Antibodies to Influenza Virus A/H1N1 Hemagglutinin in Type 1 Diabetes Children Diagnosed Before, During and After the SWEDISH A(H1N1)pdm09 Vaccination Campaign 2009–2010

Authors


Abstract

We determined A/H1N1-hemagglutinin (HA) antibodies in relation to HLA-DQ genotypes and islet autoantibodies at clinical diagnosis in 1141 incident 0.7-to 18-year-old type 1 diabetes patients diagnosed April 2009–December 2010. Antibodies to 35S-methionine-labelled A/H1N1 hemagglutinin were determined in a radiobinding assay in patients diagnosed before (n = 325), during (n = 355) and after (n = 461) the October 2009–March 2010 Swedish A(H1N1)pdm09 vaccination campaign, along with HLA-DQ genotypes and autoantibodies against GAD, insulin, IA-2 and ZnT8 transporter. Before vaccination, 0.6% patients had A/H1N1-HA antibodies compared with 40% during and 27% after vaccination (< 0.0001). In children <3 years of age, A/H1N1-HA antibodies were found only during vaccination. The frequency of A/H1N1-HA antibodies during vaccination decreased after vaccination among the 3 < 6 (= 0.006) and 13 < 18 (= 0.001), but not among the 6 < 13-year-olds. HLA-DQ2/8 positive children <3 years decreased from 54% (15/28) before and 68% (19/28) during, to 30% (9/30) after vaccination (= 0.014). Regardless of age, DQ2/2; 2/X (n = 177) patients had lower frequency (= 0.020) and levels (= 0.042) of A/H1N1-HA antibodies compared with non-DQ2/2; 2/X (n = 964) patients. GADA frequency was 50% before, 60% during and 51% after vaccination (= 0.009). ZnT8QA frequency increased from 30% before to 34% during and 41% after vaccination (= 0.002). Our findings suggest that young (<3 years) along with DQ2/2; 2/X patients were low responders to Pandemrix®. As the proportion of DQ2/8 patients <3 years of age decreased after vaccination and the frequencies of GADA and ZnT8QA were enhanced, it cannot be excluded that the vaccine affected clinical onset of type 1 diabetes.

Introduction

In June 2009, the World Health Organization (WHO) declared the first human influenza pandemic in the 21st century [1]. The first RT-PCR confirmed patient with A(H1N1)pdm09 in Sweden was reported in May 2009, although as in many European countries [2], the disease activity peaked in mid-November. During May 2009 to June 2010, 15% of the 11,009 laboratory-verified patients reported in Sweden were hospitalized and 31 patients died [3]. As seen in other countries [reviewed in [4, 5]], the attack rates (cumulative infection incidence during an outbreak) were highest in children, but they had a low case fatality rate [3]. The Swedish Institute for Communicable Disease Control (SICDC) estimated that approximately 6% of the population [3] were exposed to A(H1N1)pdm09, which is lower than that reported by the Centers for Disease Control and Prevention [6] and the WHO [7].

The Swedish A(H1N1)pdm09 vaccination campaign started in mid-October 2009, offering vaccine to all people older than 6 months. The only vaccine used in Sweden was an inactivated, split virion vaccine with AS03 adjuvant (Pandemrix®; GlaxoSmithKline, Middlesex, UK). Children under 13 years were offered two vaccine injections, each containing half of the adult dose [8]. The mean vaccination coverage in Sweden was 60% [3].

Type 1 diabetes (T1D) is a chronic metabolic disease characterized by autoimmune cell-mediated destruction of the pancreatic islet beta-cells. Autoantibodies against the beta-cell autoantigens, glutamic acid decarboxylase (GADA), islet antigen-2 (IA-2A), insulin (IAA) and three variants of ZnT8 transporter (ZnT8R-W-QA) both predict [9-11] and increase the diagnostic sensitivity [12] of T1D. The human leucocyte antigen (HLA) haplotypes, DQ2 (DQ A1*05:01-B1*02:01) and DQ8 (DQ A1*03:01-B1*03:02), are strongly associated with T1D, where particularly the DQ2/DQ8 genotype is the main genetic risk marker [13]. After Finland, Sweden has the second highest T1D incidence in the world [14]. The incidence rates are increasing in many countries, with the fastest rise occurring among the youngest children (≤4 years) [14]. Genetic factors alone may not explain this increase, supporting the hypothesis of contributing environmental factors.

The triggering of islet autoimmunity and clinical onset of T1D has been associated with virus infections [for reviews, see [15, 16]]. The Swedish A(H1N1)pdm09 vaccination campaign was unique in history. Although adverse events were considered rare, there were reports of serious adverse events demonstrating an increased incidence of narcolepsy, particularly in Sweden [17] and Finland [18]. This disease has a strong association with HLA-DQ6.2 (DQ A1*01:02-B1*06:02), suggesting a role of autoimmunity in the pathogenesis [18]. It can therefore not be excluded that the epidemic and mass vaccination may have affected immunogenetic, islet autoantibody and clinical measurements at the time of diagnosis of T1D. To our knowledge, this particular association has not been studied before. One Italian study [19] found an increased frequency of T1D cases during the 2009/2010 compared with the 2004/2005 influenza season. However, a Swedish retrospective cohort study did not find a difference in risk for T1D between vaccinated and unvaccinated subjects during the 8-to 10-month follow-up period [20]. However, it cannot be excluded that GADA, IA-2A, IAA and ZnT8A levels and frequencies were affected by exposure of the A(H1N1)pdm09 virus, Pandemrix®, or both. Other studies have shown that not only the number of autoantibodies [9-11] but also levels [21] may be associated with progression to clinical onset of T1D. The standard conventional tests for antibodies against A(H1N1)pdm09 are hemagglutination inhibition and microneutralization assays. In contrast, all six islet autoantibodies are conformation dependent, and we therefore developed a similar radiobinding assay to detect antibodies against A/H1N1-hemagglutinin (HA) (A/H1N1-HAAb). Hence, our assay is asking specific questions related to conformation-dependent antigenicity rather than hemagglutination inhibition.

The aims of the present study were to determine the presence and levels of A/H1N1-HAAb in children and adolescents with newly diagnosed T1D before, during and after the Swedish A(H1N1)pdm09 vaccination campaign (October 2009–March 2010). It cannot be excluded that some of these children developed A/H1N1-HAAb because they had been infected. We therefore tested the hypothesis that possible infection, Pandemrix®, or both, altered any characteristics (age, gender, frequencies or levels of GADA, IA-2A, IAA or ZnT8A, symptoms, as well as metabolic indicators) of T1D at time of diagnosis as well as the number of new patients diagnosed before, during and after the vaccination campaign. We also investigated whether these proposed effects varied in relation to T1D-related HLA-DQ genotypes, ethnic origin or both.

Materials and methods

Study design

The study period (1 April 2009 to 31 December 2010) was subdivided into three periods, before (1 April 2009–11 October 2009), during (12 October 2009–31 March 2010) and after (1 April 2010–31 December 2010) vaccination (Table 1) based on the timing of the Swedish A(H1N1)pdm09 vaccination campaign, which started 12 October 2009. We considered 31 March 2010, as the end of the vaccination campaign since the SICDC registered only few vaccinations after that date.

Table 1. Demographic details of newly diagnosed T1D children before, during and after vaccination
 Before vaccinationDuring vaccinationAfter vaccinationP-value
  1. The dividing of the periods was based on the timing of the Swedish A(H1N1)pdm09 vaccination campaign, which started on 12 October 2009 and ended on 31 March 2010.

  2. a

    Origin known for 87% (998/1141).

  3. bThe number of new patients (patients/days/105) comparing before–during (= 0.0008) as well as during–after (P = 0.0054) vaccination was different.

Time period1 April 2009–11 October 200912 October 2009–31 March 20101 April 2010–31 December 2010 
Number325355461 
A/H1N1-HAAb % positive0.6 (2/325)40 (143/355)27 (125/461)<0.0001
Gender -% male54 (176/325)53 (189/355)55 (254/461)0.869
Median age years (range)9.8 (1.3–18.0)10 (0.7–17.9)10 (1.2–17.7)0.902
HLA-DQ genotype
Frequency%%% 
DQ2/836 (116/325)33 (116/355)30 (138/461)0.235
DQ8/X36 (116/325)38 (134/355)41 (189/461)0.304
DQ2/X15 (48/325)18 (62/355)15 (67/461)0.471
DQX/X14 (45/325)12 (43/355)15 (67/461)0.598
Autoantibodies
Frequency positive%%% 
GADA50 (162/325)60 (214/355)51 (235/461)0.009
IAA38 (122/325)31 (109/355)35 (159/461)0.169
IA-2A75 (224/325)72 (257/355)72 (332/461)0.606
ZnT8RA56 (182/325)60 (214/355)62 (285/461)0.251
ZnT8WA49 (158/325)49 (175/355)48 (220/461)0.904
ZnT8QA30 (96/325)34 (120/355)41 (190/461)0.002
Known origina
Frequency positive%%% 
Swedish74 (210/284)74 (233/314)72 (286/400)0.665
Non-Swedish6.3 (18/284)8.9 (28/314)7.0 (28/400)0.447
Mixed20 (56/284)17 (53/314)22 (86/400)0.302
Number of new patients34 (204 days)44 (171 days)36 (275 days)
Study population

A total of 1208 newly diagnosed patients with childhood diabetes were identified in the national Swedish Better Diabetes Diagnosis (BDD) study [22]. All Swedish paediatric diabetes clinics are participating in the BDD study. The criteria of the American Diabetes Association [23] were used to determine diabetes phenotypes, and only T1D patients, clinically confirmed 6 months following diagnosis, were included in this analysis. A total of 1141 patients with T1D were subdivided into three groups, before (n = 325), during (n = 355) and after (n = 461) vaccination (Fig. 1). All BDD patients were at the time of diagnosis analysed for HLA-DQ genotype and islet autoantibodies GADA, IA-2A, IAA, ZnT8RA, ZnT8WA and ZnT8QA [24]. Demographic and clinical data as well as ethnicity (defined by country of birth of parents and grandparents) were obtained from a questionnaire administered at diagnosis. The patients were divided into five age groups being <3 (n = 86), 3 < 6 (n = 217), 6 < 9 (n = 191), 9 < 13 (n = 334) and 13 < 18 (n = 313) years of age.

Figure 1.

Flow chart of study patients. All diabetes patients up to 18 years of age in the BDD study in Sweden between 1 April 2009 and 31 December 2010. A total of 1141 patients were identified with T1D, divided into the study periods before (n = 325), during (n = 355) and after (n = 461) vaccination.

Ethnic origin (available for 998/1141; 87% of patients) was subgrouped into Swedish, non-Swedish and mixed (Swedish and non-Swedish origin) (Table 1). A patient was considered non-Swedish if both parents and all four grandparents were born outside Sweden [22].

A(H1N1)pdm09 epidemic and the Swedish A(H1N1)pdm09 vaccination campaign

We utilized data from the SICDC vaccination statistics by age groups (available only in age groups <3, 3 < 13, 13 < 65 and >65 years of age for 11/21 counties of Sweden). The number of new patients during the period October–March was calculated from data of the national register for childhood and adolescent diabetes (SWEDIABKIDS). Relevant population statistics were obtained from Statistics Sweden.

A/H1N1-HA antibodies

A classical radiobinding assay (RBA) widely used for autoantibody detection [25, 26] was established for A/H1N1-HAAb based on in vitro transcription translation of antigen cDNA [27]. The influenza A virus (A/California/04/2009/) segment 4 hemagglutinin (HA) gene was synthesized and cloned into the pJ201 vector by DNA2.0 (San Diego, CA, USA). The HA gene was subcloned into the pTNT vector (Promega, Southampton, UK) and verified by sequencing. Recombinant A/H1N1-HA was labelled with 35S-methionine (PerkinElmer Life and Analytical Sciences, Brussels, Belgium) by in vitro transcription translation in the TNT SP6-coupled reticulocyte lysate system (Promega, Southampton, UK) as described [27]. About 16% of 35S-methionine was incorporated into the expected Mr 60K HA protein. A/H1N1-HAAb were analysed in a standard RBA, separating antibody bound from free antigen with protein A-Sepharose [27]. One A/H1N1-HAAb positive sample and four negative samples were included in all assays. The intra-assay coefficient of variation (CV) for duplicates was 5.6% for the negative samples and 4.9% for the positive sample. All samples with a CV≥12 (49/1352, 3.6%) were reanalysed. The interassay CV of A/H1N1-HAAb levels was 14% for the negative samples and 15% for the positive sample. A serum with high level A/H1N1-HAAb was identified and used as an in-house standard to express A/H1N1-HAAb levels in RU/ml. The in-house standard 500 RU/ml was equivalent to 50 μg/ml IgG of a rabbit polyclonal antibody against recombinant A/H1N1 (A/California/04/2009) HA (Sino Biological Inc., Beijing, China) Recombinant HA (Sino Biological Inc., Beijing, China) was used in competition experiments to demonstrate specific binding of A/H1N1-HAAb in the RBA (Fig. 2). The cut-off level for A/H1N1-HAAb positivity was assigned at 71 RU/ml, corresponding to the 99th percentile of children (n = 325) who were diagnosed with T1D before the Swedish A(H1N1)pdm09 vaccination campaign (Fig. 3, panel A). The 99th percentile was used to minimize false-positive results.

Figure 2.

Displacement of 35S-methionine-labelled A/H1N1-HA antibody binding by non-radioactive A/H1N1-HA. 35S-methionine-labelled A/H1N1-HA was incubated with an in-house A/H1N1-HA standard (image), commercially available human serum with low-level binding (image) as well as a control serum (image) and 0.01–10 μg/ml non-radioactive (cold) recombinant A/H1N1-HA protein from Sino Biological Inc., Beijing, China. Recombinant A/H1N1-HA specifically displaced the binding of 35S-methionine labelled to the A/H1N1-HA in the in-house standard serum.

Figure 3.

Panel A–C. Q-Q plots before, during and after vaccination of levels (RU/ml) of A/H1N1-HAAb. Cut-off value (71 RU/ml) represents the 99th percentile in the population before vaccination, corresponding to 0.6% of patients before, 40% during and 27% after vaccination (< 0.0001). The cut-off is displayed with a line in all three plots.

Islet autoantibodies

Serum GADA, IA-2A, IAA, ZnT8RA, ZnT8WA and ZnT8QA were analysed at the time of diagnosis as described [24].

HLA-DQ typing

HLA-DQ A1 and B1 were typed by amplifying DNA in dried blood spots collected at time of diagnosis (≤3 days) by polymerase chain reaction (PCR) followed by allele-specific hybridization as described [24]. A total of 67 different HLA genotypes were identified and grouped based on the presence of DQ8 and DQ2 haplotypes into four genotype groups: (i) DQ2/8 (n = 370), (ii) DQ8/X which includes DQ8/8 and 8/X (where X is any other DQ haplotype except DQ2; n = 439), (iii) DQ2/X which includes DQ2/2 and 2/X (where X is any other DQ haplotype except DQ8; n = 177) and (iv)) DQX/X (where X is any other DQ haplotype except DQ8 or DQ2; n = 155).

Statistical analysis

SPSS 20® statistical package (SPSS Inc. IBM, Chicago, IL, USA) was used for statistical analyses to determine distribution of age, gender, A/H1N1-HAAb, HLA genotypes, islet autoantibodies and origin. Pearson chi-squared test of independence (including Yate's correction for continuity) and Fischer's exact test were used to determine possible associations between groups in relation to study time periods, A/H1N1-HAAb positivity and other indicators. The Mann–Whitney U or Kruskal–Wallis test was used to measure differences in levels of A/H1N1-HAAb or islet autoantibodies in relation to other variables. Finally, logistic regression models (tested at 95% confidence interval) were used to asses whether A/H1N1-HAAb were independently associated with age groups, HLA-DQ genotypes, islet autoantibodies or ethnic origin. The significance (α) level was set to 0.05.

Results

A/H1N1-HA antibodies

Two patients showed significant binding of the 35S-A/H1N1-HAAb before the Swedish A(H1N1)pdm09 vaccination campaign (Fig. 3, panel A). The binding distribution was markedly changed during and after vaccination (Fig. 3, panel B and C). Using the 99th percentile as cut-off level revealed that 0.6% had A/H1N1-HAAb before compared with 40% during and 27% after vaccination (< 0.0001) (Table 1). Median levels rose from 6 RU/ml before, to 43 RU/ml during and decreased to 30 RU/ml after vaccination (< 0.0001) (Fig. 4). In the youngest age group (<3 years), A/H1N1-HAAb were only detected during vaccination demonstrating that 11% (3/28) were positive during vaccination (Fig. 5). During vaccination and excluding the <3-year-olds, there was no difference between other age groups in the frequency of A/H1N1-HAAb (33–46%). After vaccination, A/H1N1-HAAb showed lower frequency values in all age groups compared with during vaccination (Fig. 5). The A/H1N1-HAAb frequencies during vaccination: 46% for both the 6 < 9 and the 9 < 13-year-olds decreased after vaccination to 33% and 42%, respectively (n.s.). In contrast, both the younger (3 < 6 years of age; 33% during vaccination) and the older (13 < 18 years of age; 43%) patients showed statistically lower frequencies after vaccination (14%; = 0.006 and 22%; = 0.001, respectively) (Fig. 5). The A/H1N1-HAAb levels did not differ between during and after vaccination for age groups older than 3 years of age (data not shown).

Figure 4.

A/H1N1-HAAb levels (log scale) before, during and after vaccination. Median levels as indicated were 6, 43 and 30 RU/ml, respectively (< 0.0001).

Figure 5.

Frequencies of A/H1N1-HAAb, in relation to age group before, during and after vaccination. The total number of patients positive for A/H1N1-HAAb before, during and after vaccination is shown within the boxes. Error bars show 95% confidence intervals. A/H1N1-HAAb were detected in only 2/325 patients before vaccination. In the youngest age group (<3 years), no one was positive for A/H1N1-HAAb before or after vaccination. During vaccination, there was no difference between age groups when excluding the youngest <3 years. For age groups 3 < 6 and 13 < 18, significantly fewer were positive for A/H1N1-HAAb after compared with during vaccination (= 0.006 and = 0.001, respectively).

HLA-DQ

The overall distribution of HLA-DQ genotypes before, during and after vaccination did not differ (Table 1). However, among the youngest patients (<3 years), DQ2/8 was found to vary from 54% before and 68% during to 30% after vaccination (= 0.014) (Fig. 6).

Figure 6.

Frequencies of the HLA-DQ genotypes in relation to age as well as before, during and after vaccination. The total number of patients in each HLA-DQ genotype group (DQ2/8, DQ8/X (DQ8/8 and 8/X, where X is any other DQ haplotype except DQ2) DQ2/X (DQ2/2 and 2/X, where X is any other DQ haplotype except DQ8) and DQX/X (X is neither DQ2 nor DQ8)) out of all patients before, during and after vaccination is shown within the boxes. Error bars show 95% confidence intervals. The proportion of DQ2/8 carriers among the youngest (<3 years) patients were reduced after the vaccination (P = 0.014).

Analyses of A/H1N1-HAAb in relation to HLA-DQ during and after vaccination combined (during separated from after vaccination was not significant) revealed that patients with DQ2/X had a lower frequency of A/H1N1-HAAb (24%) compared with DQ2/8, DQ8/X and DQX/X patients combined (non-DQ2/X patients; 35%; = 0.02). Combining during and after vaccination, the median A/H1N1-HAAb level in the DQ2/X patients was 28 RU/ml compared with 36 RU/ml in the non-DQ2/X patients (= 0.042) (see also Fig. 7).

Figure 7.

A/H1N1-HAAb levels (log scale) in relation to the HLA-DQ genotypes in the T1D patients during and after vaccination combined. The median levels for each genotype (DQ2/8, DQ8/X (DQ8/8 and 8/X, where X is any other DQ haplotype except DQ2) DQ2/X (DQ2/2 and 2/X, where X is any other DQ haplotype except DQ8) and DQX/X (X is neither DQ2 nor DQ8)) are indicated. DQ2/X patients had lower median level (28 RU/ml) compared with the DQ2/8, DQ8/X and DQX/X patients combined (36 RU/ml; = 0.042).

Islet autoantibodies

The frequency of GADA was 50% before, 60% during and 51% after vaccination (= 0.009) (Table 1). The median GADA levels varied between 49 RU/ml before, 99 RU/ml during and 53 RU/ml after vaccination (= 0.031). To correct for a possible seasonal variation, the BDD register was used to identify all patients diagnosed with GADA during October 12–March 31 (winter season) representing the 5.5 winter seasons of May 2005–December 2010. There was no difference in GADA frequency between the four winter seasons before, the one during and the remaining 3 months of the 2010 winter season (= 0.367). It was noted, however, that GADA for all years collectively was more common during the winter (1079/1833; 59%) than the summer season (974/1839; 53%; < 0.0001).

The frequency of ZnT8QA, but neither ZnT8RA nor ZnT8WA, increased from 30% before, to 34% during and 41% after vaccination (= 0.002) (Table 1). The median ZnT8QA levels increased accordingly from 16 RU/ml before, to 22 RU/ml during and to 42 RU/ml after vaccination (< 0.0001). There was no difference in ZnT8QA frequency between the winter and summer seasons (May 2005–December 2010; data missing in 2.3% of patients) (= 0.43). However, in contrast to GADA, ZnT8QA frequencies differed between the 5.5 winter seasons (< 0.0001) (Table 2).

Table 2. ZnT8QA in winter seasons 2005–2010
Winter season05/0606/0707/0808/0909/1010
  1. ZnT8A varied between the winter seasons (October 12–March 31); < 0.0001.

ZnT8QA % positive41 (127/312)31 (91/295)26 (89/348)27 (93/343)34 (120/357)36 (54/151)

The frequencies of IAA and IA-2A before, during and after vaccination did not differ (Table 1). Similarly, the number of islet autoantibodies before, during and after vaccination was not affected by A(H1N1)pdm09 vaccination, neither did they differ over the 5.5 winter seasons 2005–2010 (Fig. S1). Frequencies of multiple islet autoantibodies (≥2 autoantibodies) and of autoantibody negative patients did not differ before, during and after vaccination (data not shown). However, after vaccination, but not before or during, it was found that A/H1N1-HAAb were less frequent in children positive for IAA (34/159; 21%) compared with IAA-negative children (91/302; 30%) (= 0.045). No other association before, during or after vaccination was found for A/H1N1-HAAb and any combination of islet autoantibodies.

Origin

During and after vaccination combined, significantly fewer non-Swedish (12/54; 22%) were positive for A/H1N1-HAAb in comparison with Swedish (187/519; 36%) patients (= 0.043). Accordingly, the median level of A/H1N1-HAAb was lower in non-Swedish (22 RU/ml) compared with Swedish patients (37 RU/ml; = 0.024).

Clinical symptoms and other variables

There were no differences before, during and after vaccination in reporting clinical symptoms (polyuria, polydipsia and weight loss), presentation (ketoacidosis) or other clinical variables (gender, age at onset, p-glucose and HbA1c). Neither was there any association between the above variables and A/H1N1-HAAb, analysed during and after vaccination separately. Age at clinical onset did not differ between the 6 winter seasons of 2005–2011 (= 0.397) (data obtained from the SWEDIABKIDS register; Fig. S2).

Logistic regression

Different logistic regression models were performed using A/H1N1-HAAb positivity as dependent variable and age groups, HLA-DQ groups, origin and autoantibodies in various arrangements as independent variables, before, during and after vaccination. During vaccination, age was the main predictor for A/H1N1-HAAb as the <3-year-olds showed a negative association with A/H1N1-HAAb [OR = 0.14 (95% CI: 0.04–0.53), = 0.003] when compared to the 9 < 13-year-olds. Similarly, after vaccination, both the 3 < 6-year-olds [OR = 0.18 (95% CI: 0.08–0.39), < 0.0001] and the 13 < 18-year-olds [OR = 0.36 (95% CI: 0.2–0.66) = 0.001] were negatively associated with A/H1N1-HAAb compared with 9 < 13-year-olds. The analysis gave no valid result for <3-year-olds because there were no children in this age group positive for A/H1N1-HAAb after vaccination. After vaccination but not during, origin predicted A/H1N1-HAAb positivity, as being non-Swedish was negatively associated [OR = 0.2 (95% CI: 0.04–0.91), = 0.037] compared with being Swedish.

Combining during with after vaccination and keeping origin in the model, it was observed that DQ2/X was negatively associated with A/H1N1-HAAb when compared to DQ2/8 [OR = 0.54 (95% CI: 0.32–0.91), = 0.022]. Non-Swedish was no longer significantly negatively associated with A/H1N1-HAAb [OR = 0.52 (95% CI: 0.26–1.04), = 0.065] compared with Swedish patients. Adding islet autoantibodies, as a group or separately, to the model did not change the outcome of the above analyses.

Number of new patients before, during and after vaccination

The number of new patients expressed as patients per number of days per 105 1-to 18-year-olds was 34 (204 days) before, 44 (171 days) during and 36 (275 days) after the vaccination (= 0.0008 before compared with during vaccination and P = 0.0054 during compared with after vaccination) (Table 1). However, comparing the number of new patients during the winter seasons (October–March) 2005–2012, there was no indication of a major increase in the 2009/2010 winter season (Table 3).

Table 3. Number of new T1D patients October to March (winter season) 2005–2012
Winter season05/0606/0707/0808/0909/1010/1111/12
  1. Number of new patients (patients/6 months/105) during winter seasons 2005 until 2012. There are no obvious differences between the different winter seasons. Data were obtained from the SWEDIABKIDS register.

Number of new patients19.418.820.320.221.220.820.4

Discussion

Our results indicate that the Swedish A(H1N1)pdm09 vaccination campaign in conjunction with the A(H1N1)pdm09 pandemic had subtle effects on children and young adults who were diagnosed with T1D during and after vaccination. First, before vaccination, only two T1D children had A/H1N1-HAAb. This may be explained by the observation that the pandemic did not peak until November 2009 [3]. Second, the highest frequencies and levels of A/H1N1-HAAb found during vaccination among 6 < 13-year-olds suggest that children in this age group were good responders to the vaccine and also maintained high prevalence of A/H1N1-HAAb with elevated levels after the vaccination period. It cannot be excluded that some of these children developed A/H1N1-HAAb because they had been infected. Third, the lowest frequencies of A/H1N1-HAAb in the very young (<3 years) during vaccination may have several explanations. In calculating the vaccination frequency in Sweden (11/21 counties see 'Materials and methods'), children <3 years had a vaccination coverage of 60% and a second dose frequency of 57% compared with 80% and 72%, respectively, in the 3 < 13-year-olds. These differences in vaccination coverage were not as large as the differences in A/H1N1-HAAb frequencies and therefore unlikely to be the only explanation to the low frequency of A/H1N1-HAAb in the very young. It is possible that older children could have been more exposed to different strains of influenza viruses. Our data therefore suggest that children <3 years of age were low responders to Pandemrix® as detected in our immunoprecipitation A/H1N1-HAAb assay. Fourth, the lower levels and frequencies in the patients who were diagnosed with T1D after vaccination were expected as these patients were vaccinated a month or longer before their diagnosis. Indeed, information from the SICDC on the official vaccination frequencies shows that most vaccinations were executed before the end of January 2010. Among newly diagnosed T1D patients, it can therefore be concluded that levels of A/H1N1-HAAb after vaccination were not sustained except in the 6 < 13-year-olds. This conclusion is consistent with the observations that age (Fig. 5), HLA-DQ (Fig. 7) and origin seemed to contribute to low-level A/H1N1-HAAb.

The finding that the DQ2/X genotype was associated with somewhat lower frequency and levels of A/H1N1-HAAb in newly diagnosed T1D is a novel finding. Our previous analysis of BDD patients demonstrated that DQ2/X was unevenly distributed between Swedish and non-Swedish patients [22]. Therefore, we have to consider the possibility that non-Swedish patients might not have been vaccinated to the same extent as Swedish. It is interesting to note that GADA in newly diagnosed T1D patients was associated with DQ2 [22, 28], IA-2A and IAA with DQ8 [28] and ZnT8A with both DQ8 and DQ6.4 [12, 24]. GADA was associated with DQ2 in the general population [29]; however, the possible association between A/H1N1-HAAb and HLA-DQ in the Swedes remains to be determined.

The interpretation of our results is complicated by the complex aetiology and pathogenesis of T1D. At present, T1D is viewed as a two-step disease. In the genetically at-risk (DQ2, 8 or both), the first step is that islet autoimmunity reflected by the appearance of one or several islet autoantibodies is triggered, probably by environmental factor(s) [reviewed in [30, 31]]. Islet autoantibodies may be triggered and detected from about 1 year of age [11]. Several mechanisms have been proposed to explain the triggering of islet autoimmunity by infectious agents [15, 16]. Findings support the development of islet autoimmunity after enterovirus infection [32]. Therefore, the first question was whether A(H1N1)pdm09 infection or vaccination triggered islet autoantibodies resulting in an immediate or rapid onset of diabetes. The present data in newly diagnosed T1D patients do not allow us to answer this question, as our BDD patients were not identified until they developed diabetes.

The second step in the T1D pathogenesis is the progression to clinical onset of diabetes in children with islet autoantibodies. Several studies of children indicate that the second step can be anything from months to several years [9-11]. Recent studies demonstrated an increased risk of clinical onset in children with islet autoimmunity if infected with enterovirus [33]. The second question was therefore whether A(H1N1)pdm09 infection or vaccination accelerated the disease process in islet autoantibody-positive children. The observation that both GADA and ZnT8QA increased during as well as after (only ZnT8QA) vaccination may indicate an accelerated disease process. The interpretation of these results is complicated by the observations that GADA was found to vary with season (higher in the winter) and that ZnT8QA differed over the winter seasons of 2005–2010. The difference in the number of new patients diagnosed during and before/after vaccination (Table 1) may be taken as acceleration; however, this conclusion is complicated by reports demonstrating higher incidence rates during winter seasons [34, 35].

Viral infections have in this regard also been discussed to be protective for the development of T1D [15, 16]. An equally important question was therefore whether an A(H1N1)pdm09 infection or vaccination decelerated the second step disease process. The proposed mechanisms involve induction of apoptosis of auto-aggressive lymphocytes and enhancement of counter-regulatory processes [15, 16]. The unexpected decrease in the proportion of DQ2/8 children <3 years after vaccination would be consistent with deceleration (Fig. 6). The DQ2/8 genotype imposes the highest genetic risk and is the most prevalent genotype among the youngest children developing T1D [13]. We therefore speculate that the Pandemrix® vaccination delayed the clinical onset in the youngest DQ2/8 positive. This possible effect may not be dependent on A/H1N1-HAAb levels and the response to virus or vaccine but rather to the AS03 adjuvant used in Pandemrix®. The possibility that adjuvant such as alum or squalene (present in Pandemrix®) may affect the second step in the disease process cannot be excluded.

The results of this study differed from other studies on immunogenicity in younger children [36-38], who found that half the adult dose with AS03 adjuvanted vaccine was enough to provoke an adequate antibody response. Besides, a serologic study made in Sweden [3] showed that as many as 76% of 3-to 14-year-olds were positive for A/H1N1-HAAb in May 2010. However, these studies used hemagglutination inhibition or microneutralization for A/H1N1-HAAb, which is different from the RBA employed in this study. Furthermore, the study population differed as our study solely concerned newly diagnosed T1D patients. It is therefore possible to speculate that children with subclinical islet autoimmunity responded differently to the vaccine than healthy children in the general population.

Our study included a large number of population-based patients with a high degree of ascertainment [24, 39]. A distinct weakness was that there was no information if, or when, the individual child had received the Pandemrix® vaccine. Also, it was not possible to establish whether children with A/H1N1-HAAb had been infected by A(H1N1)pdm09, vaccinated or both. The Swedish A(H1N1)pdm09 vaccination campaign was a unique event in the recent history of Swedish public health. The magnitude of the effort and the extensive coverage of the population made it possible that children with subclinical islet autoimmunity may have been affected.

Conclusion

Among patients recently diagnosed with T1D, children <3 years of age and possibly DQ2/X patients had a low frequency of A/H1N1-HAAb and may have responded poorly to the Pandemrix® vaccine. The T1D high-risk DQ2/8 patients <3 years were fewer than expected after vaccination, and the frequencies and levels of GADA and ZnT8QA were increased. Even though the number of new patients seems not to have been affected, it cannot be excluded that the Swedish A(H1N1)pdm09 vaccination campaign affected levels and frequencies of the islet autoantibodies known to herald the clinical onset of T1D. It can therefore not be excluded that the vaccination campaign may have delayed the clinical onset in young high-risk children but induced an earlier diagnosis in others.

Acknowledegment

The authors are indebted to Hamideh Rastkhani, Mea Pelkonen, Ali Shalouie, Britt Bruveris Svenburg, Jennifer Hemmendal, Ida Jönsson, Annika Winkvist, Barbro Gustavsson, Rasmus Håkansson, Theodosia Massadakis, Qefsere Brahimi and Ingrid Wigheden of the BDD laboratory at Lund University/CRC at the Skåne University Hospital Malmö for expert technical assistance.

Author contributions

M.S. researched the data and wrote the manuscript. A.N. researched, developed and wrote the A(H1N1)pdm09 method, and contributed to the manuscript. A-L.N. contributed to the manuscript. A.D. researched the data and contributed to the statistical analyses and to the manuscript. H.E.-L., A.C., G.F., S.A.I., J.L., I.K., C.M., U.S., E.Ö. were responsible for the BDD protocol and study and critically reviewed the manuscript. Å.L. contributed to the design of the study and edited the manuscript together with M.S. Å.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

The Swedish Research Council (54X-15312), Swedish Child Diabetes Foundation, Skåne County Council Research and Development and Lund University Medical Faculty.

Conflict of interest

The authors declare no conflict of interests.

Appendix

Members of the BDD Study Group were as follows: Helena Desaix (Borås), Kalle Snellman (Eskilstuna), Anna Olivecrona (Falun), Åke Stenberg (Gällivare), Lars Skogsberg (Gävle), Nils Östen Nilsson (Halmstad), Jan Neiderud (Helsingborg), Åke Lagerwall (Hudiksvall), Kristina Hemmingsson (Härnösand), Karin Åkesson (Jönköping), Göran Lundström (Kalmar), Magnus Ljungcrantz (Karlskrona), Eva Albinsson (Karlstad), Karin Larsson (Kristianstad), Christer Gundewall (Kungsbacka), Rebecka Enander (Lidköping), Agneta Brännström (Luleå), Maria Nordwall (Norrköping), Lennart Hellenberg (Nyköping), Elena Lundberg (Skellefteå), Henrik Tollig (Skövde), Britta Björsell (Sollefteå), Björn Rathsman (Stockholm/Sacchska), Torun Torbjörnsdotter (Stockholm/Huddinge), Björn Stjernstedt (Sundsvall), Nils Wramner (Trollhättan), Ragnar Hanås (Uddevalla), Ingemar Swenne (Uppsala), Anna Levin (Visby), Anders Thåström (Västervik), Carl-Göran Arvidsson (Västerås), Stig Edvardsson (Växjö), Björn Jönsson (Ystad), Torsten Gadd (Ängelholm), Jan Åman (Örebro), Rein Florell (Örnsköldsvik) and Anna-Lena Fureman (Östersund).

Ancillary