Criteria for considering studies for this review
We considered randomized controlled trials with individuals of any age and clinically diagnosed tuberculous meningitis, comparing linezolid to standard treatment.
We grouped outcomes as follows:
1. Death: number of deaths at the study’s endpoint, at one month, and at three months after the beginning of therapy;
2. Neurological disability: score measured in the modified Rankin Scale (mRS) at the study’s endpoint, at one month, and at three months after the beginning of therapy;
3. Adverse effects: adverse events reported by authors and according to the Common Toxicity Criteria for Adverse Events (CTCAE).
Search methods for identification of studies
We searched both published and unpublished data from trials. Up to 5 January 2024, we searched Pubmed, EMBASE, and the Cochrane Library using the terms (tuberculosis OR 'Koch`s disease' OR 'M. tuberculosis infection' OR 'Mycobacterium tuberculosis infection' OR 'TB (tuberculosis)' OR 'TB case' OR 'TB cases' OR 'TB disease' OR 'TB infection' OR 'active TB' OR 'active tuberculosis' OR 'case of TB' OR 'cases of TB' OR 'chronic tuberculosis' OR 'infection by M. tuberculosis' OR 'infection by Mycobacterium tuberculosis' OR 'infection due to M. tuberculosis' OR 'infection due to Mycobacterium tuberculosis' OR 'infection of M. tuberculosis' OR 'infection of Mycobacterium tuberculosis' OR 'minimal tuberculosis' OR 'minimum tuberculosis' OR 'tuberculosis' OR 'tuberculous infection' OR 'tuberculous lesion') AND linezolid AND (randomized controlled trial OR 'controlled trial, randomized' OR 'randomised controlled study' OR 'randomised controlled trial' OR 'randomized controlled study' OR 'randomized controlled trial' OR 'trial, randomized controlled'). There were no publication date or language restrictions.
Authors of ongoing trials were contacted by email asking whether they could provide preliminary findings or information about the study’s stage.
Data collection and analysis
Two authors (GL and GC) independently screened the title and abstract of the identified records. If one of the authors considered the record eligible, both assessed the full text. Disagreements were solved by discussion and consulting a third reviewer (CB).
We tried to contact the authors of ongoing trials. An email was sent asking for further details about the stage of the trial and whether there was any additional relevant data they would like to share. If there was no answer, we sent a second email within two weeks of the first one.
The same authors were responsible for selecting studies extracted information about study characteristics to a prestructured file. Data was extracted from five domains: study characteristics, methods, participants, interventions, and outcomes. They included authorship, funding, conflicts, study design, country, methodological qualities, unit of allocation, number of study arms, period of the study, methods used for diagnosis, population description, HIV status, tuberculous meningitis severity, inclusion, and exclusion criteria, number of participants, interventions, dosage and duration of intervention, duration of follow up, concomitant therapies, primary and secondary outcomes as reported by the authors, potential confounders and sources of bias and the results of interest to this review (number of deaths, measurement of neurological disability, occurrence of adverse events and grading of the severity of the adverse event according to CTCAE).
The number of randomized patients in the studied arms and the number that experienced the event of interest was recorded for dichotomous outcomes.
Risk of bias assessment in included studies
Two independent reviewers (GL and GC) used the Cochrane risk-of-bias tool for randomized trials (RoB 2)[14], version of August 2019, to assess the risk of bias/quality of the included studies.
Synthesis methods
We included all trials that analyzed the addition of linezolid to standard antituberculous treatment (rifampicin, isoniazid, pyrazinamide, and ethambutol or streptomycin) compared with antituberculous treatment alone for the management of patients with tuberculous meningitis.
We used an inverse-variance random effects model, reporting dichotomous data as Relative Risk (RR) with 95% Confidence or Credible intervals (CI). Statistical heterogeneity was assessed by calculating tau (between study variance), I² (percentage of total variation across studies that is due to heterogeneity), and Cochran's Q (test assessing the null hypothesis that all studies in the analysis share a common effect size) statistics; I² higher than 50% was considered as a significant indicator of heterogeneity or a p-value lower than 0.10.
Analyses of the trials’ results with a Bayesian approach were designed under a weakly informative neutral prior (μ = 0, σ = 4). Prior selection and strength of belief are presented in the supplementary material. We used funnel plots to assess publication bias. A sensitivity analysis was performed, accounting for the follow-up losses with worst-case scenario analysis.
Analyses were conducted in RevMan Web, version 7.2.0, and R software, version 4.3.3, with the meta, metafor, and bayesmeta packages[15, 16].
Investigation of heterogeneity and subgroup analysis
Clinical diversity and methodological diversity were described in the study’s PICO table, accounting for variability in patients (tuberculous meningitis severity and HIV status), interventions (duration of therapy and drug dosing), and risk of bias (for which we conducted a sensitivity analysis comparing studies at a high vs low/some concerns risk of bias). A subgroup analysis including trials exclusively with HIV-positive patients was conducted. There was insufficient data to allow statistical analysis of outcomes according to age group (< or > 18 years), linezolid dose, and duration of therapy, as planned per protocol.
Sensitivity analysis
Losses to follow-up were investigated with a worst-case scenario analysis. We considered the worst outcome for every follow-up loss in the intervention arm (death or bad mRS score) and the best outcome for losses in the control arm.
Certainty of the evidence assessment
The overall quality of the body of evidence was assessed by one reviewer (GL) using the GRADE approach, formulating a summary of findings and certainty of evidence tables containing the critical outcomes investigated. The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) handbook was followed to determine the evidence profile of the analyses conducted[17]. In this evaluation, evidence can be divided into four grades based on how certain the outcomes are: high (we are very confident that the true effect lies close to that of the estimate of the effect), moderate (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 (the true effect may be substantially different from the estimate of the effect) and very low (the true effect is likely to be substantially different from the estimate of effect). The tables were generated in the GRADEpro GDT software[18].