Carbonic anhydrase inhibitors in patients with respiratory failure and metabolic alkalosis: a systematic review and meta-analysis of randomized controlled trials

Background Metabolic alkalosis is common in patients with respiratory failure and may delay weaning in mechanically ventilated patients. Carbonic anhydrase inhibitors block renal bicarbonate reabsorption, and thus reverse metabolic alkalosis. The objective of this systematic review is to assess the benefits and harms of carbonic anhydrase inhibitor therapy in patients with respiratory failure and metabolic alkalosis. Methods We searched the following electronic sources from inception to August 2017: the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, and SCOPUS. Randomized clinical trials were included if they assessed at least one of the following outcomes: mortality, duration of hospital stay, duration of mechanical ventilation, adverse events, and blood gas parameters. Teams of two review authors worked in an independent and duplicate manner to select eligible trials, extract data, and assess risk of bias of the included trials. We used meta-analysis to synthesize statistical data and then assessed the certainty of evidence using the GRADE methodology. Results Six eligible studies were identified with a total of 564 participants. The synthesized data did not exclude a reduction or an increase in mortality (risk ratio (RR) 0.94, 95% confidence interval (CI) 0.57 to 1.56) or in duration of hospital stay (mean difference (MD) 0.42 days, 95% CI −4.82 to 5.66) with the use of carbonic anhydrase inhibitors. Carbonic anhydrase inhibitor therapy resulted in a decrease in the duration of mechanical ventilation of 27 h (95% CI −50 to −4). Also, it resulted in an increase in PaO2 (MD 11.37 mmHg, 95% CI 4.18 to 18.56) and a decrease in PaCO2 (MD −4.98 mmHg, 95% CI −9.66, −0.3), serum bicarbonate (MD −5.03 meq/L, 95% CI −6.52 to −3.54), and pH (MD −0.04, 95% CI −0.07 to −0.01). There was an increased risk of adverse events in the carbonic anhydrase inhibitor group (RR 1.71, 95% CI 0.98 to 2.99). Certainty of evidence was judged to be low for most outcomes. Conclusion In patients with respiratory failure and metabolic alkalosis, carbonic anhydrase inhibitor therapy may have favorable effects on blood gas parameters. In mechanically ventilated patients, carbonic anhydrase inhibitor therapy may decrease the duration of mechanical ventilation. A major limitation of this finding was that only two trials assessed this clinically important outcome. Electronic supplementary material The online version of this article (10.1186/s13054-018-2207-6) contains supplementary material, which is available to authorized users.

We searched the following electronic databases from inception to August 2017: the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, and SCOPUS. Tables 1 to 4 in additional file detail the electronic search strategy. There were no language or date restrictions.
We also screened the reference lists of included trials and identified related systematic reviews.

Selection process
Teams of two review authors (BT, CN and HI, MO) screened in duplicate and independently the abstract and title of every record captured by the searches for potential eligibility. We retrieved the full texts for all citations judged as potentially eligible by at least one of the reviewers.
The teams of two reviewers then assessed in duplicate and independently the full texts for eligibility using a standardized and pilot tested screening form. They then compared their results and resolved any disagreements by consensus and, when unsuccessful, with the help of a third reviewer (SM or PBK). Before starting the selection process, BT, HI, CN, and MO conducted calibration exercises to ensure the validity of the selection process.
A PRISMA (preferred reporting items for systematic reviews and metaanalyses) flow-chart was used to summarize the results of the selection process.(9)

Data extraction
For each included study, the teams of two reviewers independently and in duplicate abstracted relevant information using standardized data extraction forms. They then compared their results and resolved any disagreements by discussion and, when unsuccessful, with the help of a third reviewer.
We extracted information about the study design; the clinical characteristics of the trial (population, intervention, comparator, and outcomes); funding, and conflicts of interest of authors.

Assessment of risk of bias
The two teams of reviewers assessed the risk of bias of each included trial independently and in duplicate. Disagreements were resolved by consensus and, when unsuccessful, with the help of a third reviewer.

Risk of bias was assessed using The Cochrane Collaboration's Risk of
Bias tool.(10, 11) The following criteria were used: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants, providers, data collectors, outcome adjudicators, and data analysts (performance bias and detection bias), incomplete outcome data (attrition bias), selective outcome reporting (reporting bias), other bias (including early stopping for benefit).
We judged Risk of bias criteria as 'low risk', 'high risk' or 'unclear risk' and evaluated individual bias items as described in the Cochrane Handbook for Systematic Reviews of Interventions.(10)

Data analysis
For dichotomous data, we used risk ratio (RR) or hazard ratio (HR) with 95% confidence intervals (CI). For continuous outcomes data, we used, whenever possible, the mean change score from baseline to follow-up for each intervention group.
One of the of the included trials (Nelson 1965) did not report standard deviations (SD) in the assessment of the outcomes PaCO2 and serum bicarbonate.(12) Therefore, we used the median SD from the other included trials that reported SDs for these outcomes, as described in Furukawa 2006.(13) In another trial (Faisy 2016),(14) the authors did not report means and SDs, so these were extrapolated respectively from the reported medians (mean = median) and interquartile ranges (IQR) (SD= IQR/1.35).(10) In Hacki 1983, outcomes data were extracted from a graph in the report using the WebPlotDigitizer tool. (15,16) We pooled data using the random-effects model for the primary metaanalyses.(17) Heterogeneity (inconsistency) between study results was assessed by visual inspection of the forest plots and by using the I 2 statistic. We considered an I 2 value of 50% or more to indicate a considerable level of heterogeneity.(10) In order to explain any heterogeneity, we planned to conduct subgroup analyses based on the following variables: specific type and dose of CAI, etiology of respiratory failure, spontaneously breathing patients or on MV or NIPPV, and severity of metabolic alkalosis.
We also planned to perform sensitivity analyses to explore the influence of the following factors (when applicable) on pooled effect sizes: restricting the analyses to studies with low risk of bias, restricting the analyses to studies with longer follow-up, and assessing the impact of missing data. (18-21).

Assessment of certainty of the evidence
Certainty of the evidence for each outcome was assessed using the