Twenty-six studies were included in the review, of which 12 were divided into sub-trials as they described the results of different trials and/or contained more than two treatment arms. Therefore, 53 trials (or sub-trials) were included, with a total of 46,581 participants. Forty-seven trials were classified as randomised controlled trials (RCTs) and 6 trials were classified as quasi-RCTs.
The mean quality assessment score was 0.39 out of a maximum possible score of one (range: 0.13 to 0.77) when using the Chalmers scale, and 2.52 out of a maximum possible score of 5 when using the Jadad scale. More recent studies scored higher for quality.
There was a statistically significant reduction in clinically confirmed cases of influenza with vaccination (49 trials; RR 0.78, 95% CI: 0.72, 0.84, P<0.0005; random-effects model). When using the fixed-effect model, the pooled RR was 0.83 (95% CI: 0.80, 0.86). There was evidence of statistically significant heterogeneity between studies (chi-squared 142.39, d.f.=48, P<0.0005).
There was a statistically significant reduction in laboratory confirmed cases of influenza with vaccination (25 trials; RR 0.37, 95% CI: 0.29, 0.47, P=0.001; random-effects model). When using the fixed-effect model, the pooled RR was 0.38 (95% CI: 0.33, 0.43). There was evidence of statistically significant heterogeneity between studies (chi-squared 52.19, d.f.=24, P=0.001).
Higher quality studies had a smaller effect size than lower quality studies. However, there was still a statistically significant reduction in both clinically confirmed (28 trials; RR 0.86, 95% CI: 0.80, 0.91) and laboratory confirmed (14 trials; RR 0.41, 95% CI: 0.31, 0.55) cases of influenza with vaccination in the higher quality trials (results presented according to Chalmers scale; the results were similar using the Jadad scale).
An aerosol inactivated vaccine was more effective at preventing clinically confirmed cases of influenza (6 trials; RR 0.45, 95% CI: 0.28, 0.73) than a parenteral inactivated vaccine (35 trials; RR 0.77, 95% CI: 0.70, 0.85) or a live aerosol vaccine (8 trials; RR 0.85, 95% CI: 0.77, 0.92).
Vaccination was more effective in patients aged younger than 33 years than in patients aged 33 and older. The RRs for clinically confirmed and laboratory confirmed cases were 0.54 (15 trials; 95% CI: 0.44, 0.67) and 0.22 (5 trials; 95% CI: 0.13, 0.37), respectively, for those younger than 33 years and 0.89 (23 trials; 95% CI: 0.85, 0.94) and 0.43 (16 trials; 95% CI: 0.33, 0.57) for those aged 33 and older.
Other covariates tested (placebo control versus B vaccine control, matching of vaccine and circulating strains, recommended vaccine strains used) had smaller effects on the results.
Statistically significant publication bias was present amongst trials that assessed clinically confirmed cases of influenza (P=0.002 adjusted rank correlation test; P<0.0005 regression asymmetry test), but not amongst trials that assessed laboratory confirmed cases of influenza (P=0.072 adjusted rank correlation test; P=0.083 regression asymmetry test).