Sara van den Berg

57 3 CMV on influenza vaccination: a systematic review Secondly, the odds ratio (OR) with 95% CI for the association between a response to influenza vaccination and CMV-serotatus was investigated (outcome b). The ORs were calculated using the numbers of responders and non-responders presented in the studies, where response to vaccination was defined as a ≥ four-fold increase, a post-vaccination titer ≥ 40 hemagluttinating units (HAU), or both. Meta-analysis was performed in R 3.3.3 using the ‘metafor’ package [35], to compare and pool the ORs of the studies. Random effects meta-analysis was performed in R 3.3.3 using the ‘metafor’ package [35]. The pooled OR and 95% CI using the DerSimonian-Laird method was calculated for the total group and for the predefined subgroups of young (<60 years of age) and old (>60 years of age) participants [36]. Heterogeneity among studies was assessed by the χ ²-based Q test and I² statistics. To analyze the influence of the quality of the studies, a sensitivity analysis was done by calculating the pooled OR for the highest-quality studies which had least potential for bias and confounding. The presence of possible publication bias was assessed using funnel plot regression [37]. The possible presence of any undetected studies and an effect estimate adjusted for publication bias were calculated using the trim-and-fill function in the ‘metafor’ package [38] in R 3.3.3. Thirdly, the role of CMV antibody levels, instead of CMV-serostatus, was investigated by extracting associations between influenza antibody titers and CMV antibody levels (outcome c). Correlations between CMV antibody levels and influenza antibody titers were tabulated, and outcomes of regression models incorporating CMV antibody levels and influenza antibody titers were narratively synthesized. No restriction for this outcome was placed on reporting post-vaccination titers or fold increase influenza titers to vaccination. Likewise, no restriction was placed on the performed statistics. RESULTS Selection and quality assessment of retrieved articles Study selection and characteristics The comprehensive EMBASE search for articles on CMV and influenza vaccination retrieved 689 individual publications ( Figure 2 ). The first selection, based on screening of title and abstract reduced this to 83 articles. Reasons for exclusion were for example no reporting of influenza vaccination response, no CMV-serostatus measurement or no primary research article. The selected 83 articles were assessed in full text and 15 articles of these were included in this systematic review. Articles were excluded if the data came from acute CMV infection or CMV disease or from an immune compromised population, or when CMV infection was only studied in vitro or in an animal model. The agreement on the independently performed study selection between authors KW and SB was large with disagreement on only three studies. These differences could easily be resolved by discussion and judgement by a third person was not needed. References of the 15 included articles were checked for additional relevant articles, but none were found. One article (Furman et al ), that contained the outcomes of three different study populations, was assessed as three individual studies, which brought the total number of studies up to 17. Whenever possible, study records were