Surgical Factors Influence the Outcome After Ivor-Lewis Esophagectomy with Intrathoracic Anastomosis for Adenocarcinoma of the Esophagogastric Junction: A Consecutive Series of 240 Patients at an Experienced Center
- First Online:
- Cite this article as:
- Ott, K., Bader, F.G., Lordick, F. et al. Ann Surg Oncol (2009) 16: 1017. doi:10.1245/s10434-009-0336-5
- 275 Views
Despite a considerable number of randomized studies, the surgical approach to locally advanced adenocarcinoma of the esophagogastric junction (AEG) I and II is still discussed controversially. Thus, we evaluated the surgical risk and outcome after an abdominothoracic esophagectomy (Ivor-Lewis) with intrathoracic anastomosis as standard procedure.
Between 1998 and 2006, a total of 240 consecutive patients underwent standardized right thoracoabdominal esophagectomy with two-field lymphadenectomy and intrathoracic anastomosis (Ivor-Lewis operation) for AEG I (n = 206) or AEG II (n = 34). A total of 157 patients (65.4%) had neoadjuvant chemotherapy.
Postoperative morbidity occurred in 17.9% (43 of 240). Overall mortality was 3.8% (9 of 240). The majority of patients (4 of 9) died because of severe pulmonary complications (44.4%) irrespective of surgical complications. Neoadjuvant chemotherapy did not increase morbidity or mortality. The median overall survival was 51 months. Multivariate analysis including age >75 years, clinical response to chemotherapy, complications, R-category and N-category revealed R-category (P = .005; relative risk [RR] 0.32, 95% confidence interval [95% CI] 0.14–0.70) and complications (P < .001, RR 0.16, 95% CI 0.08–0.35) as independent prognostic factors for all patients. Complications was the only independent prognostic factor (P < .001, RR 0.09, 95% CI 0.08–0.35) for the R0 resected patients.
At an experienced center, Ivor-Lewis resection is a safe surgical procedure. Outcome of patients was significantly influenced by surgical factors such as complete resection and complications. Neoadjuvant chemotherapy did not lead to higher morbidity and mortality. The high mortality from non-surgery–related complications emphasizes the importance of careful preoperative evaluation of comorbidities and patient selection.
Despite commendable advances in multimodal treatment regimes for esophageal cancer, surgical resection still remains the cornerstone of therapy for adenocarcinomas of the esophagogastric junction (AEG).1 Surgical strategies in the treatment of esophageal cancer are generally aimed at improving prognosis by achieving the best possible local tumor control by radical resection and adequate lymphadenectomy. At the same time, perioperative morbidity and mortality must be kept as low as possible. Opinions differ on how to conjoin these intents. Two main surgical strategies are advocated: the cervicoabdominal (transhiatal) esophagectomy without formal thoracic lymphadenectomy and the combined (cervico) thoracoabdominal resection with wide excision of the primary tumor and extended lymph node dissection in the posterior mediastinum and upper abdomen (transthoracic esophagectomy with extended en bloc lymphadenectomy) with either intrathoracic or cervical anastomosis.
Transhiatal esophagectomy intends to decrease postoperative morbidity for the price of limited radicality regarding the extent of lymphadenectomy. In contrast, the transthoracic approach enables a wide excision of the tumor and peritumoral tissue as well as extended lymph node dissection in the posterior mediastinum and upper abdomen, thus associated with higher morbidity. A meta-analysis addressing this question revealed a significantly higher early morbidity and mortality for the transthoracic approach and similar 5-year survival rate of 20% for both surgical strategies in esophageal cancer.2 The randomized study by Hulscher et al. showed a higher morbidity, an identical in-hospital mortality, but a trend toward improved long-term survival for patients undergoing a transthoracic resection compared with the transhiatal approach.3 Other randomized studies did not show statistically significant differences in morbidity and short- to median-term survival.4–6
Regarding quality of life of patients undergoing either transhiatal or transthoracic resection, no long-term differences were found. Compared with baseline, quality of life declined after the operation but was restored within a year in both types of resection.7 A recently published long-term follow-up indicated similar 5-year survival (34% versus 36%) for the thoracoabdominal approach compared with the transhiatal approach in the whole study cohort, but revealed significantly improved overall and disease-free survival for thoracoabdominal resection in case of tumors with one to eight positive lymph nodes.8 Thus, patients with a limited number of positive lymph nodes seem to benefit from an extended transthoracic resection.8 Furthermore, a better swallowing function argues for an intrathoracic anastomosis.9
We prospectively evaluated the risk and outcome of an abdominothoracic resection (Ivor-Lewis procedure) with high intrathoracic anastomosis combining a radical approach with an acceptable quality of life as a standard procedure for adenocarcinomas of the esophagus and esophagogastric junction. In addition, the influence of neoadjuvant chemotherapy on complications was analyzed.
Materials and Methods
The prospective study was performed at a single center performing more than 100 esophagectomies per year at that time. Patients had to be older than 18 years and to be fit for extended surgery (Karnofski ≥80) as well as for neoadjuvant chemotherapy in case of locally advanced disease (≥u: ultrasonography T3). Eligible patients with histologically proven AEG I staged u: ultrasonography/c: clinicalT1-4 cNany and cM0 or M1a were included. AEG II were included if endoscopy revealed that the upper extension of the tumor toward the distal esophagus clearly prohibits a transhiatal extended gastrectomy. Patients with locally irresectable disease or previous surgery prohibiting a gastric tube reconstruction were excluded.
Patients staged u/cT3/4 cNany and cM0/M1a were given neoadjuvant therapy. Induction chemotherapy was performed either with platinum/leucovorin/5-fluorouracil (PLF) or with the addition of paclitaxel in young and fit patients as previously described.10,11
Pretreatment Diagnostics and Staging
The pretreatment evaluation included a detailed physical examination, precise history, and functional examination of the patients. The preoperative risk score including general status, renal function, pulmonary function, hepatic function, and cardiac function had to be less than 21 points predicting a low or intermediate risk for esophagectomy for the included patients.12,13 Patients with a high risk score for surgical mortality were excluded.
Staging comprised chest radiography, upper gastrointestinal endoscopy with multiple biopsies to verify the histological diagnosis, endoluminal ultrasound to evaluate the uT-category, computed tomography (CT) scan of the neck, chest, abdomen and the pelvis to rule out distant metastasis. Fluorodeoxyglucose positron emission tomography (FDG-PET) scan was not used routinely for staging, but in patients with locally advanced disease sequential PET investigations were part of response-evaluation study programs.
Response Evaluation for Neoadjuvant Treated Patients
Endoscopy with endoscopic ultrasound and CT scan were used after the first and second cycle of chemotherapy as recently described.10,14,15 Clinical response was defined as an at least 50% reduction in size of the primary tumor as measured by endoscopy and imaging studies. These criteria were used in previously published studies and have been shown to be of prognostic relevance.10,11,14–16
All histopathological analyses of the resecting specimen were performed by an experienced pathologist. Tumor regression was assessed semiquantitatively according to a recently published scoring system.17 For the purpose of this study, all patients with less than 10% residual tumor cells were classified as histopathological responders.
All patients underwent an abdominothoracic en bloc esophagectomy (Ivor-Lewis) via a right transthoracic approach. The procedure starts with a transverse abdominal incision in the upper epigastrium, division of the diaphragm, mobilization of the lower portion of the esophagus, mobilization of the stomach, and formation of a gastric tube (2–3 cm diameter) along the greater curvature. After mobilization of the stomach, consecutive tubular transformation and division of the left gastric artery at its origin, the lymphadenectomy along the celiac axis and suprapancreatic region was performed (according to a D2-lymphadenectomy for gastric carcinoma). A transgastric dilatation of the pylorus by a clamp over about 5 s was accomplished in all patients. The abdomen was closed and the patient was then turned to the left for a right-sided posterolateral thoracotomy. An en bloc esophagectomy including the thoracic duct, the azygos vein in the area of the tumor, ipsilateral pleura, and all periesophageal tissue was carried out including the resection of large portions of both diaphragmatic crura around the esophageal hiatus. The specimen included the lower and middle mediastinal, subcarinal, and right-sided paratracheal lymph nodes (en bloc dissection). The thoracic duct was identified and distally closed under visual control. The small gastric tube was then pulled through into the thorax. The transsection of the esophagus was always performed clearly above the azygos vein. The reconstruction of the intestinal passage was performed with an esophagogastrostomy (end-to-side) in the tip of the right pleural cavity by a circular stapler.9 The stapling device was inserted through the blind end of the gastric tube. After the completion of the anastomosis, the closure of the gastric tube was achieved by a linear cutter. By oversuturing the anastomosis and the blind end of the stomach a functional end-to-end anastomosis was created. The chest was drained with one tube parallel to the gastric tube in the posterior mediastinum with its superior tip at the anastomosis and one tube in the recessus of the diaphragm.
Epidural anesthesia was used intraoperatively and postoperatively to minimize pulmonary complications. All patients were extubated immediately after the end of the operation. Patients were transferred to the intensive and intermediate care unit for 4 to 7 days. Intensive physiotherapy, mobilization and bronchoscopy, if necessary, were initiated from day 1 after surgery. The lower chest tube was removed after 3 days, the tube located at the anastomosis after 8 days. Oral feeding started on day 4 after removing the nasogastric tube on day 3. The anastomosis was not checked by endoscopy or contrast study routinely.
Only those patients who showed clinical signs for leakage of the esophagogastrostomy underwent CT scan of the chest and abdomen as well as endoscopy. When appropriate, covered stents were inserted endoscopically, and an additional CT driven drainage of the mediastinum was placed.18 Covered Choostent™ (length 80 or 110 mm, diameter 18 mm) were used (Choosent™, M.I. Tech, Seoul, Korea). The stents were removed 6 weeks after the application. None of them had to be reinserted. A feeding tube allocated in the duodenum was used only where small leakages occurred. Leakages of the stapler line of the gastric tube were primarily treated by endoscopic injection of fibrin glue or clipping.
The resected specimen was presented to a pathologist intraoperatively. In cases of doubtful clear resection margins, a frozen section was performed. Patients were postoperatively classified according to the American Joint Committee on Cancer Staging’s TNM classification system.19
Patients were followed at 3-month intervals by CT scan of the chest and abdomen, as well as an endoscopy during the first year. Thereafter, follow-up visits were done at 6-month intervals in the second and third year, and afterwards at 12-month intervals. Overall survival was calculated from the day of study inclusion. No patient was lost to follow-up.
All quantitative data are expressed as mean ± standard deviation (SD). Differences in proportions of patients were analyzed by Fisher exact test or χ² test where appropriate. Survival rates were estimated using the Kaplan–Meier method. Statistical comparisons between different groups of patients were performed using the log-rank test and the proportional hazard model. All tests were two-sided and performed at the 5% level of significance using SPSS version 11.50 (SPSS, Chicago, IL).
59 years (range, 33–82 years)
13/227 (5.4% vs 94.6%)
Distant organs (hep/pul/adr/os)
12 (8/2/1/1) 4.9% (3.3%/0.8%/0.4%/0.4%)
Body cavities (per/pleura)
5 (3/2) 2.0% (1.2%/0.8%)
6.2 months (0.1–88.9 months)
Complication (68 events in 43 patients)
Thromboembolic (2 peripheral vene, 1 portal vene, 1 lung embolism)
Alcohol abstinence delirium
Leakage stapler line gastric tube
Necrosis gastric tube
Bleeding (8 thoracic, 2 anastomosis, 4 abdominal)
Abdominal wall dehiscence
Injury of hepatic artery
Leakage of transverse-transversostomy
Incarceration due to incisional hernia
Within these 43 patients with complications, 23 required surgical reintervention (53.5%). In four patients two reoperations were necessary. A total of 14 rethoracotomies in 14 patients were performed. These were due to bleeding in six cases, chylothorax in four cases, leakage of the anastomosis in three cases (two placements of a thoracic tube and one oversuturing of the gastric stapler line), and ischemia of the graft in one case. Eleven relaparotomies in nine patients were necessary. Reasons for relaparotomy were an enterothorax with resection of the transverse colon in two cases, abdominal bleeding in two cases, the resection of the left colon due to a leakage, and the resection of the right colon with ileostomy due to ischemia, abdominal lavage and drainage in three cases (pancreatic fistula and two instances of peritonitis), an incarcerated incisional hernia in one case, and an abdominal wall rupture in one case. Additionally, two patients who already had previous reoperations underwent a thrombectomy and a tracheotomy.
Surgery-related complications comprised leakage of the intrathoracic anastomosis in 20 patients (8.3%), whereas 18 of them were seen at the stapler anastomosis of the esophagogastrostomy and 2 at the stapler line of the gastric tube. All leakages of the esophagogastrostomy could be treated conservatively apart from two patients who needed an additional tube for drainage in the chest requiring a reoperation. Conservative treatment consisted of endoscopic stenting in 15 patients and a nasogastric feeding tube in 3 patients. The two leakages of the gastric tube required reoperation in one case. The other case could be controlled by endoscopic fibrin injection.
Analysis of patients’ mortality
in (%) n = patients
Complications leading to death
Pulmonary without surgical complication (n = 4)
Anastomotic/gastric tube leakage (n = 20)
Chylothorax (n = 4)
Esophagobroncheal fistula (n = 1)
Type of complication management and mortality
Death after conservative treatment
Death after reoperation
Number of complications and mortality
One complication (n = 30)
>1 complication (n = 13)
Influence of Clinicopathological Factors on Morbidity and Mortality
Risk factors for complications and mortality (P value: χ2 test or Fisher exact test)
AEGI versus II
(0/1/2 versus 3/4)
(y)pN category (0 versus +)
Period of operation (I n = 31 1998–2000, II n = 70 2001–2003, III n = 139 2004–2006)
Morbidity, Mortality and Induction Chemotherapy
Correlation of response to neoadjuvant treatment, clinical and histopathological response (P value: χ2 test or Fisher exact test)
Neither clinical nor histopathological nonresponse was associated with higher complication rate or mortality (Table 5).
Prognostic factors for (a) all patients n = 240/51 patients died and (b) completely resected patients n = 196/34 patients died (P value, Kaplan–Meier: log rank, P: univariable and multivariable analysis by Cox regression analysis)
(y)pT0/1/2 vs 3/4
(y)pT0/1/2 vs 3/4
Multivariable analyses including relevant prognostic factors established by univariate analysis revealed complete resection (P = .005, RR 0.32, 95% CI 0.14–0.70) and the incidence of complications (P < .001, RR 0.16, 95% CI 0.076–0.35) as the only two independent prognostic factors (Table 6a).
An R0-resection could be achieved in 196 patients (81.7%). The median survival of the completely resected patients is not yet reached. The 1-, 2-, 3-, 4-, and 5-year survival rates were 87.7%, 75.8%, 68.2%, 59.3% and 44.2%, respectively. In the subgroup of R0 resected patients, the incidence of complications is the only independent prognostic factor (Table 6b).
The benchmark set for surgical treatment of esophageal cancer using an Ivor-Lewis procedure with mortality between 0% and 9.2% and a complication rate between 27.3% and 57% is high in experienced centers.2,3,20–28 The results of our homogeneous prospective study collective including only adenocarcinomas and more than 65% patients after preoperative chemotherapy are well in line with other experienced centers with an overall mortality of 3.8% and a relatively low complication rate of 17.9%. Interestingly, two surgical factors, complete resection and the incidence of complications strongly determine the outcome of the patients in this series. These findings are partially contradictory to those published by Ferri et al.29 who found that technical failure did not influence the outcome in 434 patients with squamous cell cancer. The difference could be explained by the relatively high mortality due to pulmonary complications without technical failure in our series. Another important difference is of course that Ferri included only patients with squamous cell carcinoma in his analysis while we studied patients with AEG type I and II. In our study, despite meticulous pretherapeutic risk evaluation, 4 out of 9 patients died of medical complications without any surgery-related problem. This fact highlights the paramount importance of meticulous pretherapeutic patient selection in esophageal surgery.12 In contrast to squamous cell cancer of the esophagus, neither the application of neoadjuvant chemotherapy nor nonresponse to the treatment had an influence on the incidence and the type of complication or mortality.30
One could argue that the price that has to be paid for a more radical approach is a potentially higher perioperative morbidity and mortality. As in other series, most postoperative deaths arose from pulmonary complications.21,31,32 Even if in our study the overall mortality was relatively low (3.8%), nearly 50% of all deaths were caused by pulmonary complications. Pulmonary complications have been shown to be associated with surgery of the upper abdomen and thoracic surgery.33 Combining both approaches means that pulmonary complications remain the main cause of the reported morbidity after esophagectomy.34 However, preoperative selection, early extubation, aggressive pulmonary rehabilitation including intensive physiotherapy, mobilization on the first day after surgery, bronchoscopy, if necessary, and effective analgesia through thoracic epidural catheters significantly lower the rate of severe pulmonary complications.12,13
Most reports concerning the management of surgical complications have focused on anastomotic leakage because historically the incidence was high and often fatal. There is a wide variation in reported leakage rate after esophagectomy, although much of this variation can be contributed to the different techniques and sites of the anastomosis. The incidence of 8.3% anastomotic leakages is comparable to that reported in the recent literature for a similar intrathoracic stapling technique.35–37 The low rate of gastric tube necrosis may be related to the narrow gastric tube design, respecting the blood supply of the stomach.38 Furthermore, a small gastric tube 2 to 3 cm in diameter guarantees the possibility of endoscopic stenting in case of an anastomotic leakage.18 The key to lower morbidity and mortality after an intrathoracic anastomotic leakage is early and stringent complication management. The new treatment option with long covered stents combined with CT-guided drainage of the mediastinum allowed conservative management of anastomotic leakages with favorable outcome in most cases, avoiding reoperation and therefore further immunosuppression and severe course of sepsis.18
Reoperation for hemorrhage is a result of a technical failure, but fortunately did not lead to death. The rate of bleeding is higher compared with other series and might be related to the radical abdominal and thoracic lymphadenectomy performed at our institution.21,39
The incidence of chylothorax is similar to that reported in other series, where routine ligation of the thoracic duct is used. All four patients with chylothorax had high-volume output and were reoperated by thoracotomy. In one patient the ligation was not successful, and he died of septic complications. Conservative treatment by lymphography and embolization was not performed in our department, but could probably reduce mortality in the treatment of chylothoraces by avoiding reoperation.40,41
The incidence of nonmalignant tracheobroncheal fistulas is lower in our series compared with series reported for squamous cell cancer of the esophagus,12,42,43 This is certainly related to the more favorable localization in the distal part of the esophagus below the tracheal bifurcation and the avoidance of preoperative radiotherapy for AEG in our department. Tracheal lesions in AEG are therefore most likely related to surgical devascularization of the trachea and proximal main stem bronchi caused by lymphadenectomy. Despite the low incidence, the postoperative mortality remains high despite the use of modern double-stenting techniques.12,42,44
Of the 13 patients with more than one complication, 5 patients (38.5%) died, compared with 4 of 30 patients (13.3%) with merely one complication. The early identification and treatment of complications and the avoidance of additional complications seems to be of utmost importance. Even reoperation, if indicated, seems not to harm the patients. Mortality of patients with reoperation was 8.7% compared with 35.0% with conservative treatment in our series. These results might be biased by patient selection before reoperation.
Besides early detection and appropriate treatment of complications, the prevention of complications is of utmost interest. Morbidity and mortality after esophageal surgery was significantly decreased by the application of preoperative risk scores and pretherapeutical patient selection.12,13 Our study included only patients with low or intermediate risk. Pretherapeutic factors predicting complications or death are the presence of diabetes and age.21,23,45 Pulmonary complications are predicted by active smoking and low vital capacity or low 1-second expiratory volume.21 Our results showed no increased risk for complications in the subset of older patients.46 However, in cases of a complication the likelihood of death is significantly increased for older patients compared with younger ones.
Interestingly, tumor-related factors such as pTNM category and resection category are not associated with the development of postoperative complications. Even advanced tumor categories, positive lymph nodes, and the incidence of distant metastases at the time of resection did not influence morbidity.
Because neoadjuvant treatment gains importance especially in the treatment of locally advanced adenocarcinoma,1,47 it is an important question whether the application of preoperative treatment influences morbidity and mortality. In our series the application of neoadjuvant chemotherapy did not influence the complication rate or mortality, similar to three other studies.48–50 Furthermore, we could not find any difference in morbidity and mortality for responding and nonresponding patients. Therefore, a radical surgical approach for locally advanced AEG seems to be appropriate even in nonresponding patients.
Despite the relatively short follow-up, the established prognostic factors for resected esophageal cancer including pT-, pN-, R-category, and clinical response could be reproduced in this series.51–54 However, to date current literature provides no comparable data based only on adenocarcinoma operated with an Ivor-Lewis procedure. Because current literature provides evidence to favor the transthoracic approach for patients with a limited number of positive lymph nodes,8 the Ivor-Lewis procedure seems appropriate for locally advanced tumors, which have a high probability for lymph node metastases, even after neoadjuvant treatment.
At an experienced center the Ivor-Lewis procedure is a safe surgical approach for AEG I and II with relatively low morbidity and mortality. Almost 50% of the mortality in our series arose from pulmonary complications irrespective of surgical complications. Thus, careful preoperative patient selection remains of paramount importance. Surgical factors such as complete resection and the incidence of complications significantly influence the outcome of patients. In contrast to other series, induction chemotherapy does not increase morbidity or mortality even in nonresponding patients. Based on these data and evidence from other series the Ivor-Lewis procedure still remains one of the standard procedures and is an excellent option in the surgical treatment of locally advanced AEG I even following neoadjuvant treatment.