Prevalence of IgA-Antiendomysium and IgA-Antigliadin Autoantibodies at Diagnosis of Insulin-Dependent Diabetes Mellitus in Swedish Children and Adolescents
Objective. This study was conducted to investigate the prevalence of celiac disease (CD) in children and adolescents at diagnosis of insulin-dependent diabetes mellitus (IDDM) before insulin treatment was started.
Material and Methods. At diagnosis of IDDM, and before treatment was started, 115 children and adolescents were screened for IgA- antiendomysium (EMA) and IgA-antigliadin antibodies (AGA). Those found to be EMA-positive and/or AGA-positive were investigated further with intestinal biopsy.
Results. Of the 115 patients, 2 had known CD at diagnosis of IDDM; of the remainder of patients, 6% (7/113) were found to be EMA-positive and 9% (10/113) were found to have AGA levels above normal. Of the 6 patients who underwent biopsy, 5 manifested villous atrophy. In addition, 2 patients with high EMA and AGA antibody titers refused biopsy, and 4 patients with low EMA and/or AGA titers were found to have normal titers at control before biopsy decision.
Conclusion. Because the prevalence of CD at diagnosis of IDDM would seem to be 6% to 8%, screening for CD seems to be justified among patients with newly diagnosed IDDM.
- insulin-dependent diabetes mellitus
- celiac disease
- IgA-antigliadin antibodies
- IgA-antiendomysium antibodies
- IDDM =
- insulin-dependent diabetes mellitus •
- CD =
- celiac disease •
- EMA =
- antiendomysium antibodies •
- AGA =
- antigliadin antibodies •
- AU =
- arbitrary units
Insulin-dependent diabetes mellitus (IDDM) is a chronic disorder that results from autoimmune destruction of the insulin-producing pancreatic β cells. The incidence in Sweden is 30/100 000 children,1 compared with 13/100 000 in the United States.2 It is well known that different autoimmune disorders occur in association with each other. The most common example of such coexistence is IDDM occurring in conjunction with autoimmune thyroiditis.3 Even in the absence of clinical disease, organ-specific autoantibodies are often found in patients with IDDM,3–5 and studies performed close to the diagnosis of IDDM have shown that the immunologic abnormalities may be transient and decline with time.4 ,5 Coexistence of celiac disease (CD) and IDDM has been described in several studies,6–8 and figures reported for its prevalence have ranged from 1% to 7%,5–13 rates that are higher than the rates of CD in the general population, estimated recently to be ∼.3% in Sweden,14 compared with .01% in the United States.9
CD is heterogeneous in its clinical and pathologic manifestations,15 and typical gastrointestinal symptoms such as poor weight gain, diarrhea, and abdominal distension seem to be rare in patients with IDDM.5 ,16 The increasing numbers of reports of milder, less symptomatic forms of CD without overt signs of malabsorption suggest that screening is necessary to detect CD in patients with IDDM. IgA-antigliadin autoantibodies (AGA) help to identify CD;10 ,11 but with a sensitivity and specificity of only ∼80%, an AGA test is insufficient for screening purposes.17 It is also known that the prevalence of AGA may be high in patients with IDDM without revealing CD with a flat mucosa.5 ,8 ,10 ,13
In many studies, IgA-antiendomysium autoantibodies (EMA) have been shown to be a more reliable marker than AGA for the diagnosis of CD,18–20 because the sensitivity and specificity of the test are much higher except in children younger than 2 years of age.21
EMA and AGA were not measured at the diagnosis of IDDM before insulin treatment in any of the studies noted above. The goal of our study was to investigate the prevalence and clinical relevance of EMA and AGA at the onset of IDDM, before insulin treatment was started.
MATERIAL AND METHODS
Malmö, Sweden, a city with a population of 240 000, is served by a single tertiary hospital and a single department of pediatrics. Between 1976 and 1998, at diagnosis of IDDM and before insulin treatment was started, sera were obtained from 115 patients (51 girls and 64 boys), all younger than 15 years of age (mean age at diagnosis: 8.6 years; range: .8–14.9 years). Since 1993, rescreening has been performed annually. Sera were stored at −20°C until analyzed.
The control group was made up of 384 healthy school children (199 girls and 185 boys) between the ages of 11 and 13 years.
A commercial enzyme-linked immunosorbent assay (Gluten IgA EIA; Pharmacia, Uppsala, Sweden) microplate was used. Microplate strips were coated with gliadin. Patient serum was diluted 1/200. Patient IgA was detected using galactosidase-conjugated rabbit antihuman IgA and the chromogenic substrate nitrophenyl-β-galactoside, the amount of specific IgA being proportional to the absorbance measured at 405 nm.
As recommended by the manufacturer, results were expressed in arbitrary units (AU), each unit representing 1% of the optical density value of the positive control. The cutoff point for a positive result was set at 25 AU. In children 5 to 15 years old, this cutoff level will give a sensitivity of 58% and a specificity of 93% (as shown by the manufacturer's analysis of sera from 84 CD patients and 205 patients with nonceliac gastrointestinal disease between the ages of 5 and 15 years). In children <5 years of age, the sensitivity of the method has been shown to be 92% and the specificity 84% (based on tests of 85 patients with CD and 144 patients without CD).
The AGA data of the control group sera have been published separately22 and were obtained using an in-house enzyme-linked immunosorbent assay with the same gliadin type of antigen that was used for the Pharmacia assay. The cutoff level of the in-house method was set to give a sensitivity of 56% and a specificity of 100%, ie, figures similar to those obtained with the cutoff adopted by us for the Pharmacia assay.
Using both the Pharmacia and the in-house methods, we have analyzed sera from 99 children (<15 years) with gastrointestinal symptoms, obtaining identical results in 96% of cases (16 children positive with both tests, 79 negative in both tests, 3 children positive with the in-house method only, and 1 child positive with the Pharmacia method only). Therefore, data for the large control group22 have also served as control data in the present study.
Indirect immunofluorescence analysis was performed using commercially available fixed sections of monkey esophagus (distal 3rd part) (BioSystems, Barcelona, Spain) as antigen substrate. Patient serum was diluted 1/5 in phosphate-buffered 0.15 mNaCl (pH 7.6) and 1% bovine serum albumin. Patient endomysium-specific IgA was detected with a fluorescein isothiocyanate-labeled antihuman IgA conjugate (BioSystems). The result was expressed as the highest dilution factor giving a positive fluorescence pattern. All sera manifesting fluorescence (titer ≥5) were considered to be positive. The great majority of all EMA studies have shown this serologic marker to be associated with very high specificity (98%–100%), and somewhat lower sensitivity (78%–100%) in series of pediatric patients from North America23 or Sweden.19 ,20 In one of these studies,20 the sensitivity in children <5 years of age was found to be similar to the sensitivity in children ≥5 years of age (95% vs 100%).
We routinely include in-house negative and positive control subjects for both AGA and EMA in every analysis. Moreover, for these two tests our laboratory takes part in Swedish National and the United Kingdom National External Quality Assessment Schemes.
A small bowel biopsy was performed with a Watson capsule under fluoroscopic control at the level of the ligament of Treitz. The specimens were put in formaldehyde solution and examined histologically at the Department of Pathology, University Hospital, Malmö, Sweden. The intestinal mucosa was classified as normal, subnormal (villous length/crypt length <2, increased number of inflammatory cells in the musosa with or without damage to the surface epithelium, and brush border), or total villous atrophy (flat mucosa). The revised criteria for the diagnosis of CD were used.24
Fisher's exact test (one-tailed) and the Mann-WhitneyU test were used for statistical analysis of the data.P values < .05 were considered significant.
Of the IDDM series as a whole, 2% (2/115) had known CD before the diagnosis of IDDM; a girl had known CD since age 5 years and a boy had known CD since he was age 10 months (Table 1).
Of the remaining patients with IDDM, 6% (7/113; 5 girls, 2 boys) were found to be EMA-positive, compared with .3% (1/384) of the control subjects (P = .0002) (Fig 1; Table 1). The mean age of the these 7 patients who were EMA-positive was 5.4 years (range: 1.1–11.4).
Of the patients positive for EMA, 5 were also positive for AGA, and their mean age was 5.4 years (range: 1.1–11.4 years). However, 5 of the AGA-positive children were found to be EMA-negative (mean age: 4.7; range: .8–11.9 years), and 2 of the children (4.1 and 7.8 years of age) were EMA-positive but AGA-negative.
At follow-up, the AGA level was found to have normalized in 3 of the 5 AGA-positive/EMA-negative patients and in 1 EMA-positve/AGA-negative patient.
Intestinal biopsy was performed in 6 patients (3 girls, 3 boys): 3 EMA/AGA-positive, 1 EMA-positive/AGA-negative, and 2 EMA-negative/AGA- positive. Of the biopsies, 5 revealed villous atrophy. One 2.5-year-old boy with normal biopsy findings had slightly increased AGA (30 AU) and EMA (1/5) titers that normalized within 2 years. A 1.1-year-old girl (EMA 1/1280 and AGA 74 AU) and a 4.2-year-old boy (EMA 1/1600 and AGA 64 AU) had villous atrophy. The 2 AGA-positive/EMA-negative patients, a 1.8-year-old boy with Down syndrome, and a .8-year-old girl, both were found to have flat mucosa at biopsy, as had the 7.8-year-old girl with an EMA titer of 1/1280 but no AGA autoantibodies. The mean age of the children with villous atrophy was 3.1 years (range .8–7.8), compared with the mean age of 8.8 years (range: 1.3–14.9) of the children with no CD (P= .0025).
Two girls (7.9 and 11.4 years of age), both EMA-positive (1/320 and 1/1280, respectively) and AGA-positive (168 AU and 86 AU, respectively) refused biopsy (Table 1). If these girls also had a flat mucosa, the mean age of the population with CD would have been 5.0 instead of 8.8 years (P = .031).
In summary, 5 patients were found to have CD when screened at diagnosis of IDDM, and 2 patients had known CD. They all have since normalized their autoantibody levels on a gluten-free diet. The 2 girls, refusing biopsy despite high levels of both EMA and AGA probably suffer from CD. This means that the prevalence was at least 6% but probably as high as 8%.
At follow-up, 2 patients (1 girl, 1 boy) who at diagnosis of IDDM were both AGA-negative and EMA-negative were found to have developed these autoantibodies within the 1st year after diagnosis. At biopsy, 1 patient was found to have a flat mucosa, and the other patient had a mucosa infiltrated with lymphocytes consistent with an early phase of CD. This means that the prevalence of CD in the entire population with IDDM may be as high as 10%.
The present study confirmed that children and adolescents at diagnosis of IDDM are characterized by a high prevalence of EMA (6%) and AGA (9%), compared with the general population. The prevalence of EMA-positivity that we found is similar to that found in other IDDM series.5 ,9 ,12 The prevalence of AGA-positivity was in the same range (5%–10%) as figures reported by others,10–12 although their studies were not restricted to patients investigated before insulin therapy was started. One study4 conducted near to the time of diagnosis of IDDM reported a prevalence of 2.6% for AGA and 3.0% for EMA, and another study13 in which reticulin and gliadin antibody screening was performed within 3 weeks after the diagnosis of IDDM reported a prevalence of 5.3% for reticulin antibodies and 2.3% for AGA antibodies. Lorini and co-workers,5 who investigated 51 of 168 patients at diagnosis of IDDM, found 33% (17/51) of the patients to be AGA-positive and 10% (5/51) of the patients to be EMA-positive.
A noteworthy finding in the present study was that in 3 of 9 AGA-positive/EMA-negative children and 1 of the EMA-positive/AGA-negative children, the abnormal titers had normalized by subsequent control. In the study noted above,5 the AGA titers had normalized in 13 of the 17 AGA-positive children at follow-up. This may reflect an immunologic chaos existing at onset of IDDM, although the presence of other gastroenterologic disease may be an alternative explanation.
In our study, at least 5, but probably 7, children had unknown CD at diagnosis of IDDM, ie, a prevalence of ∼6.0%. Including the patients with known CD before IDDM diagnosis and the 2 children who developed CD, the prevalence would seem to be 8% to 10%, which is higher than the findings from other studies conducted both in Sweden8and in other countries.9–11 The higher prevalence that we found may be the result of using the combination of AGA and EMA. We found 1 EMA-positive/AGA-negative patient with flat mucosa. On the other hand, we also found 2 AGA-positive/EMA-negative patients with flat mucosa. Of these children, 1 was a 21-month-old boy with Down syndrome, and the other was an 8-month-old girl. This is to be expected because it has been shown that EMA has decreased specificity in children <2 years of age.21 The high specificity of EMA has been demonstrated in numerous studies,17–20but in our study 1 of 6 EMA-positive patients had a normal biopsy result (titer 1/5), and 1 had a normalized EMA titer at follow-up 3 months later. This indicates that EMA is better for screening than AGA and reticulin antibodies in IDDM, because a transient increase in these antibodies is detected relatively frequently.13 As did other researchers,11 we found that the onset of IDDM generally occurred at an earlier age in the subgroup with CD-associated antibodies, which may reflect the presence of a more aggressive form of the autoimmune disease.
The prevalence of CD among pediatric patients with newly diagnosed IDDM is 20 times higher than that for the general Swedish population.14 It is unknown whether the association between CD and IDDM is determined genetically or whether one disease predisposes patients to the other.
Because CD seldom precedes IDDM, it has been suggested that IDDM might constitute an immunologic trigger for CD.25 Others have proposed that untreated CD or silent CD might induce IDDM.26 Reports have been published indicating an increased prevalence of autoimmune markers for diabetes mellitus in patients with CD.27 ,28 In our study, 2 patients had known CD before the IDDM manifested. A total of 5 patients had antibodies and a flat mucosa when biopsy was performed at the diagnosis of IDDM. This may suggest that in some patients, CD may trigger the onset of IDDM.
Both IDDM and CD are associated with certain HLA alleles of the DQB1 and DQA1 locus. Susceptibility for IDDM is primarily attributable to DQB1*0302 and *0201,29 whereas CD DQA1*0501/DQB1*0201 provides the primary association.30 The HLA alleles may explain some of the co-existance of the two diseases, but the reason IDDM diagnosis precedes CD diagnosis still is unknown. It has been postulated that in patients newly diagnosed with IDDM, there may be a nonspecific activation of the immune system directed toward several dietary proteins attributable to defective immunoregulation or loss of immunologic tolerance of a variety of ingested antigens.31
In this study, we found a high prevalence of CD at the diagnosis of IDDM. It is important to treat subclinical CD, because patients with undiagnosed CD are at risk of long-term CD complications, such as infertility, anemia, osteoporosis, and malignancy.32 Of course, the well-being of the patients and optimal gut function in patients with IDDM are preferable. It also is important to consider that some patients without markers of CD at the onset of IDDM will develop the disease later. Therefore, we suggest that children and adolescents should be screened for CD at diagnosis of IDDM and then annually using the combination of AGA and EMA testing for children <2 to 3 years of age and the EMA test alone for those who are older.
This study was supported by grants from the Faculty of Medicine, University of Lund; the Health Services Administration, University Hospital, Malmö; the Novo Nordisk Fund; the Malmö Branch of the Swedish Diabetic Association; the Childhood Diabetic Fund, the Sven Jerring Fund; Lion Club International District 101-S; and the Swedish Medical Research Council (Grant 27X-12274).
- Received July 12, 1998.
- Accepted October 19, 1998.
Reprint requests to (A.K.C.) Department of Pediatrics, University Hospital, S-20502 Malmö, Sweden.
- Mäki M,
- Hällström O,
- Huupponen T,
- Vesikari T,
- Visakorpi JK
- Barera G,
- Bianchi C,
- Calisti L,
- et al.
- De Vitis I,
- Ghirlanda G,
- Gasbarrini G
- Ascher H,
- Holm K,
- Kristiansson B,
- Mäki M
- Corrao G,
- Corazza GR,
- Andreani ML,
- et al.
- Bürgin-Wolff A,
- Gaze H,
- Hadziselimovic F,
- et al.
- European Society of Paediatric Gastroenterology and Nutrition
- ↵Holmes GKT. Long-term health risks for unrecognized coeliac patients. In: Auricchio S, Visakorpi JK, eds. Common Food Intolerance. I. Epidemiology of Coeliac Disease. Basel, Switzerland: Karger; 1992:105–118
- Copyright © 1999 American Academy of Pediatrics