Efficacy of glucocorticoids for the treatment of macrolide refractory mycoplasma pneumonia in children: meta-analysis of randomized controlled trials

Background Mycoplasma pneumoniae is one of the most common pathogens causing community acquired pneumonia in children. Although the rate of macrolide-refractory Mycoplasma pneumoniae (MRMP) has increased, systemic glucocorticoids as a treatment option has not been validated yet. The purpose of this study was to assess the efficacy of glucocorticoids add-on in the treatment of MRMP in children through systematic review and meta-analysis. Methods Data sources A systematic literature search was conducted using ten electronic bibliographic databases including English, Korean, Chinese and Japanese languages, up to March 8, 2018. Study selection The study was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist and selected randomized control trials which compared the efficacy of glucocorticoids add-on to macrolide in the treatment of MRMP in children. Data extraction Two independent reviewers extracted: primary outcomes as hospital days, fever duration, and change in C-reactive protein (CRP) and main analysis was performed through meta-analysis with random effects model. Results Twenty-four unique randomized controlled trials met the inclusion criteria. The mean length of hospital stay in glucocorticoids treatment group was significantly shorter than that in conventional macrolide-treatment group (Weighted mean difference (WMD) = − 4.03 days). The mean length of fever duration was significantly shorter in the glucocorticoid treatment group in comparison with the conventional treatment group (WMD = -3.32 days). Level of CRP after treatment was significantly lower in the glucocorticoid treatment group than that in the conventional treatment group (WMD = -16.03). Sensitivity analysis and subgroup analysis showed no significant improvement in heterogeneity. As limitations of the study, most of the studies included were from a single country and we were unable to control for heterogeneity across interventions, lack of standardized measures, and different time points of assessments across studies. Conclusions Glucocorticoid add-on treatment for MRMP can significantly shorten the duration of fever and hospital stay and decrease the level of CRP. These results should be confirmed by adequately powered studies in the future.


Background
Mycoplasma pneumoniae (M. pneumoniae) is one of the major pathogens causing community acquired pneumonia and bronchitis in children. Treatment of M. pneumoniaerelated respiratory infection is based on symptomatic treatment with antibiotics. Macrolides have been used as first line treatment. However, macrolide-refractory M. pneumoniae (MRMP) strains are increasing abruptly, particularly in East Asian countries including Korea, Japan, and China [1][2][3].
Most patients with MRMP do not show improvement of fever when they are treated with macrolides. Some may develop refractory or severe clinical course that requires additional treatment. Treatment for MRMP includes tetracyclines, fluoroquinolones, and systemic glucocorticoids [4].
Secondary antibiotics such as tetracyclines and fluoroquinolones are considered as effective alternatives in the treatment of MRMP. However, they are of limited use due to safety-concerns of teeth discoloration and musculoskeletal toxicity, particularly in children. Glucocorticoids can be also considered as alternative treatment options due to two reason. First, the pathogenesis of M. pneumoniae infection is associated with amplified host immune response and virulence of M. pneumoniae [5]. Second, adverse effects of glucocorticoids have been well established. Further risk is not likely to be added in the treatment of MRMP. However, previous studies on the effect of glucocorticoids have shown conflicting results [6][7][8].
The objective of this study was to assess the efficacy of glucocorticoids for treatment of MRMP in children through systematic review and meta-analysis.

Methods
Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 were used for this systematic review and metaanalysis [9]. The Population-Intervention-Comparison-Outcome (PICO) question used for our search strategy was: "Does use of glucocorticoids help improve the outcome of MRMP in children?"

Search strategy
We performed a systematic search utilizing a protocol designed by two independent medical librarians (D.W.S. and M.L) specifically for this study with 10 electronic databases: PubMed, EMBASE, Cochrane Library, and Core journal (Korean, Japanese, and Chinese Journal) Full-text Database. The search encompassed articles published from January, 1990 to March 8, 2018. We used search terms listed in Additional file 1 to search PubMed, Cochrane, EMBASE, and database of core countries. We imposed no language or publication restrictions.
The first screening was executed by two independent reviewers (H.S.K. and I.S.S.) who evaluated the titles and abstracts obtained from the search. Records were managed using Endnote (version X8; Clarivate Analytics, Philadelphia, PA, USA). From this initial screening, articles that did not focus on glucocorticoid use in MRMP and review articles were immediately excluded. After initial exclusion process, full texts of the remaining articles were reviewed independently by two authors (H.S.K. and I.S.S.) to determine whether any articles met the predetermined eligibility criteria described in the next section. Disagreements between the two reviewers regarding the inclusion or exclusion of particular studies were settled by consultation with a third reviewer (Y.J.L.).

Eligibility
The following inclusion criteria were applied: (1) randomized control trial (RCT) which compare the efficacy of glucocorticoids add-on to macrolide alone in children with MRMP, (2) MRMP which was diagnosed with serology or polymerase chain reaction, and that refractories were defined clinically, (3) only included children < 18 years of age, and (4) outcome measures with hospital days, fever duration, and level of C-reactive protein (CRP) change. Review articles, published abstracts without full-text publications, and case-study reports with 10 participants or less were excluded. Our search strategy included non-English articles in our initial search results. Non-English articles were then translated and included for evaluation.

Study selection
Two reviewers (H.S.K. and I.S.S.) independently screened titles and abstracts of the studies identified in our systematic search. Studies focusing on MRMP were included after review of abstracts. Full texts from included studies were reviewed to evaluate for eligibility. Reference lists of selected studies and previous reviews were also examined to determine any relevant publications overlooked by the electronic search. Disagreements between the two reviewers in the selection of particular studies were settled after discussion with a third reviewer (Y.J.L.).

Risk of bias assessment
Cochrane Collaboration Risk of Bias Tool was used by the two reviewers (H.S.K. and I.S.S.) who independently evaluated the risk of bias in included studies [10]. Risk of bias was determined as hi, low or unclear by evaluating random sequence generation, blinding of participants and personnel, incomplete outcome data, and selective reporting. Disagreements between the two reviewers regarding the risk of bias assessment of particular studies were settled after discussion with a third reviewer (Y.J.L.).

Data extraction
Two reviewers (H.S.K. and I.S.S.) used a structured form to extract data from each eligible study. Data extracted from each study could be characterized as characteristics of the sample, intervention details, and measurement of outcomes. Disagreements between the two reviewers regarding the data extraction of particular studies were settled by consultation with a third reviewer (Y.J.L.). Primary outcomes of the current study were hospital days, fever duration, and change in CRP.

Statistical analysis
The estimated mean effect of glucocorticoid add-on treatment on hospital days, fever duration, and change in C-reactive protein (CRP) and the associated 95% confidence intervals (CIs) were extracted or calculated for the 24 studies included in the meta-analysis with Review Manager 5.3 (The Cochrane Collaboration, London, United Kingdom). Random-effects model was used for studies included in the analysis.
Heterogeneity was calculated using I 2 statistic. The I 2 statistic threshold should always be interpreted with care. A rough estimate of 25% denotes low heterogeneity while 50% denotes moderate heterogeneity, and 75% denotes high heterogeneity [11]. We conducted sensitivity analyses when heterogeneity was noted. This was performed by removing a study from the analysis to determine changes in I 2 values and assess which studies play a significant role resulting in heterogeneity [11]. To assess the risk of publication bias, we used funnel plots for visual inspection, and Egger test and trim-and-fill method were performed for statistical identifying. All statistical

Systematic literature search results
A total of 1829 citations were identified initially. Of these, 1773 studies were discarded after reviewing titles and abstracts, leaving 56 articles for full text review (Fig. 1). A total of 32 studies were excluded after full text review due to no proper subject, inadequate study protocol, review article, or no relevant outcome. A total of 24 studies were included in our systematic review and meta-analysis . All studies were RCTs. (Additional file 1 for search strategies for database, Additional file 2 for PRISMA checklist).

Sample characteristics
Participants of studies enrolled in our meta-analysis was a total of 2365 patients. All these studies examined fever duration. Fifteen studies examined hospital days while 14 studies examined CRP level after treatment (Table 1).

Fever duration
The mean length of total duration of fever was significantly shorter in the glucocorticoid add-on group than that in the conventional treatment group (weighted mean difference, WMD = − 3.32, 95%CI: − 4.16 --2.48, Z = 7.72, P < 0.00001). However, there was a high between-study heterogeneity of this effect (I 2 = 98%; Fig. 2).

Hospital days
The mean length of hospital stays in the glucocorticoid add-on treatment group was significantly shorter than that in the conventional treatment group (WMD = − 4.03, 95% CI: − 4.89 --3.18, Z = 9.26, P < 0.00001). However, there was a high between-study heterogeneity of this effect (I 2 = 90%; Fig. 3).

Reduction of CRP level after treatment
The level of CRP after treatment was significantly lower in the glucocorticoid add-on treatment group than that in the conventional treatment group (WMD = − 16.03, 95%CI: − 22.56 --9.50, Z = 4.81, P < 0.00001). However, there was a high between-study heterogeneity of this effect (I 2 = 100%; Fig. 4). Sensitivity analyses was performed because of high level of heterogeneity. We removed a study from the analysis to determine which studies contributed most significantly to the heterogeneity by determining the changes in I 2 values. We found that I 2 values of fever duration, hospital days, and CRP level did not change.

Subgroup analysis
Use of glucocorticoids included the use of any type of glucocorticoids (e.g., methylprednisolone, dexamethasone, and Fig. 2 Comparison of fever duration between glucocorticoid therapy and macrolide therapy prednisolone). The use of different types of glucocorticoids was different across studies. This might have contributed to the heterogeneity in the overall use of glucocorticoids. Thus, we stratified the meta-analysis by subgroup analyses. In subgroup meta-analysis for use of methylprednisolone compared with the use of other glucocorticoids for the length of hospital stay, the use of other steroids did not show any significant subgroup differences (Figs. 5, 6 and 7).

Publication bias
All funnel plots were symmetric, indicating an absence of significant publication bias within these studies except    for CRP outcome. Egger test results were − 1.73 (P = 0.09) for fever duration, 0.59 (P = 0.56) for hospital days, and − 3.19 (P = 0.008) for CRP. Trim-and-fill method for adjusting publication bias on CRP outcome was performed. The mean difference changed from -3.27 (P = 0.35) to − 16.03 (P < 0.001). These results indicated that there was substantial evidence of publication bias in CRP outcome (Fig. 8).

Discussion
This systematic review and meta-analysis identified and assessed RCTs on the use of glucocorticoids in children with MRMP. We specifically investigated effects of glucocorticoids on fever duration, length of hospital stay, and CRP level after treatment in comparison with conventional macrolide therapy. Results revealed positive effects of glucocorticoid treatment on all outcome measures. M. pneumoniae is a common pathogen causing community acquired pneumonia. The clinical course of M. pneumoniae infection is diverse, ranging from selflimiting to severe pneumonia with extra-pulmonary complications [36]. Macrolide is considered the first-line treatment for M. pneumoniae infection [37]. In adults, one study reported that prednisone treatment for 7 days in patients with community-acquired pneumonia admitted to hospital can shorten time to clinical stability without increase in complications [38]. However, another study reported that glucocorticoid use did not show any benefits in children [8]. Pulmonary injury associated with severe mycoplasma pneumonia could be caused by host immune response rather than by direct microbial damage [39,40]. Overly active cell-mediated immunity and cytokine responses play a significant role in MRMP [41]. Severe M. pneumoniae infections and MRMP show similar laboratory findings with severe acute respiratory syndrome such as increased levels of non-specific markers of inflammation such as serum CRP, lactate dehydrogenase, and D-dimer [42]. Since MRMP can be considered as an immune-mediated disease, use of immune modulatory therapy could seem rational. For cases with MRMP infection, secondary treatment options are limited due to adverse effects of tetracycline or fluoroquinolones, especially in children [43]. The use of systemic steroids, in addition to antimicrobial therapy, to diminish host immune response in MRMP has been used in children and demonstrated satisfactory effects [7,8,14]. Glucocorticoids have shown to improve clinical features and reduce lung injury in children and adults [5,44,45]. The use of glucocorticoids in M. pneumoniae infection showed reduced lung histopathologic score by reducing cytokines and decreasing inflammatory response ultimately leading to lower mortality [5,44,45]. Therefore, glucocorticoids have been considered suitable as adjunctive therapy to macrolide therapy. Although previous studies have reported the efficacy of glucocorticoids in the treatment of severe MRMP, treatment protocols varied from oral prednisolone of 1 mg/kg/day to intravenous methylprednisolone 30 mg/kg/day for 3 days [7,17]. These variable treatment protocols are due to the lack of guidelines for glucocorticoids treatment in severe MRMP to date [7,46].
In our study, the mean length of total duration of fever, mean length of hospital stay, and level of CRP after treatment in the glucocorticoid treatment group were significantly shorter or lower than that those in the conventional treatment group. However, heterogeneity in all outcome measures was high and sensitivity analysis resulted in no significant difference. Subgroup analysis showed no significant decrease in heterogeneity. In spite of high level of heterogeneity, all studies reported favorable outcomes of hospital stays and fever duration after the use of glucocorticoids in MRMP. For CRP level, only one study reported unfavorable result after the use of glucocorticoids in MRMP. Therefore, we could consider that the use of glucocorticoids has significantly favorable effects on outcomes of MRMP.
The strength of this review is that it summarized the current evidence on the efficacy of glucocorticoids from RCT. Conducting RCTs offers the potential to produce the most definitive evidence to confirm or refute whether glucocorticoids can help improve outcomes of MRMP.
This review has some limitations. Firstly, almost all studies included in this review had at least one methodological flaw. Secondly, publication bias was significant for change in CRP level. It might be due to small sample sizes in most studies and the fact that most studies were conducted in a single country. As a result, the robustness of these outcomes should be explored with future studies. Thirdly, most of the studies included were from a single country. Lastly, we could not evaluate other outcomes such as improvement in chest x-ray infiltration or other symptoms such as cough because these data were not available in each study.
Despite these limitations, with our review and metaanalysis, this is the first study to date to synthesize the efficacy of glucocorticoids in MRMP.

Conclusions
In conclusion, even though some studies have reported the efficacy and effectiveness of systemic glucocorticoids in the treatment of MRMP , this is the first systematic review and meta-analysis to investigate the effectiveness of glucocorticoids in MRMP. We found that the use of glucocorticoids could shorten hospital days, shorten fever duration, and lower CRP levels after treatment.
However, these results should be interpreted cautiously, and future studies should also assess other outcomes to clarify the effect of glucocorticoids in MRMP.