Twenty-one articles, which were split into 32 sub-studies (number of cases unclear) were eligible for inclusion in the review.
The mean quality score was 0.37 (range: 0.13 to 0.72) when using the Chalmers scale and 2.0 when using the Jadad scale. Overall, few trials reported adequate methods for randomisation and allocation concealment and few were truly double-blind according to the criteria of Chalmers and Jadad. Only 5 studies adequately considered the issue of withdrawals and drop-outs.
Incidence of LCCs of influenza (18 comparisons, n=8,574).
There were significantly fewer LCCs in children receiving vaccinations as compared with control groups (RR 0.33, 95% CI: 0.22, 0.49, p<0.00001; I-squared 87.2%).
Incidence of CCCs of influenza (19 comparisons, n=247,517).
There were significantly fewer CCCs in children receiving vaccinations as compared with control groups (RR 0.64, 95% CI: 0.60, 0.69, p=<0.00001; I-squared 88.9%).
Incidence of AOM (11 comparisons, n=11,349).
There were significantly fewer cases of AOM in children receiving vaccinations as compared with control groups (RR 0.49, 95% CI: 0.30, 0.79, p=0.004; I-squared 82.5%).
Subgroup and sensitivity analyses.
Factors that were found to affect the efficacy of vaccination included the participants' age, study quality (e.g. double-blinding, allocation concealment), and whether or not the study was carried out in the former Soviet Union. However, the effects varied across outcomes, and many of the subgroup and sensitivity analyses showed significant heterogeneity.
The funnel plot showed some evidence of potential publication bias for studies assessing CCCs and AOM, but the Begg and Egger tests for publication bias were not significant for trials that assessed CCCs. Only the Egger test found significant asymmetry for studies assessing AOM.