Postoperative non‐steroidal anti‐inflammatory drugs and anastomotic leakage after gastrointestinal anastomoses: Systematic review and meta‐analysis

Abstract Aim Non‐steroidal anti‐inflammatory drugs (NSAIDs) are commonly used to control postoperative pain; however, their postoperative use has been associated with anastomotic leakage after gastrointestinal surgery. This systematic review and meta‐analysis aimed to determine the correlation between the use of NSAIDs and anastomotic leakage. Methods We conducted a comprehensive electronic literature search up to August 2018 to identify studies comparing anastomotic leakage in patients with and without postoperative NSAID use following gastrointestinal surgery. We then carried out a meta‐analysis using random‐effects models to calculate odds ratios (OR) with 95% confidence intervals (CI). Results Twenty‐four studies were included in this meta‐analysis, including a total of 31 877 patients. Meta‐analysis showed a significant association between NSAID use and anastomotic leakage (OR 1.73; 95% CI = 1.31‐2.29, P < .0001). Subgroup analyses showed that non‐selective NSAIDs, but not selective cyclooxygenase‐2 inhibitors, were significantly associated with anastomotic leakage. However there was no significant subgroup difference between selective and non‐selective NSAIDs. Conclusion Results of this meta‐analysis indicate that postoperative NSAID use is associated with anastomotic leakage following gastrointestinal surgeries. Caution is warranted when using NSAIDs for postoperative analgesic control in patients with gastrointestinal anastomoses.


| INTRODUC TI ON
Anastomotic leakage has long been a concern among gastrointestinal surgeons. Its occurrence not only causes postoperative morbidity and mortality, but also lengthens hospital stay and increases hospital costs. 1,2 Importantly, anastomotic leakage worsens oncological outcomes in patients with resectable and curable malignancies, leading to poorer disease-free survival, overall survival, and functional outcome. 3,4 Multiple factors contribute to anastomotic leakage, and its incidence varies depending on the location of the anastomosis.
Esophageal anastomoses have the highest incidence of leakage, and gastric anastomoses the lowest incidence, whereas the incidence of colorectal anastomotic leakage differs among publications and anastomosis sites, ranging from 1% to 20%. 5 The early recovery after surgery protocol has been proposed to reduce postoperative stress. The protocol aims to promote postoperative recovery, reduce hospital stay and, most importantly, reduce postoperative complications, especially cardiovascular and pulmonary complications. 6 Non-steroidal anti-inflammatory drugs (NSAIDs) play a major part in this protocol as a means of postoperative pain control. However, application of the early recovery after surgery protocol has been associated with an increased incidence of anastomotic leakage, 7 and it has been suggested that NSAIDs may be a causative factor in impaired anastomotic healing.
Many potential mechanisms have been proposed to explain how postoperative NSAID use may cause anastomotic leakage. NSAIDs decreased protective prostaglandins, and inhibited mucosal cyclooxygenase (COX)-1, intestinal epithelial cell migration, and mucosal restitution in animal models 8 which, in turn, reduced anastomotic tensile strength and collagen deposition causing delayed anastomotic healing. [9][10][11] Previous reviews have examined the correlation between postoperative NSAID use and anastomotic leakage, but most have considered colorectal anastomoses only. 7,12 However, we suggest that the mechanisms shown in animal models may be applicable to all gastrointestinal anastomoses. Furthermore, it is also possible that selective COX-2 inhibitors may be safer than non-selective NSAIDs in terms of preventing anastomotic leakage based on the above-mentioned mechanism.
The primary objective of this systematic review and meta-analysis was to determine the effect of postoperative NSAID use on gastrointestinal anastomotic leakage, regardless of the site of anastomosis.
The secondary objective was to compare the anastomotic leakage risk between non-selective NSAIDs and selective COX-2 inhibitors.

| Search strategy
We conducted a literature search of the Medline, PubMed, Cochrane Library, clini caltr ial.gov, and Web of Science databases up to August 2018. The search was limited to English language and human studies.
The search terms used were "Anastomosis or anastomotic leakage" AND "NSAIDs" [MesH term]. Additional articles were retrieved by manually searching the reference lists of the included studies and other reviews.

| Selection criteria
Studies were included if they met the following criteria: (i) study with anastomosis of the gastrointestinal tract; (ii) study compared postoperative NSAID use with non-use; and (iii) investigations reported anastomotic leakage. Case reports or reports with incomplete data were excluded.

| Data extraction
The studies were independently and critically assessed by two authors using a standard protocol and discrepancies were resolved by consensus. Extracted data included study design, number of institutes, definition of anastomotic leakage, operative diagnosis, location of anastomosis, urgency of surgery, type of NSAIDs, sample size, and numbers of anastomotic leakage per group.

| Quality assessment
Qualities of the included studies were assessed using the Jadad score 13 and the Newcastle-Ottawa scale (NOS) 14 for randomized controlled trials (RCT) and observational studies, respectively.
Studies were considered to be high quality if they had a Jadad score ≥3 or NOS ≥7.

| Data synthesis and meta-analysis
Meta-analysis was done by computing the OR from the original data using the Cochrane-Mantel-Haenszel method, with 95% CI. P ≤ .05 was considered significant in all analyses. Data analysis was carried out using Review Manager (RevMan) v5.3 software (Cochrane Collaboration) and a random-effect model was used for graphical presentation. Statistical heterogeneity was quantified using I 2 statistics and Cochrane Q tests. I 2 values >50% indicated heterogeneity. 15 In the presence of heterogeneity, we conducted subgroup and meta-regression analyses to determine if the inter-study variation could be explained by certain co-variates, including type of study, NSAID class, NSAID administration, urgency of surgery, location of anastomosis, and operative diagnosis. Sensitivity analyses were done to assess the impact of individual potential confounding variables. Publication bias was assessed visually by funnel plot, and asymmetry was assessed formally by rank correlation test (Begg's test). 16 Publication bias was analyzed using WINPEPI software. 17

| Study selection
The initial systematic search identified 430 studies and an additional search for reviews identified a further five studies. After adjusting for duplicates and critical assessment, a total of six RCT 18-23 and 18 observational studies [24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41] were included in the meta-analysis. The PRISMA flow diagram of the detailed literature search and selection process is shown in Figure 1. Of 27 full-text article reviews, three were excluded from the quantitative analysis because we could not extract the original data from two, and the other study compared multimodal interventions in which NSAIDs were also distributed to the control group.

TA B L E 1 Characteristics of included studies to determine the correlation between the use of NSAIDs and anastomotic leakage
Author Abbreviations: AL, anastomotic leakage; IBD, inflammatory bowel disease; NR, not reported; NSAIDs, non-steroidal anti-inflammatory drugs; PCA, patient controlled analgesia; POD, postoperative day; RCT, randomised controlled trial.
a Quality assessment for RCT and observational studies using Jadad score and Newcastle-Ottawa scale (NOS) for randomised controlled trials (RCTs) and observational studies, respectively.

TA B L E 1 (Continued)
F I G U R E 2 Forrest plot of meta-analysis between randomized controlled trials (RCT) and observational studies. NSAIDs, non-steroidal anti-inflammatory drugs  (Figure 2).

| Protocol-based versus non-systematic NSAIDs use
To investigate the effect of NSAID dose on anastomotic leakage,  Table 1.
The protocol-based group had a significantly higher anastomotic leakage rate compared with non-users (pooled OR 4.67, 95% CI 2.84-7.67, P < .001) without evidence of heterogeneity (I 2 = 5%, Cochrane Q test P = .40), whereas the non-systematic group also had a significantly increased risk for anastomotic leakage compared with non-users (pooled OR 1.38, 95% CI 1.06-0.181, P = .02), but with evidence of heterogeneity (I 2 = 82%, Cochrane Q test P < .001). However, there was a statistically significant subgroup difference between the protocol-based group and the nonsystematic group (P < .001) (Figure 4).

| Non-selective NSAIDs versus selective COX-2 inhibitors
Among all the included studies, we extracted information on nonselective NSAID use from 15 (n = 4110) and on selective COX-2 inhibitor use from eight (n = 1063) studies. Subgroup analysis showed that patients who received postoperative non-selective NSAIDs had a significantly higher rate of anastomotic leakage than patients who did not receive NSAIDs (pooled OR 1.80, 95% CI 1.12-2.91,

F I G U R E 4
Forrest plot of meta-analysis between protocol-based non-steroidal anti-inflammatory drugs (NSAIDs) use and non-systematic NSAIDs use P = .02) with evidence of heterogeneity (I 2 = 85%, Cochrane Q test P < .00001). In contrast, the anastomotic leakage rate in patients taking selective COX-2 inhibitors was not significantly higher than in those not taking NSAIDs (pooled OR = 1.67, 95% CI 0.90-3.13, P = .11), with evidence of heterogeneity (I 2 = 67%, Cochrane Q test P = .004). However, comparison between users of non-selective and selective NSAIDs showed no significant subgroup difference (P = .85) ( Figure 5).

| Colorectal anastomoses versus other gastrointestinal anastomoses
We with evidence of heterogeneity (I 2 = 72%, Cochrane Q test P = .002). There were no subgroup differences between the two groups of studies (P = .85) ( Figure S1).
Furthermore, separate stratified and meta-regression analyses showed no significant differences in the OR of anastomotic leakage rates after postoperative NSAID use in relation to the type of study, NSAID class, urgency of surgery, or operative diagnosis (Table 2).

F I G U R E 5 Forrest plot of meta-analysis between non-selective non-steroidal anti-inflammatory drugs (NSAIDs) and selective COX-2 NSAIDs
Sensitivity analyses were carried out to assess the impact of low-quality studies ( Table 1). Exclusion of the two low-quality studies did not affect the significance of the results (pooled OR 1.61, 95% CI 1.22-2.11, P < .001).  In animal models, adverse effects of NSAIDs were found in both small intestine and colon resulting in increased anastomotic leakage rate. 8,9,11,42,44 In human studies, consistent results were also reported regardless of anastomotic site; however, the majority were colorectal anastomoses. In our study, studies with colorectal anastomoses had significantly increased anastomotic leakage rates when perioperative NSAIDs were used (pooled OR 1.80, 95% CI  33,39 of which majority of anastomosis site is not colorectal, were also included in our meta-analysis. These results support our hypothesis that NSAIDs were associated with increased anastomotic leakage in all gastrointestinal anastomoses.

| D ISCUSS I ON
The present study had several limitations. First, our conclusions were mainly based on observational studies; however, subgroup analysis showed no significant subgroup difference between RCT and observational studies, suggesting that this potential bias was not significant. Second, there was statistical heterogeneity, and the included observational studies were clinically heterogenous in terms of patient characteristics, indications for surgery, and location of anastomoses. Although stratified and meta-regression analyses showed no significant differences, heterogeneity decreased the validity of the results. Third, most of the included studies (17/24) only considered colorectal anastomoses, and the implication of the results for all gastrointestinal anastomoses might not be completely accurate.
In conclusion, postoperative NSAID use appears to be associated with an increased incidence of anastomotic leakage following gastrointestinal surgery. Selective COX-2 inhibitors might be safer than non-selective NSAIDs, although the results were inconclusive. Caution is warranted when using NSAIDs for postoperative analgesic control in patients with gastrointestinal anastomoses.

ACK N OWLED G EM ENT
We thank Susan Furness, PhD, from Edanz Group (www.edanz editi ng.com/ac) for editing a draft of this manuscript.

D I SCLOS U R E
Conflicts of Interest: Authors declare no conflicts of interest for this article.