Meta-analysis: peri-operative anti-TNFα treatment and post-operative complications in patients with inflammatory bowel disease

Authors

  • N. Narula,

    1. Division of Gastroenterology, Department of Medicine, Farncombe Family Digestive Health Research Institute, McMaster University, Hamilton, ON, Canada
    Search for more papers by this author
  • D. Charleton,

    1. Division of Gastroenterology, Department of Medicine, Farncombe Family Digestive Health Research Institute, McMaster University, Hamilton, ON, Canada
    Search for more papers by this author
  • J. K. Marshall

    Corresponding author
    • Division of Gastroenterology, Department of Medicine, Farncombe Family Digestive Health Research Institute, McMaster University, Hamilton, ON, Canada
    Search for more papers by this author

  • As part of AP&T's peer-review process, a technical check of this meta-analysis was performed by Mr M. Siddiqui.

Correspondence to:

Dr J. K. Marshall, Division of Gastroenterology (2F59), McMaster University Medical Centre, 1280 Main Street West, Hamilton, ON, L8S 4K1, Canada.

E-mail: marshllj@mcmaster.ca

Summary

Background

The impact of peri-operative use of TNFα antagonists on post-operative complications such as infection and wound healing is controversial.

Aim

To conduct a systematic review and meta-analysis to assess the impact of peri-operative use of TNFα antagonists on post-operative complications such as infection and wound healing in patients with inflammatory bowel disease (IBD).

Methods

A literature search identified studies that investigated post-operative outcomes in patients with IBD using TNFα antagonists. The primary outcome was the rate of post-operative infectious complications. Secondary outcomes included the rates of non-infectious complications and total complications. Odds ratios (OR) with 95% confidence intervals (CI) are reported.

Results

Overall, 18 studies with 4659 participants were eligible for inclusion. Patients with IBD using preoperative anti-TNFα therapies had significant increases in post-operative infectious [OR 1.56 (95% CI, 1.09–2.24)], non-infectious [OR 1.57 (95% CI, 1.14–2.17)] and total complications [OR 1.73 (95% CI, 1.23–2.43)]. Studies limited to patients with Crohn's disease demonstrated a statistically significant increase in infectious (OR 1.93, 95% CI 1.28–2.89) and total (OR 2.19, 95% CI 1.69–2.84) complications, and a trend towards increase in non-infectious complications (OR 1.73, 95% CI 0.94–3.17). Studies of patients with ulcerative colitis did not demonstrate significant increases in infectious (OR 1.39, 95% CI 0.56–3.45), non-infectious (OR 1.40, 95% CI 0.68–2.85), or total complications (OR 1.10, 95% CI 0.81–1.47).

Conclusion

Anti-TNFα therapies appear to increase the risk of post-operative complications. The increase in risk is small, and may well reflect residual confounding rather than a true biological effect. Nevertheless, physicians should exercise caution when continuing biological therapies during the peri-operative period.

Introduction

Inflammatory bowel diseases (IBD), including Crohn's disease (CD) and ulcerative colitis (UC), affect approximately 1 in 200 individuals.[1] Despite optimal medical management, up to 30% of UC patients and 75% of CD patients will require surgery for refractory disease or its complications.[2, 3] Furthermore, many people also require surgery for conditions unrelated to their IBD.

Inhibitors of tumour necrosis factor alpha (TNF-α) such as infliximab, adalimumab and certolizumab have revolutionised the management of IBD. In various jurisdictions they are indicated for induction and maintenance of remission of luminal IBD and fistulizing CD refractory to conventional therapy.[4] Despite this, many patients using these therapies still require surgery either to manage their IBD or for unrelated medical conditions.[2] The impact of TNFα inhibitors on surgical outcomes is controversial, and observational studies evaluating this association have yielded conflicting results. Accordingly, it is not clear whether these agents need to be discontinued prior to surgery.

TNFα plays a role in angiogenesis and collagen synthesis and inhibition of these pathways may impair wound healing in patients after surgery.[5] It has been shown in a murine model that addition of TNFα to wounds promotes healing, and inhibition of TNFα decreases wound strength.[6] Furthermore, surgery represents a complex stress on the body that leads to widespread reduction in cell-mediated immunity.[7] Preoperative treatment with TNFα antagonists and other immunosuppressive therapies may, in turn, increase post-operative infections and other complications.

A systematic review and meta-analysis were performed to identify observational studies and clinical trials that explored peri-operative anti-TNFα therapy to determine their impact on the risk of short-term post-operative complications in patients with IBD.

Methods

Study selection

A systematic literature search was conducted to identify studies that investigated the association of peri-operative anti-TNFα with post-operative complications in patients with IBD. We identified sources from the MEDLINE, Embase and PubMed databases from the years 1950 to December 2012. There were no language restrictions. Keywords used were (infliximab, adalimumab, or anti-TNF) and (CD, UC, or IBD) and (post-operative or surgical complications). In addition, the reference lists of any studies meeting inclusion criteria were reviewed manually to identify additional relevant publications. Furthermore, an electronic search of the abstracts available online from annual congresses of the American Gastroenterological Association (2009–2012), United European Gastroenterology Federation (2006–2012) and Canadian Association of Gastroenterology (2007–2012) was also performed.

Inclusion/exclusion criteria

For inclusion in the review, studies were required to meet the following criteria: (i) observational design (prospective or retrospective cohort or case control) or treatment design (randomised or nonrandomised); (ii) association between preoperative treatment with any anti-TNFα agent and post-operative complications after any surgical intervention reported; (iii) similar outcomes reported in a control group; and (iv) complications either reported as infectious or non-infectious, or listed individually to allow classification by the reviewer. Where studies did not provide sufficient information, authors were contacted to obtain additional data.

Outcomes of interest

The primary outcome was the rate of subjects experiencing infectious complications within 30 days of surgery among those receiving vs. not receiving TNFα inhibitors. Secondary outcomes included the rates of non-infectious complications and overall complications. Table S1 reports definitions of infectious and non-infectious complications for each study included in the meta-analysis.

Data extraction and quality assessment

Data extraction was carried out independently by two investigators (NN and DC) with discrepancies resolved by consensus in consultation with the senior author (JKM). The quality of study reports was evaluated using the Newcastle–Ottawa scale (NOS), a tool which allows for assessment of nonrandomized studies for the purpose of quality appraisal in meta-analyses.[8] The highest score is 9. We deemed studies with a score of 7 or higher as high quality, consistent with several other meta-analyses.[9-11]

Statistical analysis

Meta-analysis was conducted by combining the odds ratios of individual studies into a pooled odds ratio using a fixed effects model. We tested for heterogeneity using the chi-squared test and the I2 test. The I2 test describes the percentage of variability in effect estimates that is due to heterogeneity rather than chance, wherein a I2 test greater than 50% suggests significant heterogeneity.[12] If significant heterogeneity existed, we used a random effects model, as this provides a more conservative estimate than a fixed effects model. Analyses were performed with RevMan 5.1 (Review Manager (RevMan) [Computer program] Version 5.1. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2011).

Results

Search results

The literature search identified 406 citations, of which 372 were excluded on review of the title and abstract. A further five studies were excluded after careful review of the full text; one examined peri-operative infliximab for fistula closure,[13] one did not include an unexposed control group,[14] one was an abstract with insufficient data for extraction,[15] one only looked at post-operative complications of post-operative infliximab,[16] and one had a duration of follow-up that exceeded our inclusion criteria, without reporting short-term data that could be used in our meta-analysis.[17] For some studies where only partial information for our outcomes of interest was disclosed, the authors were contacted and asked to provide the relevant information. The study by Indar et al. looked at peri-operative immunosuppression as a whole.[18] The authors responded to our inquiry to determine the outcomes of patients on anti-TNFα therapies peri-operatively compared to those not using anti-TNFα medications and provided relevant data that are included in the meta-analysis. Two abstracts were excluded based on insufficient data in the abstracts and lack of response from the authors, who were contacted for more information.[19, 20] The abstract by Syed et al. looked at preoperative anti-TNF-α use in patients with CD.[21] The authors responded to our inquiry for detailed information and provided us with detailed outcomes and methodology regarding their study. Their study was subsequently published.[22] A total of ten studies reported on patients with CD, three of which were excluded based on a low NOS score.[18, 23, 24] Thirteen studies were found on patients with UC, five of which were excluded based on a low NOS score.[25-29] Four studies reported a mixed cohort of IBD patients, one of which was excluded due to a low NOS score.[30] Overall, 18 studies with 4659 participants were eligible for meta-analysis.[22, 31-47] Infectious complication rates could be calculated for 16 of these studies, non-infectious complications for 14, and total complication rates for 15 studies.

Characteristics of included studies

Characteristics of studies included are outlined in Tables S1 and S2. All studies were retrospective comparative cohort studies, except that by Kasparek et al., which was a prospective case-control study.[35] Most studies assessed patients, who had received infliximab, but four included patients on adalimumab, and two also enrolled patients on certolizumab.[33, 43, 46, 47] Individual study characteristics, including use of concomitant medications, duration of follow-up and NOS rating are presented in Table S2. The Marchal et al. study included two follow-up periods; only data from the 10-day follow-up period were extracted, as the second follow-up period fell outside our inclusion criteria.[31] Study endpoints are also presented in Table S1. Infliximab was given in the 12 weeks prior to surgery in ten studies,[31, 32, 34-36, 40, 42, 43, 45, 46] 8 weeks prior in four studies,[22, 33, 37, 41] and permitted beyond 12 weeks but largely given within 12 weeks in four studies.[38, 39, 44, 47] Most studies that included patients with adalimumab and certolizumab did not specify the interval from last dose to surgery, except for one where it was given in the 4 weeks preceding surgery.[46]

The type of surgical intervention studied was quite variable. Some studies included patients undergoing only one specific type of surgery,[32, 36, 39] whereas some others included patients undergoing any intestinal surgery. Details are provided in Table S2. As the definitions and types of surgery included were variable, subgroup analyses for specific operative interventions were not conducted.

Infectious post-operative complications

The 15 studies that reported infectious complications included 987 subjects who received anti-TNF therapies preoperatively and 2257 controls. The pooled OR for infectious post-operative complications was 1.56 (95% CI, 1.09–2.24, χ2 = 30.92, I2 = 55%), demonstrating a significant increase (Figure 1). The rate of infectious complications was 21.7% compared to 14.5% in the control group, for an absolute risk increase of 7.2% and a number needed to harm of 14.

Figure 1.

Forrest plot of studies that assessed post-operative infectious complications on peri-operative anti-TNFα therapies compared to controls.

Non-infectious post-operative complications

The 11 studies that reported post-operative non-infectious complications included 813 subjects who received anti-TNF therapies preoperatively and 1683 controls. The pooled OR for post-operative non-infectious complications was 1.57 (95% CI, 1.14–2.17, χ2 = 17.10, I2 = 42%), demonstrating a significant increase (Figure 2). The rate of non-infectious complications was 27.9% compared to 18.8% in the control group, for an absolute risk increase of 9.1% and a number needed to harm of 11.

Figure 2.

Forrest plot of studies that assessed post-operative non-infectious complications on peri-operative anti-TNFα therapies compared to controls.

Total post-operative complications

The thirteen studies that reported total post-operative complications included 1059 subjects who received anti-TNF therapies preoperatively and 2781 controls. The pooled OR for total post-operative complications was 1.73 (95% CI, 1.23–2.43, χ2 = 40.19, I2 = 70%), demonstrating a significant increase (Figure 3). The rate of total complications was 43.8% compared to 28.8% in the control group, for an absolute risk increase of 15% and a number needed to harm of 7.

Figure 3.

Forrest plot of studies that assessed post-operative total complications on peri-operative anti-TNFα therapies compared to controls.

Sensitivity analysis and publication bias

Sensitivity analysis revealed that the increase in infectious complications (OR 1.49, 95% CI 1.05–2.11) remained statistically significant after the study with the largest treatment effect (Mor et al.)[39] was excluded. The increase in non-infectious complications also persisted (OR 1.51, 95% CI 1.10–2.06) after removal of the study with the largest effect (Rizzo et al.).[46] Finally, the increase in total complications remained significant (OR 1.43, 95% CI 1.21–1.69) after exclusion of the study with the largest disparity between the treatment and control groups (Appau et al.).[32] The increase in infectious (OR 1.45 (95% CI 1.00–2.10) and non-infectious (OR 1.37 (95% CI 1.05–1.78) complications also persisted when the Appau et al. study was removed from analysis. When only studies limited to patients with CD were included, there remained a statistically significant increase in infectious (OR 1.93, 95% CI 1.28–2.89), and total (OR 2.19, 95% CI 1.69–2.84) complications and trend towards increase in non-infectious complications (OR 1.73, 95% CI 0.94–3.17). However, when the analysis was limited to studies of patients with UC, the pooled OR were not significant for infectious (OR 1.39, 95% CI 0.56–3.45), non-infectious (OR 1.40, 95% CI 0.68–2.85), or total complications (OR 1.10, 95% CI 0.81–1.47). No publication bias was detected in comparing infectious, non-infectious, and total complications, respectively, based on the symmetry of the funnel plots (Figures 4-6).

Figure 4.

Funnel plot of published studies reporting infections complications for detecting publication bias.

Figure 5.

Funnel plot of published studies reporting non-infections complications for detecting publication bias.

Figure 6.

Funnel plot of published studies reporting total complications for detecting publication bias.

Discussion

The results of this meta-analysis suggest that peri-operative treatment with anti-TNFα medications modestly increases the post-operative risk of infectious, non-infectious and total complications in patients with IBD. These conclusions are similar to those of other systematic reviews. A recent meta-analysis found a statistically significant increased risk of infectious complications (OR 1.50, 95% CI 1.08–2.08), and a nonsignificant trends towards increased non-infectious (OR 2.00, 95% CI 0.89–4.46) and overall complications (OR 1.72, 95% CI 0.93–3.19) among patients with CD undergoing abdominal surgery.[48] Another found no increased risk of infectious (OR 1.10, 95% 0.51–2.38), non-infectious (OR 1.10, 95% CI 0.76–1.59), or total complications (OR 1.09, 95% CI 0.87–1.37) among patients with UC treated preoperatively with infliximab.[9] Our meta-analysis is the first to include data from mixed cohorts of IBD patients.[45-47] We also include more recently published manuscripts[22, 36] and excluded lower quality studies.[18, 23-29]

There are several limitations to the current meta-analysis. Of all eligible studies, only one assessed outcomes prospectively.[35] Observational studies cannot control for all potential confounding factors, and this is even more challenging with retrospective designs. Only a few studies adjusted for concurrent immunosuppressive use and age. Other factors such as disease severity and the nature, indication and timing of surgery may also influence outcome and patients treated with anti-TNFα therapies are more likely to have severe disease. Notably, seven studies that reported increased risks of infectious complications also reported higher concurrent use of immunomodulators or corticosteroids in the group treated peri-operatively with infliximab.[31, 32, 37, 38, 45-47] Accordingly, increases in the rate of complications may reflect differences in morbidity rather than an effect of anti-TNFα therapy.

Most studies reported both infectious and non-infectious complications. Where it was not stated clearly whether these outcomes were mutually exclusive, we contacted the authors for clarification. Waterman et al. provided detailed information about subjects who experienced more than one complication,[47] and our analysis accounts for this. However, Appau et al. did not respond to our inquiry[32] and we analysed their data as reported. Sensitivity analysis showed that exclusion of these data did not affect the findings of our meta-analysis. There is also variability among studies in definitions of infectious and non-infectious complications. We accepted the original authors’ classification but, where this was not provided, classified outcomes based on the data provided. The classification of different outcomes is detailed in Table S1.

Most studies did not report clearly the interval between the last dose of anti-TNFα therapy and surgery. Several studies included patients within a peri-operative infliximab group if they had received the last dose within the 12 weeks of surgery.[22, 31-37, 40-43, 45, 46] The study by Waterman et al. found no difference in outcomes whether anti-TNFα therapy was given in the 14 days preceding surgery vs. 15–30 days before surgery or 31–180 days before surgery.[47] It is noteworthy that most studies have evaluated the effect of infliximab, and it is not clear whether surgical outcomes are similar after treatment with other inhibitors of TNFα.

Large, prospective, randomised, placebo-controlled, clinical trials will be needed to clarify the true benefits and risks of continuing anti-TNFα therapy in the peri-operative period. This meta-analysis suggests that recent exposure to anti-TNFα therapies in patients with IBD is associated with small increases in post-operative infectious, non-infectious and total post-operative complications. However, the small increase in apparent risk may well reflect residual confounding rather than true cause and effect. Interruption of therapy can reactivate IBD and/or promote the formation of neutralising antibodies. Furthermore, Waterman et al. suggest that increasing the time interval between the last dose of biological therapy and surgery does not lower the risk of post-operative infection.[16] Accordingly, our data are insufficient to justify pre-operative interruption of biological therapy. Clinicians managing patients with IBD must continue to make individual decisions that are cognizant of these trade-offs in potential risk.

Authorship

Guarantor of the article: Neeraj Narula.

Author contributions: Neeraj Narula – study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; statistical analysis. Dan Charleton – acquisition of data. John K Marshall – study supervision. All authors approved the final version of the manuscript.

Acknowledgements

The study investigators thank Dr Jonathan Efron (Phoenix, AZ), Dr Mark Flasar (Baltimore, MD) and Dr. Matti Waterman (Haifa, Israel) for providing additional data for use in the meta-analysis.

Declaration of personal and funding interests: None.

Ancillary