Background
Systematic reviews use explicit, pre-specified methods to identify, appraise and synthesize all available evidence related to a (clinical) question of research interest. If appropriate, systematic reviews may include a quantitative data synthesis (i.e. meta-analysis), which is the statistical combination of results from ≥ 2 individual studies [
1]. However, systematic reviews conventionally compare only 2 interventions, despite having the existence of more than two interventions for a disease of interest. For instance, a randomised controlled trial (RCT) on antibiotics for treating leptospirosis included three arms [
2]. As such, a conventional pairwise meta-analysis may be conducted, but the comparative effectiveness of all available interventions for a given condition will not be addressed [
3]. Individual pair-wise comparisons, which in isolation fall short of informing clinical decisions when there are a greater number of treatment options available [
4]. A network meta-analysis (NMA), also known as mixed treatment comparison or multiple treatment comparison, is a method for simultaneous comparison of multiple treatments in a single meta-analysis [
3]. It expands the scope of a traditional (conventional) pairwise meta-analysis by analysing simultaneously both direct comparisons of interventions within RCTs and indirect comparisons across trials based on a common comparator [
5‐
7]. The multivariate approach, therefore, allows one to ‘borrow strength’ across correlated outcomes, to potentially reduce the impact of outcome reporting bias [
8].
Leptospirosis is a zoonosis caused by infection with pathogenic
Leptospira species that has a global distribution with a significant health impact, particularly in resource-poor tropical countries [
9]. The clinical course in humans ranges from mild to lethal with a broad spectrum of symptoms and clinical signs [
10]. A recent systematic review estimated that there are 1.03 (95% CI 0.43–1.75) million cases of leptospirosis worldwide each year and 58,900 deaths (95% CI 23,800–95,900) [
11,
12], which corresponds to an estimated 2.9 million disability-adjusted life years per annum, including 2.8 million years of life lost due to premature death [
9]. Thus far, the optimal treatment of leptospirosis remains a subject of debate, mainly due to the wide and biphasic clinical spectrum of the disease and the distinct pathogenesis in these two phases [
13,
14].
Taken together, the objective of this study was to characterise the conceptual understanding and the rational for the use of NMA in assessing drug efficacy. As such, we used results from RCTs of antibiotics for the treatment of leptospirosis as a case study.
Discussion
Meta-analyses comparing multiple treatments are feasible and should be considered as the bedrock for decisions when several treatments are available [
2,
4]. NMA in its standard form makes an assumption of ‘consistency’ [
19] that estimates of treatment effects from direct and indirect evidence are in agreement [
17,
19]. The current NMA could hold the key assumption of consistency.
The results of this NMA showed that it is possible to assess the efficacy of cephalosporin compared to a placebo even though this direct comparison was not performed in any of the included trials. Our results predicted that a cephalosporin antibiotic would have comparable efficacy to penicillin in reducing mortality in human leptospirosis. Indications for the use of cephalosporin antibiotics for the treatment of leptospirosis are included in the WHO guideline for management of leptospirosis [
28] as well as some national guidelines for management of leptospirosis in some countries such as Malaysia, as an example [
29].
A Cochrane review on seven RCTs [
30] as well as a non-Cochrane review on ten RCTs [
31] performed pairwise analyses of the efficacy of antibiotics for the treatment of human leptospirosis. Both reviews reported comparable efficacy of antibiotics in preventing mortality as an outcome as well as an effect on the duration of illness. The Cochrane systematic review concluded that there were insufficient evidence to advocate for or against the use of antibiotics for the treatment of treating leptospirosis [
30] and the review by Charan and associates [
31] showed that there was no significant difference between mortality in groups given penicillin compared to control groups.
The WHO treatment guidelines still recommend administration of antibiotics for leptospirosis regardless of the stage or severity of the disease [
28]. The optimal treatment of leptospirosis remains a major clinical dilemma, for which limited data from clinical studies exist [
14]. Penicillin G sodium (penicillin G) is generally recommended as the first choice treatment for severe leptospirosis. It is important to evaluate alternatives to penicillin G because its use has potential drawbacks. Antibiotic resistance has compromised the efficacy of penicillin G against many important bacterial pathogens, and it is intrinsically inactive against coinfected
Rickettsiosis that are common in tropical areas such as Thailand [
26]. In addition, Jarisch-Herxheimer Reaction (JHR) is a known complication associated with the use of penicillin G for the treatment of leptospirosis [
32,
33]. Therefore, penicillin G administration might pose a great burden in critically ill patients [
33]. Of note is that the small number of included studies and these being not recent is a reflection of the limited scientific interest in performing clinical trials in this field. There may be a number of reasons for this. For instance, the lack of a widely available, sensitive and rapid method of laboratory confirmation of leptospirosis has been an important impediment [
2] and this compromises the recruitment of patients for the clinical trials. Moreover, there may be a concern whether the clinical manifestation of leptospirosis would become worse after the initiation of antibiotic therapy due to the development of JHR. A systematic review of 27 studies in JHR had reported the development of JHR in 92 of 976 leptospirosis patients within 1 to 48 h after administration of the first dose of antibiotic [
32]. It is also noted that a higher proportion of JHR occurred in early stage leptospirosis, suggesting a higher probability of the (adverse) event before the natural clearance of spirochetes [
32].
Other classes of antibiotic may provide better alternatives to penicillin G. Doxycycline has the advantage that it can be administered orally but it is not suitable in pregnant women. Like penicillin, most cephalosporin act on the bacterial cell wall synthesis, with some exceptions that act on protein synthesis [
34]. Ceftriaxone can be administered once daily, which is an advantage over another third generation cephalosporin such as cefotaxime [
2] and no dosage adjustment, was required for renal failure. In addition, there is no reported evidence of JHR in patients with leptospirosis. Moreover, ceftriaxone can give extra benefit of being an excellent empirical therapy for other infections (e.g.
Streptococcus pneumonia) which mimic the clinical presentation of leptospirosis [
14].
This is the first time that an indirect evaluation of the efficacy of an antibiotic treatment for leptospirosis using NMA has been performed. This is an important additional work because the evaluation of antibiotic treatments for leptospirosis using double-blind RCTs is complicated by ethical considerations associated with the provision of a placebo to severely affected patients [
26]. Therefore, study designs that permit the use of indirect analyses of efficacy such as the NMA would allow an assessment of cephalosporin (ceftriaxone in this case) compared to a placebo control.
The indirect comparisons in the current review revealed that the antibiotics did not differ from each other with regards to their ability to reduce mortality, supporting the findings of earlier reviews [
30,
31]. However, the NMA provided slightly different results compared to the more simplistic direct comparison using conventional pairwise meta-analysis efficacy estimates. This shows the potential advantage of NMA because it can incorporate both direct and indirect comparisons, decreasing the risk for possible sponsorship bias [
35], which often is an issue for drug trails.
There are some limitations that needed to acknowledge. We did not find evidence of inconsistency in the results from our indirect comparison analysis. However, these findings should be interpreted with caution as only a small number of trials could be identified for inclusion in the current analysis. Nevertheless, our findings agree with the earlier reviews, indicating no significant difference between the antibiotics for mortality as an end point. The current network meta-analysis could hold the key assumption of consistency. The indirect comparisons presented in this study add to the current body of evidence in literature.
Acknowledgements
We are grateful to the participants and researchers of the primary studies. We thankfully acknowledge the reviewers and the editors for the comments given and the helpful inputs to improve the manuscript. We thank the International Medical University in Malaysia and the University of Queensland/School of Public Health in Australia for allowing us to perform this study.