Diabetologia

, 49:1785

HLA-DQB1 genotypes, islet antibodies and beta cell function in the classification of recent-onset diabetes among young adults in the nationwide Diabetes Incidence Study in Sweden

Authors

  • E. Bakhtadze
    • Department of Clinical Sciences Malmö, Division of Endocrinology and DiabetesLund University, Malmö University Hospital
  • H. Borg
    • Department of Clinical Sciences Malmö, Division of Endocrinology and DiabetesLund University, Malmö University Hospital
  • G. Stenström
    • Department of Medicine, Kungsbacka HospitalGöteborg University
  • P. Fernlund
    • Clinical Chemistry, Department of Laboratory Medicine MalmöLund University, Malmö University Hospital
  • H. J. Arnqvist
    • Department of Medicine and CareLinköping University
  • A. Ekbom-Schnell
    • Department of MedicineUppsala Academic Hospital
  • J. Bolinder
    • Department of MedicineKarolinska University Hospital Huddinge
  • J. W. Eriksson
    • Department of MedicineUmeå University Hospital
  • S. Gudbjörnsdottir
    • Department of MedicineSahlgrenska University Hospital
  • L. Nyström
    • Department of MedicineUmeå University Hospital
  • L. C. Groop
    • Department of Clinical Sciences Malmö, Division of Endocrinology and DiabetesLund University, Malmö University Hospital
    • Department of Clinical Sciences Malmö, Division of Endocrinology and DiabetesLund University, Malmö University Hospital
Article

DOI: 10.1007/s00125-006-0293-5

Cite this article as:
Bakhtadze, E., Borg, H., Stenström, G. et al. Diabetologia (2006) 49: 1785. doi:10.1007/s00125-006-0293-5

Abstract

Aims/hypothesis

The World Health Organization considers an aetiological classification of diabetes to be essential. The aim of this study was to evaluate whether HLA-DQB1 genotypes facilitate the classification of diabetes as compared with assessment of islet antibodies by investigating young adult diabetic patients.

Subjects and methods

Blood samples were available at diagnosis for 1,872 (90%) of the 2,077 young adult patients (aged 15–34 years old) over a 5-year period in the nationwide Diabetes Incidence Study in Sweden. Islet antibodies were measured at diagnosis in 1,869 patients, fasting plasma C-peptide (fpC-peptide) after diagnosis in 1,522, while HLA-DQB1 genotypes were determined in 1,743.

Results

Islet antibodies were found in 83% of patients clinically considered to have type 1 diabetes, 23% with type 2 diabetes and 45% with unclassifiable diabetes. After diagnosis, median fpC-peptide concentrations were markedly lower in patients with islet antibodies than in those without (0.24 vs 0.69 nmol/l, p<0.0001). Irrespective of clinical classification, patients with islet antibodies showed increased frequencies of at least one of the risk-associated HLA-DQB1 genotypes compared with patients without. Antibody-negative patients with risk-associated HLA-DQB1 genotypes had significantly lower median fpC-peptide concentrations than those without risk-associated genotypes (0.51 vs 0.74 nmol/l, p=0.0003).

Conclusions/interpretation

Assessment of islet antibodies is necessary for the aetiological classification of diabetic patients. HLA-DQB1 genotyping does not improve the classification in patients with islet antibodies. However, in patients without islet antibodies, HLA-DQB1 genotyping together with C-peptide measurement may be of value in differentiating between idiopathic type 1 diabetes and type 2 diabetes.

Keywords

HLA-DQB1 genotypesClassificationC-peptideIslet antibodies

Abbreviations

Ab+

positive for islet antibodies

Ab−

negative for islet antibodies

DASP

Diabetes Autoantibody Standardization Program

DISS

Diabetes Incidence Study in Sweden

fpC-peptide

fasting plasma C-peptide

GADA

GAD antibodies

IA-2A

protein tyrosine phosphatase-like protein antibodies

ICA

islet cell antibodies

JDF

Juvenile Diabetes Foundation

OR

odds ratio

Introduction

Type 1 and type 2 diabetes are different in terms of aetiology and clinical course. However, if only a clinical assessment is used for diagnosis, it is difficult to distinguish the two main types of diabetes from each other [14]. Among patients with the type 2 diabetes phenotype, depending on age at onset, 8–30% have islet antibodies, indicating that the correct diagnosis would be autoimmune type 1 diabetes [57]. Besides autoimmune markers, certain HLA genotypes confer increased risk of type 1 diabetes [810]. Compared with islet antibodies and fasting plasma C-peptide (fpC-peptide) concentration as a measure of beta cell function, the value of type 1 diabetes-associated HLA genotypes in the classification of diabetes among young adults has not been established.

The aim of this study was to evaluate the diagnostic value of HLA-DQB1 genotypes in the classification of diabetes as compared with islet autoantibodies and fpC-peptide among young adults, using a 5-year cohort of incident diabetic patients in the Diabetes Incidence Study in Sweden (DISS).

Subjects and methods

Subjects

Over a 5-year period (1 Jan 1998 to 31 December 2002), 2,077 young adult diabetic patients (15–34 years of age) were reported to the DISS. Type of diabetes, age, sex, height, body weight, symptoms at diagnosis, duration of symptoms, date of diagnosis, family history of diabetes and blood glucose values at diagnosis were recorded on a special form by the reporting physician. The patients were invited to donate a blood sample for determination of islet antibodies (islet cell antibodies [ICA], GAD antibodies [GADA], protein tyrosine phosphatase-like protein antibodies [IA–2A]) and HLA-DQB1 genotypes at diagnosis, and a blood sample for determination of fpC-peptide after diagnosis (median=4 months, Q1=3 months, Q3=6 months). Blood samples were available for 1,872 (90%) of the 2,077 patients. Islet antibodies were measured in 1,869, fpC-peptide in 1,522, while HLA-DQB1 genotypes were determined in 1,743 patients. Figure 1 shows a flow chart of the different assessments conducted.
https://static-content.springer.com/image/art%3A10.1007%2Fs00125-006-0293-5/MediaObjects/125_2006_293_Fig1_HTML.gif
Fig. 1

Flow chart of DISS patients recruited during the 5-year study period (1998–2002 inclusive). The upper part of the figure shows the 2,077 incident cases of young adult diabetic patients (aged 15–34 years) included in the study, with all available collected clinical and biological data. The lower part of the figure shows the 2,018 clinically classified diabetic patients and collected data on HLA-DQB1 genotypes, islet antibodies and follow-up fpC-peptide

Using a computer-based patient administrative register as a second source, the level of ascertainment in DISS during 1983–1987 was estimated at 78% in women and at 79% in men in the two southernmost counties, covering 9.2% of the population at risk. For clinical type 1 diabetes the level was 86% [11]. A similar study in the county of Västerbotten in northern Sweden, covering 2.9% of the population at risk, found no trend in the level of ascertainment during 1986–1997. The level of ascertainment for clinical type 1 diabetes was 91%. Furthermore, the median level of ascertainment for clinical type 1 diabetes at six diabetic clinics continuously assessed using DIABASE software (Kungälv, Sweden) was 82% for their patients. Therefore, the ascertainment rate in DISS during the years of study seems to be constant. Level of ascertainment was assessed using the two-sample capture-recapture method [11]. The ethics committee at the Karolinska Institute (Stockholm) approved the study, which was conducted after the patients had given informed consent.

HLA-DQB1 genotyping

Using a primer pair with biotinylated 3′ primers, the 158-bp second exon of HLA-DQB1 was amplified by PCR. The amplification product was bound to streptavadin-coated microtitration plates and denatured with NaOH. After washing, bound DNA was assessed using two different hybridisation mixtures with lanthanide (III) chelate-labelled DNA probes specific for the HLA-DQB1 alleles. One mixture contained europium (Eu)-labelled internal reporter probe for DQB1*0602 and *0603 alleles (*0602–*0603), samarium (Sm)-labelled probe for *0603 and *0604 alleles (*0603–*0604), and terbium (Tb)-labelled consensus sequence-specific probe (Tb-DQB1 control) as a control for PCR amplification. The other mixture contained Tb-, Sm- and Eu-labelled probes specific for DQB1*0201, *0301 and *0302 alleles, respectively. To measure probe hybridisation, microtitration plates were evaluated by time-resolved fluorescence (Delfia Research Fluorometer; Wallac OY, Turku, Finland). Different emission wavelengths and delay times were used to distinguish the signals of each lanthanide label [12]. Volunteers without diabetes (n=216) from the county of Skaraborg, Sweden were used as control subjects for HLA-DQB1 genotyping. In control subjects, the HLA-DQB1 locus was amplified by PCR, followed by dot blotting onto nitrocellulose filters, hybridisation using the radioactively end-labelled sequence-specific oligonucleotide probes, and autoradiography [13].

ICA

ICA were determined by a prolonged two-colour immuno-fluorescence assay [14]. The detection limit for ICA was four Juvenile Diabetes Foundation (JDF) units for the first pancreas used in samples tested up to April 1999, and five JDF units for the second pancreas used in samples tested from April 1999 and onwards. In the last ICA Proficiency Test (13th), our ICA assay performed with 100% sensitivity and 100% specificity (ICA is not included in the Diabetes Autoantibody Standardization Program [DASP]).

GADA

GADA were measured by a radioligand binding assay based on human 35S-labelled recombinant GAD65 [15]. The results are presented as an index: GADA index =100×(un)/(pn), where u is the cpm (mean activity of all four measurements for a sample) of the unknown sample, n is cpm of the negative control, and p is cpm of the positive control. A GADA index >4.6 was considered positive (97.5 percentile of 165 non-diabetic control subjects aged 7–34 years). In the first DASP (in 2000), our GADA assay showed a sensitivity of 80% and a specificity of 96%; in the second (in 2002), a sensitivity of 88% and a specificity of 87%; and in the third DASP (in 2003), a sensitivity of 82% and a specificity of 93%.

IA–2A

IA–2A were measured by an assay similar to that for GADA, which was based on human 35S-labelled recombinant IA-2 [16]. An IA–2A index >1.0 was considered positive (97.5 percentile of 165 non-diabetic controls aged 7–34 years). In the first DASP (in 2000), our IA–2A assay showed a sensitivity of 58% and a specificity of 100%; in the second (in 2002), a sensitivity of 62% and a specificity of 100%; and in the third DASP (in 2003), a sensitivity of 64% and a specificity of 100%.

Plasma C-peptide

A RIA was used to determine fpC-peptide. The detection limit was 0.10 and 0.25–0.75 nmol/l was considered the normal range. A fpC-peptide concentration 0.1–0.25 nmol/l was considered low and <0.10 nmol/l as immeasurable (beta cell failure) [17].

Clinical classification

At diagnosis, based on clinical judgment by the reporting physician, the patient was classified as having type 1, type 2 or unclassifiable diabetes.

Statistical analysis

Comparison of genotype frequencies between diabetic patients and control subjects was tested by two-tailed Fisher’s exact test or χ2 test with Bonferroni adjustment (0.05/19 [number of genotypes in the study]) of p values for multiple comparisons (p<0.0026 [0.05/19] was considered significant). Odds ratios (ORs) and 95% CIs were calculated using the formula (a×d)/(b×c), where a is the number of diabetic patients with one of the genotypes, b is the number of control subjects with the corresponding genotype, c is the number of patients without this genotype, and d is the number of control subjects without the corresponding genotype. Differences in continuous variables between groups were assessed by non-parametric Wilcoxon signed-rank test. Regression analysis was used for multiple comparisons. All statistical tests were performed by SPSS (version 11.0; SPSS, Chicago, IL, USA) or JMP (version 5; SAS Institute, Cary, NC, USA) for MAC OS X. In univariate analysis p<0.05 was considered significant, whereas in multivariate analysis p<0.0026 was considered significant. Continuous data are presented as median and 25th and 75th percentiles (Q1 and Q3), and dichotomous data as absolute values and percentage.

Results

Of the 2,077 incident diabetic patients, 2,018 were classified by the reporting physicians: 1,395 (69%) were given a diagnosis of clinical type 1 diabetes, 366 (18%) of clinical type 2 diabetes and 257 (13%) could not be classified (unclassifiable diabetes). There was a clear male preponderance among clinical type 1 diabetic patients (ratio of men:women=1.9) (Table 1). Patients with clinical type 1 diabetes were younger (24 vs 30 and 28 years, respectively, p<0.0001) and had a lower BMI (22 vs 32 and 25, respectively, p<0.0001) than patients with clinical type 2 or unclassifiable diabetes. A family history of diabetes was more frequent in clinical type 2 (53%) than in clinical type 1 (22%, p<0.0001) or unclassifiable (38%, p=0.004) diabetes (Table 1).
Table 1

Characteristics of 2,018 patients in DISS (1998–2002 inclusive) according to clinical classification as provided by the patients to DISS reporting physicians

 

Incident cases (n=2,077)

Clinical type of diabetesa

p value

Type 1 (n=1,395)

Type 2 (n=366)

Unclassifiable (n=257)

Sex, ratio of men:women

1.6

1.9

1.3

1.0

<0.0001

Age at onset (years), median (Q1, Q3)

26 (20, 30)

24 (19, 29)b

30 (26, 32)c

28 (24, 32)

<0.0001

BMI (kg/m2), median (Q1, Q3)

23 (21, 27)

22 (20, 24)d

32 (27, 37)

25 (22, 30)

<0.0001

Family history of diabetese

 Yes

407 (30)

211 (22)b

130 (53)f

66 (38)

<0.0001

 No

971 (70)

746 (78)

116 (47)

107 (62)

 

Treatment regimeng

 Insulin

1694 (83)

1339 (97)d

142 (39)

169 (66)

<0.0001

 No insulin

354 (17)

40 (2.9)

221 (61)

86 (34)

 

Patients with antibodies (Ab+)h

1250 (67)

1056 (83)d

79 (23)

113 (45)

<0.0001

Patients without antibodies (Ab−)

619 (33)

211 (17)

269 (77)

137 (55)

 

No. of antibodies detected among those Ab+

 All three Ab+

  ICA+GADA+IA–2A

591 (47)

504 (48)

32 (41)

55 (49)

 

 Two Ab+

  ICA+GADA

216 (17)

180 (17)

16 (20)

19 (17)

 

  ICA+IA–2A

76 (6.1)

69 (6.5)

5 (6.3)

1 (0.9)

 

  GADA+IA–2A

78 (6.2)

68 (6.4)

6 (7.6)

4 (3.5)

0.36

 One Ab+

  ICA

32 (2.6)

26 (2.5)

3 (3.8)

3 (2.7)

 

  GADA

226 (18)

186 (18)

14 (18)

26 (23)

 

  IA–2A

31 (2.5)

23 (2.2)

3 (3.8)

5 (4.4)

 

Antibody levels

  ICA+, median (Q1, Q3)

58 (26, 130)

58 (26, 130)i

58 (26, 130)

45 (17, 87)

0.03

  GADA+, median (Q1, Q3)

54 (20, 103)

51 (19, 102)

65 (17, 110)

65 (29, 101)

0.16

  IA–2A+, median (Q1, Q3)

92 (30, 115)

95 (36, 115)j

87 (15, 115)

51 (7, 114)

0.04

Follow-up fpC-peptidek, median (Q1, Q3)

0.31 (0.17, 0.59)

0.24 (0.14, 0.40)d

0.74 (0.48, 1.14)

0.48 (0.26, 0.84)

<0.0001

Data are presented as absolute numbers (%) or median, 25th percentile (Q1), and 75th percentile (Q3). The χ2p values for comparison between three groups (type 1, type 2 and unclassifiable) are shown. Using a computer-based patient administrative register as a second source, the level of ascertainment in DISS during 1983–1987 was estimated at 78% in women and at 79% in men in the two southernmost counties, covering 9.2% of the population at risk. aThere were 2,018 clinically classified patients; bp<0.0001 for type 1 vs type 2 and type 1 vs unclassifiable; cp=0.002 for type 2 vs unclassifiable; dp<0.0001 for type 1 vs type 2, type 1 vs unclassifiable, and unclassifiable vs type 2; e1,378 patients had information regarding possible family history of diabetes; fp=0.004 for type 2 vs unclassifiable; g2,048 patients had information regarding treatment; h1,869 patients had islet antibodies; ip=0.008 for type 1 vs unclassifiable; jp=0.01 for type 1 vs unclassifiable; k1522 patients had measured follow-up fpC-peptide; follow-up C-peptide was measured at ≤6 months after diagnosis in 1,522 patients and >6 months in 370 patients

Among all patients, 1,250 (67%) were positive for islet antibodies (Ab+). The prevalence of islet antibodies was significantly higher in clinical type 1 diabetic patients than in unclassifiable patients (83 vs 45%, p<0.0001), who, in turn, showed a higher prevalence of islet antibodies than patients with clinical type 2 diabetes (45 vs 23%, p<0.0001). Irrespective of the clinical classification, almost half of Ab+ individuals had three different antibodies. Among patients positive for ICA, the median ICA titre was significantly higher in patients with clinical type 1 diabetes than in patients with unclassifiable diabetes (58 [Q1=26, Q3=130] vs 45 [Q1=17, Q3=87] JDF units, respectively, p=0.008). Similarly, among IA–2A-positive patients, the median IA–2A titre was significantly higher in type 1 diabetic patients than in those with unclassifiable diabetes (index values 95 [Q1=36, Q3=115] vs 51 [Q1=7, Q3=114], respectively, p=0.01) compared with those with unclassifiable diabetes. Similar differences were not seen for GADA.

HLA-DQB1 genotypes

Table 2 shows that, irrespective of their clinical classification, Ab+ patients had a significantly higher prevalence of risk-associated HLA-DQB1 genotypes. However, among clinical type 1 diabetic patients, some protective and neutral HLA-DQB1 genotypes were significantly more frequent among those who were Ab− than those who were Ab+.
Table 2

Frequencies of HLA-DQB1 genotypes among clinical type 1, clinical type 2, and unclassifiable, Ab+ versus Ab− diabetic patients

HLA-DQB1 genotypes

Control subjects n (%)

Type of diabetes

Clinical type 1

Clinical type 2

Unclassifiable

Total

Ab+ n (%)

Ab− n (%)

p value

Total

Ab+ n (%)

Ab− n (%)

p value

Total

Ab+ n (%)

Ab− n (%)

p value

Risk-associated

*0201/*0302

13 (6.0)

291 (24)a

271 (27)

20 (10)

<0.0001

31 (9.9)

16 (22)

15 (6.3)

0.0005

43 (19)a

33 (31)

10 (8.2)

<0.0001

*0302/X

9 (4.2)

287 (24)a

257 (26)

30 (15)

0.0018

48 (15)a

19 (26)

29 (12)

0.009

43 (19)a

29 (27)

14 (11)

0.0036

*0302/*0604

4 (1.9)

78 (6.5)b

74 (7.4)

4 (2.0)

0.0039

6 (1.9)

3 (4.1)

3 (1.3)

NS

12 (5.2)

10 (9.4)

2 (1.6)

0.014

Protective

*0602/X

21 (9.7)

11 (0.9)a

3 (0.3)

8 (4.1)

<0.0001

17 (5.4)

1 (1.4)

15 (6.3)

NS

8 (3.5)b

2 (1.9)

6 (4.9)

NS

*0201/*0602

12 (5.6)

6 (0.5)a

4 (0.4)

2 (1.0)

NS

13 (4.2)

0 (0)

13 (5.5)

0.043

6 (2.6)

1 (0.9)

5 (4.1)

NS

*0302/*0602

11 (5.1)

12 (1.0)a

8 (0.8)

4 (2.0)

NS

9 (2.9)

0 (0)

9 (3.8)

NS

3 (1.3)b

1 (0.9)

2 (1.6)

NS

*0301/*0602

10 (4.6)

9 (0.8)a

2 (0.2)

7 (3.6)

<0.0001

14 (4.5)

2 (2.7)

12 (5.1)

NS

1 (0.4)b

0 (0)

1 (0.8)

NS

*0201/*0301

16 (7.4)

32 (2.7)a

25 (2.5)

7 (3.6)

NS

10 (3.2)b

2 (2.7)

8 (3.4)

NS

11 (4.8)

1 (0.9)

10 (8.2)

0.012

*0602-3-4/X

14 (6.5)

16 (1.3)a

5 (0.5)

11 (5.6)

<0.0001

17 (5.4)

0 (0)

17 (7.2)

0.016

12 (5.2)

2 (1.9)

10 (8.2)

0.038

X/X

14 (6.5)

17 (1.4)a

8 (0.8)

9 (4.6)

0.0005

17 (5.4)

3 (4.1)

14 (5.9)

NS

14 (6.1)

3 (2.8)

11 (9.0)

NS

*0201/X

43 (20)

180 (15)

147 (15)

32 (16)

NS

47 (15)

11 (15)

35 (15)

NS

25 (11)b

12 (11)

13 (11)

NS

Neutral

*0301/X

17 (7.9)

49 (4.1)b

27 (2.7)

22 (11)

<0.0001

25 (8.0)

3 (4.1)

22 (9.3)

NS

15 (6.6)

3 (2.8)

12 (9.8)

0.035

*0301/*0603

4 (1.9)

3 (0.3)b

0 (0)

3 (1.5)

0.004

8 (2.6)

0 (0)

8 (3.4)

NS

4 (1.8)

0 (0)

4 (3.3)

NS

*0301/*0604

3 (1.4)

17 (1.4)

14 (1.4)

3 (1.5)

NS

7 (2.2)

1 (1.4)

6 (2.5)

NS

1 (0.4)

0 (0)

1 (0.8)

NS

*0301/*0302

8 (3.7)

70 (5.9)

59 (5.9)

11 (5.6)

NS

16 (5.1)

5 (6.8)

11 (4.6)

NS

10 (4.4)

5 (4.7)

5 (4.1)

NS

*0201/*0603

5 (2.3)

17 (1.4)

11 (1.1)

6 (3.1)

0.046

9 (2.9)

1 (1.4)

8 (3.4)

NS

3 (1.3)

1 (0.9)

2 (1.6)

NS

*0201/*0604

2 (0.93)

46 (3.8)b

39 (3.9)

7 (3.6)

NS

5 (1.6)

1 (1.4)

4 (1.7)

NS

7 (3.1)

3 (2.8)

4 (3.3)

NS

*0302/*0603

3 (1.4)

39 (3.3)

33 (3.3)

6 (3.1)

NS

8 (2.6)

5 (6.8)

3 (1.3)

0.02

5 (2.2)

0 (0)

5 (4.1)

NS

*0604/X

7 (3.2)

17 (1.4)

13 (1.3)

4 (2.0)

NS

6 (1.9)

1 (1.4)

5 (2.1)

NS

6 (2.6)

1 (0.9)

5 (4.1)

NS

Total

216

1197

1000

196

 

313

74

237

 

229

107

122

 

Data are presented as absolute numbers (%). p value indicates significance for Ab+ vs Ab− for clinical type 1, clinical type 2, and unclassifiable diabetic patients. A p value <0.0026 is considered significant. Comparisons between clinical type 1, clinical type 2 and unclassifiable patients vs control subjects in terms of HLA-DQB1 genotypes: ap<0.0026; bp<0.05. X denotes either homozygous or other HLA-DQB1 allele not identified by our genotyping method

Patients with a young age at onset of diabetes (≤25 years) had a two-fold increased risk of having the HLA-DQB1*0201/*0302 genotype (OR=1.6, 95% CI 1.3–2.1, p<0.0001) relative to those with an older age at onset (>25 years); however, this effect disappeared in multivariate analysis when islet antibodies were included. In this analysis, age at onset was included as a dependent categorical, and HLA-DQB1 genotypes and islet antibodies were included as independent categorical variables.

Islet antibodies in relation to HLA-DQB1 genotypes, age at onset and sex

Patients with the HLA-DQB1*0201/*0302 genotype had a three-fold increased risk of having GADA (OR=3.4, 95% CI 2.4–4.7, p<0.0001) relative to patients with protective/neutral genotypes. In contrast, patients with HLA-DQB1*0302/X (where X denotes either homozygous or other HLA-DQB1 allele not identified by our genotyping method) had a three-fold increased risk of having IA–2A (OR=3.2, 95% CI 2.3–4.4, p<0.0001) relative to patients with protective/neutral genotypes, and in patients with HLA-DQB1*0302/*0604, the risk was increased five-fold (OR=5.6, 95% CI 3.1–10.2, p<0.0001) than patients with protective/neutral genotypes.

Being Ab+ was significantly associated with young age at onset of diabetes (≤25 years) as compared with Ab− (654 [52%] vs 180 [29%], p<0.0001). Age at onset and sex (independent categorical variables) in relation to different islet antibody combinations (dependent categorical variable) included in a multinomial logistic regression analysis showed that young age at onset (≤25 years) was independently and significantly (p<0.0001) associated with islet antibodies, particularly with IA–2A, alone (OR=7.2, 95% CI 3.2–16.4, p<0.0001) or in combination with ICA (OR=7.0, 95% CI 4.1–12.0, p<0.0001). In addition, male sex was independently associated with IA–2A, alone (OR=5.7; 95% CI 1.7–18.9, p=0.005) or in combination with ICA (OR=4.0; 95% CI 2.0–8.0, p<0.0001), whereas female sex was independently associated with GADA in combination with ICA (OR=1.7, 95% CI 1.3–2.4, p=0.001). Among patients positive for GADA, the median GADA concentration was significantly higher in women than in men (index values 82 [Q1=27, Q3=112] vs 39 [Q1=16, Q3=91], p<0.0001).

fpC-peptide in relation to islet antibodies and HLA-DQB1 genotypes

Table 1 shows that the median fpC-peptide concentration after diagnosis was significantly (p<0.0001) lower among patients with clinical type 1 diabetes (0.24 nmol/l) than among those with clinical type 2 diabetes (0.75 nmol/l) or unclassifiable (0.48 nmol/l) diabetes. Figure 2 shows that Ab+ patients had a significantly lower median fpC-peptide concentration than Ab− patients (0.24 vs 0.70 nmol/l, p<0.0001), irrespective of the presence of risk-associated HLA-DQB1 genotypes or not. On the other hand, Ab− patients with risk-associated HLA-DQB1 genotypes had a significantly lower median fpC-peptide concentration than patients with protective/neutral genotypes (0.51 vs 0.74 nmol/l, p=0.0003). After stratification for islet antibody presence, multiple regression analysis was performed in each group (Ab+ and Ab−). This analysis included fpC-peptide (dependent continuous variable), fpC-peptide follow-up time (≤6 months or >6 months) and HLA-DQB1 genotypes (independent categorical variables). The analysis showed that among Ab− patients, risk-associated HLA-DQB1 genotypes were significantly (p=0.001) associated with decreased fpC-peptide concentration, irrespective of when the follow-up sample was taken. In contrast, among Ab+ patients, a long fpC-peptide follow-up time (>6 months) was significantly (p<0.0001) associated with decreased fpC-peptide concentration, irrespective of risk-associated HLA-DQB1 genotypes.
https://static-content.springer.com/image/art%3A10.1007%2Fs00125-006-0293-5/MediaObjects/125_2006_293_Fig2_HTML.gif
Fig. 2

Levels of fpC-peptide measured at follow-up in relation to islet antibodies and HLA-DQB1 genotypes. Altogether, complete data on follow-up fpC-peptide, HLA-DQB1 genotypes and islet antibodies were available for 1,455 patients. The figure shows median fpC-peptide concentrations in bold (nmol/l), and 25th percentile (Q1) and 75th percentile (Q3) values in brackets. The median fpC-peptide concentration was significantly (p<0.0001) lower in Ab+ vs Ab− patients, irrespective of HLA-DQB1 genotypes. However, Ab− patients with risk-associated HLA-DQB1 genotypes had significantly lower (p=0003) median fpC-peptide levels than Ab− patients with protective/neutral genotypes

Categorisation of fpC-peptide according to low (≤0.25 nmol/l) or normal (>0.25 nmol/l) concentrations, respectively, showed that a low fpC-peptide (≤0.25 nmol/l) concentration was significantly associated with islet antibodies (p<0.0001) but not with HLA-DQB1 genotypes in multinomial logistic regression analysis. In the regression analysis, nominal fpC-peptide was included as a dependent categorical variable and islet antibodies, HLA-DQB1 genotypes, follow-up time, age at onset, sex, BMI and family history of diabetes as independent categorical variables.

Ketonuria, acidosis, diabetic symptoms in relation to islet HLA-DQB1 genotypes, islet antibodies, fpC-peptide, age at onset and sex

At diagnosis, 1,138 (60%) of diabetic patients had ketonuria, 149 (10%) had acidosis and 1,835 (93%) had diabetic symptoms; 1,401 (75%) had a short duration of diabetic symptoms (<3 months). Multinomial regression analysis showed that presence of ketonuria was significantly (p<0.0001) associated with islet antibodies, particularly ICA combined with IA–2A (OR=4.0, 95% CI 1.9–8.1) or three antibodies (OR=3.2, 95% CI 2.3–4.6) or male sex (OR=1.8, 95% CI 1.4–2.3), but not with risk-associated HLA-DQB1 genotypes. Prevalence of diabetic symptoms was significantly (p<0.0001) associated with three antibodies (OR=4.3, 95% CI 2.0–9.1) and male sex (OR=2.5; 95% CI 1.5–4.0). Moreover, short length of symptomatic period (<3 months) was significantly associated (p<0.0001) with three antibodies (OR=2.4; 95% CI 1.6–3.6). In the regression analysis, presence of ketonuria, the prevalence of diabetic symptoms and length of symptomatic period were separately included as dependent categorical variables, whereas islet antibodies, HLA-DQB1 genotypes, fpC-peptide, follow-up time, sex, age at onset, BMI and family history of diabetes were included as independent categorical variables.

Discussion

This study of 2,077 young adults (15–34 years old) with recently diagnosed diabetes, 1,869 of whom had islet antibodies measured, shows that 1,250 of 1,869 (67%) had autoimmune type 1 diabetes if islet antibodies were used as an objective diagnosis of autoimmune type 1 diabetes. Among those classified as having type 1 diabetes, 83% had objective type 1 diabetes; however, 23% of those classified with type 2 diabetes and 45% of those classified with unclassifiable diabetes also had objective type 1 diabetes. If immeasurable or low fpC-peptide after diagnosis was included as a further objective criterion for clinical type 1 diabetes, only another 70 patients without islet antibodies were added to the clinical type 1 diabetes group. Hence, if patients with low or immeasurable fpC-peptide were to be included, the proportion with type 1 diabetes would increase by only 4%, i.e. 71% of all incident young adults between 15 and 34 years of age developing diabetes would have objective type 1 diabetes. Accordingly, islet antibodies, not fpC-peptide, are most important in the classification of diabetes among young adults. Nevertheless, our study has shown that non-autoimmune diabetes (presumably mostly type 2 diabetes) is not rare among young adult diabetic patients, being identified in about every third patient. Hence, the differential diagnosis between clinical type 1 and clinical type 2 diabetes is a major issue among incident diabetic patients aged 15–34 years.

Based on islet antibodies at diagnosis, we have previously reported that around 25% of clinical type 2, and 50% of unclassifiable young adult diabetic patients, should be considered as having type 1 diabetes [1, 18]. It is noteworthy that the current DISS study, conducted from the start of 1998 to the end of 2002, gives similar frequencies of islet antibodies among clinical type 2 and unclassifiable diabetic patients, as shown in the previous DISS studies performed in 1987–1988 and 1992–1993. This highlights the fact that, among those aged 15–34 years, an objective classification based on islet antibodies is necessary to achieve a reliable classification.

Our study emphasises that, although risk-associated HLA-DQB1 genotypes are closely associated with islet antibodies, these HLA-DQB1 genotypes do not contribute to the classification of type 1 diabetes per se. Indeed, the only suggestion that risk-associated HLA-DQB1 genotypes may contribute to an aetiological classification was among patients without islet antibodies. In line with the Belgian Diabetes Registry [3], we found that among Ab− individuals, those with risk-associated HLA-DQB1 genotypes had a lower median fpC-peptide concentration than those with protective/neutral genotypes, irrespective of the timing of the follow-up fpC-peptide measurement. However, patients with islet antibodies and samples taken 6 months after diagnosis had low fpC-peptide concentrations, irrespective of risk-associated HLA-DQB1 genotypes, again underlining that in the presence of islet antibodies, genetic risk assessment based on HLA is not important in the classification. It was also observed previously that in the presence of islet antibodies, genetic risk or protection does not matter in the prediction of diabetes development [1921]. Our observation that low fpC-peptide was associated with risk-associated HLA genotypes in Ab− participants indicates that risk-associated HLA-DQB1 genotypes may themselves be related to impaired beta cell function. This corresponds to previous reports that the presence of a risk-associated genotype in Ab− patients confers an increased risk of insulin requirement at a later follow-up [22]. However, antibodies not detected by current assays may be present in patients without islet antibodies but with risk-associated HLA-DQB1 genotypes and low fpC-peptide. Another option is a later development of islet antibodies. Previous studies have shown that up to 10% of young adult-onset diabetic patients without islet antibodies convert to positivity after the diagnosis of diabetes [2326]. The disappearance of previous islet antibodies should also be considered.

Interesting associations between risk-associated HLA-DQB1 genotypes, age at onset of diabetes and islet antibodies were detected. It has been reported that diabetic patients with a young age at onset (≤25 years) have increased frequencies of HLA-DQB1*0201/0302 genotypes [2730]; however, according to our study, significant associations between HLA-DQB1 genotypes and young age at onset were due to islet antibodies. In agreement with previous observations [3133], GADA were associated with HLA-DQB1*0201/*0302, whereas IA–2A were associated with HLA-DQB1*0302/X and *0302/*0604, respectively. Indeed, IA–2A concentration was highest among patients with the HLA-DQB1*0302/X genotype. IA–2A are known to be associated with a rapid onset of type 1 diabetes, as well as with young age at onset [3436]. Hence, our study infers that the association between IA–2A and rapid onset may be dependent on the HLA-DQB1 locus. Our study also confirms that IA–2A were associated with male sex, whereas the presence and high levels of GADA were associated with female sex [37, 38]. However, logistic regression analysis showed that the associations between sex and age at onset to the types of islet antibodies were not related to risk-associated HLA-DQB1 genotypes, as has been previously shown [39]. Thus, the well known increased incidence of type 1 diabetes among young adult men [4043], as also demonstrated in our study, does not seem to be related to HLA-DQB1, but an effect of sex in itself. This is emphasised by our finding that ketonuria and diabetic symptoms were clearly associated with male sex.

A clear finding in our study was that normal fpC-peptide up to 6 months after diagnosis does not exclude type 1 diabetes. This finding gives further support to the concept that islet antibodies are the method of choice in the classification of diabetes. The preserved beta cell function in most patients with islet antibodies demonstrates that the process of beta cell destruction is not always fast among young adult diabetic patients. Indeed, it fits with the previous observation that it may take 12 years before severe beta cell failure develops in adult patients with islet antibodies [44].

It has been reported, that high concentrations of islet antibodies are associated with low fpC-peptide values [45], but no such association was shown in our study. We found no correlation between the number of islet antibodies and fpC-peptide, as previously reported [46, 47]. This most likely reflects that beta cell failure was not yet frequent among our study patients. Prospective follow-ups may in the future show high concentrations and/or a high number of islet antibodies in association with beta cell failure among our patients.

The major strength of this study is that we recruited a large sample of incident population-based and representative diabetic patients aged 15–34 years from a whole country. Indeed, since 1983 DISS have included >9,000 15–34-year-old patients at diagnosis of diabetes. In this 5-year study conducted between the start of 1998 and the end of 2002, blood samples were taken in most incident cases and we were able to compare complete data on HLA-DQB1 genotypes, islet antibodies and fpC-peptide in 1,455 newly diagnosed young adult diabetic patients (Fig. 2). It can be argued that we did not determine HLA-DQA1 genotypes in our patients. No doubt that extended genotyping is helpful for relative risk estimation. The sensitivity increases when new genotypes conferring risk are included. However, the linkage disequilibrium between alpha and beta chain alleles is very strong. Additional information obtained by typing for HLA-DQA1 would thus be of limited importance for our study [12]. Furthermore, our association study of HLA-DQB1 loci may be considered a cost-effective way of identifying the contribution of HLA to the classification of diabetes in individuals aged 15–34 years.

In conclusion, this study shows that (1) irrespective of clinical classification, 67% of patients with newly diagnosed diabetes at the age of 15–34 years have autoimmune type 1 diabetes; (2) islet antibodies strongly contribute to the aetiological classification of diabetes; (3) islet antibodies are more closely associated with beta cell impairment than increased-risk HLA-DQB1 genotypes; (4) risk-associated HLA-DQB1 genotypes are associated with islet antibodies and do not contribute to the classification of diabetes in Ab+ individuals; (5) risk-associated HLA-DQB1 genotypes are, however, associated with low fpC-peptide concentrations in the absence of islet antibodies, presumably identifying non-autoimmune type 1 diabetes; and (6) absence of islet antibodies and high fpC-peptide concentrations predict a type 2 diabetes phenotype. Taken together, the data re-emphasise the need to measure islet antibodies for the diagnosis of autoimmune diabetes in young adults in clinical practice, whereas HLA-DQB1 genotyping may be of interest in patients without islet antibodies.

Acknowledgement

We thank all Swedish diabetologists and diabetes nurses who contributed to DISS. C. Rosborn, U. Gustavsson, A. Radelius, G. Gremsperger and J. Pilz are acknowledged for expert technical assistance. The Juvenile Diabetes Foundation-Wallenberg Diabetes Research Program (K 98-99 JD- 128 13), the Swedish Diabetes Association, the Swedish Medical Research Council (72X-14531), the Albert Påhlsson Foundation and the Research Fund at Malmö University Hospital are acknowledged for support of DISS.

Duality of interest

L. C. Groop has been a consultant and served on advisory boards for Aventis-Sanofi, Bristol-Myers Squibb, Kowa and Roche.

Copyright information

© Springer-Verlag 2006