Comparative efficacy of Chinese herbal injections combined with azithromycin for mycoplasma pneumonia in children: A Bayesian network meta‐analysis of randomized controlled trials

Abstract What is known and objective An increasing macrolide resistance leads to complex clinical treatment schemes in mycoplasma pneumonia in children. Chinese herbal injection (CHI) is widely used to treat it and may provide a new treatment regimen. This study was conducted to systematically evaluate the efficacy of CHIs combined with azithromycin for treating mycoplasma pneumonia in children by Bayesian network meta‐analysis. Methods Randomized controlled trials (RCTs) of CHIs combined with azithromycin for mycoplasma pneumonia in children were searched in electronic databases and related references from initiation to 30 October 2018. Two researchers conducted data extraction and risk of bias assessment. WinBUGS software and STATA software were adopted to analyse the data. Results A total of 167 RCTs were included with 5 CHIs involving 16 144 patients. All CHIs combined with azithromycin had superior effects than azithromycin only among overall outcomes. Yanhuning injection combined with azithromycin ranked highest in four different outcomes and second in two based on surface under the cumulative ranking probabilities (SUCRA). Meanwhile, the results of MD and 95% CIs of concerned outcomes indicated that only Yanhuning injection combined with azithromycin had better response than other CHIs combined with azithromycin. Moreover, cluster analysis results revealed Reduning injection combined with azithromycin was associated with a positive effect on the three group outcomes. Similarly, it was found to be the top three ranking in all outcomes based on SUCRA. What is new and conclusion Yanhuning injection combined with azithromycin and Reduning injection combined with azithromycin were found to be preferable treatments based on the data of this study.


| WHAT IS K NOWN AND OBJEC TIVE
Mycoplasma pneumoniae is an atypical bacterium that can cause severe respiratory tract infections. 1,2 In addition, it is a significant cause leading to the hospitalization of children with communityacquired pneumonia. 3,4 The incidence of mycoplasma pneumonia in children is high, which accounts for 16%-30% of communityacquired pneumonia in children. [5][6][7][8] It is generally mild, but it can be very serious, even involves extra pulmonary spread, such as central nervous system, mucosa and other organs, which seriously endangered the health of children. 4 Due to the serious side effects of tetracycline and fluoroquinolones on children, macrolides have become first choice for the treatment of mycoplasma pneumonia in children. 3,9 Azithromycin, a macrolide antibiotic, is a preferred decision because of its long half-life and clear target-cell effect. 10 However, macrolide resistance has emerged worldwide with a significant increase, which increasingly lead to complex clinical treatment schemes. 2,[11][12][13][14][15] In recent years, with the increase of people' recognition of traditional Chinese medicine, combining Chinese Medicine and Western Medicine to treat diseases has been an acceptable treatment method. 16,17 This provides another idea to treat mycoplasma pneumonia in children.
Chinese herbal injection (CHI) is a new formulation with rapid action and high bioavailability and made from components extracted from traditional Chinese herb using modern technology. 18 It is widely used in the treatment of cardiovascular diseases, cancers and respiratory diseases. [19][20][21][22][23][24] Although numerous traditional pairwise metaanalytic reviews have reported on the efficacy of many CHIs that could be used to treat mycoplasma pneumonia in children, [25][26][27][28] the effect of which kind of injection is the best remains unclear. This is because the traditional pairwise meta-analysis can only achieve direct comparison between the two interventions and cannot comprehensively evaluate the efficacy of various interventions. Network meta-analysis is an extension of traditional pairwise meta-analysis to multiple treatment comparisons and is a good choice when there is a lack of direct evidence, because it allows indirect comparisons between different types of treatment. 29,30 Therefore, this study used Bayesian network meta-analysis to systematically evaluate the efficacy of CHIs for the treatment of mycoplasma pneumonia in children and obtain the ranking of their effect in different outcomes, to provide evidence of evidence-based medicine for the clinicians to choose a more suitable therapy.

| ME THODS
The study was conducted according to the PRISMA extension statement (Appendix S1). 31 In addition, the design idea and concise workflow of this study are presented in Figure 1. Outcomes. The following outcomes attracted our attention: clinical effective rate, disappearance time of fever, disappearance time of cough, disappearance time of pulmonary rale, average hospitalization time, disappearance time of pulmonary shadows in X-ray, serum level of TNF-α and IL-6. Clinical effective rate = (number of total patients -number of invalid patients)/number of total patients × 100%. The patients whose clinical symptoms were unchanged and objective indicators were not adjusted were deemed as invalid. Study design. Only randomized controlled trials (RCTs) were enrolled.

| Eligibility criteria
F I G U R E 1 Design idea and concise workflow However, RCTs that lacked course of treatment and correct generation of random sequence were excluded. The detail of search terms is showed in Appendix S3.  33 Each item had three levels: low risk, unclear and high risk. Discordance of two researchers resolved by consensus or a third opinion.

| Data extraction and risk of bias assessment
Ethical approval was unnecessary, as private information of patients was not involved in this study.

| Literature retrieval and study characteristics
A total of 1181 articles were identified at first, and 167 RCTs were

| Risk of bias assessment
A total of 34 RCTs generated random sequence through random number table and 1 RCT through coin tossing. Therefore, their selection bias of random sequence generation was regarded as low risk. Attrition bias of incomplete outcome data was assessed as low risk due to all the research data was complete. The remaining biases were estimated as unclear because there were too few details could obtain to make a decision. In short, the quality of eligible RCTs was general ( Figure 4).

| Results of overall outcomes
The result data of all outcomes are presented in Table 1. As for clinical effective rate, all CHIs combined with azithromycin were superior to azithromycin only. When compared with azithromycin, all treatments, except Xixinnao injection combined with azithromycin, provided significant benefits in reducing disappearance time of fever, cough and pulmonary rale, and average hospitalization time.
Furthermore, Yanhuning injection combined with azithromycin was found to be more efficacious than Xiyanping injection combined with azithromycin in decreasing disappearance time of cough and average hospitalization time, and other CHIs combined with azithromycin in lessening disappearance time of pulmonary rale. In terms of disappearance time of pulmonary shadows in X-ray, Reduning injection combined with azithromycin, Xiyanping injection combined with azithromycin and Yanhuning injection combined with azithromycin yielded significant effects than azithromycin alone. As for concentration of TNF-α and IL-6, only Reduning injection combined with azithromycin could achieve a better efficacy than azithromycin. The significant difference was merely observed between above groups.

| Ranking results based on SUCRA
SUCRA results of eight outcomes are depicted in Table 2.
Azithromycin was the worst treatment in all outcomes. Tanreqing injection combined with azithromycin had a 74.6% probability to be the best intervention measure for clinical effective rate, successively fol-

| Cluster analysis
Three groups of cluster analysis were performed in this study, includ-

| Publication bias
Comparison-adjusted funnel plot for clinical effective rate was drawn to test the publication bias. Figure 6 shows that the points located in the funnel plot were asymmetrical based on the zero line, and the angle between the adjusted auxiliary line and the zero line was larger. Therefore, there may be a small publication bias.

| Safety
Among the all of 167 RCTs, 95 RCTs (56.89%) reported adverse drug reactions (ADRs) during the course of treatment. Nine out of them declared there was not ADRs, and 86 out of them described the detail of ADRs. The incidence of different types of ADRs in different interventions is counted in Table 3. The remaining 72 (43.11%) did not monitor ADRs during treatment. We could reach two main points from Table 3: first, patients with azithromycin alone had the highest incidence of ADRs; second, the incidence of gastrointestinal reactions was the highest among all competing interventions.

| WHAT IS NE W AND CON CLUS I ON
In summary, this study shows the comparative efficacy of CHIs combined with azithromycin for mycoplasma pneumonia in children.
Among the 6 interventions, Yanhuning injection combined with azithromycin and Reduning injection combined with azithromycin were found to be more effective treatments. However, because of the limitations of this study, the results should be verified by more high-quality large-sample RCTs.

ACK N OWLED G EM ENTS
This study has received funding by the National Natural Science Foundation of China (No. 81473547 and No. 81673829).

CO N FLI C T S O F I NTE R E S T
The authors have no conflicts of interest to declare.