Sara van den Berg

182 Chapter 7 The majority of CMV+ individuals were CMV+ during entire follow-up The duration of CMV infection in the 155 individuals who were CMV+ at endpoint was determined by measuring CMV-specific antibody levels in the preceding 25 years. The vast majority (85.9%, n=136) of these participants had been CMV+ during the entire follow-up, and are therefore referred to as long-term CMV+ ( LT CMV+). The other CMV+ participants (n=19, of which 14 were women and 5 men) seroconverted during follow-up and are referred to as short-term CMV+ ( ST CMV+) ( Figure 1B , Supplementary Figure 2A ). An overview of the CMV-specific antibody levels over time of all CMV+ participants per 5 years age category, for men and women separately, is shown in Figure 1B . CMV-seroconverters were seen in all age categories and showed a sharp increase in antibody levels from undetectable to positive ( Figure 1B ). The average CMV-seroconversion rate was 0.56% per year in this adult population, and was higher for women than for men (0.88% versus 0.27% per year, p=0.03) ( Supplementary Figure 2B ). ST CMV+ individuals did not differ from LT CMV+ individuals in terms of age at endpoint or educational level ( Table 1 ). CMV-specific antibody levels within CMV+ individuals increase over time We investigated whether CMV-specific antibody levels increased with age. To this end, we selected samples at all time points of all CMV+ participants. We confirmed data from previous cross-sectional studies by showing a significant, positive correlation between geomean CMV-specific antibody levels and age (P<0.0001, r=0.15) ( Figure 2A, Supplementary Figure 3A ). We then used a median-based linear model to estimate the fold-change in CMV-specific antibody levels over time per individual using all available time points. This showed a significant increase within individuals over time ( Figure 2B ). On average, the increase was 1.01 (95%CI: 1.01-1.02) fold per year, or 1.42 (95%CI: 1.26-1.60) fold in 25 years (P<0.001), although in some individuals CMV-specific antibody levels decreased ( Figure 2B ) . Together, these data demonstrate a slight increase in CMV-specific antibody levels within CMV+ individuals over time. Variation in CMV-specific antibody levels at endpoint is largely explained by baseline CMV-specific antibody levels Antibody levels at endpoint showed considerable variation between individuals ( Figure 2G ). We studied which factors explained these differences using a random forest prediction algorithm. Explained variance in the algorithm to predict CMV-specific antibody levels at endpoint was 66%. The most important variable to predict turned out to be the “baseline” CMV-specific antibody level, where baseline is defined as the time of the first CMV+ observation, which was on average 27 years ago ( Figure 2H ). The increase in antibody levels over time was second in ranking of variable importance. Other factors, such as age at endpoint, age at seroconversion, duration of CMV infection, sex ( Figure 2H ) and educational level (data not shown) were much less predictive for CMV- specific antibody levels at endpoint.