Comparison | MD or OR (95% CI) | p-value | I2 (%) | τ2 | K | GRADE# | Classification of intervention¶ |
Re-intubation | 36 | 0.120 | |||||
HFNO versus COT | 0.60 (0.43–0.84) | 0.003 | 11 | Moderatea | Large beneficial effect | ||
NIV versus COT | 0.61 (0.46–0.81) | <0.001 | 15 | Moderatea | Large beneficial effect | ||
NIV versus HFNO | 0.98 (0.69–1.40) | 0.844 | 5 | Moderateb | |||
VAP | 0 | 0 | |||||
HFNO versus COT | 0.36 (0.18–0.71) | 0.003 | 3 | Moderatea | Large beneficial effect | ||
NIV versus COT | 0.49 (0.34–0.73) | <0.001 | 7 | Moderatea | Large beneficial effect | ||
NIV versus HFNO | 0.74 (0.39–1.38) | 0.855 | 1 | Moderateb | |||
Discomfort | 49 | 0.493 | |||||
HFNO versus COT | 2.60 (0.79–8.60) | 0.117 | 7 | Lowa,b | Large harmful effect | ||
NIV versus COT | 13.14 (5.94–29.04) | <0.001 | 1 | Moderatea | Large harmful effect | ||
NIV versus HFNO | 0.20 (0.06–0.62) | 0.050 | 1 | High | |||
ICU mortality | 6 | 0.017 | |||||
HFNO versus COT | 0.81 (0.44–1.48) | 0.489 | 5 | Lowa,b | Small beneficial effect | ||
NIV versus COT | 0.64 (0.39–1.05) | 0.075 | 8 | Moderatea | Large beneficial effect | ||
NIV versus HFNO | 1.26 (0.63–2.52) | 0.834 | 1 | Lowa,b | |||
Hospital mortality | 6 | 0.017 | |||||
HFNO versus COT | 0.86 (0.62–1.18) | 0.431 | 5 | Moderatea | Small beneficial effect | ||
NIV versus COT | 0.64 (0.47–0.87) | 0.004 | 11 | Moderatea | Large beneficial effect | ||
NIV versus HFNO | 1.35 (0.97–1.86) | 0.194 | 4 | Moderateb | |||
ICU length of stay | 87 | 1.048 | |||||
HFNO versus COT | −0.12 (−0.85–0.61) | 0.741 | 9 | Very lowa,b,c,d | Trivial beneficial effect | ||
NIV versus COT | −0.72 (−1.44–0.00) | 0.049 | 14 | Very lowa,c,d | Small beneficial effect | ||
NIV versus HFNO | 0.60 (−0.33–1.53) | 0.320 | 4 | Very lowb,c,d | |||
Hospital length of stay | 86 | 2.230 | |||||
HFNO versus COT | −0.04 (−1.27–1.18) | 0.264 | 8 | Very lowa,b,c | Trivial beneficial effect | ||
NIV versus COT | −2.38 (−3.69– −1.07) | <0.001 | 10 | Lowa,c | Large beneficial effect | ||
NIV versus HFNO | 2.34 (0.80–3.88) | 0.189 | 3 | Very lowa,b,c | |||
Time to re-intubation | 89 | 87.485 | |||||
HFNO versus COT | 2.99 (−8.13–14.10) | 0.564 | 3 | Very lowa,b,c | Trivial harmful effect | ||
NIV versus COT | 9.17 (−1.06–19.40) | 0.105 | 4 | Very lowa,b,c | Small harmful effect | ||
NIV versus HFNO | −6.18 (−18.86–6.5) | 0.371 | 1 | Very lowa,b,c |
MD: mean difference; OR: odds ratio; τ2 and I2: loop heterogeneity; K: number of studies providing direct evidence; GRADE: Grading of Recommendations Assessment, Development and Evaluations; HFNO: high-flow nasal oxygen; COT: conventional oxygen therapy; NIV: noninvasive ventilation; VAP: ventilator-associated pneumonia; ICU: intensive care unit. #: GRADE certainty of evidence: a: lowered one level for risk of bias; b: lowered one level for imprecision as confidence intervals do not allow excluding harm; c: lowered one level for inconsistency; d: lowered one level for incoherence. ¶: for dichotomous outcome variables, large intervention effects were considered for OR >1.3 or <0.7, moderate intervention effects for 1.2<OR≤1.3 or 0.7≤OR<0.8, small intervention effects for 1.1<OR≤1.2 or 0.8≤OR<0.9 and trivial intervention effects for 1.0<OR≤1.1 or 0.9≤OR<1.0; for length of stay, large intervention effects were considered for MD>2, moderate intervention effects for 1<MD≤2, small intervention effects for 0.5<MD≤1 and trivial intervention effects for 0<MD≤0.5; for time to re-intubation, large intervention effects were considered for MD>24, moderate intervention effects for 12<MD≤24, small intervention effects for 6<MD≤12 and trivial intervention effects for 0<MD≤6. COT was considered as the reference intervention. Bold indicates p<0.05.