Seventeen RCTs were included (n=938).
Twelve studies had adequate allocation concealment, all had clear inclusion criteria, fifteen used strategies to standardise cointerventions and all had complete follow-up details for all participants. However, eleven failed to describe blinding strategies and only one clearly blinded physicians, nurses and patients to the intervention. Four did not report an intention-to-treat analysis and six had inadequate or unclear outcome definitions.
CPAP versus SMT.
Mortality (10 RCTs, n=490): there was a statistically significant reduction in risk in the CPAP group (RD -13%, 95% CI: -22, -5, p=0.003).
Need for ETI (10 RCTs, n=490): there was a statistically significant reduction in risk in the CPAP group (RD -22%, 95% CI: -34, -10, p=0.0004). There was statistically significant heterogeneity associated with this finding (p=0.0004; I2=70.1%). However, the direction of effect was consistent for all but one study.
Risk of AMI (3 RCTS, n=112): there was no statistically significant difference between the groups for this outcome.
NPPV versus SMT.
Mortality (6 RCTS, n=315): there was no statistically significant difference between the groups for this outcome, though there was a trend for benefit in the NPPV group that was of borderline statistical significance (RD 7%, 95% CI: -14, 0, p=0.06).
Need for ETI (6 RCTs, n=315): there was a statistically significant reduction in risk in the CPAP group (RD -18%, 95% CI: -32, -4, p=0.010). There was statistically significant heterogeneity associated with this finding (p=0.02; I2=62.9%). However, the direction of effect was consistent for all but one study.
Risk of AMI (5 RCTs, n=278): there was no statistically significant difference between the groups for this outcome.
CPAP versus NPPV.
There was no statistically significant difference between the groups for the outcomes of mortality (7RCTs, n=299), need for ETI (7 RCTs, n=299) or risk of AMI (6 RCTs, n=263).
Subgroup analyses did not show increased benefit from NPPV compared with CPAP, either when using higher levels of pressure support ventilation, or for hypercapnic patients.
For the comparison of CPAP versus NPPV, the funnel plot suggested a lack of small studies favouring CPAP. There was less indication of publication bias for other comparisons and in all cases Begg's test was non significant.