After a preliminary screening through browsing the abstracts, 157 articles were obtained. Of 157 potentially relevant articles, 118 were excluded, including 56 articles with improper control group, 18 non-RCTs articles, 13 articles with a lack of infection information, 14 articles with short follow-up time, 12 repeated studies, 3 articles with subjects under 18, 1 article with unknown classifications of infections. A total of 34 RCTs (39 articles) were eligible for inclusion. Flow diagram of including RCTs is as Figure1(Page27).
Summary of characteristics of included RCTs
Of the 34 RCTs, 11 infliximab trails[2–9,17–20], 7 etanercept trails[21–29] and 16 adalimumab trails[30–46] were included, with a total of 14,166 subjects. Among subjects, 8,758 in the experimental group and 5,408 in the control group were followed up between 14weeks–104weeks. The basic characteristics are shown in Table1[Page30], the incidence of severe infections and the detailed infections is shown in Appendix1 [see Additional file 1].
Risk of bias
Risk of bias graph is as Appendix2 [see Additional file 1], showing no high risk of bias in any of the bias item. Risk of bias summary is shown in Appendix3 [see Additional file 1].
Statistical analysis results
The funnel plot of the included studies is shown in Appendix4 [see Additional file 1], showing no obvious asymmetry and indicating no obvious publication bias.
The forest plot about the risk of serious infections of the 34 included studies is shown in Figure2[Page28]. The value of inconsistency between trails was 0% (p = 0.59), indicating that studies were not statistically heterogeneous; we can use the fixed effect model to combine the effect size OR. Results showed that subjects using TNF-α inhibitors had a higher risk of serious infections than the control group, with a serious infection rate of 3.4% and 2.5%, respectively. The results were statistically significant with the pooled OR of 1.29 (95%CI1.04–1.60, p = 0.02).
Comparing the incidence rate of serious infections in infliximab group (4.7%) or adalimumab group(2.7%) to control group(3.2% or 1.8%), the differences were statistically significant with the pooled OR of 1.48(95%CI 1.05–2.08, p = 0.03) and 1.47 (95%CI 1.04–2.07, p = 0.03), respectively. The incidence rate of serious infections in etanercept group(2.2%) was compared with that in the control group (3.0%), and the results were not statistically significant with the pooled OR of 0.77 (95%CI 0.48–1.23, p = 0.28).
A total of 18 tuberculosis events were reported in 10 of the 34 included studies (Figure3) [Page28]. Subjects in these studies had been initially screened for a history of tuberculosis through the PPD skin test or chest X-ray. Only one of the 18 subjects with tuberculosis was in control group, and the remaining 17 subjects were all subjects of the experimental group. The consistency test showed a low heterogeneity, but the difference was not statistically significant with a pooled OR of 2.31 (95%CI 0.90–5.94, p = 0.08).
Of the 34 included studies, 20 reported pneumonia with a pooled OR of 1.72 (95%CI 1.04–2.83, p = 0.03)(Figure4) [Page29]. Subgroup analysis showed that only infliximab subgroup had statistically significant. In the 42 cases of pneumonia, only 1 case was from the control group, and the combined OR was 4.48(95%CI 1.66–12.09, p = 0.003), which meant that infliximab increased the risk of pneumonia by 4.48 times. The combined OR of Etanercept and adalimumab groups was 1.52(p = 0.43) and 0.65(p = 0.30), respectively.
Upper Respiratory Infection
There were 11 RCTs reporting upper respiratory infection (Figure5) [Page29], The incidence rate of it between experimental group and control group were 14.3% and 9.1%, respectively. TNF-α inhibitors were shown to mildly increase the risk of upper respiratory infection with a combined OR of 1.38(95%CI 1.10–1.73, p = 0.006).
A total of 16 studies reported nasopharyngitis (Figure6) [Page30]. Forest plot showed that TNF-α inhibitors mildly increased the infection rate of nasopharyngitis, with pooled OR of 1.33(95%CI 1.11–1.60, p = 0.002). According to the subgroup analysis, etanercept showed no statistical significance (OR1.23, p = 0.49), and the Odds ratio of adalimumab was at a critical value (OR1.26, p = 0.05). Conservative results of adalimumab subgroup were obtained using the random effect model (OR1.25, p = 0.09). Finally, only infliximab subgroup showed statistical significance with the pooled OR of 1.58(95%CI 1.08–2.31, p = 0.02). What should be noted is that the results of two adalimumab RCTs[2,32] were statistically significant, with OR of 4.04(p = 0.01) and 1.77(p = 0.04), respectively.
A total of 11 studies reported bronchitis(Appendix5) [see Additional file 1], and the incidence rate of bronchitis in experimental group(1.1%) was compared with that in the control group(0.7%), but the results were not statistically significant with a combined OR of 1.39(95%CI 0.76–2.56, p = 0.29).
5 RCTs reported sinusitis (Appendix6) [see Additional file 1]. Due to the high heterogeneity (I2 = 61%), a random effect model was used and the pooled OR was 1.05(95%CI 0.41–2.65, p = 0.93). It is important to noted that one RCT[5,18,19] showed statistical significance with the OR of 3.41 (p = 0.01).
A total of 7 studies reported gastroenteritis (Appendix7) [see Additional file 1], but the results were not statistically significant with a combined OR of 1.65(95%CI 0.56–4.83, p = 0.36). Meanwhile, none of the included RCTs was statistically significant.
Urinary Tract Infection
Of the 34 included studied, 6 studies reported urinary tract infection (Appendix8) [see Additional file 1]. The incidence rate were 1.3% and 0.6% among experimental and control group, with a pooled OR of 1.81(95%CI 0.87–3.74, p = 0.11).
5 studies reported cellulitis (Appendix9) [see Additional file 1]. The results were not statistically significant with the pooled OR of 2.62(95%CI 0.65–10.50, p = 0.17). However, since all the 11 reported cellulitis cases were from TNF-α inhibitor group, it could not be ignored that TNF-α may increase the risk of cellulitis.
A total of 4 studies reported sepsis (Appendix10) [see Additional file 1], but the results were not statistically significant with a pooled OR of 1.19(95%CI 0.38–3.70, p = 0.76).
A total of 4 studies reported abscess (Appendix11) [see Additional file 1], but the results were not statistically significant with a pooled OR of 0.77(95%CI 0.35–1.66, p = 0.50).
The information of infections reported in 34 studies was extracted for meta-analysis. Results showed that the application of TNF-α inhibitors lead to increased risk of pneumonia, upper respiratory infection and nasopharyngitis. Odds ratios and confidence intervals were shown in Table2.