We found a 12.7-fold risk of narcolepsy in 4–19-year-old individuals within approximately 8 months after Pandemrix vaccination as compared to unvaccinated individuals in the same age group. This translates into a vaccine attributable risk of 1
Our study covers the entire population of Finland and is based on comprehensive data on individual Pandemrix vaccinations, diagnoses of narcolepsy and linkage of the two using unique personal identification codes assigned to all residents in Finland. Vaccination records were retrieved from primary health care databases. The high accuracy of the exposure data was confirmed through a validation check on a random sample. Newly diagnosed cases of narcolepsy were identified via a systematic nationwide search from the hospital registers, and the diagnoses were verified through a systematic stepwise expert review procedure.
Some parents may have been tempted to recall the onset of symptoms as occurring after, rather than before their child received the pandemic vaccine. Therefore, we used different definitions for disease onset to evaluate the significance of the timing of onset on the observed association. In the primary analysis, the earliest note of EDS in the patient's medical records was used to limit recall bias.
A particular concern is that the observed association is a result of increased detection of narcolepsy among vaccinated children. According to such a view, a similar increase in narcolepsy among unvaccinated children has occurred but is yet to be observed. This argument, however, is not supported by the factual circumstances. In early 2010, narcolepsy was a rare disease unknown to most parents. Also, very few primary care physicians had seen a narcoleptic child, and no beliefs, even less conviction associated narcolepsy with the pandemic vaccine. Yet considerable numbers of Pandemrix vaccinated children were already referred to specialist before the end of February 2010 and later diagnosed with narcolepsy. The sudden surge of referrals during the first months of 2010 can hardly be explained by increased awareness and changes in diagnostic practices alone. Awareness was aroused and referrals to specialist and diagnostic workup expedited only after the media attention from Sweden broke out in August 2010.
Should a confounding factor instead of vaccination be the true cause of the association, it would have to be even more strongly associated with narcolepsy than the pandemic vaccination as we now report. In addition, such a risk factor should have a strong and time dependent positive correlation with the vaccination itself. A recent study in China found a 3–4-fold greater than predicted occurrence of narcolepsy onset following the 2009–10 H1N1 pandemic season, which was independent of vaccination 
. In our study, there was no evidence of change in the incidence among the unvaccinated 4–19-year-olds after the first H1N1 epidemic in Finland, whereas a considerably increased risk was associated with vaccination. As H1N1 infection was hardly more common in the vaccinated than in the unvaccinated population, our findings contradict the Chinese observation. We can think of several infectious, environmental, social or psychological factors that could modify the strength of the association seen in this study but none that could completely undo an association of this magnitude.
Our finding is supported by the recent results from Sweden, where a cohort study covering the entire population reported an almost 7-fold incidence of narcolepsy with cataplexy in children vaccinated with Pandemrix compared to those in the same age group who were not vaccinated 
. The incidence in the unvaccinated (0.64/100,000 person-years) compares well to that seen in our study. Preliminary passive reporting system data from France, Norway and Ireland also indicate higher than expected number of cases in children and adolescents after Pandemrix vaccination 
. On the other hand, it is perplexing that both Canada and the United Kingdom lack the signal. In these two countries, genetic susceptibility to narcolepsy is as common as in the Nordic countries. This suggests multifactorial nature of the observed phenomenon.
The biological plausibility for a vaccine contributing to the increased risk of narcolepsy particularly in the signal-generating age group is based firstly on the immunomodulatory effects of vaccination and secondly on the fact that narcolepsy is strongly linked to the HLA DQB1*0602 allele 
. An analogous example of a similar disease process affecting children and adolescents in particular is provided by type 1 diabetes, in which insulin-producing beta-cells are destroyed by immunological mechanisms in genetically predisposed individuals with HLA DQB1*0302 and 02 alleles 
. Neither an increase nor an imbalance between the vaccinated and unvaccinated in the incidence of narcolepsy was seen in the population older than 19 years 
. It is noteworthy that the HLA DQB1*0602 allele is approximately twice as common in northern than in southern Europe 
and that apart from the Nordic countries, Ireland and Canada, the AS03 adjuvanted vaccine was not widely used in the age group from 4 to 19 years. It should therefore not be surprising that the signal was detected in Sweden and Finland.
Vaccinations may induce bystander activation of immunological responses especially due to function of adjuvants. The age-related differences in the immune responsiveness to Pandemrix vaccination may be of importance in the induction of the bystander activation of immune system 
. Pandemrix vaccination could have accelerated an on-going disease process rather than triggered narcolepsy associated autoimmunity. As computer search for peptide homologies between H1N1 virus and neuron-specific proteins did not reveal any potential molecular mimicry 
, it seems unlikely that H1N1 virus infection or vaccination induced cross-reactive autoimmunity against hypocretine-producing neurons.
Our finding raises concerns of lipid containing adjuvants. Animal models have suggested that squalene, although at higher doses than used in human vaccines, is capable of contributing to the development of autoimmunity 
. In humans, the epidemiological data available until now has not supported the induction of autoimmunity by squalene containing adjuvants. Adjuvanted vaccines are much needed to enhance immune responses, especially in immune compromised persons. The large scale use of new adjuvanted vaccines in human populations calls for further research of their association with adverse effects, such as autoimmunity.
Further studies are urgently needed to determine whether the association between adjuvanted pandemic vaccinations and narcolepsy can be demonstrated in other populations. The underlying immunological mechanism also warrants further research.