Association between vedolizumab and risk of clostridium difficile infection in patients with ulcerative colitis: A systematic review and meta-analysis

Abstract

Background:

The medical treatment of ulcerative colitis (UC) includes the use of biological agents such as vedolizumab, a gut-selective alpha4beta7 (ɑ4β7) antagonist. The mechanism of action of vedolizumab involves interfering with leukocyte trafficking into the gut vasculature, which halts inflammation. Due to this mechanism of action, concerns have arisen regarding an increased risk of gut infections, specifically, clostridium difficile infection (CDI). The aim is to provide clarity regarding the association between the use of vedolizumab as a therapy for ulcerative colitis and the risk of developing CDI.

Methods:

A systematic literature review was conducted, starting with the scoping search, followed by backward snowballing parallel with keyword-based search to identify related articles. A quality assessment was conducted on the initially selected articles and excluded low-quality papers.

Results:

Pooled analyses indicated that there was no significant association between the use of vedolizumab and the risk of developing CDI (effect size = 0.03 [-0.02, 0.07]).

Conclusions:

Vedolizumab does not increase the risk of CDI in patients with UC. Further studies are needed to confirm these findings.

INTRODUCTION

Ulcerative colitis (UC) is a condition characterized by chronic inflammation of the colonic mucosa, primarily affecting the distal segment of the colon. This inflammatory bowel disease (IBD) has no known cause and can occur at any age. Common symptoms include recurrent episodes of bloody diarrhea and abdominal pain. The illness typically follows a pattern of flare-ups and periods of remission.[1]

Clostridium difficile infection (CDI) is a notable worry among individuals with UC as they face a higher susceptibility compared to the general population.[2] CDI is an enteric bacterial infection that tends to occur on the background of using antibiotics and can lead to severe, potentially life-threatening diarrhea in some patients. Previous studies have demonstrated that patients with IBD, both UC and Crohn’s disease (CD), are up to 4.8 times more likely to experience morbidity and mortality because of CDI compared with patients without IBD.[3,4]

Vedolizumab (VDZ) is a monoclonal antibody designed to target α4β7 integrin found specifically on activated gut-homing T-lymphocytes. By blocking its interaction with the mucosal address cell adhesion molecule 1 (MAdCAM-1) expressed on the endothelium of gastrointestinal tract blood vessels, it becomes a gut-targeted therapy that helps to control inflammation in the gastrointestinal tract without inducing systemic immunosuppression.[5,6] This targeted approach and the precise method of action of VDZ might help maintain a healthier balance of gut bacteria and immune responses, reducing the conditions that may foster and encourage the growth of CDI in UC patients. However, the risk of enteric infections and serious sequelae from lymphocyte homing blockade to the gastrointestinal tract remains unclear.

Given the established risks of mortality and morbidity associated with CDI in IBD patients and the importance of developing safer alternatives in managing those particular populations, this systematic review and meta-analysis examined the potential risk of developing CDI in UC patients after using VDZ.

MATERIALS AND METHODS

The PICO format was used to structure the research question framework to investigate the impact of VDZ intervention on the risk of CDI among patients aged 18 years and above with UC. The comparison group was none. The primary outcome measured was the risk of CDI, which was determined based on the presence of symptoms along with a positive test result using both polymerase chain reaction (PCR) and enzyme-linked immunosorbent assay (ELISA). This format was specifically chosen to refine the research question and abolish the hypothetical ambiguity of clinical issues.[7]

This article aims to contribute significantly to the understanding of developing C. diff in UC patients treated with VDZ. To achieve this, a systematic snowball strategy was employed concurrently with a keyword-based search method. The reason for this choice is the challenge of selecting appropriate keywords due to the limited availability of papers addressing C. diff as a complication of VDZ in UC patients only. We ran the snowball technique parallel with the classical search to verify no publications were missed. The systematic snowball technique involves gathering articles closely related to the subject of study, known as “the seeds’’, and then screening the bibliographies of these papers for any relevant articles. This process is repeated for each new article found until no further relevant articles are found, known as “backward snowballing (BSB)”.

To perform a high-quality systematic review, various search databases were considered. A scoping search was done to ensure the novelty of the work, (refer to Supplementary Data for more details). Ultimately, Google Scholar (GS) was chosen for the snowballing search based on a comparison between GS and other reliable sources of information, such as the Cochrane Database Systematic Review and The Journal of the American Medical Association (JAMA). The result indicated the adequacy of coverage and high sensitivity in finding relevant articles when using GS by itself.[8,9]

At the outset, three primary papers were selected as the seed set of papers to initiate the SB technique. Recognizing that starting sets have their limitations, we ensured that for those three papers addressing the area that we wanted to study, we used a backward snowballing technique that would widely cover the literature, followed by a keyword-based search to avoid overlooking any related articles.

Conducting backward snowballing

To conduct a backward systematic snowball review, we used the bibliographies of the (n = 3) seeds as a starting point. We analyzed the references of those three articles in three phases, employing specific inclusion and exclusion criteria to review the bibliographies of each paper. The inclusion criteria consisted of papers published from the inception of using VDZ as a treatment for UC up to the 2023 timeframe, peer-reviewed and in English. The exclusion criteria were as follows: Case reports, case series, and abstracts were excluded, as well as studies on patients less than 18 years old. UC patients should have received at least 3 doses of 300 mg IV of VDZ to be included in the study.

Using these criteria, a total of 14 new relevant papers were identified during the bibliographic snowballing process, as illustrated in Figure 1. In the process of deep analysis, we found that 4 articles out of the 14 new articles were merging the analysis of CD and UC patients and therefore were excluded. Ultimately, 14 identified papers were used in the analysis. Encountered duplications were removed during the process. The snowball search was conducted by two separate reviewers, and the last run of citation search was conducted on September 4, 2023.

Figure 1.

Backward snowballing search process: (i) The initial phase—scanning of the bibliographies of the initial set of papers was demonstrated including Dalal et al., Ng Sc et al., and Colombel et al. The relevant papers were identified based on title, abstract, introduction, and results when needed, and sometimes, more sections were reviewed. In this phase, the review of the whole text was not done; (ii) The second phase—application of the inclusion and exclusion criteria; (iii) The third phase—in-depth analysis of n = 14 new papers was conducted to confirm relevance. Then, it was run

Conducting keyword-based search

Five search key phrases (KP1-5) were set to identify relevant articles:

KP1: “vedolizumab AND ulcerative colitis”

KP2: “vedolizumab AND UC”

KP3: “vedolizumab AND IBD”

KP4: “vedolizumab AND CDI”

KP5: “vedolizumab AND Clostridium”

A classic database search was conducted after the snowball search to ensure broad coverage of the literature. We searched PubMed/MEDLINE, Embase, and Web of Science using the five key phrases stated previously; this method did not show publications not included during the snowballing search.

Before retrieving information from the included articles, we evaluated the quality of the papers. Two independent reviewers (MA and DA) assessed the quality of the included studies using the Cochrane Risk-of-Bias Tool for Randomized Trials (ROB-2)[10] and the Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I).[11] Any discrepancies were resolved by consensus and by consulting the senior author (MM).

Statistical analysis

A Meta-analysis was performed to investigate the association between the use of VDZ and the risk of developing CDI, using R statistical software with the “metafor” package (version 4.2-0). We employed a random effect model to aggregate the results of the included studies. This statistical method incorporates variations within each study and heterogeneity between studies, thereby providing a more reliable estimation of the overall effect. We evaluated the goodness of fit and the appropriateness of our statistical model using various statistics and information criteria including the Maximum-Likelihood estimation yielded log-likelihood, Akaike information criterion (AIC), Bayesian Information Criterion (BIC), and the corrected Akaike criterion (AICc).

Heterogeneity statistics

The level of heterogeneity among the included studies was initially assessed through visual inspection of forest plots [Figure 2]. The absence of statistical heterogeneity was observed as indicated by the I2 test, with a P- value of 0.973, considered insignificant.

Figure 2.

Forest plot of pooled studies

Publication bias

Various tests were used to assess publication bias in this meta-analysis, including the Fail-Safe N, Kendall’s Tau, and Egger’s regression. This assessment was conducted through the backward snowballing (BSB) process similar to the initial set of papers with three phases of analysis, continuing until no further papers were found.

RESULTS

Characteristics of the included studies

Fourteen relevant publications were identified for assessing the risk of developing CDI associated with using vedolizumab in 4160 patients with active UC.[12,13,14,15,16,17,18,19,20,21,22,23,24,25,26] Figure 1 illustrates the process of identification and selection of the study. Table 1 summarizes the main characteristics of the studies that have been included.

Table 1.

Final list of included studies (Total n=14)

Study	Study design	Year of publication	No. of patients with active UC	Male gender (%)	Montreal classification (N)							Use of steroids (%)
					E1	E2	E3	S0	S1	S2	S3
Uri Kopylov et al.[13]	Prospective cohort study	2017	74	56.8%	2	38	34	-	-	24	26	48.6%
Amy Lightner et al.[12]	Retrospective cohort study	2016	22	-	-	-	-	-	-	-	-	39%
Joseph Meserve et al.[14]	Retrospective cohort study	2019	437	45%	-	-	-	-	-	-	-	49%
Siew Ng et al.[15]	Double-blinded randomized clinical trial	2018	894	-	-	-	-	-	-	-	-	0%
Narula et al.[16]	Retrospective cohort study	2018	321	59%	16	125	178	-	32	180	109	60.75%
William J. Sandborn et al.[17]	Double-blinded randomized clinical trial	2020	54	57.4%	-	-	-	-	-	17	37	38.9%
Baumgart et al.[18]	Prospective cohort study	2016	115	57.4%	7	30	62	-	-	-	-	83.5%
Sands et al.[19]	Double-blinded randomized clinical trial	2019	383	60.8%	-	-	-	-	-	-	-	36.1%
Singh et al.[20]	Retrospective cohort study	2022	672	49.3%	-	-	-	-	-	-	-	77.7%
Wright et al.[21]	Retrospective cohort study	2018	8	100%	-	-	-	-	-	-	-	100%
Lukin et al.[22]	Retrospective cohort study	2022	454	50.2%	22	161	270	-	52	253	149	53.5%
Colombel et al.[23]	Retrospective cohort study	2017	1107	58%	-		-	-	-	-	-	53%
Dalal et al.[24]	Retrospective cohort study	2022	195	44.6%	-	158	-	-	-	-	-	43.1%
Alshahrani et al.[25]	Retrospective cohort study	2023	96	-	-	-	-	-	-	-	-	4.4%

Severity, S0: remission, no symptoms; S1: mild symptoms; S2: moderate symptoms; S3: severe symptoms; Extensity, E1: ulcerative proctitis; E2: left-sided UC, distal colitis; E3: extensive UC, pancolitis

Risk of developing CDI with VDZ

The Forest plot summarizes the results of our meta-analysis demonstrating the effects of VDZ on outcomes of CDI in UC patients. The pooled effect size showed no statistical significance (effect size = 0.03 [-0.02, 0.07]) [Table 2].

Table 2.

Random-effects model (k=14)

	Estimate	SE	Z	P	CI
					Lower Bound	Upper Bound
Intercept	0.0257	0.0231	1.11	0.267	-0.020	0.071

Tau² Estimator: Restricted Maximum-Likelihood

Heterogeneity statistics

The calculated Tau value and I squared (I² = 0%) indicate no observed heterogeneity among the effect sizes. This suggests that the variations in the effect sizes across the included studies are solely due to chance and there are no significant sources of variations that can be attributed to factors other than random variation.

Model fit statistics and information criteria

The Restricted Maximum Likelihood (REML) approach is strongly supported by comparing it with the Maximum Likelihood (ML) model. The REML model demonstrates a lower log-likelihood (13.331 vs 14.859), indicating that it is less likely to have generated the observed data by chance. Additionally, the REML model exhibits a lower deviance (-26.663 vs 5.093), signifying a superior fit to the data. Furthermore, the REML model outperforms the ML model in terms of both AIC and BIC, with lower values for both (-25.718 and -24.627 vs -22.663 and -24.440, respectively). These lower AIC and BIC scores highlight that the REML model provides a better fit for the data while maintaining a lower complexity level than the ML model.

Publication bias

Upon evaluating the data for publication bias, inconsistent findings were revealed [Table 3]. While Kendall’s Tau and the Fail-Safe N test indicate some degree of concern regarding publication bias, Egger’s regression presents less compelling evidence. A funnel plot was used to visually detect publication bias [Figure 3].

Table 3.

Publication bias assessment

Test Name	Value	P
Fail-Safe N	4.000	0.033
Kendall’s Tau	0.692	<0.001
Egger’s Regression	1.829	0.067

Fail-safe N calculation using the Rosenthal approach

Figure 3.

Funnel plot displaying risk of publication bias

DISCUSSION

The medical management of UC revolves around counteracting inflammation, which is driven mainly by immune dysfunction. Opposing inflammation has historically been achieved through locally active anti-inflammatory agents such as 5ASA derivatives, which lead to response/remission in up to 50% of patients. This class of medication delivers the advantage of being safe and free of any major adverse events. However, patients who fail to respond to this category of medications are typically escalated to drugs that influence the immune system, namely, corticosteroids for induction and thiopurines for maintenance of remission. The use of such medication is restricted by their potential to cause systemic side effects such as metabolic derangements and increased risk of infection and hence do not achieve the hope of providing successful treatment outcomes. In the late nineties, the first class of biologic therapy was introduced as a therapeutic option for moderate to severely active UC. Infliximab was the first TNF-α inhibitor to be approved for UC after the two landmark studies ACT-1 and ACT-2 demonstrated efficacy and safety. Subsequently, two more TNF-α inhibitors were approved for UC, adalimumab, and golimumab. TNF-α inhibitors were considered revolutionary therapies, but with time, through careful monitoring and follow-up of patients, it was determined that they were associated with certain serious adverse events such as increased risk of opportunistic infections, demyelinating diseases, and malignancy. For this reason, the search for safe and effective therapies continued.

VDZ is a gut-selective leukocyte trafficking inhibitor that interferes with the interaction between α₄β₇ integrins and MAdCAM-1. The gut-specific nature of this interaction guarantees no systemic adverse effects with this class of medications. This has been proven through multiple phase 4 studies and meta-analyses. On the other hand, a concern about a theoretical increase in the risk of enteric infections, specifically CDI, has been proposed and supported by several case reports and case series. Following results from the phase 3 induction and maintenance studies for VDZ in UC and CD, which showed a low risk of enteric infections (2% for UC and 6% for CD), surveillance for gut-associated infections continues.[26,27] Theoretically, enteric infections can flourish when leukocyte trafficking is blocked because this may interfere with multiple important immune-mediated interactions necessary to combat infections. First, impaired immune surveillance makes it difficult for vital immune cells like neutrophils and macrophages to efficiently patrol the intestines for approaching infections. Second, the body’s capacity to engulf and digest pathogens is weakened by diminished phagocytosis, which allows it to survive and thrive in the intestines. Third, limiting inflammation impairs the immune response by prohibiting the recruitment of more immune cells to infection sites[28]. The risk of CDI is known to be higher in patients suffering from UC compared to the general population; whether this risk is amplified using VDZ remains questionable. For this reason, we sought to examine the association between VDZ and the risk of CDI in patients with UC. Upon systematically searching the literature, we identified 14 studies that were eligible for pooling of data. The pooled effect size showed no statistical significance (effect size = 0.03 [-0.02, 0.07]). This result supports the notion that VDZ does not increase the risk of CDI in patients with UC. One plausible explanation would be that the most important factor for the emergence of CDI is the disease-associated dysbiosis, and the improvement in disease is associated with improved microbial health that increases the resilience to enteropathogens and hence the reduction of CDI events (compared with untreated IBD).

The SB method is more effective than a database search when using general terms and reducing the amount of noise[29]. Additionally, it has been reported to be more efficient at discovering pertinent articles, with a discovery rate of 85%, compared to a database search, which only identified 45.9% of relevant papers. We ran a parallel keyword-based search which supports these findings as no further articles have been identified. Ultimately, the SB technique has demonstrated greater dependability when the initial set of papers is appropriately selected in contrast to a database search.[8] In the current IBD meta-analysis landscape, this technique has become pervasive with numerous studies and publications showcasing its effectiveness.[28,29,30,31]

One of the main limitations of this study design is the reliance on available published studies, which can introduce publication bias. Aside from this, this meta-analysis is dependent on the reporting and quality of the studies included. Additionally, the number of included studies was small, and some studies did not report specific details related to the factors that can affect the risk of developing CDI for UC patients on VDZ. The small number of studies included in the analysis might affect the outcome; therefore, reanalysis is required when more data become available.

In conclusion, according to this systematic review and meta-analysis, VDZ does not increase the risk of CDI in patients with UC. Further prospective randomized controlled trials are needed to consolidate these findings.

Financial support and sponsorship

Nil.

Conflicts of interest

Dr. Mahmoud Mosli has served on advisory boards, and received speaker fees, and/or consultation fees from Abbvie, Takeda, Janssen, Ferring, Falk, Sandoz, Hikma, Pfizer, BMS, and Orgenon. Dr. Mosli has received research funding from Celgene, Pfizer, and Takeda.

SUPPLEMENTARY DATA

Supporting Figure 1

PRISMA chart

Supplementary Table 1.

Risk of Bias in Non-randomized Studies using ROBINS-1 assessment tool

ROBINS-1 (cohort-type studies)	Bias due to confoundings	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from intended interventions	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall bias
Singh S, 2021	Moderate	Moderate	Low	Low	NI	Low	Low	Moderate
Baumgart DC, 2016	Moderate	Low	Low	Low	Low	Low	Low	Low
Amiot A, 2017	Low	Low	Low	Low	NI	Low	Low	Low
Narula N, 2018	Moderate	Low	Low	Low	NI	Low	Low	Low
Kopylov U, 2017	Moderate	Low	Low	Low	NI	Low	Low	Low
Wright AP, 2017	Moderate	Moderate	Low	Low	NI	Low	Low	Moderate
Lightner AL, 2017	Moderate	Moderate	Low	Low	NI	Low	Low	Moderate
Lukin D, 2022	Low	Low	Low	Low	Low	Low	Low	Low
Meserve J, 2019	Moderate	Low	Low	Low	Low	Low	Low	Low
Alshahrani, 2023	Moderate	Low	Low	Low	Low	Low	Low	Low

Supplementary Table 2.

Risk of Bias in Randomized Controlled Trials

Study name	D1	D2	D3	D4	D5	Overall
Sands BE, 2019	L	L	L	L	L	L

L: Low risk, U: Unclear risk, H: High risk Domains: D1: Bias arising from the randomization process, D2: Bias due to deviations from intended interventions, D3: Bias due to missing outcome data, D4: Bias in measurement of outcome, D5: Bias in the selection of reported results