Postoperative outcomes after preoperative ustekinumab exposure in patients with Crohn's disease: a systematic review and meta-analysis.

BACKGROUND: Recent studies have reported conflicting data on the risk of postoperative complications in patients with Crohn's disease (CD) exposed to ustekinumab (UST) preoperatively. We performed a systematic review and meta-analysis to better assess and quantify the risk of postoperative complications in this population undergoing major abdomino-pelvic surgery.
METHODS: We conducted a comprehensive search of multiple electronic databases and conference proceedings (earliest inception through October 2020) to identify studies that reported the postoperative outcomes in CD patients with preoperative UST exposure. We estimated and compared the pooled rates of postoperative complications, including intra-abdominal sepsis, surgical site infection, any infection, any adverse event, readmission, and reoperation.
RESULTS: A total of 5 studies were included in the analysis. The last dose of the drug was at most 16 weeks prior to abdomino-pelvic surgery. A total of 172 CD patients (61% female; median age 35 years) were included. The pooled rate of any complication and any infectious complications was 23.5% (95% confidence interval [CI] 16-33.1) and 20.2% (95%CI 10.3-35), respectively. There was no difference in rates of intra-abdominal sepsis between the UST group (7.2%, 95%CI 3-16.4) and the anti-tumor necrosis factor (TNF) group (11.9%, 95%CI 5.9-22.5; P = 0.4). The rates of readmission and reoperation in the UST group were 17.4% (95%CI 7.9-34) and 14.6% (95%CI 9-22.7), respectively.
CONCLUSIONS: The postoperative complication rate in patients with preoperative UST exposure may be similar to that for anti-TNF medication. Preoperative exposure to UST does influence postoperative complication risk. Future prospective studies are needed to validate these findings. Copyright: © Hellenic Society of Gastroenterology.

Introduction

Crohn’s disease (CD) is chronic inflammatory condition affecting the gastrointestinal tract [1,2]. There has been significant expansion of medical therapy in last 3 decades following the initial approval of tumor necrosis factor (TNF) inhibitors in 1998 [3,4]. There are now multiple monoclonal antibody mechanisms approved for CD, including TNF inhibitors (infliximab, adalimumab, certolizumab pegol), integrin inhibitors (natalizumab, vedolizumab), and an interleukin 12/23 inhibitor, ustekinumab (UST). UST has shown good efficacy and safety similar to anti-TNF agents and vedolizumab in the treatment of CD [5-8].

Despite the availability of multiple medical therapies, as many as 75% of CD patients will undergo major abdominal surgery in their lifetime [9]. Furthermore, approximately 30-50% of patients are on biologic therapy at the time of surgery [10]. There has been much interest in the impact of biologic medications on postoperative outcomes, given the theoretical risks of immune modulation throughout the perioperative period. Initially anti-TNFs were thought to increase the risk of anastomotic leaks, but later studies refuted these findings and the literature remains controversial [11,12]. Subsequently, retrospective studies assessing the impact of vedolizumab on postoperative outcomes suggested vedolizumab exposure was also a potential risk factor for adverse postoperative outcomes; however, further data has contradicted these early findings [13-15].

Most recently, studies have reported conflicting data on the risk of postoperative complications in patients receiving UST for CD [16-18]. Therefore, we performed a systematic review and meta-analysis to better assess and quantify the risk of postoperative complications in CD patients exposed to UST prior to abdomino-pelvic surgery.

Materials and methods

Search strategy

A comprehensive search of multiple databases was conducted from inception to October 2020. These databases were Ovid MEDLINE®, Scopus, Embase, Cochrane Central Register of Controlled trials, and Cochrane Database of Systematic Reviews. The keywords used were: “ustekinumab”, “complications”, “postoperative”, “infections”, “crohn’s disease”. The MOOSE checklist was followed and is provided in Supplementary Table 1 [19,20].

Study selection

In this meta-analysis, we included studies that evaluated postoperative complications in the setting of preoperative UST exposure in adult patients. We included studies irrespective of the sample size, geography and clinical setting, as long as data were provided for analysis.

Our exclusion criteria were pediatric population (age <18 years), and studies published in a language other than English.

Data abstraction and quality assessment

All data from individual studies on outcomes were abstracted on a standardized form by 2 authors (RG, ALL). In addition, 2 authors (RG, BPM) did the quality scoring independently. We contacted primary study authors via email for further information or clarification of data if needed.

In the case of data from the same author or institution(s), we contacted the study authors to ascertain the potential for data duplication. If overlap existed, we included overlapping studies if there were differences in outcomes or data reported. To limit potential bias and assess the influence of this approach, we excluded overlapping outcomes when possible, and performed sensitivity analyses excluding the potentially overlapping data.

We used the Newcastle-Ottawa scale for cohort studies to assess the quality of studies [21]. It consists of 8 questions; details are provided in Supplementary Table 2.

Outcomes

The following outcomes and definitions were included in CD patients who underwent abdomino-pelvic surgery with preoperative UST use: pooled rates of any postoperative complication, any infectious complication, intra-abdominal sepsis, readmission, and reoperation. The term “infectious complication” comprised any infections, including surgical site infections, urinary tract infections or non-surgical site infections such as pneumonia. The definition was consistent across all studies. Because of study variability and limited data, further subgrouping of infectious complications was not possible. Intra-abdominal sepsis was defined as the combination of anastomotic leak and intra-abdominal abscess, as previously reported [16]. For studies that reported comparator populations, these were included in the subgroup analysis if at least 3 studies presented similar comparison groups. All definitions were assigned by individual study authors.

Other data variables

We extracted baseline data on study type, last dose of drug prior to surgery, sex, age, tobacco use and preoperative steroid use. In addition, procedural data including ostomy creation and type of procedure, either laparoscopic or open, were also collected. All the data abstracted are shown in Tables 1 and 2 as population characteristics.

Table 1

Baseline and operative characteristics of studies included in the analysis

Table 2

Data on assessed outcomes of each study included in the analysis

Statistical analysis

Meta-analysis techniques were used to calculate the pooled estimates for each outcome, using the logit transformed proportion and random-effects model suggested by DerSimonian and Laird [22]. A continuity correction of 0.5 was added if the incidence of an outcome was zero before statistical analysis [23]. Heterogeneity was assessed between study-specific estimates using the Cochran Q statistical test for heterogeneity and the I2 statistics along with 95% confidence intervals (CI) [24,25]. The I2 value signifies what proportion of the dispersion is true vs. chance [26]: low, moderate, substantial, and considerable heterogeneity were suggested by values of <30%, 30-60%, 61-75%, and >75%, respectively [27]. Publication bias was ascertained qualitatively, by visual inspection of funnel plots, and quantitatively, by the Egger test [28]. A P value of <0.05 was used to define statistically significant difference. For outcomes and variables of interest, further meta-regression analyses were also performed to identify predictors. The choice of variables was based on data availability and all variables were considered for inclusion. Subgroup analysis was performed if the outcome of interest and the variables were reported in at least 3 studies. Comprehensive Meta-Analysis (CMA) software, version 3 (BioStat, Englewood, NJ) was used to perform all analysis.

Results

Search results and population characteristics

The initial search resulted in 94 studies. Sixty records were screened after removing duplicates and 11 full-length articles were assessed after initial screening. Six studies were excluded as the outcome of interest was not present, leaving 5 studies in the final analysis that reported postoperative outcomes in CD patients with preoperative UST use [16,18,29-31]. Fig. 1 shows the schematic diagram of study selection.

Figure 1

PRISMA flow diagram showing search strategy for meta-analysis

CD, Crohn’s disease; UST, ustekinumab

A total of 1422 patients were included; 189 (13.2%) received preoperative UST, 481 (33.8%) anti-TNF, 200 (14.0%) vedolizumab, and 552 (38.8%) no biologics. Sex was reported in 4 studies in the UST group and the majority of patients were female (61%). The median age was 35 years, with a general age range of 34.5-36.7 years. Two studies compared UST and anti-TNF, one study compared UST and no biologic, one study compared UST and vedolizumab, and one study compared UST, anti-TNF, vedolizumab and no biologic to each other for rates of intra-abdominal sepsis. Patients received UST within 12 weeks of surgery in 4 studies, and within 16 weeks in one study. The follow-up period was 30 days in 3 studies, up to 6 months in one, and 90 days in one study. Among patients with UST exposure, 36.4% of patients were on preoperative steroids and 18.8% were smokers. The surgical procedure ranged from total abdominal colectomy to strictureplasty, depending on the individual patient and operating surgeon. A majority (63.8%) of patients had ostomy creation and only 39.4% had laparoscopic procedures. The baseline and operative characteristics are described in Table 1.

Characteristics and quality of included studies

All the included studies were retrospective in nature. Three studies were multicenter and the other 2 were single-center studies. Based on the Newcastle-Ottawa scale, all studies were of high quality.

Meta-analysis outcomes

A total of 189 patients exposed to UST preoperatively were included in the analysis from 5 studies. Data on assessed outcomes are described in Table 2.

There were 3 studies reporting adverse events in the UST group. The pooled rate of any postoperative complication was 23.5% (95%CI 16-33.1), I2 = 0% (95%CI 0-34.4) (Fig. 2A). The pooled rate of any infectious complications was 20.2% (95%CI 10.3-35.9), I2 = 67 (95%CI 5.1-88.8) from 4 studies (Fig. 2B). Because of the study variability and limited data, further subgrouping of infectious complications was not possible.

Figure 2

Pooled rate of any postoperative complication (a) and any infectious complication (b) in patients with preoperative ustekinumab use

CI, confidence interval

The pooled rate of intra-abdominal sepsis in UST group was 7.2% (95%CI 3.0-16.4), I2 = 15 (95%CI 0-86.9) from 4 studies, whereas the pooled rate of intra-abdominal sepsis in the anti-TNF group was 11.9% (95%CI 5.9-22.5), I2 = 71 (95%CI 53.3-94.8) from 3 studies (Fig. 3). There was no significant difference between the 2 groups (P = 0.4). We also performed a second analysis including 3 studies that directly compared UST and anti-TNF. The odds ratio of intra-abdominal sepsis comparing preoperative UST use to anti-TNF use was 0.41 (95%CI 0.12-1.23), I2 = 14 (95%CI 0-91; P = 0.11) (Supplementary Fig. 1).

Figure 3

Pooled rate of postoperative intra-abdominal sepsis in patients who received preoperative anti- TNF and UST

TNF, tumor necrosis factor; UST, ustekinumab; CI, confidence interval

The pooled rate of readmission and reoperation in the UST group was 17.4% (95%CI 7.9-34), I2 = 72 (95%CI 20-90) (Fig. 4A) and 14.6% (95%CI 9-22.7), I2 = 12 (95%CI 0-86.8), (Fig. 4B) respectively, from 4 studies.

Figure 4

Pooled rates of readmission (A) and reoperation (B) in patients with preoperative ustekinumab use

CI, confidence interval

Meta-regression

Meta-regression was performed for any adverse outcomes, infectious complications, intra-abdominal sepsis and reoperation. The variables included were preoperative steroid use and stoma creation. Preoperative steroid use and ostomy creation were not significant predictors for any of the outcomes. The results of meta-regression with their coefficient and 95%CI are summarized in Supplementary Table 3.

Validation of meta-analysis results

Sensitivity analysis

Sensitivity analysis was performed by excluding one study at a time and analyzing its effect on the main summary estimate. On this analysis, no single study significantly affected the outcome or the heterogeneity.

To assess for potential data duplication from overlapping authors or institution(s), we contacted the study authors and determined that there was patient overlap in 3 studies included in the meta-analysis, but differences in outcomes were reported. One study [30] had only rates of intra-abdominal sepsis whereas another 2 [16,29] contributed to other outcomes. When the first overlapping study was excluded [29], the rates of infectious complications, readmissions and reoperations were 23.4% (95%CI 10.9-43.2), 15.6% (95%CI 4.2-43.4) and 11.4% (95%CI 4.3-26.9), respectively, and when the second overlapping study was excluded [16], the respective rates were 15.6% (95%CI 8.0-28.0), 13.5% (95%CI 7.5-23.1) and 10.9% (95%CI 4.5-24.6). There were no studies with overlap in the outcomes regarding overall complications.

Heterogeneity

We assessed the dispersion of the calculated rates using the I2 percentage values. The calculated I2 values are reported with the pooled results. There was low heterogeneity in overall complication, intra-abdominal sepsis and reoperation outcomes; however, in view of the wide 95%CIs, high heterogeneity cannot be excluded. Heterogeneity was substantial in the infectious complications and readmissions outcomes. This was probably due to baseline patient characteristics, as we were unable to do a subgroup analysis of patients on steroids and other CD medications preoperatively. In addition, the number of procedures and types of surgery also contributed to heterogeneity in the study population. We also acknowledge that I2 values have limited utility to detect heterogeneity when the number of studies is small.

Publication bias

Publication bias was not assessed as fewer than 10 studies were included in the analysis.

Discussion

Our study demonstrated that the surgical complication rate in patients with preoperative UST exposure is similar to that for anti-TNFs. Even though we were not able to directly compare UST with vedolizumab and other biologics, the reported rate of postoperative complications is comparable to that of other biologics, including vedolizumab. Preoperative exposure to ustekinumab does not seem to influence postoperative complication risk. Identifying risk factors for postoperative complications is of paramount importance to optimize surgical and disease-related outcomes. We are likely to see a higher number of patients with medically refractory disease who received UST preoperatively.

The current literature on postoperative complications and biologic use is controversial, limited by observational studies and significant heterogeneity. The reported rates of overall postoperative complications and infectious complications after preoperative anti-TNF exposure are 42.3% and 27.2%, respectively, and the respective rates after preoperative vedolizumab exposure are 30.4% and 22.4% [15,32]. Based on our study, we reported that preoperative UST exposure was associated with similar rates of any adverse event (23.5%, 95%CI 16-33.1), and any infectious complications (20.2%, 95%CI 10.3-35.9). Two systematic reviews did not find any significant risk of postoperative infection and complications in patients exposed to vedolizumab preoperatively, compared to anti-TNF and no biologic therapy [15,32]. On the other hand, the data on anti-TNF are much more conflicting, with some studies reporting higher rates of postoperative complications and others reporting no difference compared to no biologic therapy [12,33-36].

While the current study could only make limited comparator assessments, overall, the data on preoperative UST exposure in CD are reassuring. One study reported a higher rate of intra-abdominal sepsis in UST exposed patient compared to no biologic therapy [16]. In that study, there was significantly more use of preoperative immunomodulators in the UST group, while a greater number of patients underwent laparoscopic procedures and primary anastomoses in the no biologic group, which may have confounded the results. These results were not consistent in a large multicenter study that compared preoperative no biologic, anti-TNF, vedolizumab and UST exposure and did not report an increased risk of intra-abdominal sepsis in the UST-exposed group. On multivariate analysis, preoperative steroid use and combination immunosuppression with steroids remained an independent predictor of intra-abdominal sepsis, suggesting steroids as a likely confounding agent [29,37]. Similarly, Novello et al in a case-matched analysis, did not find a higher risk of postoperative complications (any postoperative complications, infectious complications, readmission and reoperation) in the UST group as compared to preoperative vedolizumab use [31]. Another multicenter study from Canada also did not show any higher risk of postoperative complications in patients who received UST, compared to preoperative anti-TNF use [18]. A similar trend is seen for vedolizumab, with earlier studies reporting a higher risk of adverse events and later studies with larger sample size reporting no greater risk of postoperative complications after preoperative vedolizumab use [13,15]. Taken together, these studies suggest the relative safety of UST in a perioperative setting.

The limitations of this study include the fact that it evaluated only small, tertiary-care referral center studies, potentially restricting the generalizability of the results. Studies were retrospective in nature, potentially contributing to selection bias and confounding. The time period of preoperative UST administration was not consistent throughout the study period. In addition, there was no standardization of the surgical procedure in the studies evaluated. UST patients were more likely to undergo ostomy creation rather than primary anastomosis, given their severe disease and complex phenotype. Small sample sizes and studies with overlapping cohorts further contributed to the low power and low impact of meta-regression results in our study. Because of limited data availability, further analysis with subgroups, such as the number of surgeries already performed, was not possible. Data limitations also did not allow for adequate assessment of confounders, comparator populations, or additional subgroup analysis. We also could not control for other factors such as nutrition, disease severity, prior biologic use, and type of surgery.

The strengths of this review include the systematic literature search with well-defined inclusion criteria, careful exclusion of redundant studies, inclusion of good quality studies, detailed extraction of data and rigorous evaluation of study quality. This is the first meta-analysis reporting the effect of preoperative UST use on postoperative complications in CD.

In conclusion, our meta-analysis demonstrates that the rate of postoperative complications in UST-exposed CD surgical patients may be similar to that of other biologics. Together, perioperative biologics may appear safe. Future prospective studies are needed to validate these findings and determine their influence on surgical decision-making.

What is already known:

Current data on the risk of postoperative complications in patients receiving ustekinumab for Crohn’s disease undergoing surgery are conflicting

What the new findings are:

In a meta-analysis of 189 patients exposed to ustekinumab preoperatively, the surgical complication rate was similar to that of anti-tumor necrosis factor

The reported rate of postoperative complications after preoperative ustekinumab exposure may be comparable to other biologics, including vedolizumab

Click here for additional data file.