Methods
Study Type and Registration
We conducted a systematic review of randomized controlled trials (RCTs) in accordance with a previously registered protocol (PROSPERO, registration number CRD42012003162). The presented review was performed according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement.
Identification of Relevant Studies
Four databases (EMBASE, MEDLINE, The Cochrane Central Register of Controlled Trials (CENTRAL) and LILACS) were systematically searched for relevant trials published from inception until 15 February 2013, without language restriction. Personnel files and reference lists of relevant review articles for additional trials were also reviewed. Detailed search strings are listed in Additional file 1.
Eligibility Criteria
Inclusion criteria with respect to the patient, population or problem, intervention, comparison, outcomes, and setting (PICOS) criteria were as follows: 1) population: clinical diagnosis of acute lung injury (ALI) or ARDS as defined by the American-European consensus conference in 1994, or the Berlin definition 2012, or a ratio of arterial partial pressure of oxygen/fraction of inspired oxygen (PaO2/FiO2) lower than or equal to 300 during invasive mechanical ventilation, or some indication that the majority of patients would meet this criterion. Patients also had to be older than 16 years; 2) intervention: patients submitted to or requiring fluid therapy (main intervention or co-intervention); 3) comparison: colloids, independently of molecular weight compared to crystalloids must represent one of the fluid therapies; 4) outcome: primary outcome parameters were respiratory mechanics (compliance, plateau pressure) or gas exchange (arterial carbon dioxide partial pressure (PaCO2), PaO2/FiO2) or parameters of lung inflammation (bronchoalveolar lavage fluid neutrophils or IL-8 levels) and damage, hospital mortality; secondary outcome parameters: kidney function (creatinine, neutrophil gelatinase-associated lipocalin (NGAL) or need for renal replacement therapy (intermittent or continuous hemodialysis, hemofiltration or hemodiafiltration), hemodynamic stabilization (time and amount of fluid), ICU length of stay; and 5) design: RCT. Trials were excluded if they were uncontrolled, pseudo-randomized, published only as an abstract, or if all intervention groups received colloid therapy.
Trial Selection and Data Abstraction
The articles for this review were selected by examining titles, abstracts, and the full text if a potentially relevant trial was identified. We translated non-English reports into English, as necessary. Two reviewers (CU, PLS), independently and in duplicate, abstracted the data on the a priori-defined inclusion criteria (population, intervention, comparison, clinical outcomes and design). Trial data presented in figures only were extracted using Engauge Digitizer (Version 5.1., http://digitizer.sourceforge.net).
Risk of Bias Assessment and Strength of Evidence
In duplicate and independently, two reviewers assessed trial methodological quality using the risk-of-bias tool recommended by the Cochrane Collaboration. For each trial, the risk of bias was reported as low risk, unclear risk, or high risk in the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. For each outcome, we independently rated the overall quality of evidence (confidence in effect estimates) in duplicate using the GRADE approach in which trials begin as high-quality evidence, but may be rated down by one or more of five categories of limitations: risk of bias, inconsistency, indirectness, imprecision, and reporting bias. Finally, the overall risk of bias for an individual trial was categorized as low (if the risk of bias was low in all domains), unclear (if the risk of bias was unclear in at least one domain, with no high risk of bias domains), or high (if the risk of bias was high in one or more domains). Disagreement was resolved by discussion and consensus. Attempts were made to contact the authors and request for any necessary information not contained in the publications.
Data Synthesis
All analyses were performed using STATA (Version MP 11, Stata Corp LP, Lake Drive, TX, USA). To calculate the pooled risk ratio (RR) and 95% CIs of binary outcomes (mortality) trial data were combined using the Mantel-Haenszel estimator with estimation of variances according to the DerSimonian and Laird random-effect model. For continuous outcomes, the pooled weighted mean difference (WMD) in the gas exchange (PaO2/FiO2) was calculated weighting the effect in respect to sample size. Statistical heterogeneity was assessed by the I statistic. Substantial heterogeneity was predefined as I >50%.