Eighteen RCTs (n=16,996 patients) were included in the review. Sample sizes ranged from 186 to 6,184 patients. Trial quality varied with nine RCTs judged to be at low risk of bias and nine RCTs where this was unclear. Trial duration ranged from 24 weeks to three years.
There was a statistically significantly higher incidence of pneumonia (pooled RR 1.60, 95% CI 1.33 to 1.92; 18 RCTs) and serious pneumonia (RR 1.71, 95% CI 1.46 to 1.99; 16 RCTs) in the inhaled corticosteroid groups. Subgroup analyses of trials with long-acting beta-agonists and placebos showed a similar effect for the outcome of serious pneumonia. There was no statistically significant difference between the groups for the outcomes of pneumonia-related mortality and overall mortality. Statistical heterogeneity was low in all analyses. Sensitivity analyses did not significantly affect the findings. The number needed to harm for the outcome of serious pneumonia was 47 (95% CI 34 to 73). The failsafe number was calculated as 16 trials with a mean N=945.