Antibiotics for the treatment of lower respiratory tract infections in children with neurodisability: Systematic review

Determine the optimal antibiotic choice for lower respiratory tract infection (LRTI) in children with neurodisability.


| BACKG ROU N D
2][3] Between 7% and 16% of this cohort are likely to have at least one respiratory admission per year, with some having multiple. 4,5Most admissions in CP are for those who are gross motor functional classification scale (GMFCS) V, unable to maintain head control or control a wheelchair.In addition, respiratory admissions for children with CP are longer and more costly when compared to children admitted for respiratory reasons who do not have CP. 2,3Longer and more costly respiratory admissions are also reported for children with other neurodisabilities. 6spiratory-related admissions are variably coded as pneumonia (aspiration, community-acquired and hospital-acquired), asthma, lower respiratory tract infection (LRTI), exacerbation of chronic lung disease or reflux. 7,8The variability of coding for LRTIs adds to the difficulties in researching and understanding this disease and its treatments in this population.
A recent single-centre study of children with non-ambulant CP documented 27 different antibiotic regimens used to treat acute LRTI during 90 admissions over a 5-year timeframe.Antibiotics were administered either enterally, intravenously or inhaled. 8For children with neurodisability admitted for pneumonia, antibiotics are more likely to have a broader spectrum if a diagnosis of aspiration pneumonia is made compared to non-aspiration. 9In children with neurodisability admitted with LRTI or pneumonia, asthma treatment is commonly given alongside antibiotics. 9,10pert consensus for the prevention of respiratory disease in children with CP at highest risk, recommends early antibiotic administration for any signs of acute LRTI. 11The guidelines for the most effective antibiotic regimen, route and duration are not currently available as the evidence is too limited. 5,11Experts disagree on the choice of firstline antibiotic for LRTI in this cohort, compounded by the inability of children to expectorate sputum and clinical instability precluding routine bronchoalveolar lavage, preventing confirmation of the pathogen, causing deterioration and exacerbating uncertainty about antibiotic resistance.Anaerobic bacteria from the mouth are often inferred to be causative, but they are rarely cultured. 12Pseudomonas aeruginosa is commonly grown from airway samples in children that aspirate. 13w the cause of community-acquired pneumonia, Streptococcus pneumoniae, is thought to be the most likely cause. 14termining the most effective antibiotic is important to optimise treatment for children and families, streamline guidelines for clinicians and minimise antibiotic resistance. 15Due to the breadth of treatments identified in children with CP, we aimed to conduct a systematic review to determine which antibiotic is the most effective at treating LRTI in children and adolescents with neurodisability.

| ME THODS
The preferred reporting items for systematic reviews and metaanalyses (PRISMA) guidelines were followed. 16The systematic review protocol was registered with PROSPERO (CRD42022323907) before the search.
We searched the databases of Embase (1974 to 2023 January), Ovid Emcare (1995 to 2023 week 2) and Ovid MEDLINE (1946 to January 2023).Two reviewers (RM and NG) cross-referenced relevant literature, including identified trials, existing review articles, grey literature and reference lists.Experts in the area were also consulted.
The following search strategy was used: Medical Subject Headings and keywords: (1) child AND (2) lower respiratory tract infection AND (3) neurodisability AND (4) antibiotics.Studies reported in English or that were able to be translated by a machine translator engine were included.See Appendix S1 for the detailed search strategy and methods.
Studies were included if they met the following inclusion criteria: at least 50% of participants were aged 0-18 years, with at least 50% having a non-progressive neurodisability.Neurodisability was defined as a group of congenital or acquired long-term conditions attributed to a disturbance of the brain that creates physical limitations.We included all types of studies except for case reports, due to the small number of studies available.LRTI is defined as all types of pneumonia, bronchitis and pneumonitis.No radiological imaging for diagnosis was required for the studies to be included.
for what constitutes LRTI in this cohort is required before designing prospective randomised trials.

K E Y W O R D S
antibiotics, disability, paediatric, pneumonia, respiratory

Key Notes
• Anaerobic antibiotics are effective for the treatment of lower respiratory tract infection (LRTI) in children with neurodisability.
• Pseudomonas aeruginosa cover may be considered if the initial response to treatment is inadequate or when children are more unwell at presentation.
• Anaerobic cover seems to be required, suggesting dental disease may predispose to LRTI.
We included all antibiotics given to treat or prevent LRTI.
Antibiotics were classified according to their spectrum into the following four groups: anaerobic and gram-positive cover; grampositive and gram-negative cover; anaerobic, gram-positive and gram-negative cover, and lastly, anaerobic, gram-positive, gramnegative and P aeruginosa cover (see Table S1).The main outcome was broadly defined as clinical efficacy and this could be measured as the length of stay, length of treatment, length of fever, intensive care admission and mortality.

| Data screening and extraction
An automated screening tool (Research Screener) was used to reduce the screening time of abstracts. 17 Articles that remained post-research screeners were inserted into Covidence for full-text review 18 (see Figure S1).Conflicts were discussed and resolved with an independent reviewer (AW).
Studies found to meet all eligibility criteria had data extracted into a spreadsheet.The following items were extracted independently and then compared: study design, antibiotic treatment, study population and clinical efficacy.The study intervention characteristics were then compared against each other in terms of clinical efficacy.

| Risk of bias and GRADE assessment
Two authors independently appraised the methodological quality of included studies using the Joanna Briggs Institute checklist, 19 appropriate for the study type (quasi-experimental appraisal or randomised control trial), with disagreements reached by consensus. [22]

| Data synthesis
A narrative summary of all the results reported in the included studies was conducted.Further data was sought through author correspondence when desired outcomes were not available in publications.

| RE SULTS
From the 701 articles identified, 52 full-text articles were assessed for eligibility and nine were included (Figure S1).Excluded studies after full-text review with reasons for exclusion are in Appendix S3.

| Included studies
4][25][26][27][28][29] More than 90% of the patients came from one study. 29One article reported a randomised controlled trial, another was a prospective cohort study, and the remainder were retrospective cohort studies.The mean or median age varied between 2.5 and 11 years, and the percentage of males varied between 39% and 64%.An overview of the study characteristics is shown in Table S2, with a more detailed version in Appendix S4.

| Risk of bias and GRADE in included studies
Summaries of the risk of bias are shown in Table S3.Overall, we assessed the included studies as having a moderate risk of bias with all studies scoring at least two domains of study quality as 'unclear' or 'no'.Only two studies were funded by government organisations. 27,29[25] Summaries of the GRADE framework are shown in Table S4. 20,21erall, we assessed the included outcomes with low to very low certainty.Given that most studies were observational, evidence is downgraded to low certainty automatically.The GRADE certainty rating means that the true effect of antibiotics could be anywhere from markedly different to close to the reported outcome.

| Effects of interventions
All nine studies focused on the length of treatment or time to clinical improvement, see Table S4 for a comparison of outcomes.27]29 This varied between 3 and 8 days, with the shortest being 3.4 days for anaerobic and gram-positive cover (e.g., clindamycin) in Jacobson et al. 25 The length of treatment was reported in two studies, 23,25 with treatment being between 7 and 16 days long.The length of fever was reported in all Brook studies, 12,23,24 with a mean of 3-6 days.
Only two studies reported on intensive care admission. 8,29The timing of intensive care admission was different between the studies.Thomson et al. found that patients transferred to intensive care were more likely to receive gram-positive and gram-negative cover, while Dayman et al. found them more likely to receive anaerobic, gram-positive, gram-negative and P. aeruginosa cover.
Only two studies reported on the mortality rate. 8,25Both reported one death.Four of the studies did not report adverse events; therefore, mortality rate may be underestimated.
In terms of different antibiotic covers, most patients received anaerobic cover (85%, n = 4342/5115), with anaerobic, gram-positive and gram-negative cover (55%, n = 2794/5115) being the most common type of antibiotic cover.This was seen in six studies 8,12,[26][27][28][29] and it was frequently adequate.Thomson et al. found that anaerobic, gram-positive and gram-negative cover was better than gram-positive and gram-negative cover. 29Gram-positive and gramnegative cover had the most ICU transfers after day 2 of admission and longest length of stay.Thomson et al. also found that anaerobic, gram-positive and gram-negative cover had little benefit over anaerobic and gram-positive cover.
In terms of P. aeruginosa cover, Thomson et al. found that some patients improved faster on anaerobic, gram-positive, gramnegative and P. aeruginosa cover, particularly if the patient had failed to respond to anaerobic, gram-positive and gram-negative cover (deteriorated within 24 h or no improvement within 48 h), if more unwell at presentation (needing non-invasive ventilation or invasive ventilation within 24 h of presentation) and if more disabled (more comorbidities and/or more technology). 29Unfortunately, we were unable to compare illness severity between studies.

| DISCUSS ION
Our systematic review has identified a limited number of studies published on the treatment of LRTI in children with neurodisability.
Anaerobic, gram-positive and gram-negative cover were acceptable treatments in most patients, supporting the empiric use of oral amoxicillin-clavulanic acid in this population.Given that all regimens had clinical success, P. aeruginosa as a first-line empirical treatment may not be necessary.However, given the low to very low certainty, each child needs to be assessed and reviewed with their comorbidities in mind.Of note, in the era of antibiotic resistance, clinicians should always consider the narrowest possible spectrum of antibiotics.
Due to sampling difficulties, the microbiological aetiology of LRTI in children with neurodisability remains unclear.P. aeruginosa is commonly cultured from airway samples in children with pulmonary aspiration. 13,30,31We would suggest that P. aeruginosa cover be considered when children with neurodisability present critically unwell (requiring non-invasive ventilation or invasive ventilation) at presentation, in more complicated children (four or more comorbidities and use of medical technology assistance, like gastrostomy, etc. 32 ) or if there is no improvement after 2 days. 26,29piration pneumonia is now thought to be part of a continuum involving both community and hospital-acquired pneumonia. 14eatment with narrow-spectrum antibiotics, like penicillin, may be enough if the underlying aetiology is S. pneumoniae.However, Thomson et al. found that children treated with only gram-positive and gram-negative cover did worse compared to those who also received anaerobic cover. 29This may point to the underlying cause being anaerobic bacteria, as seen in dental disease.Lin et al. found   the only intervention to prevent repeat admission for severe pneumonia in children with neurological impairment was routine dental care. 33e closest well-understood pathophysiology to LRTI in neurodisability is bronchiectasis, not associated with cystic fibrosis.It is therefore not unreasonable to draw some inferences from this group in trying to understand the pathophysiology of LRTI in neurodisability.Prospective studies have shown that more than half of children with bronchiectasis have positive respiratory virus studies during exacerbations, and virus-positive exacerbations are more likely to need hospitalisations for further treatment than virus-negative exacerbations. 34Also, antibiotics compared to placebo in children with nonsevere exacerbations of bronchiectasis are more effective. 35sed on a Delphi process with 85% agreement, a group of 60 experts recommended that all children with neurodisability with signs of a LRTI are promptly started on antibiotics. 11This recommendation to start antibiotics early in children with neurodisability is supported by the rationale that the pathophysiology of LRTI is similar to that in bronchiectasis, not associated with cystic fibrosis.Early treatment of LRTI in children with neurodisability is likely to benefit the most severely disabled children by preventing the worsening of exacerbation, the need for intravenous antibiotics and hospital admission.However, it remains controversial since this approach may drive antibiotic resistance.Of note, there is no evidence for either early treatment or watchful waiting in neurodisability.
As well as a poor understanding of the pathophysiology of LRTI in neurodisability, there is no accepted definition of neurodisability or LRTI in children with neurodisability.This makes comparing studies difficult.Should this entity be called aspiration pneumonia, given the lack of clarity regarding causation in most cases?In adult studies, the diagnosis of aspiration pneumonia is made based on clinical history (witnessed macro-aspiration), risk factors and compatible findings on X-ray. 14,36Risk factors in adults are very similar to those described in children with CP with recurrent respiratory symptoms. 4,5In addition to questions about terminology, studies addressing the need and type of airway samples, length of antibiotic treatment, route of administration and when to start antibiotics are also required. 11A clear definition of treatment outcomes and consideration of chest imaging and bacterial cultures are also required.Given the findings of lowquality studies with heterogeneous outcome measures, we would recommend that length of stay be most appropriate for determining antibiotic treatment success as it is patient-centred.
The limitations of the review include the low quality of the studies included, which were at moderate risk of bias and had low to very low certainty in terms of the GRADE framework.Most studies had small numbers of patients, with only one being randomised and another conducted prospectively.An important outcome of this systematic review has been identifying the barriers to conducting vigorously controlled studies in a heterogeneous and vulnerable population like neurodisability, where utilising a placebo would be unethical and could result in rapid deterioration.

| CON CLUS ION
The evidence available on which antibiotics are most effective in treating LRTI in children with neurodisability is poor.In this context, we have documented that anaerobic, gram-positive and gramnegative cover is as effective as P. aeruginosa cover.The anaerobic, gram-positive and gram-negative cover should be used empirically to minimise the risk of antibiotic resistance.This review shows that Before using Research Screener, 28 seed articles were identified (see Appendix S2).These included articles on LRTI in children with neurodisability.The Research Screener removes duplicates and orders abstracts to show the most similar first.All articles were independently reviewed by two authors (RM and NG).