Twenty-two papers were included in the meta-analysis, giving a total of 11,264 observations: 5,416 serological observations, 2,858 observations of local reactions, and 2,990 observations of systemic reactions.
Serology: when pooling all the data, regardless of vaccine type, influenza (sub)type and other stratifications, the PRD was not significant (PRD -0.6%, 95% CI: -3, 1.9). However, this value was influenced by the quality score: the PRD values were -9.4%, -0.3% and 1.1% for poor-, medium- and high-quality papers, respectively. The main reason for the poor quality of 2 papers was that their study design did not include random allocation of the vaccines. These papers were excluded from further calculations to avoid a non-randomised bias.
The test for homogeneity on treatment effects allowed for subsequent meta-analysis calculations, except for the comparisons of SU versus WV influenza B vaccines. The authors stated that heterogeneity was probably caused by age differences. This was supported by the finding that after using a stratification on age, homogeneity was regained.
SU and SPL vaccines were found to be immuno-equivalent. For all comparisons between SU and SPL vaccines, PRD values were positive, i.e. in favour of the SU vaccines (range: 0.7 to 8%), but not statistically significant.
For the comparisons between the SU and WV vaccines, all but one of the PRD values were positive, although non significant. Only for the H1N1 subtype in patients born after 1957 was the WV vaccine significantly more immunogenic than the SU vaccine (PRD -12.9%, 95% CI: -23.7, -2.1).
Reactogenicity: compared with SPL and WV vaccines, SU vaccines showed a lower frequency of both local and systemic reactions; the PRD values varied between -8.9% (95% CI: -17.4, -0.3) and -3.0% (95% CI: -6.1, -0.1). Three of the four comparisons were significantly different from zero.
The SU versus WV comparisons suffered from heterogeneity which was caused by a few studies with extreme rate differences. When these meta-analysis groups were excluded, homogeneity was re-established. Subsequently, the PRD values were still negative (borderline significance), indicating a slightly lower frequency of reactions after SU vaccination, compared with the WV vaccine.