L was termed Barrett s esophagus (BE) after the

Transcription

1 ORIGINAL ARTICLES Barrett s Esophagus With High-Grade Dysplasia: An Indication for Esophagectomy? Manuel Pera, MD, Victor F. Trastek, MD, Herschel A. Carpenter, MD, Mark S. Allen, MD, Claude Deschamps, MD, and Peter C. Pairolero, MD Section of General Thoracic Surgery, Department of Surgery, and Department of Laboratory Medicine and Pathology, Mayo Clinic and Mayo Foundation, Rochester, Minnesota Between 1982 and 1991, 19 patients (17 men and 2 women) with Barrett s esophagus, 10 of whom were in a surveillance program, were found to have high-grade dysplasia without evidence of invasive carcinoma. Median age was 66 years (range, 30 to 79 years). Heartburn was the most common presenting symptom. Esophagoscopy at the time of high-grade dysplasia diagnosis demonstrated normal Barrett s mucosa in 10 patients (53%), shallow ulcers in 3, slight mucosal irregularities in 2, small mucosal nodules in 2, stricture in 1, and shallow ulcer with stricture in 1. Eighteen patients underwent esophagectomy. There were no operative deaths. Nine patients (50%) had invasive carcinoma. Postsurgical stage was stage 0 in 9 patients, stage I in 6, stage IIA in 2, and stage IIB in 1. Median follow-up was 34 months (range, 2 to 116 months). Recurrent cancer developed in 2 patients. Overall 5-year survival was 66.7%; 5-year survival for patients with stage 0 disease was 100% and for stage I and I1 disease, 35.7%. We conclude that high-grade dysplasia is an indication for esophageal resection because of the high rate of associated early invasive carcinoma and that resection can be done safely with the expectation of excellent long-term survival. Because of these findings, we continue to recommend endoscopic surveillance in all patients with Barrett s esophagus. (Ann Thorac Surg 1992;54: ) ining of the lower esophagus by columnar epithelium L was termed Barrett s esophagus (BE) after the presentation by Normal Barrett in 1957 [l]. Although this finding has been attributed to Barrett, the work of others, including Tileston [2] and Allison and Johnstone [3], preceded Barrett s description. Other complications involving this condition have been subsequently reported, and all have included the eponym Barrett s. Today, Barrett s ulcer, Barrett s stricture, and Barrett s carcinoma are all well recognized, implying that this process and its inherent complications can best be described as Barrett s disease. The association of adenocarcinoma with Barrett s disease was first reported by Morson and Belcher in Since then, adenocarcinoma has increased in incidence, especially during the past two decades [5, 61. Dysplasia has been observed in 83% to 100% of resected esophageal adenocarcinomas [7-91 and supports the hypothesis of a dysplasia to carcinoma sequence. Others have suggested high-grade dysplasia (HGD) as a marker of associated invasive adenocarcinoma [lo]. The natural history, however, of Barrett s disease is not completely understood, and, consequently, management of those patients with HGD remains controversial. Esophagectomy has been recommended, whereas others have proposed surveillance if no evidence of concomitant invasive carcinoma is present [ 11-13]. Unfortunately, the paucity of reports concerning Barrett s disease leaves this controversy unre- Presented at the Twenty-eighth Annual Meeting of The Society of Thoracic Surgeons, Orlando, FL, Feb 3-5, Address reprints to Dr Trastek, 200 First St, SW, Rochester, MN solved, and the proof that dysplasia is premalignant has been mostly circumstantial. In an effort to provide additional data, we reviewed our experience with HGD in Barrett s disease. Material and Methods Between January 1982 and September 1991, 19 patients with HGD in BE were evaluated at the Mayo Clinic. No patients had evidence of invasive adenocarcinoma. The diagnosis of BE was based on the histologic finding of at least 3 cm of columnar epithelium in the distal esophagus or of intestinal-type goblet cells observed in any esophageal biopsy specimen. All patients had a second biopsy to confirm the initial diagnosis of HGD. Because of the presence of HGD, esophagectomy was offered to all patients. All microscopic tissue was reexamined by one of us (H.A.C.). Dysplasia was defined as unequivocal neoplastic nuclear abnormalities in the absence of invasion of the lamina propria. Nuclear dysplasia comprised a spectrum of progressive abnormalities characterized by nuclear enlargement, elongation, hyperchromatism, and increased numbers of mitoses, usually associated with crowding of cells, loss of polarity, and stratification of nuclei. Dysplasia was graded using the criteria established for inflammatory bowel disease [13, 141. Low-grade dysplasia consisted of mild to moderate nuclear dysplasia combined with nuclear stratification limited to the lower half of the epithelium (Fig 1A). High-grade dysplasia consisted of by The Society of Thoracic Surgeons /92/$5.00

2 200 PERAETAL Ann Thorac Surg 1992;54: B Fig 1. (A) Specialized mucosa demonstrating low-grade dysplasia. The dysplasia nuclei are enlarged, hyperchromatic, and stratified in the lower half of the epithelium. (B) Specialized mucosa demonstrating high-grade dysplasia. The dysplastic nuclei are stratified throughout the entire thickness of the epithelium. (C) lnvasive (intramucosal) adenocarcinoma. (D) Higher magnification. The neoplastic glands are crowded and infiltrating the lamina propria. (All slides stained with hematoxylin and eosin. Magnification: A, X200; B, X200; C, X40; D, x400; all before 40% reduction.) D either mild to moderate nuclear dysplasia combined with nuclear stratification extending through the entire thickness of the epithelium or severe nuclear dysplasia with or without nuclear stratification (Fig 1B). Carcinoma in situ was classified as HGD. Intramucosal carcinoma was invasive carcinoma limited to the lamina propria (Figs lc, 1D). All tumors were classified postsurgically by the TNM classification system of the American Joint Committee on Cancer Staging [15]. Operative mortality included all patients who died within the first 30 postoperative days and those who died later but during the same hospitalization. Survival was estimated by the method of Kaplan and Meier [16] using the date of esophagectomy as the starting point. Results There were 19 patients with HGD arising in BE. None had any evidence of invasive carcinoma. Seventeen were men and 2 were women. All were white. Median age at the time of HGD diagnosis was 66 years (range, 30 to 79 years). Heartburn was the most common presenting symptom, occurring in 14 patients for a median of 20 years (range, 4 to 40 years). Six patients had dysphagia; 2 of them had stricture. Two patients presented with chest pain, and 1 with postprandial discomfort. Five patients (26.3%) presented with anemia or hematemesis, 1 of whom had a previous esophagogastrectomy for HGD elsewhere 3 months previously. Postoperative esophagoscopy in this latter patient demonstrated a residual 3 cm segment of Barrett s metaplasia with HGD. Two patients had previous antireflux procedures. Upper gastrointestinal roentgenography was done in all patients; it demonstrated a diaphragmatic hernia in 14 and midesophageal stricture in 2, and it was normal in 3. All patients had esophagogastroduodenoscopy, which demonstrated normal-appearing Barrett s mucosa in 10 patients; the remaining 9 had subtle mucosal findings that were not considered indicative of cancer: shallow ulcer, 3; slight mucosal irregularity, 2; small mucosal nodule, 2; shallow ulcer below stricture, 1; and stricture, 1. The

3 Ann Thorac Surg 1992;5419%204 PERAETAL 201 Table 1. Postsurgical Esophageal Stage Patient Endoscopic Postsurgical No. Findings Pathologic Findings Stage Outcome 1 Normal BE HGD Tis NO, stage 0 Alive 4 mo 2 Normal BE HGD Tis NO, stage 0 Alive 7 mo 3 Normal BE HGD Tis NO, stage 0 Alive 2 mo 4 Stricture and shallow ulcer HGD Tis NO, stage 0 Alive 14 mo 5 Normal BE HGD Tis NO, stage 0 Alive 38 mo 6 Normal BE HGD Tis NO, stage 0 Alive 42 mo 7 Small nodule HGD Tis NO, stage 0 Alive 64 mo 8 Normal BE HGD Tis NO, stage 0 Alive 106 mo 9 Shallow ulcer HGD Tis NO, stage 0 Alive 116 mo 10 Normal BE ACA T1 NO, stage I Died 29 mo, no recurrence 11 Shallow ulcer ACA T1 NO, stage I Died 72 mo, recurrence 12 Normal BE ACA T1 NO, stage I Alive 5 mo 13 Small nodule ACA T1 NO, stage I Alive 29 mo 14 Irregular mucosa ACA T1 NO, stage I Alive 46 mo 15 Normal BE ACA T1 NO, stage I Alive 56 mo 16 Shallow ulcer ACA T2 NO, stage IIA Alive 41 mo 17 Stricture ACA T2 NO, stage IIA Alive 84 mo 18 Irregular mucosa ACA T1 N1, stage IIB Died 31 mo, recurrence ACA = carcinoma; BE = Barrett s esophagus; HGD = high-grade dysplasia median length of BE was 6.6 cm; BE length ranged from 2 to 13 cm. High-grade dysplasia in 10 patients was diagnosed through surveillance. Median time interval between initial diagnosis of BE and diagnosis of HGD was 29 months and ranged from 11 to 97 months. These 10 patients had a total of 42 endoscopic procedures (10 initially and 32 sequentially). The remaining 9 patients had HGD diagnosed during their initial evaluation for gastrointestinal symptoms. All 19 patients were offered esophageal resection. One patient refused operation. The remaining 18 patients underwent esophagectomy, 3 of whom, however, had a prolonged delay after diagnosis of HGD (7, 14, and 18 months). An Ivor Lewis esophagogastrectomy was the procedure most commonly performed (12 patients). Five patients had a transhiatal esophagectomy, and 1 had colon interposition. The entire length of Barrett s mucosa was resected in all patients. There were no operative deaths. Complications developed in 4 patients (22.2%), which included a cervical leak in 2, both of which resolved with conservative management; chylothorax in 1, which required ligation of the thoracic duct; and colon necrosis in 1, which was managed with diversion and delayed reconstruction with a cervical esophagojejunostomy. Median hospitalization was 16.5 days and ranged from 12 to 44 days. At dismissal, all patients were eating a soft, general diet. Microscopic examination demonstrated invasive adenocarcinoma in 9 patients (50%). Six patients were postsurgically classified as stage I (T1 NO MO); 2 as stage IIA (T2 NO MO); and 1 as stage IIB (T1 N1 MO). The remaining 9 patients, who were all considered to have in situ carcinoma, were classified as stage 0 (Tis NO MO). Specialized epithelium was present in 16 patients and specialized and cardiac in 2. The 3 patients who had delayed esophageal resection had stage 0 disease. Three patients (33.3%) with normal-appearing Barrett s mucosa and 6 patients (66.7%) with abnormal-appearing Barrett s mucosa had invasive carcinoma (Table 1). Five of 9 patients undergoing resection who were found to have HGD in a surveillance program were postsurgical stage 0, and 4 were stage I. In contrast, 3 of 9 patients who were found to have HGD at the time of initial evaluation were postsurgical stage 0, 3 were stage I, 2 were stage IIA, and 1 was stage IIB (Table 2). Postsurgical stage varied with the length of BE. The 9 patients with stage 0 disease had a median length of 6.5 cm (range, 2 to 10 cm), and the 6 patients with stage I disease had a median length of 5.0 cm (range, 2.8 to 8.0 cm). In contrast, the 3 patients with stage IIA and IIB disease had a median length of 12 cm (range, 10 to 13 cm). Follow-up was complete in all patients and ranged from 2 to 116 months (median, 34 months). The 1 patient who refused operation is still under surveillance and is alive and well at 29 months without evidence of invasive cancer. Recurrent cancer developed in 2 patients, 1 at 57 months after resection (stage I) and 1 at 19 months (stage IIB). Both died, 1 at 72 months and the other at 31 months, respectively. The remaining 16 patients had no evidence of recurrent disease; 1, however, died of myocardial infarction at 29 months. Estimated overall 5-year survival for all 18 patients was 66.7% (Fig 2). Five-year survival for the 9 patients with stage 0 disease was 100%; 5-year survival for the 9 with either stage I or I1 disease was 35.7% (Fig 3). Two patients, both of whom had transhiatal esophagectomy, had development of dysplasia secondary to anastomotic narrowing and required dilations during the first year with complete relief of symptoms. Two other patients had mild dumping symptoms and a third had mild gastric outlet obstruction, which responded to balloon

4 202 PERA ETAL Ann Thorac Surg 1992;54: Table 2. Surveillance Time Elapsed Time Elapsed From HGD No. of From BE to HGD Diagnosis to Endoscopies Patient Diagnosis Esophagectomy During No. (mo) (mo) Surveillance Stage I I I I refused, a Patient numbers same as in Table 1. BE = Barrett s esophagus; HGD = high-grade dysplasia. dilation of the pylorus. At follow-up, all patients were eating a general diet. Comment Barrett and other early investigators who developed the concept of the columnar-lined lower esophagus could not have known that their contribution would play such a large role in later attempts to determine the pathophysiology of esophageal adenocarcinoma. The importance of understanding this process is undeniably clear; hopefully, this knowledge will lead to early cancer detection with improved survival in what is otherwise a fatal disease. Just why Barrett s metaplasia develops in a subset of patients with gastroesophageal reflux and why dysplasia and cancer develops in yet a smaller group still needs further elucidation. What is known, however, is that Barrett s metaplasia is probably due to both gastric and duodenal content reflux [17, 181, that adenocarcinoma of the esophagus and gastroesophageal junction has increased during the last two decades, and that most, if not all, adenocarcinomas are associated with Barrett s mucosa [5, 6, 9, 191. Whatever the reason for this increase, we believe that the ability to detect and intervene in this process of carcinogenesis will result in improved survival for these patients; thus, the identification of markers, such as HGD, is extremely important. Most pathologists agree that HGD represents carcinoma in situ. The interval from the discovery of Barrett s metaplasia to the development of HGD is unknown; however, in our limited patient population who had surveillance, the median interval from the diagnosis of BE to the development of HGD was 29 months. Similarly, the length of time from HGD to invasive cancer is not known. Our study demonstrated that at least half of all patients with HGD had invasive carcinoma at the time of HGD diagnosis. We also demonstrated that HGD can persist for years without progressing, as observed in 3 of our patients who had delayed resection but still were in postsurgical stage 0. Others have reported similar findings [ll, 13, Consequently, we believe that once HGD is confirmed, resection should be performed if the patient is otherwise in good health, as our study demonstrates that surveillance and resection for HGD results in a larger number of patients with stage 0 and I disease. We recognize that this recommendation is controversial and that a management spectrum exists, ranging from those who prefer to continue surveillance until an invasive carcinoma is confirmed [12, 13, 221 to those who suggest early esophagectomy [lo, 11, 21, 23, 241. Although the risk of esophagectomy is not minimal, mortality has steadily decreased over the past few decades [ and was zero in this study; early morbidity, however, continues to be substantial, but long-term function is satisfactory [26]. Consequently, we believe that the opportunity for early intervention in an otherwise hopeless disease outweighs the risks of resection. The goals of resection are to completely remove all areas of Barrett s metaplasia and to thoroughly examine lymph nodes histologically. Complete removal must be emphasized because adenocarcinoma arising in residual BE after esophagectomy has been reported [28, 291. Both the Ivor Lewis and transhiatal esophagectomy attain these goals. Our preference depends on the length of Barrett s mucosa, and most often either a high thoracic or neck esophagogastric anastomosis is required. The goal of cancer surveillance in patients with Barrett s I Stage 0 Percent 60 Survival 40 5 *O 0 I Years after esophagectomy Fig 2. Estimated survival of 18 patients with Barrett s esophagus undergoing esophagectomy for high-grade dysplasia. Zero time on abscissa is date of operation. Percent Survival 4o 2o F ob I I I \ Stage I & II Years after esophagectomy Fig 3. Estimated survival of 9 patients with postsurgical stage 0 esophageal carcinoma undergoing esophagectomy for high-grade dysplasia as compared with survival of 9 patients with either postsurgical stage 1 or stage 11 disease. Zero time on abscissa is date of operation. I I

5 Ann Thorac Surg 1992; PERAETAL 203 ESOPHAGECTOMY FOR BE WITH DYSFLASIA disease is to detect esophageal neoplasm in an early curable stage. The importance of surveillance cannot be overemphasized, and we prefer an annual four-quadrant biopsy every 2 cm of Barrett s metaplasia. We believe that in this group given enough time, HGD will develop in many patients, and invasive carcinoma (hopefully detected at an early stage) will then develop in most of these, a fact suggested by the development of invasive cancer in 44% of our surveillance patients, none of whom were in stage 11. In contrast, 67% of patients with HGD found at the time of initial evaluation had invasive cancer and half were in stage 11. Identification of high-risk patients may also lead to earlier diagnosis and treatment. All of our patients were white and most of them were male; the median age was 66 years. Most had a long history of gastroesophageal reflux. Others have observed similar findings [7, 8, 301. Thus, evaluating white, older men with a long history of heartburn specifically for the presence of BE appears reasonable, and, if BE is found, yearly surveillance should be instituted. Patients who have subtle endoscopic mucosal abnormalities have a greater probability of having advanced disease. In our series, two-thirds of these patients had invasive carcinomas, half of which were stage 11, as compared with only one-third invasive cancers, all of which were stage I, in those patients with normalappearing mucosa. Similarly, the length of Barrett s metaplasia may also correlate with the extent of invasive carcinoma. The median length in our series was 6.6 cm and was similar to that reported by others [ll]. Although our patient numbers are far too small to make any conclusions, the length of BE did increase with advanced stages of carcinoma. However, others have not found any relationship, and, in fact, have reported invasive carcinoma arising in short segments of Barrett s esophagus [19,31]. Thus, the endoscopic findings of either abnormal Barrett s mucosa or extended length may well indicate the need for increased surveillance with more frequent examinations. However, normal-appearing BE does not exclude an early invasive carcinoma, a fact confirmed in our study and reported by others [lo, 13, 321. Hopefully, data from surveillance, when combined with new techniques of molecular cytologic analysis such as flow cytometry [33] and p53 oncogene identification [34], will ultimately determine which patients are at risk for the development of carcinoma in the future. In conclusion, we believe that HGD is an indication for esophageal resection because of the high rate of associated early invasive carcinoma and that the resection can be performed safely with the expectation of excellent long-term survival. Because of these findings, we continue to recommend surveillance in all patients with BE. Dr Manuel Pera is supported by a grant from the Fundacion Ramon Areces (1990/1991), Madrid, Spain. References 1. Barrett NR. The lower esophagus lined by columnar epithelium. Surgery 1957;41: Tileston W. Peptic ulcer of the oesophagus. Am J Med Sci 1906;132: Allison PR, Johnstone AS. The oesophagus lined with gastric mucous membrane. Thorax 1953;8: Morson BC, Belcher JR. Adenocarcinoma of the oesophagus and ectopic gastric mucosa. Br J Cancer 1952;6: Blot WJ, Devesa SS, Kneller RW, et al. Rising incidence of adenocarcinoma of the esophagus and gastric cardia. JAMA 1991;265: Pera M, Cameron AJ, Trastek VF, et al. Esophageal adenocarcinoma: a population-based study of its incidence and the role of Barrett s esophagus [Abstract]. Gastroenterology 1991;100:A Haggitt RC, Tryzelaar J, Ellis FH, et al. Adenocarcinoma complicating columnar epithelium-lined (Barrett s) esophagus. Am J Clin Pathol 1978;70: Smith RR, Hamilton SR, Boitnott JK, et al. The spectrum of carcinoma arising in Barrett s esophagus. A clinicopathologic study of 26 patients. Am J Surg Pathol 1984;8: Thompson JJ, Zeinsser KR, Enterline HT. Barrett s metaplasia and adenocarcinoma of the esophagus and gastroesophageal junction. Hum Pathol 1983;14: Hamilton SR, Smith RRL. The relationship between columnar epithelial dysplasia and invasive adenocarcinoma arising in Barrett s esophagus. Am J Clin Pathol 1987; Altorki NK, Sunagawa M, Little AG, et al. High-grade dysplasia in the columnar-lined esophagus. Am J Surg 1991; 161: DeMeester TR, Attwood SE, Smyrk T, et al. Surgical therapy in Barrett s esophagus. Ann Surg 1990;212: Reid BJ, Weinstein WM, Lewin KJ, et ai. Endoscopic biopsy can detect high-grade dysplasia or early adenocarcinoma in Barrett s esophagus without grossly recognizable neoplastic lesions. Gastroenterology 1988;94: Riddle RH, Goldman H, Ransohoff DF. Dysplasia in inflammatory bowel disease: a standardized classification with provisional clinical applications. Hum Pathol 1983; American Joint Committee on Cancer. Manual for staging of cancer, edition 3. Philadelphia: Lippincott, Kaplan EL, Meir P. Nonparametric estimation from incomplete observation. J Am Stat Assoc 1958;53: Iascone C, DeMeester TR, Little AG, et al. Barrett s esophagus: functional assessment, proposed pathogenesis, and surgical therapy. Arch Surg 1983;118: Attwood SEA, DeMeester TR, Bremner C, et al. Alkaline gastroesophageal reflux: implication in the development of complications in Barrett s columnar-lined esophagus. Surgery 1989;106: Hamilton SR, Smith RRL, Cameron JL. Prevalence and characteristics of Barrett s esophagus in patients with adenocarcinoma of the esophagus or esophagogastric junction. Hum Pathol 1988;19: Lee RG. Dysplasia in Barrett s esophagus: a clinicopathological study of six patients. Am J Surg Pathol 1985;9: Robertson CS, Mayberry JF, Nicholson DA, et al. Value of endoscopic surveillance in the detection of neoplastic change in Barrett s oesophagus. Br J Surg 1988;75: Hameeteman W, Tytgat GNJ, Houthoft HJ, et al. Barrett s esophagus: development of dysplasia and adenocarcinoma. Gastroenterology 1989;96: Palley SL, Sampliner RE, Gazewal HS. High-grade dysplasia in Barrett s esophagus. J Clin Gastroenterol 1989;11: Ellis FH. Barrett s esophagus: a continuing conundrum. Postgrad Med 1991;90: Ellis FH, Gibb SP, Watkins E. Esophagogastrectomy. A safe, widely applicable and expeditious form of palliation for uatients with carcinoma of the esouhaeus and cardia. Ann I Y kurg 1983;198: King RM.. Pairolero PC. Trastek VF, et al. Ivor Lewis esouhagogastrectomy for carcinoma of the esophagus: early and late functional results. Ann Thorac Surg 1987;44: Orringer MB. Transthoracic versus transhiatal esophagectomy: what difference does it make? Ann Thorac Surg 1987;44: Hamilton SR, Hutcheon DF, Ravich WJ, et al. Adenocarcinoma in Barrett s esophagus after elimination of gastroesophageal reflux. Gastroenterology 1984;86:

6 204 PERAETAL Ann Thorac Surg 1992;54: Berenson MM, Riddell RH, Skinner DB. Malignant transformation of esophageal columnar epithelium. Cancer 1978;41: Skinner DB, Walter BC, Riddell RH. Barrett s esophagus. Comparison of benign and malignant cases. Ann Surg 1983; 32. Schmidt HG, Riddle RH, Walter B, et al. Dysplasia in Barrett s esophagus. Cancer Res Clin Oncol1985;110: Reid BJ, Haggitt RC, Rubin CE, et al. Barrett s esophagus. Correlation between flow cytometry and histology in detection of patients at risk for adenocarcinoma. Gastroenterology ;93: Schnell TG, Sontag SJ, Chejfec G. Adenocarcinoma arising in 34. Blount PL, Ramel S, Raskind WH, et al. 17p allelic deletion tongues or short segments of Barrett s esophagus. Dig Dis Sci and p53 protein overexpression in Barrett s adenocarcinoma. 1992;37: Cancer Res 1991;51: DISCUSSION DR F. HENRY ELLIS, JR (Boston, MA): All of us involved in the care of patients with carcinoma of the esophagus have long been anxious to improve long-term survival rates. Some have favored a super radical operation in certain favorable cases, whereas others have vigorously pursued various neoadjuvant protocols. There has been no major impact on long-term survival by any of these approaches. It has long been my belief that only early diagnosis will have any meaningful impact on long-term survival, a view that is supported by the data from mainland China where patients at high risk in areas where there is a high prevalence of esophageal cancer have been screened periodically using endoscopic brushings. Such a program has led to the early diagnosis of cancer resulting in an operative mortality rate approaching 100% and 5-year survival rates in the neighborhood of 90%. Patients with Barrett s esophagus constitute another group of people at risk for cancer of the esophagus, 75 times that of the normal population in our experience. The inevitable progression of high-grade dysplasia, which most pathologists now equate with carcinoma in situ, to invasive carcinoma suggests that such patients be treated by resection. This is the issue that Dr Pera and associates have so clearly addressed today. Their data show not only that half of their patients with high-grade dysplasia on biopsy exhibit invasive carcinoma in the resected specimen, but also that early resection improves long-term survival. Their survival rate of 67% at 5 years is certainly commendable and substantiates their view that resection should be offered to these patients. My associate, Dr Streitz, has accumulated data on 79 patients with adenocarcinoma in Barrett s esophagus who were seen from 1973 to 1992; in 19 a prior diagnosis of benign Barrett s esophagus had been made and they were followed up endoscopically at varying intervals, the median interval being 6 months. Eight of them had high-grade dysplasia or carcinoma in situ on biopsy but only 1 had invasive carcinoma in the resected specimen. Eighteen of these patients were operated on. One patient with high-grade dysplasia refused operation and returned a year later with a nonresectable tumor, illustrating the danger of waiting until invasive carcinoma occurs. When the results in patients who were under surveiilance are compared with those in whom cancer developed at the time the diagnosis of Barrett s esophagus was made there is a marked difference in the stage of the resected specimen. Seventy-nine percent of those under surveillance were in stages 0, I, and 11, whereas 57% of those not surveyed were in stages I11 and IV. There is also a marked difference in the 5-year survival rate: 67% of those under surveillance, identical to the Mayo experience, as compared with 22% of those not under surveillance. I would like to compliment Dr Pera and associates on a very important study and ask them to tell us how frequently patients with Barrett s esophagus should undergo endoscopy. Should patients with mild dysplasia undergo endoscopy more often than those without dysplasia, and do they think that frequent endoscopic surveillance will prove to be cost-effective? There are a number of published articles suggesting that such an approach is not cost-effective. DR PERA: Thank you, Dr Ellis, for your comments. I would like to say that currently at the Mayo Clinic, patients with Barrett s esophagus without dysplasia are recommended to have endoscopy every 12 months. When we find low-grade dyspiasia on a biopsy, the patient undergoes endoscopy at least every 6 to 12 months and maybe more frequently. Currently, I would say that doing surveillance for all patients with Barrett s esophagus, although it incurs costs, is the only way to shift the extent of the disease to an earlier stage, therefore improving survival. It is apparent that we need to continue to document our work in this area very closely to help answer these questions. DR DAVID B. SKINNER (New York, NY): I too would like to congratulate Dr Pera not only for this report but for his elegant experimental work on carcinogenesis in this condition. Two years ago Dr Altorki presented our results with esophagectomy when the sole indication was high-grade dysplasia, and exactly as in this report, 50% of our patients had invasive carcinoma in the resected specimen. So we strongly agree with Pera and associates contention that these people should have esophagectomy when the diagnosis is made and should be treated as cancer patients. About 10 years ago we reported that when reflux was controlled, some of the patients responded by growing back areas of squamous epithelium that would spread across the area of Barrett s epithelium. We performed an endoscopic study in a physician who had had Barrett s esophagus for some time and was asymptomatic on medical therapy. On the left we could see the Barrett s segment and way in the distance there was a white area, and as we went down the esophagus this white area was the spreading squamous epithelium. However, when the squamous epithelium was present the Barrett s mucosa remained underneath the squamous epithelium. This patient did progress to high-grade dysplasia, underwent resection, and had microinvasive carcinoma. We have all been wondering whether or not the spreading back of squamous epithelium would relieve us of the obligation to continue these people under surveillance, but this case tells us that is not the case.