Skip to main content

Table 3 Assessment of benefits and disadvantages in asthma exacerbation

From: Does zafirlukast reduce future risk of asthma exacerbations in adults? Systematic review and meta-analysis

Clinical outcomes

Comparisons

Therapy type

Illustrative comparative risk

Relative effect (95% CI)

No. of participants (studies)

Quality of evidence (GRADE)

With comparator

With intervention

Asthma exacerbation

Zafirlukast vs. placebo

First-line

92/1000

62/1000 (44 to 92)

OR = 0.68 [0.45, 1.00]

1,255 (n = 4)

Low1

Zafirlukast vs. ICs

First-line

31/1000

64/1000 (42 to 96)

OR = 2.11 [1.35, 3.30]

1,963 (n = 6)

Moderate2

Zafirlukast vs. ICs

Add-on

83/1000

92/1000 (46 to 175)

OR = 1.12 [0.53, 2.34]

353 (n = 2)

Low3

Zafirlukast vs. placebo

Add-on

121/1000

120/1000 (69 to 200)

OR = 0.96 [0.54, 1.81]

383 (n = 3)

Low4

Asthma exacerbation requiring systemic corticosteriod

Zafirlukast vs. placebo

First-line

108/1000

86/1000 (40 to 186)

OR = 0.76 [0.45, 1.29]

544 (n = 2)

Low5

Zafirlukast vs. ICs

First-line

18/1000

64/1000 (32 to 123)

OR = 3.71 [1.82, 7.59]

1,089 (n = 3)

Moderate6

Asthma exacerbation requiring emergency treatment

Zafirlukast vs. ICs

First-line

22/1000

23/1000 (10 to 53)

OR = 1.07 [0.46, 2.51]

994 (n = 3)

Low7

 

Zafirlukast vs. placebo

Add-on

60/1000

44/1000 (11 to 159)

OR = 0.72 [0.18, 2.99]

163 (n = 3)

Low8

  1. CI,= confidence intervals; ICs,= inhaled corticosteroids; OR, odds ratio.
  2. 1 (−1 limitations) (−1 publication bias). Four trials (Boushey HA 2005, Busse W 2001, Fish JE 1997, Nathan RA 1998) have high quality, but all of them fail to adhere to an intention-to-treat analysis, so suggesting high likelihood of bias (−1 of quality). Four trials were included, from the funnel plot we strongly suspected there was publication bias.
  3. 2 (−1 limitations). Busse W 2001 and Nathan RA 1998 fail to adhere to an intention-to-treat analysis, so suggesting high likelihood of bias (−1 of quality).
  4. 3 (−1 limitations) (−1 publication bias). Busse W 2001 and Boushey HA 2005 fail to adhere to an intention-to-treat analysis, so suggesting high likelihood of bias (−1 of quality). Only two trials were included, from the funnel plot we strongly suspected there was publication bias.
  5. 4 (−1 limitations) (−1 publication bias). Three trials (Boushey HA 2005, Busse W 2001, Huang CJ 2003) have high quality, but all of them fail to adhere to an intention-to-treat analysis, so suggesting high likelihood of bias (−1 of quality). Three trials were included, from the funnel plot we strongly suspected there was publication bias.
  6. 5 (−1 limitations) (−1 publication bias).Busse W 2001 and Nathan RA 1998 fail to adhere to an intention-to-treat analysis, so suggesting high likelihood of bias (−1 of quality). Only two trials were included, from the funnel plot we strongly suspected there was publication bias.
  7. 6 (−1 limitations) (−1 publication bias) (+1 large effect). Busse W 2001 fails to adhere to an intention-to-treat analysis, so suggesting high likelihood of bias (−1 of quality). Three trials were included, from the funnel plot we strongly suspected there was publication bias. Large effect (M-H pooled OR = 3.712) in the absence of other methodological limitations), so upgrading quality of evidence.
  8. 7 (−1 limitations) (−1 publication bias).Boushey HA 2005 fails to adhere to an intention-to-treat analysis, so suggesting high likelihood of bias (−1 of quality). Trials were included, from the funnel plot we strongly suspected there was publication bias.
  9. 8 (−1 limitations) (−1 publication bias).Boushey HA 2005 and Huang CJ 2003 fail to adhere to an intention-to-treat analysis, so suggesting high likelihood of bias (−1 of quality). Trials were included from the funnel plot we strongly suspected there was publication bias.