Search strategy and selection criteria
An extensive search of PubMed (up to September 2013), Cochrane Central Register, and Scopus of RCTs was performed to identify relevant RCTs for our meta-analysis. Search terms included “linezolid,” “child,” “randomized clinical trials,” “efficacy,” and “safety.” No language restrictions were used. Specific search terms for each database were the following:
PubMed search terms (linezolid [substance] OR “zyvox” [all fields]) AND (child [all fields] OR neonate [all fields] OR infant [all fields]) AND (randomized controlled trial [pt] OR controlled clinical trial [pt] OR randomized controlled trials [mh] OR random allocation [mh] OR double-blind method [mh] OR single-blind method [mh] OR clinical trial [pt] OR clinical trials [mh] OR (“clinical trial” [tw]) OR ((singl* [tw] OR doubl* [tw] OR trebl* [tw] OR tripl* [tw]) AND (mask* [tw] OR blind* [tw])) OR (“Latin square” [tw]) OR placebos [mh] OR placebos [mh] OR placebo* [tw] OR random* [tw] OR research design [mh:noexp] OR follow-up studies [mh] OR prospective studies [mh] OR cross-over studies [mh] OR prospectiv* [tw] OR volunteer* [tw]) NOT (animal [mh] NOT human [mh])
CENTRAL (Cochrane) search terms (linezolid AND child)
SCOPUS search terms (linezolid AND (child OR neonate OR infant) AND clinical trial).
Randomized controlled trials were selected using a methodological filter . To identify relevant completed studies that were unpublished, we searched clinical trial registries (ISRCTN Register, Netherland’s Trial Register, UMIN Clinical Trials Registry, Australian New Zealand Clinical Trials Registry, and ClinicalTrial.gov) up to September 2013. We obtained the results of these studies by contacting the manufacturer through one of their expert opinion leaders.
Two reviewers (MI and FA) independently searched the literature and examined relevant RCTs for further assessment of data on effectiveness and safety. Eligible studies were randomized controlled trials that assessed the clinical efficacy and safety of linezolid in children in comparison with other antimicrobial agents for treatment of infections caused by Gram-positive bacteria. However, no eligible studies evaluating only efficacy or safety were retrieved from the databases. Hospital admission of patients was not required for eligibility. We did not place any restrictions on participants, length of follow-up, or the class of antimicrobial agents in the control group. Experimental trials and trials focusing on pharmacokinetic or pharmacodynamic variables, trials referring only to the in vitro activity of linezolid, and unpublished studies that were incomplete were excluded.
The following data were extracted from every study: year of publication, clinical setting, patient population, number of patients [by intention to treat (ITT), and those assessed clinically and microbiologically], characteristics of patients (age, sex, ethnic origin, and disease), antimicrobial agents and doses used, clinical and microbiological outcomes, adverse effects, treatment duration, time from treatment to test of cure, and funding. The ITT population comprised patients who received at least one dose of the medications studied in the individual RCTs. The clinically assessed population comprised patients that fulfilled all inclusion and exclusion criteria in the individual RCTs, had complete follow-up with a final assessment at the preset test of cure visit, and received one course of study drug for the minimum duration, with no more than one dose of potentially effective concomitant therapy after the first dose of linezolid. The microbiologically assessed population was a subset of the clinically assessable population, and had also microbiologically documented infections.
Two reviewers (MI and FA) independently extracted the relevant data. Any disagreement was resolved by consensus in meetings with all investigators.
Definitions of infections
Infections were defined according to the definitions used in the individual randomized controlled trials. Definition of infection did not differ substantially between the studies included in the meta-analysis.
For skin and soft tissue infections (SSTIs), both complicated and uncomplicated SSTIs could be included in the meta-analysis. At least one of the following symptoms of infection needed to have been present: drainage or discharge, erythema, fluctuance, heat or localized warmth, pain or tenderness to palpation, swelling, or induration.
For catheter-related bacteremia, presence of indwelling venous or arterial catheter, at least one positive blood culture obtained from the catheter growing a Gram-positive organism and clinical profile was to be compatible with a diagnosis of catheter-related bacteremia with at least two of the following findings: fever, hypothermia, leukocytosis, leukopenia or left shift of band neutrophils, increased pulse (>98th percentile for age), and increased respiratory rate (>2SD of normal for age). In addition to the inclusion signs and symptoms, other signs of septic shock (decreased peripheral perfusion or hypotension) and petechiae or purpura could also be used to meet inclusion for patients with catheter-related bacteremia.
With regard to the definition of bacteremia of unknown origin, a clinical profile compatible with a diagnosis of bacteremia without an identified source was considered as equivalent to a diagnosis of bacteremia of unknown origin. In addition to the inclusion signs and symptoms, other signs of septic shock (decreased peripheral perfusion or hypotension) and petechiae or purpura could also be used to meet inclusion for patients with bacteremia of unknown origin.
For a diagnosis of pneumonia, baseline chest radiograph needed to show new or progressive infiltrates, consolidation with or without effusion, and two of the following signs and symptoms: cough, new/worsened purulent sputum production, rales, pulmonary consolidation, or signs of respiratory distress (e.g., dyspnea, tachypnea, cyanosis, intercostal retractions, labored breathing, grunting, or nasal flaring). Patients with pneumonia could also have, as part of these two inclusion criteria, fever, hypothermia, leukocytosis, leukopenia or left shift of band neutrophils, increased pulse (>98th percentile for age), increased respiratory rate (>2SD of normal for age), requirement for mechanical ventilation, an increase in ventilator settings, altered mental status, lethargy, or irritability in infants <1 year of age.
We assessed the quality of every published randomized controlled trial on the basis of five elements of study design and reporting the following: generation of random numbers, details of double-blinding procedure, information on withdrawals, allocation concealment, and discussion of other potential sources of bias . One point was awarded for the specification of each criterion, with a maximum score of 5. High-quality RCTs scored 3 or more points, whereas low-quality RCTs scored 2 or fewer points, according to a modified Jadad score .
The primary outcome was (1) treatment success in the clinically assessable population, which was defined as cure (resolution of symptoms and signs of infection) or clinically significant improvement of patients and (2) treatment success in the microbiologically assessable population. Secondary outcomes included adverse events or discontinuation of treatment that was probably related to effects of the study drug and eradication of pathogens present at baseline. Treatment success and adverse events were assessed in all patients.
Data analysis and statistical methods
The meta-analysis was performed with random effects models in Review Manager (version 5.1). We chose to use random effect models because of the obvious heterogeneity across the trials included in the meta-analysis (different infections, control drugs). Pooled odds ratios (ORs) and 95 % confidence intervals (CIs) for all primary and secondary outcomes (including ITT, clinically assessed and microbiologically assessed populations) were calculated using the Mantel-Haenszel method.
The statistical heterogeneity between RCTs was assessed by Q statistic generated from the x
2 test (p < 0.10 was defined to indicate significant heterogeneity). The extent of heterogeneity was estimated with the I
2 measure, and published guidelines were used to define low (I
2 = 25–49 %), moderate (I
2 = 50–74 %), and high (I
2 > 75 %) heterogeneity . For each analysis, the I
2 and p values of statistical heterogeneity are presented.
Assessment of publication bias was not undertaken because of the small number of included studies.