Risk factors for postoperative complications after elective ileocolic resection for Crohn’s disease: a retrospective study

Introduction

Crohn’s disease (CD) is a chronic inflammatory condition that follows a course of alternating periods of relapse and remission and can affect any part of the gastrointestinal tract [1]. In 1998, the approval of infliximab, the first monoclonal antibody against tumor necrosis factor (TNF)-α, changed the modern conservative treatment of inflammatory bowel diseases (IBD) [2]. Despite immense progress in the era of biologic agents regarding the medical treatment of CD, the need for intestinal resection remains high: at 1, 5 and 10 years after diagnosis the risk for intestinal surgery is 16%, 33% and 47%, respectively [3].

Current evidence suggests high rates for postoperative complications (POC); up to 37% of CD patients who undergo surgery will experience POC [4]. Thus, several studies have tried to highlight the risk factors related to POC and research has been mainly focused on the preoperative immunosuppressive medication as a potential predictor of POC. However, the majority of current data regarding the association between medication and POC are extracted from observational, retrospective studies with high heterogeneity regarding POC classification (overall, anastomotic-related, septic vs. non-septic, minor vs. major), medication dosage, interval between last administration and surgery, and the examination of confounding factors such as nutritional status or coadministration of other immunosuppressive regimens [5]. Another confounding factor that should be taken into account is surgery in the emergency/urgent setting. These patients cannot be properly optimized preoperatively in order to minimize the risk for postoperative morbidity [6].

While the findings for certain categories of medication, such as steroids or immunomodulators (IM), are highly consistent in the literature [7,8], data regarding biologic agents remain conflicting [9]. The purpose of this study was to evaluate the impact of anti-TNF therapy on the postoperative course and to identify other risk factors for overall and intra-abdominal septic POC in CD patients who undergo elective ileocolic resection (ICR).

Patients and methods

Results

Baseline characteristics

The demographic, clinical and surgical characteristics of 153 patients are presented in Table 1. Our population had a median age of 36 years (IQR 28-47 years), and included 86 male patients (56.2%) and 61 active smokers (39.9%). Single ICR was performed in most cases (86.9%), whereas 20 patients underwent concomitant resection and/or strictureplasty. A laparoscopic approach was preferred in 19 patients (12.4%), whereas EEA was the most common anastomotic configuration (n=128, 83.7%). Severe anemia (hemoglobin [Hb] <8 g/dL) was not recorded in any patient and the majority of patients (86.5%) had serum albumin >3.0 g/dL. According to the ASA score, the study population had no severe comorbidities (ASA I/II, 93.5%). Most patients had no perianal involvement (77.1%) or previous resections (84.3%).

Table 1 Overall baseline characteristics (n=153)

Preoperative medical treatment was described according to the kind of medication and therapeutic regimen; anti-TNF was the most commonly used category (39.9%), as either monotherapy (25.5%) or combination treatment (14.4%). Vedolizumab (3.3%) and ustekinumab (2.0%) were used only as monotherapy regimens. Steroid therapy was taken by 23.5% of patients. Homogeneity among patients who were under anti-TNF therapy or not was analyzed (Table 2). Statistically significant differences between the 2 groups were observed for median age (34 vs. 38.5 years, P=0.016), age at diagnosis according to Montreal classification (P=0.010), and steroid use, which was less frequent in anti-TNF patients (13.1% vs. 30.4%, P=0.013).

Table 2 Comparison of baseline characteristic between patients under anti-TNF therapy and not

POC

Overall, 35 patients (22.9%) experienced POC (Table 3). Anastomotic leakage was the most common POC (n=18, 11.8%) and IASC was recorded in 19 patients (12.4%). Postoperative ileus was the most frequent non-IASC complication (3.9%), while postoperative hemorrhage and unexplained fever were recorded in 4 patients (2.6%). Non-IASC septic complications included wound infections (2.0%), UTI (1.3%), and RTI (0.7%). In total, 22 patients required reintervention (14.4%); 17 with IASC (2 cases with anastomotic leak were treated conservatively), 4 with hemorrhage and 1 with postoperative bowel obstruction. No patient died during the 30-day postoperative period.

Table 3 Postoperative complications

Univariate analysis

The analysis of preoperative medications showed no statistically significant differences in the rates of overall POC and IASC (Fig. 1). Steroid therapy was associated with a higher incidence of overall POC (33.3%), although statistical significance was not reached (P=0.09). Patients with preoperative administration of anti-TNF agents had a non-significantly lower percentage of overall POC (21.3% vs. 23.9%, P=0.71) and a non-significantly higher percentage of IASC (13.1% vs. 12.0%, P=0.83). Similarly, the analysis according to the therapeutic regimen revealed no significant differences in overall and IASC rates (Fig. 2). Anti-TNF treatment was not associated with overall POC or IASC, either as monotherapy or in combination regimens. The administration of steroids, either as monotherapy or as a combination treatment with IM or anti-TNF, showed the highest incidences of overall POC, although the differences were statistically non-significant (31.6%, P=0.35; 33.3%, P=0.41; 37.5%, P=0.37, respectively). As for IASC, combined steroid and anti-TNF therapy had the highest rate (25.0%, P=0.23). Patients under a combined anti-TNF and IM regimen showed insignificantly lower percentages of complications (7.1%, P>0.99). The number of patients receiving vedolizumab (n=5) or ustekinumab (n=3) was too small for analysis.

Figure 1 Univariate analysis for the correlation of overall (A) and intra-abdominal septic (B) postoperative complications with preoperative medication

IM, immunomodulators; TNF, tumor necrosis factor

Figure 2 Univariate analysis for the correlation of overall (A) and intra-abdominal septic (B) postoperative complications with preoperative therapeutic regimen. All P-values are vs. none. Overall P=0.75 (A) and 0.80 (B)

IM, immunomodulators; TNF, tumor necrosis factor

Demographic, clinical and surgical variables were used in the univariate analysis (Table 4). Three variables were associated with significantly higher rates of overall POC and IASC: perianal disease (37.1%, P=0.022 and 25.7%, P=0.016, respectively), a history of previous intestinal resection (54.2%, P<0.001 and 37.5%, P<0.001, respectively), and disease duration (P=0.012 and 0.029, respectively). Statistical significance for the “disease duration” variable was noted after the post hoc analysis for patients with >10 years of diagnosed CD compared to those with <5 years (36.4% vs. 16.1%, P=0.012 for overall POC and 21.8% vs. 6.5%, P=0.016 for IASC).

Table 4 Univariate analysis for detection of risk factors for postoperative complications

Multivariate analysis

Preoperative medication variables and others with a P-value <0.25 in the univariate analysis were inserted into a binary logistic regression model (Table 5). Previous resections were the only significant independent predictor for overall POC and IASC: OR 3.90, 95%CI 1.38-11.06; P=0.010, and OR 4.56, 95%CI 1.51-13.77; P=0.007, respectively. In contrast to the univariate analysis, perianal disease and duration of CD >10 years were not independently associated with a higher OR for POC. Preoperative medication was also non-significantly correlated with POC.

Table 5 Multivariate analysis for possible risk factors

Discussion

In this retrospective study we tried to detect and analyze risk factors for POC in a Greek cohort of CD patients operated in a high-volume tertiary center, paying special attention to the patients’ preoperative immunosuppressive therapy. In total, 22.9% of our patients experienced some kind of POC; this finding is comparable with complication rates reported in other studies [15,18,19]. Anastomosis-related complications, usually presenting as free peritoneal leakage or intra-abdominal abscess leading to intra-abdominal sepsis, represent the category of POC causing most concern among colorectal-IBD surgeons. The recognition of patients at a higher risk for IASC is crucial and affects the surgical strategy; anastomotic defunction with a temporary stoma is the preferred option for these patients. In our series the incidence of IASC was 12.4%, also comparable with the current literature [17,20,21].

According to this study’s findings, anti-TNF treatment within 3 months before surgery was not associated with a higher risk for overall POC or IASC. The great concern regarding the safety of anti-TNF agents is that their immense immunosuppressive effect could not only affect the patient’s defense against infections, but theoretically impair the healing cascade as well, leading to septic and anastomotic complications. Clinical data from a large number of non-randomized, mainly retrospective studies are conflicting. Some are in accordance with our findings, concluding that preoperative anti-TNFs significantly correlate with POC [22-24]. A multicenter retrospective study with 231 patients found that overall complications, anastomotic leak and IASC rates were not increased in patients under infliximab or adalimumab [17]. Furthermore, a nationwide prospective study from the REMIND group assessed the impact of anti-TNF agents by evaluating the preoperative serum trough levels; it was found that POC were not increased, regardless of the interval from last dosing or the serum levels before surgery [15]. Similarly, recently published data from another multicenter prospective cohort (PUCCINI trial) with 955 IBD patients, although not separating patients with CD from those with ulcerative colitis (UC), showed that anti-TNF medication within 3 months preoperatively was not an independent predictor for infectious complications, even when there were detectable serum drug levels at the time of surgery [25].

On the other hand, many studies have documented an increased risk for postoperative morbidity, especially septic POC [26,27]. Recent retrospective data from Leuven and Amsterdam, concerning 538 patients who had undergone ileocecal resection, reported an OR of 3.30 (95%CI 1.09-9.99, P=0.035) for anastomotic leakage for patients receiving anti-TNF medication [19]. In 2018, a prospective study of a nationwide cohort of 592 patients from France was published, in which anti-TNF within 3 months before ICR was an independent risk factor for overall POC [28]. Similarly, Kulaylat et al analyzed a large database of 2364 patients, adjusting for confounders such as emergency surgery, and concluded that the overall and septic POC are significantly higher in anti-TNF exposed patients [18].

A critical question is whether a surgeon should postpone an elective surgery according to the timing of the last drug infusion. Since evidence for the safety of anti-TNFs is conflicting, our practice was to perform elective ICRs when a period of at least 4 weeks had elapsed since the last administration of anti-TNF; we could assume that this time interval seems to be safe. A Danish nationwide cohort with 2293 CD patients concluded that anti-TNF administration within 14 days before surgery was not associated with a higher risk for POC compared to longer intervals [22]. A meta-analysis conducted for the comparison of different time intervals (within 4, 8 and 12 weeks) also found no significant differences [29]. Lau et al, in contrast to the PUCCINI trial and the prospective cohort from the REMIND group, found that anti-TNF serum levels at surgery of ≥3 μg/mL correlated with increased overall postoperative morbidity and infectious complications, while levels of ≥8 μg/mL correlated with overall postoperative morbidity and readmissions [30]. Nevertheless, in this study a major limitation was that data regarding the interval between last infusion and surgery were not obtained and therefore the optimal timing for last dosing could not be estimated. Overall, it seems that the preoperative serum levels of the drug do not help discriminate patients at risk and a safe time interval from last administration to surgery cannot be defined.

To date, since the conduction of well-designed randomized controlled trials is an unmet need because of ethical considerations, the literature provides clinicians with conflicting evidence about the perioperative safety of anti-TNF agents. This heterogeneity is reflected in the discrepancy of results from several meta-analyses that used the same pool of data. An interesting meta-analysis in 2013, by El-Hussuna et al, found a 1.6-fold increase in relative risk for anastomotic complications only when the analysis included studies with a low risk of bias [31]. Yang et al found an increased risk for overall and infectious complications in their meta-analysis and stated that the most appropriate definition with lower heterogeneity among studies would be infectious and non-infectious POC [32]. As for IASC, Huang in 2014 found that steroid use, hypoalbuminemia, previous surgery and preoperative abscess were predictors, but anti-TNF treatment was not [33]. In a 2018 meta-analysis with separate analysis for every category of POC described in previous studies (overall, major/minor, septic/non-septic, surgical/medical) anti-TNFs were not associated with any kind of POC [34], whereas a more recent study in 2019, which included 7115 patients, concluded that there is a higher risk for infectious POC [35]. The European Crohn’s and Colitis Organization published in 2020 the updated guidelines for CD, stating that current evidence does not support any correlation of anti-TNF agents with POC after abdominal surgery [36].

In the present study, none of the other medications had a significant impact on POC. However, we detected a trend for increased overall POC in patients under steroid treatment, with borderline significance (P=0.09). Nevertheless, this finding is biased by the fact that patients under high doses of steroids for a long period and who needed ICR were protected by a diverting ileostomy and were thus excluded from this cohort. The dose-related association of steroids with POC is definite [7,37]. Moreover, we tried to evaluate the effect of combined immunosuppressive therapy, but statistical significance was not reached for any possible combination of anti-TNFs, steroids and immunomodulators. However, there was a trend for increased overall POC and IASC when steroids were added to any other class of medication (Fig. 2). In accordance with our results, the REMIND Group study documented no association of anti-TNFs in combination with other immunosuppressants [15]. A recent large retrospective study found that double immunosuppression of steroids and biologicals was an independent risk factor for IASC [38]. Furthermore, McKenna’s cohort in 2018 recorded an increased risk for IASC only with triple immunosuppression [16]. The 2 last studies included anti-TNFs as well as vedolizumab and ustekinumab as the “biological part” of the combined regimen; therefore, they could not conclude if a specific biological class, when combined, increased the risk. In our cohort, no patient was under a triple immunosuppressive regimen.

A history of previous resections was the only independent risk factor for overall POC and IASC in this cohort. The correlation of repeated resections with postoperative morbidity is not well documented in the literature. A multicenter retrospective study was the first to report this variable as an independent risk factor, but only for anastomotic leakage and not for IASC [17]. A subsequent single-center study by Johnston et al tried to confirm this finding and, although statistical significance was reached in univariate analysis, multivariate analysis was not conducted because of the limited data [39]. Our results are in accordance with 2 large, recently published, retrospective studies by the same team in the Mayo Clinic that documented previous resections as an independent risk factor for IASC [16,38]. Interestingly, this finding was confirmed by Huang’s abovementioned meta-analysis in 2015, although none of the included studies reported previous surgery as a risk factor for IASC [33]. We therefore believe that our findings increase the evidence for the negative effect of previous resections on the postoperative course of CD patients. We speculate that this association could be attributed to the technical difficulties encountered in a previously operated abdomen, which could compromise the blood supply in bowel cut edges, in combination with the aggressive disease pattern in patients who require reoperations; both could impair anastomotic healing.

In the present cohort, patients were operated in the elective setting and were therefore adequately optimized with proper anemia and malnutrition management. The mean Hb concentration was 12.7 g/dL, with no difference between the “anti-TNF” and “no anti-TNF” groups, whereas no patient had severe anemia. Similarly, 86.5% of patients had serum albumin >3 g/dL and 89.1% of them BMI >18.5 kg/m2. Thus, anemia and malnutrition were not confirmed as established risk factors [6,20,21]. Anastomotic technique is a controversial risk factor in the literature; current evidence, although weak, supports the superiority of a stapled SSA configuration over the EEA in terms of lower risk for anastomotic leak [36]. In our study, EEA was the prevailing configuration, all anastomoses were hand-sewn and differences between EEA and SSA were not confirmed for either overall POC or IASC. Last, the penetrating-B3 phenotype as a predictor was also not confirmed [20,33]; patients with severe intra-abdominal sepsis and abscesses were either drained percutaneously or diverted.

This study has certain limitations that should be highlighted. First, the retrospective, single-center design and the consequent selection bias is a flaw. Furthermore, our population was not completely homogenous between anti-TNF and non-anti-TNF patients in terms of age, disease duration and steroid medication. Second, adjustment for certain confounders, such as disease activity, was not feasible. Although inclusion criteria were selected in order to include CD patients with mild disease, non-septic and under proper optimization strategy, disease severity data with validated scores might be useful. Last, although a (probably arbitrary) minimum interval of 4 weeks before elective surgery was kept as a rule of thumb, precise data could not be obtained regarding the last administration of anti-TNFs in order to include it as a variable in our analysis. However, we believe that this study adds more evidence in the evolving problem of POC in CD patients.

In conclusion, in the present study, previous resection was an independent risk factor for overall POC and IASC after elective ileocolectomy, adding more evidence about this variable to the existing literature. Preoperative medications, focusing especially on patients exposed to anti-TNF drugs, were not associated with postoperative morbidity, either as monotherapy regimens or as immunosuppressive combination treatments. Thus, we believe that a surgeon should take into account all possible risk factors for POC and should try to optimize the patient, rather than postpone an elective resection considering only the anti-TNF therapy status.