1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. Supporting information
  9. References
  10. Supporting Information


Oesophageal malignancy is a disease with a poor prognosis. Oesophagectomy is the mainstay of curative treatment but associated with substantial morbidity and mortality. Although mortality rates have improved, the incidence of perioperative morbidity remains high. This study assessed the impact of postoperative morbidity on long-term outcomes.


A prospective database was designed for patients undergoing oesophagectomy for malignancy from 1998 to 2011. An observational cohort study was performed with these data, assessing intraoperative technical complications, postoperative morbidity and effects on overall survival.


Some 618 patients were included, with a median follow-up of 51 months for survivors. The overall complication rate was 64·6 per cent (399 of 618), with technical complications in 124 patients (20·1 per cent) and medical complications in 339 (54·9 per cent). Technical complications were associated with longer duration of surgery (308 min versus 293 min in those with no technical complications; P = 0·017), greater operative blood loss (448 versus 389 ml respectively; P = 0·035) and longer length of stay (22 versus 13 days; P < 0·001). Medical complications were associated with greater intraoperative blood loss (418 ml versus 380 ml in those with no medical complications; P = 0·013) and greater length of stay (16 versus 12 days respectively; P < 0·001). Median overall and disease-free survival were 41 and 43 months. After controlling for age, tumour stage, resection margin, length of tumour, adjuvant therapy, procedure type and co-morbidities, there was no effect of postoperative complications on disease-specific survival.


Technical and medical complications following oesophagectomy were associated with greater intraoperative blood loss and a longer duration of inpatient stay, but did not predict disease-specific survival. Copyright © 2012 British Journal of Surgery Society Ltd. Published by John Wiley & Sons, Ltd.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. Supporting information
  9. References
  10. Supporting Information

Oesophageal resection for malignant disease is associated with high rates of morbidity and mortality, especially in low-volume centres of care1. Although mortality rates have fallen2, 3, morbidity rates are still high. Contemporary large series report overall morbidity rates of 26–66·7 per cent, with major morbidity in 26–36 per cent of patients undergoing surgery4–9.

Improvements in perioperative mortality have been attributed to better patient selection, modern anaesthesia, intensive care and postoperative support, perioperative nutrition and treatment in high-volume centres1, 3, 10. Life-threatening complications may be better managed. Surgical techniques have also altered over time, with a trend towards more extensive nodal dissection and the introduction of minimally invasive oesophagectomy4, 6, 7, 11, 12. There is little evidence that any of these aspects of management has lowered complication rates. It is important to understand the impact of postoperative complications on long-term patient outcomes, to see whether measures that might reduce complication rates are likely to impact on survival.

Morbidity arising from oesophagectomy is generally considered to be technical or medical13–16. Technical complications relate to intraoperative events that damage structures or result in inadequate reconstruction13–16. Medical complications are typically pulmonary, cardiac or infective, and arise, in part, from the systemic insult of the surgery itself13, 14, 16.

There is conflicting evidence regarding the effects of perioperative complications on long-term survival after oesophagectomy13–18. There is weak evidence for other cancer operations, in particular hepatic and colorectal surgery, that links postoperative complications with poorer survival19–21. Although there is no direct evidence, it has been suggested that the inflammatory and immune modulatory effects of surgery and surgical morbidity have a negative effect on the host response to malignant disease18–21.

This study was undertaken to define the incidence of complications in a large series of patients undergoing oesophagectomy for cancer, and to assess the effect of complications on long-term survival.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. Supporting information
  9. References
  10. Supporting Information

A prospective database (Microsoft® Access version 1.0 1992; Microsoft, Washington, DC, USA) has been maintained of all patients with oesophageal carcinoma under the care of surgeons in the Department of Upper Gastrointestinal and Soft Tissue Tumour Surgery of the Princess Alexandra Hospital (PAH). Permission to collect and use the information was approved by the hospital ethics committee (PAH HREC/03/QPAH/165). Patient details were recorded at presentation, during all treatments, and at follow-up visits until patient death, or to October 2011 for those patients still alive. Clinicians and a research nurse, whose primary responsibility was data collection, completed all records.

Patients who had a resection for squamous cell carcinoma or adenocarcinoma of the oesophagus or oesophagogastric junction between January 1998 and June 2011 were included.

Preoperative morbidity was defined by the following criteria. Cardiac disease included the requirement for pharmaceutical or mechanical intervention (percutaneous stent insertion, coronary artery bypass grafting or pacemaker) for cardiac ischaemia or arrhythmia. Respiratory disease was defined as the use of any long-term medication to treat a specific pulmonary disorder. Renal disease was defined as a baseline creatinine level above 90 µmol/l. Diabetes was defined as the use of oral hypoglycaemic medication or insulin. The severity of co-morbidities was not recorded. Anaesthetic fitness was graded according to the American Society of Anesthesiologists' classification22, 23.

Preoperative staging included endoscopy and computed tomography (CT) of the chest, abdomen and pelvis, and, from 2004, fluorodeoxyglucose–positron emission tomography. Endoscopic ultrasonography was used selectively. Treatment was directed by local protocols. Patients with early tumours or those medically unsuitable for neoadjuvant therapy proceeded directly to surgery. Patients receiving preoperative chemoradiation had a radiotherapy dose varying from 35 to 45 Gy according to trial or institutional protocols24. Preoperative chemotherapy consisted of the combination of cisplatin and infusional 5-fluorouracil (5-FU) (2 cycles), or preoperative and postoperative epirubicin, cisplatin and 5-FU25.

Pathological stage was determined according to the International Union Against Cancer (UICC) staging system (7th edition)26. Completeness of resection was classified as R0 for negative margins, R1 for microscopically positive margins (involved or within 1 mm) and R2 for macroscopically positive margins.

Postoperative morbidity was grouped as technical or medical to maintain consistency with previous reports, but not stratified by severity13, 14, 16.

Technical complications were classified as damage to surrounding structures (chyle leak, recurrent laryngeal nerve palsy, postoperative haemorrhage, tracheo-oesophageal fistula) or inadequate reconstruction (anastomotic leak, conduit necrosis).

Chyle leak was defined as excessive drain output (more than 500 ml in 24 h), persistent for more than 2 days and confirmed by triglyceride analysis. Recurrent laryngeal nerve palsy was confirmed by laryngoscopy in patients with altered phonation or aspiration pneumonia. Postoperative haemorrhage was defined as the requirement for blood transfusion within 24 h of surgery, or reoperation for bleeding. Anastomotic leak was defined as any evidence of leak, as shown clinically by the loss of gastrointestinal contents into a drain or as seen on contrast swallow routinely performed on postoperative day 6 or 7. Patients with an anastomotic leak were analysed as a single group. Conduit necrosis was diagnosed at endoscopy or surgery. Fistula from airway to conduit or oesophagus was confirmed by endoscopy and bronchoscopy.

Postoperative medical complications were grouped into respiratory, cardiac, wound infection, urinary, gastroparesis, delirium and thromboembolic complications. Pulmonary complications included respiratory failure, pneumonia and acute respiratory distress syndrome. Respiratory failure was defined as prolongation of intubation or as reintubation to allow facilitation of oxygenation. Pneumonia was defined as a febrile illness with consistent clinical findings and consolidation on radiological imaging. Acute respiratory distress syndrome was defined as a non-infective respiratory decompensation with consistent radiological findings.

Cardiac complications included arrhythmias, defined as abnormalities or changes from the preoperative electrocardiogram, and myocardial ischaemia, defined as increased cardiac enzyme levels with a characteristic clinical picture. Wound infection was defined as persistent erythema around a surgical wound or as purulent discharge from a wound requiring antibiotic therapy and/or drainage. Urinary complications included acute urinary retention requiring recatheterization and urosepsis confirmed by urine culture. Postoperative gastroparesis was defined as poor emptying of the gastric conduit for more than 5 days and after the return of normal bowel function requiring nasogastric tube decompression. Delirium was defined as disorientation and altered mental state that persisted for more than 24 h. Thromboembolic complications included deep vein thrombosis or pulmonary embolism confirmed by venous ultrasonography, ventilation/perfusion scan or high-resolution CT pulmonary angiography.

Data allowing stratification of the severity of complications were not recorded. A separate analysis for patients who had a reoperation for technical complications was performed.

Procedure-related mortality was defined as any death within 90 days of surgery, or as death before discharge from hospital.

Various combinations of open and minimally invasive surgery were used and have been reported elsewhere4, 11. These included: open transthoracic oesophagogastrectomy; thoracoscopically assisted three-field oesophagectomy with the abdominal procedure performed via a laparotomy; thoracoscopic and laparoscopic oesophagectomy; and an open three-field resection (McKeown). All patients in the latter three groups had a cervical anastomosis. A jejunostomy was constructed in all patients, and feeding commenced on the first postoperative day.

Standard dissection included removal of the nodal stations from the subcarinal, posterior mediastinal, paraoesophageal, diaphragmatic, right and left paracardial, left gastric artery, coeliac axis, suprapancreatic region and common hepatic regions. These correspond to the following lymph node stations in accordance with the Japanese Research Society for Gastric Cancer27: 1, 2, 3, 4Sb, 7, 11p, 11d, 12a, 110, 111 and 112. The superior mediastinal, recurrent laryngeal and cervical nodes were not removed, nor was the thoracic duct.

After discharge, all patients visited outpatients every 3 months for 2 years, every 6 months for 4 years, and once a year thereafter. Survival was calculated from the date of operation, with the analysis including postoperative deaths.

Statistical analysis

The analysis of results was performed with the entire group as well as separate analyses for the surgery-alone and neoadjuvant groups. To identify the clinicopathological factors associated with outcome, univariable analyses were undertaken for disease-specific and overall survival, separately for the surgery-alone and neoadjuvant therapy groups. Patients with stage IV disease were excluded from these analyses.

Continuous variables were expressed as median (range) and compared using the Mann–Whitney U test. Comparison of categorical variables was done by contingency table analysis using χ2 tests. Comparison of means was based on Student's t test. Survival data were summarized using the Kaplan–Meier method. Unadjusted comparisons of survival times were based on the log rank test. Factors found significant after univariable analysis were included in multivariable analyses. Multivariable analyses of survival times were based on Cox regression. Predictors of complications were assessed by logistic regression. For all calculations, P < 0·050 was considered significant. Data analysis was performed using SPSS® version 20 (IBM, Armonk, New York, USA).


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. Supporting information
  9. References
  10. Supporting Information

A total of 618 patients were included in this series. Median follow-up for survivors was 51 (range 2–151) months. Perioperative morbidity and survival data were available for all patients. Patient characteristics are summarized in Tables 1 and 2. Patients in the neoadjuvant therapy group had more tumours larger than 6 cm and a higher UICC stage. Patients in the surgery-alone group were older than those in the neoadjuvant therapy group (66 (27–84) versus 62 (16–79) years respectively; P < 0·001) and had a higher incidence of medical co-morbidities.

Table 1. Patient characteristics
 Entire group (n = 618)Surgery alone (n = 316)Adjuvant therapy (n = 302)P*
  • Values in parentheses are percentages. ASA, American Society of Anesthesiologists; UICC, International Union Against Cancer.

  • *

    χ2 test.

Sex ratio (M:F)523:95264:52259:430·502
ASA grade   0·668
 I21 (3·4)12 (3·8)9 (3·0) 
 II430 (69·6)213 (67·4)217 (71·9) 
 III163 (26·4)88 (27·8)75 (24·8) 
 IV3 (0·5)2 (0·6)1 (0·3) 
 Missing1 (0·2)1 (0·3)0 (0) 
 Cardiac150 (24·3)99 (31·3)51 (16·9)< 0·001
 Respiratory132 (21·4)94 (29·7)38 (12·6)< 0·001
 Diabetes54 (8·7)31 (9·8)23 (7·6)0·393
 Renal53 (8·6)39 (12·3)14 (4·6)0·001
Type of oesophagectomy   0·108
 Thoracoscopically assisted436 (70·6)212 (67·1)224 (74·2) 
 Transthoracic137 (22·2)74 (23·4)63 (20·9) 
 Minimally invasive28 (4·5)18 (5·7)10 (3·3) 
 Other17 (2·8)12 (3·8)5 (1·7) 
Histology   0·208
 Adenocarcinoma507 (82·0)265 (83·9)242 (80·1) 
 Squamous cell carcinoma110 (17·8)50 (15·8)60 (19·9) 
 Missing1 (0·2)1 (0·3)0 (0) 
Tumour length (cm)   0·001
 ≤ 6493 (79·8)260 (82·3)233 (77·2) 
 > 687 (14·1)29 (9·2)58 (19·2) 
 Missing38 (6·1)27 (8·5)11 (3·6) 
Tumour location   0·315
 Proximal third1 (0·2)1 (0·3)0 (0) 
 Middle third54 (8·7)22 (7·0)32 (10·6) 
 Distal third433 (70·1)224 (70·9)209 (69·2) 
 Oesophagogastric junction130 (21·0)69 (21·8)61 (20·2) 
UICC stage   < 0·001
 077 (12·5)41 (13·0)36 (11·9) 
 I110 (17·8)83 (26·3)27 (8·9) 
 II188 (30·4)73 (23·1)115 (38·1) 
 III236 (38·2)114 (36·1)122 (40·4) 
 IV7 (1·1)5 (1·6)2 (0·7) 
R0 resection547 (88·5)272 (86·1)275 (91·1)0·059
In-hospital mortality7 (1·1)5 (1·6)2 (0·7)0·451
90-day mortality6 (1·0)4 (1·3)2 (0·7)0·686
Table 2. Patient characteristics according to type of oesophagectomy
 Thoracoscopically assisted (n = 436)Transthoracic (n = 137)Minimally invasive (n = 28)P*
  • Values in parentheses are percentages. ASA, American Society of Anesthesiologists; UICC, International Union Against Cancer.

  • *

    χ2 test.

Sex ratio (M:F)364:72122:1524:40·282
ASA grade   0·010
 I15 (3·4)1 (0·7)4 (14) 
 II307 (70·4)95 (69·3)18 (64) 
 III113 (25·9)38 (27·7)6 (21) 
 IV1 (0·2)2 (1·5)0 (0) 
 Missing0 (0)1 (0·7)0 (0) 
 Cardiac100 (22·9)32 (23·4)11 (39)0·142
 Respiratory82 (18·8)31 (22·6)10 (36)0·077
 Diabetes40 (9·2)9 (6·6)2 (7)0·613
 Renal36 (8·3)11 (8·0)4 (14)0·528
Histology   0·001
 Adenocarcinoma344 (78·9)127 (92·7)24 (86) 
 Squamous cell carcinoma91 (20·9)10 (7·3)4 (14) 
 Missing1 (0·2)0 (0)0 (0) 
Tumour length < 6 cm362 (83·0)93 (67·9)25 (89)0·006
Tumour location   < 0·001
 Proximal third1 (0·2)0 (0)0 (0) 
 Middle third50 (11·5)1 (0·7)2 (7) 
 Distal third336 (77·1)67 (48·9)22 (79) 
 Oesophagogastric junction49 (11·2)69 (50·4)4 (14) 
UICC stage   < 0·001
 060 (13·8)9 (6·6)8 (29) 
 I90 (20·6)13 (9·5)3 (11) 
 II137 (31·4)41 (29·9)6 (21) 
 III143 (32·8)73 (53·3)11 (39) 
 IV6 (1·4)1 (0·7)0 (0) 
R0395 (90·6)116 (84·7)23 (82)0·081
In-hospital mortality6 (1·4)0 (0)1 (4)0·203
90-day mortality5 (1·1)0 (0)1 (4)0·187

There was no difference in median operative blood loss between the surgery-alone group and the neoadjuvant therapy group (350 (0–2000) versus 350 (0–2300) ml respectively; P = 0·605), or in median operating time (290 (150–503) versus 285 (165–540) min; P = 0·495).

The overall median length of stay (LOS) was 14 (7–123) days. There was no difference between the surgery-alone and neoadjuvant therapy groups (median 14 (7–123) versus 13 (7–97) days respectively; P = 0·784).

The incidence of postoperative complications is shown in Table 3. Overall, complications occurred in 64·6 per cent of patients. The only significant difference between the two treatment groups was in the incidence of urinary complications, which were more frequent in patients treated by surgery alone (P = 0·037).

Table 3. Postoperative morbidity
 Entire group (n = 618)Surgery alone (n = 316)Neoadjuvant therapy (n = 302)P*
  • Values in parentheses are percentages.

  • *

    χ2 test.

Total complications399 (64·6)203 (64·2)196 (64·9)0·867
Technical complications124 (20·1)60 (19·0)64 (21·2)0·547
 Anastomotic leak69 (11·2)36 (11·4)33 (10·9)0·899
 Chyle leak39 (6·3)16 (5·1)23 (7·6)0·247
 Haemorrhage13 (2·1)6 (1·9)7 (2·3)0·784
 Necrotic conduit9 (1·5)4 (1·3)5 (1·7)0·747
 Vocal cord palsy8 (1·3)3 (0·9)5 (1·7)0·496
 Tracheo-oesophageal fistula5 (0·8)2 (0·6)3 (1·0)0·680
Medical complications339 (54·9)178 (56·3)161 (53·3)0·468
 Respiratory230 (37·2)123 (38·9)107 (35·4)0·405
 Cardiac91 (14·7)44 (13·9)47 (15·6)0·572
 Infection60 (9·7)32 (10·1)28 (9·3)0·786
 Gastroparesis8 (1·3)6 (1·9)2 (0·7)0·287
 Urinary12 (1·9)10 (3·2)2 (0·7)0·037
 Delirium31 (5·0)18 (5·7)13 (4·3)0·465
 Thromboembolic11 (1·8)4 (1·3)7 (2·3)0·374
Other complications30 (4·9)18 (5·7)12 (4·0)0·353
Multiple complications154 (24·9)87 (27·5)67 (22·2)0·137
Reoperation for technical complication40 (6·5)20 (6·3)20 (6·6)1·000

Technical complications were associated with longer median operating time (308 (range 165–510) versus 293 (150–540) min; P = 0·017) and greater median intraoperative blood loss (448 (50–2300) versus 389 (50–2000) ml; P = 0·035) than in patients with no technical complications. Technical complications were not associated with patient age, ASA grade, UICC stage, tumour location, neoadjuvant therapy, procedure type or pre-existing co-morbidities (Table S1, supporting information).

Medical complications were not associated with longer median operating time (294 (175–540) versus 299 (150–503) min; P = 0·113), but were associated with greater intraoperative blood loss (417 (0–2000) versus 380 (0–2300) ml; P = 0·013) and longer length of stay (16 (8–123) versus 12 (7–68) days; P < 0·001).

Of the 123 patients with UICC stage 0–III disease who had a technical complication, 67 (54·5 per cent) also experienced a medical compliation. This was similar to the rate of medical complications in patients without technical complications (269 of 488, 55·1 per cent; P = 0·919). Specifically relating to pulmonary complications, there were 51 (41·4 per cent) of 123 in the technical complication group, compared with 177 (36·3 per cent) of 488 in the group without technical complications (P = 0·298). No other subtype of medical complication was associated with technical complications (Table S2, supporting information).

Patients with no complications had a median LOS of 12 (range 7–25) days, compared with 16 (8–123) days for those with complications (P < 0·001). In patients without a technical complication, median LOS was 13 (7–123) days versus 22 (8–97) days when a technical complication occurred (P < 0·001).

The in-hospital mortality rate was 1·1 per cent and the 90-day mortality rate was 1·0 per cent. In-hospital death occurred in six (1·5 per cent) of 399 patients after any complication and in one (0·5 per cent) of 219 patients with no complication (P = 0·431). In-hospital mortality occurred in two (1·6 per cent) of 124 patients with technical complications and in five (1·0 per cent) of 494 patients without (P = 0·632).

Median (range) overall and disease-specific survival times were 41 (2–151) and 43 (1–151) months respectively.

In the surgery-alone group, multivariable analysis using the significant univariable factors found that age, cardiac complications, R1/2 resection and UICC stage were independent prognostic factors for overall survival (Table 4). In terms of technical complications, postoperative tracheo-oesophageal fistula was a predictor of poor overall but not disease-free survival. No other postoperative surgical complications were significant prognostic factors. For disease-free survival, UICC stage and R1/2 resection were the only significant prognostic factors (Table 4).

Table 4. Univariable and multivariable analysis of overall and disease-free survival in the surgery-alone group (excluding patients with stage 4 cancer)
  Overall survivalDisease-free survival
 nLength (months)Univariable P*Multivariable PHazard ratioLength (months)Univariable P*Multivariable PHazard ratio
  • Values in parentheses are 95 per cent confidence intervals. UICC, International Union Against Cancer; NA, data not available.

  • *

    Log rank test, except

  • Mann–Whitney U test;

  • Cox regression analysis.

Age (years)  0·0010·0061·0 (1·0, 1·0) 0·120  
Sex  0·601   0·506  
 M26048   57   
 F5139   47   
Histology  0·644   0·808  
 Adenocarcinoma26048   53   
 Squamous cell carcinoma5031   47   
Tumour length (cm)  0·0020·520  0·0020·445 
 ≤ 625752   58   
 > 62724   24   
UICC stage  < 0·001< 0·001  < 0·001< 0·001 
 041NA  1·0NA  1·0
 I83NA 0·0503·4 (1·0, 11·5)NA 0·1053·5 (0·8, 15·7)
 II7334 < 0·00111·8 (3·6, 38·3)39 < 0·00116·3 (3·9, 67·8)
 III11417 < 0·00120·4 (6·4, 65·6)18 < 0·00131·4 (7·6, 129·3)
Resection margin  < 0·001   < 0·001  
 R027157  1·079  1·0
 R1–24015 0·0022·0 (1·3, 3·0)15 0·0071·8 (1·2, 2·8)
 Any complication201360·0110·535 470·0480·443 
 No complication11074   NA   
 Technical complications60240·0110·982 270·0120·987 
 No technical complication25153   65   
  Anastomotic leak36220·0160·516 220·0120·251 
  No anastomotic leak27552   65   
  Chyle leak16470·530  470·467  
  No chyle leak29544   53   
  Haemorrhage6140·307  270·585  
  No haemorrhage30547   53   
  Necrotic conduit4150·0230·413 240·068  
  No necrotic conduit30748   53   
  Vocal cord palsy3NANA  NA0·647  
  No vocal cord palsy30847   53   
  Tracheo-oesophageal fistula2100·0460·0265·2 (1·2, 22·2)100·0280·051 
  No tracheo-oesophageal fistula30947  1·053   
 Medical complications177360·198  480·453  
 No medical complication13457   65   
  Respiratory122470·640  530·990  
  No respiratory complication18947   52   
  Cardiac44300·0410·0401·6 (1·0, 2·5)350·2230·077 
  No cardiac complication26752  1·058   
  Infection32240·186  240·090  
  No infection27948   57   
  Gastroparesis6210·624  210·954  
  No gastroparesis30547   53   
  Urinary10NANA  NA0·404  
  No urinary complication30147   53   
  Delirium18360·138  360·431  
  No delirium29348   53   
  Thromboembolic4110·879  110·716  
  No thromboembolic complication30747   53   
 Other complication17530·476  NA0·315  
 No other complication29444   52   
 Multiple complications87300·0130·420 350·059  
 None or single complication22457   67   
 Reoperation for technical complication20240·217  260·283  
 No technical complication25153   65   

For the neoadjuvant group, multivariable analysis revealed UICC stage to be the only independent prognostic factor for both overall and disease-free survival (Table 5). Technical complications, individual complications and complication groups were not associated with disease-free survival.

Table 5. Univariable and multivariable analysis of overall and disease-free survival in the neoadjuvant therapy group (excluding patients with stage 4 cancer)
  Overall survivalDisease-free survival
 nLength (months)Univariable P*Multivariable PHazard ratioLength (months)Univariable P*Multivariable PHazard ratio
  • Values in parentheses are 95 per cent confidence intervals. UICC, International Union Against Cancer; NA, data not available.

  • *

    Log rank test, except

  • Mann–Whitney U test;

  • Cox regression analysis.

Age (years)  0·292   0·844  
Sex  0·646   0·294  
 M25740   34   
 F4347   61   
Histology  0·814   0·172  
 Adenocarcinoma24037   34   
 Squamous cell carcinoma6055   55   
Tumour length (cm)  0·0490·133  0·091  
 ≤ 623147   41   
 > 65827   27   
UICC stage  < 0·001   < 0·001  
 036NA < 0·0011·0NA < 0·0011·0
 I27NA < 0·0015·3 (2·5, 11·5)NA < 0·00117·5 (4·3, 71·4)
 II11540 < 0·0016·5 (2·6, 16·1)49 < 0·0015·0 (2·2, 11·6)
 III12227 0·0031·7 (1·2, 2·3)24 0·0031·7 (1·2, 2·4)
Resection margin  0·0360·362  0·0210·226 
 R027544   37   
 R1/22517   17   
 Any complication194410·631  360·546  
 No complication10638   35   
 Technical complications63400·938  300·447  
 No technical complication23741   36   
  Anastomotic leak32410·818  320·674  
  No anastomotic leak26841   36   
  Chyle leak23270·664  250·336  
  No chyle leak27741   36   
  Haemorrhage7400·481  NA0·472  
  No haemorrhage29341   35   
  Necrotic conduit5NANA  NA0·086  
  No necrotic conduit29538   35   
  Vocal cord palsy5250·551  250·628  
  No vocal cord palsy29541   36   
  Tracheo-oesophageal fistula3250·630  250·707  
  No tracheo-oesophageal fistula29741   36   
 Medical complications159440·462  410·101  
 No medical complications14138   33   
  Respiratory106410·910  310·826  
  No respiratory complication19441   36   
  Cardiac46340·425  370·614  
  No cardiac complication25443   35   
  Infection28720·518  720·297  
  No infection27238   35   
  Gastroparesis2NANA  NA0·515  
  No gastroparesis29841   35   
  Urinary2NANA  NA0·174  
  No urinary complication29840   35   
  Delirium13530·475  NA0·236  
  No delirium28741   35   
  Thromboembolic7320·905  NA0·474  
  No thromboembolic complication29341   35   
 Other complication12440·616  440·990  
 No other complication28840   35   
 Multiple complications66440·756  410·261  
 None or single complication23440   35   
 Reoperation for technical complication20440·469  400·763  
 No technical complication23741   36   

After stratification by procedure type (thoracoscopically assisted oesophagectomy, transthoracic oesophagectomy, minimally invasive oesophagectomy), multivariable analysis of postoperative morbidity did not reveal any prognostic factors for overall or disease-free survival (Table S3, supporting information). Technical complications requiring reoperation also failed to be a prognostic factor for overall or disease-free survival for the surgery-alone or neoadjuvant therapy group (Tables 4 and 5).


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. Supporting information
  9. References
  10. Supporting Information

The incidence of postoperative morbidity was similar to that seen in other large series13–18. Technical complications were seen in 20·1 per cent and medical complications in 54·9 per cent of the cohort. There was no difference in the rate of technical complications in patients having neoadjuvant therapy compared with those progressing directly to surgery. Although this finding is subject to selection bias, it is similar to that found in other series13, 14. Patients in the surgery-alone group were significantly older and had a higher incidence of medical co-morbidities, but neither age nor co-morbidities were independent predictors of postoperative or technical complications.

Technical complications were associated with surrogate markers of difficult surgery, with longer operating times and greater intraoperative blood loss. As shown by others13, 14, 16, technical complications were associated with prolonged LOS.

It may be expected that a higher incidence of medical complications would follow technical complications, and, although other studies have supported this14, 16, the present study did not.

Technical complications overall had no effect on overall or disease-free survival rates in patients with oesophageal cancer. Only two patients in the surgery-alone group developed tracheo-oesophageal fistulas, so this specific complication became a predictor of poor overall survival. It is difficult to draw conclusions, except to comment that this problem is a substantial physiological challenge requiring further operative intervention.

The outcomes from this study and three other reports13, 14, 17, including data from more than 1600 patients, suggest that technical complications are not associated with poorer overall survival. In contrast, two large series have described a link between technical complications and worse survival16, 18. One of these was a population-based prospective study in which almost half of the patients had surgery at low-volume (9 or fewer patients per year) institutions18. This introduces variations in selection criteria for surgery, and in the diagnosis and management of postoperative morbidity. These differences can affect perioperative mortality adversely and, although patients who died in the perioperative period were excluded from analysis, there may be a persisting impact on long-term survival18. The patient group in the other series16 was similar to that in the present study. As neither patient selection nor complication severity was stratified objectively, these factors may be important predictors of the ability to tolerate the impact of complications after surgery.

In terms of medical morbidity, cardiac complications were predictive of poorer overall but not disease-free survival in the surgery-alone group. Wherever possible, steps should be taken to minimize the risk and/or severity of perioperative cardiac complications. Otherwise, the data failed to show an effect for any other medical complication on overall or disease-free survival.

A significant shortcoming of this database was the lack of objective stratification of complication severity. The Accordion28 and Clavien29 severity grading systems have become widely used in recent years. It is possible that a relationship exists between more severe morbidity and survival that could be identified through such grading systems. Using reoperation as a surrogate for complication severity in the present series did not, however, show any predictive value for overall or disease-free survival. Despite the size of this study, the incidence of several complications (for example tracheo-oesophageal fistula) was low, so for some specific complications this database may not have sufficient power for analysis.

Despite the lack of an effect on disease-free survival following technical complications, reducing their incidence should remain an important target for improvement1. The impact of complications on other long-term measures, such as quality of life, may be considerable and merits further attention. In addition, agreed clear definitions of the complications specifically related to oesophagectomy will allow better comparison between studies.

Supporting information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. Supporting information
  9. References
  10. Supporting Information

Additional supporting information may be found in the online version of this article:

Table S1 Identification of variables predictive of technical complications (Word document)

Table S2 Incidence of medical complications in patients with and without technical complications (Word document)

Table S3 Multivariable analysis for predictors of survival by procedure type (Word document)

Please note: John Wiley & Sons Ltd is not responsible for the functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. Supporting information
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. Supporting information
  9. References
  10. Supporting Information
bjs_8973_sm_suppinfo.doc387KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.