If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address for reprints: Smita Sihag, MD, MPH, Thoracic Service, Department of Surgery, Memorial Sloan Kettering Cancer Center, 1275 York Ave, C-881, New York, NY 10065.
As endoscopic approaches become more widely used to treat early-stage esophageal cancer, reliably identifying patients with less-aggressive tumors is paramount. We sought to identify risk factors for recurrence in patients with completely resected T1 esophageal adenocarcinoma.
Methods
We retrospectively analyzed a single-institutional database for all patients with completely resected pathologic T1 esophageal adenocarcinoma (1996-2016). Risk factors for recurrence were identified using competing-risk regression methods. Risk stratification was performed on the basis of known preoperative clinicopathologic factors; this model's discriminative power for overall survival was evaluated using a Cox proportional hazards model.
Results
Of 243 patients, 32 experienced recurrence. At a median follow-up among survivors of 4 years (range, 0.05-19 years), the 5-year cumulative incidence of recurrence was 15%, and median time to recurrence was 2 years (range, 0.26-6.13 years). On univariable analysis, submucosal invasion, N1 disease, poor differentiation, tumor length, lymphovascular invasion, and multicentricity were significantly associated with recurrence. On multivariable analysis, N1 disease (hazard ratio, 2.93; 95% confidence interval, 1.17-7.34; P = .022) and tumor length (hazard ratio, 1.44; 95% confidence interval, 1.12-1.86; P = .004) were independently associated with recurrence. Risk stratification showed that patients without lymphovascular invasion and a with median tumor length of 0.8 cm (range, 0.10-1.70 cm) had a <10% risk of recurrence and improved survival.
Conclusions
Pathologic T1 tumors have a 5-year cumulative incidence of recurrence of 15%. Nodal involvement and tumor length were independent risk factors for recurrence, whereas tumors <2 cm in length without lymphovascular invasion were associated with a low risk of recurrence.
T1 esophageal lesions shorter than 2 cm, without evidence of lymphovascular invasion, carry a low (<10%) recurrence risk and are associated with improved survival at 5 years.
Risk factors for recurrence of pathologic T1 esophageal tumors, essential for selecting patients for organ-sparing therapy, remain poorly defined. Our retrospective analysis identified tumor length <2 cm and absence of lymphovascular invasion as predictive of low (<10%) risk of recurrence and improved survival at 5 years.
See Commentaries on pages 1280, 1281, and 1282.
Although the majority of esophageal cancers are diagnosed at advanced stages, more regimented endoscopic surveillance for Barrett's dysplasia has contributed to an increase in the detection of early-stage esophageal cancers.
During the past 15 years, the treatment paradigm for early-stage esophageal adenocarcinoma has shifted away from radical esophagectomy toward organ-sparing techniques, such as endoscopic mucosal resection (EMR) and endoscopic submucosal dissection, in carefully selected patients.
This change is reflected in the current recommendations for clinical T1a N0 disease, which include endoscopic resection followed by eradication therapy for residual Barrett's metaplasia or dysplasia.
These recommendations are based on the findings of several retrospective institutional series demonstrating that overall survival in these patients is comparable between esophagectomy and EMR as long as a negative margin is achieved.
Esophagectomy versus endoscopic resection for patients with early-stage esophageal adenocarcinoma: a National Cancer Database propensity-matched study.
The efficacy of organ-sparing approaches for T1a N0 disease relates to the low risk of lymph node metastasis in this setting, typically reported to be 7% or less,
although the lack of pathologic lymph node evaluation represents a major limitation of endoscopic resection. In clinical T1b disease, the risk of nodal metastasis increases to 20% or higher,
and therefore esophagectomy is still favored. Prior studies have shown that certain key histopathologic factors, such as depth of invasion, tumor grade, tumor length, and lymphovascular invasion (LVI), are strongly correlated with the risk of lymph node metastasis in T1 tumors.
Although nodal metastasis has been consistently found to be independently associated with recurrence and poor prognosis in early-stage esophageal adenocarcinoma, accurately predicting nodal metastases on the basis of these factors preoperatively remains a significant challenge.
Identifying subsets of patients with low-risk T1 lesions with less aggressive tumor behavior is paramount in selecting appropriate candidates for esophageal preservation. Thus, we sought to identify risk factors for recurrence in patients with completely resected pathologic T1 esophageal adenocarcinoma. We hypothesized that certain prognostically significant clinicopathologic characteristics that are knowable preoperatively with greater certainty may differentiate high- and low-risk groups to further delineate and validate current eligibility criteria for endoscopic therapy versus esophagectomy.
Methods
Following approval from our institutional review board (informed consent waived), we retrospectively reviewed a prospectively maintained database to identify all patients with pathologic T1 esophageal adenocarcinoma who were treated with esophagectomy only at Memorial Sloan Kettering Cancer Center between 1996 and 2016. Patients were excluded from our analysis if resection yielded positive margins, they received neoadjuvant or adjuvant therapy, and/or they were followed up elsewhere postsurgery and surveillance data were not available (Figure 1). Data collected included demographic information, submucosal invasion, nodal status, grade, tumor length, presence of Barrett's metaplasia or dysplasia, Siewert type, presence of LVI, multicentricity, proximal and distal margins, number of lymph nodes resected, procedure type, and incidence of anastomotic leak. Tumor length included only neoplasia and excluded Barrett's metaplasia or dysplasia. All specimens were reviewed by an expert gastrointestinal pathologist at our center, and pathologic staging was based on the seventh edition of the American Joint Committee on Cancer staging manual.
Figure 1Consolidated standards of reporting trials diagram.
Our surveillance protocol included computed tomography imaging every 3 to 6 months for 2 years and annually thereafter. All recurrences were confirmed by biopsy. Locoregional recurrences were defined as those to the esophagus, stomach, perigastric, celiac, mediastinal, or supraclavicular lymph nodes, whereas distant recurrences were defined as those to all other organs.
Demographic characteristics are presented as descriptive statistics: median and range for continuous variables and prevalence with percentage for categorical variables. Time to recurrence was calculated from the date of surgery to the date of recurrence, death, or last follow-up. Cumulative incidence of recurrence (CIR) and site-specific recurrences were estimated using cumulative incidence functions such that death without recurrence and, for site-specific recurrence, recurrence at a different site were considered competing risk events.
Risk factors for recurrence were evaluated using Fine-Gray competing risk regression methods.
Statistically significant variables from univariable analyses were then incorporated into a multivariable model, using a forward selection approach for competing risk regression based on the Akaike information criterion to allow for a maximum of three variables to be retained.
We then further refined this multivariable model to include only relevant clinicopathologic variables that may be known preoperatively, to stratify patients by risk of recurrence. Risk stratification into high- and low-risk groups was performed on the basis of evaluation of the distribution of predicted 5-year CIR probabilities using the variables retained in this final model.
Overall survival after surgery was estimated using Kaplan-Meier methods, and its association with recurrence risk groups was evaluated using Cox proportional hazards regression. Statistical tests were 2-sided. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Inc, Cary, NC) and R version 3.2.4 (R Foundation for Statistical Computing, Vienna, Austria).
Results
After exclusion criteria were applied, 243 patients composed our cohort for analysis. Median age was 65 years (range, 22-89 years), and 83.5% were men. Overall, 158 patients (65%) had submucosal invasion, 32 patients (13%) had pathologic N1 disease, 66 patients (27%) had evidence of poor tumor differentiation, and 56 patients (23%) had LVI; all of these features were more prevalent among the 32 patients who experienced recurrence during the study period. Median tumor length was 1.7 cm (range, 0.1-9 cm). Median proximal and distal margin lengths, extent of lymphadenectomy, surgical approach, and rates of anastomotic complications were similar between patients whose tumors did and did not recur (Table 1). The rate of lymph node metastasis was 3.5% in T1a patients and 18.4% in T1b patients.
Overall survival was 68.6% (95% confidence interval [CI], 62.1%-75.7%) at 5 years. Median follow-up among survivors was 4.0 years (range, 0.05-19 years), and median interval to recurrence was 2.0 years (range, 0.26-6.13 years). CIR plateaued at 15% (95% CI, 10%-20%) at 5 years (Figure 2, A). Most recurrences were distant; the cumulative incidence of distant recurrence was 11% (95% CI, 7%-15%) at 5 years (Figure 2, B), and that of locoregional recurrence was only 4% (95% CI, 2%-8%). The most common sites of distant recurrence were liver and lung; the most common sites of locoregional recurrence were mediastinal or paratracheal lymph nodes (Figure 2, C).
Figure 2Cumulative incidence and patterns of recurrence. Cumulative incidence of (A) overall recurrence and (B) locoregional (LR) versus distant recurrence (DR) since surgery. C, Site-specific patterns of recurrence. CI, Confidence interval.
We performed a univariable competing risk regression analysis to identify risk factors for recurrence in our study population. Of the variables considered (see Methods and Table 2), submucosal invasion (hazard ratio [HR], 3.1; 95% CI, 1.22-7.92; P = .018), pathologic N1 disease (HR, 5.23; 95% CI, 2.60-10.52; P < .001), poor tumor differentiation (HR, 2.37; 95% CI, 1.19-4.72; P = .014), tumor length (HR, 1.59; 95% CI, 1.31-1.93; P < .001), LVI (HR, 3.72; 95% CI, 1.64-6.52; P = .001), and multicentricity (HR, 3.08; 95% CI, 1.18-8.02; P = .022) were significantly associated with recurrence (Table 2). In a multivariable analysis, pathologic N1 disease (HR, 2.93; 95% CI, 1.17-7.34; P = .022) and tumor length (HR, 1.44; 95% CI, 1.12-1.86; P = .004) were significant independent risk factors for recurrence in this cohort, whereas LVI was not, despite being retained in this model. Multicentricity was excluded from this multivariable analysis owing to incomplete pathologic data.
Table 2Risk factors for recurrence identified by univariable and multivariable analysis
To risk-stratify patients, we constructed a final multivariable model using only clinicopathologic variables that may be known preoperatively to guide decision making: submucosal invasion, LVI, and tumor length. In this model, only LVI and tumor length were significant and retained. We then defined low- and high-risk groups on the basis of this final model at a threshold of 10% predicted probability of recurrence. The low-risk group had 6 recurrences, whereas the high-risk group had 25 recurrences. CIR at 5 years was significantly higher in the high-risk group (23%; 95% CI, 15%-32%) than in the low-risk group (7%; 95% CI, 2%-14%) (Figure 3, A). Furthermore, patients in the high-risk group had a significantly increased risk of death, compared with patients in the low-risk group (HR, 1.67; 95% CI, 1.08-2.58; P = .022). Five-year overall survival in the high-risk group was 58% (95% CI, 49%-69%), compared with 75% (95% CI, 65%-86%) in the low-risk group (Figure 3, B). Of 94 patients in the low-risk group, none had LVI or tumors longer than 1.7 cm (Table 3). Of 125 patients in the high-risk group, all patients with tumors <2 cm in length also had presence of LVI. Furthermore, patients with high-risk tumors without LVI had a tumor length >3 cm.
Figure 3Discriminatory power of high- and low-risk groups based on relevant preoperative clinicopathologic factors. A, Cumulative incidence of recurrence by risk group. B, Overall survival (OS) by risk group. Low-risk patients with <10% cumulative incidence of recurrence are those without lymphovascular invasion with tumor length <2 cm. CI, Confidence interval.
Table 3Characteristics of risk-stratified patients based on a multivariable model of relevant preoperative clinicopathologic factors associated with recurrence
Using multivariable analysis followed by a risk-stratification model, we found that patients at the lowest risk of recurrence following complete resection of pT1 esophageal adenocarcinoma are those without evidence of LVI and with tumors shorter than 2 cm (Figure 4). This patient population had a <10% risk of recurrence, with 5-year survival of 75%. Thus, our study suggests that these patients may represent good candidates for endoscopic resection. High-risk patients are those with LVI or tumors >3 cm—esophagectomy is most appropriate in this setting, given the higher risk of recurrence (≥10%) and significantly lower survival, compared with low-risk patients.
Figure 4Defining low-risk lesions in early-stage esophageal carcinoma. CIR, Cumulative incidence of recurrence; CI, confidence interval; LVI, lymphovascular invasion.
Our identification of N status as an independent predictor of recurrence confirms the results of multiple prior studies, which found that the presence of lymph node metastasis is the strongest independent predictor of poor overall and disease-free survival in localized esophageal adenocarcinoma.
Unfortunately, predicting nodal metastasis in early-stage tumors before resection remains a significant clinical challenge due to the inaccuracy of available diagnostic tools. Luu and colleagues
showed that concordance between preoperative clinical staging by endoscopic ultrasound and final surgical pathologic staging is only 53% in early-stage disease. Moreover, van Westreenen and colleagues
observed that the pooled sensitivity of positron emission tomography for detection of locoregional lymph node metastases across several studies is relatively low, at 0.51. Thus, in early-stage disease, current staging modalities may be only marginally better than the flip of a coin, and nodal metastasis prediction nomograms have high false-positive rates and may thus lead to unnecessary esophagectomies.
Because clinical nodal status cannot be knowable with a high degree of certainty preoperatively, we did not include this variable in our final multivariable model for risk stratification.
Our findings that tumor length and LVI correlate with risk of recurrence and overall survival in pT1 esophageal adenocarcinoma are in line with prior studies showing their association with nodal metastasis as well as overall and disease-free survival.
identified tumor length >3 cm as the sole independent predictor of lymph node metastasis and poor survival, similarly in pT1 disease. Our study further supports this cutoff, as high-risk tumors that did not have evidence of LVI were >3 cm in length.
By allowing determination of these features preoperatively, EMR may complement clinical staging of superficial tumors, even if endoscopic therapy ultimately proves inadequate. Supporting this conclusion, Mohiuddin and colleagues
Outcomes of surgical resection of T1bN0 esophageal cancer and assessment of endoscopic mucosal resection for identifying low-risk cancers appropriate for endoscopic therapy.
were able to determine both the presence of LVI and the degree of differentiation in all 51 specimens evaluated, although depth of invasion was not assessable in most cases. Similarly, EMR appears to allow reliable assessment of T stage because concordance with final surgical pathologic assessment was 72% in a series of 58 patients from our institution, which was superior to that for endoscopic ultrasound (35%).
However, in our multivariable analysis, submucosal invasion was not independently associated with recurrence in the presence of N1 disease or LVI.
Our findings may help inform long-term surveillance practices following esophagectomy or endoscopic therapy for pT1 disease, which are not well-defined and frequently vary by institution. Neither axial imaging nor endoscopy is routinely recommended following esophagectomy, but it should be considered in the setting of symptoms.
For endoscopically resected tumors, surveillance strategies usually consist of serial endoscopy every 3 months initially after disease eradication and then every 6 to 12 months for an additional 2 to 3 years.
CIR data herein indicate that, for pT1 tumors, surveillance with axial imaging at least annually may be appropriate in this patient population, but it does not need to extend past 5 to 6 years, because the curve plateaus at that point. As local recurrence rates for T1a lesions following EMR are higher (up to 12%) than those following esophagectomy and the rate of distant recurrence is <10%,
both routine endoscopy and axial imaging, although at a lower frequency, seem appropriate in this setting.
Although patients with pT1 disease typically have a favorable prognosis following resection, with prolonged disease-specific survival and lower rates of postoperative morbidity and mortality, we observed a slightly lower than expected overall survival (68.6% at 5 years) as well as a higher than expected rate of postoperative anastomotic complications (17%) in this cohort. This may reflect the fact that our cohort excluded patients who received adjuvant or neoadjuvant chemotherapy and/or radiotherapy because these modalities have an undefined role in the setting of endoscopic resection and could influence our risk-stratification model. Consequently, the only patients with pathologic nodal disease (13% overall) were those who were ineligible for systemic therapy due to reasons such as poor performance status, excessive comorbidities, or postoperative complications. Such patients would be expected to have worse than average outcomes. Others have suggested that a higher rate of postoperative anastomotic or infectious complications may increase the risk of recurrence, but we did not see any indication of this in our univariable analysis.
This study has several limitations due to its retrospective nature. Our primary outcome of interest, disease recurrence, was identified on retrospective chart review. In addition, this outcome was relatively infrequent; only 32 patients experienced recurrence over the follow-up period, although some recurrences may have been missed, especially if patients were lost to follow-up. As a result, our multivariable model was restricted to 3 variables, and in contrast to prior studies, submucosal invasion was not found to be independently associated with recurrence. Lastly, instances of incomplete pathologic data, with respect to variables such as multicentricity and measured depth of submucosal invasion in microns, were not possible to re-review with our pathology team, given the lack of available tissue in our biorepository because many patients were resected >10 years ago.
Conclusions
Our results show that despite an overall good prognosis, early-stage tumors treated with esophagectomy have a 5-year CIR of 15%. Although nodal disease and tumor length were independently associated with recurrence, nodal disease cannot be definitively known for preoperative decision making. Therefore, on the basis of a multivariable model of only clinicopathologic factors that may be known preoperatively, patients at low risk of recurrence (<10%) are those with tumors <2 cm in length without LVI. These findings, using a unique data set from a high-volume center with long-term follow up, further validate the findings of other retrospective studies in the literature and may begin to assist clinicians in selecting appropriate patients for esophageal preservation.
Conflict of Interest Statement
Dr Jones serves as a senior medical advisor for Diffusion Pharmaceuticals and a consultant for Merck and AstraZeneca. All other authors reported no conflicts of interest.
The Journal policy requires editors and reviewers to disclose conflicts of interest and to decline handling or reviewing manuscripts for which they may have a conflict of interest. The editors and reviewers of this article have no conflicts of interest.
References
Codipilly D.C.
Chandar A.K.
Singh S.
Wani S.
Shaheen N.J.
Inadomi J.M.
et al.
The effect of endoscopic surveillance in patients with Barrett's esophagus: a systematic review and meta-analysis.
Esophagectomy versus endoscopic resection for patients with early-stage esophageal adenocarcinoma: a National Cancer Database propensity-matched study.
Outcomes of surgical resection of T1bN0 esophageal cancer and assessment of endoscopic mucosal resection for identifying low-risk cancers appropriate for endoscopic therapy.
This research was supported, in part, by National Institutes of Health Cancer Center Support Grant No. P30 CA008748. Dr Sihag is supported by a grant from the Fiona and Stanley Druckenmiller Center for Lung Cancer Research at Memorial Sloan Kettering Cancer Center and the David J. Sugarbaker Research Scholarship from the American Association for Thoracic Surgery.
In this issue of the Journal, Shihag and colleagues1 identify risk factors for recurrence among patients with completely resected T1 esophageal adenocarcinoma. The main impetus for this analysis is the paradigm shift of treatment for early-stage esophageal adenocarcinoma away from radical esophagectomy and toward organ-sparing techniques, such as endoscopic mucosal resection (EMR) and endoscopic submucosal dissection (ESD). The promise of treating esophageal cancer with endoscopy certainly has the promise of alchemy and has been supported by several retrospective institutional series demonstrating comparable overall survival between EMR and esophagectomy as long as a negative margin is achieved.
Organ preservation in the treatment of early esophageal cancer is an accepted paradigm. These techniques are appropriate for carefully selected patients who have early-stage and less aggressive tumors or as a mitigating measure.1 Unlike some other malignancies for which the aggressiveness of the cancer, such as lung cancer, can span the spectrum of indolent to aggressive, esophageal adenocarcinoma mostly has aggressive and very aggressive biology. As such, only the earliest of cancers have a low enough risk of metastatic spread at the time of diagnosis that they can be treated with local therapies.
Over 20 years ago, the first reports of endoscopic mucosal resection (EMR) for mucosal-based esophageal cancers appeared in the gastrointestinal literature1 and shortly after in the thoracic surgery literature.2 This initiated a paradigm shift in how early esophageal adenocarcinomas, particularly T1a tumors, were being managed. Endoscopic eradication therapy, mucosal resection to remove the cancerous lesion, and radiofrequency ablation3 to eradicate the precursor lesion—Barrett's esophagus—has emerged as an important treatment strategy that allows for organ preservation in patients with Barrett's-related early cancers.
Your session will expire shortly. If you are still working, click the ‘Keep Me Logged In’ button below. If you do not respond within the next minute, you will be automatically logged out.
33097-X&id=fx1.jpg){kind=link}
33097-X&id=fx2.jpg){kind=link}
33097-X&id=gr1.jpg){kind=link}
33097-X&id=gr2.jpg){kind=link}
33097-X&id=gr3.jpg){kind=link}
33097-X&id=gr4.jpg){kind=link}