Introduction
Type 1 diabetes is a chronic autoimmune disease with both genetic and environmental contributions. Viruses may influence susceptibility and trigger autoimmunity in individuals genetically predisposed to diabetes [
1,
2]. Enteroviruses and other viruses have been most frequently studied in relation to type 1 diabetes [
3]. Recently, respiratory virus infections have also been associated with the development of islet autoimmunity and the first manifestations of clinical symptoms of type 1 diabetes [
4‐
8]. Influenza virus can infect human pancreatic cell lines, and can cause pancreatitis and hyperglycaemia in animal models [
9]. Influenza A (H1N1) virus infection has also been associated with acute pancreatitis [
10‐
12] and type 1 diabetes [
13,
14] in human case-report studies.
Influenza spreads yearly across the continents, and when a new influenza virus emerges and transmits among humans, an influenza pandemic can occur [
15]. In April 2009, the World Health Organization detected an outbreak of a new influenza virus [Influenza A(H1N1)pdm09] in Mexico. This influenza, the swine flu, was declared a pandemic in June 2009, and this lasted until August 2010 [
16]. The virus has subsequently continued to circulate as one of the seasonal influenza strains.
Two small retrospective studies have suggested that pandemic influenza may be associated with type 1 diabetes [
17,
18]. These showed a concomitant increase in type 1 diabetes during the pandemic influenza period among children. However, as the typical time from induction of islet autoimmunity to clinical onset of type 1 diabetes is several years [
19], studies with longer follow-up after influenza are necessary to elucidate the role of pandemic influenza in development of diabetes.
No previously published studies have addressed whether pandemic influenza diagnosis is associated with the development of type 1 diabetes. We have recently reported that there was no association between vaccination against the 2009 H1N1 pandemic influenza virus with the AS03 adjuvanted Pandemrix vaccine and type 1 diabetes [
20]. In Norway, the vaccination campaign and the main influenza wave occurred simultaneously [
21,
22]. Here, we investigate a potential relationship between pandemic or seasonal influenza in the years 2006–2014 and subsequent type 1 diabetes in a nationwide register-based cohort study from Norway.
Discussion
In this nationwide study of all residents in Norway below 30 years of age, we found a twofold higher risk of developing type 1 diabetes in individuals who had a specialist care diagnosis or a laboratory-confirmed infection with pandemic influenza.
This is the first study using national registries to address the long-term risk of type 1 diabetes after a pandemic influenza diagnosis. A main strength of the study is the large sample size and the complete inclusion of all residents in Norway. Nearly all persons with type 1 diabetes are diagnosed and treated in the public health system in Norway, and consultations and medications are free for children with type 1 diabetes until 16 years of age. Dispensing of insulin registered in the Norwegian Prescription Database is likely to detect nearly all cases of type 1 diabetes [
26], and our algorithm for diagnosis of type 1 diabetes combining registers is likely to ensure true type 1 diabetes in the vast majority of cases [
27]. By restricting the study population to those below 30 years of age, we reduced the risk of misclassifying type 2 diabetes as type 1 diabetes. However, as misclassification of diabetes may have occurred in a small number of cases, we performed sensitivity analyses in which we restricted our sample to those below 15 years of age. In this age group, such misclassification is highly unlikely [
27].
The main limitation of this study is that we did not capture all cases of influenza, as we only have data on those who sought healthcare and received an influenza diagnosis. It is probable that many people infected with influenza did not seek healthcare, especially those with milder illness. Furthermore, males and females, or different age groups, might seek medical help following different patterns. Administration of antivirals without prescription during the pandemic may also have contributed to fewer people visiting a physician for treatment of influenza infection [
28]. Probably those with more severe symptoms were more likely to seek healthcare [
29], and our estimates are therefore likely to relate to more severe influenza infections. Infection severity or host response could be important, as influenza diagnosed in primary care did not show a clear association. A possible explanation is that this group contains misclassified cases that might not have been infected with influenza, or that susceptible individuals had more severe symptoms. In our study population, 3% were diagnosed with influenza during the pandemic. Likewise, a report from Sweden suggested that around 6% of the population was diagnosed with H1N1 [
30], an estimate probably restricted to those with symptomatic infection. It has been estimated that around 20–30% of the Norwegian population were infected during the 2009–2010 pandemic [
31]. Serum A H1N1 antibody positivity can occur after clinical influenza, after mild/quiescent non-clinical influenza or after vaccination with Pandemrix, making it difficult to estimate the true proportion affected by clinical H1N1 infection in a population. A study from Norway conducted in January 2010 showed influenza A (H1N1) serum antibody positivity in up to 65% of younger age groups [
32]. Many of these are probably positive because of vaccination with Pandemrix, as around 40% of the total population were vaccinated during the pandemic.
Most individuals were not tested for pandemic influenza as the capacity in laboratories was stretched, and at the point when the pandemic strain was considered to be the primary cause of influenza-like illness, it was regarded as unnecessary to test individuals in primary care when making a diagnosis. Therefore, we do not have laboratory confirmation for the majority of the pandemic influenza diagnoses in our study. During the pandemic period in Norway, there were periods in which non-influenza viruses may have given influenza-like symptoms and resulted in an influenza diagnosis. However, when restricting the exposure period to the pandemic peak period (October to December 2009), when no other influenza strains were circulating and most individuals under age 30 with influenza symptoms were likely to have pandemic influenza, we found similar results.
Previous studies of influenza and type 1 diabetes have been limited to a few small retrospective studies, showing a temporal relationship between H1N1 influenza infection and increased type 1 diabetes incidence [
17,
18]. In one study, influenza A antibodies were not associated with initiation of islet autoimmunity in children, but this study did not investigate pandemic influenza A (H1N1), included asymptomatic influenza infections and did not study clinical type 1 diabetes as outcome [
33]. In theory, viral infections may affect the progression from islet autoimmunity to clinical diabetes in the small proportion of individuals who are positive for islet autoantibodies [
34].
Any association with type 1 diabetes could in theory be due to non-specific immunological mechanisms associated with infections. Respiratory infections in early life and type 1 diabetes have been linked [
4‐
8], and we can consider influenza-like illness as a respiratory viral infection caused by influenza or other viruses. These, and other common viruses causing infection with fever, could be important as cytokine inducer and T cell activators [
3]. Our finding that pandemic influenza diagnosis in specialist healthcare was more strongly associated with type 1 diabetes may possibly indicate that an association with type 1 diabetes is stronger with severe infections (needing hospitalisation or other type of specialist care). It is plausible that those with more severe illness had their infection confirmed by laboratory test. Also, it may be that individuals with preclinical diabetes have an underlying higher risk of developing severe influenza. It is likely that most individuals with type 1 diabetes experience a short period of hyperglycaemia before clinical diagnosis of diabetes. The duration of this period is unlikely to last for much more than a few months [
35], a period shorter than the average time between influenza and diagnosis of type 1 diabetes in our current analysis. The increased risk of diabetes ascertained from laboratory-confirmed cases could also have occurred by chance. Even though we have a very large dataset, the actual numbers of laboratory-confirmed cases were small.
The dominating circulating influenza types usually differ by season, and it is difficult to discern from our data whether any specific strain tends to be more strongly associated with type 1 diabetes. We found the strongest evidence for pandemic influenza. However, there was also increased risk of type 1 diabetes after influenza in the following season (2010–2011), where influenza A (H1N1)pdm09 virus and influenza B co-dominated (albeit with slightly more influenza B) [
36]. It is also possible that the 2009 pandemic influenza strain has stronger tropism for pancreatic cells than other influenza strains [
37]. The immunological response to influenza infection with different severity and with different virus strains may differ and is not well understood [
38]. More studies are needed to conclude on the role of different seasonal influenza viruses in type 1 diabetes aetiology.
Our large register cohort does not include information on pre-diagnostic diabetes associated autoantibodies. Hence, we could not investigate whether influenza infections induced or accelerated autoimmunity. Viruses could contribute to the development of clinical diabetes through stress and inflammation in individuals with autoimmunity (non-specific effect of virus infections) [
39].
We could speculate that preventing viral infections, for example through influenza vaccination, could help reduce the incidence of type 1 diabetes. In a recent paper from The Environmental Determinants of Diabetes in the Young (TEDDY) study, the Pandemrix vaccination was associated with a lower risk of islet autoimmunity in children at increased genetic risk in Finland, whereas no difference was seen in Sweden [
40]. Unfortunately, we do not have access to data on autoimmunity in our register-based study. In line with the Swedish data we did not find any association with Pandemrix and type 1 diabetes in our study [
20]. In Norway, many of those who were vaccinated with Pandemrix had already been infected, or were infected with influenza A (H1N1) after vaccination, but before effective protective antibodies had been induced [
21].
In conclusion, we could not demonstrate a clear association between clinically reported pandemic influenza infection and incident type 1 diabetes in this register-based cohort study. We did, however, find a twofold excess of incident diabetes in the subgroups with laboratory-confirmed pandemic influenza A (H1N1) or pandemic influenza diagnosed in specialist healthcare. This suggests that respiratory infections may play a role in the aetiology of type 1 diabetes, but more studies are warranted.