We searched the Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 3), MEDLINE, the Cumulative Index to Nursing and Allied Health Literature (CINAHL), Embase, Web of Science, proceedings from four conferences, and clinical trials registries; and we handsearched reference lists of relevant studies. We conducted searches from January 2000 to March 2016 and reran the searches in December 2016. We added four new studies of potential interest to a list of 'Studies awaiting classification' and will incorporate them into formal review findings during the review update.
SELECTION CRITERIA: Two review authors independently assessed studies for inclusion, extracted data, and assessed risk of bias.
MAIN RESULTS: We included 11 studies with 1972 participants. Participants in six studies had
respiratory failure, and in five studies required
oxygen therapy after extubation. Ten studies compared HFNC versus low-flow
oxygen devices; one of these also compared HFNC versus CPAP, and another compared HFNC versus BiPAP alone. Most studies reported randomization and allocation concealment inadequately and provided inconsistent details of outcome assessor blinding. We did not combine data for CPAP and BiPAP comparisons with data for low-flow
oxygen devices; study data were insufficient for separate analysis of CPAP and BiPAP for most outcomes. For the primary outcomes of treatment failure (1066 participants; six studies) and mortality (755 participants; three studies), investigators found no differences between HFNC and low-flow
oxygen therapies (risk ratio (RR), Mantel-Haenszel (MH), random-effects 0.79, 95% confidence interval (CI) 0.49 to 1.27; and RR, MH, random-effects 0.63, 95% CI 0.38 to 1.06, respectively). We used the GRADE approach to downgrade the certainty of this evidence to low because of study risks of bias and different participant indications. Reported adverse events included
nosocomial pneumonia,
oxygen desaturation, visits to general practitioner for respiratory complications,
pneumothorax, acute pseudo-obstruction,
cardiac dysrhythmia,
septic shock, and cardiorespiratory arrest. However, single studies reported adverse events, and we could not combine these findings; one study reported fewer episodes of
oxygen desaturation with HFNC but no differences in all other reported adverse events. We downgraded the certainty of evidence for adverse events to low because of limited data. Researchers noted no differences in ICU
length of stay (mean difference (MD), inverse variance (IV), random-effects 0.15, 95% CI -0.03 to 0.34; four studies; 770 participants), and we downgraded quality to low because of study risks of bias and different participant indications. We found no differences in oxygenation variables: partial pressure of arterial
oxygen (PaO2)/fraction of inspired
oxygen (FiO2) (MD, IV, random-effects 7.31, 95% CI -23.69 to 41.31; four studies; 510 participants); PaO2 (MD, IV, random-effects 2.79, 95% CI -5.47 to 11.05; three studies; 355 participants); and oxygen saturation (SpO2) up to 24 hours (MD, IV, random-effects 0.72, 95% CI -0.73 to 2.17; four studies; 512 participants). Data from two studies showed that oxygen saturation measured after 24 hours was improved among those treated with HFNC (MD, IV, random-effects 1.28, 95% CI 0.02 to 2.55; 445 participants), but this difference was small and was not clinically significant. Along with concern about risks of bias and differences in participant indications, review authors noted a high level of unexplained statistical heterogeneity in oxygenation effect estimates, and we downgraded the quality of evidence to very low. Meta-analysis of three comparable studies showed no differences in
carbon dioxide clearance among those treated with HFNC (MD, IV, random-effects -0.75, 95% CI -2.04 to 0.55; three studies; 590 participants). Two studies reported no differences in
atelectasis; we did not combine these findings. Data from six studies (867 participants) comparing HFNC versus low-flow
oxygen showed no differences in respiratory rates up to 24 hours according to type of
oxygen delivery device (MD, IV, random-effects -1.51, 95% CI -3.36 to 0.35), and no difference after 24 hours (MD, IV, random-effects -2.71, 95% CI -7.12 to 1.70; two studies; 445 participants). Improvement in respiratory rates when HFNC was compared with CPAP or BiPAP was not clinically important (MD, IV, random-effects -0.89, 95% CI -1.74 to -0.05; two studies; 834 participants). Results showed no differences in patient-reported measures of comfort according to
oxygen delivery devices in the short term (MD, IV, random-effects 0.14, 95% CI -0.65 to 0.93; three studies; 462 participants) and in the long term (MD, IV, random-effects -0.36, 95% CI -3.70 to 2.98; two studies; 445 participants); we downgraded the certainty of this evidence to low. Six studies measured dyspnoea on incomparable scales, yielding inconsistent study data. No study in this review provided data on
positive end-expiratory pressure measured at the pharyngeal level, work of breathing, or cost comparisons of treatment.
AUTHORS' CONCLUSIONS: