Opportunistic Infections With Anti-TNF Therapy in IBD
Opportunistic Infections With Anti-TNF Therapy in IBD
The search strategy identified 20,563 citations, 20,534 of which were excluded after examining the title and abstract (Figure 1). Twenty-nine articles reporting on the efficacy of anti-TNFα therapies in IBD were retrieved and evaluated in more detail. Eight of these were excluded, three because they used only a single infusion of anti-TNFα therapy, three because they were induction of remission trials in which some participants went on to be enrolled in a subsequent maintenance of remission trial that was also eligible for inclusion, and two because there was no placebo arm.
(Enlarge Image)
Figure 1.
Flow diagram of assessment of studies identified in the systematic review. CD, Crohn's disease; TNFα, tumor necrosis factor-α; UC, ulcerative colitis.
This left 21 articles eligible for inclusion, reporting on 22 separate RCTs. Of these, 15 articles studied the efficacy of anti-TNFα therapies in CD and 6 in UC, one of which reported two separate trials. Duration of follow-up ranged between 2 and 56 weeks. Two of the studies randomized patients to receive active azathioprine and placebo infliximab infusions, placebo azathioprine and active infliximab infusions, or active azathioprine and active infliximab infusions. For the purposes of our analysis we only extracted data from the active azathioprine and placebo infliximab infusions, and active azathioprine and active infliximab arms of these trials. Detailed trial characteristics are provided in Supplementary Table S1 online. Only four of the RCTs we identified were judged as being at low risk of bias. Individual risk values of bias items for each trial are provided in Supplementary Table S2 online.
Rates of individual opportunistic infections within each trial arm are detailed in Table 1 . Among patients receiving active therapy these included eight cases of Mycobacterium tuberculosis, eight cases of herpes simplex infection, six cases of oral or esophageal candidiasis, six cases of herpes zoster infection, two cases of varicella-zoster virus infection, two cases of cytomegalovirus or Epstein-Barr virus infection, and one case of Nocardia infection.
The 22 trials contained a total of 7,054 patients with IBD. Of these, 4,135 (58.6%) were randomized to receive anti-TNFα therapy, and 2,919 to placebo. In total, there were 39 (0.9%) opportunistic infections among patients allocated to anti-TNFα therapy, compared with 9 (0.3%) among those assigned to placebo. The RR of developing an opportunistic infection was significantly higher with anti-TNFα therapy than with placebo (2.05; 95% CI 1.10–3.85) (Figure 2), with no statistically significant heterogeneity detected between studies (I=0%, P=0.99), and no evidence of funnel plot asymmetry (Egger test, P=0.20). The NNH with anti-TNFα therapy to cause one opportunistic infection was 500 (95% CI 200–1,567). Only one of the patients experiencing opportunistic infection died. This individual received golimumab as part of the PURSUIT-Maintenance trial, and the cause of death was disseminated tuberculosis.
(Enlarge Image)
Figure 2.
Forest plot of opportunistic infections in randomized controlled trials of anti-tumor necrosis factor-α (TNFα) therapy vs. placebo in inflammatory bowel disease.
We conducted a priori subgroup analyses ( Table 2 ). The RR of opportunistic infections with anti-TNFα therapy was higher among trials conducted in CD (2.34; 95% CI 0.98–5.57) compared with UC (1.78; 95% CI 0.72–4.42), but this was not statistically significant (Cochran Q=0.18, P=0.67). The RR of opportunistic infections did not appear to differ according to the anti-TNFα agent used (Cochran Q=0.31, P=0.96). When duration of therapy was examined, the RR was also higher in trials with <8 weeks of treatment (5.09; 95% CI 1.13–23.0), compared with those of >8 weeks duration (1.70; 95% CI 0.85–3.39), but again this was not statistically significant (Cochran Q=1.68, P=0.19). There were no significant differences in RR of opportunistic infection according to exposure to immunosuppressants, or anti-TNFα therapy during induction of remission (Cochran Q=0.06, P=0.80 and Cochran Q=0.03, P=0.85 respectively). The RR of opportunistic infection was no higher in the eight trials that stratified their randomization process according to glucocorticosteroid usage, or in which all patients were receiving concomitant glucocorticosteroids (2.18; 95% CI 0.81–5.87), compared with those in which there were potential inequalities in glucocorticosteroid use between the arms (1.97; 95% CI 0.88–4.44) (Cochran Q=0.02, P=0.88). Finally, in RCTs that were judged as being at low risk of bias, the RR was higher (5.09; 95% CI 1.13–23.0) than in trials at unclear risk of bias (1.70; 95% CI 0.85–3.39), but again this difference was not statistically significant (Cochran Q=1.68, P=0.19).
Oral or esophageal candidiasis and herpes zoster infection can be considered as less severe opportunistic infections. When we excluded these from the definition of opportunistic infection, there were 27 (0.65%) opportunistic infections among patients allocated to anti-TNFα therapy, compared with 7 (0.2%) among those assigned to placebo. The RR of opportunistic infection was no longer significantly higher with anti-TNFα therapy, although it approached statistical significance (RR=1.95; 95% CI 0.97–3.90) (Figure 3), with no statistically significant heterogeneity detected between studies (I=0%, P=1.00) and no evidence of funnel plot asymmetry (Egger test, P=0.46).
(Enlarge Image)
Figure 3.
Forest plot of more serious opportunistic infections in randomized controlled trials of anti-tumor necrosis factor-α (TNFα) therapy vs. placebo in inflammatory bowel disease.
In total, there were 8 (0.2%) cases of Mycobacterium tuberculosis infection among patients allocated to anti-TNFα therapy, compared with none among those assigned to placebo. The RR of developing Mycobacterium tuberculosis infection was not significantly higher with anti-TNFα therapy than with placebo (2.52; 95% CI 0.62–10.21), again with no heterogeneity between studies (I=0%, P=0.98) and no evidence of funnel plot asymmetry (Egger test, P=0.16).
Results
The search strategy identified 20,563 citations, 20,534 of which were excluded after examining the title and abstract (Figure 1). Twenty-nine articles reporting on the efficacy of anti-TNFα therapies in IBD were retrieved and evaluated in more detail. Eight of these were excluded, three because they used only a single infusion of anti-TNFα therapy, three because they were induction of remission trials in which some participants went on to be enrolled in a subsequent maintenance of remission trial that was also eligible for inclusion, and two because there was no placebo arm.
(Enlarge Image)
Figure 1.
Flow diagram of assessment of studies identified in the systematic review. CD, Crohn's disease; TNFα, tumor necrosis factor-α; UC, ulcerative colitis.
This left 21 articles eligible for inclusion, reporting on 22 separate RCTs. Of these, 15 articles studied the efficacy of anti-TNFα therapies in CD and 6 in UC, one of which reported two separate trials. Duration of follow-up ranged between 2 and 56 weeks. Two of the studies randomized patients to receive active azathioprine and placebo infliximab infusions, placebo azathioprine and active infliximab infusions, or active azathioprine and active infliximab infusions. For the purposes of our analysis we only extracted data from the active azathioprine and placebo infliximab infusions, and active azathioprine and active infliximab arms of these trials. Detailed trial characteristics are provided in Supplementary Table S1 online. Only four of the RCTs we identified were judged as being at low risk of bias. Individual risk values of bias items for each trial are provided in Supplementary Table S2 online.
Details of Individual Opportunistic Infections
Rates of individual opportunistic infections within each trial arm are detailed in Table 1 . Among patients receiving active therapy these included eight cases of Mycobacterium tuberculosis, eight cases of herpes simplex infection, six cases of oral or esophageal candidiasis, six cases of herpes zoster infection, two cases of varicella-zoster virus infection, two cases of cytomegalovirus or Epstein-Barr virus infection, and one case of Nocardia infection.
Overall Risk of Opportunistic Infections With Anti-TNFα Therapy vs. Placebo in IBD
The 22 trials contained a total of 7,054 patients with IBD. Of these, 4,135 (58.6%) were randomized to receive anti-TNFα therapy, and 2,919 to placebo. In total, there were 39 (0.9%) opportunistic infections among patients allocated to anti-TNFα therapy, compared with 9 (0.3%) among those assigned to placebo. The RR of developing an opportunistic infection was significantly higher with anti-TNFα therapy than with placebo (2.05; 95% CI 1.10–3.85) (Figure 2), with no statistically significant heterogeneity detected between studies (I=0%, P=0.99), and no evidence of funnel plot asymmetry (Egger test, P=0.20). The NNH with anti-TNFα therapy to cause one opportunistic infection was 500 (95% CI 200–1,567). Only one of the patients experiencing opportunistic infection died. This individual received golimumab as part of the PURSUIT-Maintenance trial, and the cause of death was disseminated tuberculosis.
(Enlarge Image)
Figure 2.
Forest plot of opportunistic infections in randomized controlled trials of anti-tumor necrosis factor-α (TNFα) therapy vs. placebo in inflammatory bowel disease.
We conducted a priori subgroup analyses ( Table 2 ). The RR of opportunistic infections with anti-TNFα therapy was higher among trials conducted in CD (2.34; 95% CI 0.98–5.57) compared with UC (1.78; 95% CI 0.72–4.42), but this was not statistically significant (Cochran Q=0.18, P=0.67). The RR of opportunistic infections did not appear to differ according to the anti-TNFα agent used (Cochran Q=0.31, P=0.96). When duration of therapy was examined, the RR was also higher in trials with <8 weeks of treatment (5.09; 95% CI 1.13–23.0), compared with those of >8 weeks duration (1.70; 95% CI 0.85–3.39), but again this was not statistically significant (Cochran Q=1.68, P=0.19). There were no significant differences in RR of opportunistic infection according to exposure to immunosuppressants, or anti-TNFα therapy during induction of remission (Cochran Q=0.06, P=0.80 and Cochran Q=0.03, P=0.85 respectively). The RR of opportunistic infection was no higher in the eight trials that stratified their randomization process according to glucocorticosteroid usage, or in which all patients were receiving concomitant glucocorticosteroids (2.18; 95% CI 0.81–5.87), compared with those in which there were potential inequalities in glucocorticosteroid use between the arms (1.97; 95% CI 0.88–4.44) (Cochran Q=0.02, P=0.88). Finally, in RCTs that were judged as being at low risk of bias, the RR was higher (5.09; 95% CI 1.13–23.0) than in trials at unclear risk of bias (1.70; 95% CI 0.85–3.39), but again this difference was not statistically significant (Cochran Q=1.68, P=0.19).
Risk of Opportunistic Infections (Excluding Mild Infections) With Anti-TNFα Therapy vs. Placebo in IBD
Oral or esophageal candidiasis and herpes zoster infection can be considered as less severe opportunistic infections. When we excluded these from the definition of opportunistic infection, there were 27 (0.65%) opportunistic infections among patients allocated to anti-TNFα therapy, compared with 7 (0.2%) among those assigned to placebo. The RR of opportunistic infection was no longer significantly higher with anti-TNFα therapy, although it approached statistical significance (RR=1.95; 95% CI 0.97–3.90) (Figure 3), with no statistically significant heterogeneity detected between studies (I=0%, P=1.00) and no evidence of funnel plot asymmetry (Egger test, P=0.46).
(Enlarge Image)
Figure 3.
Forest plot of more serious opportunistic infections in randomized controlled trials of anti-tumor necrosis factor-α (TNFα) therapy vs. placebo in inflammatory bowel disease.
Risk of Mycobacterium tuberculosis Infection With Anti-TNFα Therapy vs. Placebo in IBD
In total, there were 8 (0.2%) cases of Mycobacterium tuberculosis infection among patients allocated to anti-TNFα therapy, compared with none among those assigned to placebo. The RR of developing Mycobacterium tuberculosis infection was not significantly higher with anti-TNFα therapy than with placebo (2.52; 95% CI 0.62–10.21), again with no heterogeneity between studies (I=0%, P=0.98) and no evidence of funnel plot asymmetry (Egger test, P=0.16).
Source...