Antibiotics for hospital‐acquired pneumonia in neonates and children

Abstract

Background

Hospital‐acquired pneumonia is one of the most common hospital‐acquired infections in children worldwide. Most of our understanding of hospital‐acquired pneumonia in children is derived from adult studies. To our knowledge, no systematic review with meta‐analysis has assessed the benefits and harms of different antibiotic regimens in neonates and children with hospital‐acquired pneumonia.

Objectives

To assess the beneficial and harmful effects of different antibiotic regimens for hospital‐acquired pneumonia in neonates and children.

Search methods

We searched CENTRAL, MEDLINE, Embase, three other databases, and two trial registers to February 2021, together with reference checking, citation searching, and contact with study authors to identify additional studies.

Selection criteria

We included randomised clinical trials comparing one antibiotic regimen with any other antibiotic regimen for hospital‐acquired pneumonia in neonates and children.

Data collection and analysis

Three review authors independently assessed studies for inclusion, extracted data, and assessed risk of bias. We assessed the certainty of the evidence using the GRADE approach. Our primary outcomes were all‐cause mortality and serious adverse events; our secondary outcomes were health‐related quality of life, pneumonia‐related mortality, non‐serious adverse events, and treatment failure. Our primary time point of interest was at maximum follow‐up.

Main results

We included four randomised clinical trials (84 participants). We assessed all trials as having high risk of bias.

We did not conduct any meta‐analyses, as the included trials did not compare similar antibiotic regimens.

Each of the four trials assessed a different comparison, as follows: cefepime versus ceftazidime; linezolid versus vancomycin; meropenem versus cefotaxime; and ceftobiprole versus cephalosporin.

Only one trial reported our primary outcomes of all‐cause mortality and serious adverse events. Three trials reported our secondary outcome of treatment failure. Two trials primarily included community‐acquired pneumonia and hospitalised children with bacterial infections, hence the children with hospital‐acquired pneumonia constituted subgroups of the total sample sizes.

Where outcomes were reported, the certainty of the evidence was very low for each of the comparisons. We are unable to draw meaningful conclusions from the numerical results.

None of the included trials assessed health‐related quality of life, pneumonia‐related mortality, or non‐serious adverse events.

Authors' conclusions

The relative beneficial and harmful effects of different antibiotic regimens remain unclear due to the very low certainty of the available evidence. The current evidence is insufficient to support any antibiotic regimen being superior to another. Randomised clinical trials assessing different antibiotic regimens for hospital‐acquired pneumonia in children and neonates are warranted.

Plain language summary

Antibiotics for hospital‐acquired pneumonia in newborns and children

Review question

Which antibiotic regimen is safer and more effective in treating neonates (newborns) and children with hospital‐acquired pneumonia?

Background

Hospital‐acquired pneumonia is an inflammation of the tissue of one or both lungs caused by an infection that occurs during a hospital stay (i.e. 48 hours or more after hospital admission). It is one of the most common hospital‐acquired infections in children worldwide, and is associated with a high death rate. Most of our understanding of hospital‐acquired pneumonia in children is drawn from adult studies. To our knowledge this is the first review with meta‐analysis that assesses the benefits and harms of different antibiotic regimens in newborns and children with hospital‐acquired pneumonia.

Search date

The evidence is current to February 2021.

Study characteristics

We included four trials randomising 84 children with hospital‐acquired pneumonia to different antibiotic regimens. Three trials were multicentre trials from the USA, Latin America, Europe, and South Africa. The South African trial included one site in Malaysia. Each of the four included trials compared different antibiotic regimens, as follows: cefepime versus ceftazidime; linezolid versus vancomycin; meropenem versus cefotaxime; and ceftobiprole versus cephalosporin.

Study funding sources

Three trials were funded by pharmaceutical companies (Zeneca Pharmaceuticals, Pharmacia Corp, and Basilea Pharmaceutica International Ltd.), indicating a possible risk of bias related to a vested interest risk.

Key results

Each of the four included trials compared different antibiotic regimens, as follows: cefepime versus ceftazidime; linezolid versus vancomycin; meropenem versus cefotaxime; and ceftobiprole versus cephalosporin.

Only one trial reported our primary outcomes of death from all causes and serious adverse events (major complications). Three trials reported our secondary outcome of treatment failure. Two trials primarily included community‐acquired pneumonia and hospitalised children with bacterial infections, hence the children with hospital‐acquired pneumonia constituted only subgroups of the total study populations.

Where outcomes were reported, the certainty of the evidence was very low for each of the comparisons. We were unable to draw any meaningful conclusions from the numerical results.

None of the included trials assessed health‐related quality of life, pneumonia‐related death, or non‐serious adverse events (minor complications).

Conclusions

The available evidence does not suggest that one antibiotic regimen is safer and more effective than another in treating newborns and children with hospital‐acquired pneumonia. Further research is needed.

Certainty of the evidence

The certainty of evidence is very low. All four included trials had high risk of bias (i.e. the studies were designed in such a way that the results may have been skewed). In addition, the included trials involved few participants, which is likely to have led to inaccurate results.

Background

Description of the condition

Hospital‐acquired pneumonia (also known as nosocomial pneumonia) is defined as pneumonia that occurs 48 hours or more after hospital admission (Eccles 2014; Kalil 2016; Torres 2017).

Epidemiology

Hospital‐acquired infection is a serious complication of hospitalisation worldwide in adults and children (Polin 2012; Zingg 2017). The incidence of hospital‐acquired infections is between 0.17% and 36% of hospitalised paediatric patients (Polin 2012; Vijay 2018). Variations in incidence may be due to differences in diagnostic criteria as well as differences in local risk factors for the development of hospital‐acquired infections (Polin 2012; Vijay 2018). The highest incidences are seen in neonatal intensive care units (NICUs) and paediatric intensive care units (PICUs) (Iosifidis 2018; Polin 2012; Stein 1994; Zingg 2017). European studies suggest that the incidence of hospital‐acquired infections is higher in paediatric surgical wards (17%) compared with general paediatrics wards (2.5%) (Li 2019).

Hospital‐acquired pneumonia is one of the most common hospital‐acquired infections in children worldwide (Alvares 2019). Hospital‐acquired pneumonia in neonates and children accounts for 6.8% to 32.3% of all hospital‐acquired infections (Polin 2012; Stein 1994; Zingg 2017). It is therefore a frequent cause of hospital‐acquired infection in patients in the NICU or PICU, only surpassed by catheter‐associated bloodstream infections (Bigham 2009; Cernada 2013; Polin 2012; Richards 1999; Zingg 2017). The incidence is particularly high amongst premature neonates or neonates with low birth weight (Apisarnthanarak 2003; Tan 2014). Paediatric hospital‐acquired pneumonia has been shown to be associated with increased mortality (Bigham 2009; Iosifidis 2018). Hospital‐acquired pneumonia is associated with even higher mortality and morbidity in preterm neonates (Apisarnthanarak 2003).

The vast majority of hospital‐acquired pneumonia is ventilator‐associated pneumonia, a subtype of hospital‐acquired pneumonia. Ventilator‐associated pneumonia is defined as "pneumonia that occurs 48 hours or more after endotracheal intubation" (Cernada 2013; Iosifidis 2018; Joram 2012; Kalil 2016; Torres 2017). The incidence of ventilator‐associated pneumonia is reported to be 2.9 to 11.6 cases per 1000 ventilator days (de Neef 2019; Jarvis 1991; Joram 2012).

Even though most research is focused on ventilator‐associated pneumonia, non‐ventilatory hospital‐acquired pneumonia has similar, or even higher mortality rates and financial costs than ventilator‐associated pneumonia, whilst its incidence could be underestimated (Davis 2012; Giuliano 2018).

Risk factors

Risk factors for hospital‐acquired pneumonia are prolonged hospitalisation, mechanical ventilation, serious underlying illnesses (e.g. lung disease, immune deficiency), bloodstream infections, recent antimicrobial therapy, genetic syndromes, immunosuppression, use of steroids, prematurity, low birth weight, reintubation or self‐extubation, and bronchoscopy (Aelami 2014; Liu 2013; Stein 1994). Newborns, preterms, and infants are especially prone to infections, due to a developmental deficiency in the innate, adaptive immune systems, usage of endotracheal and orogastric tubes, exposure to broad‐spectrum antibiotic agents, and parenteral nutrition (Aelami 2014; Polin 2012; Tan 2014). This broad range of risk factors increases the risk of hospital‐acquired pneumonia; however, they are associated with different kinds of pathogens (Mourani 2017; Polin 2012), therefore one antibiotic regimen for all patients might not be warranted.

The onset of ventilator‐associated pneumonia is also a risk factor associated with specific pathogens and prognosis (Ewig 1999; Kalil 2016; Safdar 2005). Early‐onset ventilator‐associated pneumonia and late‐onset ventilator‐associated pneumonia are distinguished by whether the ventilator‐associated pneumonia occurs before or after the first four days of hospitalisation (Langer 1987). Early‐onset ventilator‐associated pneumonia is associated with a better prognosis than late‐onset ventilator‐associated pneumonia (Kalil 2016; Safdar 2005).

Pathophysiology

Hospital‐acquired pneumonia is most often caused by aspiration of bacteria from the pharynx, oral cavity, or the upper gastrointestinal tract (Polin 2012). The increased risk of ventilator‐associated pneumonia after intubation is caused by endotracheal tubes bypassing the initial host barrier defence mechanisms (Polin 2012). In the absence of the endotracheal tube as a direct portal of entry for pathogens, non‐ventilatory hospital‐acquired pneumonia could be caused by the contiguous spread of a primary infection at a distant site (Polin 2012), or by specific conditions of susceptibility of the patient. For example, hospital‐acquired pneumonia is more frequent in patients who are subjected to several emergency procedures, or who have skin and mucous lesions, which cause a disruption of natural membrane defences, with an increased risk of the infection spreading. Hence, there is a higher rate of hospital‐acquired pneumonia in paediatric patients hospitalised for an injury including the head and neck, and those with firearm or pulmonary injuries (Cutler 2017). Moreover, the trauma itself generates an impairment of immunological defences of the patients, making them more prone to infections (Pories 1991).

Microbiology

The most common pathogens involved in hospital‐acquired pneumonia worldwide are Enterobacteriaceae,Pseudomonas aeruginosa, and Staphylococcus aureus (Jones 2010; Patel 2000; Srinivasan 2009; van der Zwet 2005; Weiner‐Lastinger 2020). Gram‐negative bacteria cause 67.5% of hospital‐acquired pneumonia in children, whereas gram‐positive bacteria and respiratory viruses cause 13% and 12.6% of hospital‐acquired pneumonia in children, respectively (Wang 2010). However, when comparing different geographical regions, the pathogens, their antibiotic susceptibility, the burden of disease, and diagnostic methods vary (Bigham 2009; Iosifidis 2018; van der Zwet 2005). In particular, some studies show that viruses such as rhinovirus, influenza, and parainfluenza could be as common as bacterial pathogens in causing hospital‐acquired pneumonia in non‐ventilated children and adults (Shorr 2017; Zinna 2016). Several observational studies show that infections caused by multidrug‐resistant (MDR) pathogens increase the risk of death, length of hospital stay, and healthcare costs (Su 2020).

Furthermore, rhinovirus and enterovirus are the most commonly recognised pathogens causing hospital‐acquired viral respiratory infection in both adults and paediatric patients (Chow 2017; Zinna 2016).  

Diagnosis

The diagnosis of hospital‐acquired pneumonia and ventilator‐associated pneumonia is based upon a combination of imaging test evidence of a lung disease plus clinical evidence that the infiltrate is of an infectious origin (Kalil 2016). 

• Radiological test (e.g. X‐ray image) could show a new and persistent or progressive and persistent lung infiltrate, consolidation, cavitation, or pneumatocele (in infants younger than one year old) (Gunalan 2021; Magill 2013).

• Sign and symptoms may vary depending on the age of the patient, as follows. 

  • "For children > 1 year old or ≤ 12 years old: fever, leukocytosis, new onset of purulent sputum or change in character of sputum, increased respiratory secretions or increased suctioning requirements, new onset or worsening cough, dyspnoea, apnoea, or tachypnoea, rales or bronchial breath sounds, worsening gas exchange" (Gunalan 2021; Magill 2013).

  • "For children < 1 year old worsening of gas exchange with increased oxygen requirements is the most common presentation, in association with temperature instability, leukopenia (≤ 4000 WBC/mm³) or leukocytosis (≥ 15,000 WBC/mm³), new onset of purulent sputum or change in character of sputum, increased respiratory secretions or increased suctioning requirements, apnoea, tachypnoea, nasal flaring, wheezing, cough, bradycardia (< 100 beats/min) or tachycardia (> 170 beats/min)" (Gunalan 2021; Magill 2013).

The clinical symptoms of hospital‐acquired pneumonia are non‐specific, and no combination of signs and symptoms has been found to be highly sensitive or specific for the diagnosis (Fabregas 1999; Ferrer 2019). Nevertheless, no gold standard exists for the diagnosis of hospital‐acquired pneumonia (Chang 2016; Iosifidis 2018).

Description of the intervention

The treatment of hospital‐acquired pneumonia can be either empirical (initiation of an antibiotic regimen before the aetiological pathogen is known) or based on the results of microbiologic studies. The decision to treat empirically is based primarily on the clinical presentation of the patient (Kalil 2016; Torres 2017). Early initiation and appropriate antimicrobial therapy of hospital‐acquired pneumonia has been shown to significantly reduce morbidity and mortality in adults (Kelly 2019). Current guidelines for adults recommend that the choice of antibiotics should be based on local antibiograms, local distribution of pathogens, and individual risk factors for serious infection, MDR pathogens, or if P aeruginosa is suspected (Kalil 2016; Kelly 2019; Torres 2017).

Patients assessed as being at low risk of antibiotic resistance and early‐onset hospital‐acquired pneumonia or ventilator‐associated pneumonia are recommended for initial empiric therapy with a narrow‐spectrum antibiotic, whereas high‐risk patients will require broader therapy with a combination of different classes of antimicrobials (Kelly 2019; NICE 2019; Torres 2017).

In the case of low risk of methicillin‐resistant S aureus (MRSA), the American Thoracic Society guidelines recommend piperacillin‐tazobactam, cefepime, levofloxacin, imipenem, or meropenem for S aureus, P aeruginosa, and other gram‐negative bacilli (the last only for patients suspected of having ventilator‐associated pneumonia) (Kalil 2016).

If there is a risk of MRSA, the American Thoracic Society guidelines recommend vancomycin or linezolid (Kalil 2016). Whether to initiate monotherapy or combination therapy depends on the risk of gram‐negative bacteria or risk of antimicrobial resistance, or both (Kalil 2016; Weiss 2020).

Antibiotics such as aminoglycoside and colistin are not recommended, unless alternative agents with adequate gram‐negative activity are unavailable (Kalil 2016).

The role of viruses in causing hospital‐acquired pneumonia in neonates and children might also be taken into account. The confirmation of a viral organism when routine cultures are negative might facilitate antibiotic discontinuation (Shorr 2017).

Guidelines for the treatment of hospital‐acquired pneumonia focus primarily on adults (Kalil 2016; Kelly 2019; NICE 2019; Torres 2017); however, it should be noted that children differ from adults with hospital‐acquired pneumonia due to differences in pathogenesis, pharmacokinetics, and types of pathogens (Fernandez 2011; Jain 2015; Stephenson 2005). Consequently, evidence from adult studies cannot be directly transmitted to treatment regimens in children.

How the intervention might work

Hospital‐acquired pneumonia could be both a viral or bacterial infection. Considering that viral pneumonia does not require antibiotic therapy unless a mixed infection or secondary bacterial infection is suspected, one of the main objectives of empirical treatment of hospital‐acquired pneumonia is to kill the bacteria. Antibiotics are therefore an essential part of the treatment of hospital‐acquired pneumonia.

Antibiotics may be classified by their: "1) mechanism of action (bactericidal or bacteriostatic); 2) bacterial spectrum (broad or narrow); and 3) chemical structure (e.g. penicillins, aminoglycosides, macrolides, glycopeptides, or quinolones)" (Bérdy 2005; Korang 2021b; Korang 2021c).

The empirical treatment for suspected hospital‐acquired pneumonia should provide coverage for the most likely bacteria. This may result in antibiotic combination therapy if there is a suspicion of either MDR pathogens or severe infection (Kalil 2016; Weiss 2020). The rationale of combination therapy is to widen the spectrum of the empirical antibiotic regimen to increase the likelihood of covering the causative bacteria. Theoretically, combination therapy might also suppress the occurrence of resistant subpopulations (Allan 1985; Milatovic 1987). A recent guideline has been created to determine whether to continue or stop the empirical antibiotic after 48 to 72 hours of treatment (Shein 2019)

An optimal empirical antibiotic treatment would ideally reduce disease progression of the pneumonia and avoid the development of sepsis and septic shock (Chang 2016; Weiss 2020). This would in turn reduce the risk of death and complications (Chang 2016). By clearing the pathogen, an optimal antibiotic regimen would also speed up the recovery and thereby reduce the discomfort and work of breathing that a child may experience during such an infection.

Why it is important to do this review

Hospital‐acquired pneumonia is one of the most common nosocomial infections amongst neonates and children (Cernada 2013; Polin 2012). Current guidelines are directed solely towards adults (Kelly 2019; Martin‐Loeches 2018). Most of our understanding of hospital‐acquired pneumonia in children is derived from adult studies; however, there exist many differences between neonates/children and adults with respect to hospital‐acquired pneumonia (such as the pattern of causative agents isolated, risk factors, and diagnostic methods) (Iosifidis 2018; Vijay 2018). The certainty of evidence from adult studies will also generally tend to be downgraded due to the indirectness of the evidence (Guyatt 2011a; Weiss 2020). No previous systematic review with meta‐analysis has assessed the benefits and harms of different antibiotic regimens for children with hospital‐acquired pneumonia. There is a need for a systematic review with meta‐analysis to provide the necessary evidence for the effects of antibiotics in children with hospital‐acquired pneumonia.

Objectives

To assess the beneficial and harmful effects of different antibiotic regimens for hospital‐acquired pneumonia in neonates and children.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised clinical trials reported as full text, abstract only, and unpublished data. We excluded trials with a cross‐over design and cluster‐randomised trials.

Types of participants

We included neonates (< 28 days old) and children (< 18 years of age) suspected of, or diagnosed with, hospital‐acquired pneumonia (as defined by the trialists).

Types of interventions

We included trials comparing one antibiotic regimen with any other antibiotic regimen or placebo. We included the following antibiotic groups.

1. Beta‐lactam antibiotics

   1. Narrow‐spectrum penicillins (penicillin G, oxacillin, dicloxacillin, cloxacillin, nafcillin, and methicillin).

   2. Broad‐spectrum penicillins (e.g. amoxicillin, ampicillin, piperacillin, ticarcillin, mezlocillin, and carbenicillin).

   3. Penicillins combined with beta‐lactamase inhibitors (e.g. piperacillin/tazobactam and amoxicillin/clavulanic acid).

   4. Cephalosporins (e.g. cefuroxime, cefotaxime, ceftazidime, cefazolin, cefalexin, cefotetan, cefoxitin, ceftriaxone, cefepime, cefazolin, ceftobiprole, and cefoperazone).

   5. Carbapenems (e.g. meropenem, imipenem, doripenem, and ertapenem).

   6. Monobactams (aztreonam).

2. Aminoglycosides (e.g. amikacin, tobramycin, and gentamicin).

3. Quinolones (e.g. ciprofloxacin, ofloxacin, temafloxacin, garenoxacin, gatifloxacin, grepafloxacin, sparfloxacin, levofloxacin, and moxifloxacin).

4. Macrolides (e.g. azithromycin, clarithromycin, and erythromycin).

5. Glycopeptides (e.g. vancomycin and teicoplanin).

6. Lincosamides (e.g. clindamycin).

7. Antibacerial oxazolidinone agents (e.g. linezolid).

8. Nitroimidazoles (e.g. metronidazole) (Korang 2019).

We also planned to assess any antibiotic regimen (such as either piperacillin‐tazobactam, cefepime, levofloxacin, or meropenem/imipenem) that covers patients at low risk of having an MDR pathogen compared to an antibiotic regimen (such as a combination of either piperacillin‐tazobactam, cefepime/ceftazidime, levofloxacin/ciprofloxacin, meropenem/imipenem, or amikacin/gentamicin/tobramycin plus either vancomycin or linezolid) that covers patients at high risk of having an MDR pathogen.

Types of outcome measures

Primary outcomes

1. All‐cause mortality.

2. Proportion of participants with one or more serious adverse events. We used the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use ‐ Good Clinical Practice (ICH‐GCP) definition of a serious adverse event, which is any untoward medical occurrence that resulted in death, was life‐threatening, required hospitalisation or prolonging of existing hospitalisation, and resulted in persistent or significant disability or jeopardised the participant (ICH‐GCP 2016). If the trialists did not use the ICH‐GCP definition, we included the data if the trialists used the term 'serious adverse event'. If the trialists did not use the ICH‐GCP definition or this term, then we included the data if the event clearly fulfilled the ICH‐GCP definition for a serious adverse event. We planned to assess each type of serious adverse event separately (Korang 2021b; Korang 2021c).

Secondary outcomes

1. Health‐related quality of life (any continuous scale used by the trialists).

2. Pneumonia‐related mortality (as defined by trialists).

3. Proportion of participants with one or more non‐serious adverse event (any adverse event which was not classified as "serious" or which did not clearly fulfilled the ICH‐GCP definition for a serious adverse event ). We planned to assess each reported adverse event separately.

4. Proportion of participants with treatment failure. We defined treatment failure as clinical deterioration or recurrence of clinical signs leading to any modification of the assigned empirical antibiotic treatment (we accepted similar definitions as defined by the trialists).

We used the trial results reported closest to one month as our primary time point of interest for all outcomes.

Search methods for identification of studies

Electronic searches

We searched the following databases from inception to present.

1. The Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library (1 February 2021, Issue 2).

2. MEDLINE Ovid (from 1946 to 1 February 2021).

3. Embase Ovid (from 1974 to 1 February 2021).

We also searched the following databases.

1. CINAHL via EBSCOhost (Cumulative Index to Nursing and Allied Health Literature) (from 1961 to 1 February 2021).

2. PsycINFO via EBSCOhost (from 1967 to 1 February 2021).

3. Science Citation Index Expanded (Web of Science) (from 1900 to 1 February 2021) and Conference Proceedings Citation Index – Science (Web of Science) (from 1990 to 1 February 2021).

4. LILACS (Latin American and Caribbean Health Science Information database) (from 1982 to 1 February 2021).

We used the search strategy described in Appendix 1 to search MEDLINE. We combined the MEDLINE search with the Cochrane Highly Sensitive Search Strategy for randomised trials: sensitivity and precision‐maximising version (2008 revision) (Lefebvre 2011).

We also conducted a search of the US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov) (1 February 2021) and the World Health Organization International Clinical Trials Registry Platform (apps.who.int/trialsearch/) (1 February 2021).

Searching other resources

We checked the reference lists of all primary studies and review articles for additional references. We contacted experts in the field to identify additional unpublished materials.

In an effort to identify unpublished trials, we searched clinical trial registers of Europe and the USA and the websites of pharmaceutical companies, the US Food and Drug Administration (FDA), and the European Medicines Agency (EMA).

We searched for errata or retractions from the included studies published in full text on PubMed (23 March 2021) (www.ncbi.nlm.nih.gov/pubmed).

Data collection and analysis

Selection of studies

Three review authors (SKK, CN, SPM) independently screened the titles and abstracts of records identified by the search for potential inclusion in the review. We retrieved selected full‐text study reports/publications, and three review authors (SKK, CN, SPM) independently screened the full‐texts and identified trials for inclusion, and identified and recorded reasons for exclusion of the ineligible studies. Any disagreements were resolved through discussion or by consulting a fourth review author (JCJ) if required. We excluded duplicates and collated multiple reports of the same study so that each study, rather than each report, was the unit of interest in the review. We recorded the selection process in sufficient detail to complete a PRISMA flow diagram (Figure 1) and Characteristics of excluded studies table (Moher 2009). We did not impose any language or publication restrictions.

1.

Study flow diagram.

Data extraction and management

We used a data collection form to record study characteristics and outcome data that we piloted on at least one study in the review. One review author (SKK or CN or SPM) extracted trial characteristics from the included trials. We extracted the following trial characteristics.

1. Methods: trial design, total duration of trial, number of trial centres and location, trial setting, withdrawals, and date of trial.

2. Participants: number of participants, mean age, age range, sex, microbial agent isolated, severity of condition, diagnostic criteria, baseline lung function, smoking history (of participants or parents, or both), inclusion criteria, and exclusion criteria.

3. Interventions: intervention (including dosage, route of administration, and length of empirical treatment), comparison, co‐interventions, and excluded medications.

4. Outcomes: primary and secondary outcomes specified and collected, and time points reported.

5. Notes: funding for trial, and notable conflicts of interest of trial authors.

Three review authors (SKK, CN, SPM) independently extracted outcome data from the included trials. We noted in the Characteristics of included studies table if outcome data were not reported in a useable way. Any disagreements were resolved by consensus or by consulting a fourth review author (JCJ). One review author (SKK) entered the data into Review Manager 5 software (Review Manager 2020). We double‐checked that data were entered correctly by comparing the data presented in the systematic review with the study reports.

Assessment of risk of bias in included studies

Three review authors (SKK, CN, SPM) independently assessed the risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017). Any disagreements were resolved by discussion or by involving another review author (JCJ). We assessed risk of bias according to the following domains.

1. Random sequence generation.

2. Allocation concealment.

3. Blinding of participants and personnel.

4. Blinding of outcome assessment.

5. Incomplete outcome data.

6. Selective outcome reporting.

7. Other bias.

We graded each potential source of bias as low, high, or unclear and provided a quote from the study report together with a justification for our judgement in the risk of bias table. We summarised the risk of bias judgements across different studies for each of the domains listed. We considered the domains blinding of outcome assessment, incomplete outcome data, and selective outcome reporting separately for different key outcomes, where necessary. Where information on risk of bias related to unpublished data or correspondence with a trialist, we noted this in the risk of bias table.

When considering treatment effects, we took into account the risk of bias for the studies that contributed to that outcome.

Overall risk of bias

We assessed overall risk of bias as follows.

1. Low risk of bias: we classified the outcome of a trial as overall 'low risk of bias' only if all domains were classified as at low risk of bias.

2. Unclear risk of bias: we classified the outcome of a trial as overall 'unclear' risk of bias if one or more domains were classified as unclear, and no domain was at high risk of bias.

3. High risk of bias: we classified the outcome of a trial as overall 'high risk of bias' if at least one domain was classified as high risk of bias.

See Appendix 2 for further details.

We planned to assess confidence in network meta‐analysis results using CINeMA (Confidence in Network Meta‐Analysis) (Nikolakopoulou 2020; Papakonstantinou 2020).

Assessment of bias in conducting the systematic review

We conducted the review according to the published protocol (Korang 2021a), and reported any deviations from it in the Differences between protocol and review section.

Measures of treatment effect

We entered the outcome data for each trial into the data tables in Review Manager 5 to calculate the treatment effects (Review Manager 2020).

Dichotomous outcomes

We calculated risk ratios (RRs) with 95% confidence interval (CI) for dichotomous outcomes.

Continuous outcomes

We planned to calculate the mean differences (MDs) and the standardised mean difference (SMD) with 95% CI for continuous outcomes.

We planned to perform meta‐analysis only if the treatments, participants, and the underlying clinical question were similar enough for pooling to make sense.

Unit of analysis issues

The unit of analysis was the participating children in individually randomised trials.

Dealing with missing data

We contacted trial investigators to obtain missing outcome data where possible. If the missing data were unobtainable, we explored the impact of the missing data in a sensitivity analysis (Sensitivity analysis).

If numerical outcome data such as standard deviations or correlation coefficients were missing, and they could not be obtained from the trial authors, we would calculate them from other available statistics such as P values according to the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021).

We did not impute missing values for any outcomes in our primary analysis. We planned to impute data in two sensitivity analyses.

Assessment of heterogeneity

We planned to visually inspect forest plots for signs of heterogeneity, and to explore possible heterogeneity in our prespecified subgroup analyses. We inspected trial characteristics across trials to identify clinical heterogeneity. We planned to assess the presence of statistical heterogeneity by the Chi² test (threshold P < 0.10) and to measure the quantities of heterogeneity using the I² statistic (Higgins 2002; Higgins 2003). If we detected moderate or high heterogeneity, we explored possible causes (e.g. differences in study design, participants, interventions, or completeness of outcome assessments) (Korang 2019).

We defined the level of heterogeneity as:

1. 0% to 40%: might not be important;

2. 30% to 60%: may represent moderate heterogeneity;

3. 50% to 90%: may represent substantial heterogeneity; and

4. 75% to 100%: may represent considerable heterogeneity.

We would evaluate whether a meta‐analysis should be avoided if the level of heterogeneity indicated that pooling of data was not justified (Higgins 2021).

Assessment of reporting biases

We planned to use a funnel plot to assess publication bias only if we included 10 or more trials. We planned to visually inspect funnel plots to assess risk of bias. We planned to test asymmetry with the Harbord test (Harbord 2006).

Data synthesis

We planned to pool data from trials that we judged to be clinically homogeneous. We planned to perform meta‐analysis only if more than one trial provided relevant data in any single comparison.

Meta‐analysis

We planned to undertake meta‐analyses according to the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions, Higgins 2021, and our protocol (Korang 2021a). We used Review Manager 5 software (Review Manager 2020).

We planned to assess our intervention effects with both fixed‐effect and random‐effects meta‐analyses (DeMets 1987; DerSimonian 1986). We planned to use the more conservative point estimate of the two. We considered the point estimate closest to zero effect as the more conservative point (Jakobsen 2014). As we have chosen two primary outcomes, we considered a P value of 0.033 or less as the threshold for evidence of a difference (Jakobsen 2014). We planned to use the eight‐step procedure provided by Jakobsen and colleagues to assess if the threshold for any evidence of a difference was crossed (Jakobsen 2014). Where data were available from only one trial, we planned to use Fisher’s exact test for dichotomous data (Fisher 1922).

We planned that if the ranking of the identified interventions was unclear based on aggregating the meta‐analysis results, we would perform a network meta‐analysis (see Appendix 3).

In addition to the primary meta‐analysis, we planned to use Trial Sequential Analysis (TSA) as a secondary analysis (see Appendix 4).

Subgroup analysis and investigation of heterogeneity

We planned to carry out the following subgroup analyses.

1. Trials at high risk of bias compared to trials at low risk of bias.

2. Age: newborn (less than 1 month), infants (1 month to 1 year), children of preschool age (1 to 5 years), children of school age (5 to 12 years), adolescents (older than 12 years).

3. Trials from high‐income countries compared to trials from low‐ and middle‐income countries, as defined by the World Bank (World Bank 2020).

4. Suspected hospital‐acquired pneumonia without radiological verification or culture of respiratory specimens compared to hospital‐acquired pneumonia with radiological verification or culture of respiratory specimens at randomisation.

5. Empirical compared to targeted treatment based on bacterial cultures, if possible.

6. Hospital‐acquired pneumonia compared to ventilator‐associated pneumonia.

7. Early‐onset compared to late‐onset, defined as onset of ventilator‐associated pneumonia before or after four days.

8. Length of antibiotic treatment: three days or shorter, four to seven days, or longer than seven days.

9. Participants without underlying diseases compared to participants with underlying diseases such as genetic syndromes, lung disease, or immune deficiency.

We planned to use the Chi² test to test for subgroup interactions in Review Manager 5 (Review Manager 2020).

Sensitivity analysis

To assess the potential impact of missing data, we planned to perform two sensitivity analyses on the primary outcomes, as follows.

1. ‘Best‐worst‐case’ scenario: we assumed that all participants lost to follow‐up in the experimental group survived and had no serious adverse event. We assumed that all of those with missing outcomes in the control group did not survive and had a serious adverse event.

2. 'Worst‐best‐case’ scenario: we assumed that all participants lost to follow‐up in the experimental group did not survive and had a serious adverse event. We assumed that all participants lost to follow‐up in the control group survived and had no serious adverse event (Jakobsen 2014).

Summary of findings and assessment of the certainty of the evidence

We created two summary of findings tables (Table 1; Table 2) reporting our primary and secondary outcomes. We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness, and publication bias) to assess the certainty of a body of evidence as it relates to the studies which contribute data to the meta‐analyses for the prespecified outcomes (Atkins 2004). We used the methods and recommendations in Section 8.5 and Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), employing GRADEpro GDT software (GRADEpro GDT). We justified all decisions to down‐ or upgrade the quality of the evidence using footnotes and made comments to aid the reader's understanding of the review where necessary.

Summary of findings 1. Cefepime compared with ceftazidime for hospital‐acquired pneumonia in neonates and children.

Cefepime compared with ceftazidime for hospital‐acquired pneumonia in neonates and children
Patient or population: neonates and children with hospital‐acquired pneumonia Settings: hospital Intervention: cefepime Comparison: ceftazidime
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No. of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Ceftazidime	Cefepime
All‐cause mortality Maximum follow‐up (Time point was not described.)	200 per 1000	28 per 1000 (2 to 510)	RR 0.14 (0.01 to 2.55)	30 (1 study)	⊕⊝⊝⊝ Very lowa	OIS: 3262 (RR 0.80, α 0.033, β 0.20, Pc 20%)
Serious adverse events Maximum follow‐up (Time point was not described.)	200 per 1000	28 per 1000 (2 to 510)	RR 0.14 (0.01 to 2.55)	30 (1 study)	⊕⊝⊝⊝ Very lowa	OIS: 3262 (RR 0.80, α 0.033, β 0.20, Pc 20%) Serious adverse events were deaths.
Treatment failure Maximum follow‐up (Time point was not described.)	400 per 1000	200 per 1000 (60 to 656)	RR 0.50 (0.15 to 1.64)	30 (1 study)	⊕⊝⊝⊝ Very lowa	OIS: 1272 (RR 0.80, α 0.033, β 0.20, Pc 40%)
Health‐related quality of life Maximum follow‐up	‐	‐	‐	‐	‐	This outcome was not reported.
Pneumonia‐related mortality Maximum follow‐up	‐	‐	‐	‐	‐	This outcome was not reported.
Non‐serious adverse events Maximum follow‐up	‐	‐	‐	‐	‐	This outcome was not reported.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; OIS: optimal information size; RR: risk ratio; Pc: percentage in control group
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aDowngraded by 1 level for study limitations due to serious risk of bias, and two levels for imprecision due to very small information size.

Summary of findings 2. Linezolid compared with vancomycin for hospital‐acquired pneumonia in neonates and children.

Linezolid compared with vancomycin for hospital‐acquired pneumonia in neonates and children
Patient or population: neonates and children with hospital‐acquired pneumonia Settings: hospital Intervention: linezolid Comparison: vancomycin
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No. of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Vancomycin	Linezolid
All‐cause mortality Maximum follow‐up (up to 35 days)	‐	‐	‐	‐	‐	This outcome was not reported.
Serious adverse events Maximum follow‐up (up to 35 days)	‐	‐	‐	‐	‐	This outcome was not reported.
Treatment failure Maximum follow‐up (up to 35 days)	154 per 1000	315 per 1000 (75 to 1000)	RR 2.05 (0.49 to 8.63)	32 (1 study)	⊕⊝⊝⊝ Very lowa	OIS: 4438 (RR 0.80, α 0.033, β 0.20, Pc 15.4%)
Health‐related quality of life Maximum follow‐up	‐	‐	‐	‐	‐	This outcome was not reported.
Pneumonia‐related mortality Maximum follow‐up	‐	‐	‐	‐	‐	This outcome was not reported.
Non‐serious adverse events Maximum follow‐up	‐	‐	‐	‐	‐	This outcome was not reported.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; OIS: optimal information size; RR: risk ratio; Pc: percentage in control group
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aDowngraded by 1 level for study limitations due to serious risk of bias, and 2 levels for imprecision due to very small information size.

Had we performed a network meta‐analysis, we would also have used CINeMA to assess the certainty of a body of evidence (Guyatt 2008; Guyatt 2011b; Schünemann 2003).

Results

Description of studies

For study details, see Characteristics of included studies, Characteristics of excluded studies, and Characteristics of ongoing studies.

We based our assessment of the included trials on the recommendations provided in the Cochrane Handbook for Systematic Reviews of Interventions, Higgins 2021, and our protocol (Korang 2021a).

Results of the search

We searched seven databases (see Electronic searches) and retrieved 1846 records. Our searches of the trial registers identified four further studies. Our searches of other resources identified no additional studies appearing to meet the inclusion criteria. Screening reference lists of the included publications revealed two potentially relevant studies. We therefore retrieved a total of 1852 records, which amounted to 1499 records after de‐duplication. We excluded 1466 records based on title and abstract, and obtained the full texts of the remaining 33 records. We excluded 28 studies (see Characteristics of excluded studies). We identified one ongoing trial that might include children with hospital‐acquired pneumonia (Shahrin 2020).

We included four trials reported in eight articles. Our screening process is illustrated in a PRISMA flow diagram (see Figure 1).

Included studies

Four trials met our inclusion criteria (Bosheva 2021; Jantausch 2003; Schuler 1995; Shahid 2008). Four additional papers were included as secondary publications to Jantausch 2003 (Deville 2003; Kaplan 2003; Meissner 2003; Saiman 2003). For study details, see Characteristics of included studies. Three trials were multicentre trials and included 59 sites in the USA and Latin America (Jantausch 2003), 12 sites in Europe (Bosheva 2021), 22 sites in Europe and South Africa (Schuler 1995), and one site in Malaysia (Shahid 2008). Three trials included children with different infections, of which a minor portion was hospital‐acquired pneumonia (Bosheva 2021; Jantausch 2003; Schuler 1995). Two trials primarily included community‐acquired pneumonia, Bosheva 2021, and hospitalised children with bacterial infections (Schuler 1995), hence the children with hospital‐acquired pneumonia constituted only subgroups of the total sample sizes. One trial included participants with late‐onset ventilator‐associated pneumonia only (Shahid 2008).

Participants

The four trials randomised a total of 84 participants. The studies included participants of the following age groups:

1. under one year (Shahid 2008);

2. birth to 12 years (Jantausch 2003);

3. 3 months to 12 years (Schuler 1995);

4. 3 months to 18 years (Bosheva 2021).

Interventions

The four trials compared four different antibiotic regimens, as follows:

1. cefepime versus ceftazidime (Shahid 2008);

2. linezolid versus vancomycin (Jantausch 2003);

3. ceftobiprole versus standard of care (cephalosporin) (Bosheva 2021);

4. meropenem versus cefotaxime (Schuler 1995).

One trial administered metronidazole in cases of mixed aerobic/anaerobic infection, but only for the control group (Schuler 1995). One trial described the use of vancomycin in the control group when MRSA was confirmed or suspected, and similarly amikacin, gentamicin, or tobramycin was administered in the control groups when infection by P aeruginosa was confirmed or suspected (Bosheva 2021).

Co‐interventions

One trial provided concomitant treatment in both groups with amikacin, vancomycin, gentamicin, or tobramycin for confirmed or suspected infection caused by P aeruginosa, which could be added at the discretion of the blinded investigator (Bosheva 2021). The remaining trials did not report co‐interventions (Jantausch 2003; Schuler 1995; Shahid 2008).

Outcomes

Only one trial reported sparse data on all‐cause mortality and serious adverse events (Shahid 2008). This trial did not report serious adverse events according to the ICH‐GCP, but reported the number of deaths in each group. Three trials reported treatment failure (Jantausch 2003; Schuler 1995; Shahid 2008). None of the included trials assessed health‐related quality of life, pneumonia‐related mortality, or non‐serious adverse events.

Antibiotic resistance in included trials

One trial reported that resistance to study medications was not found in pathogens isolated at baseline in either group (Jantausch 2003). The remaining trials did not report on antibiotic resistance amongst the culture‐positive participants having hospital‐acquired pneumonia (Bosheva 2021; Schuler 1995; Shahid 2008)

Excluded studies

We assessed 28 trials as potentially relevant upon review of the abstract, but that were later excluded after full‐text review. The reasons for exclusion were as follows.

1. Wrong participant population, such as adults or participants that did not have hospital‐acquired pneumonia (25 trials) (Agweyu 2015; Amonova 2011; Awad 2014; Barradas 1989; Bassetti 1991; Begue 1998; Bhavnani 2012; Chaudhary 2008; Chuchalin 1997; Cometta 1994; Dickstein 2016; Giamarellos‐Bourboulis 2014; Grudinina 2002; Iakovlev 2000; Iakovlev 2006; Joshi 1999; Kollef 2012; Lacy 2015; Mohamed 2018; Muscedere 2012; Nassar 2018; Norrby 1993; Rodloff 1996; Straneo 1990; Tucker 2017).

2. Not randomised (1 trial) (Berman 2004).

3. The same antibiotic assessed in the intervention and control groups (2 trials) (Fatehi 2019; Patel 2015).

For details, see Characteristics of excluded studies.

Risk of bias in included studies

We assessed all of the included trials as at overall high risk of bias (see Figure 2; Figure 3).

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

3.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Allocation

Three trials used a computer‐generated assignment sequence resulting in 'low risk of bias' (Bosheva 2021; Schuler 1995; Shahid 2008). One trial did not describe how allocation sequence generation was performed and was therefore judged as 'unclear risk of bias' (Jantausch 2003).

None of the trials described allocation concealment, therefore we judged all four trials as 'unclear risk of bias' (Bosheva 2021; Jantausch 2003; Schuler 1995; Shahid 2008).

Blinding

Two trials were unblinded and were assessed as 'high risk of bias' in both domains (Jantausch 2003; Schuler 1995). One trial did not describe blinding and was assessed as 'unclear risk of bias' in both domains (Shahid 2008). The remaining trial blinded the investigators but not the treatment providers/participants, resulting in a judgement of 'low risk of bias' for blinding of outcome assessment and 'high risk of bias' for blinding of participants and personnel (Bosheva 2021).

Incomplete outcome data

Two trials had a high number of dropouts and were assessed as 'high risk of bias' (Schuler 1995; Shahid 2008). Two trials did not describe dropouts and were assessed as 'unclear risk of bias' (Bosheva 2021; Jantausch 2003).

Selective reporting

One trial reported both mortality and serious adverse events, resulting in a judgement of 'low risk of bias' (Shahid 2008). Three trials did not have a protocol, nor did they report mortality and serious adverse events, resulting in a judgement of 'unclear risk of bias' (Bosheva 2021; Jantausch 2003; Schuler 1995).

Other potential sources of bias

We found that three trials were funded by the pharmaceutical companies that produce the studied antibiotics (Bosheva 2021; Jantausch 2003; Schuler 1995).  Amongst those, we found one trial to be at high risk of other bias due to for‐profit bias (Bosheva 2021) as the pharmaceutical company had a major involvement in designing the study and in the acquisition of data, statistical analysis, and article preparation. The other two (Jantausch 2003; Schuler 1995) were funded and may be influenced by vested interests, but no involvement of the funders was described. We found no other potential sources of bias in the remaining trials (Jantausch 2003; Schuler 1995; Shahid 2008). 

Effects of interventions

See: Table 1; Table 2

The four trials assessed different comparisons of antibiotic regimens, therefore we did not perform any meta‐analyses, trials sequential analyses, or subgroup analyses on any our outcomes. Of the two trials that had under 10 participants, one did not report our prespecified outcomes (Bosheva 2021), and the other trial had an unclear comparison (Schuler 1995); we therefore decided not to include a summary of findings table for these two comparisons.

Three trials compared two beta‐lactam antibiotics (Bosheva 2021; Schuler 1995; Shahid 2008), and one trial compared a glycopeptide with an antibacterial agent (oxazolidinone) (Jantausch 2003).

Cefepime compared with ceftazidime for late‐onset ventilator‐associated pneumonia

Primary outcomes

All‐cause mortality

One trial showed that the effect of cefepime on mortality compared to ceftazidime is very uncertain (risk ratio (RR) 0.14, 95% confidence interval (CI) 0.01 to 2.55; 1 trial, 30 participants; very low‐certainty evidence; Analysis 1.1) (Shahid 2008). We calculated the optimal information size based on a relative risk reduction (RRR) of 20%, an alpha of 3.3%, a beta of 20%, and the observed incidence in the control group (15.4%) (Guyatt 2011c; Higgins 2011). Calculation of the optimal information size showed that we did not have sufficient information to confirm or reject that cefepime compared with ceftazidime reduced the risk of death by 20% or more.

1.1. Analysis.

Comparison 1: Cefepime compared with ceftazidime, Outcome 1: All‐cause mortality

This outcome was assessed as high risk of bias (Figure 2), and the certainty of the evidence was very low (Table 1).

Proportion of participants with one or more serious adverse events

One trial showed that the effect of cefepime on serious adverse events compared to ceftazidime is very uncertain (RR 0.14, 95% CI 0.01 to 2.55; 1 trial, 30 participants; very low‐certainty evidence; Analysis 1.2) (Shahid 2008). We calculated the optimal information size based on an RRR of 20%, an alpha of 3.3%, a beta of 20%, and the observed incidence in the control group (15.4%) (Guyatt 2011c; Higgins 2011). Calculation of the optimal information size showed that we did not have sufficient information to confirm or reject that cefepime versus ceftazidime reduced the risk of having a serious adverse event by 20% or more.

1.2. Analysis.

Comparison 1: Cefepime compared with ceftazidime, Outcome 2: Serious adverse events

This outcome was assessed as high risk of bias (Figure 2), and the certainty of the evidence was very low (Table 1).

Secondary outcomes

Shahid 2008 did not report health‐related quality of life, pneumonia‐related mortality, or non‐serious adverse events for participants with hospital‐acquired pneumonia.

Proportion of participants with treatment failure

One trial showed that the effect of cefepime on serious adverse events compared to ceftazidime is very uncertain (RR 0.50, 95% CI 0.15 to 1.64; 1 trial, 30 participants; very low‐certainty evidence; Analysis 1.3) (Shahid 2008). We calculated the optimal information size based on an RRR of 20%, an alpha of 3.3%, a beta of 20%, and the observed incidence in the control group (40%) (Guyatt 2011c; Higgins 2011). Calculation of the optimal information size showed that we did not sufficient information to confirm or reject that cefepime versus ceftazidime reduced the risk of having treatment failure by 20% or more.

1.3. Analysis.

Comparison 1: Cefepime compared with ceftazidime, Outcome 3: Treatment failure

This outcome was assessed as high risk of bias (Figure 2), and the certainty of the evidence was very low (Table 1).

As we only included one trial in this comparison, we did not perform a subgroup analysis, sensitivity analysis, or funnel plot.

Linezolid compared with vancomycin for hospital‐acquired pneumonia

Primary outcomes

Jantausch 2003 did not report all‐cause mortality or the proportion of participants with one or more serious adverse events for hospital‐acquired pneumonia.

Secondary outcomes

Jantausch 2003 did not report health‐related quality of life, pneumonia‐related mortality, or non‐serious adverse events for participants with hospital‐acquired pneumonia.

Proportion of participants with treatment failure

One trial showed that the effect of linezolid on treatment failure compared to vancomycin is very uncertain (RR 2.05, 95% CI 0.49 to 8.63; 1 trial, 32 participants; very low‐certainty evidence; Analysis 2.1) (Jantausch 2003). We defined treatment failure as participants not being clinically cured. We calculated the optimal information size based on an RRR of 20%, an alpha of 3.3%, a beta of 20%, and the observed incidence in the control group (15.4%) (Guyatt 2011c; Higgins 2011). Calculation of the optimal information size showed that we did not have sufficient information to confirm or reject that linezolid versus vancomycin reduced the risk of having treatment failure by 20% or more.

2.1. Analysis.

Comparison 2: Linezolid compared with vancomycin, Outcome 1: Treatment failure

This outcome was assessed as high risk of bias (Figure 2), and the certainty of the evidence was very low (Table 2).

As we only included one trial in this comparison, we did not perform a subgroup analysis, sensitivity analysis, or funnel plot.

Ceftobiprole compared with standard of care (cephalosporin) for hospital‐acquired pneumonia

Primary outcomes

Bosheva 2021 did not report all‐cause mortality or the proportion of participants with one or more serious adverse events for hospital‐acquired pneumonia.

Secondary outcomes

Bosheva 2021 did not report health‐related quality of life, pneumonia‐related mortality, the proportion of participants with one or more non‐serious adverse events, or the proportion of participants with treatment failure for hospital‐acquired pneumonia.

As we only included one trial in this comparison, we did not perform a subgroup analysis, sensitivity analysis, or funnel plot.

Meropenem compared with cefotaxime for hospital‐acquired pneumonia

Primary outcomes

Schuler 1995 did not report all‐cause mortality or the proportion of participants with one or more serious adverse events for hospital‐acquired pneumonia.

Secondary outcomes

Schuler 1995 did not report health‐related quality of life, pneumonia‐related mortality, or non‐serious adverse events for participants with hospital‐acquired pneumonia.

Proportion of participants with treatment failure

One trial showed that the effect of meropenem on treatment failure compared to cefotaxime is very uncertain (RR 1.80, 95% CI 0.10 to 31.52; 1 trial, 6 participants; very low‐certainty evidence; Analysis 3.1) (Schuler 1995). We defined treatment failure as participants not being clinically cured. It was not possible to calculate the optimal information size, as we did not observe any incidence in the control group (Guyatt 2011c; Higgins 2011). However, as the trial included only six participants, it is fair to assume that there was insufficient information to confirm or reject that meropenem versus cefotaxime reduced the risk of having treatment failure by 20% or more.

3.1. Analysis.

Comparison 3: Meropenem compared with cefotaxime, Outcome 1: Treatment failure

This outcome was assessed as high risk of bias (Figure 2), and the certainty of the evidence was very low.

As we only included one trial in this comparison, we did not perform any subgroup analysis, sensitivity analysis, or funnel plot.

Discussion

Summary of main results

We included four trials randomising a total of 84 participants. We assessed all trials as having high risk of bias.

We did not conduct any meta‐analyses as the included trials all compared different antibiotic regimens. None of the comparisons reached the optimal information size (Table 1; Table 2).

Where outcomes were reported, the certainty of the evidence was very low for each of the comparisons. We were unable to draw any meaningful conclusions from the numerical results. None of the included trials assessed health‐related quality of life, pneumonia‐related mortality, or non‐serious adverse events.

Overall completeness and applicability of evidence

The relative beneficial and harmful effects of different antibiotic regimens remain unclear due to a lack of well‐powered trials and high risk of systematic errors. The current evidence is insufficient to support any antibiotic regimen being superior to another. Large randomised clinical trials assessing different antibiotic regimens for hospital‐acquired pneumonia in children and neonates are warranted.

Quality of the evidence

Heterogeneity

We did not perform a meta‐analysis, therefore we did not assess heterogeneity.

Risk of systematic error ('bias')

All of the included trials were at high risk of bias.

It was not possible to assess publication bias, as only four trials were included in the review.

Risk of random error ('play of chance')

None of the trial results reached the optimal information size. It was not possible to perform trial sequential analysis, as we performed no meta‐analysis.

GRADE

We assessed the certainty of the evidence for each of the outcomes using the GRADE approach (Table 1; Table 2). The evidence for each reported outcome was of very low certainty. The rationale for the GRADE assessment is given in footnotes (Table 1; Table 2).

Potential biases in the review process

The main limitation of this review is the low number of randomised participants, and the very low certainty of the available evidence. A further limitation is that we could not perform a meta‐analysis, as none of the trials compared similar antibiotics.

There might also be a difference between the pathogens and their antibiotic susceptibility in different countries. The optimal antibiotic regimen will therefore vary according to country and local risks of antibiotic resistance. The number of included trials was insufficient to confirm or reject this presumption.

The lack of a gold standard for the diagnosis of hospital‐acquired pneumonia could result in clinical heterogeneity between studies. Some participants may have had an unrecognised viral aetiology of hospital‐acquired pneumonia. Although these cases of viral infection must be assumed to be equally distributed in the intervention groups, they might lead to an underestimation of the differences between different antibiotic regimens.

Agreements and disagreements with other studies or reviews

Although hospital‐acquired pneumonia is one of the most common nosocomial infections amongst the paediatric population, we are not aware of any other reviews assessing the effects of different antibiotic regimens in paediatric patients with hospital‐acquired pneumonia.

In the latest review conducted by the National Institute for Health and Care Excellence to develop the therapeutic guideline for hospital‐acquired pneumonia (NICE 2019), studies with a mixed population of hospital‐acquired pneumonia and community‐acquired pneumonia were excluded, unless ≥ 75% were a hospital‐acquired pneumonia population. Moreover, NICE 2019 excluded studies with a mixed population of ventilator‐ and non‐ventilator‐associated pneumonia where data could not be analysed separately. As such, no studies involving children were identified and included, even if they were recognised as a subgroup of interest in the study protocol. Instead, the authors decided to make recommendations for paediatric therapy on the basis of higher‐quality evidence on adults.

Overall, whilst including six randomised clinical trials and one post hoc analysis, the NICE 2019 authors did not find any difference in their analysis of the effectiveness of antibiotic regimens in adults (NICE 2019). Likewise, they did not find any statistical difference in adverse effects between antibiotics or classes of antibiotics in people with hospital‐acquired pneumonia.

Although there are similarities between children and adults with hospital‐acquired pneumonia (e.g. pathogenesis), evidence from adult studies cannot be transmitted to treatment regimens in children with certainty. The spectrum of responsible bacteria may differ, as colonisation in the pharynx and trachea varies, particularly in young children, who are less commonly colonised with S aureus, P aeruginosa, and MDR pathogens, compared to adults (Jain 2015). Furthermore, antibiotic pharmacokinetics vary between adults and children (Fernandez 2011; Stephenson 2005), and children have less frequent underlying lung diseases and known risk factors such as chronic lung disease and chronic renal failure, which may influence the severity and treatment response (Sopena 2014). Choice of antibiotics, as well as dosing and treatment duration, should thus be evaluated in children with hospital‐acquired pneumonia.

Authors' conclusions

Implications for practice.

Based on the currently available evidence, we were unable to confirm or reject whether one antibiotic regimen is superior to another.

Implications for research.

Randomised clinical trials are needed to assess the effects of different antibiotic regimens for hospital‐acquired pneumonia. Such trials should:

1. randomise a sufficient number of participants to demonstrate reliable results;

2. assess treatment failure, all‐cause mortality, and serious adverse events; and

3. be conducted such that there is a low risk of bias.

History

Protocol first published: Issue 1, 2021

Appendices

Appendix 1. Search strategy

Cochrane Central Register of Controlled Trials in the Cochrane Library (CENTRAL; 2021, Issue 2):

#1 MeSH descriptor: [Anti‐Bacterial Agents] explode all trees

#2 (antibiot* or antimicrob*)

#3 MeSH descriptor: [Aminoglycosides] explode all trees

#4 MeSH descriptor: [Carbapenems] explode all trees

#5 MeSH descriptor: [Cephalosporins] explode all trees

#6 MeSH descriptor: [Glycopeptides] explode all trees

#7 MeSH descriptor: [Lincosamides] explode all trees

#8 MeSH descriptor: [Macrolides] explode all trees

#9 MeSH descriptor: [Monobactams] explode all trees

#10 MeSH descriptor: [Nitroimidazoles] explode all trees

#11 MeSH descriptor: [Penicillins] explode all trees

#12 MeSH descriptor: [Quinolones] explode all trees

#13 (Aminoglycosides or Antibacerial oxazolidinone agents or Beta‐lactam antibiotics or Carbapenems or Cephalosporins or Glycopeptides or Lincosamides or Macrolides or Monobactams or Nitroimidazoles or Penicillins or Quinolones or amikacin or amoxicillin or ampicillin or azithromycin or aztreonam or carbenicillin or cefazolin or cefepime or cefoperazone or cefotaxime or cefotetan or cefoxitin or ceftazidime or ceftobiprole or ceftriaxone or cefuroxime or cephalexin or ciprofloxacin or clarithromycin or clavulanic acid or clindamycin or Cloxacillin or Dicloxacillin or doripenem or ertapenem or erythromycin or garenoxacin or gatifloxacin or gentamycin or grepafloxacin or imipenem or levofloxacin or linezolid or meropenem or Methicillin or metronidazole or mezlocillin or moxifloxacin or Nafcillin or ofloxacin or Oxacillin or penicillin G or piperacillin or sparfloxacin or tazobactam or teicoplanin or temafloxacin or ticarcillin or tobramycin or vancomycin)

#14 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13

#15 MeSH descriptor: [Healthcare‐Associated Pneumonia] explode all trees

#16 MeSH descriptor: [Pneumonia, Ventilator‐Associated] explode all trees

#17 ((pneumonia* and (((hospital or ventilator or health‐care or health care) and (aquired or associated)) or nosocomial)) or HAP or VAP)

#18 #15 or #16 or #17

#19 MeSH descriptor: [Adolescent] explode all trees

#20 MeSH descriptor: [Child] explode all trees

#21 MeSH descriptor: [Infant] explode all trees

#22 (child* or P*ediat* or infant* or bab* or pre*school or lactant* or neonat* or adolesc* or school*child or youth* or toddler* or teen* or boy* or girl* or student* or juvenile* or minor* or pubescen* or young* or newborn)

#23 #19 or #20 or #21 or 322

#24 #14 and #18 and #23

MEDLINE Ovid (1946 to February 2021)

1. exp Anti‐Bacterial Agents/

2. (antibiot* or antimicrob*).mp. [mp=title, abstract, original title, name of substance word, subject heading word, floating sub‐heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]

3. exp Aminoglycosides/

4. exp Carbapenems/

5. exp Cephalosporins/

6. exp Glycopeptides/

7. exp Lincosamides/

8. exp Macrolides/

9. exp Monobactams/

10. exp Nitroimidazoles/

11. exp Penicillins/

12. exp Quinolones/

13. (Aminoglycosides or Antibacerial oxazolidinone agents or Beta‐lactam antibiotics or Carbapenems or Cephalosporins or Glycopeptides or Lincosamides or Macrolides or Monobactams or Nitroimidazoles or Penicillins or Quinolones or amikacin or amoxicillin or ampicillin or azithromycin or aztreonam or carbenicillin or cefazolin or cefepime or cefoperazone or cefotaxime or cefotetan or cefoxitin or ceftazidime or ceftobiprole or ceftriaxone or cefuroxime or cephalexin or ciprofloxacin or clarithromycin or clavulanic acid or clindamycin or Cloxacillin or Dicloxacillin or doripenem or ertapenem or erythromycin or garenoxacin or gatifloxacin or gentamycin or grepafloxacin or imipenem or levofloxacin or linezolid or meropenem or Methicillin or metronidazole or mezlocillin or moxifloxacin or Nafcillin or ofloxacin or Oxacillin or penicillin G or piperacillin or sparfloxacin or tazobactam or teicoplanin or temafloxacin or ticarcillin or tobramycin or vancomycin).mp. [mp=title, abstract, original title, name of substance word, subject heading word, floating sub‐heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]

14. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13

15. exp Healthcare‐Associated Pneumonia/

16. exp Pneumonia, Ventilator‐Associated/

17. ((pneumonia* and (((hospital or ventilator or health‐care or health care) and (aquired or associated)) or nosocomial)) or HAP or VAP).mp. [mp=title, abstract, original title, name of substance word, subject heading word, floating sub‐heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]

18. 15 or 16 or 17

19. exp Adolescent/ or exp Child/ or exp Infant/

20. (child* or P*ediat* or infant* or bab* or pre*school or lactant* or neonat* or adolesc* or school*child or youth* or toddler* or teen* or boy* or girl* or student* or juvenile* or minor* or pubescen* or young* or newborn).mp. [mp=title, abstract, original title, name of substance word, subject heading word, floating sub‐heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]

21. 19 or 20

22. 14 and 18 and 21

23. (randomized controlled trial or controlled clinical trial).pt. or clinical trials as topic.sh. or trial.ti.

24. (random* or blind* or placebo* or meta‐analys*).mp. [mp=title, abstract, original title, name of substance word, subject heading word, floating sub‐heading word, keyword heading word, organism supplementary concept word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]

25. 22 and (23 or 24)

Embase Ovid (1974 to February 2021)

1. exp antiinfective agent/

2. (antibiot* or antimicrob*).mp. [mp=title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword, floating subheading word, candidate term word]

3. exp aminoglycoside/

4. exp carbapenem derivative/

5. exp cephalosporin derivative/

6. exp glycopeptide/

7. exp lincosamide/

8. exp macrolide/

9. exp monobactam derivative/

10. exp nitroimidazole derivative/

11. exp penicillin derivative/

12. exp quinolone derivative/

13. (Aminoglycosides or Antibacerial oxazolidinone agents or Beta‐lactam antibiotics or Carbapenems or Cephalosporins or Glycopeptides or Lincosamides or Macrolides or Monobactams or Nitroimidazoles or Penicillins or Quinolones or amikacin or amoxicillin or ampicillin or azithromycin or aztreonam or carbenicillin or cefazolin or cefepime or cefoperazone or cefotaxime or cefotetan or cefoxitin or ceftazidime or ceftobiprole or ceftriaxone or cefuroxime or cephalexin or ciprofloxacin or clarithromycin or clavulanic acid or clindamycin or Cloxacillin or Dicloxacillin or doripenem or ertapenem or erythromycin or garenoxacin or gatifloxacin or gentamycin or grepafloxacin or imipenem or levofloxacin or linezolid or meropenem or Methicillin or metronidazole or mezlocillin or moxifloxacin or Nafcillin or ofloxacin or Oxacillin or penicillin G or piperacillin or sparfloxacin or tazobactam or teicoplanin or temafloxacin or ticarcillin or tobramycin or vancomycin).mp. [mp=title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword, floating subheading word, candidate term word]

14. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13

15. exp health care associated pneumonia/

16. exp ventilator associated pneumonia/

17. ((pneumonia* and (((hospital or ventilator or health‐care or health care) and (aquired or associated)) or nosocomial)) or HAP or VAP).mp. [mp=title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword, floating subheading word, candidate term word]

18. 15 or 16 or 17

19. exp Adolescent/ or exp Child/ or exp Infant/

20. (child* or P*ediat* or infant* or bab* or pre*school or lactant* or neonat* or adolesc* or school*child or youth* or toddler* or teen* or boy* or girl* or student* or juvenile* or minor* or pubescen* or young* or newborn).mp. [mp=title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword, floating subheading word, candidate term word]

21. 19 or 20

22. 14 and 18 and 21

23. Randomized controlled trial/ or Controlled clinical study/ or trial.ti.

24. (random* or blind* or placebo* or meta‐analys*).mp. [mp=title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword, floating subheading word, candidate term word]

25. 22 and (23 or 24)

LILACS (Bireme; 1982 to February 2021)

(antibiot$ or antimicrob$) or (Aminoglycosides or Antibacerial oxazolidinone agents or Beta‐lactam antibiotics or Carbapenems or Cephalosporins or Glycopeptides or Lincosamides or Macrolides or Monobactams or Nitroimidazoles or Penicillins or Quinolones or amikacin or amoxicillin or ampicillin or azithromycin or aztreonam or carbenicillin or cefazolin or cefepime or cefoperazone or cefotaxime or cefotetan or cefoxitin or ceftazidime or ceftobiprole or ceftriaxone or cefuroxime or cephalexin or ciprofloxacin or clarithromycin or clavulanic acid or clindamycin or Cloxacillin or Dicloxacillin or doripenem or ertapenem or erythromycin or garenoxacin or gatifloxacin or gentamycin or grepafloxacin or imipenem or levofloxacin or linezolid or meropenem or Methicillin or metronidazole or mezlocillin or moxifloxacin or Nafcillin or ofloxacin or Oxacillin or penicillin G or piperacillin or sparfloxacin or tazobactam or teicoplanin or temafloxacin or ticarcillin or tobramycin or vancomycin) [Words] and ((pneumonia$ and (((hospital or ventilator or health‐care or health care) and (aquired or associated)) or nosocomial)) or HAP or VAP) [Words] and (child$ or P$ediat$ or infant$ or bab$ or pre$school or lactant$ or neonat$ or adolesc$ or school$child or youth$ or toddler$ or teen$ or boy$ or girl$ or student$ or juvenile$ or minor$ or pubescen$ or young$ or newborn) [Words]

Science Citation Index EXPANDED (1900 to February 2021) and Conference Proceedings Citation Index – Science (1990 to February 2021) (Web of Science)

#8 #7 AND #6

#7 TI=(random* or blind* or placebo* or meta‐analys* or trial*) OR TS=(random* or blind* or placebo* or meta‐analys*)

#6 #5 AND #4 AND #3

#5 TS=(child* or P*ediat* or infant* or bab* or pre*school or lactant* or neonat* or adolesc* or school*child or youth* or toddler* or teen* or boy* or girl* or student* or juvenile* or minor* or pubescen* or young* or newborn)

#4 TS=((pneumonia* and (((hospital or ventilator or health‐care or health care) and (aquired or associated)) or nosocomial)) or HAP or VAP)

#3 #2 OR #1

#2 TS=(Aminoglycosides or Antibacerial oxazolidinone agents or Beta‐lactam antibiotics or Carbapenems or Cephalosporins or Glycopeptides or Lincosamides or Macrolides or Monobactams or Nitroimidazoles or Penicillins or Quinolones or amikacin or amoxicillin or ampicillin or azithromycin or aztreonam or carbenicillin or cefazolin or cefepime or cefoperazone or cefotaxime or cefotetan or cefoxitin or ceftazidime or ceftobiprole or ceftriaxone or cefuroxime or cephalexin or ciprofloxacin or clarithromycin or clavulanic acid or clindamycin or Cloxacillin or Dicloxacillin or doripenem or ertapenem or erythromycin or garenoxacin or gatifloxacin or gentamycin or grepafloxacin or imipenem or levofloxacin or linezolid or meropenem or Methicillin or metronidazole or mezlocillin or moxifloxacin or Nafcillin or ofloxacin or Oxacillin or penicillin G or piperacillin or sparfloxacin or tazobactam or teicoplanin or temafloxacin or ticarcillin or tobramycin or vancomycin)

#1 TS=(antibiot* or antimicrob*)

CINAHL (Ebsco host; February 2021) (317 hits)

S11 S9 AND S10

S10 TX (random* or blind* or placebo* or meta‐analys*)

S9 S4 AND S8

S8 S5 OR S6 OR S7

S7 TX ((pneumonia* and (((hospital or ventilator or health‐care or health care) and (aquired or associated)) or nosocomial)) or HAP or VAP)

S6 MH pneumonia, ventilator‐associated

S5 MH healthcare‐associated pneumonia

S4 S1 OR S2 OR S3

S3 TX (Aminoglycosides or Antibacerial oxazolidinone agents or Beta‐lactam antibiotics or Carbapenems or Cephalosporins or Glycopeptides or Lincosamides or Macrolides or Monobactams or Nitroimidazoles or Penicillins or Quinolones or amikacin or amoxicillin or ampicillin or azithromycin or aztreonam or carbenicillin or cefazolin or cefepime or cefoperazone or cefotaxime or cefotetan or cefoxitin or ceftazidime or ceftobiprole or ceftriaxone or cefuroxime or cephalexin or ciprofloxacin or clarithromycin or clavulanic acid or clindamycin or Cloxacillin or Dicloxacillin or doripenem or ertapenem or erythromycin or garenoxacin or gatifloxacin or gentamycin or grepafloxacin or imipenem or levofloxacin or linezolid or meropenem or Methicillin or metronidazole or mezlocillin or moxifloxacin or Nafcillin or ofloxacin or Oxacillin or penicillin G or piperacillin or sparfloxacin or tazobactam or teicoplanin or temafloxacin or ticarcillin or tobramycin or vancomycin)

S2 TX (antibiot* or antimicrob*)

S1 MH antibiotics

Appendix 2. Risk of bias assessment

Allocation sequence generation

1. Low risk: if sequence generation was achieved using a computer random number generator or a random numbers table. Drawing lots, tossing a coin, shuffling cards, and throwing dice are also considered adequate if performed by an independent adjudicator.

2. Unclear risk: if the method of randomisation was not specified, but the trial is still presented as being randomised.

3. High risk: if the allocation sequence was not randomised or was only quasi‐randomised.

Allocation concealment

1. Low risk: if the allocation of participants was performed by a central, independent unit, onsite locked computer, identical‐looking numbered, sealed envelopes, drug bottles or containers prepared by an independent pharmacist or investigator.

2. Unclear risk: if the trial was classified as randomised, but the allocation concealment process was not described.

3. High risk: if the allocation sequence was familiar to the investigators who assigned participants.

Blinding of participants and treatment providers

1. Low risk: if the participants and the treatment providers were blinded to intervention allocation, and this was described.

2. Unclear risk: if the blinding procedure was insufficiently described.

3. High risk: if blinding of participants and treatment providers was not performed.

Blinding of outcome assessment

1. Low risk of bias: if it was mentioned that outcome assessors were blinded, and this was described.

2. Unclear risk of bias: if blinding of outcome assessors was not mentioned, or the extent of blinding is insufficiently described.

3. High risk of bias: if no blinding or incomplete blinding of outcome assessors was performed.

Incomplete outcome data

1. Low risk of bias: if missing data were unlikely to make treatment effects depart from plausible values. This could be either:

   1. there were no dropouts or withdrawals for all outcomes; or

   2. the numbers and reasons for the withdrawals and dropouts for all outcomes were clearly stated and are similar between groups. Generally, the trial was judged as at low risk of bias due to incomplete outcome data if dropouts were less than 5%; however, this cut‐off was not definitive.

2. Unclear risk of bias: if there was insufficient information to assess whether missing data were likely to introduce bias into the results.

3. High risk of bias: if the results were likely to be biased due to missing data, either because the pattern of dropouts could be described as differing between the two intervention groups, or the trial used improper methods in dealing with the missing data (e.g. 'last observation carried forward').

Selective outcome reporting

1. Low risk of bias: if a protocol was published before or at the time the trial was begun, and the outcomes specified in the protocol were reported on. If there is no protocol, or the protocol was published after the trial was begun, reporting of all‐cause mortality and serious adverse events granted the trial a grade of low risk of bias.

2. Unclear risk of bias: if no protocol was published, and the outcomes all‐cause mortality and serious adverse events were not reported.

3. High risk of bias: if the outcomes in the protocol were not reported on.

Other bias

1. Low risk of bias: if the trial appears to be free of other components that could put it at risk of bias (e.g. academic bias or for‐profit bias).

2. Unclear risk of bias: if the trial may or may not be free of other components that could put it at risk of bias.

3. High risk of bias: if there are other factors in the trial that could put it at risk of bias (e.g. the authors have conducted trials on the same topic, for‐profit bias, etc.).

Overall risk of bias

1. Low risk of bias: we classified the outcome of a trial as overall 'low risk of bias' only if all domains were classified as at low risk of bias.

2. Unclear risk of bias: we classified the outcome of a trial as overall 'unclear' risk of bias if one or more domains were classified as unclear, and no domain was at high risk of bias.

3. High risk of bias: we classified the outcome of a trial as overall 'high risk of bias' if at least one domain was classified as high risk of bias.

We graded each potential source of bias as low, high, or unclear, and provided a justification for our judgement in the risk of bias table. We planned to perform a sensitivity analysis considering trials with domains at unclear risk of bias as overall high risk of bias because meta‐epidemiologic studies suggest that they tend to overestimate positive intervention effects and underestimate negative effects in the same way as domains with high risk of bias (Hróbjartsson 2012; Hróbjartsson 2013; Hróbjartsson 2014; Moustgaard 2020; Savovic 2018). We summarised the risk of bias judgements across different trials for each of the domains listed. Where information on risk of bias relates to unpublished data or correspondence with a trialist, we noted this in the risk of bias table. When considering treatment effects, we took into account the risk of bias for the trials that contributed to that outcome.

Appendix 3. Network meta‐analysis

We planned to obtain information about the antibiotic regimens of interest either from head‐to‐head trials, or from trials comparing an antibiotic regimen with another antibiotic regimen, or placebo. Hence, the synthesis comparator set consisted of all the antibiotic regimens listed in Types of interventions as well as a placebo. We analysed each specific antibiotic regimen separately.

We generated descriptive statistics for each treatment comparison describing important clinical and methodological characteristics (e.g. publication year, participant age). Each outcome data set would be presented in a different network diagram, where the size of the nodes was proportional to the total number of randomised participants, and the width of each edge was weighted according to the number of studies comparing the connected treatments. We planned to additionally plot the edges of each network according to the average risk of bias per treatment comparison, using green for low, yellow for moderate, and red for high risk of bias. We anticipated that any participant who met the inclusion criteria was, in principle, equally likely to be randomised to any of the interventions in the synthesis comparator set. We would perform network meta‐analysis using Stata 16.1 (command: mvmeta) under the frequentist framework (Stata 2019), using the network suite of commands (White 2015). The network meta‐analysis synthesises evidence for the comparative effectiveness of more than two alternative interventions for the same condition (Korang 2020; Shim 2017).

We planned only to perform network meta‐analysis if a connected network of trials could be conducted (Mills 2013).

If network meta‐analysis was possible, we would assess a priori the two prerequisite assumptions: transitivity and consistency. We planned to assess for the transitivity assumption across treatment comparisons in the network using box plots, and evaluate the assumption of consistency using the design‐by‐treatment interaction model as a global test (Higgins 2003; Shim 2017). Effect modifiers would be age, ethnicity (based on country of participants), type of pneumonia (hospital‐acquired pneumonia or ventilator‐associated pneumonia), onset of pneumonia (early or late onset), existence of underlying diseases (e.g. genetic syndromes, lung disease, or immune deficiency), length of treatment (3 days or shorter, 4 to 5 days, 6 to 7 days, or longer than 7 days). We planned to evaluate the transitivity assumption for carrying out a network meta‐analysis using these effect modifiers. We would also explore these through network subgroup meta‐analyses. If we concluded that the transitivity and consistency assumptions were not met, we would not perform network meta‐analysis, but would present direct and indirect evidence separately.

We would report the estimation of each treatment comparison separately using the relevant effect size (risk ratio), a 95% confidence interval, and a 95% prediction interval. We planned to use the network forest plot to illustrate the summary effect size of the comparative effectiveness amongst the antibiotic regimens. Along the estimated effect sizes, we would present the ranking probabilities for each antibiotic regimen being at each possible rank, as well as the surface under the cumulative ranking curve (SUCRA) (Räcker 2015; Salanti 2011). We planned to use a rank‐heat plot to depict the SUCRA values (and their 95% confidence interval) across all outcomes (Veroniki 2016).

We planned to conduct a random‐effects network meta‐analysis, assuming a common within‐network heterogeneity for each analysis, since the nature of the antibiotic regimens in the network is similar (Mills 2013; White 2015).

Appendix 4. Trial Sequential Analysis

Cumulative meta‐analyses are at risk of producing random errors due to sparse data and multiple testing of accumulating data (Brok 2008; Brok 2009; Higgins 2011; Pogue 1997; Thorlund 2009; Wetterslev 2009; Wetterslev 2017). Trial Sequential Analysis (TSA), CTU 2011, can be applied to control these random errors and to assess the risks of imprecision (Castellini 2018; Gartlehner 2019; Jakobsen 2014; Thorlund 2011). The required information size calculated by TSA takes into account the event proportion in the control group, the assumption of a plausible relative risk reduction (RRR), and the heterogeneity of the meta‐analysis (Turner 2013; Wetterslev 2009).

For dichotomous outcomes, we have not identified valid previous data on effect sizes, so we have chosen an RRR of 20% as anticipated intervention effect. We estimated the required information size based on the proportion of participants with an outcome in the control group and an RRR of 20%, an alpha of 2.5%, a beta of 20%, and a variance suggested by the trials in a random‐effects meta‐analysis (diversity‐adjusted required information size) (Jakobsen 2014; Wetterslev 2009). In case there is some evidence of effect of the intervention, a supplementary TSA used the limit of the confidence interval closest to 1.00 as the anticipated intervention effect (Jakobsen 2014). We additionally calculated the TSA‐adjusted confidence interval.

For continuous outcomes, we have not identified valid previous data on effect sizes on quality of life, so we have chosen to use standard deviation (SD)/2 as anticipated intervention effect. Hence, we estimated the required information size based on the SD observed in the control group of trials with low risk of bias or lower risk of bias and a minimal relevant difference of the observed SD/2, an alpha of 2.5%, a beta of 20%, and a diversity suggested by the trials in the meta‐analysis (Jakobsen 2014; Wetterslev 2009). In case there is some evidence of effect of the intervention, a supplementary TSA used the limit of the confidence interval closest to 0.00 as the anticipated intervention effect (Jakobsen 2014). We additionally calculated the TSA‐adjusted confidence interval.

Data and analyses

Comparison 1. Cefepime compared with ceftazidime.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 All‐cause mortality	1	30	Risk Ratio (M‐H, Fixed, 95% CI)	0.14 [0.01, 2.55]
1.2 Serious adverse events	1	30	Risk Ratio (M‐H, Fixed, 95% CI)	0.14 [0.01, 2.55]
1.3 Treatment failure	1	30	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.15, 1.64]

Comparison 2. Linezolid compared with vancomycin.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Treatment failure	1	32	Risk Ratio (M‐H, Fixed, 95% CI)	2.05 [0.49, 8.63]

Comparison 3. Meropenem compared with cefotaxime.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Treatment failure	1	6	Risk Ratio (M‐H, Fixed, 95% CI)	1.80 [0.10, 31.52]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Bosheva 2021.

Study characteristics
Methods	Design: randomised, multicentre trial Duration: 27 November 2017 to 16 March 2020 Country: 12 sites in Europe (Bulgaria, Georgia, Hungary, and Romania) Setting: hospital (no detail as to which unit or department) Date of publication: 21 January 2021
Participants	Randomised 138 participants with pneumonia (8 children with nosocomial pneumonia) to either ceftobiprole or standard of care (cephalosporin). Mean age (range): not reported Sex (M/F): NA Type of HAP: HAP Inclusion criteria: individuals 3 months to 18 years of age with a body weight of ≥ 5 kg and a diagnosis of HAP (pneumonia acquired after ≥ 48 hours of hospitalisation). Exclusion criteria: use of systemic antibacterial treatment for > 24 hours in the 48 hours before randomisation for the current episode of pneumonia (except patients with CAP who failed to improve after at least 48 hours of prior antibiotic therapy and required a change in treatment), mechanical ventilation for > 48 hours, viral pneumonia without bacterial superinfection, and known resistance to study antibiotic treatments.
Interventions	Intervention group: ceftobiprole Control group: cephalosporin Co‐interventions: vancomycin (10 to 15 mg/kg as an IV infusion every 6 hours) was also administered when MRSA was confirmed or suspected. Concomitant treatment with amikacin, gentamicin, or tobramycin for confirmed or suspected infection caused by Pseudomonas aeruginosa could be added at the discretion of the blinded investigator. Length of intervention: 7 to 14 days.
Outcomes	Primary outcomes: the cumulative incidence of adverse events during the first 3 days of study treatment and at the end of treatment, test‐of‐cure, and last follow‐up visits. Secondary outcomes: comparison of early clinical response at day 4 and clinical cure rates at the end of treatment. Clinical and microbiologic relapse rates at the last follow‐up visit were also compared (all efficacy populations). Microbiologic eradication rates at the test‐of‐cure visit, duration of IV antibiotic treatment, time to oral antibiotic switch, and duration of hospitalisation. Follow‐up: up to 35 days
Notes	Missing data: mean age and gender distribution for HAP participants exclusively. Email: kamal.hamed@basilea.com ClinicalTrials.gov ID: NCT03439124
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was carried out using a central interactive web‐based response system based on a computer‐generated randomisation schedule.
Allocation concealment (selection bias)	Unclear risk	Allocation concealment was not described.
Blinding of participants and personnel (performance bias) All outcomes	High risk	All other study site staff, including the principal investigator, pharmacists, and nursing staff, were unblinded. The participant and their parent/guardian were also unblinded and were reminded at each interaction with the blinded investigator not to disclose the treatment assignment.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The blinded investigator was also responsible for determining the duration of IV treatment, the decision to discontinue IV treatment, and the time point to switch to an oral antibiotic. To maintain blinding, the blinded investigator did not observe the participant at times when the study antibiotics were being administered.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Dropouts were unclear.
Selective reporting (reporting bias)	Unclear risk	They did not have a protocol and did not report outcome data on HAP participants separately.
Other bias	High risk	The study was supported by Basilea Pharmaceutica International Ltd., Basel, Switzerland. Basilea Pharmaceutica International Ltd. designed the study and aided in the acquisition of data, statistical analysis, and article preparation. Under the direction of the authors, medical writing support for the article was provided by Stephanie Carter of Arc, a division of Spirit Medical Communications Group Limited, funded by Basilea Pharmaceutica International Ltd.

Jantausch 2003.

Study characteristics
Methods	Design: randomised, open‐label, multicentre trial Duration: February to December 2001 Country: the USA and Latin America Setting: hospital (no detail regarding which unit or department) Date of publication: 9 April 2003
Participants	Randomised 40 children with nosocomial pneumonia to either linezolid or vancomycin. Mean age (range): NA Sex (M/F): NA Type of HAP: antibiotic‐resistant gram‐positive HAP Inclusion criteria: nosocomial pneumonia as defined by local institutions or PRSP (penicillin MIC 2 g/mL). Chest radiograph at baseline consistent with a diagnosis of pneumonia. At least 2 of the following: cough, new/worsened purulent sputum production, rales, pulmonary consolidation, or signs of respiratory distress (e.g. dyspnoea, tachypnoea, cyanosis, intercostal retractions, laboured breathing, grunting, or nasal flaring). Exclusion criteria: previous treatment for > 24 h with a potentially effective antibiotic within 48 h of study enrolment (unless the treatment failed or the pathogen showed resistance to the antibiotic). Furthermore, patients with pulmonary conditions such as cystic fibrosis or general underlying conditions such ischaemic ulcers, necrotising fasciitis, gas gangrene, etc., were excluded.
Interventions	Intervention group: linezolid Control group: vancomycin Co‐interventions: not reported Length of intervention: 10 to 21 days
Outcomes	Primary outcomes: clinical status at the test‐of‐cure follow‐up visit (cured, failure, indeterminate or missing). Secondary outcomes: pathogen eradication rates and changes in clinical signs and symptoms of infection. Follow‐up: up to 35 days
Notes	Missing data: mean age and gender distribution for HAP participants exclusively. The review authors defined treatment failure as participants who did not experience clinical cure. Funded by Pharmacia Corp.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Described as being randomised, but did not specify the sequence generation.
Allocation concealment (selection bias)	Unclear risk	Described as being randomised, but did not specify the allocation concealment.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Unblinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	Unblinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	4 participants in the linezolid group were excluded in the ITT analysis.
Selective reporting (reporting bias)	Unclear risk	They did not have a protocol and did not report all‐cause mortality and serious adverse events for participants with HAP.
Other bias	Low risk	No other biases were identified.

Schuler 1995.

Study characteristics
Methods	Design: randomised, multicentre trial Duration: 7.1 days (mean) in each group Country: 22 sites in Europe (Belgium, the Czech Republic, France, and Hungary) and South Africa Setting: hospital (no detail regarding which unit or department) Date of publication: 1995
Participants	Randomised 119 hospitalised children with bacterial infections (6 children with nosocomial pneumonia) to either meropenem or cefotaxime. Mean age (range): not reported Sex (M/F): NA Type of HAP: HAP Inclusion criteria: hospitalised children aged 3 months to 12 years, with clinical signs and symptoms of a bacterial infection requiring a parenteral antibiotic. Exclusion criteria: body weight < 5 to 6 kg in order to exclude neonates; hypersensitivity to any beta‐lactam antibiotic; administration of another antibiotic within the 3 days before enrolment (unless the pathogen was resistant or persisted) or another investigational drug within the 30 days before enrolment; clinically manifest hepatic failure or hepatic coma; renal function impairment (creatinine clearance rate ≤ 51 mL/min); history of seizures, meningitis, or cystic fibrosis; neutrophil count < 1 x 109/L; and the presence of severe underlying disease likely to prevent completion of at least 48 hours of study drug therapy.
Interventions	Intervention group: meropenem (10 or 20 mg/kg up to a maximum of 1 g) Control group: cefotaxime 100 to 150 mg/kg/day divided into 2 to 4 equal doses. Metronidazole (7.5 mg/kg every 8 hours) was added to the cefotaxime regimen in cases of suspected mixed aerobic/anaerobic infection. Amikacin (15 mg/kg/day in 2 to 3 equal doses) was added to the cefotaxime regimen for the treatment of UTI in France, according to local guidelines, but only for the control group. Co‐interventions: not reported Length of intervention: maximum 28 days Follow‐up: up to 6 weeks
Outcomes	Primary outcomes: safety Secondary outcomes: clinical and bacteriological efficacy
Notes	Missing data: mean age and gender distribution for HAP participants exclusively. Funded by Zeneca Pharmaceuticals.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated code used.
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Unblinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	Unblinded
Incomplete outcome data (attrition bias) All outcomes	High risk	High number of dropouts
Selective reporting (reporting bias)	Unclear risk	They did not have a protocol and only reported on clinical cure/treatment failure for HAP participants.
Other bias	Low risk	No other biases were identified.

Shahid 2008.

Study characteristics
Methods	Design: randomised, single‐centre clinical trial Duration: from April 2004 to August 2005 Country: Malaysia Setting: hospital (no detail regarding which unit or department) Date of publication: 2 April 2007
Participants	Randomised 30 children with late‐onset VAP to either cefepime or ceftazidime Mean age (SE): 1.56 year (0.7) Sex (M/F): 11/19 Type of HAP: late‐onset VAP Inclusion criteria: age < 1 year and late‐onset VAP. New or progressive pulmonary infiltrates and at least 2 of the following: a body temperature of > 38 °C or < 36 °C; > 10,000 or < 4000 leucocytes/uL blood; purulent tracheo‐bronchial secretions; and/or a decrease in the partial pressure of oxygen. Exclusion criteria: disseminated intravascular coagulation, organ failure, immunosuppression, or known hypersensitivity to cephalosporins and preterm newborns.
Interventions	Intervention group: cefepime Control group: ceftazidime Co‐interventions: not reported
Outcomes	Primary outcomes: clinical response (cure, improvement, failure, or death) Secondary outcomes: microbiological response (eradication, persistence, superinfection, or unable to determine) Follow‐up: not reported
Notes	The mean age of VAP participants seems to conflict with the age limit mentioned in the inclusion criteria. Email: sukhbir5@lycos.com
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random table used.
Allocation concealment (selection bias)	Unclear risk	Did not describe allocation concealment.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Did not describe blinding.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Did not describe blinding.
Incomplete outcome data (attrition bias) All outcomes	High risk	8/40 participants had missing data for the only outcome (treatment failure) we were able to extract.
Selective reporting (reporting bias)	Low risk	They did not have a protocol. However, all‐cause mortality and serious adverse events were reported.
Other bias	Low risk	No other biases were observed.

AEs: adverse events
CAP: community‐acquired pneumonia
EOT: end of treatment
HAP: hospital‐acquired pneumonia
ITT: intention‐to‐treat
IV: intravenous
MIC: minimal inhibitory concentration
MRSA: methicillin‐resistant Staphylococcus aureus
NA: not available
PRSP: penicillin‐resistant Streptococcus pneumoniae
SE: standard error
UTI: upper tract infection
VAP: ventilator‐associated pneumonia

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Agweyu 2015	Unclear if children had HAP. Authors did not respond to our request for this information.
Amonova 2011	Only included adult participants
Awad 2014	Only included adult participants
Barradas 1989	Only included adult participants
Bassetti 1991	Only included adult participants
Begue 1998	Did not include participants with HAP
Berman 2004	Not randomised
Bhavnani 2012	Only included adult participants
Chaudhary 2008	Only included adult participants
Chuchalin 1997	Only included adult participants
Cometta 1994	Only included adult participants
Dickstein 2016	Only included adult participants
Fatehi 2019	Same antibiotic (colistin) in both intervention groups
Giamarellos‐Bourboulis 2014	Only included adult participants
Grudinina 2002	Only included adult participants
Iakovlev 2000	Only included adult participants
Iakovlev 2006	Only included adult participants
Joshi 1999	Only included adult participants
Kollef 2012	Only included adult participants
Lacy 2015	Only included adult participants
Mohamed 2018	Only included adult participants
Muscedere 2012	Only included adult participants
Nassar 2018	Only included adult participants
Norrby 1993	Only included adult participants
Patel 2015	Same antibiotic (amoxicillin) in both intervention groups
Rodloff 1996	Only included adult participants
Straneo 1990	Only included adult participants
Tucker 2017	Only included adult participants

HAP: hospital‐acquired pneumonia

Characteristics of ongoing studies [ordered by study ID]

Shahrin 2020.

Study name	Injectable amoxicillin versus injectable ampicillin plus gentamicin in the treatment of severe pneumonia in children aged 2 to 59 months: protocol for an open‐label randomised controlled trial
Methods	This randomised, controlled, open‐label, non‐inferiority trial is being conducted in Dhaka Hospital of the International Centre for Diarrheal Disease Research, Bangladesh. A sample size of 308 children with severe pneumonia will give adequate power to this study. Children aged 2 to 59 months are randomised to either IV ampicillin or IV amoxicillin, plus IV gentamicin in both study arms. The monitoring of patients is carried out according to the WHO protocol for the treatment of severe pneumonia. The primary objective is the rate of treatment failure, defined as the persistence of danger signs of severe pneumonia beyond 48 hours or deterioration within 24 hours of initiation of therapy. The secondary objectives are: improvement in or the resolution of danger signs since enrolment; length of hospital stay; death during hospitalisation; rate of nosocomial infections.
Participants	Children aged 2 to 59 months are eligible for study enrolment upon meeting clinical criteria of severe pneumonia, as defined by the WHO classification updated in 2014.
Interventions	In the ampicillin arm, the participant receives a 50 mg/kg dose of IV ampicillin every 6 hours and a 7.5 mg/kg dose of IV gentamicin once daily for 5 to 7 days. In the intervention arm (amoxicillin arm), the participant receives a 40 mg/kg dose of IV amoxicillin every 12 hours and a 7.5 mg/kg dose of IV gentamicin once daily for 5 to 7 days.
Outcomes	Primary outcome variable: the percentage of children with treatment failure, as determined either by the persistence of danger signs over 48 hours or by the appearance of new danger signs within 24 hours of the study intervention. Secondary outcome variables: time to resolution of danger signs of severe pneumonia; length of hospital stay; rate of nosocomial infections; death during or after discharge. The secondary outcome measurement variables are the time (in hours) of disappearance of danger signs, time (in days) to discharge from the acute phase, and rate of suspected nosocomial infections (a nosocomial infection will be diagnosed based on the appearance of new signs of infection, such as fever, cough, or respiratory distress, diarrhoea, or crying during urination, after 48 hours of admission or within 72 hours of discharge from the hospital).
Starting date	1 January 2018
Contact information	Email: lubabashahrin@icddrb.org
Notes	ClinicalTrials.gov ID: NCT03369093

IV: intravenous
WHO: World Health Organization

Differences between protocol and review

There were no differences between the protocol, Korang 2021a, and the review.

Contributions of authors

Steven Kwasi Korang (SKK), Chiara Nava (CN), Sutharshini Punniyamoorthy Mohana (SPM), Ulrikka Nygaard (UN), Janus C Jakobsen (JCJ)
Conceiving the protocol/review: SKK and CN
Co‐ordinating the protocol/review: SKK
Writing the review: SKK, CN, and SPM
Designing the protocol: SKK, CN, and JCJ
Guarantor for the review (one author): SKK
Revising the review: SKK, CN, SPM, UN, and JCJ
Responsible for reading and checking the review before submission: SKK, CN, SPM, UN, and JCJ

All authors agreed on the final version of the review.

Sources of support

Internal sources

• The Copenhagen Trial Unit, Denmark

  The review was conducted during work hours.

External sources

• The Danish State, Denmark

  Funds the Copenhagen Trial Unit and thus funded this review indirectly

Declarations of interest

The performance of this review is free of any real or perceived bias introduced by receipt of any benefit in cash or kind, on any subsidy derived from any source that may have or be perceived to have an interest in the outcomes of the review.

Steven Kwasi Korang: has declared that they have no conflict of interest.
Chiara Nava: has declared that they have no conflict of interest.
Sutharshini Punniyamoorthy Mohana: has declared that they have no conflict of interest.
Ulrikka Nygaard: has declared that they have no conflict of interest.
Janus C Jakobsen: has declared that they have no conflict of interest.

New