Thirteen RCTs with 31 comparisons were included (n at least 139,973; the total number was unclear because of the potential double-counting of patients in the control groups).
The results reported below were obtained from random-effects models.
Culture-confirmed influenza (5 RCTs with 12 comparisons, n=6,668): all 5 studies were set in the USA. The influenza vaccine significantly reduced culture-confirmed influenza (OR 0.21, 95% CI: 0.13, 0.34). Significant heterogeneity was found (P<0.0001). The one study showing a different direction of treatment effect had the lowest influenza attack rate amongst controls.
The OR for the inactivated vaccine was 0.28 (95% CI: 0.17, 0.47; no significant heterogeneity found, P=0.14) and the OR for the live-attenuated vaccine was 0.16 (95% CI: 0.08, 0.34; significant heterogeneity found, P=0.0002). There was significant heterogeneity between vaccines (P=0.01).
Serologically-confirmed influenza (4 RCTs with 11 comparisons, n=3,581): 2 studies were set in Russia and two in the USA. The influenza vaccine significantly reduced serologically-confirmed influenza (OR 0.30, 95% CI: 0.20, 0.45). Significant heterogeneity was found (P=0.0003).
The OR for the inactivated vaccine was 0.26 (95% CI: 0.18, 0.38; no significant heterogeneity found, P=0.20) and the OR for the live-attenuated vaccine was 0.36 (95% CI: 0.18, 0.74; significant heterogeneity found, P=0.0001). There was no significant heterogeneity between vaccines (P=0.17).
Clinical influenza-like illness (6 RCTs with 16 comparisons, n=161,390): the studies were conducted in five different countries. The influenza vaccine significantly reduced clinical influenza-like illness (OR 0.54, 95% CI: 0.50, 0.59). Significant heterogeneity was found (P<0.0001).
The OR for the inactivated vaccine was 0.51 (95% CI: 0.40, 0.64; significant heterogeneity found, P=0.0009) and the OR for the live-attenuated vaccine was 0.54 (95% CI: 0.49, 0.59; significant heterogeneity found, P<0.0001). The P-value for difference between vaccines was 0.06.
Estimated efficacy was 74% for culture-confirmed influenza, 59% for serologically-confirmed influenza and 33% for clinical influenza-like illness.
There was no evidence of publication bias (P-values for Begg and Muzumdar and Egger's tests ranged from 0.08 to 0.94 across outcomes).
The results were similar when data from the control groups were used only once.