Month of birth is associated with multiple sclerosis but not with HLA-DR15 in Finland

Abstract

Background: Spring birth has been found to be associated with increase in the risk of MS. However, this effect is reportedly found particularly in HLA-DR15 positive patients.

Objective: Examine the month of birth (MOB) effect in Finnish MS patients and its association with HLA-DR15.

Methods: Using the national register on hospitalizations during 1979–2004 we identified 8359 MS patients. We used a χ²-analysis to compare the distribution of MOB in MS patients versus age-adjusted controls. Patients were subdivided into 10-year birth cohorts to analyse temporal variations in MOB. A subgroup of HLA-typed MS patients (n = 657) and a population sample of healthy donors from the Finnish Bone Marrow Donor Registry (n = 19,805) were included to analyse connections between HLA-DR15 and MOB.

Results: An April birth was associated with a 9.4% increased risk of MS (p = 0.008), whereas November birth was associated with an 11.1% decreased risk (p = 0.004) compared with the expected birth distribution. No temporal clustering in the MOB effect was found during 1900–1988. Both HLA-DR15 positive (34% increase, p = 0.043) and DR15 negative (48% increase, p = 0.0095) MS patients exhibited an increase in spring births (in May). HLA-DR15 in healthy subjects was not associated with MOB.

Conclusions: We confirm that spring birth increases MS risk. This phenomenon was not associated with HLA-DR15 and there was no evidence of temporal clustering. The MOB effect suggests that early developmental processes may modulate MS risk.

Keywords

epidemiology HLA multiple sclerosis season of birth vitamin D

Introduction

Multiple sclerosis (MS) is an immune-mediated chronic demyelinating and neurodegenerative disease of the central nervous system. The aetiology of MS is unclear but both genetic and environmental factors are known to be involved. Recent studies in both northern and southern hemispheres have suggested that season or month of birth is associated with the risk of MS, which points to environmental influences during gestation or early life. In the northern hemisphere, MS patients are born more often than expected in the spring months (April–May) and less than expected in November.^1–3 An opposite pattern has been reported in the southern hemisphere with highest MS risk in November–December births and a decreased risk for May–June.⁴ It has been suggested that this month of birth (MOB) effect is found primarily in relapsing–remitting cases⁵ and, more recently, that it is associated with HLA-DR15.⁶ Low maternal UV-light exposure and low vitamin D status during the last trimester has been considered a possible aetiological factor in this MOB effect.⁷ The HLA-DR15 haplotype is thus far the strongest known genetic risk factor for MS^8,9 and, interestingly, a vitamin-D response element has been identified at the promoter of the HLA-DRB1*1501 allele,¹⁰ which is part of the risk haplotype DRB5*0101-DRB1*1501-DQA1*0102-DQB1*0602. It is of additional interest that narcolepsy, another HLA-DR15 haplotype associated disorder (primary risk allele possibly DQB1*0602),¹¹ exhibits a somewhat similar MOB effect (peak in March, nadir in September).¹²

Here, we have performed a Finnish nationwide study on the MOB effect in MS. Finland is well suited for such a study for several reasons. First, the incidence and prevalence of MS is high in Finland.^13,14 Second, the public health-care system enables us to capture virtually all patients through nationwide hospital registers.¹⁴ Third, Finland is ethnically and culturally relatively homogeneous. Fourth, Finland is situated between approximately 60°N and 70°N latitude. Due to the northern location and low winter sun exposure, vitamin-D deficiency is common in Finland.^15,16 We have analysed whether the reported MOB effect is found in the Finnish population, whether this effect is found particularly in HLA-DR15 positive patients as recently proposed⁶ and whether HLA-DR15 per se is associated with MOB given that two DR15-associated disorders exhibit such an association.

Materials and methods

MS patients were identified from the nationwide Hospital Discharge Register maintained by the National Research and Development Centre for Welfare and Health (abbreviation STAKES in Finnish; http://www.stakes.fi/EN/index.htm). The Hospital Discharge Register collects information about diagnoses on admission. We collected diagnoses for the years 1979–2004. International Classification of Diseases, ICD-8 (until 1987), ICD-9 (1987–95), and ICD-10 (1996 onwards) were used. To avoid false diagnoses (e.g. due to clerical errors or wrong initial suspicions) we used the following criteria for case ascertainment: 1. cases that occurred in the register ≥ 2 times with a diagnosis of MS; 2. MS-diagnosis preceded by a diagnosis consistent with a clinically isolated syndrome such as optic neuritis, myelitis or other demyelinating disorder. The latter included the following ICD-codes: 341.08 and 367.0* (ICD-8), 3418X and 3773* (ICD-9), G37.* and H46 (ICD-10). There were 70,589 hospitalizations with at least one of the above diagnoses during 1979–2004. The number of subjects hospitalized was 12,439, of whom 8359 fulfilled the above case ascertainment criteria for MS (5411 females and 2948 males, female/male ratio 1.835). The age of inclusion (second appearance of MS-diagnosis in the register) was 16–60 years in 93.9% of cases, while 0.9 % were < 16 years and 5.2% > 60 years of age. Social security numbers were used as an identity code for hospitalizations in each MS patient. Every person in Finland has a unique social security number, which is in the format ddmmyycnnnx, where the first six numbers refer to birthday, c corresponds to the century of birth, nnn represents an identity number (which is an odd number in males, even number in females) and x (a letter or a number) is a check mark. Permission for data collection was obtained from the Ethical Committee of the Hospital District of Helsinki and Uusimaa and from STAKES. The study protocol was also reviewed by The Office of the Data Protection Ombudsman.

The control group comprised all recorded births in Finland from 1900 to 1988 (7,014,435). Monthly birth records were obtained from the Finnish statistical centre (Statistics Finland). MS patients were removed from the control population. There were relatively more subjects in early years (1900–1940) as well as in later years (1960–88) in the control population than in the MS population. Since MOB effect may be influenced by year of birth we adjusted the control population so that the year of birth distribution was matched to the MS population. This adjustment was accomplished by applying a correction factor for each year. The relative contribution of each year was thus similar in the MS and control populations.

The influence of HLA-DR15 on MOB was analysed in two datasets. First, we analysed a previously HLA-typed set of 657 MS patients who had taken part in MS genetics projects.^9,17,18 In the Finnish population HLA-DR15 is associated with MS with an odds ratio (OR) in the range of 3–4, which is similar to other populations of Northern European descent. In a recent genome-wide association study (GWAS) using an HLA-DR15 tagging single nucleotide polymorphism (SNP) as a marker the OR was 3.03.¹⁹ In the present study the OR is 4.01 when the HLA-typed patients are compared with the HLA-typed bone marrow donors (see below). Here, HLA typing was performed by sequence-based oligonucleotide probes¹⁷ in 303 MS patients and by HLA-DR15 allele-specific PCR²⁰ in 360 subjects. As a quality-control measure 179 patients were typed by both methods. Six samples (3.4%) exhibited contradictory results in the two assays; these subjects were excluded from the analysis. The patients with definite MS were selected by neurologists in Helsinki, Turku, Tampere, Kuopio, Seinäjoki and Oulu. These patients were born in years 1940–1983, female/male ratio was 2.13 in the total population, 2.35 in HLA-DR15+, and 1.90 in the HLA-DR15- group. HLA-DR15 frequencies in the year-of-birth quartiles 1940–50 (n = 103), 1951–61 (n = 293), 1962–72 (n = 220) and 1973–83 (n = 41) were 48.5%, 54.9%, 57.7% and 56.1%, respectively. The control population comprised all recorded births from 1940 to 1983 (3,461,242). The second dataset was used to test the influence of HLA-DR15 per se on MOB effect and was obtained from the Finnish Bone Marrow Donor Registry.²¹ This data comprised 19,805 healthy individuals, born between 1951 and 1988. Serological HLA class II (DR) typing was performed with the complement-mediated lymphocytotoxicity test on lymphocytes until the year 2003 (typing sera from Pel-Freez Inc. and Biotest Inc.). From 2004 onwards, all donors were typed with sequence-based oligonucleotide probes.¹⁷ The control population comprised all recorded births from 1951 to 1988 (2,780,905).

Results

As shown in Table 1, 9.4 % more MS patients were born in April than expected (χ² = 7.10, p = 0.0077). Conversely, 11.1 % fewer MS patients were born in November than expected (χ² = 8.36, P = 0.0039). These differences were contributed to by both sexes. p-values for April were 0.056 and 0.051, and for November, 0.032 and 0.049, in males and females respectively. These results confirm previous reports,^2,3 indicating increased MS births in spring (April–May) and decreased births in autumn (November). Female/male ratio was 1.785 in patients born in April and 1.927 in patients born in November; not a significant difference (χ² = 0.43, 1 df, p = 0.51). We also observed decreased births (by 9.1%) in January (χ² = 6.33, p = 0.012), a novel finding which was not significant after Bonferroni correction (p_c12 = 0.144) (Table 1).

Table 1.

Observed and expected numbers of monthly births in MS patients as compared with age-adjusted standard population.

Month	Observed	Expected	Observed/expected	χ²	p-value
Jan	640	704	0.91	6.326	0.012
Feb	666	655	1.02	0.120	0.655
Mar	772	749	1.03	0.7664	0.381
April	791	723	1.09	7.101	0.0077
May	750	738	1.02	0.1996	0.6550
June	750	721	1.04	1.198	0.274
July	740	734	1.01	0.041	0.840
Aug	684	704	0.97	0.646	0.422
Sept	691	693	1.00	0.0079	0.929
Oct	622	646	0.96	0.984	0.321
Nov	559	629	0.89	8.360	0.0039
Dec	694	660	1.05	1.813	0.178

In order to find out whether the MOB effect was the result of a few outlying decades, we subdivided the material into nine birth cohorts (Table 2). Increased April births were seen in all except one (1910–1919) of the nine birth cohorts (Table 2). Decreased November births were found in the majority of the birth cohorts (in seven of the nine cohorts). April/November ratios were consistently above that observed in the control population (Table 2). This descriptive data suggests that the overall effect in April and November is a consistent phenomenon and is not caused by just a few outlier decades.

Table 2.

Temporal variation in MS patients’ month of birth during 1900–1988.

Decade (% of data)	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	Apr/Nov
1900–09 (1.5%)	0.51	0.81	0.63	1.24	1.26	0.99	0.93	0.73	0.42	1.14	0.92	1.47	2.00 (1.35)
1910–19 (4.1%)	0.80	1.15	1.26	1.00	1.12	1.06	1.24	0.66	0.84	1.06	0.65	1.14	1.31 (1.53)
1920–29 (9.6%)	0.86	0.99	1.10	1.08	1.05	1.10	0.95	0.96	1.02	0.90	1.00	1.00	1.54 (1.08)
1930–39 (14.0%)	0.80	1.06	1.18	1.07	0.96	0.97	1.05	1.00	1.03	0.87	1.00	1.05	1.34 (1.07)
1940–49 (23.1%)	0.97	0.94	0.92	1.14	1.00	1.03	1.09	1.10	0.95	0.95	0.83*	1.06	1.49 (1.37)
1950–59 (21.9%)	0.83*	0.94	1.07	1.06	1.17*	0.93	0.92	1.02	1.08	1.07	0.94	1.00	1.30 (1.13)
1960–69 (16.7%)	1.06	1.18	1.96	1.13	0.90	1.09	0.93	0.96	1.01	0.83	0.86	1.10	1.32 (1.36)
1970–79 (7.6%)	1.02	1.02	0.95	1.02	0.89	1.21	1.14	0.66**	0.87	1.19	0.94	1.10	1.65 (1.09)
1980–88 (1.5%)	1.06	1.20	1.24	1.19	0.86	1.46	0.46	1.23	0.85	0.68	0.72	0.99	1.44 (1.65)
SD	0.17	0.12	0.34	0.07	0.13	0.15	0.21	0.19	0.19	0.15	0.11	0.14	0.21 (0.21)
1900–88 (100%)	0.91*	1.02	1.03	1.09**	1.02	1.04	1.01	0.97	1.00	0.96	0.89**	1.05	1.42 (1.22)

Observed/expected ratio is given for each decade and month. Additionally, April/November ratios are given for each decade as absolute ratios and conditioned to the control population ratios (in brackets). Conditioned ratios were calculated as ratio (MS)/ratio (controls).

SD: standard deviation, Apr: April, Nov: November

p < 0.05

p < 0.01

To test the effect of HLA-DR15, we analysed a subpopulation of Finnish definite MS cases (n = 657) that were previously HLA-typed.^9,17 Among this subgroup of patients an increased number of births was found in May (Table 3, 40% increase, χ² =10.49, p=0.0012), whereas the November effect was not significant (Table 3, 4% decrease, χ² = 0.064, p = 0.80). Of these patients 361 (55%) were HLA-DR15 positive. Both HLA-DR15+ and DR15- subgroups exhibited a significant increase in May births. In DR15+ patients there was a 34% increase (Table 3, χ² = 4.08, p = 0.043) and in DR15- patients there was a 48% increase in May births (Table 3, χ² = 6.73, p = 0.0095). November births did not deviate significantly from the expected figures in either HLA-DR15+ or HLA-DR15- patients (Table 3).

Table 3.

Month of birth effect in HLA-typed MS-patients compared with the birth-year matched controls.

Month	All observed number of births (n = 657)	All obs/exp births	All p-value	DR15+ observed no. of births (n = 361)	DR15+ Obs/exp births	DR15+ p-value	DR15- Observed no. of births (n = 296)	DR15-Obs/exp births	DR15-p-value
Jan	56 (8,5%)	1.03	0.82	33 (9,1%)	1.10	0.55	23 (7,8%)	0.94	0.75
Feb	51 (7.8%)	0.99	0.93	26 (7,2%)	0.92	0.65	25 (8,5%)	1.07	0.70
Mar	59 (9.0%)	0.99	0.91	36 (10,0%)	1.10	0.57	23 (7,8%)	0.85	0.42
Apr	65 (9.8%)	1.12	0.32	32 (8,9%)	1.01	0.96	33 (11,2%)	1.27	0.15
May	82 (12,5%)	1.40	0.0012*	43 (11,9%)	1.34	0.043	39 (13,2%)	1.48	0.0095
Jun	42 (6,4%)	0.75	0.048	23 (6,4%)	0.75	0.14	19 (6,4%)	0.75	0.19
Jul	62 (9,4%)	1.11	0.44	36 (10,0%)	1.16	0.35	26 (8,8%)	1.02	0.90
Aug	40 (6,1%)	0.73	0.037	25 (6,9%)	0.83	0.33	15 (5,1%)	0.61	0.042
Sept	52 (7,9%)	0.95	0.72	30 (8,3%)	1.00	1.00	22 (7,4%)	0.89	0.59
Oct	49 (7,5%)	0.95	0.72	27 (7,5%)	0.95	0.80	22 (7,4%)	0.95	0.80
Nov	48 (7,3%)	0.97	0.80	24 (6,6%)	0.88	0.51	24 (8,1%)	1.07	0.73
Dec	51 (7,8%)	0.98	0.89	26 (7,2%)	0.91	0.62	25 (8,5%)	1.07	0.73

P = 0.014 with Bonferroni correction

May/Nov ratios: All, 1.71; DR15+, 1.79; DR15-, 1.63; Apr/Nov ratios: All 1.35, DR15+ 1.33, DR15- 1.38.

To test the effect of HLA-DR15 per se on MOB, we analysed a population sample of 19,805 HLA-typed apparently healthy stem cell donors (see Methods). In this material there were 4619 (23.3%) HLA-DR15+ individuals. We did not find any significant (Bonferroni corrected) effects of HLA-DR15 on MOB (Table 4). We compared the MOB effect in DR15+ patients (n = 361) versus DR15+ controls (n = 4691) and DR15- patients (n = 296) versus DR15- controls (n = 15,114). A non-significant increase in MS patients’ May births was found in the DR15+ group (27% increase, χ² = 2.89, 1 df, p = 0.089), while a significant increase was also found in the DR15- group (44% increase, χ² = 6.18, p = 0.013).

Table 4.

Observed and expected numbers of monthly births in HLA-DR15+ stem cell donors as compared with age-adjusted standard population.

Month	Observed	Expected	Observed/expected	χ²	p-value
Jan	360	390	0.92	2.56	0.11
Feb	354	363	0.96	0.23	0.63
Mar	442	416	1.06	1.85	0.17
Apr	425	401	1.06	1.62	0.20
May	425	410	1.04	0.64	0.42
Jun	373	400	0.93	1.93	0.16
Jul	390	406	0.96	0.71	0.40
Aug	353	388	0.91	3.44	0.064
Sept	389	381	1.02	0.16	0.69
Oct	393	356	1.10	4.09	0.043
Nov	336	346	0.97	0.29	0.59
Dec	379	363	1.04	0.75	0.39

Discussion

In this nationwide analysis of 8739 Finnish MS patients we confirmed previous findings of increased spring (April–May) and decreased autumn (November) births. We also analysed temporal variations in the MOB effect and found it relatively consistently across the decades during 1900–1988. To our knowledge the temporal variation of this phenomenon has not been analysed previously. In contrast to one previous report,⁶ we did not find different patterns of MOB effect in HLA-DR15+ and DR15- patients. In the population-based stem cell donor population HLA-DR15 as such did not vary by MOB. This is a relevant result since a MOB effect has been reported in two DR15-associated disorders (MS and narcolepsy),^2,12 and it suggests that DR15 is not a confounder in these findings.

In the main analysis against the population controls we observed similar and significant MOB effects in both DR15+ and DR15- patients. In DR15+ patients there was a 34% increase (p = 0.043,) and in DR15- patients there was a 48% increase in May births (p = 0.0095). In the analysis against the smaller HLA-matched control group increased May births were found in MS patients, but with reduced statistical significance. In the DR15+ group there was a 27% increase (p = 0.089) and in the DR15- group a 44% increase in May births (p = 0.013). In the HLA-typed subpopulation the peak increase in births occurred in May, whereas in the total population it was April. This difference is largely explained by an age effect. In the 1950–59 cohort a relatively prominent increase in May births was observed (Table 2) and this cohort was the largest in the HLA-typed subpopulation.

Our results differ from those reported by Ramagopalan et al. in 4834 HLA-typed MS patients from Canada, Sweden and Denmark.⁶ They reported an increase in April births in HLA-DR15+ patients as compared with HLA-DR15- patients (10.3% vs. 7.8%, p = 0.004) and, conversely, decreased November births in the respective comparison (6.0% vs. 7.9%, p = 0.023). In our dataset both April and May ‘high-season’ births were, on the contrary, decreased in HLA-DR15+ patients as compared with DR15- patients (Table 3). November births did not show significant difference from controls in the HLA-typed MS patients. DR15+ patients had a lower frequency of November births than DR15- patients (6.6% vs. 8.1%), but our dataset is clearly not powered to detect MOB differences in DR15+ versus DR15- patients. A more powerful analysis including > 9,499 patients from 12 countries was recently published.¹⁹ In this analysis a SNP tagging DRB15 showed no evidence for association with MOB (p = 0.6). However, the ethnic background was clearly more heterogeneous than in the Ramagopalan et al.⁶ study, making direct comparison difficult.

These discordant results indicate that the interesting link proposed between the MOB effect and the vitamin-D response element (GGGTGGAGGGGTTGA) at the DRB1*1501 promoter appears uncertain.^6,7,10 Nevertheless, the vitamin-D hypothesis is still valid since Ramagopalan et al. have also demonstrated that vitamin-D receptor binding sites are widely dispersed in the human genome (2776 genomic sites) and have identified at least 229 genes that are regulated by vitamin D.²² Other candidate genes underlying the MOB effect, however, should still be investigated, although not detectable in the recent GWAS.¹⁹ There are several possible explanations for the different results regarding the HLA-DR15 association with MOB. First, it is possible that in the population studied by Ramagopalan et al. HLA-DR15 represented a marker of a genetic subpopulation, where the MOB effect was especially prominent. Such subpopulations may exist since it has been previously found that the MOB effect is stronger in familial than in sporadic cases.² In a more homogeneous population (like in Finland) the HLA-DR15 ancestral haplotype might not be needed to dissect any subpopulation. Second, our Finnish study is the northernmost single population study of the MOB effect, and our northern location may have driven the enrichment of specific D-vitamin responsive alleles other than DR15 (selection effect). Third, it is also possible that the quantitative lack of sunlight is so profound in Finland that we cannot detect an association between HLA-DR15 and MOB in MS (threshold effect).

A somewhat similar MOB effect has been reported in type 1 diabetes in Great Britain²³ and Ukraine,²⁴ where the highest risk was associated with spring births and the lowest with late autumn births. However, this phenomenon has not been demonstrated in other European countries.²⁵ Nutritional factors, vitamin D deficiency and infections were proposed as possible explanations.²⁶ HLA-DR15,DQ6 haplotype is a strong protective factor against type 1 diabetes, while HLA-DR4,DQ8 and DR3,DQ2 are the most common predisposing HLA-haplotypes.²⁷

A relative lack of maternal vitamin D during winter (as well as other sunlight driven factors), within the last months of pregnancy remains an attractive hypothesis for explaining the increased risk of MS in spring births.⁷ Seasonal variations in infection, pollen, diet and life style are obvious alternative hypotheses. The MOB effect is reproducible in different populations and provides support for the concept that very early developmental processes play a role in MS. The unravelling of these processes may offer important new insights into the prevention of MS.

Footnotes

Acknowledgments

We thank Olli Taskinen for help in data management.

Funding

This work was supported by the Finnish Academy, the Sigrid Juselius Foundation, Helsinki University Central Hospital and The Social Insurance Institution of Finland.

Conflict of interest statement

None declared.

References

Templer

Trent

Spencer

. Season of birth in multiple sclerosis. Acta Neurol Scand 1992; 85: 107–109.

Willer

Dyment

Sadovnick

Rothwell

Murray

Ebers

; Canadian Collaborative Study Group. Timing of birth and risk of multiple sclerosis: population based study. BMJ 2005; 330: 120.

Bayes

Weir

O’Leary

. Timing of birth and risk of multiple sclerosis in the Scottish population. Eur Neurol 2009; 63: 36–40.

Staples

Ponsonby

Lim

. Low maternal exposure to ultraviolet radiation in pregnancy, MOB, and risk of multiple sclerosis in offspring: longitudinal analysis. BMJ 2010; 340: c1640.

Sadovnick

Duquette

Herrera

. A timing-of-birth effect on multiple sclerosis clinical phenotype. Neurology 2007; 69: 60–62.

Ramagopalan

Link

Byrnes

. HLA-DRB1 and month of birth in multiple sclerosis. Neurology 2009; 73: 2107–2111.

Handunnetthi

Ramagopalan

Ebers

. Multiple sclerosis, vitamin D, and HLA-DRB1*15. Neurology 2010; 74: 1905–1910.

Olerup

Hillert

. HLA class II-associated genetic susceptibility in multiple sclerosis: a critical evaluation. Tissue Antigens 1991; 38: 1–15.

Lincoln

Montpetit

Cader

. A predominant role for the HLA class II region in the association of the MHC region with multiple sclerosis. Nat Genet 2005; 37: 1108–1112.

10.

Ramagopalan

Maugeri

Handunnetthi

. Expression of the multiple sclerosis-associated MHC class II Allele HLA-DRB1*1501 is regulated by vitamin D. PLoS Genet 2009; 5: e1000369.

11.

Chabas

Taheri

Renier

Mignot

. The genetics of narcolepsy. Annu Rev Genomics Hum Genet 2003; 4: 459–483.

12.

Dauvilliers

Carlander

Molinari

. Month of birth as a risk factor for narcolepsy. Sleep 2003; 26: 663–665.

13.

Sumelahti

Tienari

Wikström

. Regional and temporal variation in multiple sclerosis incidence in Finland during 1979–93. Neuroepidemiology 2000; 19: 67–75.

14.

Sumelahti

Tienari

Wikström

. Increasing prevalence of multiple sclerosis in Finland. Acta Neurol Scand 2001; 103: 153–158.

15.

Kokkonen

Koivisto

Kirkinen

. Seasonal variation in serum-25-OH-D3 in mothers and newborn infants in northern Finland. Acta Paediatr Scand 1983; 72: 93–96.

16.

Lamberg-Allardt

Ala-Houhala

Ahola

Parviainen

Räsänen

Visakorpi

. Vitamin D status of children and adolescents in Finland. Ann Nutr Metab 1986; 30: 267–272.

17.

Laaksonen

Pastinen

Sjöroos

. HLA class II associated risk and protection against multiple sclerosis – a Finnish family study. J Neuroimmunol 2002; 122: 140–145.

18.

Bonetti

Koivisto

Pirttilä

. A follow-up study of chromosome 19q13 in multiple sclerosis susceptibility. J Neuroimmunol 2009; 208: 119–124.

19.

International Multiple Sclerosis Genetics Consortium, Wellcome Trust Case Control Consortium 2, Sawcer

. Genetic risk and a primary role for cell-mediated immune mechanisms in multiple sclerosis. Nature 2011; 476: 214–219.

20.

Westman

Kuismin

Partanen

Koskimies

. An HLA-DR typing protocol using group-specific PCR-amplification followed by restriction enzyme digests. Eur J Immunogenet 1993; 20: 103–109.

21.

Sirén

Sareneva

Lokki

Koskimies

. Unique HLA antigen frequencies in the Finnish population. Tissue Antigens 1996, 48: 703–707.

22.

Ramagopalan

Heger

Berlanga

. A ChIP-seq defined genome-wide map of vitamin D receptor binding: associations with disease and evolution. Genome Res 2010; 20: 1352–1360.

23.

Rothwell

Staines

Smail

. Seasonality of birth of patients with childhood diabetes in Britain. BMJ 1996; 312: 1456–1457.

24.

Vaiserman

Carstensen

Voitenko

. Seasonality of birth in children and young adults (0–29 years) with type 1 diabetes in Ukraine. Diabetologia 2007; 50: 32–35.

25.

Rothwell

Gutnikov

McKinney

Schober

Ionescu-Tirgoviste

Neu

. Seasonality of birth in children with diabetes in Europe: multicentre cohort study. BMJ 1999; 319: 887–888.

26.

Åkerblom

Vaarala

Hyöty

. Environmental factors in the etiology of type 1 diabetes. Am J Med Genet 2002; 115: 18–29.

27.

Thorsby

Rønningen

. Role of HLA genes in predisposition to develop insulin-dependent diabetes mellitus. Ann Med 1992; 24: 523–531.