Induction therapy

The results of the NMAs for induction therapy are summarized in Table 52 and Table 53. The NMA conducted by Singh et al. included a biologic-naive population only and suggests no significant difference between vedolizumab and adalimumab, with an RR of 0.47 (95% CrI, 0.13 to 1.75), for inducing clinical remission in the induction phase. However, the NMA conducted by Singh et al. does provide estimates of effect favouring treatment with infliximab over vedolizumab for inducing clinical remission in the induction phase, with an RR of 0.23 (95% CrI, 0.06 to 0.78). Furthermore, the NMA conducted by Hazelwood et al. included a mixed population of participants with and without prior experience with TNF alpha antagonists and suggests no significant difference between vedolizumab and adalimumab or infliximab for inducing clinical remission, with an OR of 0.67 (95% CrI, 0.33 to 1.5) and 0.70 (95% CrI, 0.25 to 1.5), respectively.

Table 52

Clinical Remission or Response in the Induction Phase.

Table 53

Induction Phase Efficacy End Points in the Network Meta-Analysis Submitted to Nice.

The IDC conducted by Miligkos et al. included a mixed population of participants with and without prior experience with TNF alpha antagonists and suggests no significant difference between adalimumab and vedolizumab, with an OR of 1.27 (95% CI, 0.69 to 2.33), for inducing clinical response in the induction phase. However, the IDC conducted by Miligkos et al. does provide estimates of effect favouring treatment with infliximab over vedolizumab for inducing clinical response in the induction phase, with an of 2.81 (95% CI, 1.10 to 7.20). With regard to clinical remission in the induction phase, the IDC conducted by Miligkos et al. suggests no significant differences between both infliximab and adalimumab compared with vedolizumab, with ORs of 1.41 (95% CI, 0.74 to 2.67) and 1.09 (95% CI, 0.60 to 1.98), respectively.

The NMA submitted to NICE provided only ORs versus placebo and did not present any data indirectly comparing active treatments to vedolizumab. In the anti-TNF-naive population, the NMA submitted to NICE did provide estimates of effect favouring treatment with vedolizumab, adalimumab, and infliximab when compared with placebo for inducing clinical remission in the induction phase, with ORs of 0.29 (95% CrI, 1.5 to 6.0), 4.1 (95% CrI, 1.8 to 10) and 26 (95% CrI, 4.0 to 425), respectively. Similarly, estimates of effect favouring treatment with vedolizumab, adalimumab, and infliximab when compared with placebo were observed for both enhanced clinical response and clinical response in the induction phase.

The NMA submitted to NICE also provided results based on an anti-TNF–experienced/failure population. The results for this population suggest no significant difference between vedolizumab and placebo for inducing clinical remission in the induction phase (OR 1.4; 95% CrI, 0.8 to 2.6). However, the IDC submitted to NICE does provide estimates of effect favouring treatment with adalimumab over placebo for inducing clinical remission in the induction phase with an OR of 3.6 (95% CrI, 1.8 to 7.1). Similar to the anti-TNF-naive treatment population, estimates of effect favouring treatment with vedolizumab and adalimumab when compared with placebo were observed for both enhanced clinical response and clinical response in the induction phase for the anti-TNF–experienced/failure population.

The NMA submitted to NICE provided no information with respect to infliximab in the induction phase for the anti-TNF–experienced/failure population.

Maintenance therapy

The results of the NMAs for maintenance therapy are summarized in Table 54 and Table 55. The NMA conducted by Singh et al. included a biologic-naive population only and suggests no significant difference between vedolizumab and adalimumab or infliximab for maintaining clinical remission in the maintenance phase, with RRs of 0.43 (95% CrI, 0.05 to 3.36) and 0.67 (95% CrI, 0.06 to 5.64), respectively. Furthermore, the NMA conducted by Hazelwood et al. included a mixed population including participants with and without prior experience with anti-TNF treatments and suggests no significant difference between vedolizumab and infliximab for maintaining clinical remission in the maintenance phase, with ORs of 0.77 (95% CrI, 0.39 to 1.5) and 0.70 (95% CrI, 0.25 to 1.5), respectively. However, the NMA conducted by Hazelwood et al. does provide estimates of effect favouring treatment with adalimumab over vedolizumab for maintaining clinical remission in the maintenance phase, with an odds ratio of 0.42 (95% CrI, 0.22 to 0.85).

Table 54

Clinical Remission or Response in the Maintenance Phase.

Table 55

Maintenance Phase Efficacy End Points in the Network Meta-Analysis Submitted to Nice.

The IDC conducted by Miligkos et al. included a mixed population of participants with and without prior experience with TNF alpha antagonists and suggests estimates of effect favouring treatment with infliximab and adalimumab over vedolizumab for maintaining clinical response in the maintenance phase, with ORs of 1.95 (95% CI, 1.03 to 3.70) and 2.23 (95% CI, 1.26 to 3.93), respectively. With regard to clinical remission in the maintenance phase, the IDC conducted by Miligkos et al. suggests no significant differences between infliximab compared with vedolizumab, with an OR of 1.61 (95% CI, 0.80 to 3.22). However, the IDC conducted by Miligkos et al. does provide estimates of effect favouring treatment with adalimumab over vedolizumab for maintaining clinical remission in the maintenance phase, with an odds ratio of 2.22 (95% CI, 1.20 to 4.09).

The NMA submitted to NICE provided only ORs versus placebo and did not present any data indirectly comparing active treatments to vedolizumab. In the anti-TNF–naive population, the NMA submitted to NICE did provide estimates of effect favouring treatment with vedolizumab and infliximab when compared with placebo for maintaining clinical remission in the maintenance phase, with ORs of 2.9 (95% CrI, 1.4 to 6.1) and 2.5 (95% CrI, 1.3 to 5.2), respectively. Similarly, estimates of effect favouring treatment with vedolizumab and infliximab were observed for maintaining clinical response in the maintenance phase when compared with placebo.

The NMA submitted to NICE provided no information with respect to adalimumab in the maintenance phase for the anti-TNF–naive population nor did it provide any information with respect to the maintenance phase for any biologics in the anti-TNF–experienced/failure population.

Safety

The results of the IDCs for safety are summarized in Table 56 and Table 57. The IDC conducted by Hazelwood et al. included a mixed population including participants with and without prior experience with anti-TNF treatments and suggests no significant difference between vedolizumab and adalimumab or infliximab in total withdrawals, with ORs of 2.1 (95% CrI, 1.0 to 4.6) and 1.3 (95% CrI, 0.57 to 2.9), respectively. When considering withdrawals due to adverse events, the NMA conducted by Hazelwood et al. suggests no significant difference between vedolizumab and adalimumab, with an odds ratio of 1.4 (95% CrI, 0.72 to 2.8). However, the NMA conducted by Hazelwood et al. does suggest a significant difference favouring treatment with vedolizumab over infliximab with respect to withdrawals due to adverse events, with an odds ratio of 0.24 (95% CrI, 0.12 to 0.51).

Table 56

Total Withdrawals and Withdrawals Due to Adverse Events in Hazelwood et al.

Table 57

Withdrawals Due to Adverse Events in the Network Meta-Analysis Submitted to NICE.

The IDC conducted by Miligkos et al. included a mixed population including participants with and without prior experience with anti-TNF treatments and suggests no significant difference between vedolizumab and infliximab or adalimumab with respect to adverse events, with ORs of 1.46 (95% CI, 0.75 to 2.84) and 0.75 (95% CI, 0.44 to 1.28) in the induction phase, and 0.83 (95% CI, 0.35 to 1.96) and 1.04 (95% CI, 0.54 to 2.04) in the maintenance phase, respectively.

The NMA submitted to NICE provided only ORs versus placebo and did not present any data indirectly comparing active treatments to vedolizumab. In the anti-TNF–naive population, the NMA submitted to NICE did provide significant estimates of effect favouring treatment with adalimumab when compared with placebo when considering withdrawals due to adverse events in the induction phase, with an OR of 0.0 (95% CrI, 0.0 to 0.7); however, vedolizumab was not associated with a significant difference with an OR of 1.4 (95% CrI, 0.3 to 7.4). In contrast, when considering the anti-TNF–experienced/failure population, the NMA submitted to NICE did provide estimates of effect favouring treatment with vedolizumab but not adalimumab when compared with placebo when considering withdrawals due to adverse events in the induction phase, with ORs of 0.4 (95% CrI, 0.1 to 0.9) and 0.5 (95% CrI, 0.1 to 2.4), respectively.

The NMA submitted to NICE also provided results based on an anti-TNF–naive population during the maintenance phase. The results for this population provide no significant estimates of effect for treatment with vedolizumab when compared to placebo when considering withdrawals due to adverse events in the maintenance phase with an OR of 0.8 (95% CrI, 0.3 to 2.7); however, infliximab was associated with a significant increase with an OR of 3.4 (95% CrI, 1.3 to 10.0).

Critical appraisal of the indirect cmparisons

Summaries of the limitations and heterogeneities of the included NMAs are presented in Table 59 and Table 60.

Table 59

Summary of Heterogeneities of Induction Studies Included in the Indirect Comparisons.

Table 60

Summary of Heterogeneities of Maintenance Studies Included in the Indirect Comparisons.

Study characteristics

The majority of trials for vedolizumab were conducted later than the trials for the comparators, particularly for infliximab (T16 concluded in 1996;⁸² ACCENT I concluded in 2001⁶⁷). It is possible that the clinical management of Crohn’s disease has evolved over the period since the introduction of the first biologic, introducing heterogeneity between the included studies.

Some of the studies included in the induction phase NMAs for infliximab, adalimumab, and vedolizumab utilized doses that are not reflective of the induction dosage regimens recommended in the Canadian product monographs (e.g., T16⁸²). These below-recommended dosages of active treatment may bias the study results in favour of vedolizumab for efficacy end points and against vedolizumab for safety end points. In the induction phase in some of the studies included in the NMAs, for maintenance end points, patients in the placebo groups received varying amounts of active treatment prior to randomization. For example, patients in the placebo group of ACCENT I⁶⁷ received only a single infusion of active treatment (i.e., at week 0) compared with the two infusions of active treatment in the GEMINI II²⁸ trial (i.e., at weeks 0 and 2). This difference in exposure is a significant source of heterogeneity across trials included in the NMAs and could contribute to the differences in placebo-response rates.

Although the dosing in at least one treatment group of the maintenance phase of the adalimumab trials was consistent with recommendations in the Canadian product monograph, the doses provided in the induction phase of the maintenance trials were below the recommended doses: patients in CHARM,⁸⁶ CLASSIC II,⁵⁴ and Watanabe⁸⁵ could have also received suboptimal induction doses. Similar to the infliximab comparison, these differences in exposure to active treatment within the placebo groups is a significant source of heterogeneity between the vedolizumab and adalimumab trials and could contribute to the reduced placebo-response rates reported in the CHARM⁸⁶ and Watanabe⁸⁵ trials compared with those reported in GEMINI II.²⁸ For all safety comparisons, using less than the recommended doses of Crohn’s disease treatment could underestimate the comparative harms associated with these treatments (with the exception of those associated with disease exacerbation).

Induction of clinical remission was evaluated at four weeks in the adalimumab and infliximab trials and six weeks in the vedolizumab trials. In addition, the efficacy end points in the maintenance phase studies were also evaluated at different time points (46 weeks with vedolizumab, 52 weeks with infliximab, and 52 to 56 weeks with adalimumab). Given that patients who failed to complete the trials were considered to be nonresponders in all of the included studies and that the proportion of patients who withdrew for any reason (including loss of efficacy and patients who were lost to follow-up) increases with time, having an earlier end point evaluation in the maintenance phase could favour vedolizumab treatment compared with the alternatives.

Study populations

The population of interest for the current CDR submission is patients with moderately to severely active Crohn’s disease who have had an inadequate response to alternative therapies (as per the indication under review for vedolizumab). Mean baseline CDAI scores for the induction phase studies were all within the moderate to severe range and were generally similar across the different studies. However, there is substantial heterogeneity in the characteristics of the different study populations, including clinically relevant parameters such as prior exposure to TNF alpha antagonists and concomitant use of Crohn’s disease treatments.

Patients were enrolled in the maintenance phase studies only if they had demonstrated a response to the active treatment in the induction phase. This introduces variation within the placebo groups across the studies, as the patients who were randomized to receive placebo in the maintenance phase had been previously treated with a different biologic therapy (e.g., vedolizumab, adalimumab, or infliximab). There were also differences in the placebo-response rates for maintaining clinical remission across the studies. The reason for these differences in the baseline risk for inducing and maintaining clinical remission is unclear; however, the manufacturer of vedolizumab has suggested that the differences in the maintenance phase could be attributed to the longer-lasting effect of vedolizumab compared with the TNF alpha antagonists (i.e., remission induced as a result of vedolizumab treatment is maintained longer than remission induced with the TNF alpha antagonists following removal of active treatment). Overall, these differences are an important source of between-study heterogeneity, and the implications for the results of the NMAs are unclear.

The placebo-response rate for inducing clinical remission was lower in the infliximab trials compared with the trials for vedolizumab and adalimumab. Similar to the maintenance phase analyses, the reasons for the differences in placebo-response rates are unclear for the induction phase.

As infliximab was the first biologic to be approved for use in the treatment of Crohn’s disease, all patients enrolled in the infliximab trials were naive to biologic therapy for Crohn’s disease. In contrast, the study populations for many of the other trials used in the NMAs were composed of a mix of treatment-naive and treatment-experienced patients who had previously failed at least one TNF alpha antagonist. In addition, a significant proportion of the patients in the vedolizumab trials had failed treatment with two TNF alpha antagonists, and a small proportion had failed treatment with three TNF alpha antagonists. Some of the adalimumab trials included patients with prior exposure to TNF antagonists, although few would have failed multiple TNF alpha antagonists as those enrolled in the vedolizumab trials. These differences in prior exposure to biologic therapy for Crohn’s disease may be clinically relevant and may be an indication that the study populations included in the vedolizumab trials are composed of patients with Crohn’s disease that is more refractory to treatment.

In general, there were differences between the induction studies and maintenance studies in the proportion of patients using concomitant Crohn’s disease treatments at baseline. Overall, there is insufficient evidence to evaluate whether or not these across-trial imbalances in concomitant Crohn’s disease treatment usage could influence the results of the IDCs, particularly in the induction phase analyses, where the differences were most pronounced.

Systematic review methods

The methods for the literature search were incomplete in the NMA conducted by Hazelwood et al. and in the IDC conducted by Miligkos et al., which were missing information about the reporting strategy. Furthermore, the NMA submitted to NICE and the NMA by Hazelwood et al. as well as the IDC by Miligkos et al. lack any description about how definitions of moderate to severe Crohn’s disease were considered in the study selection process. The NMA submitted to NICE did not provide a description of whether or how study durations were considered in the study selection process. In addition, the IDC conducted by Miligkos et al. appears to base the systematic review on published literature only, and makes no reference to the utilization of unpublished materials, which can lead to publication bias.

Analysis methods

The methodological descriptions of the Bayesian analyses were adequately reported in the all of the NMAs. Both direct (placebo comparisons) and indirect (active comparisons) estimates of effect are presented in the NMAs conducted by Singh et al. and Hazelwood et al.; however, the NMA submitted to NICE lacked the indirect (active comparisons) estimates of effect and reported only comparisons against placebo. The results of the NMAs for both efficacy and safety end points were adequately reported in summary tables, as RRs or ORs with 95% CrIs. In contrast, the methodological descriptions of the Bucher analyses were inadequately reported in the IDC conducted by Miligkos et al. The methods sections in the IDC conducted by Miligkos et al. lack any description about the Bucher analyses used to evaluate the results from the pooled trials included in the analysis. In addition, when considering trials with multiple active treatment arms, the IDC by Miligkos et al. combines the multiple treatment arms regardless of the dosing regimen in the evaluation of the pooled efficacy and safety, creating uncertainty in the overall dose-effect relationship. Furthermore, it appears that some of the included trials in the IDC by Miligkos et al. did not include consistent definitions for the measured outcomes (i.e., adverse events as defined in each of the included trials may have been different); consequently, any interpretation based on the pooling of safety outcomes must be made with caution, as this may not have been appropriate given the variation in definitions for safety outcomes. The results of the IDC for both efficacy and safety end points were adequately reported in summary tables, as ORs with 95% CIs.