Pattern of recurrence following complete resection of esophageal carcinoma and factors predictive of recurrent disease




Despite increasingly radical surgery for esophageal carcinoma, a large number of patients still experience recurrent disease soon after operation. The current study was undertaken to evaluate the pattern of recurrence after curative esophagectomy for cancer of the thoracic esophagus and to identify factors predictive of recurrent disease.


A total of 439 consecutive patients discharged from the authors' institution following R0 resection between January 1982 and July 2002 were followed for evidence of recurrence over a mean interval of 37.3 (range, 1–207) months.


Overall 1-, 3- and 5-years survival rates were 91%, 54%, and 41%, respectively. Some 230 patients (52.4%) developed proven recurrence, of whom 24 were alive and 206 were dead at the time of writing. The median time to recurrence was 12.0 (range, 6–96) months, with a median survival thereafter of 7.0 (range, 0–83) months. The pattern of recurrence was local in 12.1%, regional in 20.5% (cervical 3.6%, mediastinal 14.8%, and abdominal 2.1%), and distant in 19.8%, respectively. The overall pattern of dissemination was significantly different according to the histologic subtype (P = 0.021). One hundred five (45.7%) of all recurrences occurred within 12 months of surgery, with local, regional, and distant recurrence occurring at a median of 14.0 (range, 6–77), 13.5 (range, 6–73), and 11.0 (range, 6–96) months, respectively; A factor predictive of recurrent disease was histologic tumor depth invasion (P = 0.001).


Depth of tumor invasion should be used to identify patients who will have recurrence within 12 months of operation, so that these patients may be either entered into trials of multimodality treatment or offered nonsurgical palliation. Cancer 2003;97:1616–23. © 2003 American Cancer Society.

DOI 10.1002/cncr.11228

Carcinoma of the esophagus is often far advanced at the time of detection, and only a small number of patients are considered for curative resection and long-term survival. Resection rates range from 19% to 64% and 5-year survival from 10% to 55% depending on stage.1 Surgical resection remains the primary treatment modality for esophageal carcinoma, as it provides sustained palliation of dysphagia and the best chance of cure.2 Over the past 10 years advances in surgical and anaesthesic techniques, together with improvements in perioperative management, have reduced the risks of esophageal resection to an acceptable level. This reduction in the postoperative mortality rate has, however, failed to translate itself into any significant benefit in long-term survival, as large numbers of patients continue to present with recurrent disease following apparently curative surgery. Therefore attempts have been made to improve the outcome of surgical resection by increasing the radicality of nodal dissection or by combining surgery with other treatment modalities. Before embarking into more extensive surgical procedures or designing adjuvant therapeutic regimens, it is important to understand the causes and characteristics of recurrence after esophagectomy. The suppression of postoperative recurrences will improve the prognosis of patients with esophageal carcinoma undergoing curative resection. Thus, it is of clinical importance to predict postoperative recurrences.

The current article presents the results of R0 resection for esophageal carcinoma with the aim of determining the pattern and the timing of local, regional, and distant recurrences and of assessing the factors that predict the recurrent disease.


Patient Population

The records of 1800 patients who underwent surgical evaluation for esophageal malignancy in our institution between January 1982 and July 2002 were reviewed. Of these, 460 consecutive patients had subtotal esophagectomy with two-field lymphadenectomy and R0 resection. Overall 30 day and in-hospital mortality rates were 2.4% (n = 11) and 4.5% (n = 21), respectively. Patients who had rare tumors were excluded.

In the 439 patients who had undergone R0 tumor resection and who were discharged from the hospital, the median followup was 25.0 [standard deviation (SD), 35.7; range, 1–207)] months. The male to female ratio was 7.8:1 and the median age 57.6 (SD, 9.4; range 32–77) years. Squamous cell carcinoma (SCC) was the predominant histologic subtype compared with adenocarcinoma, with a ratio of 4.7:1. The majority of SCC were located at the middle third of the thoracic esophagus, whereas the majority of adenocarcinomas were located within the lower third (Table 1).

Table 1. Distribution of Tumors According to Histologic Subtype and Site
 Upper third (%)Middle third (%)Lower third (%)Total (%)
  1. SCC: squamous cell carcinoma.

SCC61 (13.8)232 (52.8) 69 (15.7)362 (82.5)
Adenocarcinoma 1 (0.3) 13 (3.0) 63 (14.4) 77 (17.5)
Total62 (14.1)245 (55.8)132 (30.1)439 (100)

Surgical Approach

The detailed resection techniques have been described elsewhere.3 Surgical resection consisted, in a transthoracic esophagectomy for tumor of the middle third or lower third of the esophagus, completed with a cervical incision for anastomosis in case of tumor of the upper third of the thoracic esophagus. The surgical approach included an abdominal lymphadenectomy and an extended en bloc mediastinal lymphadenectomy (two-field lymphadenectomy). No cervical lymphadenectomy was undertaken. Abdominal lymphadenectomy comprised en bloc removal of all lymphatic tissue in the lower posterior mediastinum, in the left and right paracardial regions, along the lesser curve, and along the left gastric artery.

Lymphadenectomy of the paratracheal, carinal, and left and right bronchial nodes was performed followed by en bloc resection of the thoracic duct together with the para-aortic nodes. The nodes in the aortoplumonary window were removed, but a full dissection of the left recurrent laryngeal nerve chain was not routinely carried out. Neoadjuvant radiochemotherapy was proposed for locally advanced tumors or within the framework of therapeutic protocols.

Histopathologic Assessment of the Removed Specimen and Lymph Nodes

All nodal material was dissected separately from the specimen at the end of the procedure by the surgeon and analyzed according to the guidelines of the Japanese Society for Esophageal Diseases.4 The histologic staging was based on the pTNM classification,5 defined as follows: esophageal cancer Stage I [pTis-T1N0M0]: tumor restricted to either mucosa [pTis] or submucosa [pT1], no regional lymph node [N0], no distant metastasis [M0]; Stage IIA [pT2-T3N0M0]: tumor restricted to muscularis propria [pT2] or adventia [pT3], no regional lymph node [N0], no distant metastasis [M0]; Stage IIB: [pT1-T2N1M0]: tumor restricted to mucosa [pT1] or to muscularis propria [pT2], regional lymph node metastasis [N1], no distant metastasis [M0]; Stage III: [pT3N1 or pT4NxM0]: tumor restricted to adventia [pT3], regional lymph node metastasis [N1] or tumor invaded the neighboring structures [pT4], with or without regional lymph node metastasis [Nx], no distant metastasis [M0].

Recurrence Identification

All patients surviving operation were followed until death or the time of writing at the end of the first month, at six-month intervals in years one and two, and annually thereafter. Clinical review consisted of history and abdominal examination. Abdominal ultrasonography was realized twice a year, chest X-ray, endoscopy, and indirect laryngoscopy once a year. If recurrence was suspected, patients underwent barium-swallow, ultrasonography, chest X-rays, thoracoabdominal computed tomography (CT), and endoscopic examination with biopsies. More selective investigations such as cervical ultrasonography, bone scintigraphy, and cerebral CT were carried out based on specific symptomatology, clinical examination, and biochemical profile.

Follow-up was complete for all patients. By definition, the timing of recurrence was always above six months after surgery. Before six months, evidence of tumor was considered as persistent neoplastic disease. The mode of recurrence was classified into three patterns: local recurrence was defined as anastomotic recurrence; regional recurrence was defined as that occurring either in the mediastinum or upper abdomen at the site of previous esophageal resection and nodal clearance or in the cervical area where no lymphadenectomy had been performed; and distant recurrence was defined as hematogenous if it developed within a solid organ or within the peritoneal cavity. When a patient had recurrence in multiple distant organs, the initial detected organ was chosen as the recurred organ. Histologic, cytologic, or unequivocal radiologic proof was required before a diagnosis of recurrence was made. Recurrence supported by clinical impression alone was not included.

Statistical Analysis

The survival status of patients was ascertained in July 2002. Followup was complete for all 439 patients. Data are shown as prevalence, mean (SD), or median (SD). Comparison of continuous data between groups was done by the student t test and of ordinal data by chi-square test or Fischer exact test when appropriate. In analyzing survival time, we used the Statistical Package for Social Sciences (SPSS, Chicago, IL). The survival function was estimated by the actuarial method after excluding the postoperative deaths. The log rank test was used for comparison of survival curves. To determine predictors of distant recurrences, we used the stepwise logistic regression model, adjusting all the covariates simultaneously. The 0.1 level was defined for entry into the model. Differences were considered to be significant at 5% alpha risk.


Pattern of Recurrence

At the time of writing, 264 patients had died during the follow-up period, 206 as a result of proven recurrent esophageal carcinoma and 58 from causes other than recurrence. Of the remaining 209 patients who were alive at the time of writing, 24 were known to have recurrent disease. Thus, 52.4% (230 of 439) of all patients discharged following potentially curative esophageal resection developed recurrent esophageal carcinoma. Local recurrence was recognized in 53 patients (12.1%), regional recurrence in 90 (20.5%; cervical in 16, mediastinal in 65, and abdominal in 9), and distant metastases in 87 (19.8%; liver in 35, lung in 21, bone in 18, cerebral in 7, kidney in 2, adrenal gland in 2, peritoneal in 1, and skin in 1), respectively.

The patterns of recurrence according to the clinical and histopathologic factors are shown in Table 2, with a similar proportion of local, regional, and distant dissemination, except for histologic subtype and mean number of positive lymph node metastases. Distant dissemination rather than local or regional recurrences occurred more frequently with lower-third tumors (predominantly adenocarcinoma, 27.6%; 18.9% and 11.1%, respectively) whereas the opposite was true for upper or middle-third lesions (predominantly SCC, 72.4%, 81.1% and 88.9%, respectively; P = 0.02). For pT1 tumors, no recurrence occurred in patients with tumor restricted to mucosa (n = 12) whereas recurrence was observed in 25 (31.8%) patients with tumor restricted to submucosa. The mean number of positive lymph node metastases was significantly higher in the distant recurrence group rather than in the regional group (P = 0.04). Cervical recurrence occurred in 16 patients (3.6%); all of them had SCC, six of the upper third, six of the middle third, and four of the lower third of the esophagus. Seven patients with cervical recurrence had node-positive tumors at the time of the operation. The overall incidence of recurrent disease in areas outside the operative field was 23.5%. In patients with carcinoma of the upper esophagus, recurrence was observed frequently in the neck. Conversely, in the cases with carcinoma of the lower esophagus, intraabdominal lymph node recurrence was more frequently observed than in those patients with carcinoma of other parts of the esophagus. The number of involved lymph nodes significantly associated with recurrence was found to be one or more.

Table 2. Clinical and Histopathologic Factors and Patterns of Recurrence
  Pattern of recurrence
Local (n = 53)Regional (n = 90)Distant (n = 87)Total (n = 230)
  1. SD: standard deviation; SCC: squamous cell carcinoma.

Mean (SD) age at presentation in years58.1 (10.5)56.7 (10.2)56.4 (7.8)57.6 (9.4)
Gender ratio(M:F)6.6:110.3:116.4:110.5:1
Location of the main tumorUpper third6171134
 Middle third365347136
 Lower third11202960
Weight loss<10%407468182
Neoadjuvant chemoradiotherapyYes284740115
Anastomotic siteCervical17292369
Anastomotic leakYes58720
Respiratory complicationYes13191547
Histologic subtypeSCC438063186
 Moderately or poorly15212460
Depth of invasionpT110171643
Lymph node metastasispN020403191
Mean (SD) number of lymph nodes with positive metastasis2.2 (3.3)1.7 (2.2)2.5 (3.4)2.1 (3.0)
pTNM stageI7131030

With regard to the distant recurrence group, there was a positive correlation between the pulmonary recurrence and either the tumor localization at the upper third of the esophagus (P = 0.05), the subgroup of cervical anastomosis (P = 0.01), or the SCC histologic subtype (P = 0.02).

Prognostic factors associated with the development of recurrent disease are shown in Table 3. There was no significant difference in the overall pattern of recurrence in relation to the level of the primary tumor. The only clinical factor associated with recurrent disease was dysphagia at the initial presentation (P = 0.002). With regard to histopathologic factors, increased depth of tumor invasion, and the presence and the mean number of positive nodal metastases were found to correlate with an increased incidence of recurrent disease, same as the pTNM stage (P < 0.001), although there was no association with the histologic subtype or histologic differentiation.

Table 3. Prognostic Factors Associated with the Development of Recurrent Disease
  Recurrence (n = 230)No recurrence (n = 209)P
  1. SD: standard deviation; SCC: squamous cell carcinoma.

Mean (SD) age at presentation56.9 (9.4)58.3 (9.3)0.118
Gender ratio(M:F)
Location of the main tumorUpper third3428 
 Middle third1361090.161
 Lower third6072 
Weight loss< 10%1821660.988
 ≥ 10%4843 
Neoadjuvant chemoradiotherapyYes1151140.341
Anastomotic siteCervical69560.457
Anastomotic leakYes20140.434
Respiratory complicationYes47520.266
Histologic subtypeSCC1861760.358
 Moderately or poorly6056 
Depth of invasionpT143104 
 pT24142< 0.001
Lymph node metastasispN091134< 0.001
Mean (SD) number of lymph node with positive metastasis2.1 (3.0)0.9 (1.7)< 0.001
pTNM stageI3077 
 IIa5955< 0.001

Time of Recurrence

Mean (SD) time to recurrence after operation was 17.8 (15.3; range, 6–96) months, with 45.7% (105 of 230) of all recurrences developing within 12 months of operation (43.4%, 40.0%, and 52.9% for local, regional, and distant recurrences, respectively). The median delay before recurrence was 12.0 months, and the median time between recurrence and death was 7.0 months. The median (SD) time to local recurrence was 14.0 (13.5; range, 6–77) months, the median (SD) time to regional recurrence was 13.5 (14.2; range, 6–73) months, and the median (SD) time for distant recurrence was 11.0 (17.5; range, 6–96) months. Median time to recurrence was shorter in the distant recurrence group compared with local or regional recurrence groups (P = 0.04). Ninety percent of recurrences occurred before the 38th month. In the recurrence group, the mean (SD) recurrence times of each pTNM stage were 12.9 (17.0) months in Stage I, 11.8 (13.3) months in Stage IIA, 7.0 (5.6) months in Stage IIB, and 8.1 (8.8) months in Stage III. Mean recurrence times were significantly lower for Stage III compared with Stages I or IIA (P = 0.04 and P = 0.03, respectively) and for Stage IIB compared with Stage IIA (P = 0.04). Even with Stage I esophageal carcinoma, recurrence within 12 months was recognized in 17 of 30 cases, and it was also observed in 37 of 59 cases in Stage IIA. Distant recurrence within 12 months was observed in 4 and 10 patients with Stage I and IIA, respectively.

Treatment of Recurrent Disease

Thirty-five patients (24.5%) with local or regional recurrence had no specific treatment other than pain control. Patients with dysphagia due to anastomotic, conduit, or hiatal recurrence were palliated with dilatation and endoprothesis insertion with or without radiotherapy, depending on their general condition. The majority of patients with distant organ metastasis received no definitive therapy other than symptomatic relief. Adjuvant therapy for recurrent disease was performed in 28 of 115 patients (24.3%) who had received neoadjuvant radiochemotherapy. Twenty-six patients (46.4%) with liver or pulmonary metastases received palliative chemotherapy with 5-fluorouracil and cisplatin, six (33.3%) with bone metastases underwent palliative radiotherapy or orthopaedic fixation, and two (28.6%) with cerebral localization underwent surgery with external radiotherapy. No survival advantage was observed in patients who were selected for interventional therapy compared with those who had symptomatic treatment alone. Recurrent laryngeal nerve palsies resulting from cervical recurrence were treated with Teflon injection and later with thyroplasty. Patients with isolated asymptomatic recurrent cervical node disease had ipsilateral radical cervical lymph node clearance and postoperative radiotherapy.

Survival and Predictive Factors of Recurrence

The median follow-up interval was 25.0 (35.7; range, 1–207) months. The median (SD) survival of patients was 38.6 (3.2) months and median (SD) time free of disease was 37.0 (2.5) months. The overall and disease free 1-, 3-, and 5-year survival rates were 91%, 54%, 41%, and 91%, 53%, 37%, respectively. Stratification of recurrence free survival according to the pTNM staging system confirmed a significantly worse survival for more advanced tumors (Fig. 1).

Figure 1.

Overall survival following discharge for 439 patients undergoing R0 resection for esophageal carcinoma according to the TNM stage. The number of subjects at risk at each interval is shown in the table at the bottom of the graph.

Median (SD) survival following the detection of recurrent esophageal carcinoma was 7.0 (0.5) months and was significantly longer for regional dissemination [7.0 (0.8) months] than for local [5.0 (1.0) months] or distant [5.0 (0.9) months] recurrence (P = 0.01). Median survival (SD) in patients with cervical recurrence was 26 (7.3) months and was significantly longer than in those with recurrence at other sites (P = 0.049).

There was no difference in median of disease-free interval between SCC and adenocarcinoma (38.0 and 36.0 months, respectively, P = 0.19), or between upper-third, middle-third, and lower-third lesions (38.0, 37.0, and 37.0 months, respectively, P = 0.90).

The only predictive factor of recurrent disease, obtained by multivariate analysis, was the depth of the tumor invasion on the histologic assessment (pT) (X2 = 10.7, risk ratio = 1.9, 95% confidence interval [CI] 1.3–2.7, P = 0.001; Table 4). We did not find any significant factor predictive of recurrence for the either subgroups, local, regional, or distant recurrence, neither for recurrence occurring within 12 months after curative surgery.

Table 4. Multivariate Analysis: Predictive Factors of Recurrent Disease
VariablesPX2Relative risk95% confidence interval
Depth of tumor invasion0.00110.71.91.3–2.7
Lymph node metastasis0.–11.6
pTNM stage0.600.31.20.4–1.8


At a median follow-up of 37.3 months, recurrent esophageal carcinoma had developed in almost half of the patients who underwent an apparently curative resection, with an overall 5-year survival rate of only 41%. These dismal results occurred in an already highly selected patient group who had been staged having localized and resectable tumors. The treatment of recurrent disease is disappointing, and there is no evidence that currently available treatment can prolong survival.6

The results of the current study suggest that patients at high risk of developing recurrence need to be identified in the preoperative staging process, since more than 50% of all recurrences occurred within 12 months after surgery, either in local, regional, or distant sites. Tumor depth invasion was the only factor predictive of recurrent disease (P = 0.001).

With the aim of decreased incidence of recurrent disease in esophageal carcinoma, many approaches were proposed. Early diagnosis is essential to improve the curability and the resectability of this cancer. While it is evident from the stage-specific survival figure presented here that early tumors (pT1) are associated with a low incidence of recurrence and a significant survival advantage, they are found in only a few patients undergoing surgery (33.5%). Even if the primary tumor is limited to the submucosal layer, it may proceed to lymph node6 or distant metastases (31.8% in the current study). Moreover, pretherapeutic exploration is hampered by the difficulty in determining exactly the tumoral stage. Like other authors,7 pT classification seems to appear to us as an important prognostic factor of recurrent disease, with increased prevalence of lymph node metastases with increasing tumor depth.8 Instead, strategies aimed at decreasing the recurrence rate and improving survival must continue to focus on the majority of patients who present advanced esophageal carcinoma. Radical lymphadenectomies, such as en bloc esophagectomy8–12 and three-field dissection,7, 13 have been developed hoping that meticulous nodal dissection would improve pathologic tumor staging, enhance local control, and provide some survival benefit in terms of delay to recurrence and even long-term cure.

Dresner et al.14 have reported a 21% local recurrence rate following radical resection for middle-third and lower-third tumors, which is lower than that described in other series where no formal lymphadenectomy was performed.15 There are some conflicting reports suggesting that lymphadenectomy confers no survival benefit.16 Furthermore, despite the strong association between nodal disease and tumor recurrence, in the current study, some node-positive patients were long-term survivors, with 5-year survival rates of 39% and 19% for Stages IIB and III tumors, respectively.

In an effort to improve the results of surgery still further, Japanese surgeons have advocated the addition of cervical lymphadenectomy to conventional thoracoabdominal lymph node dissection. In a recent prospective study,17 an overall rate of 36% for cervical nodal disease was documented by three-field nodal dissection, with no statistically significant differences in overall survival. In the current series, isolated cervical recurrence occurred in only 16 patients (3.6%), suggesting a much lower incidence of positive cervical nodes than described previously.18, 19 Any potential benefit from cervical lymphadenectomy must be balanced against the additional morbidity and mortality, particularly related to laryngeal nerve damage.17

Alternative strategies to decrease recurrence and improve survival have therefore turned away from increasingly radical surgery and focused more on multimodality therapy. With regard to distant recurrence, its frequency after curative esophagectomy was similar compared with regional recurrence (37.8% vs 39.1% in the current experience). However, the median time for distant recurrence is significantly shorter than that of regional recurrence (11.0 and 13.5 months, respectively). These findings suggest that microscopic metastasis often already occurs at the time of operation or that metastatic lesions show an extremely rapid growth after operation.15 Moreover, it suggests that lymph node recurrence did not primarily occur before distant dissemination, as suggested by some authors, who think that lymphatic and hematogenic extension occur independently.20, 21 This could explain the 40.4% of patients without any lymph node metastases who developed recurrence in the current study. Indeed, one recent prospective study in which trephine was performed in patients with operable esophageal carcinoma showed micrometastases in 88% of rib and 15% of iliac bone marrow aspirates.22 Furthermore, distant recurrence within 12 months also occurred even in the cases with early stage carcinoma, with 4 and 10 patients with Stages I and IIA esophageal carcinoma, respectively, in the current series.

Surgery as a single modality therapy can be considered only as a local treatment. Systemic treatment of potentially disseminated disease may be accomplished by the use of chemotherapy, particularly if administered in the neoadjuvant setting. In the current series we found no benefit in neoadjuvant treatment regarding recurrence (P = 0.34). Results of randomized controlled trials23–27 performed to date are conflicting and open to criticism. So far, six reported randomized trials comparing preoperative combined chemotherapy and radiotherapy followed by surgery with surgery alone in resectable esophageal carcinoma23–28 have been reported. When using neoadjuvant chemoradiotherapy, only the Walsh et al. trial23 showed an improvement in survival, and only one study has suggested an improvement of survival29 when using neoadjuvant chemotherapy.

Finally, two randomized trials,30, 31 in which postoperative chemotherapy was used, failed to show an improvement. However, the use of postoperative chemotherapy in patients at high risk of recurrence is still in research stages in Europe.

The high incidence, the shorter median time to recurrence, and the predominant lower-third esophageal tumors in the distant recurrence group suggest genuine biologic differences between adenocarcinoma and SCC. There are two groups of patients whom we must try to identify. They are those who already have metastatic disease at the time of diagnosis but who are not being identified by current staging techniques and those who have micrometastases. The former group are unlikely to be cured by surgery and would die early without benefiting from radical treatment. In such cases, consideration should be given to nonoperative treatment. For the latter group inclusion in trials of systemic, neoadjuvant therapy should be considered. Moreover, other histologic or biologic tools should be used to select patients who will benefit from this strategy.32–34

Despite radical surgery and in some case preoperative chemoradiation, patient death is often related to local or regional recurrence and distant dissemination. Patients at high risk of recurrence need to be identified accurately before operation so that they may be offered entry into trials of multimodality therapy or alternatively considered for nonsurgical palliation. Further oncologic research is needed to predict distant recurrence.