Sara van den Berg

29 2 CMV on influenza vaccination Hemagglutination-inhibition (HI) assay HI assays were performed in the pandemic season for A/California/7/2009(H1N1pdm09) and in season 2010-2011 for A/California/7/2009(H1N1pdm09) and A/Perth/16/2009(H3N2) to determine influenza virus-specific antibody titers before and after vaccination. Briefly, a dilution series of cholera filtrate-treated serum samples was incubated with four Hemagglutinin Units (HAU) of influenza virus for 20 minutes, 0.25% turkey erythrocytes for 45 minutes and scored for agglutination [39]. The influenza antibody titer is the inverse of the last dilution of the serum that completely inhibited hemagglutination. A detectable influenza antibody body titer is defined as > 5 HAU. Statistical analysis Antibody responses to H1N1pdm influenza vaccination in the pandemic season were expressed in two different ways: (1) influenza antibody titer and (2) protection rate (antibody titer ≥ 40 HAU). For all statistical analyses, influenza antibody titers were log (base 2) transformed, and presented as geometric mean titer (GMT) with 95% confidence interval (CI) in the figures. First, a two-tailed Student’s t-test (for two groups) or one-way ANOVA (for three or more groups) was used to explore group differences in influenza antibody titers (e.g. between low, medium and high CMV IgG groups). For the two-tailed T test, equality of variances was tested with Levene’s test for equality. Group differences in categorical variables were compared with the Fisher exact test. Second, we investigated the effect of latent CMV infection in a multivariate context; the effect of CMV infection on influenza antibody titers was adjusted for potential confounders using a Generalized Estimating Equations (GEE) regression model ( Supplementary Table 1 ) [40]. This model takes repeated measurements for the same individuals into account. For influenza antibody titer the normal distribution and for protection the binomial distribution of the model was used. The effect of CMV infection was investigated in two ways; (a) CMV-serostatus: CMV-seropositive individuals were compared to CMV-seronegative individuals, and (b) anti- CMV IgG groups: low, medium and high anti-CMV IgG levels were compared within CMV- seropositive individuals. Confounders included were age, sex, time and various variables concerning vaccination history (see Supplementary Table 1 ). The model yielded a beta regression coefficient for each variable, that reflects how a category (e.g. highest age group) compares to the reference category (e.g. lowest age group). Regression coefficients of the GEE models are given in Supplementary Table 2-7 . The model also yielded adjusted results (i.e. influenza antibody titers or protection rates) for each time point at which comparisons between CMV-serostatus or anti-CMV IgG group were performed, by including an interaction term between time and CMV-serostatus or anti-CMV IgG group in the GEE models. The adjusted outcomes of the models and pairwise comparisons are presented in the figures. These analyses were also performed for the influenza vaccine response in season 2010- 2011 for H1N1pdm and H3N2. P values of ≤ 0.10 were considered a trend and of ≤ 0.05 were