Nonsteroidal anti-inflammatory drugs (NSAIDs) exert a competitive inhibitory action on the enzymatic isoforms of cyclo-oxygenase by preventing the pro-nociceptive inflammatory action of prostaglandins through the dorsal horn of the spinal cord causing postoperative hyperalgesia. The enzymatic inhibition on COX1 and COX2 accounts also for their main side effects due to a homeostatic dearrangement mainly on the gastrointestinal and renal systems [24]. To the local anti-inflammatory action at the abdominal level in the prevention of peritoneal adhesions and in reduction of post-operative ileus [6, 25], new evidence suggests direct oncological effects both on chemoprevention of polypoid or dysplastic forms and on recurrent free-suvival, refuting a putative dominant role of the pro-inflammatory environment in tumor growth and progression [8].
The perioperative administration of NSAIDs for colorectal surgery represents a cornerstone for enhanced recovery after surgery programs. Since the ancillary descriptions from Dahl and Kehlet in 1991 [26], numerous studies have validated the beneficial aspects of NSAIDs-based analgesia in gastrointestinal surgery. In the concept of a multimodal and balanced strategy, the adoption of opioid-sparing strategies results in a reduction of nausea, vomiting and postoperative ileus with comparable pain control, rapid mobilization, reduction of postoperative complications and hospital stay [6]. In 2014, the French Society of Anesthesia and Rianimation (SFAR) and the French Society of Gastrointestinal Surgery (SFCD) jointly published recommendations for colorectal surgery where use of NSAIDs was probably recommended although doubts persisted about the risk of anastomotic fistulas [27]. Similarly, the 2018 Guidelines for Perioperative Care in Elective Colorectal Surgery and Enhanced Recovery After Surgery [6] recommended multimodal analgesia avoiding opioids with a moderate quality of evidence and strong racommendation grade.
Anastomotic leakages are the most feared and common complications in colorectal surgery with an incidence ranging from 5–19% [2] and several predisposing factors such as patients’conditions (age, comorbidities, nutritional status, alcohol, smoking, corticosteroid therapies), surgical indications (tumor, diverticulitis, chronic inflammatory bowel diseases, prophylactic polyposis surgery), the anastomotic level (ileo-colic, colo-colic, colorectal), previous adjuvant therapies as well as intraoperative factors such as operative time, blood loss, maintenance fluid therapy, thermal homeostasis and perioperative hypoxia [28].
However, a real shift from experimental model to clinical practice in order to clarify hypothetical pharmacological mechanisms contributing for the risk of anastomotic fistulas have not been demonstrated and still raises debate. Cahill et al [29], speculating about the effects of cyclo-oxigenase 2 inhibitors on colonic anastomosis in an animal experimental model, demonstated the administration of rofecoxib increased leak risk (p = 0.048) and, on tensiometry, bowel segments were weaker in both tensile ans bursting pressure (p = 0.043 and p = 0.019, respectively). Inan et al [30] similarly demonstrated a lower anastomotic strenght in bursting pressure in NSAID cohorts (Diclofenac vs control: 40.12 ± 6.72 mmHg vs 54.13 ± 8.11 mmHg) as far as a reduction in hydroxyproline concentration in the perianastomotic tissue (1.97 ± 0.17 µg / mg vs 2.16 ± 0.12 µg / mg tissue), suggesting an inhibitory effect of post-surgical reparative processes.
The inhibition of leukocyte chemotaxis, the reduced synthesis of vascular endothelial growth factor and a reduction collagen deposition would favor an unpaired postoperative tissue remodeling response which could justify an increased anastomotic dehiscence rate after colorectal surgery [31, 32]. Finally, by reducing prostaglandin synthesis, NSAIDs would interfere with the production of hyaluronic acid necessary for the formation and remodeling of collagen in the anastomosis healing. With an overall postoperative anastomotic leakage incidence of 4.69%, our systematic review failed to demonstrate the perioperative administration of NSAIDs could significanty increase the risk of anastomotic leakage after colorectal cancer (OR: 1.24, 95%CI: 0.93–1.66; p = 0.14). The results were consistent with random effects model and a large heretogeneity among articles (I2 = 83%, Tau2 = 0.20). Similarly, Burton et al [33], in a meta-analysis including six randomized controlled trials with the aim to assess the 30-day incidence of anastomotic dehiscence after colorectal surgery, did not found any significant difference between intervention and control arms, although heterogeneity between the trials and the lack of statistical power of the analysis emerged.
However, evidences from the literature are quite discordant, fragmentary and often misleading and too far from strong levels of evidence and degrees of recommendation. Most studies are flawed and may be describing pre-existing biases [34–36]. In this scenario, Huang et al [37] demonstrated a significant lower rate of anastomotic dehiscence in the no-NSAIDs group (OR:2.00, 95%: 1.48–2.71, p < 0.001) in a pooled mixed study among observational and randomized controlled studies; but, at the subgroup analysis, similar leakage rates between arms when only RCTs were considered (p = 0.17).
Bhangu et al [38], in a recent meta-analysis combining both human and experimental RCTs and observational studies, concluded anastomotic leak risk was doubled in intervention group (OR: 2.14; p < 0.001). But, there was a strong evidence of selection bias among elected studies and additional incosinstent definition protocols and outcomes. In reality. similar limitations are also found in our systematic review. In fact, in three studies it was not possible to derive any pharmacological administration protocol [16, 21, 23] and in four [16, 21–23] any drug dosage.
Although studies have shown conficting evidence on putative negative effects of non-selective NSAIDs over COX-2 selective ones on gastrointestinal anastomosis healing, our results struggled with current concepts. Including six studies (Non-selective NSAIDs patients: 2,468 vs Selective NSAIDs patients 1,106), the perioperative administration of non-selective drugs appeared to significantly reduce the incidence of anastomotic leakages with consistent trends for each study at the Forrest plot (OR: 0.51, 95% CI: 0.40–0.66, p < 0.001) and results appear to be in agreement with those reported by Saleh et al [15] that, in an observational cohort study, found no significant predictor role of NSAIDs on postoperative dehiscence (OR: 1.21, 95%CI: 0.52–2.84; p = 0.660). On the other hand, in a large report from the Surgical Care and Outcomes Assessment Program (SCOAP), Hakkarainen et al [39] demonstrated NSAIDs-based analgesia protocols were associated with a 24% increased risk. However, after risk adjustment, the same Authors recognised results could have been influenced by emergency colorectal surgery cases where there was almost a two-fold increase in incidence in the experimental group than the control one. Concerning with COX2 inhibitors, Holte et al [40] demonstrated celecoxib as an independent detrimental factor of anastomotic leak. Similar results were reported by Zittel et al [41].
Routine administration of NSAIDs for pain relief is a crucial aspect of post-operative recovery implementation programs. The different pharmacodynamic properties were evaluated in our analysis. As regards for diclofenac, a correlation close to statistical significance was found with an increased risk of anastomotic fistulas (OR: 1.99, 95% CI: 0.09–4.41; p = 0.09); while in the case of both the use of ibuprofen and ketorolac no noteworthy trends were identified (Ibuprofen: OR: 0.91, 95% CI: 0.39–2.14, p = 0.82; Ketorolac: OR: 1.14, 95% CI: 0.67–1.94, p = 0.63). Interestingly, the scatter plot analysis identified excellent pharmacological safety profiles at maximal dosage for all the molecules investigated, as no daily dosage significantly exposed to an augmented incidence of anastomotic fistulas greater than those reported in the literature (ranges:5–19%) [2]; but, trends towards a rapidly increasing risk at incremental dosages for ketorolac were reported. Our results somehow appear to struggle with what reported by Klein et al [12] that, in a case-control study based on 75 undergoing laparoscopic colorectal surgery patients, demonstrated a significant increase of adverse events in the diclofenac arm. Subendran et al reported same conclusions [42], showing that administration of keterolac doubled the risk of fistulas in colorectal cancer patients.
Limitations of the study
The study was conducted based upon a rigorous statistical methodology (PRISMA®statements, Cochrane Collaboration Tools, Quality assessment tools). Notwithstanding analyses showed population heterogeneity, current systematic review presents some limitations. First four studies [16, 21–23] lacked an exhaustive methodological report on protocols and dosages, although a rigorous comparison between experimental and control arms was declared. Second, restriction for selected cohorts of patients according to drug administration could lead to a type 2 statistical error (β > 20%) due to the weight percentage of two studies in both the ibuprofen [14, 18] and ketorolac [15, 20] analyses. In addition, conceptual and design limitations emerge from inadequate control of factors that could have contributed to anastomotic fistulas such as the location of colic versus rectal resection (except for one study [17]), the absence of details regarding bowel preparation, concurrent loop ileostomy procedures and rates of adherence to ERAS programs as far as expertises among surgical and anesthesiological teams.