Professional Documents
Culture Documents
Original Article: Key Words
Original Article: Key Words
Original Article: Key Words
DOI: 10.1093/dote/doaa060
Original Article
John V. Reynolds,1 Noel Donlon,1 Jessie A. Elliott,1 Claire Donohoe,1 Narayanasamy Ravi,1 Madhan Kumar
Kuppusamy,2 Donald E. Low2
1
National Esophageal and Gastric Cancer Center, St James’s Hospital and Trinity College, Dublin, Ireland, and
2
Virginia Mason Medical Center, Seattle, WA, USA
SUMMARY. The ECCG developed a standardized platform for reporting operative complications, with consensus
definitions. The Dutch Upper Gastrointestinal Cancer Audit (DUCA) published a national comparison against
these benchmarks. This study compares ECCG data from the Irish National Center (INC) with both published
benchmark studies. All patients undergoing multimodal therapy or surgery with curative intent from 2014 to 2018
inclusive were studied, with data recorded prospectively and entered onto a secure online database (Esodata.org).
219 patients (mean age 67; 77% male) underwent open resection, 66.6% via transthoracic en bloc resection. 30-
day and 90-day mortality were 0.0 and 0.9%,nrespectively. Anastomotic leak rate was 5.4%, pneumonia 18.2%,
respiratory failure 10%, ARDS 2.7%, atrial dysrhythmia 22.8%, recurrent nerve injury 3%, and delirium in 5%
of patients. Compared with both ECCG and DUCA, where MIE constituted 47 and 86% of surgical approaches,
respectively, overall complications were similar, as were severity of complications; however, anastomotic leak rate
was several-fold less, and mortality was significantly lower (P < 0.001). In this consecutive series and comparative
audit with benchmark averages from the ECCG and DUCA publications, a low mortality and anastomotic leak
rate were the key differential findings. Although not risk stratified, the severity of complications from this ‘open’
series is consistent with series containing large numbers of total or hybrid MIE, highlighting a need to adhere to
these strictly defined definitions in further prospective research and randomized studies.
KEY WORDS: esophageal cancers, esophagectomy, complications.
Address correspondence to: John V. Reynolds, Trinity Centre, St. James’s Hospital, Dublin 8, Ireland. Email: reynoljv@tcd.ie
Conflicts of interest: The authors declare no conflict of interest.
Funding: None declared.
© The Author(s) 2020. Published by Oxford University Press on behalf of International Society for Diseases of the Esophagus. All rights
reserved. For permissions, please e-mail: journals.permissions@oup.com 1
2 Diseases of the Esophagus
academic centers across 14 countries. The Esodata lymphadenectomy could be performed or for those
(Esodata.org) collaboration evolved from such an ini- with multifocal in situ cancer. It was also considered
tiative, creating an enormously powerful dataset of for higher risk operative cases, mainly due to respira-
contemporary international practice and outcomes.11 tory morbidity and concern of tolerance of one lung
The ECCG definitions are already established nation- anesthesia, or age, in particular over 80. All cases had
ally within the Dutch Upper Gastrointestinal Can- a pyloroplasty. Patients with AEG type III tumors
cer Audit (DUCA) and are being used in current were excluded unless they underwent a transthoracic
Sex
Female 51 23.3% 607 22.4% 0.777 388 24% 0.818
Male 168 76.7% 2,096 77.5% 1,228 76%
†P-values reported for BMI <25 versus 25–30 versus >30; ASA 1–2 versus 3–5; OG junction, distal esophagus and fundus versus upper
esophagus; R0 versus R1–2
December 2016, and the DUCA series collected from In comparison with ECCG, the age and
January 2016 to December 2017. A significance level co-morbidity profile was similar, with 35% and 37%
of 0.05 was used for all analyses, and all P values being ASA 3 in INC and ECCG, respectively. Obesity
reported are two tailed. Association of categorical was evident in 22.9%, with 65.8% overweight or
variables was assessed using χ 2 test or Fisher’s exact obese, compared with 19.5 and 63%, respectively
test where appropriate. (P = 0.0002). In the DUCA series, the ASA 3 and
obesity rates were significantly (P < 0.001) less, at
19.5 and 16.4%, respectively, and the median age
was significantly higher. In all studies, over 80% of
RESULTS resections were for tumors in the distal half of the
esophagus or the esophagogastric junction. About,
Patient demographics and operative comparisons 27, 21, and 7% underwent upfront surgery in the INC,
In total, 219 patients underwent an esophageal ECCG, and DUCA series, respectively. In the ECCG
resection for cancer during this 5-year period, 76.7% data, 46% underwent neoadjuvant chemoradiation
male, with a median (range) age of 67 (24–86), 53 and 30% perioperative chemotherapy, compared with
(24.2%) were over 70; 77 patients (35.2%) were ASA 3, 88 and 5%, respectively, in DUCA, and 42 and 28%,
19% had COPD, 16% had a cardiac history, and 9.6% respectively, in the Irish cohort.
had diabetes mellitus. All patients had open surgery,
146 (66.6%) via en bloc transthoracic resection, 103
(70.5%) of these having a thoracic anastomosis. Complications
A third of cases had THE. The R0 rate was 94% Comparing the INC, ECCG, and DUCA, 54, 59
(Table 1). and 65%, respectively, experienced a postoperative
4 Diseases of the Esophagus
complication. In a direct comparison with ECCG rate of 0.4%, and conduit necrosis rate of 0.4%. This
data (Table 2), 46.6 and 41% experienced no compli- contrasts with 3 and 0.9%, respectively, for the ECCG
cation (P = 0.138) and 18.3 and 18.5% experienced report and 5.1 and 0.6%, respectively, in the DUCA
three or more complications, respectively (P = 0.431) series (Table 4).
Using the ECCG definitions, Table 3 shows the For chyle leak, in the INC, ECCG, and DUCA
differences between groups. Gastrointestinal compli- series, a leak occurred in 5.5, 4.7, and 8.6%, respec-
cations were significantly (P < 0.001) less in the INC tively, with type III, requiring interventional or sur-
group, in particular, the leak rate from the anasto- gical therapy, 0.4, 1.1, and 1.3%, respectively. The
mosis, staple line or localized conduit necrosis, the incidence of recurrent nerve injury in this series was
incidence was 5.4% (12/219) in the INC, 11.4% in 3.2%, with five patients (2.2%) undergoing a thyro-
the ECCG series, and 19% in DUCA. In this series, plasty. This compared with 4 and 4.3% rate in the
nine were in thoracic anastomoses, an 8% rate, with ECCG and DUCA series, with 0.2 and 0.1% under-
two cases presenting on readmission after 30 days, going a surgical intervention.
and three in cervical anastomoses, a 2.5% leak rate.
For INC, ECCG, and DUCA, respectively, conduit Complication severity as per Clavien-Dindo
failure was seen in 1.3, 1.3, and 1%, and chyle leak in
5.4, 4.7, and 8.6%. Delayed gastric emptying requiring In the INC data, 53.9% patients suffered a complica-
intervention was not observed in the INC compared tion, significantly less (P < 0.001) compared with 59%
with 6.7% in the ECCG (P < 0.0001) and 1.7% in the for ECCG and 67% for DUCA, respectively. Clavien–
DUCA populations (P = 0.999). Dindo ≥III was 30.9%, and ≥ IIIb was 14.9%, com-
The recorded pneumonia rate was similar, at 18.2% pared with 31.1 and 16.9% for ECCG, and 30, and
in the INC, compared with 14.6% in ECCG and 18% for DUCA (P = ns). In the INC series, there
21.1% in DUCA series. The ARDS/respiratory failure was no 30-day mortality, and two patients (0.9%)
rates were 2.7%/10% in INC, significantly (P = 0.041) died between 30 and 90 days, this was significantly
higher than 1.8%/7% for ECCG, and 1.5%/5.2% for (P < 0.05) different than ECCG 30-day mortality of
DUCA (P < 0.05). Pleural effusions were drained in 2.4% and 4.5% within 90 days. For DUCA, this was
15, 9.9, and 7.7% of INC, ECCG, and DUCA cohorts, 1.7% for 30-day mortality and 2.4% for 90-day mor-
respectively (P = 0.002). Atrial dysrhythmia was sig- tality, with no significant difference. The 30-day read-
nificantly (P < 0.05) higher at 22.8% in the INC com- mission rate was 10, 10.2, and 15% in the INC, ECCG,
pared with 14.5 and 13.9% for ECCG and DUCA, and DUCA series, respectively (P = ns) (Table 5).
respectively, as was sepsis (P < 0.05) at 5.9, 1.9, and
1.1%, respectively, and multisystem organ failure at 4, DISCUSSION
1, and 0.1%, respectively. The 30-day readmission was
10, 11 and 15% in the INC, ECCG, and DUCA series, The ECCG consensus represents major progress in
respectively. the standardized reporting of complications for the
complex operation of esophageal cancer resection,
with enormous potential value for national and
Complications with level of complexity defined as per international benchmarking of data, clinical trials,
ECCG and clinical research. This focus on strict definitions
The low rate of most specific complications precludes is particularly required at a time where minimally
statistical analysis due to insufficient numbers of invasive approaches, or hybrid techniques, and
observations. For leak rates, occurring in 5.4% of robotic-assisted surgery, is evolving rapidly, and where
cases, 45% required no intervention, 45% requiring the evidence-base will largely focus on a reduction in
drainage by interventional radiology, and just one complication rate and recovery of quality of life. This
patient required surgery, with an overall reoperation report from a well-structured academic high-volume
Use of ECCG definitions for comparative audit 5
conduit irrespective of Type II: localized defect requiring interventional but not 6 2.7% 131 4.8% 131 8.1%
presentation or method of surgical therapy
identification Type III: localized defect requiring surgical therapy 1 0.4% 80 3.0% 83 5.1%
Conduit necrosis/failure Grade
Definition: postoperative No conduit necrosis 216 98.6% 2,672 98.8% 1,604 99.2%
identification of conduit Type I: conduit necrosis foal identified endoscopically 2 0.9% 2 0.1% 1 0.1%
necrosis (Treatment—additional monitoring or non-surgical therapy)
Type II: conduit necrosis focal identified endoscopically 0 0% 7 0.3% 3 0.2%
& not associated with free anastomotic or conduit leak
(Treatment—surgical therapy not involving esophageal diversion)
Type III: Conduit necrosis extensive 1 0.4% 23 0.9% 9 0.6%
(Treatment—treated with conduit resection with diversion)
Recurrent laryngeal nerve Grade
injury involvement
Definition: vocal cord No recurrent laryngeal nerve injury 212 96.8% 2,595 96.0% 1,547 95.7%
dysfunction postresection. Type I: transient injury requiring no therapy 1 0.4% 87 3.2% 61 3.8%
Confirmation & assessment Type II: injury requiring elective surgical procedure, for 1 0.4% 16 0.5% 5 0.3%
should be by direct example, thyroplasty or medialization procedure
examination Type IIIa: unilateral injury requiring acute surgical 5 2.2% 2 0.1% 0 0%
intervention (due to aspiration or respiratory issues), for
example, thyroplasty or medialization procedure
Type IIIb: bilateral Injury requiring acute surgical 0 0% 4 0.1% 1 0.1%
intervention (due to aspiration or respiratory issues); for
example; thyroplasty or medialization procedure
Chyle leak severity Grade
Definition: milky discharge No chyle leak 207 94.5% 2,578 95.3% 1,478 91.4%
upon initiation of enteric Type Ia: <1 L output, Treatment—enteric dietary 3 1.3% 67 2.5% 68 4.2%
feeds and/or pleural fluid modifications
analysis demonstrating Type Ib: >1 L output, Treatment—enteric dietary 0 0% 10 0.4% 3 0.2%
triglyceride level > 100 mg/dL modifications
and/or chylomicrons in Type IIa: <1 L output, Treatment—total parenteral 7 3.2% 11 0.4% 14 0.9%
pleural fluid nutrition
Type IIb: >1 L output, Treatment—total parenteral 1 0.4% 6 0.2% 8 0.5%
nutrition
Type IIIa: <1 L output, Treatment—interventional or 0 0% 12 0.4% 2 0.1%
surgical therapy
Type IIIb:>1 L output, Treatment—interventional or 1 0.4% 20 0.7% 19 1.2%
surgical therapy
designated national center, with exclusively open sur- these other series, or the management of severe
gical approaches, represents a contemporary opportu- complications, it is not possible to conclude that the
nity to compare and contrast with the original ECCG management of severe complications is a major factor
benchmark series of 2,704 patients from 24 hospitals for low mortality; nonetheless, it may be a reasonable
and 1,607 patients from 22 hospitals in the Nether- inference.21,22
lands within DUCA, and to also seek a broader A key finding, possibly linked to the mortality rate,
interpretation with respect to modern international is a low anastomotic leak rate of 5.4%, twofold less
standards of care. than 11.4% in the ECCG, and almost fourfold less
All series reflect modern practice from high- than the 19% in the DUCA series (P < 0.001). In the
volume centers, with low mortality, yet complications Seattle series with 0.3% mortality, an anastomotic
still occur in over 50% of patients. The most dreaded leak was reported in just 3.8% of patients.20 These
outcome in surgery is a postoperative death, followed leak rates contrast with high reported anastomotic
arguably by an anastomotic leak or gastric conduit leak rates in the CROSS RCT, at 26%.4 A more recent
necrosis, which may represent a failure of technique. study, also from the Netherlands, reported a 17% leak
Prior systematic reviews of the literature pre-1980, rate for thoracic anastomoses, and 21.9% for cervical
from 1980–1988, and from 1990–2000 reported anastomoses.23 Clearly, a wide spectrum of leak rates
mortality rates of 29, 13, and 6.7%, respectively.17–19 is evident even in the leading international centers,
A modern benchmark may be 4% or less.4 In this and the optimization of technique and outcome is
INC series, no 30-day mortality was observed and clearly an imperative to further improve outcome.24
two deaths occurred within 90 days (0.9%), one from The grading of leaks within ECCG is a major
ARDS postmultimodal therapy and the other from advance, as type I and type II leaks can be managed
repeated mesenteric ischemia in a patient with atrial relatively simply; hence, an important feature of our
fibrillation, and this compares favorably (P < 0.05) experience is that type III leaks, requiring surgical
with 90-day mortality of 4.5 and 2.4%, respectively, therapy, and typically more serious, revealed the
for the ECCG and DUCA series. To our knowledge, greatest proportional differences, with 0.4% at the
the publication in 2007 of 340 consecutive patients INC, 3% within ECCG, and 5.1% in the DUCA
by Low et al, with 0.3% mortality, represents the report. There are myriad technical approaches to an
lowest published mortality rate in a high-volume anastomosis, as well as key factors including conduit
series, and the focus on a standardized clinical perfusion, anesthetic factors, and comorbidities,
pathway in a specialist center was cited as a key as well as postoperative care and the impact of
factor in achieving this outcome.20 It is the possible other complications. Notwithstanding, this series
similar factors of structure and process that apply has shown that that a simple uniform standardized
at this Center, and that an ability to rescue from sutured approach resulted in low leak rates.24,25 This
severe complications is more likely to be embedded is apposite in an MIE era. Although we do not know
in high-volume institutions.21,22 In addition to the the differential leak rate of open and MIE cases within
management of complications, a low mortality rate the ECCG and DUCA series, a recent study from
may reflect surgery on a more selected lower risk four European centers identified a mean of 119 cases
cohort or reduced complications. A comparison required in the MIE learning curve, with a reduction
of demographics, including ASA, co-morbidities, in anastomotic leak from 29.5% to less than 5%
and obesity rates, suggests that case selection was after the learning curve is complete, but an over 10%
not a relevant factor. There were also no major leakage during this learning period. In light of this
differences in the severity of complications, with learning curve, our series highlights a major modern
Clavien–Dindo ≥ IIIb and < 5 in 14% at this Center, dilemma for this Center and probably for many
14.3% in the ECCG, and 16% in DUCA. Accordingly, surgeons internationally who are intending to or are
without knowing the exact causes of mortality in transitioning to MIE or robot-assisted approaches.26
8 Diseases of the Esophagus
Conversely, a recent report from Toronto of 383 in this series, at 5%, compared with 1.9 and 1.1%
patients, where sequentially there was a transition rate in the ECCG and DUCA series, respectively.
from open to MIE approaches, with approximately These latter rates appear very low and inconsistent
a 50:50 ratio, reported similar leak rates of 12.4 with the reported leak rates and complication severity,
versus 8.9% and 90-day mortality of 6.2 and 4.6% perhaps raising broader questions about data docu-
in open and MIE, respectively, highlighting that a mentation across a large national or international net-
transitioning experience in a major specialist Center work? A THE is selectively used at this Center, partic-
of less than 1%. The study highlights the value of 19 Jamieson G G, Mathew G, Ludemann R et al. Postoperative
the ECCG definitions and Esodata project for com- mortality following oesophagectomy and problems in reporting
its rate. Br J Surg 2004; 91: 943–7.
parative analysis, and the clear potential to use the 20 Low D E, Kunz D, Schembre D, Otero H et al. Esophagec-
large contemporaneous ECCG dataset to enable audit tomy – it’s not just about mortality anymore: standardized
and quality improvement at institution, national, and perioperative clinical pathways improve outcomes in patient
with esophageal cancer. J Gastrointest Surg 2007; 11: 1395–402.
international level. It also creates an opportunity 21 Finks J F, Osborne N H, Birkmeyer J D. Trends in hopsital
to increase knowledge and to improve standards volume and operative mortality for high risk surgery. N Engl J