Long-term and Perioperative Corticosteroids in Anastomotic Leakage: A Prospective Study of 259 Left-Sided Colorectal Anastomoses

Objective To determine the risk factors for symptomatic anastomotic leakage (AL) after colorectal resection.

Design Review of records of patients who participated in the Analysis of Predictive Parameters for Evident Anastomotic Leakage study.

Setting Eight health centers.

Patients Two hundred fifty-nine patients who underwent left-sided colorectal anastomoses.

Intervention Corticosteroids taken as long-term medication for underlying disease or perioperatively for the prevention of postoperative pulmonary complications.

Main Outcome Measures Prospective evaluations for risk factors for symptomatic AL.

Results In 23% of patients, a defunctioning stoma was constructed. The incidence of AL was 7.3%. The clinical course of patients with AL showed that in 21% of leaks, the drain indicated leakage; in the remaining patients, computed tomography or laparotomy resulted equally often in the detection of AL. In 50% of patients with AL, a Hartmann operation was needed. The incidence of AL was significantly higher in patients with pulmonary comorbidity (22.6% leakage), patients taking corticosteroids as long-term medication (50% leakage), and patients taking corticosteroids perioperatively (19% leakage). Perioperative corticosteroids were prescribed in 8% of patients for the prevention of postoperative pulmonary complications.

Conclusions We found a significantly increased incidence of AL in patients treated with long-term corticosteroids and perioperative corticosteroids for pulmonary comorbidity. Therefore, we recommend that in this patient category, anastomoses should be protected by a diverting stoma or a Hartmann procedure should be considered to avoid AL.

Trial Registration trialregister.nl Identifier: NTR1258

Anastomotic leakage (AL) following colorectal resection is a feared complication. The reported incidence of AL is estimated to be between 3% and 19%,^1-5 and when it occurs, the mortality is between 10% and 20%.^4,6-8 There are several basic requirements concerning the construction of a colorectal anastomosis: blood flow should be adequate, the anastomosis should be free of tension, and the abdomen should be free of infectious disease and have minimal contamination. Nevertheless, the problem of AL has not diminished over the years.⁹ Although accurate prediction of risk is difficult, certain factors are known to contribute to the risk for AL. There is an inverse relationship between the height of the anastomosis from the anal verge and the incidence of AL, with extraperitoneal anastomoses carrying the highest risk.^10-15 For these high-risk anastomoses, studies have demonstrated the important reduction in mortality and morbidity that can be achieved by the construction of a diverting stoma.^4,16,17 Much research has been done to determine additional risk factors for AL. Reported risk factors include male sex,^12,13,16 smoking,¹⁸ radiotherapy,^1,12,19 blood transfusion,^1,15,16,20 obesity,^5,13 and atherosclerosis.²¹ Although corticosteroids are known to impair wound healing, their influence on the healing of colorectal anastomoses is unclear and studies have reported conflicting results.^22-25

The aim of this prospective study was to determine risk factors for clinical AL in patients undergoing left-sided colorectal anastomoses, together with a description of diagnostic and therapeutic approaches in patients having developed AL.

Patients included in this study all participated in the Analysis of Predictive Parameters for Evident Anastomotic Leakage study. In our prospective study, all patients undergoing left-sided colorectal resection with construction of an anastomosis were given an intra-abdominal drain. During the first 5 postoperative days, drain fluid was collected for analysis in search of a predictive parameter for AL. Results of drain fluid analysis will be reported separately. Eight centers participated in this study. The study, registered in the Netherlands Trial Register (http://www.trialregister.nl/trialreg/index.asp, study No. NTR1258), was approved by the medical ethical committees of the participating centers in accordance with the ethical standards of the Declaration of Helsinki of 1975, and all patients gave informed consent. Participating centers included patients consecutively treated between January 2007 and December 2009; centers stopped including patients earlier if their target inclusion was fulfilled.

Surgical procedures included left hemicolectomy, sigmoidectomy, high anterior resection or partial meserectal excision, low anterior resection or total mesorectal excision, and subtotal colectomy with ileorectal anastomosis. Oncologic resections as well as resections for inflammatory disease were also included. Emergency operations and colostomy reversals were excluded.

The surgical procedure was left to the surgeon's discretion. All patients received preoperative antibiotic prophylaxis and an intra-abdominal drain, but guidelines concerning bowel preparation differed by center. Patients were operated on by laparotomy or laparoscopy and the anastomosis was stapled or hand sewn, either end to end, end to side, side to end, or side to side. A diverting stoma was constructed according to the surgeon's preference.

Symptomatic AL, the end point of our analysis, was defined as clinically apparent leakage (eg, gas, pus, or fecal discharge from the pelvic drain), apparent AL during reoperation, or extravazation of endoluminally administrated water-soluble contrast on computed tomography. Radiologic examination was performed only when there was clinical suspicion of AL. All postoperative fistulas communicating with the surgical anastomosis were classified as leaks. Postoperative abscesses were classified as AL if there was extravasation of enteric contrast on an imaging study, significant perianastomotic air, or communication with the anastomosis noted after radiologic drainage. AsymptomaticAL was not considered since routine contrast enema was not performed postoperatively.

High anterior resection was defined as a colorectal resection with an intraperitoneal anastomosis. Low anterior resection was defined as a colorectal anastomosis with an extraperitoneal anastomosis.

Patients were followed up from their preoperative admission to the ward until the first postoperative follow-up at the outpatient clinic. Demographic data of the patients, operative details, postoperative events, and follow-up data were obtained through a standardized case record form and entered into a database. We used the International Classification of Diseases, Tenth Revision codes to define pulmonary and cardiac comorbidities. The diseases included in pulmonary comorbidity are chronic lower respiratory diseases (J40-J47), lung diseases due to external agents (J60-J70), other respiratory diseases principally affecting the interstitium (J80-J84), other diseases of pleura (J90-J94), and other diseases of the respiratory system (J95-J99). The diseases included in cardiac comorbidity are chronic rheumatic heart diseases (I05-I09), ischemic heart diseases (I20-I25), pulmonary heart disease and diseases of pulmonary circulation (I26-I28), and other forms of heart disease (I29-I52). Use of corticosteroids was subdivided in patients with long-term corticosteroid use for underlying pathology (herein called long-term corticosteroids) and patients having a corticosteroid scheme prescribed by the anesthesiologist or lung specialist for the reduction of postoperative pulmonary complications starting 5 days prior to the surgical intervention (herein called perioperative corticosteroids). In case of AL, the postoperative day of diagnosis was noted along with the clinical manifestation of AL, the diagnostic tool for leak detection, and the treatment. Postoperative mortality included all patients who died within 30 days after an operation either at the hospital or after discharge.

Categorical data are presented as numbers with percentages, and numerical data are presented as medians with interquartile ranges. Univariate analysis of the differences between patients having a protective ileostomy or not was performed using a χ² test for categorical data and a Mann-Whitney U test for numerical data.

Anastomotic leakage binary response (yes or no) was analyzed using a logistic regression model, reporting the odds ratios (ORs) together with 95% confidence intervals (CIs). The univariate and multivariate analyses were conducted with the following covariates: age (categorized at median = 65 years), sex, body mass index (calculated as weight in kilograms divided by height in meters squared) (categorized at 30), American Society of Anesthesiologists classification (3 or 4 vs 1 or 2), smoking (yes or no), cardiac comorbidity (yes or no), pulmonary comorbidity (yes or no), corticosteroids (long-term, perioperatively, or none), radiotherapy (yes or no), type of resection (low anterior resection, high anterior resection, sigmoidectomy, left hemicolectomy, or subtotal colectomy), height of anastomosis (>7 cm or <7 cm), anastomosis construction (stapled or hand sewn), configuration (end to end, end to side, side to end, or side to side), and protective ileostomy (yes or no). In the multivariate analysis, the initial model contained all the covariates with the grouped type of resection (low anterior resection vs others). A backward elimination procedure was applied to remove nonsignificant covariates with P ≥ .1.

A total of 259 patients underwent colorectal resection with left-sided anastomosis. Nineteen patients (7.3%) developed clinical AL. Mean (SD) age was 64.6 (0.75) years; 144 patients (56%) were men and 115 (44%) were women. Thirty-five patients (13.6%) were treated for inflammatory diseases; 220 (84.9%), for malignancy; and 4 (1.5%), for ischemic colitis. Sixty patients (23%) underwent preoperative radiotherapy, and 89 (34%) had anastomoses situated below the peritoneal reflection. In 60 patients (23%), a defunctioning stoma was constructed.

Nineteen patients (7.3%) developed clinical AL. The median postoperative day of diagnosis of AL was 6 (range, 4-10 days). Mortality was significantly increased in patients with AL, compared with patients without AL (15.8% vs 2.5%; P = .02). Four patients (21.1%) had a diverting stoma at the time of diagnosis. In 3 patients, AL manifested with a pelvic or intra-abdominal abscess, 7 had peritonitis, and 9 had signs of sepsis. In 8 patients, AL was diagnosed through computed tomography, diagnosis was made during laparotomy in 7 patients, and 4 patients had fecal discharge from the drain confirming the diagnosis. In 3 patients, treatment consisted of antibiotics, and 2 patients had drainage and irrigation of an abscess. The remaining patients were reoperated on: 4 patients had fecal diversion, 2 had reanastomosis, and the remaining 8 had the anastomosis converted into end colostomy. Patients with a diverting ileostomy did not have a clinically lighter presentation of AL nor less invasive treatment of the AL.

Table 1 shows the result of the univariate analysis for the risk for clinical AL. Sex, body mass index, American Society of Anesthesiologists classification, smoking, preoperative radiotherapy, height of the anastomosis, type of resection, and construction of a defunctioning stoma were not associated with an increased risk for AL. Two factors were significant in the univariate analysis: pulmonary comorbidity and use of corticosteroids. The incidence of AL was significantly higher in patients taking corticosteroids as long-term medication (50% AL, P = .002) and those taking corticosteroids perioperatively (19% AL, P = .001), compared with patients not taking corticosteroids (5.2% AL). In patients known to have pulmonary comorbidity, the incidence of AL was 22.6%; this was 5.3% in patients without pulmonary comorbidity (P = .002). Multivariate analysis was performed, showing a significant result for perioperative corticosteroids (OR, 26.98; standard error, 30.71; P = .004; 95% CI, 2.89-251.10) but a wide CI because of a relatively small quantity of cases; therefore, we decided to perform a multivariate analysis with long-term corticosteroids and perioperative steroids combined. From that final multivariate model, it can be concluded that taking corticosteroids increased the risk for AL by more than 7 times (OR, 7.52; standard error, 4.47; P = .001; 95% CI, 2.35-24.08). Patients with low anastomosis had a nearly 3 times greater risk for AL, compared with patients with high anastomosis (OR, 2.98; standard error, 1.65; P = .049; 95% CI, 1.01-8.83).

There was no difference in the risk for AL in patients taking corticosteroids between high and low anastomoses.

The characteristics of patients with and without a defunctioning ileostomy are demonstrated in Table 2. Significantly more patients with preoperative radiotherapy and low anastomoses or low anterior resections had a diverting stoma. A diverting stoma was equally constructed in patients receiving corticosteroids or not receiving corticosteroids.

Risk assessment for AL is important because it could lead to better management of high-risk patients. For extraperitoneal anastomoses, which often involve extensive surgical resection and neoadjuvant radiotherapy, it is known that the risk for AL is increased but that the construction of a diverting stoma reduces morbidity and mortality in cases of AL.^4,16,17 The benefit of a diverting stoma outweighs the morbidity it involves because the important complications of clinical AL can be avoided in high-risk groups. However, risk assessment and knowledge of risk factors are essential in deciding on a diverting stoma.

In this prospective study, the incidence of clinical AL was 7.5% among 259 left-sided colorectal anastomoses, and the mortality rate in patients with AL was 15.8%. These high incidences are in line with percentages of AL and associated mortality reported in literature^1,4,5,12 and again emphasize the importance of this complication in colorectal surgery. The clinical course of patients with AL showed that most patients had signs of sepsis or peritonitis. In 21% of leaks, the drain indicated leakage; in the remaining patients, computed tomography or laparotomy resulted equally often in the detection of AL. In 50% of patients with AL, a Hartmann operation was needed.

In 26% of patients in this series, a diverting stoma was constructed, significantly more often in high-risk patients who underwent radiotherapy and had low anastomoses after a total mesorectal excision surgery. In univariate analysis, low anastomoses did not have an increased risk for AL; however, in the multivariate analysis, low anastomoses did increase the risk for AL (P = .049). Having a nonsignificant effect in a univariate analysis and a significant effect in a multivariate analysis is because of an interaction effect among factors. First, there is an interaction between the height of an anastomosis and a protective ileostomy, with 80% of protective ileostomies in low anastomoses. Second, there is also an interaction with steroids; more than 90% of patients with stoma do not take steroids. Therefore, when looking at the effect of the height of anastomosis in patients not taking steroids in the multivariate analysis, the risk of leakage will be higher in low anastomoses.

In patients taking corticosteroids (long-term use of corticosteroids or perioperative use of corticosteroids), the risk for AL was significantly increased. It is known that corticosteroids impair wound healing by decreasing activation and infiltration of inflammatory cells. These inflammatory cells, macrophages and polymorph leucocytes, are essential in the first phase of wound healing.²⁶ Additionally, corticosteroids inhibit the expression of growth factors and matrix proteins such as collagen synthesis.²⁶ Known complications of glucocorticoids include gastrointestinal bleeding, peptic ulcer perforation, and sigmoid diverticular perforation,^27,28 as well as postoperative complications such as wound infection and wound dehiscence.^29-31

Despite these effects on wound healing in general, conflicting results are found in literature concerning the effect of corticosteroids on the healing of colorectal anastomoses. Most experimental studies have shown less breaking strength and collagen concentrations in the anastomoses of animals treated with steroids,^32-37 although some have found equal healing.^38,39 On the contrary, few clinical studies have found corticosteroids to be a risk factor for AL.^22,23,40,41 Among these is 1 prospective study of patients with colorectal carcinoma; other studies are retrospective, only included patients with Crohn disease, or looked at postoperative complications in general and not specifically AL. However, most clinical studies have failed to demonstrate a significant relationship between corticosteroids and impaired wound healing after colorectal surgery.^{14,24,25,42-47} Unfortunately, most studies were retrospective, had broad inclusion criteria, and often only included patients with Crohn disease. In a substantial part of these studies, postoperative complications in general were taken as outcome measures and AL was not specifically addressed.

To our knowledge, this is the first study to search for risk factors for AL that not only included long-term use of corticosteroids but also separately focused on corticosteroids prescribed perioperatively for the reduction of postoperative pulmonary complications. In a systematic review, Smetana et al⁴⁸ showed that chronic obstructive pulmonary disease is an important independent risk factor for postoperative pulmonary complications, as are specific surgical interventions including abdominal surgery and operations lasting longer than 2.5 hours. Therefore, the Dutch guidelines for the prevention of postoperative pulmonary complications recommend perioperative corticosteroids for patients having newly diagnosed, unstable, or severe chronic obstructive pulmonary disease, as well as undergoing extensive surgical interventions.⁴⁹ In these patients, it is advised to start 5 days prior to surgery administration of prednisone, 30 mg daily, which can be ended or gradually reduced 2 to 3 days postoperatively.

In our study, perioperative corticosteroids were prescribed to 8% of patients, and to our knowledge, this is the first study to demonstrate a significant relationship between patients undergoing perioperative corticosteroid treatment and clinical AL. However, perioperative use of steroids is not considered a substantial risk factor for AL according to surgeons because these patients did not receive a diverting stoma more often than patients who did not take corticosteroids.

In conclusion, our results show that the construction of a diverting stoma in patients with extraperitoneal anastomoses, who have often undergone preoperative radiotherapy and extensive total mesorectal excision surgery, is effective in reducing the risk for AL. However, in this prospective study, we did find a significantly increased incidence of AL in patients taking long-term corticosteroids and perioperative corticosteroids for pulmonary comorbidity. Therefore, we recommend that in this patient category, anastomoses should be protected by a diverting stoma or, in the case of chronic corticosteroids, a Hartmann procedure should also be considered to avoid the morbidity and strongly increased mortality associated with AL.

Correspondence: Juliette C. Slieker, MD, Room Ee-173, Laboratory of Experimental Surgery, Erasmus University Medical Centre, Postbus 2040, 3000 CA Rotterdam, the Netherlands (juliette_slieker@hotmail.com).

Accepted for Publication: September 24, 2011.

Published Online: January 16, 2012. doi:10.1001 /archsurg.2011.1690

Author Contributions:Study concept and design: Slieker, Komen, Jeekel, and Lange. Acquisition of data: Slieker, Komen, Karsten, Willemsen, and Lange. Analysis and interpretation of data: Slieker, Komen, Mannaerts, Murawska, Jeekel, and Lange. Drafting of the manuscript: Slieker, Jeekel, and Lange. Critical revision of the manuscript for important intellectual content: Slieker, Komen, Mannaerts, Karsten, Willemsen, Murawska, Jeekel, and Lange. Statistical analysis: Slieker, Murawska, and Lange. Obtained funding: Jeekel and Lange. Administrative, technical, and material support: Komen, Mannaerts, and Lange. Study supervision: Komen, Mannaerts, Jeekel, and Lange.

Funding/Support: This work was supported by a grant from Stichting Technische Wetenschappen, Utrecht, the Netherlands.