November 2008 PREDICTORS OF SURGICAL VERSUS ENDOSCOPIC THERAPY 1207 Methods The porﬁmer sodium PDT protocol at our institution has been described in detail previously.18 In general, this consists of The study was performed at Massachusetts General 2 mg/kg porﬁmer sodium intravenously on day 0, followed byHospital, an urban academic teaching hospital that serves as a 2 light exposures (day 2 and day 4). Light delivery is via atertiary referral center. For an individual patient with Barrett’s cylindric diffusing ﬁber (without a centering balloon), at aHGD or esophageal carcinoma, referral for gastroenterological, wavelength of 630 nm and energy of 150 J/cm. EMR is per-surgical, or oncologic consultation, respectively, is at the dis- formed with either a saline-assisted cap and snare technique, orcretion of the referring physician. a band ligator device (Duette; Cook Medical, Bloomington, IN). Identiﬁcation of study subjects was performed through the Statistical analysis was performed using SAS V 9.1.3 (SASResearch Patient Data Registry, an institutional database con- Institute, Cary, NC). Univariate analysis was performed withsisting of comprehensive records from more than 3.5 million testing of signiﬁcance by t test or Wilcoxon rank-sum testingpatients dating back to 1979. The study was approved by the for continuous variables (depending on assumptions of nor-institutional review board. mality of distribution), and either the chi-square or the Fisher The Research Patient Data Registry was queried to identify exact test for categoric variables. After univariate analysis, apatients with an International Classiﬁcation of Diseases–9th multiple variable logistic regression analysis was performed.revision (ICD-9) code diagnosis for Barrett’s esophagus Model inputs were selected on the basis of a priori hypotheses,(530.85). Demographic data including sex, ethnicity, date of without the use of a variable selection algorithm. For one modelbirth, and vital status also were captured. Additional search covariate, the length of Barrett’s esophagus, mean imputationcriteria included ICD-9 code diagnosis of comorbid conditions was used to account for missing data. All reported P values areincluding esophageal cancer, coronary artery disease, congestive 2-sided.heart failure, and chronic obstructive pulmonary disease (seeAppendix 1). Dates of all diagnoses were determined on thebasis of date of initial ICD-9 coding. Results After identiﬁcation and characterization of the cohort, the Patient Characteristicssearch was restricted further to identify Barrett’s esophaguspatients who had undergone esophagectomy or endoscopic The database query identiﬁed 2107 individual patientstherapy. For Barrett’s esophagus patients who had undergone with Barrett’s esophagus at our institution between January 1,either surgical therapy, endoscopic therapy, or both, individual 2003 and July 17, 2007. The mean age at diagnosis of Barrett’sendoscopy, surgical, and pathology reports were reviewed to esophagus was 61.2 14.8 years. Sixty-four percent of theconﬁrm the diagnosis of Barrett’s esophagus and the perfor- cohort were men, and 88% were Caucasian. Comorbid diag-mance of therapy targeted at Barrett’s-associated neoplasia. The noses were present as follows: coronary artery disease in 15%medical record was reviewed to conﬁrm the presence or absence (307 of 2107), congestive heart failure in 7% (150 of 2107), andof comorbid medical conditions. Patients were not excluded chronic obstructive pulmonary disease in 6% (132 of 2107).from the analysis on the basis of comorbid illness or perceivedﬁtness for surgery. Treatment Groups The length of Barrett’s esophagus was determined by the dis- Esophagectomy was performed in 82 patients, and en-tance between the squamocolumnar junction and the upper extent doscopic ablation was performed in 137 patients. Three pa-of the gastric folds, as described in either upper-gastrointestinal tients underwent esophagectomy for indications other thanendoscopy or esophagoscopy reports. Pretreatment neoplasia stage Barrett’s esophagus–associated neoplasia (Boerhaave’s syn-was determined on the basis of biopsy and endoscopic ultrasound drome or esophageal perforation, squamous cell carcinoma)(EUS) ﬁndings. In instances of discrepancy between biopsy and and were excluded from further analysis. Fifty-seven patientsEUS staging, the higher of the 2 stages was selected. Neoplasia had undergone endoscopy with ablation or destruction of tis-stage was classiﬁed into 3 groups: (1) lesions limited to the mucosa sue other than Barrett’s esophagus, most commonly resection(encompassing HGD, Tis, and T1a lesions); (2) T1 cancer with of fundic gland polyps, and were excluded from further analy-submucosal invasion (T1sm or T1b lesions); and (3) stage T2 or sis. Endoscopic ablation techniques consisted of argon plasmagreater. Biopsy and EUS ﬁndings were recorded either on the basis coagulation in 1 patient, and, in the remaining patients, eitherof biopsy and/or EUS performed at our institution, or as docu- PDT (N 37 patients, 36 with porﬁmer sodium photosensi-mented in the clinical record based on reports from evaluations tizer and 1 with aminolevulinic acid photosensitizer), EMRperformed at referring institutions. A central review of all outside (N 32 patients), or a combination of PDT and EMR (N 10pathology specimens was not performed. patients). The type of initial subspecialty consultant (gastroenterolo- Treatment designation was determined on the basis of pri-gist, surgeon, or oncologist) was deﬁned as the treating or mary treatment intent. As such, the esophagectomy group in-evaluating clinician in the earliest clinic note or endoscopy cluded 2 patients taken to the operating room for a plannedreport at our institution. If the patient’s initial documented esophagectomy, but in whom the esophagectomy was abortedmedical encounter at our institution was an endoscopy or EUS intraoperatively; and 4 patients who had undergone EMR forwith referral from a physician outside the institution, as docu- diagnostic/staging purposes before esophagectomy. The abla-mented in the endoscopy report, then the gastroenterologist tion group included 2 patients who underwent PDT, but whoperforming the endoscopy/EUS was deﬁned as the initial sub- later underwent esophagectomy.specialty consultant. However, if the referring physician for theendoscopy was a surgeon or an oncologist from our institution, Patient Characteristics by Treatment Groupthen that physician was deﬁned as the initial subspecialty con- The treatment cohort ultimately consisted of 79 pa-sultant. tients in the esophagectomy group, and 80 patients in the
1208 YACHIMSKI ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 11Table 1. Patient Characteristics by Treatment Modality greater disease were each treated for palliative purposes or without curative intent. In contrast, a wider range of pretreat- Esophagectomy Ablation P value ment cancer stage was observed among patients undergoingN 79 80 esophagectomy (Table 1). The proportion of patients undergo-Age at treatment, y 63.1 10.6 69.7 9.4 .0001 ing endoscopic ablation (vs esophagectomy) was as follows: 71%Male sex 73 (92%) 64 (80%) .02 (70 of 98) among patients with HGD/IMC, 43% (6 of 14)Vital status: deceased 10 (13%) 8 (10%) .60 among patients with T1sm disease, and 9% (4 of 47) amongLength of Barrett’s 6.6 3.7 5.4 3.7 .03 patients with T2 or greater disease. esophagus, cmaComorbid diagnoses Treatment Modality by Initial Subspecialty Coronary artery disease 19 (24%) 21 (26%) .75 Consultant Congestive heart failure 6 (8%) 5 (6%) .74 Chronic obstructive 5 (6%) 9 (11%) .27 Among patients with Barrett’s-associated neoplasia pulmonary disease who underwent endoscopic ablation, a gastroenterologist wasCancer stage the initial consultant in 95% (76 of 80) of patients. In contrast, HGD/IMC 28 (35%) 70 (88%) .0001 a wider range of initial consultants was evident in patients T1sm 8 (10%) 6 (8%) undergoing esophagectomy (Table 1). T2 43 (54%) 4 (5%) Initial consultation was performed by a gastroenterologistInitial consultant for the majority of patients with HGD or IMC (78%; 76 of 98). Gastroenterologist 32 (41%) 76 (95%) .0001 A surgeon was the initial consultant for 21 patients with HGD Surgeon 42 (53%) 4 (5%) or IMC. Among patients with HGD/IMC, crude esophagectomy Oncologist 5 (6%) 0 rates were 12% (9 of 76) if the initial consultation was per-aPretreatment length of Barrett’s esophagus not recorded in 24 pa- formed by a gastroenterologist versus 86% (18 of 21) if thetients in the esophagectomy group, and 4 patients in the ablation initial consultation was performed by a surgeon (P .0001).group. Among patients with T2 or greater disease (N 47), a gastroenterologist was the initial consultant in 23 patients and a surgeon was the initial consultant in 20 patients. Oncologistsendoscopic ablation group (Table 1). Patients undergoing were the initial consultant for relatively few patients (3%; 5 ofesophagectomy were on average younger at the time of treat- 159), irrespective of cancer stage.ment than patients undergoing endoscopic ablation, with amean age of 63.1 10.6 years in the esophagectomy group Point of Referralversus 69.7 9.4 years in the ablation group (P .0001) (Table 1). Data regarding the referring physician were availableEsophagectomy rates were highest in patients aged 60 years or for 126 patients. Referring physicians consisted of gastroenter-younger, whereas ablation rates were highest in patients aged 80 ologists in 75% of patients (95 of 126), internists or familyyears or older. practitioners in 16% of patients (20 of 126), surgeons and A greater proportion of patients were men in the esophagec- oncologists each in 4% of patients (5 of 126), and an emergencytomy group compared with the ablation group (92% vs 80%, room physician in 1 case.P .02). Patients in the esophagectomy group had longer Eighty-one percent (86 of 106) of referrals to gastroenterol-length Barrett’s esophagus than patients in the ablation group ogists at our institution came from another gastroenterologist,(6.6 3.7 vs 5.4 3.7 cm, P .03) (Table 1). 14% (15 of 106) came from an internist or family practitioner, The prevalence of documented coronary artery disease, con- and 2% (2 of 106) came from a surgeon. Forty-seven percent (8gestive heart failure, and chronic obstructive pulmonary disease of 17) of referrals to surgeons at our institution came fromdid not appear to differ between the esophagectomy and abla- gastroenterologists, 29% (5 of 17) came from internists or fam-tion groups (Table 1). ily practitioners, and 18% (3 of 17) came from another surgeon. Treatment Modality by Cancer Stage Postsurgical Pathology Among Patients With Among patients for whom both pretreatment tissue High-Grade Dysplasia/Intramucosaldata and pretreatment EUS (performed at our institution or a Carcinomareferring institution) were available, there was agreement in Thirty patients underwent esophagectomy after preop-disease stage between tissue diagnosis and EUS diagnosis in erative evaluation indicating disease limited to the mucosa84% of patients (105 of 125). Thirteen patients with HGD/ (HGD/IMC). Among these patients, disease was limited to theintramucosal carcinoma (IMC) had their pretreatment stage mucosa in surgical resection specimens for 80% (24 of 30).upgraded on the basis of EUS ﬁndings: 9 patients with tissue Cancer stage was understaged pretreatment in 20% of patientsdiagnosis of HGD/IMC were diagnosed with T1sm disease on (6 of 30), including 5 patients found to have T1sm disease inthe basis of EUS ﬁndings, and 4 patients with tissue diagnosis the surgical specimen and 1 patient with more deeply invasiveof HGD/IMC were diagnosed with T2 or greater disease on the adenocarcinoma.basis of EUS ﬁndings. Alternatively, for 7 patients, biopsy spec-imens and/or EMR showed invasive disease not detected by Logistic Regression AnalysisEUS. A logistic regression analysis was performed using a Barrett’s-associated neoplasia was staged as HGD or IMC in 4-variable model: patient age, pretreatment cancer stage, lengththe majority of patients (88%) undergoing endoscopic ablation of Barrett’s esophagus, and initial consultant seen. Unadjusted(Table 1). The 4 patients in the ablation group with stage T2 or and adjusted odds ratios are presented in Table 2.
November 2008 PREDICTORS OF SURGICAL VERSUS ENDOSCOPIC THERAPY 1209Table 2. Logistic Regression Model: Predictors of patients with high surgical risk because of comorbid illness may Esophagectomy be inappropriate surgical candidates. Speciﬁcally, among pa- tients undergoing esophagectomy for HGD at experienced cen- Unadjusted Adjusted OR OR 95% CI ters, including our own institution, a 2% surgical mortality rate has been reported as a low-end estimate.10,21Age 60 y 4.15 4.95 1.65–14.9 Endoscopic therapy, although less invasive than esophagec-Cancer stage T1sm 12.2 16.0 5.60–45.6 tomy, is not without procedural morbidity. Quality of life afterLength of Barrett’s esophagus 1.10 1.06 0.92–1.23 porﬁmer sodium PDT, for instance, may be impacted by pho-Initial consultant (surgeon vs 24.9 35.1 9.58–129 tosensitivity of several weeks’ duration. Esophageal stricture gastroenterologist) develops in 23% of 27% of patients after porﬁmer sodium PDT for HGD or IMC, and may require multiple endoscopic dila- tions for palliation.18,22 Independent predictors of esophagectomy, after adjusting This study does not address efﬁcacy or long-term outcomesfor other variables in the model, were as follows: age 60 years or of either esophagectomy or endoscopic ablation for treatmentyounger (odds ratio [OR], 4.95; 95% conﬁdence interval [CI], of Barrett’s-associated neoplasia. Although esophagectomy is1.65–14.9), cancer stage T1sm or greater (OR, 16.0; 95% CI, considered the more deﬁnitive treatment by many, caveats5.60 – 45.6), and initial consultation performed by a surgeon should accompany assurances of long-term cure for either ap-(OR, 35.1; 95% CI, 9.58 –129). proach. Recurrent Barrett’s23 and adenocarcinoma24 have been Including patient sex in the model did not signiﬁcantly alter described after esophagectomy. Alternatively, in a prospectivethe model results. Creation of an interaction variable as a model study of HGD patients receiving PDT, 15% developed cancerinput did not support the inﬂuence of cancer stage as an effect over a 5-year follow-up period.14 The risk of recurrent HGDmodiﬁer of the association between the initial consultant (sur- or cancer is a particular concern among patients with glan-geon vs gastroenterologist) and esophagectomy. dular epithelium buried beneath neosquamous mucosa after PDT.25,26 Discussion Moreover, accurate disease staging is a critical prerequisite to For patients with Barrett’s esophagus and HGD or selecting appropriate therapy for Barrett’s neoplasia. Endo-IMC, there are neither prospective controlled data nor consen- scopic ablation or resection theoretically can eradicate neopla-sus algorithms to guide treatment strategy. The goal of this sia when this is limited to the esophageal mucosa and morestudy was to identify predictors of surgical versus endoscopic advanced disease has been excluded. In the current era oftherapy in a tertiary center offering both treatment options. systematic biopsy protocols, as well as staging provided by EMROur data suggest that patient age and cancer stage indepen- and EUS, the prevalence of occult carcinoma in patients withdently predict whether a patient will receive surgical versus Barrett’s HGD appears to be declining. A retrospective analysisendoscopic therapy. Patients ages 60 years or younger and from Johns Hopkins reported that occult cancer was found incancer stage T1sm or greater are signiﬁcant predictors of sur- 43% of esophagectomy specimens for HGD between 1982 andgical versus endoscopic therapy. Our data do not support an 1994, but in 17% of specimens between 1994 and 2001.27 Ainﬂuence of cardiac or pulmonary comorbidity on choice of more recent analysis has shown that when a distinction is madetherapy, although the method of capturing comorbidity data between mucosal-limited and invasive disease among historical(by ICD-9 coding) may have been limited in its ability to gauge cohorts found to have occult carcinoma after esophagectomythe presence or severity of comorbid illness. for HGD, the prevalence of occult invasive carcinoma is 12.7%.28 Our data also suggest that choice of therapeutic modality, High rates of metastatic lymphadenopathy are present in pa-particularly for patients with HGD/IMC, depends on whether tients with submucosal tumor invasion by EMR,29 and as suchthe patient is evaluated by a surgeon or a gastroenterologist. patients with T1sm disease are not ideal candidates for endo-This likely reﬂects a referral or selection effect. Presumably, a scopic therapy.particular patient with HGD/IMC is referred for esophagec- Patients in our study did not undergo uniform pretreatmenttomy because he/she has been deemed an appropriate surgical evaluation, which was instead performed at the discretion ofcandidate. Other patients may be deemed more suitable for the treating gastroenterologist or surgeon. In some instances,endoscopic therapy. Yet the factors that inﬂuence this treat- this evaluation may have taken place at an outside institutionment decision are largely unknown. before referral. Variability in performance or interpretation of One set of factors may be related to endoscopic or biopsy histopathologic, endoscopic, or imaging ﬁndings may impactfeatures of Barrett’s esophagus. Our model controlled for the the accuracy of pretreatment cancer stage. However, our studylength of Barrett’s esophagus, which did not appear to predict deﬁned pretreatment cancer stage on the basis of the individualtreatment modality. However, our model did not control for data available to treating clinicians, which are the data thatextent of HGD or IMC. A decision for surgical or endoscopic drive treatment decisions in clinical practice.therapy could be quite different for a patient with a nodular Additional limitations of this study include its retrospectivefocus of dysplasia, as compared with a patient with a long design, although our multivariate logistic regression modelcircumferential extent of Barrett’s with extensive HGD in mul- attempted to control for some of the potential biases. Thetiple biopsy locations. results have not been validated prospectively and should there- A second set of factors may be related to a patient’s overall fore be considered hypothesis-generating. To the extent thathealth status. Among Medicare patients undergoing esophagec- availability and expertise of surgical and/or endoscopic therapytomy, surgical mortality rate varies considerably according to for Barrett’s esophagus is likely to vary among medical centers,individual surgeon and hospital surgical volumes.19,20 As such, external validity of our ﬁndings cannot be assumed. A prior
1210 YACHIMSKI ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 11survey of Barrett’s practice patterns indicated that ablation Congestive heart failure: 428, 428.0, 428.1, 428.22, 428.30,therapy is more likely to be performed in an academic rather 428.31, 428.32, 428.33, 428.9.than a community setting.5 In our analysis, 86% of HGD/IMC Chronic obstructive pulmonary disease: 491.20, 491.21,patients evaluated by a surgeon underwent esophagectomy, as 491.8, 491.22, 491.9, 492, 492.0, 492.8.compared with 12% of HGD/IMC patients evaluated by a gas- Esophagectomy: 42.40, 42.41, 42.42, 42.51, 42.54, 42.58,troenterologist. The latter esophagectomy rate is considerably 42.62.lower than the 73% to 83% esophagectomy rate recommended Endoscopic excision or destruction of lesion or tissue of thefor HGD in national surveys of gastroenterologists,12,13 al- esophagus: 42.33.though it is conceivable that national practice patterns haveevolved over the past decade. Our study failed to identify sub- Referencesjects before 2003 on the basis of the speciﬁed search criteria 1. Falk GW. Barrett’s esophagus. Gastroenterology 2002;122:because an ICD-9 code speciﬁc for Barrett’s esophagus was not 1569 –1591.used before this time. As such, our study cannot effectively 2. Shaheen NJ, Crosby NA, Bozymski EM, et al. Is there publicationassess longer-term trends in endoscopic versus surgical man- bias in the reporting of cancer risk in Barrett’s esophagus?agement of HGD at our institution. Gastroenterology 2000;119:587–589. In addition to physician recommendations, the potential 3. Rastogi A, Puli S, El-Serag HB, et al. Incidence of esophagealrole of patient preferences in shared decision making should be adenocarcinoma in patients with Barrett’s esophagus and high-considered. Cancer risk may affect health-related quality of life grade dysplasia: a meta-analysis. Gastrointest Endosc 2008;67: 394 –398.for a patient with Barrett’s.30 In a survey of Barrett’s patients 4. Rice TW, Falk GW, Achkar E, et al. Surgical management ofpresented with the hypothetical scenario of HGD, the majority high-grade dysplasia in Barrett’s esophagus. Am J Gastroenterolpreferred frequent endoscopic surveillance over either esopha- 1993;88:1832–1836.gectomy or PDT.31 Patient choice among these strategies is 5. Edwards MJ, Gable DR, Lentsch AB, et al. The rationale forlikely to be inﬂuenced by perceived posttreatment quality of esophagectomy as the optimal therapy for Barrett’s esophaguslife, including presence or absence of dysphagia and risk of with high-grade dysplasia. Ann Surg 1996;223:589 –591.cancer.32 The impact of patient self-referral for endoscopic or 6. Heitmiller RF, Redmond M, Hamilton SR. Barrett’s esophagussurgical therapy for Barrett’s could not be assessed in this with high-grade dysplasia: an indication for prophylactic esopha-retrospective analysis, yet may be important in an era in which gectomy. Ann Surg 1996;224:66 –71. 7. Ferguson MK, Mannheim KS. Resection for Barrett’s mucosa withthe Internet and other sources of patient information are high-grade dysplasia: implications for prophylactic photodynamicreadily available. therapy. J Thorac Cardiovasc Surg 1997;114:824 – 829. The magnitude of the difference in esophagectomy rates for 8. Nigro JJ, Hagen JA, DeMeester TR, et al. Occult esophagealHGD/IMC between surgeons and gastroenterologists in our adenocarcinoma: extent of disease and implications for effectivestudy might be noteworthy from the patient’s perspective, and therapy. Ann Surg 1999;230:433– 440.further underscores the requirement that treatment decisions 9. Zaninotto G, Parenti AR, Ruol A, et al. Oesophageal resection forbe based on the best available outcome, efﬁcacy, and safety data. high-grade dysplasia in Barrett’s oesophagus. Br J Surg 2000;This burden of proof may be borne increasingly by gastroen- 87:1102–1105.terologists because of the role of the gastroenterologist as 10. Reed MF, Tolis G Jr, Edil BH, et al. Surgical treatment of esoph-endoscopist and endoscopic ultrasonographer. In this study, ageal high-grade dysplasia. Ann Thorac Surg 2005;79:1110 – 1115.75% of referrals came from gastroenterologists, and 78% (76 of 11. Sujendran V, Sica G, Warren B, et al. Oesophagectomy re-98) of HGD/IMC patients were ﬁrst seen at our institution by mains the gold standard for treatment of high-grade dysplasiaa gastroenterologist. As endoscopic technologies for ablation of in Barrett’s oesophagus. Eur J Cardiothorac Surg 2005;28:Barrett’s mucosa continue to develop, the extent to which data 763–766.emerge, and the extent to which practicing gastroenterologists 12. Gross CP, Canto MI, Hixson J, et al. Management of Barrett’sbecome widely proﬁcient in the application of these techniques esophagus: a national study of practice patterns and their costmay inﬂuence the volume of both endoscopic ablations and implications. Am J Gastroenterol 1999;94:3440 –3447.esophagectomies performed for HGD/IMC. 13. Falk GW, Ours TM, Richter JE. Practice patterns for surveillance In summary, at our institution between 2003 and 2007, of Barrett’s esophagus in the United States. Gastrointest Endosc 2000;52:197–203.more patients with Barrett’s HGD/IMC underwent endoscopic 14. Overholt BF, Wang KK, Burdick JS, et al. Five-year efﬁcacy andablation than underwent esophagectomy. Among all patients safety of photodynamic therapy with Photofrin in Barrett’s high-with Barrett’s-associated neoplasia, age of 60 years or younger grade dysplasia. Gastrointest Endosc 2007;66:460 – 468.and cancer stage T1sm or greater predicted esophagectomy over 15. Prasad GA, Wang KK, Buttar NS, et al. Long-term survivalendoscopic therapy. Whether a patient is ﬁrst evaluated by a following endoscopic and surgical treatment of high-grade dys-gastroenterologist or a surgeon also appears to inﬂuence treat- plasia in Barrett’s esophagus. Gastroenterology 2007;132:ment modality; this effect persists even when controlling for age 1226 –1233.and cancer stage. 16. SSAT Patient Care Guidelines: management of Barrett’s esoph- agus. J Gastrointest Surg 2007;11:1213–1215. 17. Wang KK, Sampliner RE. Updated guidelines 2008 for the diag- Appendix 1 nosis, surveillance and therapy of Barrett’s esophagus. Am J Gastroenterol 2008;103:788 –797. ICD 9 used in database search. 18. Yachimski P, Puricelli WP, Nishioka NS, et al. Predictors of esoph- Esophageal cancer: 150.0, 150.1, 150.2, 150.3, 150.4, 150.5, ageal stricture development following photodynamic therapy. Clin150.8, 150.9. Gastroenterol Hepatol 2008;6:302–308. Epub 2008 Feb 6. Coronary artery disease: 414.0, 414.1, 414.8, 414.9, 414.00. 19. Birkmeyer JD, Siewers AE, Finlayson EVA, et al. Hospital volume
November 2008 PREDICTORS OF SURGICAL VERSUS ENDOSCOPIC THERAPY 1211 and surgical mortality in the United States. N Engl J Med invasive esophageal cancer in high-grade dysplasia in Barrett’s 2002;346:1128 –1137. esophagus overestimated? Clin Gastroenterol Hepatol 2008;6:20. Birkmeyer JD, Stukel TA, Siewers AE, et al. Surgeon volume and 159 –164. Epub 2007 Dec 21. surgical mortality in the United States. N Engl J Med 2003;349: 29. Prasad GA, Buttar NS, Wongkeesong LM, et al. Signiﬁcance of 2117–2127. neoplastic involvement of margins obtained by endoscopic mu-21. Headrick JR, Nichols FC 3rd, Miller DL. High-grade esophageal cosal resection in Barrett’s esophagus. Am J Gastroenterol dysplasia: long-term survival and quality of life after esophagec- 2007;102:2380 –2386. tomy. Ann Thorac Surg 2002;73:1697–1702. 30. Gerson LB, Ullah N, Hastie T, et al. Does cancer risk affect22. Prasad GA, Wang KK, Buttar NS, et al. Predictors of stricture health-related quality of life in patients with Barrett’s esophagus? formation after photodynamic therapy for high-grade dysplasia in Gastrointest Endosc 2007;65:31–35. Barrett’s esophagus. Gastrointest Endosc 2007;65:60 – 66. 31. Hur C, Wittenberg E, Nishioka NS, et al. Patient preferences for23. Chaves P, Pereira AD, Cruz C. Recurrent columnar-lined esopha- the management of high-grade dysplasia in Barrett’s esophagus. geal segments—study of the phenotypic characteristics using Dig Dis Sci 2005;50:116 –125. intestinal markers. Dis Esophagus 2002;15:282–286. 32. Hur C, Wittenberg E, Nishioka NS, et al. Quality of life in patients24. Wolfsen HC, Hemminger LL, DeVault KR. Recurrent Barrett’s with various Barrett’s esophagus associated health states. esophagus and adenocarcinoma after esophagectomy. BMC Health Qual Life Outcomes 2006;4:45. Gastroenterol 2004;4:18.25. Mino-Kenudson, Ban S, Ohana M, et al. Buried dysplasia and early carcinoma arising in Barrett esophagus after porﬁmer- photodynamic therapy. Am J Surg Pathol 2007;31:403– Address requests for reprints to: Patrick Yachimski, MD, Gastroin- 409. testinal Unit, Blake 4, Massachusetts General Hospital, 55 Fruit Street,26. Van Laethem JL, Peny MO, Salmon I, et al. Intramucosal carci- Boston, Massachusetts 02114. e-mail: email@example.com; fax: noma arising under squamous re-epithelialisation of Barrett’s (617) 724-6832. oesophagus. Gut 2000;46:574 –577. P.Y. and N.S.N. have received consulting honoraria from Axcan27. Tseng EE, Wu TT, Yeo CJ, et al. Barrett’s esophagus with high Scandipharm. P.Y. was supported by NIH T32 training grant grade dysplasia: surgical results and long-term outcome—an DK007191. C.H. was supported by NIH grant K07CA10706O. update. J Gastrointest Surg 2003;7:164 –170. A version of these data were presented in abstract form at28. Konda VJ, Ross AS, Ferguson SK, et al. Is the risk of concomitant Digestive Disease Week, May 17-22, 2008, San Diego, CA.