The Efficacy of Anti-TNF Agents for Ulcerative Colitis
The Efficacy of Anti-TNF Agents for Ulcerative Colitis
A flow diagram depicting the search and selection process is provided in Figure 1. Initial searches of the Medline and Embase databases yielded 486 citations. A manual search of the PubMed.gov database for pertinent systematic reviews, meta-analyses and guidelines identified four summary documents, (a review of which yielded 20 additional citations), totalling 506 citations. Title review of these two groups of citations yielded 376 unique potentially relevant articles. Abstract and/or brief manuscript review of these articles yielded 12 manuscripts appropriate for detailed evaluation. Seven of the remaining manuscripts were included in the final analysis. Note that there were six individual manuscripts included, but Rutgeerts et al. 2005 reported the results of two separate randomised, controlled trials and therefore each trial was included as a separate study. The remaining five articles were excluded because they did not meet eligibility criteria, as detailed in the flow diagram. There was 100% agreement between reviewers regarding final study selection.
(Enlarge Image)
Figure 1.
Flow diagram of the search strategy and included studies in the analysis.
Characteristics of the included studies are listed in Table 1 and efficacy data in Table 2. The seven studies in six manuscripts meeting eligibility criteria included a total of 1823 subjects for induction and 1070 subjects for maintenance. No comparative efficacy studies were identified; all included studies compared placebo to various anti-TNF therapies. Three studies compared infliximab to placebo. Among these, three studies and two studies evaluated remission and response, respectively as an endpoint for induction (n = 529) and one study evaluated remission and response as an endpoint for maintenance (n = 242). Two studies compared adalimumab to placebo, of which both evaluated remission and response as an endpoint for induction (n = 778) and 1 evaluated remission and response as an endpoint for maintenance (n = 518). Two studies compared golimumab to placebo, of which 1 evaluated remission and response as an endpoint for induction (n = 516) and the other evaluated remission and response as an endpoint for maintenance (n = 310).
Pooled analysis of the effects of infliximab, adalimumab and golimumab on induction (remission and response) and maintenance (remission and response) demonstrated significant statistical heterogeneity in only one situation, i.e. the induction of remission among infliximab studies (I = 72.8%, P = 0.025). Meta-influence analysis and visual inspection of Galbraith plots revealed that Rutgeerts 2005 (ACT 2) was responsible for this heterogeneity and was likely due to the high placebo rates.
Induction of Remission and Response. Compared to placebo, traditional meta-analysis revealed that anti-TNF agents result in a 2.45-fold higher likelihood of induction of remission and 1.65-fold higher likelihood of induction of response compared to placebo (RR: 2.45, 95% CI: 1.72–3.47 and RR: 1.65, 95% CI: 1.37–1.99 respectively). Individually, infliximab resulted in a 2.76-fold higher likelihood of inducing remission and 2-fold higher likelihood of inducing response compared to placebo (RR: 2.76, 95% CI: 1.29–5.90 and RR: 2.00, 95% CI: 1.64–2.44 respectively). Adalimumab resulted in a 1.87-fold higher likelihood of inducing remission and 1.36-fold higher likelihood of inducing response compared to placebo (RR: 1.87, 95% CI: 1.27–2.75 and RR: 1.36, 95% CI: 1.13–1.64 respectively). Golimumab resulted in a 3-fold higher likelihood of inducing remission and 1.75-fold higher likelihood of inducing response compared to placebo (RR: 3.00, 95% CI: 1.75–5.14 and RR: 1.75, 95% CI: 1.40–2.19 respectively). (Figure 2a, b).
(Enlarge Image)
Figure 2.
Panel: Meta-analysis of anti-TNFs for the treatment of ulcerative colitis. (a) Meta-analysis of the induction of remission endpoint. (b) Meta-analysis of the induction of response endpoint. (c) Meta-analysis of the maintenance of remission endpoint. (d) Meta-analysis of the maintenance of response endpoint.
Network meta-analysis of agents for induction of remission did not show significant differences between agents (RR: 2.08 for infliximab vs. adalimumab, 95% CrI: 0.32–12.03; RR: 1.18 for infliximab vs. golimumab, 95% CrI: 0.13–10.63; and RR: 1.75 for adalimumab vs. golimumab, 95% CrI: 0.17–16.86). While trends favoured infliximab over both adalimumab and golimumab, the 95% credible interval crossed 1 and was therefore not statistically significant. Infliximab was ranked the most effective drug in 55% of the simulations, while golimumab was favoured in 36% and adalimumab in 9%.
Similarly, the network meta-analysis of agents for the induction of response demonstrated no statistically significant differences between agents (RR: 2.15 for infliximab vs. adalimumab, 95% CrI: 0.73–5.80; RR: 1.48 for infliximab vs. golimumab, 95% CrI: 0.38–4.69; and RR: 1.46 for golimumab vs. adalimumab, 95% CrI: 0.42–5.38). Again, trends favoured infliximab over both adalimumab and golimumab; however, the 95% credible interval crossed 1 and was therefore not statistically significant. Infliximab was ranked the most effective drug in 80% of the simulations, while golimumab was favoured in 17% and adalimumab in 3%.
Compared to placebo, traditional meta-analysis revealed that anti-TNF agents result in a 2.00-fold higher likelihood of maintenance of remission and 1.76-fold higher likelihood of maintenance of response compared to placebo (RR: 2.00, 95% CI: 1.52–2.62 and RR: 1.76, 95% CI: 1.46–2.14 respectively). Individually, infliximab resulted in a 2.10-fold higher likelihood of maintaining remission and 2.29-fold higher likelihood of maintaining response compared to placebo (RR: 2.10, 95% CI: 1.31–3.36 and RR: 2.29, 95% CI: 1.52–3.45 respectively). Adalimumab resulted in a 2.06-fold higher likelihood of maintaining remission and 1.68-fold higher likelihood of maintaining response compared to placebo (RR: 2.06, 95% CI: 1.26–3.38 and RR: 1.68, 95% CI: 1.21–2.33 respectively). Golimumab resulted in a 1.86-fold higher likelihood of maintaining remission and 1.61-fold higher likelihood of maintaining response compared to placebo (RR: 1.86, 95% CI: 1.19–2.90 and RR: 1.61, 95% CI: 1.22–2.13 respectively – Figure 2C, D).
Network meta-analysis of agents for maintenance of remission did not show significant difference between agents (RR: 1.18 for infliximab vs. adalimumab, 95% CrI: 0.19–8.02, RR: 1.22 for infliximab vs. golimumab, 95% CrI: 0.18–8.43, RR: 1.04 for adalimumab vs. golimumab, 95% CrI: 0.16–6.96). While trends favoured infliximab over both adalimumab and golimumab, the 95% credible interval crossed 1 and was therefore not statistically significant. Infliximab was ranked the most effective drug in 45% of the simulations, while adalimumab was favoured in 28% and golimumab in 26%.
Finally, the network meta-analysis of agents for the maintenance of response demonstrated no statistically significant difference between agents (RR: 1.70 for infliximab vs. adalimumab, 95% CrI: 0.17–16.59, RR: 1.47 for infliximab vs. golimumab, 95% CrI: 0.15–14.43, RR: 1.14 for adalimumab vs. golimumab, 95% CrI: 0.11–10.92). While trends favoured Infliximab over both adalimumab and golimumab, the 95% credible interval crossed 1 and was therefore not statistically significant. Infliximab was ranked the most effective drug in 61% of the simulations, while golimumab was favoured in 22% and adalimumab in 17%.
Using data generated by our NMA, the required sample sizes for direct comparative efficacy trials between anti-TNF agents were calculated (Table 3). Comparison of infliximab to adalimumab with 90% power would require 174 and 204 subjects to determine whether superiority exists between these agents for induction and maintenance respectively. Direct comparison of infliximab to golimumab would require 214 and 1870 subjects to determine superiority for induction and maintenance, respectively.
The funnel plot asymmetry test for publication bias was negative for all outcomes using the Harbord test (P = 0.44) for induction of remission; (P = 0.64) for induction of response; (P = 0.59) for maintenance of remission; (P = 0.27) for maintenance of response.
The sensitivity analyses did not substantively change the results. Specifically, excluding outlier studies [n = 1, Rutgeers 2005 (ACT 2)], studies that use the UCSS instrument to assess the outcome measure (n = 1, Probert 2003), and studies that continued maintenance therapy on those that responded to the induction therapy (n = 1, Sandborn et al.) in separate sensitivity traditional meta-analyses did not substantively affect composite treatment effects.
Results
Literature Search
A flow diagram depicting the search and selection process is provided in Figure 1. Initial searches of the Medline and Embase databases yielded 486 citations. A manual search of the PubMed.gov database for pertinent systematic reviews, meta-analyses and guidelines identified four summary documents, (a review of which yielded 20 additional citations), totalling 506 citations. Title review of these two groups of citations yielded 376 unique potentially relevant articles. Abstract and/or brief manuscript review of these articles yielded 12 manuscripts appropriate for detailed evaluation. Seven of the remaining manuscripts were included in the final analysis. Note that there were six individual manuscripts included, but Rutgeerts et al. 2005 reported the results of two separate randomised, controlled trials and therefore each trial was included as a separate study. The remaining five articles were excluded because they did not meet eligibility criteria, as detailed in the flow diagram. There was 100% agreement between reviewers regarding final study selection.
(Enlarge Image)
Figure 1.
Flow diagram of the search strategy and included studies in the analysis.
Characteristics of Included Studies
Characteristics of the included studies are listed in Table 1 and efficacy data in Table 2. The seven studies in six manuscripts meeting eligibility criteria included a total of 1823 subjects for induction and 1070 subjects for maintenance. No comparative efficacy studies were identified; all included studies compared placebo to various anti-TNF therapies. Three studies compared infliximab to placebo. Among these, three studies and two studies evaluated remission and response, respectively as an endpoint for induction (n = 529) and one study evaluated remission and response as an endpoint for maintenance (n = 242). Two studies compared adalimumab to placebo, of which both evaluated remission and response as an endpoint for induction (n = 778) and 1 evaluated remission and response as an endpoint for maintenance (n = 518). Two studies compared golimumab to placebo, of which 1 evaluated remission and response as an endpoint for induction (n = 516) and the other evaluated remission and response as an endpoint for maintenance (n = 310).
Testing for Heterogeneity Between Eligible Studies
Pooled analysis of the effects of infliximab, adalimumab and golimumab on induction (remission and response) and maintenance (remission and response) demonstrated significant statistical heterogeneity in only one situation, i.e. the induction of remission among infliximab studies (I = 72.8%, P = 0.025). Meta-influence analysis and visual inspection of Galbraith plots revealed that Rutgeerts 2005 (ACT 2) was responsible for this heterogeneity and was likely due to the high placebo rates.
Meta-analysis Results
Induction of Remission and Response. Compared to placebo, traditional meta-analysis revealed that anti-TNF agents result in a 2.45-fold higher likelihood of induction of remission and 1.65-fold higher likelihood of induction of response compared to placebo (RR: 2.45, 95% CI: 1.72–3.47 and RR: 1.65, 95% CI: 1.37–1.99 respectively). Individually, infliximab resulted in a 2.76-fold higher likelihood of inducing remission and 2-fold higher likelihood of inducing response compared to placebo (RR: 2.76, 95% CI: 1.29–5.90 and RR: 2.00, 95% CI: 1.64–2.44 respectively). Adalimumab resulted in a 1.87-fold higher likelihood of inducing remission and 1.36-fold higher likelihood of inducing response compared to placebo (RR: 1.87, 95% CI: 1.27–2.75 and RR: 1.36, 95% CI: 1.13–1.64 respectively). Golimumab resulted in a 3-fold higher likelihood of inducing remission and 1.75-fold higher likelihood of inducing response compared to placebo (RR: 3.00, 95% CI: 1.75–5.14 and RR: 1.75, 95% CI: 1.40–2.19 respectively). (Figure 2a, b).
(Enlarge Image)
Figure 2.
Panel: Meta-analysis of anti-TNFs for the treatment of ulcerative colitis. (a) Meta-analysis of the induction of remission endpoint. (b) Meta-analysis of the induction of response endpoint. (c) Meta-analysis of the maintenance of remission endpoint. (d) Meta-analysis of the maintenance of response endpoint.
Network meta-analysis of agents for induction of remission did not show significant differences between agents (RR: 2.08 for infliximab vs. adalimumab, 95% CrI: 0.32–12.03; RR: 1.18 for infliximab vs. golimumab, 95% CrI: 0.13–10.63; and RR: 1.75 for adalimumab vs. golimumab, 95% CrI: 0.17–16.86). While trends favoured infliximab over both adalimumab and golimumab, the 95% credible interval crossed 1 and was therefore not statistically significant. Infliximab was ranked the most effective drug in 55% of the simulations, while golimumab was favoured in 36% and adalimumab in 9%.
Similarly, the network meta-analysis of agents for the induction of response demonstrated no statistically significant differences between agents (RR: 2.15 for infliximab vs. adalimumab, 95% CrI: 0.73–5.80; RR: 1.48 for infliximab vs. golimumab, 95% CrI: 0.38–4.69; and RR: 1.46 for golimumab vs. adalimumab, 95% CrI: 0.42–5.38). Again, trends favoured infliximab over both adalimumab and golimumab; however, the 95% credible interval crossed 1 and was therefore not statistically significant. Infliximab was ranked the most effective drug in 80% of the simulations, while golimumab was favoured in 17% and adalimumab in 3%.
Maintenance of Remission and Response
Compared to placebo, traditional meta-analysis revealed that anti-TNF agents result in a 2.00-fold higher likelihood of maintenance of remission and 1.76-fold higher likelihood of maintenance of response compared to placebo (RR: 2.00, 95% CI: 1.52–2.62 and RR: 1.76, 95% CI: 1.46–2.14 respectively). Individually, infliximab resulted in a 2.10-fold higher likelihood of maintaining remission and 2.29-fold higher likelihood of maintaining response compared to placebo (RR: 2.10, 95% CI: 1.31–3.36 and RR: 2.29, 95% CI: 1.52–3.45 respectively). Adalimumab resulted in a 2.06-fold higher likelihood of maintaining remission and 1.68-fold higher likelihood of maintaining response compared to placebo (RR: 2.06, 95% CI: 1.26–3.38 and RR: 1.68, 95% CI: 1.21–2.33 respectively). Golimumab resulted in a 1.86-fold higher likelihood of maintaining remission and 1.61-fold higher likelihood of maintaining response compared to placebo (RR: 1.86, 95% CI: 1.19–2.90 and RR: 1.61, 95% CI: 1.22–2.13 respectively – Figure 2C, D).
Network meta-analysis of agents for maintenance of remission did not show significant difference between agents (RR: 1.18 for infliximab vs. adalimumab, 95% CrI: 0.19–8.02, RR: 1.22 for infliximab vs. golimumab, 95% CrI: 0.18–8.43, RR: 1.04 for adalimumab vs. golimumab, 95% CrI: 0.16–6.96). While trends favoured infliximab over both adalimumab and golimumab, the 95% credible interval crossed 1 and was therefore not statistically significant. Infliximab was ranked the most effective drug in 45% of the simulations, while adalimumab was favoured in 28% and golimumab in 26%.
Finally, the network meta-analysis of agents for the maintenance of response demonstrated no statistically significant difference between agents (RR: 1.70 for infliximab vs. adalimumab, 95% CrI: 0.17–16.59, RR: 1.47 for infliximab vs. golimumab, 95% CrI: 0.15–14.43, RR: 1.14 for adalimumab vs. golimumab, 95% CrI: 0.11–10.92). While trends favoured Infliximab over both adalimumab and golimumab, the 95% credible interval crossed 1 and was therefore not statistically significant. Infliximab was ranked the most effective drug in 61% of the simulations, while golimumab was favoured in 22% and adalimumab in 17%.
Direct Comparison Sample Size Estimations
Using data generated by our NMA, the required sample sizes for direct comparative efficacy trials between anti-TNF agents were calculated (Table 3). Comparison of infliximab to adalimumab with 90% power would require 174 and 204 subjects to determine whether superiority exists between these agents for induction and maintenance respectively. Direct comparison of infliximab to golimumab would require 214 and 1870 subjects to determine superiority for induction and maintenance, respectively.
Publication Bias
The funnel plot asymmetry test for publication bias was negative for all outcomes using the Harbord test (P = 0.44) for induction of remission; (P = 0.64) for induction of response; (P = 0.59) for maintenance of remission; (P = 0.27) for maintenance of response.
Sensitivity Analyses
The sensitivity analyses did not substantively change the results. Specifically, excluding outlier studies [n = 1, Rutgeers 2005 (ACT 2)], studies that use the UCSS instrument to assess the outcome measure (n = 1, Probert 2003), and studies that continued maintenance therapy on those that responded to the induction therapy (n = 1, Sandborn et al.) in separate sensitivity traditional meta-analyses did not substantively affect composite treatment effects.
