Twenty-nine RCTs (2,674 patients; calculated 2,669 patients) were included in the review. Follow-up duration ranged from six hours to four weeks (where specified). Eleven trials had high risk of bias and eleven trials had unclear risk of bias.
Post-tonsillectomy bleeding episodes
The difference in odds of bleeding with steroid treatment versus control groups was not statistically significant (OR 0.96, 95% CI 0.66 to 1.40, 29 trials). No statistical heterogeneity was observed (Ι²=0%). Sensitivity analyses yielded similar results (reported fully in paper). According to GRADE criteria, this was evidence of very low quality.
Hospital admission
No significant statistical difference in hospital admission (due to a bleeding episode) was found between steroid and control groups (OR 1.16, 95% CI 0.68 to 2.00, 17 trials). Very low statistical heterogeneity was shown (Ι²=19%). All sensitivity analyses yielded similar results (data not reported). This evidence was rated as being very low quality (GRADE).
Reintervention for a bleeding episode
The incidence of operative reintervention for bleeding episodes was statistically significantly increased in the steroid treatment groups, compared with control groups (OR 2.27, 95% CI 1.03 to 4.99, 12 trials). No statistical heterogeneity was indicated (Ι²=0%). This evidence was rated as high quality (GRADE).
Sensitivity analyses revealed a statistically significant increase in reinterventions among children (OR 3.43, 95% CI 1.29 to 9.13, eight trials, Ι²=0%), among trials containing less than one hundred patients (OR 3.85, 95% CI 1.13 to 13.17, seven trials, Ι²=0%) and among those with a steroid dosage regimen of 0.5 mg/kg or less (OR 2.45, 95% CI 1.04 to 5.76, eight trials, Ι²=0%). A borderline statistically significant increase in reintervention among patients receiving non-steroidal anti-inflammatory drugs was shown (OR 4.10, 95% CI 0.99 to 16.97, three trials, Ι²=0%).
No significant differences in reintervention incidence were observed in any of the other sensitivity analyses (reported in paper); double blinded trials and trials with longer follow-ups suggested strong (non-significant) trends towards increased incidence among steroid groups.
Allogenic transfusion and mortality
None of the included trials reported transfusion of red blood cells or mortality.
Publication bias and quality of evidence
The authors stated that publication bias was not present.