Twenty-four randomised, double-blind placebo-controlled trials (1,087 children) were included.
Homogeneity: the authors described the included trials as being non-homogeneous because of differences in the source population, patient age, asthma severity, and type, dose, and duration of inhaled steroid. However, no formal tests of heterogeneity were presented.
Total symptom scores (15 trials): the overall weighted RIM total symptom score (inhaled steroid versus placebo) was 50% (95% CI: 49, 51).
Cough scores (11 trials): the overall weighted RIM cough score (inhaled steroids versus placebo) was 24% (95% CI: 23, 25).
Wheeze scores (16 trials): the overall weighted RIM wheeze score (inhaled steroids versus placebo) was 36% (95% CI: 35, 37).
Concomitant beta-agonist use (14 trials): the overall weighted relative decrease in mean beta-agonist use (inhaled steroid versus placebo) was 37% (95% CI: 36, 38).
Concomitant oral steroid use (12 trials): the overall weighted relative decrease in mean concomitant oral steroid use (inhaled steroid versus placebo) was 68% (95% CI: 66, 70).
PEFR estimated in litres per minute (L/min) (9 trials): the overall weighted absolute improvement in mean PEFR (inhaled steroid versus placebo) was 38 L/min (95% CI: 34.3, 41.7).
PEFR estimated as percentage predicted (5 trials): the overall weighted absolute improvement in mean PEFR (inhaled steroid versus placebo) was 11% predicted value (95% CI: 9.5, 12.5).
Hospitalisations (5 trials): there was no statistically-significant difference in hospital admission rates between the steroid and placebo groups.
Adverse effects (23 trials): all the reported adverse effects were minor; they were not sufficiently severe in any study to warrant stopping randomised treatment.
Modifying factors: compared with placebo, inhaled steroids appeared to be more effective in reducing symptoms in older children (those aged over 5 years), in children with severe disease (oral steroid-dependent), and at higher doses (greater than 400 microg/day); however, the differences were not statistically significant.