA)agus that are normally associated with progression of for use in patients with high-grade dysplasia to decreaseneoplasia in Barrett’s esophagus, and inhibition of these the risk of cancer formation. Case series of patients withpathways may prevent the development of adenocarci- high-grade dysplasia treated with PDT using sodiumnoma of the esophagus in an animal model of Barrett’s porﬁmer have found that dysplasia can be eliminated inesophagus.120 –123 In patients with Barrett’s esophagus 78% of patients with elimination of 75%– 80% of thewho were asymptomatic on acid suppressive therapy and Barrett’s mucosa, but esophageal strictures occurred inalso had normalization of intraesophageal acid exposure 34% of patients.140 Unfortunately, the residual Barrett’son pH monitoring, there was signiﬁcantly more differ- mucosa after ablation therapy has been found to containentiation and less proliferation compared with patients genetic mutations similar to those found before ablation,who continued to have abnormal intraesophageal acid suggesting that histologic improvement may not corre-exposure.124 However, even on high dosages of proton late with elimination of cancer risk.113 A prospectivepump inhibitors, acid inhibition may be difﬁcult to randomized trial studied PDT using 2 different dosagesachieve, with 4 of 25 patients having pathologic acid of 5-aminolevulinic acid (a less potent agent than sodiumreﬂux on 80 mg/day of omeprazole.119 Fundoplication is porﬁmer) followed by argon plasma coagulation com-not successful in controlling acid reﬂux on a long-term pared with argon plasma coagulation alone in 40 pa-basis and did not prevent development of cancer, as tients.141 These investigators found that argon plasmademonstrated by a follow-up to a randomized prospective coagulation alone or in combination with PDT was onlytrial comparing medical with surgical therapy.125 Epide- effective in eliminating about 70% of the Barrett’s mu-miologic studies involving more than 66,000 patients cosa.with reﬂux treated medically compared with 11,000patients treated with fundoplication in Sweden found no NSAIDsprotective effect of surgery in decreasing the develop- The rationale for the use of NSAIDs to preventment of adenocarcinoma of the esophagus.126 cancer in Barrett’s esophagus is based on experimental
1484 WANG ET AL GASTROENTEROLOGY Vol. 128, No. 5and epidemiologic evidence. No randomized controlled associated with the presence of any gastroesophagealtrials have found aspirin to be effective for prevention of reﬂux disease symptoms (29.7%; 95% CI,cancer. It has been shown that bile salts in a pH- 19.5%– 42.3%). Consumption of fruits and vegetablesdependent manner can activate the arachidonic acid less than twice a day on average had a modest PAR ofpathway and may promote neoplasia in Barrett’s esoph- 15.3% (95% CI, 5.8%–34.6%). In this population,agus.121,122 The activation of the arachidonic acid path- 78.7% (95% CI, 66.5%– 87.3%) of esophageal adeno-way leads to increased production of prostaglandin E2, carcinoma cases could be attributed to one or more ofwhich increases cellular proliferation.142,143 Inhibition of these well-established risk factors, with smoking andprostaglandin E2 normalizes proliferation in Barrett’s body mass index contributing the most.epithelial cells and decreases it to a level that is seen in Summary of Evidence—Based on level II evidence, fun-normal squamous epithelium.143 Inhibition of the cyclo- doplication should not be recommended solely to preventoxygenase enzyme with selective and nonselective cyclo- esophageal cancer in patients with gastroesophageal re-oxygenase inhibitors has been shown to reduce the risk of ﬂux disease. The use of high-dose proton pump inhibi-adenocarcinoma of the esophagus in an animal model of tors to prevent esophageal cancers in patients with gas-Barrett’s esophagus.123 Finally, cyclooxygeanse-2 expres- troesophageal reﬂux disease has also not been establishedsion has also been shown to parallel the progression of and cannot be recommended in clinical practice (basedneoplasia in human Barrett’s esophagus.144 on level III evidence). Level II evidence suggests that Epidemiologic studies have found that NSAIDs ablation therapy may decrease dysplasia in Barrett’sprotect against the risk of esophageal cancer.145–149 esophagus and could potentially decrease cancer risk.Case-controlled studies have shown a 36%–90% rel- NSAIDs should not be recommended at the current timeative risk reduction for esophageal cancer in patients solely for the prevention of esophageal adenocarcinomausing aspirin or other NSAIDs occasionally or on a but have promise as a chemopreventative agent, whichlong-term basis.145–147,149 A decision analysis model might be beneﬁcial if required for other medical indica-has been used to evaluate the feasibility and cost- tions (based on level II evidence). Patients at risk foreffectiveness of NSAIDs to prevent esophageal adeno- esophageal cancer should be counseled to adapt healthiercarcinoma. It suggested that the high incidence of lifestyles, in particular attaining normal weight andesophageal adenocarcinoma in Barrett’s esophagus eliminating tobacco use (based on level II evidence).with high-grade dysplasia renders chemoprevention acost-effective option.150 NSAIDs were not a cost-ef- Diagnosis and Staging offective measure in the general population of all pa- Esophageal Cancertients with Barrett’s esophagus, but this was sensitiveto variations in the cost of chemoprevention, the in- Diagnosiscidence of cancer, and the efﬁcacy of the NSAID in Most patients with esophageal cancer present at aprevention of cancer. In patients with gastroesopha- late stage when the diagnosis of cancer is made, withgeal reﬂux, a chemopreventive approach will only be dysphagia as the cardinal symptom.153 In particular,cost-effective if the agents are very safe and effective to persistent dysphagia that progresses from solids to liq-prevent the development of adenocarcinoma of the uids should heighten suspicion for esophageal cancer andesophagus. prompt an endoscopic evaluation. Up to 75% of patients have experienced anorexia and weight loss when seeking Lifestyle medical attention. Patients may also present with Lifestyle factors for the development of esophageal odynophagia, chest pain, or gastrointestinal bleeding.cancer have been identiﬁed through epidemiologic stud- Cough aggravated by swallowing raises the possibility ofies. Obesity seems to be one of the most important risk an esophagopulmonary ﬁstula, a devastating complica-factors associated with esophageal cancer.151,152 In a re- tion associated with a high 30-day mortality rate.154cently completed study of patients with esophageal and The diagnosis of esophageal cancer is established bygastric cancer, smoking and dietary habits were found to ﬂexible endoscopy with biopsy. The features, location,be important risk factors.152 Smoking was found to have and size of the tumor can be more accurately assessed bya population attributable risk (PAR) of 39.7% (95% endoscopy than by radiographic studies.155 Barium swal-conﬁdence interval [CI], 25.6%–55.8%). The frequency low as an initial diagnostic test is of limited value.156,157of gastroesophageal reﬂux disease symptoms also showed However, it may be useful to conﬁrm the presence ofa positive trend in the PAR, with symptoms at least once esophagopulmonary ﬁstulas or document complete lumi-per day accounting for almost one half of the PAR nal obstruction when clinically suspected. In patients
May 2005 AMERICAN GASTROENTEROLOGICAL ASSOCIATION 1485 additional staging studies are needed. In one series of 209 patients, endoscopy was 89% accurate in determin- ing tumor staging compared with surgical or EUS stag- ing.167 Patients with large bulky circumferential tumors who present with dysphagia are likely to have advanced disease and will require tests to conﬁrm this. Patients who have cancers that are 2 cm in diameter and are asymptomatic are more likely to have early disease. Al- gorithms for evaluation of advanced and early esophageal cancers are found in Figures 1 and 2. The most commonly used staging procedure in esoph- ageal cancer is computed tomography (CT). The partic- ular strength of this test is the ability to detect distant Figure 1. Algorithm for staging of advanced cancers. metastatic disease.168 The primary reason this is used as the ﬁrst staging study is because of its availability and the change in the treatment course that would occurwith advanced cancers, esophageal dilation may be re- should metastasis or invasion of other organs be detected.quired to allow for a standard (OD, 9.8 mm) endoscope On initial presentation, most patients with metastaticto traverse the obstructed lumen. Alternatively, an ul- disease had involvement of abdominal lymph nodestrathin (OD, 5.3– 6 mm) endoscope may pass through (45%), liver (35%), lung (20%), and cervical lymphthe stenosis and allow completion of the examination in nodes (18%) based on ﬁndings at esophagectomy.169 CT75% of cases.158 of the chest and abdomen correctly identiﬁed metastatic Biopsy specimens are required for histologic conﬁrma- disease in only 69% of the cases. In this series, patientstion. The diagnostic yield reaches 100% when 6 or more without metastasis on CT did not have involvement ofsamples are obtained using a standard endoscopic biopsy the bones or brain. CT has limited accuracy in stagingforceps.159,160 Biopsy specimens of necrotic or ﬁbrotic the extent of tumor invasion, with correct staging foundareas should be avoided. The adequacy of biopsy speci- in 50%–90%.170 –172 CT has been shown to produce falsemens obtained via ultrathin endoscopes has not been negatives in 30%– 60% of patients.173 Even with im-formally assessed. As an adjunct, brush cytology can be provements in technology such as the helical CT scanner,helpful in sampling tight malignant strictures, which comparisons with EUS and ﬁne-needle aspiration indi-may not be easily accessible by conventional biopsy cate that it is still not as sensitive for detecting celiactechniques.161 Brushings should be collected before bi- lymph node disease or T4 carcinoma.174opsy to maximize the diagnostic yield.162 Endoscopic EUS consists of 3 fundamentally different devices thatultrasonography (EUS) should be considered when stan- may not be available to all gastroenterologists. There aredard biopsy and/or brush cytology fail to conﬁrm the probes that can be placed within the endoscope or useddiagnosis in the setting of high clinical suspicion (e.g., as stand-alone instruments, radial scanning echoendo-submucosal tumors).163 Fine-needle aspiration or Tru-cut scopes, and linear array echoendoscopes that permit ﬁne-biopsy can be performed at the time of EUS. Staging The staging of esophageal cancer is critical toguide further therapy for the patient. Patients with can-cer conﬁned to the mucosa or superﬁcial submucosa canbe treated using surgical resection or potentially endo-scopic therapy.137,164 However, patients who have moreadvanced disease will require surgical resection or che-moradiation.165,166 None of the currently available stag-ing technologies have been shown to be able to stage allaspects of the tumor. The selection of the best stagingtests depends on the probability of detecting metastaticcancer. Although endoscopy is not a staging tool, it canbe helpful to guide the gastroenterologist toward which Figure 2. Algorithm for staging of early cancers.
1486 WANG ET AL GASTROENTEROLOGY Vol. 128, No. 5needle aspiration. EUS can be directed primarily at the actually combines the CT image with the PET image tomucosa with the use of probes that usually involve allow better tumor localization. This may increase thehigh-frequency scanning at 20 –30 MHz or can be di- speciﬁcity of the test.rected to more distant structures by scanning at 7.5–12 Other staging procedures that have been investigatedMHz with the radial echoendoscopes. The diagnostic include minimally invasive surgery or laparoscopic stag-utility of EUS is in determining the depth of tumor ing before esophagectomy. Case series have found thatinvasion and the presence of local-regional adenopathy. performance of laparoscopic- or thoracoscopic-guided bi-This technology is limited in terms of detecting distant opsies of lymph nodes increases the detection of meta-metastasis because visualization of other organ systems, static disease and can change management in 17% ofparticularly the lung, is not possible because air disrupts patients.171,190 –196 Thoracoscopy has a signiﬁcant advan-ultrasound conduction. Identiﬁcation of abnormal-ap- tage in detecting thoracic lymph nodes and when com-pearing lymph nodes has been found to be speciﬁc (85%– bined with laparoscopy appears to be superior to EUS,97%) for nodal metastasis but not very sensitive (72%– PET, and CT staging.171,192,197 These series generally do77%) in case series.175,176 If abnormalities are found, not have uniform staging procedures performed beforebiopsies can be performed on the lesions using a small the laparoscopy, which makes it difﬁcult to assess theaspiration needle positioned in a linear array instrument. role of these procedures in relation to the standard stag-The sensitivity of this technique in tumor staging has ing modalities. Bronchoscopy has been suggested as abeen reported in clinical series between 59% and 90%. staging procedure for patients with tumors that areSome of this variation can be attributed to the difﬁculty found above the tracheal bifurcation. In one prospectivein passing the ultrasound instrument through a malig- study, almost 10% of patients were found to have tra-nant stricture.177 Histologic conﬁrmation of nodal dis- cheal involvement on the basis of bronchoscopy alone,ease altered planned therapy in only 13% of patients although EUS was not used in this study.198 Magneticwith esophageal cancer in one retrospective case series.178 resonance endoscopy has also been used in staging esoph-This may be because the majority of esophageal cancers ageal cancer but as yet has not been shown to be superiorare found at an advanced stage. Miniprobes that have to existing tests.199,200 Recent cost analyses of thesehigher resolution are the best at deﬁning early-stage various staging modalities alone and in combination havecancer.179 Decision analysis models have been used to suggested that CT combined with EUS with ﬁne-needledetermine the value of EUS in assessment of esophageal aspiration if appropriate offered the least expensive andcancer. These models have found that EUS is a more reasonable efﬁcacy in terms of QALYs.201 PET and EUScost-effective initial staging strategy if the probability plus ﬁne-needle aspiration offered more QALYs but wasthat EUS can ﬁnd advanced disease is 30% or if the also more expensive at $60,544 per QALY.cost of EUS is less than 3.5 times that of CT.180 Addi- Summary of Evidence—Given the state of the currenttional analysis based on the use of EUS in a national information, we recommend the following strategy. CTdatabase found that preoperative use of this technique of the chest and upper abdomen should be the ﬁrstcould prevent 26% of patients from unnecessary com- staging procedure, followed by EUS and ﬁne-needle as-bined-modality therapy.181 piration if no evidence of distant metastasis is found on Positron emission tomography (PET) is a technology CT and the procedure is available. If surgical resection isthat uses 18F-ﬂuorodeoxyglucose for the detection of still considered, PET can be considered if available due tonodal or distant metastasis in esophageal cancer.182 This its increased sensitivity for distant metastasis. This isis used to detect neoplastic tissues because they normally based on level III evidence. In patients with potentiallymetabolize glucose at a faster rate than normal tissues. early-stage disease (tumors 2 cm and nonobstructing),However, inﬂammatory tissues are also fast glucose me- EUS with endoscopic mucosal resection may be consid-tabolizers and metastases often have to be differentiated ered as an alternative staging procedure if available forfrom inﬂamed tissue, leading to false positives.183 A histologic staging of the cancer and potential therapy.number of case series have found that PET is not assensitive as EUS or CT for locoregional disease.184 –189 Treatment of Esophageal CancerThis technology cannot deﬁne the tumor stage because itcannot resolve the layers of the esophagus. In addition, Treatment of Early Cancerspatients with hyperglycemia are not good candidates for Early esophageal cancers, those conﬁned in thePET. Because of these factors, PET cannot be envisioned mucosa or upper submucosa of the esophagus, are termedas an initial staging tool in esophageal cancer. New T1, N0, M0 by the American Joint Commission onadvances in PET technology include fusion PET, which Cancer terminology (see Table 5). There have not been