Skip to main content

Table 3 Confidence in the relative risk of each comparison and outcome assessed by the GRADE system for short-term mortality, reintubation, and post-extubation respiratory failure

From: Post-extubation oxygenation strategies in acute respiratory failure: a systematic review and network meta-analysis

  Risk of bias across studies Imprecision Heterogeneity Indirectness Publication bias Incoherence Confidence in relative risk of the event
Short-term mortality
NIV vs. COT Undetected Very seriousa
(95% CI 0.53–1.06)
No concern
(95% PI 0.51–1.11)
Low Not suggested No concern
(p = 0.33)

Low
HFNC vs. COT Undetected Seriousb
(95% CI 0.67–1.27)
No concern
(95% PI 0.64–1.33)
Low Not suggested No concern
(p = 0.33)

Moderate
HFNC vs. NIV Undetected Seriousb
(95% CI 0.61–1.08)
No concern
(95% PI 0.58–1.13)
Low Not suggested No concern
(p = 0.33)

Moderate
Reintubation
NIV vs. COT Undetected Seriousb
(95% CI 0.30–1.00)
Some concernc
(95% PI 0.16–1.84)
Low Not suggested No concern
(p = 0.58)

Moderate
HFNC vs. COT Undetected Not serious
(95% CI 0.32–0.89)
Major concernd
(95% PI 0.17–1.70)
Low Not suggested No concern
(p = 0.58)

High
HFNC vs. NIV Undetected Very seriousa
(95% CI 0.53–1.97)
No concern
(95% PI 0.29–3.55)
Low Not suggested No concern
(p = 0.58)

Low
Post-extubation respiratory failure
NIV vs. COT Undetected Very seriousa
(95% CI 0.54–1.38)
No concern
(95% PI 0.29–2.58)
Low Not suggested No concern
(p = 0.56)

Low
HFNC vs. COT Undetected Seriousb
(95% CI 0.43–1.02)
Some concernc
(95% PI 0.23–1.92)
Low Not suggested No concern
(p = 0.56)

Moderate
HFNC vs. NIV Undetected Very seriousa
(95% CI 0.79–2.14)
No concern
(95% PI 0.42–3.98)
Low Not suggested No concern
(p = 0.56)

Low
  1. CI confidence interval, COT conventional oxygen therapy, HFNC high-flow nasal therapy, NIV noninvasive ventilation, PI prediction interval
  2. aConfidence interval extends into clinically important effects in both directions
  3. bConfidence interval extends into clinically important effects
  4. cPrediction interval extends into clinically important or unimportant effects
  5. dPrediction interval extends into clinically important effects in both directions