Respiratory Tract Infections in Patients With Inflammatory Bowel Disease: Safety Analyses From Vedolizumab Clinical Trials

Abstract Background and Aims Vedolizumab, a humanised monoclonal antibody for the treatment of inflammatory bowel disease, selectively blocks gut lymphocyte trafficking. This may reduce the risk of respiratory tract infections [RTIs] compared with systemic immunosuppressive therapies. To assess this possibility, we evaluated the rates of RTIs in clinical trials of vedolizumab. Methods Patient-level data from Phase 3 randomised controlled trials [RCTs] of vedolizumab in ulcerative colitis [UC; GEMINI 1] and Crohn’s disease [CD; GEMINI 2], and a long-term safety study [UC and CD] were pooled. Cox proportional hazards models were used to estimate the incidence of upper RTIs [URTIs] and lower RTIs [LRTIs] with adjustment for significant covariates. Results In the RCTs [n = 1731 patients], the incidence of URTIs was numerically higher in patients receiving vedolizumab compared with those receiving placebo, although this difference was not statistically significant (38.7 vs 33.0 patients per 100 patient-years; hazard ratio [HR] 1.12; 95% confidence interval [CI]: 0.83–1.51; p = 0.463). The rate of LRTIs, including pneumonia, was numerically lower in the vedolizumab versus the placebo group: this difference was not statistically significant (7.7 vs 8.5 per 100 patient-years [HR 0.85; 95% CI: 0.48–1.52; p = 0.585]). Both URTIs and LRTIs were more frequent in patients with CD compared with UC. Most RTIs in patients receiving vedolizumab were not serious and did not require treatment discontinuation. Conclusions Vedolizumab therapy was not associated with an increased incidence of respiratory tract infection compared with placebo.


Introduction
Tumour necrosis factor [TNF] antagonists have revolutionised the treatment of inflammatory bowel disease [IBD]. Although these drugs are considered relatively safe, the increased risk of serious infection remains an important concern. 1-4 Upper respiratory tract infections [URTIs] are among the most common adverse events [AEs] reported in randomised controlled trials of TNF antagonists, [5][6][7][8] and an increased incidence of URTIs in comparison with placebo has been consistently observed in these studies. 7,9,10 Patients with IBD have an increased risk of pneumonia which is further intensified by TNF antagonist therapy. 11 In a retrospective cohort study, the risk of developing pneumonia was approximately 50% higher for patients with IBD [n = 108 604] than for the general population (n = 434 416; hazard ratio [HR] 1.54; 95% confidence interval [CI]: 1.49-1.60), and among those with IBD, TNF antagonist therapy was independently associated with pneumonia (odds ratio [OR] 1.28; 95% CI: 1.08-1.52). 11 In a US hospitalisation database, 27.5% of all Crohn's disease [CD]-or ulcerative colitis [UC]-related hospitalisations were attributable to infection. 12 These patients had excess mortality risks compared with patients without infection-related hospitalisations, which varied depending on the infection type; pneumonia had one of the highest excess mortality risks compared with patients without infectionrelated hospitalisation [OR 3.6; 95% CI: 2. 9-4.5]. An analysis of the TREAT™ registry, which evaluated 6273 patients with CD over a mean follow-up of 5.2 years, found that infliximab therapy was independently associated with an increased risk of serious infection (2.06 events per 100 patient-years [HR 1.43; 95% CI: 1.11-1.84; p = 0.006]). 3 The most common serious infection was pneumonia, with an incidence of 0.24 cases per 100 patient-years of follow-up in infliximab-treated patients compared with a rate of 0.14 cases per 100 patient-years of follow-up for those on treatments other than infliximab. 3 Similarly, in the ENCORE registry, which followed 1541 infliximab-treated patients with CD for up to 5 years, infliximab therapy was independently associated with an increased risk of serious infection [HR 1.64; 95% CI: 1.17-2.31], and the most common AEs were abscess and pneumonia. 13 Vedolizumab, a humanised monoclonal antibody that binds to the α 4 β 7 integrin and selectively blocks lymphocyte trafficking to the gut, 14,15 has been shown to be an effective induction and maintenance therapy for both UC and CD. 16,17 Unlike thiopurines, methotrexate, and TNF antagonists, the gut-selective mechanism of action of vedolizumab 15 does not result in systemic immunosuppression. 14 Although a previous evaluation of the safety of vedolizumab did not identify an increased risk of respiratory tract infection [RTI], 18 the estimates used in that study were based on simple incidence rates and were not adjusted for important covariates such as smoking status and age. Given the previously described increase in risk associated with TNF antagonists, it is important to accurately estimate the incidence of RTIs in patients treated with vedolizumab. We used an integrated dataset derived from placebo-controlled trials to obtain such estimates.

Data sources: GEMINI 1 and 2 studies and the GEMINI open-label extension
Data from two Phase 3, randomised, placebo-controlled clinical trials of vedolizumab, GEMINI 1 [UC] and GEMINI 2 [CD], were analysed for rates of URTIs and LRTIs. GEMINI 1 and 2 evaluated the efficacy and safety of vedolizumab 300 mg or placebo as induction and maintenance therapy for up to 52 weeks. The study designs and outcomes of these trials have been described previously. 16,17 In this analysis, the vedolizumab study population [n = 1434] included all patients who were responders to vedolizumab induction therapy and were subsequently randomised to vedolizumab maintenance therapy [every 4 or 8 weeks] at Week 6, and also those who received induction therapy with vedolizumab, did not achieve clinical response at Week 6, and were then assigned to vedolizumab every 4 weeks for the maintenance phase. The placebo population [n = 297] comprised patients who received placebo during the induction phase and were subsequently assigned to continue placebo during the maintenance phase.
To evaluate the incidence of RTI beyond 52 weeks of treatment with vedolizumab, we also included data from the GEMINI long-term safety [LTS] trial, which evaluated the safety of vedolizumab 300 mg administered every 4 weeks in adult patients: 'rollover' patients from GEMINI 1 [n = 675], 16  . SAEs were defined as those occurring at any dose, regardless of causality, which: were life-threatening or resulted in death; required inpatient hospitalization; resulted in persistent or significant disability/incapacity, congenital anomaly or birth defect; jeopardised the patient and required medical or surgical intervention to prevent any of the above; or involved suspected transmission of an infectious agent. For this analysis, 'LRTI' is used as an abbreviation for the MedDRA HLT 'lower respiratory tract and lung infection', and 'events' refers to AEs. Data for the maintenance phase only were analysed for GEMINI 1 and 2, as well as the LTS trial [including exposure from previous trials]. Exposure-adjusted incidence rates per 100 patient-years were calculated for the incidence of patients with any AE or SAE within the HLTs, as well as for each individual MedDRA preferred term within these. LRTI AEs were analysed by duration of vedolizumab exposure, for the GEMINI LTS trial to evaluate trends over time. The times to first URTI and LRTI event for the GEMINI 1, 2, and LTS trials were also evaluated.
Multivariate logistic regression and Cox proportional hazards modelling were used to identify predictors for the occurrence of any RTI, any URTI, or any LRTI in the individual and pooled GEMINI 1 and GEMINI 2 studies. Predictors assessed were: age; sex; disease duration; previous TNF antagonist use; baseline disease activity; concomitant use of narcotics, corticosteroids, or immunosuppressives; smoking history; and occurrence of surgery during the study. Both unadjusted and adjusted models were evaluated. , and common index for pooled GEMINI 1 and 2. Common index ranged from 0 to 9 to allow the combination of baseline partial Mayo and HBI scores in the pooled analysis.

Baseline characteristics
Demographics and baseline characteristics are summarised in Table 1

Overall respiratory tract infections
No statistically significant differences in time to first lower/upper RTI were observed between patients assigned to vedolizumab therapy and those who received placebo [ Figure 1A;

Upper respiratory tract infection
No statistically significant differences in time to first URTI were observed between patients assigned to vedolizumab therapy and those who received placebo [ Figure 1B; HR 1.23, log-rank Acute tonsillitis Acute sinusitis    The most frequently reported URTI, nasopharyngitis, occurred at a higher rate in the vedolizumab group than in patients who received placebo [18.6 vs 12.8 patients per 100 patient-years in the pooled GEMINI 1 and 2 populations; Table 2]. URTI was the next most common URTI [MedDRA preferred term] and the incidence rate was similar for vedolizumab-treated and placebo-receiving patients [10.5 vs 11.6 patients per 100 patient-years].

Lower respiratory tract infection
No statistically significant differences in time to first LRTI were observed between patients assigned to vedolizumab therapy and those who received placebo [ Figure 1C; HR 0.95, log-rank p = 0.851]. In unadjusted multivariable logistic regression analyses, previous TNF antagonist use [HR 2.20; 95% CI: 1.10-4.41; p = 0.027] and female sex [HR 2.11; 95% CI: 1.07-4.14; p = 0.030] were associated with LRTI in patients with UC, whereas in the CD population, current smoker status alone was associated with the occurrence of LRTI [HR 2.37; 95% CI: 1.26-4.45; p = 0.008; Table 3]. A higher proportion of patients with CD were current smokers compared with those with UC [29.2% of females and 22.3% of males in GEMINI 2; 4.2% of females and 7.5% of males in GEMINI 1]. Vedolizumab therapy, age, disease duration, baseline disease activity, or concomitant narcotic, corticosteroid, and immunosuppressive use were not associated with a significantly greater risk of LRTI. After adjustment for disease activity, none of the evaluated risk factors were significant predictors of LRTI in patients with UC [ Table 4]. However, smoking [current smokers, HR 3.43; 95% CI: 1.67-7.04; p = 0.0008] remained a significant predictor of LRTI in patients with CD.
The exposure-adjusted incidence rate of LRTIs was numerically lower for vedolizumab-treated patients than for placebo recipients in the pooled GEMINI 1 and GEMINI 2 analysis, although this difference was not statistically significant (7.7 vs 8.  Table 2). There were two LRTI-related discontinuations in the vedolizumab group [bronchitis and pneumonia, both in patients with CD] compared with none in the placebo group [Supplementary Table 2]. There was one LRTI-related death, which occurred in a patient receiving placebo who died from bronchopneumonia. Patients with UC who received treatment with vedolizumab had a statistically non-significant lower rate of LRTIs compared with those receiving placebo (6.8 vs 9.0 per 100 patient-years [HR    Table 2). Two of these events in the vedolizumab group were serious, both in patients with CD. The fatal event of bronchopneumonia in a patient with CD was the only serious event in the placebo group.

Respiratory tract infections based on frequency of vedolizumab maintenance therapy
No statistically significant differences were observed between patients assigned to vedolizumab maintenance therapy every 4 weeks vs every 8 weeks in time to first RTI [ Figure 3A; HR 0.98, log-rank p = 0.864], time to first URTI [ Figure 3B; HR 0.98, log-rank p = 0.902] or time to first LRTI [ Figure 3C; HR 1.20, log-rank p = 0.531]. Similarly, no significant between-group differences were found in exposure-adjusted incidence rates of RTI, URTI, or LRTI [ Figure 4]. Two patients who received vedolizumab maintenance therapy every 4 weeks had a serious URTI and one patient discontinued treatment due to a serious URTI [Supplementary      Table 3). Two of the pneumonia events in the vedolizumab 4-week group and one of the bronchitis events in the vedolizumab 8-week group were serious.

Respiratory tract infections in the open-label extension
In the GEMINI LTS trial, exposure-adjusted incidence rates for URTI  Figure 2 and Supplementary Table 2). Nasopharyngitis and URTI [MedDRA preferred term] were the most frequent URTI events [ Table 2]. There was no increase in the incidence rates [95% CI] of LRTIs in the GEMINI LTS trial relative to those in the pooled GEMINI 1 and 2 data analysis  Table 2]. , and common index for pooled GEMINI 1 and 2. Common index ranged from 0 to 9 to allow the combination of baseline partial Mayo and HBI scores in the pooled analysis.
In the GEMINI LTS trial, the proportion of patients with reported respiratory infection declined over time: from 24.8% during months 0-12 to 9.6% during months 60-72 for URTI events, and from 5.7% during months 0-12 to 1.1% during months 60-72 for LRTI events [ Table 5].

Discussion
Our results show that, in the GEMINI 1 and 2 studies, vedolizumab therapy was not associated with significantly higher rates of RTI. Most of the URTIs and LRTIs that occurred in patients receiving vedolizumab were not serious and did not result in treatment discontinuation. Multivariate Cox proportional hazards modelling showed that, unlike smoking or previous exposure to TNF antagonist therapy, vedolizumab therapy was not a risk factor for LRTI events, in either UC or CD. Moreover, analysis of vedolizumab serum concentrations did not reveal any significant association or consistent trends between vedolizumab quartile levels and RTIs, URTIs, or LRTIs [unpublished data]. These results are consistent with a metaanalysis of pooled data from six randomised placebo-controlled clinical trials in patients with IBD, which found no significant differences between vedolizumab and placebo in the risk of all serious infections [relative risk 1.17; 95% CI: 0.51-2.69]. 24 Pneumonia is an event of particular interest, and rates were low in these post hoc analyses, with most classified as non-serious. Of the 11/1434 vedolizumab-treated patients in GEMINI 1 and 2 who had pneumonia, the event was considered serious in two patients. In the GEMINI LTS trial, of the 49/2243 patients who had pneumonia events [0.92 per 100 patient-years], 18 were considered to have serious pneumonia [0.33 per 100 patient-years]. One patient discontinued vedolizumab due to pneumonia. The data suggest that long-term vedolizumab therapy does not have any progressive or cumulative effect on susceptibility to pneumonia. These findings are consistent with the gut-selective mechanism of action of vedolizumab and lack of systemic immunosuppressive effects. 14 The comparator group comprised patients who responded to placebo and were subsequently maintained on placebo, suggesting a relatively lower disease severity that would, if anything, bias the study toward finding an increase in pneumonia risk with vedolizumab exposure. This finding also provides greater confidence in the conclusion that vedolizumab was not associated with a greater risk of pneumonia. Although previous studies had suggested a potential increased risk of URTI for vedolizumab, based on mucosal vascular addressin cell adhesion molecule 1 [MAdCAM-1] expression in the oropharynx, our results do not support a clinically relevant corollary of this finding. 15 An increased risk of RTI is one limitation of TNF antagonist therapy. 3,[11][12][13] Due to their systemic immunosuppressive mode of action, patients exposed to these agents have an increased risk of URTI and LRTI, serious RTIs, and pneumonia-related mortality. 3,[11][12][13] It is difficult to make an indirect comparison of the absolute rates of URTI   and LRTI events observed in this analysis with those from pivotal trials of TNF antagonists in IBD [Supplementary Table 4] for several reasons. First, published reports provide only sparse details regarding these events, and rates of occurrence are expressed as simple proportions rather than exposure-adjusted rates. Second, publications often provide rates for common AEs only-typically those that occurred in >5% of patients. Usually the incidence of LRTIs does not meet this criterion. Consequently, little attention has been drawn to the issue of RTIs in patients receiving TNF antagonists despite observations that LRTIs, and specifically pneumonia, are among the most common serious infections associated with the use of these agents. 3,25 Risk factors associated with the occurrence of pneumonia include smoking 26 and TNF antagonist therapy. 3,[27][28][29] Consistent with this, our analysis identified smoking status, previous use of TNF antagonists, and female sex as independent risk factors for LRTI. In both UC and CD, cigarette smoking was identified as a risk factor for both URTI and LRTI. Although this finding requires little explanation, it is concerning that the prevalence of cigarette smoking remains high in patients with CD. A higher proportion of patients with CD were current smokers compared with those with UC. In GEMINI 1 [UC], 4.2% of females and 7.5% of males were current smokers compared with 29.2% of females and 22.3% of males in GEMINI 2 [CD]. Accordingly, measures to reduce the known risk factors for RTI should be implemented, including referral to smoking cessation programmes, administration of vaccines, 30 and judicious use of systemic immunosuppression in these patients. 31 Both the pooled GEMINI 1 and 2 population and patients with UC were more likely to have an LRTI if they had had previous TNF antagonist exposure, despite implementation of a washout period before randomisation 16,17 [Table 3]. However, there is some evidence to suggest that previous TNF antagonist therapy in patients with rheumatoid arthritis is associated with serious infection, 25 and switching from a TNF antagonist to a treatment with an alternative mechanism of action may reduce the risks of infection. 32 An explanation for these observations is not readily apparent. However, patients with previous TNF antagonist exposure may be at greater risk for RTIs because of longer disease duration, slightly greater baseline disease activity, and greater use of additional therapies such as corticosteroids, compared with patients without previous exposure [Supplementary Table 5, available as Supplementary data at ECCO-JCC online].
The GEMINI LTS trial provides insight into the risks of longterm exposure to vedolizumab (5430 patient-years to interim analysis date [ May 19, 2015]). Rates of URTI and LRTI events were generally lower than those in the GEMINI 1 and 2 studies, and exhibited similar patterns. These lower rates may reflect the improvement in general health in patients responding to longterm vedolizumab therapy. The importance of accruing data from a large number of patients with a long follow-up time, thereby improving sensitivity to rare or delayed safety concerns, should be balanced with the limitations of LTS studies, such as lack of a reference arm. Ongoing registry evaluations, such as the recently initiated ENTYVIO ® Outcomes in Realworld Bio-naïve Ulcerative Colitis and Crohn's Disease Patients [EVOLVE], will provide insight on treatment patterns and outcomes for patients.
In summary, this post hoc analysis of controlled trials and their LTS study provides a comprehensive assessment of LRTIs and URTIs in almost 6500 patient-years of exposure to vedolizumab for the treatment of IBD. Vedolizumab therapy was not associated with an increased incidence of LRTIs compared with placebo.

Funding
This work was supported by Takeda Pharmaceutical Company Ltd. Medical writing support was provided by Khalid Siddiqui of Chameleon Communications International Ltd, UK [a Healthcare Consultancy Group Company] and sponsored by Takeda Pharmaceutical Company Ltd.