Skip to main content

Table 2 Confidence assessment in the risk ratio of each comparison and outcome

From: Respiratory support strategy in adults with acute hypoxemic 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

  HFNO vs SOT

Undetected

Major concerns (a)

95% CI (0.66–1.20)

No concerns

95% PI (0.47–1.70)

Low

Not suggested

Some concerns

P=0.07*

Low

  NIV vs SOT

Undetected

No concern

95% CI (0.61–0.95)

Major concerns (b)

95% PI (0.42–1.39)

Low

Not suggested

No concerns

P=0.05*

Low

  IMV vs SOT

Undetected

Major concerns (a)

95% CI (0.57–1.78)

No concerns

95% PI (0.45–2.29)

Low

Not suggested

Major concerns(c)

P= 0.01**

Very low

  HFNO vs IMV

Undetected

Major concerns (a)

95% CI (0.47–1.65)

No concerns

95% PI (0.37–2.10)

Low

Not suggested

Major concerns(c)

P=0.01**

Very low

  NIV vs IMV

Undetected

Major concerns (a)

95% CI (0.45–1.27)

No concerns

95% PI (0.34–1.65)

Low

Not suggested

Major concern(c)

P= 0.01**

Very low

  NIV vs HFNO

Undetected

Major concerns (a)

95% CI (0.61–1.20)

No concern

95% PI (0.44–1.67)

Low

Not suggested

Major concern(c)

P=0.01*

Very low

Endotracheal intubation

  HFNO vs SOT

Undetected

Some concerns

95% CI (0.61–1.11)

Some concerns

95% PI (0.41–1.64)

Low

Not suggested

No concern

P= 0.50*

Moderate

  NIV vs SOT

Undetected

No concern

95% CI (0.51–0.79)

Major concerns (b)

95% PI (0.33–1.22)

Low

Not suggested

No concern

P= 0.85*

Low

  NIV vs HFNO

Undetected

Some concerns

95% CI (0.56–1.06)

Some concerns

95% PI (0.38–1.56)

Low

Not suggested

No concern

P=0.24*

Moderate

  1. CI confidence interval, SOT standard oxygenation therapy, HFNO high-flow nasal oxygen, IMV invasive mechanical ventilation, NIV noninvasive ventilation, PI prediction interval
  2. *We used a side-splitting approach as a local method
  3. **We used the design-by-treatment model as a global method
  4. aConfidence interval extends into clinically important effects in both directions
  5. bPrediction interval extends into clinically important effects in bothdirections
  6. cP value of inconsistency was <0.05