Antileukotrienes for the prevention and treatment of chronic lung disease in very preterm newborns: a systematic review

Background Very preterm infants are at high risk of developing chronic lung disease, which requires respiratory support and might have long-term sequelae. As lung inflammation plays an important role in pathogenesis, antileukotrienes have been explored in both clinical and animal studies. We aimed to assess the benefits and harms of antileukotrienes for the prevention and treatment of respiratory morbidity and mortality in very preterm newborns. Methods In this systematic review, we included randomized trials and non-randomized studies in humans and animals reporting the effects of antileukotrienes in very preterm infants or other mammals within 10 days of birth. Our pre-specified primary outcomes were all-cause mortality and any harm, and, for the clinical studies, incidence of chronic lung disease. Included studies underwent risk of bias-assessment and data extraction performed by two authors independently. There were no language restrictions. Results Fifteen studies totally met our inclusion criteria: one randomized trial and four non-randomized studies in humans and 10 animal studies (five in rodents, two in lambs and one in either guinea pigs, rabbits or caprinae). All five clinical studies used montelukast and had a small sample size, ranging from 4 to 77 infants. The randomized trial (n = 60) found no difference in the incidence of chronic lung disease between the groups. Only one clinical study, which enrolled four very preterm infants and had a critical overall risk of bias, reported long-term outcomes. All other studies had unclear or greater overall risk of bias and meta-analyses were therefore deemed unfeasible. Eight of ten animal studies used leukotriene receptor antagonists as antileukotriene (montelukast in three of ten studies) and seven had an experimental study design (i.e. some animals were not exposed to antileukotrienes but no randomization). Three of the ten animal studies assessed different doses. Animal studies found no effect on the outcomes mortality, growth, or lung function related surrogate outcomes. Conclusions Use of antileukotrienes in very preterm infants to prevent or treat chronic lung disease is not supported by the available evidence. Large randomized trials focusing on outcomes relevant to patients, including long-term outcomes, are needed. Studies should also minimize risk of bias.


Background
Very preterm infants (born before 32 weeks' gestational age) constitute an extremely vulnerable population and are at high risk of developing chronic lung disease [1]. Chronic lung disease is a broad term, which includes bronchial asthma and bronchopulmonary dysplasia (BPD). It has been reported that BPD is the most Open Access *Correspondence: matteo.bruschettini@med.lu.se 5 Cochrane Sweden, Dept. Research and Development, Skåne University Hospital, Clinical Science Lund, Lund University, Lund, Sweden Full list of author information is available at the end of the article common complication in extremely preterm infants [2]. Defining BPD remains a challenge [3]. This is mainly due to there being multiple factors involved in the underlying pathophysiology. Injury to the lungs, both before and after birth, may lead to an abnormal reparative response. This could cause flawed lung development, which can affect lung function into adult life [2]. Caffeine is the only drug that reduces the rate of BPD [4], mortality, and neurodevelopmental disability [5]. More interventions are therefore needed to prevent and treat BPD and its consequences.
Montelukast is the most common type in clinical use, is administered once a day, and can be taken without regard to meals [7]. Zafirlukast and pranlukast are administered twice a day. The LTRA s are processed mainly in the liver [8], metabolized mostly by CYP2C8, with the involvement of CYP2C9 CYP3A4 enzymes [9,10]. It is worth mentioning that LTRAs are substrates for transporters [11] and the s of genes In children, common montelukast induced adverse events are headaches, abdominal pain, rash, thirst, hyperkinesia, asthma and eczema [13]. Pharmacovigilance studies have also reported increased frequency sleeping disorders in infants younger than 2 years and psychiatric disorders in children aged 2 to 11 years, being more frequently reported than in adults. This led to a US FDA alert being issued for psychiatric events being associated with montelukast. Eosinophilic granulomatosis may also be associated with the use of montelukast, but the role of LTRAs in its pathogenesis is still uncertain [15].
The drug zileuton, also an antileukotriene, has a different action mechanism from LTRAs. It works as an inhibitor of 5-lipoxygenase. The most serious concern is hepatotoxicity. Zileuton is mainly metabolized through the liver, particularly via P450 enzymes, mostly by CYP3A4 [16]. This can lead to problems when using drugs such as theophylline at the same time due to impaired metabolization of theophylline. An option is to halve the dose of theophylline when starting treatment with zileuton [8]. Theophylline is an example of a methylxanthine, which are known to have a protective effect on the respiratory system [17]. Methylxanthines are natural components of cocoa-based products and beverages such as coffee, tea and yerba mate and therefore are commonly present in the human milk, thus reaching the newborn.
The properties of antileukotrienes might have the potential to be useful in the prevention and treatment of chronic lung disease in very preterm infants and they are currently used clinically based on anecdotal evidence, though not approved for this purpose. Their harms and benefits have not been assessed systematically. This systematic review aims to explore the evidence base for antileukotrienes in very preterm neonates in both clinical and animal studies.

Methods
Our methods for systematically reviewing the clinical studies are based on the template developed by the Cochrane Neonatal Review group (Resources for Review Authors, n.d.) [18]. Two separate protocols were registered in Prospero for the clinical and animal studies, respectively [19,20], since Prospero requires authors to register separate protocols for clinical and animal studies. An exploratory pilot search for animal studies was performed before submitting the protocols and our comprehensive search and data extraction.

Types of studies
We included randomized and non-randomized animal studies. Studies with a cross-over design were excluded due to our interest in long-term outcomes and the potential for carry-over effects.
For the clinical studies, we included randomized trials, quasi-randomized trials and non-randomized studies of intervention (NRSI). Again, we excluded trials with a cross-over design.

Types of participants
We included studies in any neonatal mammals, both term and pre term. "Neonatal" was defined as the first 10 days since birth, which is an arbitrary cut-off point that we pre-specified in our protocol. For the clinical studies, we included very preterm infants with a gestational age below 32 weeks and who were admitted to a neonatal department.

Types of interventions
For animal studies, we included studies using co-interventions and any route and dose of antileukotriene administration. We excluded studies where antileukotrienes were administered to mothers before birth or to lactating mothers. We also excluded studies where cointerventions were not available for all study arms. We included two types of studies (1) antileukotrienes versus control (either placebo, no intervention, or treatment as usual); (2) studies without any comparator (non-controlled studies).
For the clinical studies, we included two comparisons, i.e. (1) prevention and (2) treatment of chronic lung disease.

Animal studies
Our primary outcomes for the animal studies were: (1) survival until last follow up; (2) any harm.

Clinical studies
Our primary outcomes for the clinical studies were: (1) all-cause mortality during initial hospitalization; (2) BPD/chronic lung disease incidence: only for comparison one (i.e. prevention of chronic lung disease) according to the three definitions: [21][22][23]; (3) any harm. Secondary outcomes were: (1) all-cause neonatal (first 28 days) mortality, only for comparison one (i.e. prevention of chronic lung disease); (2) retinopathy of prematurity (any and ≥ stage 3 [24]) (3) days of respiratory support; (4) days of supplemental oxygen; (5) need for mechanical ventilation (yes/no); (6) days of hospital stay; (7) major neurodevelopmental disability: cerebral palsy, developmental delay [25,26] or Griffiths Mental Development Scale [27] assessment greater than two standard deviations (SDs) below the mean), intellectual impairment (intelligence quotient (IQ) greater than two SDs below the mean), blindness (vision less than 6/60 in both eyes), or sensorineural deafness requiring amplification. We pre-planned to assess data for children aged 18 to 24 months and aged three to five years separately; (8) each component of the composite outcome "major neurodevelopmental disability"; (9) pulmonary function test at school age (as specified by study authors).

Searches
We searched the Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library; MEDLINE via PubMed, and Embase, in September 2020. We also searched ongoing clinical trials submitted at clinicaltrials.gov and ITCRP website. We did not apply any restrictions regarding language, publication year, or publication status. Methodological filters excluding diagnostic studies were not used. Search strings for each database are listed in Appendix.

Selection of studies
Two authors independently screened titles and abstracts and retrieved the full text of potentially relevant articles. Eligibility was assessed according to our inclusion criteria. Two authors independently performed data extraction and assessed risk of bias.

Assessment of risk of bias
We used SYRCLE's risk of bias tool [28] for animal studies, which include the following seven domains: selection bias due to sequence generation, baseline characteristics or inadequate allocation concealment; performance bias due to inadequate randomization housing or blinding; detection bias due to inadequate randomization of outcome assessment or blinding; attrition bias due to incomplete outcome data; reporting bias due to selective outcome reporting; and other sources of bias.
For non-randomized clinical studies, we used the ROBINS-I [29] tool to assess the risk of bias, which include the following eight domains: bias due to confounding; bias in selection of participants into the study; bias in classification of interventions; bias due to deviations from intended interventions; bias due to missing data; bias in measurement of outcomes; bias in the reported results; and the overall risk of bias. For the domain "confounding", we took into account the following confounders: antenatal steroids, gestational age, birth weight, Apgar score, indication to start antileukotrienes and level of respiratory support at study entry.
For randomized trials, we used the Cochrane Risk of Bias 2 tool [30], which include the following five domains: bias arising from the randomization process; bias due to deviations of intended interventions; bias due to missing outcome data, bias in measurements of the outcome; bias in selection of the reported results and overall risk of bias.
Any disagreements were solved through discussion and, if necessary, by consulting a third review author.

Data analysis
We planned to use the Cochrane software RevMan 5.4 [31] to synthesize and analyze data. We planned to analyze all infants and animals on an intention-to-treat basis and to use the fixed-effect model for meta-analyses because we expected a consistent treatment effect. We planned to synthesize data with risk ratios (RR) for dichotomous outcomes and mean differences (MD) for continuous outcomes, with 95% confidence intervals (CI). The overall certainty of the evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, as outlined in the GRADE Handbook [32], for our primary outcomes.

Subgroup analyses
For the animal studies, we planned the following: type of lung injury, dose and type of antileukotrienes.

Results of the search
Our searches for animal and clinical studies (Appendix) returned 1929 unique records. One additional study was identified through other sources (online search) while completing the review. Following screening titles and abstract, 22 studies were collected and assessed in fulltext. Three animal studies were excluded because the animals were older than 10 days. Three studies were labelled as awaiting classification because the text of the conference abstracts were not available [33,34] or because a protocol registered in 2007 was apparently not followed by a publication. One ongoing uncontrolled clinical study was identified, with a planned sample size of 200 very low birth weight newborns [35]. Thus, fifteen studies were included: ten animal studies (see Table 1) and five clinical studies of which one was a randomized trial (see Table 2). Figure 1 presents the PRISMA flow chart.

Included studies Animal studies
Of the ten included animal studies, five were in rodents (three in rats and two in mice) [36][37][38][39][40] and two were in lambs [41,42]. The remaining three were in either guinea pigs [43], rabbits [44] or caprinae (sheep and goats in the same study) [45]. Two studies assessed prevention of respiratory morbidity only [36,44], while three studied treatment effects only [37,39,40]. Five studied both prevention and treatment effects [38,[41][42][43]45]. Eight of ten studies used leukotriene receptor antagonists such as antileukotrienes (montelukast used in three studies [36,37,39]), one studied a leukotriene synthesis inhibitor [40] and one studied both types of antileukotrienes [45]. Of the 10 animal studies, seven had an experimental study design (i.e. some animals were not exposed to antileukotrienes but were not randomized) and three an observational study design (i.e. all animals were exposed to antileukotrienes). Within three of the ten studies different doses were assessed [40,43,44].

Clinical studies
The five clinical studies included one randomized trial from Korea [46], a non-randomized study performed in Germany and the USA [47], and tree observational studies from Korea [48], the UK [49] and Taiwan [50]. Four studied treatment and one studied prevention of BPD [46]. All five assessed the same leukotriene antagonist; montelukast. The administered dose of montelukast ranged from 1 to 2 mg/kg body weight. Details are provided in Table 2.

Risk of bias Animal studies
Details of our risk of bias assessments are presented in Table 3. Overall, risk of bias was difficult to assess due to poor reporting and most domains were therefore "unclear" using the SYRCLE risk of bias tool [28]. As this tool is developed specifically for experimental animal studies, some domains were not applicable to the three non-controlled studies [41,42,45].
The seven experimental animal studies all had unclear risk of selection bias because the randomization process and baseline characteristics were not specified. They all had an unclear risk of performance bias because none of the studies reported on random housing. The measures   According to body weight (less than 1,000 g, 0.5 mg; 1,000 g to 1,500 g, 1.0 mg; 1,500 g to 2,000 g, 1.5 mg; greater than 2,000 g, 2 mg) used to house the animals randomly within the animal room were not reported. Blinding of the investigators was also not reported. Only one study [37] was assessed as at low risk of detection bias as they reported that "all morphometric assessments were made blindly by the same observer (except for the bronchial alveolar attachments)". All studies had unclear or higher risk of reporting bias as their protocol was not available. One study reported that data for some outcomes were not shown [42] and thus had high risk of bias. The animal studies appeared free from other sources of biases.

Clinical studies
Details of our risk of bias assessments for the randomized and non-randomized studies are presented in Tables 4  and 5, respectively. None of the included clinical studies were assessed to have low risk of bias. The single included RCT had an overall risk of bias assessed as "some concerns" [46] due to missing information about the randomization process; unclear description of the infants that were not included in the final analysis and because it was unclear whether the outcome assessors were blinded. The cohort study did not clearly define inclusion and exclusion criteria and was therefore assessed to have serious risk of bias [48]. The study by Rupprecht et al. [47] was scored with an overall critical risk of bias because of confounding as the control group consisted of children whose parents provided informed consent for participation as a control group patient but not for administration of the medication montelukast. The reasons for only allowing the child into the control group

Overall risk of bias
Kim 2015 some concerns a some concerns b low c some concerns d low e some concerns Table 5 Risk of Bias assessment with ROBINS-I tool for the included non-randomized studies a The control group consisted of children whose parents provided informed consent for participation in this study (as a control group patient) but not for administration of the medication montelukast (controls 1-5, 8,and 9; Table 1); and children in whom the planned therapy scheme was not possible due to existing or arising contraindications for the study drug (4 children, phenobarbital therapy in controls 6, 7, 10, and 11) b No significant difference between groups regarding patients' characteristics c There is no clear definition of inclusion and exclusion criteria d The study does not specify the exact time for which montelukast was given and for how long co-interventions of the conservative treatment were given, which may lead to relevant differences in co-interventions e Data appears to be complete, although no protocol was published and the study was not registered as a clinical study f Outcome "Duration for mechanical ventilation" might be biased by unblinded outcome assessor g Outcome "Need for mechanical ventilation" might be biased by unblinded outcome assessor h There is a discrepancy between text of the results section and  There was no mortality among study animals In the prevention study (normoxia), 3 out of 19 and 1 out of 6 pups died in the antileukotriene and control group, respectively. In the treatment study (hyperoxia), 3 out of 22 and 0 out of 6 pups died in the antileukotriene and control group, respectively This outcome cannot be calculated because "any loss of sample size due to deaths was made up for by random sampling  overall critical risk of bias as a historical cohort design was used without clear inclusion and exclusion criteria.

Effects of the interventions
Meta-analysis of the clinical and the animal studies was not deemed feasible for any of the outcomes since they were reported by too few studies with highly variable designs and were assessed with outcome measures which could not be meaningfully pooled. Table 6 shows the list of the outcomes reported by each study. Four controlled studies reported on mortality and found no significant effect [36,37,40,44]; two controlled studies reported on growth [36,37]; no significant effect was found in either study between combination treatment (montelukast, clarithromycin and pentoxifylline combination) versus placebo.

Animal studies
We made the post hoc decision to include the reported outcome 'lung weight to body weight ratio' , in addition to our prespecified outcome lung volume to body weight. No statistically significant difference was found in the three studies reporting on either of the two outcomes [37,39,44].
Three studies reported on inflammation markers for lungs [39,43,44]. Two studies [43,44] reported on polymorphonuclear leukocytes and protein in bronchoalveolar fluid. Phillips et al. [43] showed a reduction in the number of neutrophils and protein in the treated hypoxia group and in eosinophils in the treated normoxia group. The study by Chen et al. [39] detected a reduction in the concentration of other inflammation markers in the lung tissue of BPD mice.
Lung injury was reported in one study in which montelukast treatment decreased malondialdehyde levels and enhanced superoxide dismutase activity in the lung tissues of the BPD mice [39].
The study by Demir et al. [36] was the only study to report fibrosis and smooth muscle actin expression. They did not detect an effect of montelukast alone versus placebo; the combination treatment (montelukast, clarithromycin and pentoxifylline combination) did result in a lower actin score compared to the placebo group.
Only one study reported on behavioral tests, the Morris water maze experiment [38]. There was an improvement in escape latency in the pranlukast group and the number of times rats in the pranlukast group crossed the platform in the maze increased.
The study by Schreiber et al. [41] found a decrease in pulmonary vascular resistance in lambs after antileukotriene infusion.
Two clinical studies reported on all-cause mortality [47,48]; only one study [47], non-randomized, reported on all-cause mortality for both the intervention and the control group and found a significant reduction in allcause mortality in the montelukast group.
The two studies that reported on frequency and the severity classification of BPD showed no relevant difference between case and control group [46,48].
The occurrence of adverse events did not differ between intervention and control groups in either the randomized trial [46] or the observational study by Kim [48]. It was unclear whether the other three studies had planned to report adverse events, but they did not.
Rupprecht et al. [47] did not provide information about the timing of drug administration and therefore all-cause neonatal mortality could not be extracted from the study for our pre-defined time point. Kim et al. [46], Panjwani et al. [49] and Min Kim et al. [48] did not report all-cause neonatal mortality.
Rupprecht et al. [47] reported a significantly shorter duration of respiratory support in the group receiving montelukast compared to controls (41.2 ± 25.3 vs. 103.7 ± 90.6 days).
Two studies reported on mechanical ventilation and found no differences [46,48].
None of the included studies reported on fraction of inspired oxygen duration or pulmonary function testing at school age.

GRADE assessment
The certainty of the evidence was "very low" for all outcomes because of imprecision and high risk of bias in multiple other domains, both in clinical and animal studies.

Summary of main findings
In this systematic review, we aimed to systematically assess the effects of antileukotrienes for the prevention and treatment of chronic lung disease in very preterm newborns. We included five clinical studies and ten animal studies. The clinical studies consisted of one RCT and four non-randomized studies. These five clinical studies and three of the animal studies examined the leukotriene antagonist montelukast. We did not find it meaningful to pool results because of the differences in study design and the high overall risk of bias. Drawing definitive conclusions on basis of the existing evidence is thus not possible.

Overall completeness and applicability of evidence
The animal studies had an overall unclear risk of bias due to poor reporting. None of the experimental studies reported on sequence generation, baseline characteristics, allocation concealment, random housing, blinding of the caregivers or random outcome assessment. Only Jouvencel [37] reported adequately on blinding of the outcome assessor, and Demir [36] on completeness of the outcome data, whereas the other studies lacked information. The fact that a protocol was not available for any of the animal studies is also noteworthy. This leads to unclear risk of reporting bias and poor transparency in general. In the case of Schreiber [42] it was also mentioned that data for some outcome was not shown, which causes a high risk of bias for outcome reporting. We classified Phillips [43] as assessing both prevention and treatment effects, as we considered the pups treated with antileukotriene in normoxic conditions as the prevention group and the pups with hyperoxia as the treatment group. Seven studies had an experimental design, i.e. the animals were exposed to two or more different interventions, whereas in the remaining three studies all animals received the same intervention and were therefore defined as observational.
Only one study reported outcome data following hospital discharge [50]. Kim et al. [46] is the first prospective study of montelukast for very preterm infants. Min Kim [48] was a cohort study with a historical control group. The study did not provide a clear definition of the inclusion and exclusion criteria, which leads to a serious risk of bias in the selection of participants.

Relation to other research
A study from 2019 evaluated incidence trends of neonates born very preterm in 11 high-income countries and reported increased BPD rates in most countries [51]. Main reasons for this trend include the increased survival of extremely preterm infants and active resuscitation at lower gestational age. Additional interventions are needed to prevent and treat this condition. Of note, montelukast is already being used as a drug in infants with BPD [52]. Interestingly, in this leaflet released/ published by American Thoracic Society, montelukast is only listed as an anti-inflammatory medicine for children with BPD.
The administration of off-label drugs in neonates are a universal problem. This forces the neonatologist to rely mostly on clinical experience, expert consensus and data extrapolation from patients other than neonates when deciding upon drug choice and dosage [53]. This supports the need for additional high quality research on this topic. We identified one ongoing observational study that aims to explore the effects of montelukast on very low birth weight infants with BPD [35]. The planned sample size, 200 infants, is considerably larger than the clinical studies performed so far and might allow to better explore potential harms of antileukotrienes administration. However, a randomized design would be preferable to assess the efficacy.

Strengths and weaknesses of our review
This is the first systematic review that explores the evidence base of antileukotrienes in very preterm infants in both clinical and animal studies. The review has several strengths. We conducted a comprehensive search with no date or language restrictions. We had studies translated from Mandarin [38] and Korean to English [48]. Further, all the potentially eligible titles and abstracts were screened independently by two authors, as were data extraction and the assessment of risk of bias. We used the most recent and validated tools to assess risk of bias in trials, non-randomized studies and animal studies.
Limitations include our arbitrary definition of neonate animals, i.e. up to 10 days of life. As the definition of a newborn infant (up to 28 days of life) is not based on a specific developmental phase or level of maturation it is not possible to identify a corresponding age in animal models. Further, we did not find any metaanalyses feasible. To retrieve additional information we contacted the authors of two conference abstracts and of the registered protocol we identified in our searches, however, we did not receive any response. Therefore, we could not include these studies and this restriction, though outside of our control, is a potential source of bias.

Implications for research and practice
Refining the existing models to recapitulate the pathology at play in the infants is an urgent matter in order to better evaluate new interventions for BPD [54]. Most animal experiments are carried out to gather information about health in humans and aim to investigate new interventions that are intended for future use in humans. Differences in outcomes in animals and humans are partly due to fundamental biological differences. However, other factors such as for instance design, conduct and reporting play an equally important role [55]. Future animal studies should be designed with higher quality and aim to minimize potential sources of bias, as described in the SYRCLE tool [28]. The registration of the protocols of animal studies in free databases such as https:// precl inica ltria ls. eu/ should become a standard practice and become a formal requirement placed by journals to publish such studies, as is already commonly done for the clinical studies. Similarly, an appropriate randomization should be performed to ensure that animals in each group are in the same housing conditions (e.g. temperature, humidity, light, noise, odors) and to avoid that researchers subjectively select which animals and samples to be used for outcome assessment. Finally, animal studies should clearly report how many animals were used in each step of the experiment, from inclusion to reporting of all outcomes, so that attrition bias can be assessed. As all the animal studies included in this review failed to address these key components in conducting and reporting, the translational value is extremely limited.
Seven of the ten animal studies used an antileukotriene other than montelukast. We speculate that, unless justified by species or pharmacokinetics characteristics, in future animal studies only montelukast should be administered, as only this drug has been used in all clinical studies so far, including the large ongoing study. When choosing outcomes, the focus should be on those with clinical relevance, such as mortality, improved respiratory function and harms.
Future clinical studies should preferably be designed as large, high quality RCTs [56]. New trials are necessary as the harms are not negligible. The findings of the ongoing study [35] with a planned sample size of 200 infants are not available yet. Multicenter RCTs would be an option in order to reach a sufficient sample size. Just as for animal studies, the focus should be on clinically relevant outcomes.

Conclusions
Based on the available evidence, no reliable conclusions about the clinical relevance of antileukotriene administration to very preterm infants can be drawn. Large randomized trials that focus on outcomes relevant to patients and their families, including long-term outcomes, are needed. Animal studies should prioritize montelukast over other antileukotrienes and minimize risks of bias.