Tumor length (expressed in centimeters) determined by preoperative endoscopy for patients with and without associated Barrett mucosa.
Degree of tumor differentiation for patients with and without associated Barrett mucosa.
Actuarial survival, including deaths from all causes, in the entire series of 215 patients with adenocarcinoma. Data are given as mean (SD).
Actuarial survival, including deaths from all causes in the Barrett and no Barrett groups (61% vs 28%, P < .001).
Actuarial survival, including deaths from all causes in the Barrett and no Barrett groups for American Joint Committee on Cancer stage II (A) and American Joint Committee on Cancer stage III disease (B) (57.6% vs 61.7%, P = .89 and 19.6% vs 10.4%, P = .45, respectively).
Actuarial survival, including deaths from all causes in the Barrett and no Barrett groups when tumors at the gastroesophageal junction are excluded (n = 162) for overall survival (A) (62.7% vs 22.5%, P < .001), American Joint Committee on Cancer stage II disease (B) (51.7% vs 46.6%, P = .745), and American Joint Committee on Cancer stage III disease (C) (20.9% vs 11.8%, P = .25).
Portale G, Peters JH, Hagen JA, DeMeester SR, Gandamihardja TAK, Tharavej C, Hsieh C, DeMeester TR. Comparison of the Clinical and Histological Characteristics and Survival of Distal Esophageal–Gastroesophageal Junction Adenocarcinoma in Patients With and Without Barrett Mucosa. Arch Surg. 2005;140(6):570–575. doi:10.1001/archsurg.140.6.570
The incidence of adenocarcinoma in the distal esophagus and at the gastroesophageal junction (GEJ) has been increasing in the last decades. It has been suggested that patients in whom Barrett mucosa can be identified in the surgical specimen have a better prognosis compared with those without. This has led to the belief that patients with and without Barrett mucosa may represent 2 distinct cancer types.
Distal esophageal–GEJ adenocarcinoma with and without Barrett mucosa share the same origin, but differ only in clinical presentation and outcome.
Design and Setting
Retrospective cohort study in a university tertiary referral center.
Patients and Methods
Between 1992 and 2002, 215 patients (173 men and 42 women; median age, 66 years; age range, 26-91 years) had esophagogastrectomy for adenocarcinoma of the distal esophagus-GEJ. Patients receiving neoadjuvant chemotherapy or radiation therapy were excluded.
Main Outcome Measures
Clinical presentation, tumor characteristics, and survival were compared in patients with Barrett mucosa (n = 140) and those without (n = 75).
Patients with Barrett mucosa in the specimen had tumors that were diagnosed earlier; were smaller in size; earlier in stage, with fewer node metastases; and had a better 5-year survival.
Observed differences in survival between patients with distal esophageal–GEJ adenocarcinoma with and without Barrett mucosa can be explained by earlier diagnosis. Patients without Barrett mucosa have their tumors detected later, when the disease is more advanced. This suggests the possibility that tumors without Barrett mucosa are not of a different origin, but rather are larger tumors that may have overgrown areas of Barrett mucosa.
The relationship between gastroesophageal reflux disease (GERD) and esophageal adenocarcinoma is clear for tumors located in the tubular esophagus, regardless of whether Barrett mucosa is found in the resected esophagus. Controversy exists regarding the origin of tumors located near the gastroesophageal junction (GEJ) where 2 distinct cancer types could potentially arise. When Barrett mucosa is found adjacent to the tumor, the causative role of reflux is accepted. It has been proposed that GEJ tumors without associated Barrett mucosa represent a different tumor that originates in the gastric mucosa of the proximal stomach.1 The major evidence cited in support of this concept is the observation that prognosis differs for tumors with and without adjacent Barrett mucosa.2- 4 An alternative explanation is that when no Barrett mucosa is found in association with the tumor, the tumor may have overgrown the Barrett mucosa from which it originated.5 The aim of this study was to compare demographic information, clinical features, and tumor characteristics in patients with adenocarcinoma of the distal esophagus and GEJ with and without adjacent Barrett mucosa to determine whether these represent distinct tumor types.
From January 1992 to December 2002, 263 patients underwent esophagectomy for adenocarcinoma of the distal esophagus or GEJ. Forty-eight patients were excluded from analysis because they received neoadjuvant chemotherapy and/or radiation therapy, which is known to affect the detection of Barrett mucosa.6 The medical records of the remaining 215 patients were reviewed and demographic information, clinical features, and tumor characteristics were recorded. Pathology reports were reviewed to determine whether intestinal metaplasia consistent with Barrett epithelium was identified adjacent to the tumor. Table 1 lists the types of operation performed. Patients were followed up by the operating surgeon at regularly scheduled intervals for a median of 25 months (interquartile range [IQR], 11-48 months).
Proportions were compared using Fisher exact test. Continuous variables were compared using the Mann-Whitney test. Survival estimates were calculated according to the method of Kaplan and Meier, with comparisons performed using the log-rank test. A P<.05 was considered statistically significant.
Of the 215 patients with adenocarcinoma of the distal esophagus or GEJ included in this study, 140 (65%) had Barrett mucosa identified in the surgical specimen (hereafter referred to as the “Barrett group”). In 75 patients (35%) no Barrett mucosa could be found despite a detailed examination of the specimen (hereafter referred to as the “no Barrett group”). Demographic information and the reason for initial evaluation are summarized in Table 2. Of interest, 85% of patients in the no Barrett group presented with dysphagia or gastrointestinal bleeding, whereas most patients in the Barrett group were identified during surveillance endoscopy for Barrett esophagus or at endoscopy prompted by worsening of foregut symptoms.
Patients in the Barrett group had significantly shorter tumors on preoperative endoscopy than those in the no Barrett group (1 cm vs 5 cm, P<.001). Sixty-two percent (87/140) of the adenocarcinomas in the Barrett group were 2 cm or shorter compared with only 12% (9/75) in the no Barrett group (P<.001) (Figure 1). Tumor characteristics based on careful examination of the surgical specimen are shown in Table 3. Tumors in the Barrett group were shorter in length, less deeply invasive, and associated with fewer node metastases. This resulted in significantly earlier staged tumors in the Barrett group. On microscopy, tumors in the Barrett group were more differentiated than those in the no Barrett group (65% grade 1-2 vs 41%, P = .003) (Figure 2). Signet ring adenocarcinomas were seen in both groups but were more common in the no Barrett group (29.7% vs 5.6%, P<.001).
Actuarial survival for all 215 patients is depicted in Figure 3. Five-year survival was 49%, with a median survival of 59 months. Survival in the 140 patients in the Barrett group was significantly better than in the no Barrett group (5-year survival 61% vs 28%, P<.001) (Figure 4). Survival curves for American Joint Committee on Cancer stage II and III disease are shown in Figure 5. No significant difference was observed for either stage between the Barrett and no Barrett groups. Similar comparisons were not done for stage I and IV disease, because there was only 1 patient in the no Barrett group with a stage I tumor, and only 3 patients in the Barrett group had stage IV disease.
A similar analysis was performed limited to patients with tumors in the tubular esophagus where the Barrett etiology is less controversial. Identical results were found (Table 4 and Figure 6).
Adenocarcinoma commonly occurs in the distal esophagus or GEJ in the absence of detectable Barrett epithelium. When Barrett epithelium is identified, the reflux origin of the tumor is undisputed. Controversy persists, however, regarding the origin of adenocarcinomas near the GEJ in the absence of detectable Barrett epithelium. It has been suggested, on the basis of observed differences in survival, that tumor with or without Barrett epithelium represents 2 distinct cancer types. Although we have also observed a difference in survival between the Barrett and no Barrett groups, our results suggest an alternative explanation.
We have shown that tumors occurring in the absence of detectable Barrett epithelium are less differentiated, longer in length, more deeply invasive, more frequently associated with node metastases, and more apt to be diagnosed in an advanced clinical state. Consequently, they represent a group of more advanced staged tumors. Although overall survival in the 2 groups was significantly different, survival by tumor stage was similar. This suggests that adenocarcinomas without detectable Barrett epithelium may not be a different type of tumor but rather a more advanced stage of the same disease. It is likely that these larger more aggressive tumors have overgrown the Barrett epithelium in which they arose. This concept is supported by several biochemical and molecular studies.7- 9
Accepting the concept that adenocarcinomas with and without associated Barrett mucosa represent one disease would expand significantly the number of cancers considered to be related to reflux. Based on the most recent Surveillance, Epidemiology, and End Results (SEER) cancer registry data,10 the number would nearly double. This has several important implications. First, the potential for loss of life due to malignancy from gastroesophageal reflux disease is even greater than currently appreciated. Second, it suggests that intestinal metaplasia limited to the GEJ (cardiac intestinal metaplasia [CIM]) and short segments of Barrett deserve equal emphasis to that placed on traditional long-segment Barrett esophagus. Recognition of the malignant potential of long-segment Barrett esophagus has led to screening and surveillance programs, which have resulted in improved prognosis when a cancer develops.11- 14 It is likely that a similar approach for patients with CIM or short-segment Barrett esophagus would also result in an improved prognosis for cancers occurring near the GEJ. Third, endoscopists should apply the same biopsy protocols to patients with short segments of columnar-lined esophagus as they do to longer segments. Fourth, in patients with reflux symptoms, biopsy specimens of the squamous columnar junction should be taken routinely even if the junction appears normal. If intestinal metaplasia is found in the absence of a Helicobacter pylori infection of the stomach, surveillance should be instituted.
Correspondence: Tom R. DeMeester, MD, Department of Surgery, University of Southern California, 1510 San Pablo St, Los Angeles, CA 90033 (firstname.lastname@example.org).
Accepted for Publication: January 31, 2005.
Previous Presentation: This paper was presented at the 112th Scientific Session of the Western Surgical Association; November 8, 2004; Las Vegas, Nev; and is published after peer review and revision. The discussions that follow this article are based on the originally submitted manuscript and not the revised manuscript.
Acknowledgement:Statistical expertise: J. A. Hagen.
Claude Deschamps, MD, Rochester, Minn: Once again, Dr DeMeester’s group has provided us yet with another relevant topic to ponder. The study presented here is simple and retrospective. The implications could be far reaching. I would like to congratulate G. Portale for a fine presentation.
If I look at the message, the level of evidence of the data, and its utility, here is what I recognize:
The message is in the findings. In a retrospective study on patients who had potentially curative surgery for esophageal carcinoma, those without Barrett metaplasia identified in the resected specimen were worse off. Their tumor was more advanced, and they had a worse prognosis. They were also less likely to have been in a surveillance program.
The authors conclude that the 2 cancers are one and the same and that the observed difference in survival can be explained by earlier diagnosis. That is fair. In their article, they recommend that surveillance should be instituted in patients with intestinal metaplasia of the cardia (CIM) and further, that a biopsy specimen should be obtained even if the mucosa looks normal. Now, it would not be the first time that Dr DeMeester clearly sees a solution while the rest of us are still trying to solve the problem; in this case however, based on the level of evidence, these recommendations are slightly overenthusiastic and the implications for applying the authors’ recommendations to the population at risk are huge.
Regarding the utility of all this, I would like to direct the authors to 2 of their recent papers [Balaji NS, DeMeester SR, Wickramasinghe KS, Hagen JA, Peter JH, DeMeester TR. Etiology of intestinal metaplasia at the gastroesophageal junction. Surg Endosc. 2003;17:43-48, and DeMeester SR, Wickramasinghe KS, Lord RVN, et al. Cytokeratin and DAS-1 immunostaining reveal similarities among cardiac mucosa, CIM, and Barrett’s esophagus. Am J Gastroentrol. 2002;97:2514-2523] on the topic of the etiology of CIM and the role of molecular markers. They should probably refer to these in their articles. These important and relevant publications of 2002 and 2003 are, paradoxically, the natural follow-up to the present study. If the reader retrofit the work presented here today, the whole thing makes great sense. I will quote the DeMeester et al (2002) who wrote in 1 of these 2 papers: “Consequently, future studies that carefully assess both the type of intestinal metaplasia and the likely etiology of CIM will be necessary to clarify the natural history and malignant risk of this abnormality.” I totally agree.
The authors are pointing us in the right direction as they are trying to provide another building block towards establishing evidence that most adenocarcinomas of the distal esophagus–GEJ are related to Barrett metaplasia and reflux.
I have 3 questions. In your group of patients in a surveillance program, how long were they surveyed? Exactly 50% of the patients had a transhiatal esophagectomy. That is more than I would have expected considering your group’s strong bias for “en bloc resection.”
Were the DeMeesters out of town? Please remind us which criteria you use to pick one operation over the other? Can we use today molecular markers to follow up patients with Barrett esophagus and CIM? I thank the association for the fantastic opportunity of discussing this article and the authors for sending me the manuscript.
Dr Peters: Thank you Claude for your insightful comments, picking up on the important points of course. Let me comment on the evidence issue. This is certainly not level I evidence. I think it classifies as level III evidence as retrospective case series. We are glimpsing into the future here and the assertion that all patients should have biopsy specimens of their GEJs, which is the outcome of these data, may be a bit of a stretch. We would acknowledge that. Nevertheless, every future prediction must begin with some sort of evidence. We would certainly encourage a more rigorous prospective trial to see if our thesis does, indeed, hold true over time. We feel that it likely would.
Regarding the origin of CIM, there are basically 2 causes: Helicobactor pylori and reflux. If one obtains a biopsy specimen of the cardia and finds H pylori, we may be dealing with a different scenario, although we recognize H pylori is also a carcinogen, thus it may not make that much difference either way. With regard to your specific 3 questions, we do not have the answer to the surveillance mean follow-up time. I would suspect it is measured in years. Thirty percent of the entire population was in a surveillance group. We did not calculate the time that they were in surveillance as it was not one of the primary end points of the study. What about en bloc esophagectomy? On average, about 30% to 50% of the population have visible but curable tumors in which we recommend an en bloc esophagectomy. We still feel strongly that that is an important part of the armamentarium. We are seeing 20% to 25%, maybe even 30% of the patients with nonvisible, extremely small or biopsy-only cancers. This population has moved more toward transhiatal or vagal-sparing esophagectomy. Finally, your question regarding molecular markers. It is unclear if that will pan out to be the ultimate answer to these questions. So far my read of these data is that the molecular analysis of cardia tumors and esophageal tumors are not different enough yet for us to be able to precisely identify if some may be gastric origin. We need to do more work there. It may give us the answer in the end but I am uncertain that it will ultimately.
Kathrin Mayer, MD, Sacramento, Calif: I was also very fascinated by the publication from your group within the last couple of years about the potential regression of esophageal dysplasia. Based on your data today and on your previous work on this very important topic, how aggressively should we be screening the population with reflux and how aggressively should we obtain a biopsy specimen from these patients?
Dr Peters: I guess that gets to Dr Deschamps’ first point as to how much weight you put into class II or III evidence. We prefer an aggressive approach with routine biopsy specimens obtained at the GEJ. If you do that, you certainly are opening up Pandora’s box about where to go once you find intestinal metaplasia, and as most of you in the room know, the gastroenterology community and others feel that we should not routinely obtain a biopsy specimen because we do not have the answer to what to do with it if you find GEJ metaplasia. We feel it is better to know than not know and make decisions on that information, even though we do not have the final conclusion about how to manage it.
James Debord, MD, Peoria, Ill: One of your conclusions that the patients that did worse had bigger tumors that covered up the Barrett mucosa somewhat obviates your theory that you have 2 groups; those with and without Barrett esophagus. Maybe your 2 groups really are patients diagnosed as having Barrett esophagus who are in a surveillance program and undiagnosed Barrett that are not in a surveillance program.
Dr Peters: Actually the thesis is that we have 1 group; they all have Barrett esophagus. We are not trying to make the argument that there are 2 groups. In contrast, we believe they are 1 group, 1 entity, all arising from Barrett mucosa. As a matter of fact, there is an interesting observation from Munich, Germany, in which patients who had chemoradiotherapy started with approximately 60% of the population with Barrett and after neoadjuvant therapy ended up with 100% of the population with Barrett; that is, they shrunk the tumor and uncovered the Barrett.
Sherry Wren, MD, Palo Alto, Calif: This article supports the role of surveillance in this population. Did you analyze this specific group of patients with Barrett to compare screened vs unscreened? Did the unscreened patients have similar stages to the patients without Barrett changes?
Dr Peters: We have compared surveillance and no surveillance in the past, as have others, including some in this room. The answer is if you pick out the surveillance population, they clearly do much better. They have much earlier tumors. By and large, they have carcinoma in situ or early stage I tumors and can be cured. The demographics of those 2 populations are otherwise similar, although overall survival and tumor stage is quite different because they have been in a surveillance program. This fact has been shown by 6 or 7 different authors, but was not the aim of this study.
John Hunter, MD, Portland, Ore: Jeff, you set the stage nicely for my question in quoting the Munich data because I was going to ask whether you had any similar observations in your data set. Were patients with stage III tumors who went through preoperative neoadjuvant chemoradiation therapy more likely to have the detection of Barrett at operation than patients who did not receive neoadjuvant chemoradiation therapy?
Dr Peters: We do not know largely because we have not taken a liberal attitude toward neoadjuvant therapy. As a matter of fact, we have done the opposite. Few of our patients have neoadjuvant chemotherapy. Our current treatment algorithm generally begins with surgery, whatever the appropriate operation may be, followed by postoperative chemotherapy in node-positive patients. So we really do not have that population to look at as clearly as they did in Munich and elsewhere.