Comparative Effectiveness of Therapies in Crohn's Disease
Comparative Effectiveness of Therapies in Crohn's Disease
We conducted a systematic review and network meta-analysis of randomized controlled trials of immunosuppressants and biologics for induction and maintenance of remission in adults with Crohn's disease. Our systematic review identified a paucity of head-to-head trials in the Crohn's disease literature, which are necessary to inform clinicians on the appropriate efficacious and safe treatment regimens for Crohn's disease. By applying a network meta-analysis approach, our study integrated direct head-to-head data with indirect evidence to provide the most robust data available for inducing and maintaining remission in Crohn's disease. The purpose of comparative effectiveness research is to "assist consumers, clinicians, purchasers, and policy makers to make informed decisions that will improve healthcare at both the individual and population levels." Until more head-to-head trials are completed, these data are the best we have available to guide clinical decision making within and between classes of drug therapies.
Azathioprine/6-mercaptopurine and methotrexate were not different from placebo for induction of remission in Crohn's disease. This finding is consistent with clinical practice, whereby immunosuppressants are not prescribed as monotherapy for induction of remission. Infliximab, infliximab + azathioprine, adalimumab, and vedolizumab were superior to placebo in inducing remission. In pairwise comparisons, infliximab + azathioprine and adalimumab showed superiority to certolizumab. However, induction studies should be compared carefully because of intrinsic differences in study designs. For example, the time point in defining remission for induction trials varied from 1 to 4 months.
For several decades, azathioprine/6-mercaptopurine have been the primary treatment for maintenance of remission of moderate to severe Crohn's disease. Although this network meta-analysis confirms that azathioprine/6-mercaptopurine are superior to placebo for maintenance of remission, its efficacy was inferior to adalimumab, infliximab, and infliximab + azathioprine. In this network meta-analysis, methotrexate was not different from azathioprine/6-mercaptopurine or any anti-TNF agent. Although these data suggest that methotrexate may be an efficacious primary treatment option for Crohn's disease, the results should be interpreted cautiously because the credible intervals were wide. The uncertainty around the estimates was the result of the small sample size of methotrexate clinical trials. Furthermore, 2 of the 5 trials were at high risk of bias and therefore were excluded from our primary analysis. In addition, among the 3 trials included in the primary analysis, the dose of methotrexate used was mixed (oral vs subcutaneous dose range, 12.5–25 mg). Nonetheless, this network meta-analysis at least supports a direct head-to-head trial of conventionally dosed methotrexate vs azathioprine/6-mercaptopurine with or without an additional anti-TNF arm to further inform this important therapeutic choice.
The introduction of anti-TNF therapy has had a significant impact on the management of Crohn's disease by improving quality of life and reducing hospitalization and surgery. Although all 3 approved anti-TNF agents available in the United States were superior to placebo for maintenance of remission, important differences were observed between agents when used as monotherapy (eg, adalimumab was superior to certolizumab). Thus, the indirect evidence suggests that the efficacy of the earlier-described agents varies despite being in the same therapeutic class. This may not be surprising because 2 previous anti-TNF agents (etanercept and CDP571) were shown to be inefficacious in the treatment of Crohn's disease.
A previous network meta-analysis comparing anti-TNF monotherapies in induction and maintenance of remission showed superior efficacy of adalimumab compared with certolizumab for induction; however, anti-TNF therapies had similar efficacy for maintenance of remission. In contrast, our analysis showed the superiority of adalimumab vs certolizumab for both induction and maintenance. The difference in findings likely is explained by the greater number of trials included in our network meta-analysis, which increases the certainty of indirect comparisons.
Our results contribute to the evolving understanding of the role of combining immunosuppressants with anti-TNF therapy. In 2010, the Study of Biologic and Immunomodulator Naive Patients in Crohn's Disease (SONIC) study showed that infliximab + azathioprine was superior to each drug in isolation. In our analysis, the combination of infliximab + azathioprine was superior to certolizumab and infliximab monotherapy, but was not different from adalimumab for maintenance of remission. Although studies have not combined immunosuppressants with certolizumab or adalimumab, this combination is presumed to be beneficial owing to reduced immunogenicity. However, a meta-analysis of placebo-controlled trials with individual patient-level data showed that concomitant immunosuppressants with infliximab, but not adalimumab or certolizumab, was associated with improved clinical remission rates. Thus, replicating the SONIC study design with adalimumab and/or certolizumab would provide valuable clinical direction.
With the recent approval of vedolizumab for induction and maintenance therapy for Crohn's disease, gastroenterologists will need to decide the sequence of prescribing vedolizumab relative to immunosuppressants and anti-TNF therapies. In the network meta-analysis, vedolizumab was significantly superior to placebo for induction and maintenance of remission; however, the indirect comparison showed that vedolizumab was not different from immunosuppressants or anti-TNF therapies. Thus, the position of prescribing vedolizumab will need to balance its comparative efficacy against its potential to reduce adverse events through its gut-selective immunosuppressing properties.
Crohn's disease patients were significantly more likely to have an adverse event leading to trial withdrawal with azathioprine/6-mercaptopurine or methotrexate relative to placebo, most anti-TNF monotherapies, and vedolizumab. In contrast, we observed the fewest total withdrawals and withdrawals secondary to adverse events with adalimumab and infliximab + azathioprine. Vedolizumab was associated with fewer WDAEs when compared with infliximab and infliximab + azathioprine, but not compared with adalimumab and certolizumab. However, these data should be reviewed cautiously because randomized controlled trials have insufficient power to detect small but important differences between drugs for rare adverse events. Furthermore, because adverse events were rare we used a fixed-effects model, which can underestimate the degree of uncertainty. Finally, some withdrawals may have been secondary to Crohn's disease flares rather than true adverse events of treatment, thus reflecting a lack of efficacy rather than increased toxicity.
In network meta-analyses, comparisons are made between differences of comparators (ie, odds ratios) and, thus, the results may depend on the baseline placebo response. This is important in Crohn's disease clinical trials because of the variability of placebo response that has been observed between clinical trials. We conducted a meta-regression to adjust for the baseline placebo response rate. The meta-regression was inconsistent with our primary analysis in some situations (eg, infliximab vs azathioprine/6-mercaptopurine). However, the meta-regression should be interpreted cautiously because our meta-regression included few trials, and in any meta-regression (traditional or network) the variable evaluated in the model (eg, baseline placebo response) is not distributed randomly across the trials and therefore is prone to confounding by other trialspecific characteristics.
Our results should be interpreted in the context of limitations associated with network meta-analyses. The paucity of head-to-head trials and reliance on only indirect evidence (eg, for infliximab + methotrexate comparisons) resulted in less precise credible intervals. Also, the assumption made when pooling studies in a traditional meta-analysis is that effect modifiers are balanced across trials. With a network meta-analysis, the same assumption is made, but it is extended across different treatment comparisons. Another concern with any meta-analysis is heterogeneity across trials. The trials in our study differed in several aspects, including the disease severity of the patient populations, risk of bias, disease severity at the time of randomization, prior exposure to anti-TNF therapy, and primary end points. Although heterogeneity between clinical trials merits cautious interpretation, sensitivity analyses only showed marginal differences in the results. We also found consistency between the treatment effects in the network meta-analysis and those observed in a direct (traditional) meta-analysis, when direct evidence was available, although there were few closed loops in the evidence network (Figure 2).
The generalizability of our results is limited to the eligible populations enrolled in the included trials. For example, randomized controlled trials typically recruit Crohn's disease patients with moderate to severe disease activity and, thus, our interpretations do not necessarily relate to patients with mild Crohn's disease or severe flares requiring hospitalizations. Furthermore, treatment approaches for Crohn's disease may differ based on age, disease location, disease behavior, prior surgery, and smoking, which could not be explored using the data available in the clinical trials. Similarly, we were not able to evaluate randomized controlled trials that compared strategies of treating Crohn's disease such as comparing early use of anti-TNF therapy (ie, top-down approach) with incremental use of immunosuppressants and anti-TNF agents (step-up approach).
Discussion
We conducted a systematic review and network meta-analysis of randomized controlled trials of immunosuppressants and biologics for induction and maintenance of remission in adults with Crohn's disease. Our systematic review identified a paucity of head-to-head trials in the Crohn's disease literature, which are necessary to inform clinicians on the appropriate efficacious and safe treatment regimens for Crohn's disease. By applying a network meta-analysis approach, our study integrated direct head-to-head data with indirect evidence to provide the most robust data available for inducing and maintaining remission in Crohn's disease. The purpose of comparative effectiveness research is to "assist consumers, clinicians, purchasers, and policy makers to make informed decisions that will improve healthcare at both the individual and population levels." Until more head-to-head trials are completed, these data are the best we have available to guide clinical decision making within and between classes of drug therapies.
Azathioprine/6-mercaptopurine and methotrexate were not different from placebo for induction of remission in Crohn's disease. This finding is consistent with clinical practice, whereby immunosuppressants are not prescribed as monotherapy for induction of remission. Infliximab, infliximab + azathioprine, adalimumab, and vedolizumab were superior to placebo in inducing remission. In pairwise comparisons, infliximab + azathioprine and adalimumab showed superiority to certolizumab. However, induction studies should be compared carefully because of intrinsic differences in study designs. For example, the time point in defining remission for induction trials varied from 1 to 4 months.
For several decades, azathioprine/6-mercaptopurine have been the primary treatment for maintenance of remission of moderate to severe Crohn's disease. Although this network meta-analysis confirms that azathioprine/6-mercaptopurine are superior to placebo for maintenance of remission, its efficacy was inferior to adalimumab, infliximab, and infliximab + azathioprine. In this network meta-analysis, methotrexate was not different from azathioprine/6-mercaptopurine or any anti-TNF agent. Although these data suggest that methotrexate may be an efficacious primary treatment option for Crohn's disease, the results should be interpreted cautiously because the credible intervals were wide. The uncertainty around the estimates was the result of the small sample size of methotrexate clinical trials. Furthermore, 2 of the 5 trials were at high risk of bias and therefore were excluded from our primary analysis. In addition, among the 3 trials included in the primary analysis, the dose of methotrexate used was mixed (oral vs subcutaneous dose range, 12.5–25 mg). Nonetheless, this network meta-analysis at least supports a direct head-to-head trial of conventionally dosed methotrexate vs azathioprine/6-mercaptopurine with or without an additional anti-TNF arm to further inform this important therapeutic choice.
The introduction of anti-TNF therapy has had a significant impact on the management of Crohn's disease by improving quality of life and reducing hospitalization and surgery. Although all 3 approved anti-TNF agents available in the United States were superior to placebo for maintenance of remission, important differences were observed between agents when used as monotherapy (eg, adalimumab was superior to certolizumab). Thus, the indirect evidence suggests that the efficacy of the earlier-described agents varies despite being in the same therapeutic class. This may not be surprising because 2 previous anti-TNF agents (etanercept and CDP571) were shown to be inefficacious in the treatment of Crohn's disease.
A previous network meta-analysis comparing anti-TNF monotherapies in induction and maintenance of remission showed superior efficacy of adalimumab compared with certolizumab for induction; however, anti-TNF therapies had similar efficacy for maintenance of remission. In contrast, our analysis showed the superiority of adalimumab vs certolizumab for both induction and maintenance. The difference in findings likely is explained by the greater number of trials included in our network meta-analysis, which increases the certainty of indirect comparisons.
Our results contribute to the evolving understanding of the role of combining immunosuppressants with anti-TNF therapy. In 2010, the Study of Biologic and Immunomodulator Naive Patients in Crohn's Disease (SONIC) study showed that infliximab + azathioprine was superior to each drug in isolation. In our analysis, the combination of infliximab + azathioprine was superior to certolizumab and infliximab monotherapy, but was not different from adalimumab for maintenance of remission. Although studies have not combined immunosuppressants with certolizumab or adalimumab, this combination is presumed to be beneficial owing to reduced immunogenicity. However, a meta-analysis of placebo-controlled trials with individual patient-level data showed that concomitant immunosuppressants with infliximab, but not adalimumab or certolizumab, was associated with improved clinical remission rates. Thus, replicating the SONIC study design with adalimumab and/or certolizumab would provide valuable clinical direction.
With the recent approval of vedolizumab for induction and maintenance therapy for Crohn's disease, gastroenterologists will need to decide the sequence of prescribing vedolizumab relative to immunosuppressants and anti-TNF therapies. In the network meta-analysis, vedolizumab was significantly superior to placebo for induction and maintenance of remission; however, the indirect comparison showed that vedolizumab was not different from immunosuppressants or anti-TNF therapies. Thus, the position of prescribing vedolizumab will need to balance its comparative efficacy against its potential to reduce adverse events through its gut-selective immunosuppressing properties.
Crohn's disease patients were significantly more likely to have an adverse event leading to trial withdrawal with azathioprine/6-mercaptopurine or methotrexate relative to placebo, most anti-TNF monotherapies, and vedolizumab. In contrast, we observed the fewest total withdrawals and withdrawals secondary to adverse events with adalimumab and infliximab + azathioprine. Vedolizumab was associated with fewer WDAEs when compared with infliximab and infliximab + azathioprine, but not compared with adalimumab and certolizumab. However, these data should be reviewed cautiously because randomized controlled trials have insufficient power to detect small but important differences between drugs for rare adverse events. Furthermore, because adverse events were rare we used a fixed-effects model, which can underestimate the degree of uncertainty. Finally, some withdrawals may have been secondary to Crohn's disease flares rather than true adverse events of treatment, thus reflecting a lack of efficacy rather than increased toxicity.
In network meta-analyses, comparisons are made between differences of comparators (ie, odds ratios) and, thus, the results may depend on the baseline placebo response. This is important in Crohn's disease clinical trials because of the variability of placebo response that has been observed between clinical trials. We conducted a meta-regression to adjust for the baseline placebo response rate. The meta-regression was inconsistent with our primary analysis in some situations (eg, infliximab vs azathioprine/6-mercaptopurine). However, the meta-regression should be interpreted cautiously because our meta-regression included few trials, and in any meta-regression (traditional or network) the variable evaluated in the model (eg, baseline placebo response) is not distributed randomly across the trials and therefore is prone to confounding by other trialspecific characteristics.
Our results should be interpreted in the context of limitations associated with network meta-analyses. The paucity of head-to-head trials and reliance on only indirect evidence (eg, for infliximab + methotrexate comparisons) resulted in less precise credible intervals. Also, the assumption made when pooling studies in a traditional meta-analysis is that effect modifiers are balanced across trials. With a network meta-analysis, the same assumption is made, but it is extended across different treatment comparisons. Another concern with any meta-analysis is heterogeneity across trials. The trials in our study differed in several aspects, including the disease severity of the patient populations, risk of bias, disease severity at the time of randomization, prior exposure to anti-TNF therapy, and primary end points. Although heterogeneity between clinical trials merits cautious interpretation, sensitivity analyses only showed marginal differences in the results. We also found consistency between the treatment effects in the network meta-analysis and those observed in a direct (traditional) meta-analysis, when direct evidence was available, although there were few closed loops in the evidence network (Figure 2).
The generalizability of our results is limited to the eligible populations enrolled in the included trials. For example, randomized controlled trials typically recruit Crohn's disease patients with moderate to severe disease activity and, thus, our interpretations do not necessarily relate to patients with mild Crohn's disease or severe flares requiring hospitalizations. Furthermore, treatment approaches for Crohn's disease may differ based on age, disease location, disease behavior, prior surgery, and smoking, which could not be explored using the data available in the clinical trials. Similarly, we were not able to evaluate randomized controlled trials that compared strategies of treating Crohn's disease such as comparing early use of anti-TNF therapy (ie, top-down approach) with incremental use of immunosuppressants and anti-TNF agents (step-up approach).
