|Home | About | Journals | Submit | Contact Us | Français|
Endoscopic eradication therapy is used to treat mucosal neoplasms in Barrett’s esophagus, but cannot cure cancers that have metastasized to lymph nodes. The risk of such metastases has been proposed as a reason to consider esophagectomy rather than endoscopic therapy for esophageal mucosal neoplasia. The objective of our study was to determine the frequency of lymph node metastases in patients with high-grade dysplasia (HGD) and intramucosal carcinoma in Barrett’s esophagus.
We performed a systematic review using the PRISMA guidelines to identify studies that included patients who had esophagectomy for HGD or intramucosal carcinoma in Barrett’s esophagus, and that reported final pathology results after examination of esophagectomy specimens.
We identified 70 relevant reports that included 1,874 patients who had esophagectomy performed for HGD or intramucosal carcinoma in Barrett’s esophagus. Lymph node metastases were found in 26 patients (1.39%, 95% CI .86% – 1.92%). No metastases were found in the 524 patients who had a final pathology diagnosis of HGD, whereas 26 (1.93%, 95% CI 1.19% –2.66%) of the 1,350 patients with a final pathology diagnosis of intramucosal carcinoma had positive lymph nodes.
The risk of unexpected lymph node metastases for patients with mucosal neoplasms in Barrett’s esophagus is in the range of 1% to 2%. Esophagectomy has a mortality rate that often exceeds 2%, with substantial morbidity and no guarantee of curing metastatic disease. Therefore, the risk of lymph node metastases alone does not warrant the choice of esophagectomy over endoscopic therapy for HGD and intramucosal carcinoma in Barrett’s esophagus.
Barrett’s esophagus is the precursor of esophageal adenocarcinoma, a lethal tumor whose frequency has increased at an alarming rate over the past several decades (1). Cancers in Barrett’s esophagus are thought to evolve through a sequence of genetic and epigenetic alterations that endow cells with growth advantages and cause histological changes in the metaplastic epithelium that pathologists recognize as dysplasia (2). For patients with Barrett’s esophagus, regular endoscopic surveillance is advised primarily to detect dysplasia, which is a potentially curable form of neoplasia. The management of low-grade dysplasia remains highly controversial, but most experts agree that high-grade dysplasia (HGD) in Barrett’s esophagus poses a sufficient risk for malignancy that intervention is warranted (3). Traditionally, that intervention was esophagectomy. Recently, endoscopic eradication has become the preferred therapy for HGD. The American Gastroenterological Association’s recent position statement on the management of Barrett’s esophagus concluded that, “Most patients with high-grade dysplasia (70%–80%) can be successfully treated with endoscopic eradication therapy (4).”
Endoscopic eradication therapy can eliminate neoplasms that are confined to the esophageal mucosa, but cannot cure cancers that have successfully metastasized to lymph nodes. For patients with esophageal adenocarcinomas that involve the submucosa, the frequency of lymph node metastases is at least 20% (5) and, therefore, endoscopic eradication therapy generally is deemed inadequate as a curative treatment. This is why endoscopic mucosal resection (EMR) is advised for T staging in patients who have Barrett’s esophagus with dysplasia that is associated with visible mucosal irregularities (3, 4). If EMR of such lesions reveals no submucosal invasion, then endoscopic eradication therapy might be appropriate. Implicit in the choice of endoscopic eradication therapy for dysplasia in Barrett’s esophagus, however, is the assumption that the neoplasm does not involve regional lymph nodes.
Unlike most hollow gastrointestinal organs, the esophagus has lymphatic vessels that frequently can be seen to extend into the mucosa (6). Therefore, even a neoplasm that appears to be confined to the mucosa (like HGD and intramucosal carcinoma) might have the potential to metastasize to lymph nodes through mucosal lymphatics. When considering treatment options for patients with mucosal neoplasms in Barrett’s esophagus, it seems important to know how often those lesions are associated with lymph node metastases, since such metastases would make it unlikely that endoscopic treatment could be curative. The frequency of lymph node involvement in this situation is not clear, however, and published series of patients who had esophagectomy for HGD or intramucosal carcinoma in Barrett’s esophagus have described a surprisingly wide range of frequencies for lymph node metastases. Indeed, some investigators have found rates of unexpected lymph node metastases so high that they recommend against endoscopic therapy for patients who are good operative risks (7). The aim of our study was to perform a systematic review to determine the rate of lymph node metastases in patients with HGD or intramucosal carcinoma in Barrett’s esophagus.
We performed a systematic review using the PRISMA guidelines (8). Two members of the study team (KBD, SJS) searched the PubMed search engine of MEDLINE-indexed literature from the National Center for Biotechnology Information (http://www.pubmed.gov), Embase, and the Cumulative Index to Nursing and Allied Health Literature (CINAHL) for English language articles published through September 2011. We performed additional searches of MEDLINE using the Ovid Medline database. Search terms included the medical subject heading (meSH) search term ‘Barrett’s esophagus’ (also ‘Barrett’s esophagus’, ‘Barrett oesophagus’, or ‘Barrett’s oesophagus’) combined with the keywords ‘lymph node’, ‘dysplasia’, ‘high grade dysplasia’, ‘intramucosal adenocarcinoma’, ‘neoplasia’, ‘T1’, and ‘esophagectomy’. Titles and abstracts were screened for eligibility, and potential studies meeting the inclusion criteria were reviewed. In addition, we reviewed the reference lists of selected articles to identify additional articles not retrieved by computer searches.
We decided a priori to include reports that: i) included patients who had esophagectomy for HGD, intramucosal carcinoma and/or early adenocarcinoma in Barrett’s esophagus; ii) reported final surgical pathology results with at least one of the following descriptors: high grade dysplasia (HGD), intramucosal carcinoma or intramucosal adenocarcinoma, pT1a, T1a, Tis, carcinoma in situ, mucosal cancer, or cancers involving m1, m2, m3 layers; and iii) reported surgical pathology lymph node status as N0 or N1, or described surgical pathology lymph node status as positive or negative. We decided a priori to exclude reports that: i) did not distinguish between mucosal (T1a) and submucosal (T1b) cancers; ii) did not distinguish esophageal adenocarcinomas from squamous cell carcinomas; iii) did not link lymph node status to the T stage; iv) were single patient case reports; v) reported only lymph node micrometastases; vi) included patients who had undergone preoperative radiation or chemotherapy; and vii) stated specifically that patients had been reported in other series; in these cases, we included only the most recent or the largest series with complete data. Questions regarding the eligibility of individual reports were resolved by discussion between the reviewers. A flowchart of the search strategy and results is shown in Figure 1.
From eligible studies, the data abstracted included the number of patients with a final pathology diagnosis of intramucosal cancer (T1a) or HGD, and the lymph node status of these patients. For papers including patients with both adenocarcinoma and squamous cell carcinoma, only the data pertaining to adenocarcinoma were abstracted. Likewise, in studies that included patients with more advanced esophageal adenocarcinoma (stages II–IV, T stages T1b-T4), only data pertaining to the early stage disease (T1a, HGD) were abstracted. Finally, we calculated the total number of patients, number of patients with HGD, number of patients with intramucosal carcinoma, and number of patients with lymph node metastases.
The prevalence rates were weighted for each study’s sample size and pooled with calculation of 95% confidence intervals. A secondary analysis using more restrictive inclusion criteria was also performed. The weighted prevalence rates between the primary and secondary analyses were compared using Fisher’s exact test. Statistical analyses were performed using SAS (Cary, North Carolina).
A total of 1,583 articles were generated from computer searches, and 10 more articles were identified through a review of reference lists. After removal of duplicate articles, 899 reports remained. Of these articles, 349 were excluded as they were review articles, case reports, guidelines, or letters to the editor, and 360 were unrelated to the topic of interest (Figure 1). One-hundred ninety full text articles were reviewed for eligibility, and 70 met the inclusion criteria. Of the 120 excluded, the most common reasons were a lack of distinction between T1a and T1b adenocarcinomas, no link between T and N stage, a mixed population of adenocarcinoma and squamous cell carcinoma patients where T and N stage could not be clearly assigned to the adenocarcinoma patients, or a combination of these factors.
Among the 70 reports that met our inclusion criteria, there were a total of 1,874 patients who had a final pathology diagnosis of intramucosal adenocarcinoma or HGD (Table 1). Relatively few reports focused specifically on the frequency of lymph node metastases in patients with HGD and intramucosal carcinoma. Rather, that information was gleaned from reports that focused on a number of other issues (summarized in Table 1) such as how the depth of tumor invasion (T stage) influenced the risk for lymph node metastases, the accuracy of endoscopic biopsy protocols, the utility of endoscopic ultrasound for staging, and the results of different esophagectomy techniques. The total enrollment of the individual studies ranged from 8 to 359 patients and, in most reports, patients who were eligible for our analyses (i.e. patients with HGD or intramucosal carcinoma in Barrett’s esophagus) comprised a relatively small minority of the total patient population. This is because many reports included patients with squamous cell carcinomas and advanced adenocarcinomas. In individual reports, the number of eligible patients ranged from 3 to 150.
Information on the number of lymph nodes resected during esophagectomy was provided in 32 of the 70 studies. The number of lymph nodes resected varied widely among the reports, ranging from a median of 0 in a series of minimally invasive, vagal-sparing esophagectomy to a median of 48 in a study of en bloc esophagectomy with 3-field lymph node dissection (9, 10). Nineteen of the 70 studies reported at least one patient with lymph node metastases, with rates of positive lymph nodes ranging from 0% to 66% of all patients in the series who had esophagectomy for HGD or intramucosal carcinoma. Combining the results from all studies, 26 of 1,874 patients with HGD or intramucosal carcinoma had lymph node metastases, for an overall rate of 1.39% (95% CI .86% – 1.92%).
Thirty-nine reports included patients who had a final pathology diagnosis of HGD (Table 2). The number of patients with HGD in individual studies ranged from 2 to 38, with a total of 524 patients. No patient with a final pathology diagnosis of HGD was found to have lymph node metastases.
Sixty-one reports included patients who had a final pathology diagnosis of intramucosal carcinoma (Table 3). The number of patients with intramucosal carcinoma in individual studies ranged from 2 to 150, with a total of 1,350 patients. Twenty-six of the 1,350 patients had lymph node metastases, for a rate of 1.93% (95% CI 1.19% – 2.66%).
For 15 of the 26 patients identified as having lymph node metastases, the reports provided some description of the tumor and lymph nodes. Among 9 patients for whom detailed information on the depth of tumor invasion was available, 8 had m3 tumors (which involve, but do not penetrate through, the muscularis mucosae) and 1 patient had an m2 lesion (involving the lamina propria but not the muscularis mucosae) (5, 14–20). Among the 12 studies that reported the depth of mucosal tumor invasion as m1/m2/m3, there were a total of 170 patients with m3 lesions, 8 of whom (4.7%) had lymph node metastases (5, 14, 16, 17, 21–28).
Twelve studies provided data on the frequency of lymphovascular invasion in patients with intramucosal adenocarcinoma, which ranged from 0% to 18% (5, 14, 18, 19, 22, 23, 29–34). Among the 26 patients with lymph node metastases, lymphovascular invasion was described specifically in only 2 (5, 19). Tumor size and grade together were described for only one of the 26 patients with lymph node metastases (5); that patient had a 2.2 cm tumor that was poorly differentiated (5).
Six studies specifically reported data on the presence of a duplicated muscularis mucosae in the resected Barrett’s esophagus (5, 16, 18, 20, 22, 35). For patients with intramucosal carcinoma in these studies, the frequency of finding a duplicated muscularis mucosae ranged from 32% to 65%. There was no apparent association between the presence of a duplicated muscularis mucosae and the presence of metastatic lymphadenopathy (5, 16, 20, 22).
Among 10 patients for whom the number of positive lymph nodes was reported, 9 had only a single lymph node involved (5, 10, 13, 15, 17, 20, 34, 36, 37). For 21 of the 26 patients with lymph node metastases, the report did not specify whether the involved lymph nodes were loco-regional or distant; among 5 patients for whom this information was provided, all involved nodes were loco-regional (10, 13, 20, 36).
A number of individual reports included in this systematic review are from the same institution and share some of the same authors. This raises the possibility that some of the same patients were included in multiple reports, and it is not clear how this would affect the outcome of our systematic review. To minimize this possibility, we performed another analysis that included only one report of the largest study from each institution, unless the article provided sufficient information to exclude the possibility of patient overlap. We identified 35 such reports that included a total of 967 patients with HGD or intramucosal cancer. Twelve of these patients had lymph node metastases, for a rate of 1.24% (95% CI .54% – 1.94%). Among the 259 patients with a final pathology diagnosis of HGD, none had positive lymph nodes. Among the 708 patients with a final pathology diagnosis of intramucosal cancer, 12 (1.70%, 95% CI .74%–2.65%) had lymph node metastases. The frequency of lymph node metastases in this analysis that included only one report per institution did not differ significantly from the analysis that included all reports in the systematic review (1.39% vs. 1.24%, p = 0.65 for all patients with HGD and intramucosal carcinoma, 1.93% vs. 1.70%, P=0.62 for patients with intramucosal carcinoma).
Our systematic review identified 70 reports that described surgical pathology findings for 1,874 patients who had esophagectomy performed because of HGD or intramucosal carcinoma in Barrett’s esophagus. Lymph node metastases were found in 26 (1.39%) of those patients. No lymph node metastases were found in the 524 patients who had a final pathology diagnosis of HGD, whereas 26 (1.93%) of the 1,350 patients with a final pathology diagnosis of intramucosal carcinoma had positive lymph nodes. These results did not change significantly in a secondary analysis performed to minimize the possibility that the same patients might have been included in more than one report.
Our findings might suggest that HGD in Barrett’s esophagus has virtually no risk of lymph node metastases. However, it is important to emphasize that our estimates on the frequency of lymph node metastases are based on a systematic review of final pathology diagnoses that were rendered after examination of esophagectomy specimens. Caution should be exercised in extrapolating these results to predict the frequency of lymph node metastases in patients found to have HGD or intramucosal carcinoma in endoscopic biopsy specimens of Barrett’s metaplasia. Substantial disagreement among gastrointestinal pathologists has been reported in distinguishing HGD from intramucosal carcinoma in such biopsy specimens (38). In equivocal cases, the absence of lymph node metastases in an esophagectomy specimen might persuade a pathologist to make a diagnosis of HGD rather than intramucosal carcinoma. Alternatively, the presence of positive lymph nodes might bias that pathologist to make a diagnosis of intramucosal carcinoma rather than HGD.
Histopathological features of mucosal neoplasms that might affect the risk of lymph node metastases include the depth of tumor invasion, the presence of lymphovascular invasion, and the degree of tumor differentiation. For T1b tumors, which penetrate through the muscularis mucosae, it is well established that the frequency of lymph node metastases increases with the depth of tumor invasion into the submucosa. For example, lymph node metastases have been found in 9%–20% of patients with sm1 tumors (which involve only the upper one-third of the submucosa), whereas 24%–50% of patients with sm3 tumors (which involve the deepest one-third of the submucosa) have positive lymph nodes (20, 22, 31). Our systematic review suggests that depth of invasion is also a risk factor for lymph node metastases in T1a tumors, which do not penetrate the muscularis mucosae. Although we found only 12 studies that reported the depth of mucosal tumor invasion as m1/m2/m3, 8 of the 170 total patients with m3 lesions (4.7%) had lymph node metastases (5, 14, 16, 17, 21–28). In advanced esophageal adenocarcinomas, lymphovascular invasion and poor tumor differentiation are well established risks factors for metastatic disease (18, 21, 23, 29, 39, 40). For patients with mucosal neoplasms in Barrett’s esophagus, however, our systematic review revealed insufficient data to determine whether lymphovascular invasion and tumor differentiation significantly increase the risk of lymph node metastases.
Histological evaluation of the esophagus resected for dysplasia or cancer in Barrett’s esophagus has revealed duplication of the muscularis mucosae in up to 92% of cases (42). This finding is important because duplication of the muscularis mucosae can result in errors in the staging of superficial adenocarcinomas. Typically, the result is overstaging because the pathologist assumes mistakenly that tumor found below the more superficial of the two muscularis mucosae is involving the submucosa. Among the 70 reports included in our systematic review, only 6 provided data on the frequency of finding a duplicated muscularis mucosae in the resected Barrett’s esophagus, which ranged from 32% to 65% (5, 16, 18, 20, 22, 35). The presence of a duplicated muscularis mucosae should not affect the detection of lymph node metastases at esophagectomy, and we found no association between the presence of a duplicated muscularis mucosae and the presence of such metastases.
Our study suffers from the limitations inherent in a retrospective review of published data. Few of the reports included in our study had focused specifically on the frequency of lymph node metastases in patients with HGD and intramucosal carcinoma. Different surgeons used different techniques for esophagectomy. The number of lymph nodes removed during this operation can range from 0 to more than 50 (41), and the chances for detecting positive lymph nodes increases as the number of nodes resected increases. There was no standardized protocol for pathology review of the esophagectomy specimens, and differences among institutions in the thoroughness of this review are likely to influence the frequency of finding lymph node metastases.
A number of the reports included in our review predate the widespread availability of EMR and endoscopic ultrasonography (EUS), and there was no standardized protocol for their application in studies conducted when these techniques were available. It is not clear how the preoperative use of these techniques would influence the frequency of finding lymph node metastases at esophagectomy. EMR is better than endoscopic biopsy for detecting submucosal invasion of cancer in Barrett’s esophagus (42). Therefore, the use of EMR is likely to improve the accuracy of the preoperative diagnosis of HGD and intramucosal carcinoma. However, our systematic review focused on the final pathology report after an examination of the resected esophagus, and the availability of EMR is unlikely to have influenced those results. Today, the preoperative use of EUS might influence the frequency of finding lymph node metastases at esophagectomy for patients with HGD and intramucosal carcinoma, because only patients found to have malignant lymph nodes by EUS would be likely to have esophagectomy. Without positive lymph nodes demonstrated by EUS, the large majority of patients with HGD and intramucosal cancer now would be treated with endoscopic eradication therapy. Therefore, the preoperative use of EUS might be expected to increase the likelihood of finding lymph node metastases at esophagectomy. Reports on the utility of EUS for the detection of malignant lymph nodes have described a wide range of sensitivities and specificities(19), however, and a false-positive EUS diagnosis of lymph node metastases might result in esophagectomy with the finding of negative lymph nodes. Thus, it is not entirely clear how preoperative EUS would influence the frequency of positive lymph nodes at esophagectomy.
With some cancers, such as renal cell carcinoma, there are rare reports of metastases resolving spontaneously after resection of the primary tumor (43). We are unaware of any such reports for esophageal adenocarcinoma, and spontaneous resolution of metastases from this cancer seems unlikely. Therefore, endoscopic eradication therapy cannot be considered curative for patients with mucosal cancers in Barrett’s esophagus that have metastasized to regional lymph nodes. Such patients might be cured by an operation that removes the malignant nodes, or by the use of chemotherapy or radiation therapy aimed at lymph node metastases. These treatments can be associated with substantial morbidity and mortality, however, and none can guarantee a cure for patients with metastatic esophageal adenocarcinoma.
Our systematic review suggests that the risk of lymph node metastases for patients with mucosal neoplasms in Barrett’s esophagus is in the range of 1% to 2%. Despite this potential for metastatic disease, several tertiary referral centers have reported excellent long-term results for endoscopic eradication therapy of mucosal Barrett’s neoplasms (44–46). Metachronous neoplasms, presumably due to incomplete obliteration of the mucosal tumors, occurred frequently in those series, but those metachronous lesions generally responded readily to repeat endoscopic eradication therapy. However, a careful review of these reports reveals that some patients died from metastatic esophageal adenocarcinoma. In Pouw’s series of 169 patients with mucosal neoplasms in Barrett’s esophagus who were treated endoscopically and followed for a median of 32 months, one patient (0.6%) died of metastatic esophageal cancer (44). Prasad et al. compared the outcomes of endoscopic and surgical treatments for patients with intramucosal carcinoma (45). During a mean follow-up of 64 months, 1 of 132 patients (0.8%) treated endoscopically died of metastatic disease. It should be noted however, that 1 of 46 patients (2.1%) in the surgical group also died of metastatic esophageal cancer even though no positive lymph nodes were found in the resection specimen. In Pech’s series of 349 patients with mucosal neoplasms in Barrett’s esophagus, there were no deaths from metastatic cancer during a mean follow-up of 63.6 months(46). However, there were 58 patient deaths (not tumor-related) during that time, and 13 patients eventually had esophagectomy because endoscopic therapy had failed. With such a high rate of drop out due to non-cancer deaths and surgery, a low rate of metastatic disease easily could have been missed. These low rates of metastatic cancer for highly selected patients in tertiary referral centers support the findings of our systematic review that mucosal neoplasms in Barrett’s esophagus do metastasize, albeit infrequently.
We feel it is important that patients and clinicians appreciate that there is a small but real risk of lymph node metastases associated with mucosal neoplasms in Barrett’s esophagus. Our systematic review suggests that this risk is in the range of 1% to 2%. Thus, patients who opt for endoscopic eradication therapy have at least a 1% to 2% chance that this treatment will not be curative. The alternative treatment is esophagectomy. For older patients who have serious comorbid illnesses, the risks of esophagectomy clearly outweigh the risks of lymph node metastases. For younger and otherwise healthy patients with HGD and intramucosal carcinoma in Barrett’s esophagus, the mortality of esophagectomy probably exceeds 2%, and the procedure is associated with considerable long-term morbidity (3). As discussed above, furthermore, it is not clear that esophagectomy will be curative for patients with lymph node metastases. It has been suggested that physicians should consider a number of factors (e.g. patient age, co-morbidities, extent of metaplasia and neoplasia) when deciding between endoscopic and surgical therapies for mucosal neoplasms in Barrett’s esophagus, and we agree that the choice should be individualized. However, our systematic review suggests that the risk of lymph node metastases alone does not warrant the choice of esophagectomy over endoscopic therapy.
Financial support: supported by the Office of Medical Research, Department of Veterans Affairs and the National Institutes of Health (R01-CA134571)
Potential competing interests: None
Specific author contributions: Kerry Dunbar performed the literature search, reviewed the articles, performed data abstraction, wrote the manuscript and approved the final draft submitted. Stuart Spechler participated in identifying papers for the study, reviewed the articles, the search, edited the manuscript and approved the final draft submitted.