ORIGINAL ARTICLE

Anastomotic Leak Increases Distant Recurrence and Long-Term

Mortality After Curative Resection for Colonic Cancer
A Nationwide Cohort Study
Peter-Martin Krarup, MD, Andreas Nordholm-Carstensen, MD, Lars N. Jorgensen, MD, DMSc,
and Henrik Harling, MD, DMSc; on behalf of the Danish Colorectal Cancer Group

resection for adenocarcinoma, consistent results have demonstrated

Objective: To investigate the impact of anastomotic leak (AL) on disease
that AL is associated with an increased rate of local recurrence and
recurrence and long-term mortality in patients alive 120 days after curative
long-term mortality, whereas the rate of distant recurrence remains
resection for colonic cancer.
unaffected.811 In contrast, AL after colonic resection has previously
Background: There is no solid data as to whether AL after colonic cancer
not been associated with higher rates of local or distant recurrence.9
surgery increases the risk of disease recurrence.
However, there is evidence that AL reduces both disease-free and
Methods: This was a nationwide cohort study of 9333 patients, prospec-
overall survivals in patients with colonic cancer.2,6 Accordingly, it is
tively registered in the database of the Danish Colorectal Cancer Group and
essential to distinguish between colonic and rectal cancer surgery be-
merged with data from the Danish Pathology Registry and the National Pa-
cause the influence of AL on long-term outcomes seems to differ. The
tient Registry. Multivariable Cox regression analysis was used to adjust for
excess long-term mortality after AL could be mediated by increased
confounding.
rates of disease recurrence, higher comorbidity or the influence from
Results: The incidence of AL was 6.4%, 744 patients died within 120 days. Of
factors associated with AL.12,13
the remaining 8589 patients, 861 (10.0%) developed local recurrence with no
In this nationwide study on colonic cancer surgery, the primary
association to AL [adjusted hazard ratio (HR) = 0.78; 95% confidence interval
objective was to investigate the impact of AL on the rates of local
(CI): 0.551.12; P = 0.184]. Distant recurrence developed in 1281 (14.9%)
and distant disease recurrence and secondary objective was to do it
patients and more frequently after AL (adjusted HR = 1.42; 95% CI: 1.13
on long-term all-cause mortality. Furthermore, a subgroup analysis
1.78; P = 0.003). AL was also associated with increased long-term mortality
was conducted to evaluate the influence of AL on administration of
(adjusted HR = 1.20; 95% CI: 1.011.44; P = 0.042). In 2841 patients with
adjuvant chemotherapy (AC) in patients with stage III cancer.
stage III cancer, AL was associated with both decreased likelihood of receiving
adjuvant chemotherapy (adjusted HR = 0.58; 95% CI: 0.450.74; P < 0.001)
and a delay to initial administration (16 days; 95% CI: 1220 days; P < 0.001). METHODS
Conclusions: AL was significantly associated with increased rates of distant
recurrence and long-term all-cause mortality. Cancelled or delayed adminis- Study Population and Variables
tration of adjuvant chemotherapy may partly account for these findings. This was a nationwide cohort study with data from 3 Danish
registers: the prospective database of the Danish Colorectal Cancer
Keywords: adjuvant chemotherapy, anastomotic leakage, colon cancer, mor-
Group (DCCG), the Danish Pathology Registry, and the National Pa-
tality, nationwide, recurrence
tient Registry. Data from these registers were merged to investigate
(Ann Surg 2013;00:19) possible associations between AL and long-term outcomes. The pri-
mary and secondary outcomes were disease recurrence and all-cause
mortality, respectively.

C olorectal cancer is one of the most frequent malignancies world-

wide with an estimated 1.2 million new cases and 600,000 deaths
annually.1 Surgical resection is essential to obtain long-term disease-
All patients included in the study were recorded in the DCCG
database between May 2001 and December 2008 with a first-time di-
agnosis of colonic adenocarcinoma. The DCCG database is approved
free survival, but postoperative complications have a significant im- by the Danish Data Protection Agency (Ref no. 2000-53-0073) and
pact on the surgical outcome. Anastomotic leak (AL) is one of the includes at least 95% of all Danish patients with colorectal cancer.4
most devastating complications in these patients because of an im- All patients underwent a curative colonic resection with a primary
mense increase in short-term morbidity and mortality.2,3 The inci- anastomosis without a protecting ostomy. Trained pathologists eval-
dence of AL varies between 3% and 12%2,4,5 depending on the type uated all excisional specimens according to the guidelines of the
of surgical procedure. There are only few and contradictory studies on DCCG. The curative resection criteria required at least 2 mm from
the oncological outcome and long-term mortality in patients with AL the tumor to the nonperitonealized resection margin, as evaluated mi-
after resection of colonic cancer.2,69 In patients undergoing rectal croscopically, and with no tumor growth or distant disease left after
completed surgery. The proximal and distal resection margins were
examined microscopically when the distance from the tumor to these
From the Department of Surgery K, Bispebjerg Hospital, University of Copen- margins were less than 20 mm.
hagen, Copenhagen NV, Denmark. Data on disease recurrence and overall survival were obtained
Disclosure: Founding was received from the I. M. Dhnfeldt Foundation, the A. P.
Mller Foundation, and the Danish Medical Association. The authors have no
from the Danish Pathology Registry and the National Patient Registry
conflicts of interest to declare. on November 25, 2010. Recurrent disease was defined according to
Reprints: Peter-Martin Krarup, MD, Department of Surgery K, Bispebjerg the DCCG guidelines as local or distant recurrence diagnosed no
Hospital, University of Copenhagen, Denmark, Bispebjerg Bakke 23, earlier than 120 days after the index operation. All cases of recurrent
DK-2400, Copenhagen NV, Denmark. E-mail: krarup75@gmail.com.
Copyright C 2013 by Lippincott Williams & Wilkins
disease were confirmed histologically, by diagnostic imaging or
ISSN: 0003-4932/13/00000-0001 surgery. Patients with both local and distant recurrence were
DOI: 10.1097/SLA.0b013e3182a6f2fc classified as distant recurrence. All-cause mortality was exclusively

Annals of Surgery r Volume 00, Number 00, 2013 www.annalsofsurgery.com | 1

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Krarup et al Annals of Surgery r Volume 00, Number 00, 2013

TABLE 1. Patient Characteristics Patients with colonic cancer

assessed for inclusion
(n = 18,484)
AL No AL
n = 8589 420 (5%) 8169 (95%) Patients rejecting operation (n = 1344)
Gender
Female 171 (41%) 4331 (53%)
Male 249 (59%) 3838 (47%) Excluded procedures (n = 4501)
Age groups, yr Hartmann's procedure (n = 1311)
60 80 (19%) 1466 (18%) Total colectomy (n = 769)
Anterior resection of the rectum (n = 435)
>6070 138 (33%) 2222 (27%) Ostomy only (n = 429)
>7080 157 (37%) 2865 (35%) Self-expandable metal stents (n = 381)
>80 45 (11%) 1616 (20%) Polypectomy n = 258
Exploratory laparotomy only (n = 131)
Comorbidity Other procedures (n = 787)
Normal 266 (63%) 5342 (65%)
Moderate 63 (15%) 1274 (16%)
Severe 44 (11%) 799 (10%) Residual disease (n = 2903)
Very severe 33 (8%) 454 (6%)
Missing 14 (3%) 300 (3%)
Tumour stage Anastomosis with a defunctioning ostomy (n = 389)
UICC I 74 (18%) 1207 (15%)
UICC II 202 (48%) 3905 (48%) Mismatch between the National Patient Registry and the
UICC III 129 (31%) 2712 (33%) Database of the Danish Colorectal Cancer Group (n = 14)
Missing 15 (3%) 345 (4%)
Surgical procedure
Right hemicolectomy 169 (40%) 4090 (50%) Eligible patients (n = 9333)
Transverse colectomy 8 (2%) 214 (3%)
Left hemicolectomy 72 (17%) 823 (10%)
Sigmoid colectomy 171 (41%) 3042 (37%) Patients excluded within 120 days after surgery (n = 744)
Surgical approach Emigration (n = 4)
Open 329 (78%) 6700 (82%) Mortality (n = 740)
Laparoscopy 91 (22%) 1469 (18%)
Surgical priority
Elective 384 (91%) 7607 (93%) Patients included in the study
(n = 8589)
Emergency 36 (9%) 562 (7%)
Surgeons specialization
Gastrointestinal 324 (77%) 5976 (73%) FIGURE 1. Inclusion chart modified from Krarup et al.5
General 96 (23%) 2191 (27%)
Missing 0 (0%) 2 (<1%)
Organ resection The independent variable of interest, AL, was defined accord-
None 371 (89%) 7243 (89%) ing to the guidelines of the DCCG: Clinical symptoms suggesting
Spleen 9 (2%) 71 (<1%) AL and confirmed by contrast enema or computerized tomography
Other organs 39 (9%) 855 (11%) within 60 days from the operation. Patients with AL were identi-
Missing 1 (<1%) 0 (0%) fied in the DCCG database or National Patient Registry using the
Blood transfusion International Classification of Disease (ICD-10) for diagnosis and
No 170 (41%) 6468 (79%)
Yes 244 (58%) 1643 (20%)
reoperation codes associated with AL (DT813A, KJWF00).
Missing 6 (1%) 58 (1%) Relevant and potential confounding covariates were extracted
from the DCCG database and the National Patient Registry including

Comorbidity according to Charlson comorbidity index scores of 0 (normal), age, gender, comorbidity as assessed by the Charlson index, tumor
1 (moderate), 2 (severe), and 3 (very severe).
stage (Union for International Cancer control, UICC), type of surgery,
surgeons specialization level, resection of adjacent organs, and peri-
operative blood transfusion. The Charlson comorbidity index reflects
the cumulative increased likelihood of 1-year mortality and was cal-
investigated in patients surviving the first 120 postoperative days culated from ICD-10 diagnoses registered in the National Patient
after primary surgery. Registry before the day of surgery.14,15 The comorbidity score was
The outcome in patients with stage III colonic cancer may be categorized as normal (0), moderate (1), severe (2), and very severe
predicted by the radicality of surgery, administration of AC, and the (3).16
time to initial administration in patients who receive AC. A post hoc
analysis was therefore undertaken to investigate the impact of AL on Statistical Analysis
administration of AC in the group of patients with stage III cancer. Duration of follow-up was calculated from the day of surgery
In the subset of patients who did receive AC, the influence of AL on and analyzed by the reverse Kaplan-Meier method.17 For long-term
time to initial administration was likewise examined. Patients were outcomes, Kaplan-Meier curves were plotted and patients with and
regarded as having AC if initial administration was achieved before without AL were compared by the log-rank analysis. Univariable
postoperative day 120 according to data from the National Patient and multivariable Cox regression analyses were used to investigate
Registry. This minimized the risk of bias associated with recurrence the influence of AL on disease recurrence, all-cause mortality, and
developing before the initial administration of AC. During the study administration of AC.
period, the standard AC regimen for patients with stage III colonic Hazard ratios (HRs) with 95% confidence intervals (CIs) of
cancer was a combination of fluorouracil/leucovorin and oxaliplatin. more than 1 indicated increased likelihood of disease recurrence, fatal

2 | www.annalsofsurgery.com 
C 2013 Lippincott Williams & Wilkins

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 00, Number 00, 2013 Anastomotic Leak and Distant Recurrence

TABLE 2. Multivariable Cox Regression Analyses of Long-Term Outcomes

Local Recurrence Distant Recurrence All-Cause Mortality
HR 95% CI P HR 95% CI P HR 95% CI P
AL 0.184 0.003 0.042
No 1.00 1.00 1.00
Yes 0.78 0.551.12 1.42 1.131.78 1.20 1.011.44
Gender 0.302 0.033 <0.001
Female 1.00 1.00 1.00
Male 1.08 0.941.23 1.13 1.011.26 1.35 1.251.46
Age groups, yr <0.001 <0.001 <0.001
60 1.00 1.00 1.00
> 6070 0.72 0.610.86 <0.001 0.93 0.801.09 0.380 1.37 1.181.60 <0.001
> 7080 0.55 0.450.65 <0.001 0.90 0.771.05 0.176 2.29 1.992.64 <0.001
> 80 0.35 0.270.45 <0.001 0.62 0.500.75 <0.001 4.14 3.584.78 <0.001
Comorbidity 0.119 0.004 <0.001
Normal 1.00 1.00 1.00
Moderate 1.19 0.991.44 0.065 1.04 0.881.22 0.654 1.39 1.251.54 <0.001
Severe 1.14 0.911.43 0.266 1.20 1.001.44 0.057 1.63 1.441.83 <0.001
Very severe 0.83 0.581.18 0.293 1.45 1.171.81 <0.001 1.99 1.732.29 <0.001
Tumour stage <0.001 <0.001 <0.001
UICC I 1.00 1.00 1.00
UICC II 0.97 0.791.19 0.759 1.90 1.502.40 <0.001 1.33 1.161.52 <0.001
UICC III 1.50 1.231.84 <0.001 4.18 3.315.27 <0.001 2.28 1.992.61 <0.001
Surgical procedure 0.111 0.993 0.035
Right hemicolectomy 1.00 1.00 1.00
Transverse colectomy 1.33 0.901.96 0.154 1.03 0.711.47 0.892 1.09 0.871.36 0.441
Left hemicolectomy 0.89 0.701.12 0.313 1.00 0.831.21 0.984 0.90 0.791.04 0.154
Sigmoid colectomy 0.88 0.751.03 0.101 0.99 0.871.12 0.812 0.90 0.810.97 0.010
Surgical approach 0.015 0.789 0.641
Open 1.00 1.00 1.00
Laparoscopic 1.24 1.041.48 0.98 0.831.16 0.97 0.861.10
Surgical priority 0.051 <0.001 <0.001
Elective 1.00 1.00 1.00
Emergency 1.32 1.001.74 2.14 1.802.54 1.75 1.541.99
Surgeons specialization <0.001 0.181 0.023
Gastrointestinal 1.00 1.00 1.00
General 0.69 0.580.82 1.09 0.961.24 1.10 1.011.20
Organ resection 0.211 <0.001 <0.001
None 1.00 1.00 1.00
Spleen 1.16 0.572.37 0.690 1.33 0.792.24 0.291 1.50 1.072.10 0.019
Other organs 1.21 0.981.51 0.082 1.59 1.351.86 <0.001 1.38 1.231.55 <0.001
Blood transfusion 0.467 0.519 0.002
No 1.00 1.00 1.00
Yes 1.07 0.891.28 1.05 0.911.20 1.15 1.051.26
The number of patients included in each outcome response was N = 6900 for local recurrence, N = 7315 for distant recurrence, and N = 8164 for all cause mortality.

HR < 1 indicates decreased likelihood of disease recurrence or mortality, whereas HR > 1 indicates increased likelihood of disease recurrence or mortality.
Comorbidity according to Charlson comorbidity index scores of 0 (normal), 1 (moderate), 2 (severe), and 3 (very severe).

outcome, and administration of AC. Schoenfeld residuals were exam- -Haenszel random-effects model (Review Manager, Version 5.1,
ined to verify the assumption of proportional hazards. The adjusted Copenhagen, Denmark: the Nordic Cochrane Centre, the Cochrane
influence of AL on the time to initial administration of AC was cal- Collaboration), updating (September 2013) the systematic review
culated with multiple linear regression in the subset of patients who from Mirnezami et al9 including the present data. A sensitivity anal-
did receive AC. All variables were simultaneously included in the ysis was performed removing individual studies in a serial fashion to
multivariable analyses. Status of disease recurrence was then entered determine the influence of each study on the combined OR estimate
as a time-dependent variable to assess the potential mediating role and 95% CI.
of disease recurrence on the relationship between AL and all-cause
mortality. Missing values were disregarded and thus not replaced.
Number and percentage of missing values are provided in Table 1. RESULTS
The results of the multivariable analyses are presented as haz- A total of 18,484 patients were assessed for inclusion of whom
ard ratios (HR) with 95% confidence intervals (CI). All analyses were 9333 (50.5%) were eligible (Fig. 1). The overall incidence of AL was
2-sided and considered statistically significant if P < 0.050. Data were 593/9333 (6.4%).5 Four patients were excluded because of emigration
analyzed with SPSS Statistics Version 20 (IBM Corp, Armonk, NY). after surgery, and 740 of 9329 (7.9%) patients died within the first 120
Meta-analysis of published data on the relationship between days: 173 of 593 (29.2%) with AL and 567 of 8736 (6.5%) without
AL and distant disease recurrence was performed using the Mantel AL, P < 0.001. Thus, 8589 patients with a median age of 72 years


C 2013 Lippincott Williams & Wilkins www.annalsofsurgery.com | 3

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Krarup et al Annals of Surgery r Volume 00, Number 00, 2013

100
A 40 Distant recurrence
90
80
30
Local recurrence, %

70

Mortality, %
60 Local + distant recurrence
50
20 No anastomotic leak
40 Local recurrence
30
10 20
Anastomotic leak No recurrence
10
P < 0.001
P = 0.638 0
0 1 2 3 4 5 6 7 8 9 10
0
0 1 2 3 4 5 6 7 8 9 10 Time after surgery, yr
Numbers at risk
Time after surgery, yr No recurrence 6156 5902 5500 4427 3544 2676 1940 1249 689 227
Numbers at risk Local recurrence 861 835 779 603 430 285 203 124 76 28
Anastomotic leak 320 283 255 200 158 108 74 40 24 10 Distant recurrence 998 902 672 442 290 179 96 56 35 12
No anastomotic leak 6697 6120 5520 4383 3449 2594 1881 1214 666 217 Local + distant recurrence 283 266 218 151 113 72 43 29 15 3

FIGURE 4. Kaplan-Meier plots illustrating all-cause mortality in

B 40
patients alive 120 days after curative colonic cancer surgery.
Anastomotic leak
Mortality was increased in patients with distant recurrence (ad-
justed HR = 4.07, 95% CI 3.534.68, P < 0.001), local and
Distant recurrence, %

30 distant recurrence (adjusted HR = 3.37, 95% CI: 2.684.25, P

< 0.001), and in contrast to local recurrence (adjusted HR =
1.08, 95% CI: 0.921.27, P = 0.361).
20

No anastomotic leak
(range: 2398 years) were studied (Table 1). Median follow-up was
10 5.3 years (interquartile range [IQR]: 3.67.3 years).

P < 0.001 Disease Recurrence

0
0 1 2 3 4 5 6 7 8 9 10 The number of patients with local and distant recurrence were
Time after surgery, yr 861 (10.0%) and 1281 (14.9%), respectively. The latter group in-
Numbers at risk cluded 283 (3.3%) patients with both local and distant recurrence.
Anastomotic leak 371 322 273 211 161 110 75 40 24 10
No anastomotic leak 7066 6473 5692 4470 3499 2626 1889 1219 666 217
The number of disease-free patients alive at the end of follow-up was
4849 (56.5%). The median time to diagnosis of local and distant recur-
FIGURE 2. Kaplan-Meier plots illustrating the association be- rence was 1.2 years (IQR: 0.62.4 years) and 1.5 years (IQR: 0.92.5
tween AL and the rates of local (A) and distant (B) recurrence years), respectively. In the univariable analysis, AL was not associated
in patients alive 120 days after curative colonic cancer surgery. with local recurrence [HR = 0.92 (95% CI: 0.661.29; P = 0.638)],
which was confirmed in the multivariable analysis [HR = 0.78 (95%
CI: 0.551.12; P = 0.184)]. Covariates with a significant influence on

21

11

10

22

8
23

18
19

24

20

P I

Decreased Increased
distant recurrence distant recurrence

FIGURE 3. Forest plot on the association between AL and the incidence of distant recurrence after colonic (C) and rectal (R)
cancer surgery. The meta-analysis was updated from Mirnezami et al9 in September 2012 adding 4 additional studies6,1820 and
the present results.

4 | www.annalsofsurgery.com 
C 2013 Lippincott Williams & Wilkins

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 00, Number 00, 2013 Anastomotic Leak and Distant Recurrence

TABLE 3. Univariable and Multivariable Cox Regression Analyses of Adjuvant Chemotherapy in Patients With Stage III
Colonic Cancer
Univariable Multivariable
Adjuvant No Adjuvant
Chemotherapy Chemotherapy HR 95% CI) HR 95% CI) P
N = 2841 1725 (61%) 1116 (39%)
AL <0.001
No 1659 (96%) 1053 (94%) 1.00 1.00
Yes 66 (4%) 63 (6%) 0.65 0.510.83 0.58 0.450.74
Gender 0.449
Female 877 (51%) 650 (58%) 1.00 1.00
Male 848 (49%) 466 (42%) 1.18 1.081.29 1.04 0.941.14
Age groups, yr <0.001
60 1275 (74%) 646 (58%) 1.00 1.00
>6070 232 (13%) 225 (20%) 0.82 0.730.92 0.86 0.760.97
>7080 142 (8%) 147 (13%) 0.39 0.350.45 0.43 0.380.49
> 80 76 (4%) 98 (9%) 0.05 0.040.07 0.06 0.040.08
Comorbidity <0.001
Normal 523 (30%) 69 (6%) 1.00 1.00
Moderate 649 (38%) 165 (15%) 0.65 0.560.75 0.86 0.740.99
Severe 505 (29%) 444 (40%) 0.61 0.510.73 0.72 0.600.85
Very severe 48 (23%) 438 (39%) 0.53 0.420.67 0.62 0.490.79
Surgical procedure 0.748
Right hemicolectomy 802 (47%) 626 (56%) 1.00 1.00
Transverse colectomy 37 (2%) 31 (3%) 0.88 0.631.23 0.92 0.661.28
Left hemicolectomy 221 (13%) 109 (10%) 1.33 1.151.54 1.06 0.911.23
Sigmoid colectomy 665 (39%) 350 (31%) 1.30 1.171.44 0.98 0.881.09
Surgical approach 0.048
Open 1371 (79%) 935 (84%) 1.00 1.00
Laparoscopic 354 (21%) 181 (16%) 1.22 1.091.37 1.13 1.001.28
Surgical priority 0.029
Elective 1589 (92%) 1005 (90%) 1.00 1.00
Emergency 136 (8%) 111 (10%) 0.80 0.670.96 0.82 0.690.98
Surgeons specialization <0.001
Gastrointestinal 1391 (81%) 781 (70%) 1.00 1.00
General 334 (19%) 334 (30%) 0.69 0.620.78 0.70 0.620.79
Organ resection 0.454
None 1523 (88%) 986 (88%) 1.00 1.00
Spleen 13 (1%) 9 (1%) 0.79 0.461.36 0.71 0.401.23
Other organs 189 (11%) 121 (11%) 0.98 0.841.13 1.02 0.871.19
Blood transfusion <0.001
No 1445 (84%) 757 (69%) 1.00 1.00
Yes 268 (16%) 348 (31%) 0.52 0.460.59 0.68 0.590.78

HR < 1 indicates decreased likelihood of receiving AC, whereas HR > 1 indicates increased likelihood of AC. P, multivariable analysis.
Comorbidity according to Charlson comorbidity index scores of 0 (normal), 1 (moderate), 2 (severe), and 3 (very severe).

80 local recurrence were age, tumor stage, surgical approach, and sur-
70 geons specialization level (Table 2). In contrast to local recurrence,
No anastomotic leak
Adjuvant chemotherapy, %

there was a significant association between AL and distant recurrence

60
in both the univariable analysis [HR = 1.49 (95% CI: 1.201.85; P <
50 0.001] (Fig. 2) and the multivariable analysis [HR = 1.42 (95% CI:
40 1.131.78; P = 0.003] (Table 2). In addition, gender, age, comorbid-
30
ity, tumor stage, surgical priority, and extracolonic organ resection
Anastomotic leak
were significantly associated with distant recurrence (Table 2).
20
The association between AL and distant recurrence was put
10 in context of previous published studies, updating a meta-analysis by
P < 0.001
0 Mirnezami et al.9 Four additional studies were included in addition
0 20 40 60 80 100 120 to the present data.6,1820 The Forest plot showed a significant asso-
Numbers at risk
Time after surgery, d ciation between AL and development of distant recurrence (OR =
Anastomotic leak 129
No anastomotic leak 2712
129
2691
121
2112
94
1322
80
1122
68
1073
63
1053
1.38; 95% CI: 1.051.81; P = 0.02; Fig. 3). The sensitivity analysis
rendered a nonsignificant association between AL and distant recur-
FIGURE 5. Kaplan-Meier plots illustrating the rates of adminis- rence in the combined estimate after exclusion of the study by Law
tered AC in stage III colonic cancer patients with and without et al20 (OR = 1.26; 95% CI: 0.991.60; P = 0.06) or the present study
AL. (OR = 1.35; 95% CI: 0.971.89; P = 0.08).


C 2013 Lippincott Williams & Wilkins www.annalsofsurgery.com | 5

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Krarup et al Annals of Surgery r Volume 00, Number 00, 2013

TABLE 4. Multiple Linear Regression Analysis of Time to Administration of Adjuvant Chemotherapy in

Adjuvant Chemotherapy (Days) Stage III Cancer Patients
In the subgroup of patients with stage III disease, administra-
Estimated tion of AC was initiated in 1725 of 2841 (60.7%) patients within 120
Coefficient days. Patients with AL were less likely to receive AC according to
Variables (95% CI) P
both the univariable [HR = 0.65 (95% CI: 0.510.83; P < 0.001)]
AL <0.001 and the multivariable models [HR = 0.58 (95% CI: 0.450.74; P <
No (reference) 0 0.001] (Table 3 and Fig. 5). In addition, the time to initial adminis-
Yes 16 (1220) tration of AC was median 59 days (IQR: 4882 days) after surgery
Gender 0.804 for patients with AL compared with 43 days (IQR: 3554 days) in
Female (reference) 0 patients without AL (P < 0.001). Multiple linear regression analysis
Male 0 (1 to 2)
Age groups, yr 0.187
confirmed that administration of AC was initiated 16 days (95% CI:
60 (reference) 0 1220 days; P < 0.001) later in patients with AL compared with
> 6070 0 (3 to 1) 0.520 patients without AL (Table 4).
> 7080 1 (03) 0.218 Administration of AC in patients with AL was not associated
> 80 2 (2 to 7) 0.396 with a reduction in distant recurrence but a significant reduction in all-
Comorbidity 0.176 cause mortality (Figs. 6A, C). Administration of AC beyond 55 days
Normal (reference) 0 from primary surgery was not associated with a significant reduction
Moderate 1 (1 to 4) 0.205 in long-term mortality [HR = 0.88 (95% CI: 0.681.02] (Table 5).
Severe 3 (06) 0.060 The rates of distant recurrence and mortality in patients receiving AC
Very severe 2 (3 to 5) 0.546
Surgical procedure 0.153
within day 55 were comparable between patients with and without
Right hemicolectomy (reference) 0 AL (Figs. 6B, D).
Transverse colectomy 5 (011) 0.058
Left hemicolectomy 2 (4 to 1) 0.234
Sigmoid colectomy 0 (3 to 1) 0.376 DISCUSSION
Surgical approach 0.026 This nationwide study demonstrates an inferior oncological
Open (reference) 0 outcome in patients who develop AL after curative resection for
Laparoscopic 2 (4 to 0) colonic cancer and survive the first critical postoperative phase. This
Surgical priority <0.001 is reflected by an increased rate of distant recurrence and therefore a
Elective (reference) 0 continuing excess mortality.
Emergency 5 (28) The distinct finding, that AL was associated with a significant
Surgeons specialization 0.696
increase in distant recurrence, provides new insight into the negative
Gastrointestinal (reference) 0
General 0 (2 to 2) implications of AL in patients with colonic cancer. The issue was
Organ resection 0.012 recently addressed in a meta-analysis in which Mirnezami et al were
None (reference) 0 unable to confirm a relationship between AL and distant recurrence
Spleen 12 (321) 0.008 (overall OR = 1.38; 95% CI: 0.961.99; P = 0.083).9 However, the
Other organs 2 (1 to 4) 0.232 meta-analysis included pooled data from studies on both colonic8
Blood transfusion <0.001 and rectal cancer surgery,8,10,11 as well as mixed colorectal cancer
No (reference) 0 surgery.2124 The update of the meta-analysis by Mirnezami et al
Yes 5 (37) including the present data suggests a relationship between AL and
The analysis included a subset of patients with stage III colonic cancer who development of distant recurrence. The sensitivity analysis indicates
received adjuvant chemotherapy within 120 days from the index operation, N = that the large cohort in the present study may drive the outcome of
1725. The Intercept in this model was 44 days (95% CI: 4247 days) and predicts
the time to AC when all variables take the reference value. The coefficients were
the meta-analysis to become statistically significant. A fuller meta-
rounded to whole numbers (days) and estimate the change in time to AC given analysis including a subgroup analysis of patients with colonic cancer
a change from the reference value. is warranted to address the robustness of the conclusion that AL in-

Comorbidity according to the Charlson Comorbidity Index scores of 0 creases the risk of distant recurrence in patients with colonic cancer.
(normal), 1 (moderate), 2 (severe), and 3 (very severe).
The unadjusted results from the meta-analysis could also be subjected
to confounding because advanced tumor stage, emergency surgery,
and comorbidity were strong predictors for distant recurrence in the
present study. Nevertheless, the association between AL and distant
recurrence remained significant in the multivariable analysis. A lim-
itation of this finding is the competing risk of mortality in patients
Long-Term All-Cause Mortality dying before a potential recurrence.
AL was significantly associated with increased all-cause mor- The present study included patients from all Danish depart-
tality in the multivariable analysis [HR = 1.20 (95% CI: 1.011.44; ments conducting colorectal surgery during the study period and was
P = 0.042)]. Covariates that reached statistical significance were gen- further strengthened by an almost complete and unselected com-
der, age, comorbidity, tumor stage, surgical procedure and priority, pilation of data merged from 3 different population-based national
surgeons specialization level, extracolonic organ resection, and blood registers. There was a long follow-up period extending beyond the
transfusion (Table 2). traditional 2 years used for assessing recurrence status. However,
In contrast to local recurrence, distant recurrence was sig- radiological and endoscopic follow-up was not performed routinely
nificantly (P < 0.001) associated with all-cause mortality (Fig. 4). raising the risk that some patients with recurrence may have remained
Inclusion of recurrence status as a time-dependent variable in the unrecognized despite the study design with 3 independent sources of
multivariable model eliminated the impact of AL on mortality [HR = data. Another limitation is the allocation of patients with both local
1.10 (95% CI: 0.921.32; P = 0.289)]. and distant recurrence to the group of patients with distant recurrence.

6 | www.annalsofsurgery.com 
C 2013 Lippincott Williams & Wilkins

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 00, Number 00, 2013 Anastomotic Leak and Distant Recurrence

A 80 C 80 AL - AC
No AL - AC
70 70
Distant recurrence, %

60 60
AL - AC

Mortality, %
50 AL + AC 50
40 40
30 30 AL + AC
20 No AL+ AC 20 No AL + AC
No AL - AC
10 10
P = 0.083 P < 0.001
0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time after surgery, yr Time after surgery, yr
Numbers at risk Numbers at risk
AL + AC 61 50 39 33 22 11 7 1 1 AL + AC 66 61 53 43 30 17 11 4 4 1
AL - AC 59 44 32 18 14 13 7 3 1 1 AL - AC 63 53 44 25 20 19 9 3 1 1
No AL + AC 1399 1246 1038 792 580 422 274 163 81 9 No AL + AC 1659 1587 1415 1072 779 540 345 208 109 15
No AL - AC 973 804 635 461 342 258 194 140 88 46 No AL - AC 1053 928 749 551 407 304 231 166 106 59

B 60 D 60
AL + AC > day 55
AL + AC day 55 No AL + AC > day 55
50 50
AL + AC > day 55
Distant recurrence, %

40 40

Mortality, %
30 30

20 No AL + AC > day 55 20
No AL + AC day 55 No AL + AC day 55
10 10 AL + AC day 55
P = 0.176 P = 0.003
0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time after surgery, d Time after surgery, d
Numbers at risk Numbers at risk
AL + AC day 55 29 23 17 15 12 5 3 AL + AC day 55 30 27 24 20 16 8 5 2 2 1
AL + AC > day 55 32 27 22 18 10 6 4 1 1 AL + AC > day 55 36 34 29 23 14 9 6 2 2
No AL + AC day 55 1090 981 826 619 457 332 218 126 57 2 No AL + AC day 55 1297 1251 1124 834 610 424 270 160 78 2
No AL + AC > day 55 309 265 212 173 123 90 56 37 24 7 No AL + AC > day 55 362 336 291 238 169 116 75 48 31 13

FIGURE 6. Kaplan-Meier plots illustrating the rates of distant recurrence (A, B) and all-cause mortality (C, D) in patients with and
without ALs after surgery for stage III colonic cancer. Data were stratified for administration of AC in panels A and C, and for time
to initial administration of AC in panels B and D. The P value represents the overall log-rank analyses.

TABLE 5. Univariable and Multivariable Analyses of Time to Initial Administration of Adjuvant Chemotherapy and
All-Cause Mortality In Patients With Stage III Colonic Cancer
Univariable Multivariable
n HR (95% CI) P HR (95% CI) P
Time to chemotherapy, d <0.001 <0.001
No chemotherapy 1116 1.00 1.00
035 434 0.51 (0.420.61) 0.76 (0.620.94)
3643 367 0.43 (0.350.54) 0.63 (0.500.80)
4455 424 0.38 (0.310.47) 0.53 (0.420.66)
56120 398 0.62 (0.520.74) 0.83 (0.681.02)
Time intervals were based on quartiles. Multivariable Cox regression analysis was adjusted for AL, age, gender, comorbidity, surgical procedure, approach and priority,
surgeons specialization, extracolonic organ resection, and blood transfusion.

HR < 1 indicates decreased likelihood of fatal outcome.

This could potentially limit the analyses of local recurrence because Although AL increased the rate of distant recurrence, no in-
a subset of these patients was thus not included in the analyses. The fluence on local recurrence was observed. This is in agreement with
purpose was to avoid an overinterpretation of the impact of local re- other reports on colonic surgery but differs from rectal surgery.79 The
currence and was supported by the finding of equal mortality rates in finding illustrates the importance of analyzing long-term outcome in
any patient with distant metastasis compared with patients with local patients with colonic and rectal cancer separately. The incidence of
recurrence (Fig. 4). local recurrence is reported lower after resection of colonic cancer


C 2013 Lippincott Williams & Wilkins www.annalsofsurgery.com | 7

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Krarup et al Annals of Surgery r Volume 00, Number 00, 2013

as compared with rectal cancer, possibly because local recurrence in the poor long-term prognosis in patients with AL was promoted by
colonic cancer may remain undiagnosed until symptomatic distant increased metastatic activity. Cancelled or delayed administration of
metastases develop.25 In addition, data from a human study demon- AC because of AL may partly account for these findings.
strated significantly higher cytokine levels, especially interleukin-8,
in the region around rectal anastomoses as compared with colonic ACKNOWLEDGMENTS
anastomoses.26 interleukin-8 promotes proliferation and metastasis Dr Steffen Hgskilde extracted data from the databases of
of colorectal cancer cells and may impair anastomotic healing.26,27 the DCCG, the Danish Pathology Registry, and the National Patient
Interestingly, local recurrence did not affect the overall mortal- Registry. The authors thank Dr Birgitte Frederiksen for valuable dis-
ity. This somewhat surprising finding was also reported in a random- cussions on the potential confounding of covariates.
ized trial on the effect of preoperative radiotherapy in rectal cancer
surgery, where a significant reduction in local recurrence from 8.2%
to 2.4% was not paralleled by a reduction in mortality.28 The present REFERENCES
cohort of more than 9000 unselected patients with colonic cancer 1. Jemal A, Bray F, Center MM, et al. Global cancer statistics. CA Cancer J Clin.
2011;61:6990.
allowed a thorough analysis of the association of AL with local re-
2. Kube R, Mroczkowski P, Granowski D, et al. Anastomotic leakage after colon
currence including adjustment for confounders. The findings confirm cancer surgery: a predictor of significant morbidity and hospital mortality, and
that AL does not predict development of local recurrence in patients diminished tumour-free survival. Eur J Surg Oncol. 2010;36:120124.
with colonic cancer. 3. Khan AA, Wheeler JM, Cunningham C, et al. The management and outcome
Previously, the impact of AL on long-term mortality has been of anastomotic leaks in colorectal surgery. Colorectal Dis. 2008;10:587592.
analyzed in different ways. Five-year overall survival was decreased in 4. Bertelsen CA, Andreasen AH, Jorgensen T, et al. Anastomotic leakage af-
patients with AL,2,6 but these studies included patients dying within ter anterior resection for rectal cancer: risk factors. Colorectal Dis. 2010;12:
3743.
30 days after surgery. To avoid this potential overlap of short- and
5. Krarup PM, Jorgensen LN, Andreasen AH, et al. A nationwide study on
long-term mortality, patients who died within the first 120 days were anastomotic leakage after colonic cancer surgery. Colorectal Dis. 2012;14:
excluded from the present analysis. It was subsequently observed that e661e667.
patients with AL were still at increased risk of a fatal outcome. Ad- 6. Marra F, Steffen T, Kalak N, et al. Anastomotic leakage as a risk factor for the
justment for comorbidity, a known predictor for reduced survival in long-term outcome after curative resection of colon cancer. Eur J Surg Oncol.
patients with colorectal cancer,29 did not alter this conclusion. Inter- 2009;35:10601064.
estingly, the influence of AL on mortality did not persist with addition 7. Branagan G, Finnis D. Prognosis after anastomotic leakage in colorectal
surgery. Dis Colon Rectum. 2005;48:10211026.
of recurrence status to the multivariable analysis. This suggests that
8. Eberhardt JM, Kiran RP, Lavery IC. The impact of anastomotic leak and intra-
the effect of AL on long-term mortality was mediated by an increased abdominal abscess on cancer-related outcomes after resection for colorectal
rate of distant recurrence. cancer: a case control study. Dis Colon Rectum. 2009;52:380386.
It remains to be established which mechanisms promote dis- 9. Mirnezami A, Mirnezami R, Chandrakumaran K, et al. Increased local recur-
tant recurrence in patients with AL. To address this, the impact of rence and reduced survival from colorectal cancer following anastomotic leak:
AL on administration of AC in patients with stage III colonic cancer systematic review and meta-analysis. Ann Surg. 2011;253:890899.
was analyzed. AC increases overall survival by about 30% compared 10. den Dulk M, Marijnen CA, Collette L, et al. Multicentre analysis of oncolog-
ical and survival outcomes following anastomotic leakage after rectal cancer
with surgery alone30,31 and is thus important in these high-risk pa- surgery. Br J Surg. 2009;96:10661075.
tients. Severe inflammation due to peritonitis and septicemia after AL 11. Bertelsen CA, Andreasen AH, Jorgensen T, et al. Anastomotic leakage after
may contribute to the metastatic cascade explaining the association curative anterior resection for rectal cancer: short and long-term outcome.
between AL and distant recurrence in the present study.32 Further- Colorectal Dis. 2010;12:e76e81.
more, AL was strongly associated with cancelled administration of 12. De Marco MF, Janssen-Heijnen ML, van der Heijden LH, et al. Comorbidity
AC in the subset of patients with stage III colonic cancer. Similar find- and colorectal cancer according to subsite and stage: a population-based study.
Eur J Cancer. 2000;36:9599.
ings were reported in patients with AL after low anterior resection.33
13. Sarfati D, Hill S, Blakely T, et al. The effect of comorbidity on the use of
Recently, El Shayeb et al34 demonstrated that comorbidity and ad- adjuvant chemotherapy and survival from colon cancer: a retrospective cohort
vanced age were main reasons physicians did not recommend AC. study. BMC Cancer. 2009;9:116.
Here, the association between AL and administration of AC was 14. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic
consistent after adjustment for confounding variables including age comorbidity in longitudinal studies: development and validation. J Chronic Dis.
and comorbidity. The authors did further conclude that reasons for 1987;40:373383.
patient refusal were mostly unknown.34 Postoperative complications 15. Sundararajan V, Henderson T, Perry C, et al. New ICD-10 version of the
Charlson comorbidity index predicted in-hospital mortality. J Clin Epidemiol.
including AL could influence these patients attitude toward AC. 2004;57:12881294.
AL was also associated with a significant delay in initial admin- 16. Schmidt M, Jacobsen JB, Lash TL, et al. 25 year trends in first time hospitalisa-
istration of AC. The delay persisted after adjustment for confounding tion for acute myocardial infarction, subsequent short and long term mortality,
variables and postponed the initial administration to 8 weeks after and the prognostic impact of sex and comorbidity: a Danish nationwide cohort
primary surgery. When AC was initiated beyond day 55 after primary study. BMJ. 2012;344:e356.
surgery, there was no significant reduction in mortality compared 17. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure
time. Control Clin Trials. 1996;17:343346.
with patients who did not receive AC. Biagi et al35 recently addressed
18. Jorgren F, Johansson R, Damber L, et al. Anastomotic leakage after surgery for
the consequences of delayed AC, reporting a 14% decrease of both rectal cancer: a risk factor for local recurrence, distant metastasis and reduced
overall and disease-free survival for every 4 weeks postponement cancer-specific survival? Colorectal Dis. 2011;13:272283.
of initial administration. The present data do not support late onset 19. Katoh H, Yamashita K, Wang G, et al. Anastomotic leakage contributes to the
administration of AC in patients with AL after curative surgery for risk for systemic recurrence in stage II colorectal cancer. J Gastrointest Surg.
stage III colonic cancer. 2011;15:120129.
20. Lin JK, Yueh TC, Chang SC, et al. The influence of fecal diversion and
CONCLUSIONS anastomotic leakage on survival after resection of rectal cancer. J Gastrointest
Surg. 2011;15:22512261.
This study demonstrated increased mortality rates in patients 21. Akyol AM, McGregor JR, Galloway DJ, et al. Anastomotic leaks in colorectal
surviving the first critical phase after AL. A robust association cancer surgery: a risk factor for recurrence? Int J Colorectal Dis. 1991;6:
between AL and distant recurrence was observed suggesting that 179183.

8 | www.annalsofsurgery.com 
C 2013 Lippincott Williams & Wilkins

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 00, Number 00, 2013 Anastomotic Leak and Distant Recurrence

22. Docherty JG, McGregor JR, Akyol AM, et al. Comparison of manu- 29. Iversen LH, Norgaard M, Jacobsen J, et al. The impact of comorbidity on
ally constructed and stapled anastomoses in colorectal surgery. West of survival of Danish colorectal cancer patients from 1995 to 2006a population-
Scotland and Highland Anastomosis Study Group. Ann Surg. 1995;221: based cohort study. Dis Colon Rectum. 2009;52:7178.
176184. 30. Moertel CG, Fleming TR, Macdonald JS, et al. Levamisole and fluorouracil
23. Fujita S, Teramoto T, Watanabe M, et al. Anastomotic leakage after colorectal for adjuvant therapy of resected colon carcinoma. N Engl J Med. 1990;322:
cancer surgery: a risk factor for recurrence and poor prognosis. Jpn J Clin 352358.
Oncol. 1993;23:299302. 31. International Multicentre Pooled Analysis of Colon Cancer Trials (IMPACT)
24. Law WL, Choi HK, Lee YM, et al. Anastomotic leakage is associated with investigators. Efficacy of adjuvant fluorouracil and folinic acid in colon cancer.
poor long-term outcome in patients after curative colorectal resection for ma- Lancet. 1995;345:939944.
lignancy. J Gastrointest Surg. 2007;11:815. 32. Wu Y, Zhou BP. Inflammation: a driving force speeds cancer metastasis. Cell
25. Abulafi AM, Williams NS. Local recurrence of colorectal cancer: the problem, Cycle. 2009;8:32673273.
mechanisms, management and adjuvant therapy. Br J Surg. 1994;81:719. 33. Smith JD, Paty PB, Guillem JG, et al. Anastomotic leak is not associated with
26. Chuang D, Paddison JS, Booth RJ, et al. Differential production of cytokines oncologic outcome in patients undergoing low anterior resection for rectal
following colorectal surgery. ANZ J Surg. 2006;76:821824. cancer. Ann Surg. 2012;256:10341038.
27. Lee YS, Choi I, Ning Y, et al. Interleukin-8 and its receptor CXCR2 in the tu- 34. El Shayeb M, Scarfe A, Yasui Y, et al. Reasons physicians do not recom-
mour microenvironment promote colon cancer growth, progression and metas- mend and patients refuse adjuvant chemotherapy for stage III colon cancer: a
tasis. Br J Cancer. 2012;106:18331841. population based chart review. BMC Res Notes. 2012;5:269.
28. Kapiteijn E, Marijnen CA, Nagtegaal ID, et al. Preoperative radiotherapy com- 35. Biagi JJ, Raphael MJ, Mackillop WJ, et al. Association between time to initia-
bined with total mesorectal excision for resectable rectal cancer. N Engl J Med. tion of adjuvant chemotherapy and survival in colorectal cancer: a systematic
2001;345:638646. review and meta-analysis. JAMA. 2011;305:23352342.


C 2013 Lippincott Williams & Wilkins www.annalsofsurgery.com | 9

Copyright 2013 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.