68 Chapter 3 was partitioned in the incidence rates for both years. Subjects with unclear disease onset were excluded from our analyses. Hypersomnolence incidence rates We first repeated the locally estimated scatterplot smoothing (LOESS) models as described in Zhang et al. [159]. We analysed NT1 (N = 981) separately from the combined group of NT2 (N =189) and IH (N = 356). In summary, the incidence rate of each year was predicted by the LOESS model with 95% confidence intervals (CIs) and then the ratio R was calculated between the predicted value and the observed value in each year. LOESS models were first performed combining data from all countries and hereafter repeated separately in the countries with at least 150 individuals (the Netherlands, France, Italy and Czech Republic). The incidence rates were considered as R-fold significantly increased if the bottom of the 95% predictive CIs of R was larger than 1.3 and the P-value < 0.0001. Sensitivity analyses were performed for the combined group of NT2/ IH to test the robustness of possible incidence peaks. To test whether HLADQB1*06:02 positivity may have played a role, we first excluded the individuals that were HLA-DQB1*06:02 positive. We additionally implemented permutation testing in which we repeated the LOESS model 10,000 times, each with a 90% randomly sampled proportion of the children with NT2/IH. As an exploratory post-hoc analysis we also repeated the incidence rate analyses split for NT2 and IH in Appendix A. In addition, we divided the database into two subgroups: children cases (age of EDS onset ≤ 18 years old) and adult cases (age of EDS onset > 18 years old), and repeated the LOESS modelling in the two subgroups to further investigate whether possible increased incidence rates of the hypersomnolence disorders were age-specific. Age of symptom onset was missing in for some individuals in the databases from Innsbruck (29 NT1 and 16 IH with missing onset), Czech Republic (1 NT1) and the Slovak Republic (2 NT1) and these people were thus excluded from the analyses separating child- and adult-onset. Influenza season severity correlations For the countries with largest hypersomnolence incidence datasets (the Netherlands, France, Italy and Czech Republic, which also included data of both NT1 and NT2/IH patients) we additionally performed correlational analyses between hypersomnolence disorder incidence rates and influenza season severity defined by the number of infections in the preceding season (e.g., between the influenza season severity in 2009-2010 and narcolepsy incidence in 2010) with additional analyses on the different influenza strains.
RkJQdWJsaXNoZXIy MjY0ODMw