Objective. To investigate the prevalence of celiac disease in a large cohort of children and adolescents at the onset of type 1 diabetes and the occurrence of new cases during a 6-year follow-up.
Methods. We prospectively studied, by repeated serologic screening, 274 consecutive patients at the onset of type 1 diabetes (age [mean ± standard deviation]: 8.28 ± 4.65 years) for 6 subsequent years. One patient had a diagnosis of celiac disease before the onset of diabetes. The immunoglobulin A-antiendomysium antibody test was selected as the screening test; patients with positive results (++ or +++) or with 2 consecutive weak positive tests (+) were considered appropriate for the jejunal biopsy.
Results. At diabetes onset, 15 (5.5%) of 273 patients tested positive with the antiendomysium test; jejunal biopsy was performed in 10, and celiac disease was diagnosed in 9. The prevalence of biopsy-confirmed celiac disease at the manifestation of diabetes was 3.6% (10 of 274 patients). Twelve more patients with a negative antiendomysium antibody test at diabetes onset tested positive during the follow-up within 4 years; 10 of them had biopsies performed, and 7 had celiac disease. Therefore, the overall prevalence of biopsy-confirmed celiac disease in the entire cohort of patients was 6.2%. The age at diabetes onset in patients with and without celiac disease was not different (7.88 ± 5.69 vs 8.3 ± 4.58 years). The majority of cases of celiac disease were asymptomatic in their presentation, and no signs of overt malnutrition were documented.
Conclusions. The prevalence of celiac disease in patients with type 1 diabetes is approximately 20 times higher than in the general population. Sixty percent of cases are already present at diabetes onset, mostly undetected, but an additional 40% of patients develop celiac disease a few years after diabetes onset. Extending screening programs for celiac disease after the onset of type 1 diabetes is recommended, even in the absence of clinical symptoms.
Celiac disease (CD) and type 1 diabetes are common chronic disorders in the pediatric age group, with estimated prevalences in unselected populations of 1 in 2581 and 1 in 1000,2 respectively. Coexistence of the diseases has been previously described in several cross-sectional studies, and figures reported for its prevalence have ranged from 1% to 8%,3–9 approximately 10 times higher than expected.
Few studies5–7,10,11 have been conducted at or near the time of the onset of type 1 diabetes, and very few articles have reported follow-up evaluation for CD,5,11,12 documenting later development of disease. No longitudinal studies have been performed to investigate the development of new cases of CD during a long-term follow-up.
Most cases of CD in type 1 diabetes seem to be asymptomatic or silent and can be detected only by serologic screening.3 Many studies have shown that immunoglobulin A (IgA) antiendomysium antibodies (EMA) are reliable markers for the screening and diagnosis of CD in pediatric subjects in unselected populations13 as well as in patients with type 1 diabetes,4,14 with a high correlation with mucosal findings. Although no serologic test enjoys 100% sensitivity and specificity, positive and negative predictive values of EMA are as high as 97% and 98%, respectively.13
Our objective was to study by repeated serologic screening the prevalence of CD in a large cohort of patients with new-onset type 1 diabetes and the annual occurrence of new cases during a 6-year follow-up. We also evaluated the clinical, auxological, and nutritional features of CD in this population and the therapeutic effect of a gluten-free diet (GFD).
From January 1993 to December 1997, we prospectively recruited 274 consecutive patients with new-onset type 1 diabetes (116 female, 159 male; age [mean ± standard deviation]: 8.28 ± 4.65 years; range: 0.6–18.7 years). Patients came from different areas of northern Italy, having been referred to our pediatric endocrinology clinic. One patient had a diagnosis of CD before the onset of type 1 diabetes, and he was still on a GFD at the manifestation of type 1 diabetes (age at CD diagnosis: 3 years; age at type 1 diabetes onset: 16 years). This subject was excluded from the screening program. All 273 eligible patients were tested for EMA at the onset of type 1 diabetes (screening study) and annually up to 6 years or until 1999 (longitudinal study).
EMA was measured by indirect immunofluorescence on commercially available cryosection of monkey esophagus. Serum samples were tested diluted 1:10 in 50 mM phosphate buffer, 0.15 M NaCl, pH 7.3 (PBS) on 4-μm unfixed cryostat sections. After a 30-minute incubation with serum, sections were washed in PBS and incubated with fluorescein isothiocyanate rabbit anti-human IgA, washed again in PBS, and read on a fluorescent microscope by the same operator (E.B.). Fluorescence was compared, with positive and negative control samples tested in each assay. The slides were interpreted as positive only when a characteristic honeycomb pattern was observed around the smooth muscle fibers of the esophagus. Any positive reading in immunofluorescence (from + to +++) was considered a positive result. In a previous study that included 112 patients with celiac disease and 92 healthy control subjects, the sensitivity and specificity of our assay were 97.3% and 98.9%, respectively.15 Because the grading of positivity is subjective, patients with weak reactivity to EMA (+) were retested after 3 to 6 months while on gluten-containing diets before proceeding with the intestinal biopsy. The intestinal biopsy was recommended in all patients with positive (++ and +++) or with 2 weak positive (+) EMA results.
Serum IgA concentration was determined by standard methods in all patients, and values below 0.05 g/L in the presence of normal levels of serum IgG and IgM were regarded as selective IgA deficiency.16 These patients were tested for IgG-antiendomysium and IgG-antigliadin antibodies,15 because IgA-EMA will be falsely negative.
Intestinal Biopsy Procedure
Intestinal biopsy was performed by peroral procedure in older children (8 patients), after topical anesthetic (Xylocaine 10%) was sprayed onto the posterior pharynx. The suction capsule (2-hole biopsy catheter [Medi Tech, Microvasive, Boston, MA]) was advanced under fluoroscopy to the level of the duodenojejunal junction, and 1 or 2 specimens were obtained. Infants and young children had upper endoscopy (13 patients); meperidine and midazolam were titrated to allow conscious or deep sedation. At least 3 pinch mucosal samples were taken in the fourth portion of the duodenum.
Specimens were examined by a pediatric pathologist, who was blind to clinical symptoms and EMA results, according to the mucosal changes described by Marsh.17 The infiltrative (type 1) lesion comprises normal mucosal architecture in which the villous epithelium is markedly infiltrated by small, nonmitotic intraepithelial lymphocytes. The hyperplastic (type 2) lesion is similar to the type 1 lesion but with the addition of enlarged crypts whose epithelium, like the villi, is also infiltrated by intraepithelial lymphocytes. The destructive (type 3) lesion is characterized by some degree of villous atrophy, with inflammation and hyperplastic crypts, the classic lesion associated with celiac sprue.
The diagnosis of CD was considered the histologic demonstration of the hyperplastic or destructive mucosal lesions (type 2 or 3), followed by serologic response to a GFD, and by clinical remission in patients who exhibited any symptoms.18
Clinical and Laboratory Data
The following parameters were evaluated at type 1 diabetes onset in patients with CD and compared with those without CD: height, expressed as standard deviation score (Ht-SDS)19; body mass index, calculated as weight (kg)/height2 (m), expressed as standard deviation score (BMI-SDS)20; blood count indices, determined using an automatic analyzer: hemoglobin (normal range: 115–155 g/L), and mean red cell volume (normal range: 76–94 fL); serum albumin (normal range: 35–50 g/L), and serum ferritin (normal range: 31–337 pmol/L). These were measured by routine methods. The presence of associated autoimmune diseases and of thyroid microsomal autoantibodies was also assessed in the 2 groups of patients. Clinical, auxological, and nutritional features were studied at pretreatment baseline and after 3.42 ± 2.2 years of GFD therapy.
Paired or unpaired Student t test was used when indicated to compare variables between groups. The Cox proportional hazard model was performed to investigate the effect of age at diagnosis as a predictor of CD. Proportions and their 95% confidence intervals (95% CI) are reported. Life table analyses were computed according to Kaplan-Meier curves. All analyses were performed using SAS version 6.12 software (SAS Inc, Cary, NC). Significant differences were assumed at P < .05.
Screening Study at Type 1 Diabetes Onset
In this series of 273 patients, 2 had selective IgA deficiency, and IgG-antiendomysium and IgG-antigliadin antibody assays tested normal in both.
Fifteen individuals tested positive with the EMA assay (5.5%; 95% CI: 3.1–8.9): 9 presented EMA assays positive (++ or +++), and 6 were only weakly positive (+). Five patients with weak EMA reactivity (positive +) were negative at the second assay, and the intestinal biopsy was not performed. The intestinal biopsy was then performed in 10 individuals, showing in 9 (3.3%; 95% CI: 1.5–6.2) hyperplastic or destructive lesions (Marsh type 2 and 3), consistent with CD. Overall prevalence of CD (1 case already known, plus 9 detected by screening) in these 274 patients with new-onset type 1 diabetes was therefore 3.6% (95% CI: 1.7–6.6).
Median (first to third quartiles) time of follow-up was 24 months (12–48). The number of patients who underwent the annual rescreening by EMA, the frequency of positive results, and the prevalence of CD confirmed by biopsy during each year of follow-up are shown in (Fig 1). The cumulative prevalence of patients with at least 1 positive EMA test and of biopsy-proven CD during follow-up were 9.9% (95% CI: 6.6–14.1) and 6.2% (95% CI: 3.4–9.8), respectively.
The age at type 1 diabetes onset (mean ± standard deviation) was not different between the 17 patients with CD/type 1 diabetes compared with the 257 patients with type 1 diabetes without CD (7.88 ± 5.69 vs 8.30 ± 4.58 years; P = .72). The age at onset of type 1 diabetes was not a predictor of CD development (P = .4).
Eleven patients tested positive by EMA 1 or more times, but most of them turned out to be negative during follow-up despite a gluten-containing diet Table 1. Biopsy was indicated in 6 patients and was performed in 4. Patient 10 had 2 subsequent biopsies performed at 1-year intervals, and both showed a Marsh type 1 lesion without any progression of tissue damage. These patients were not considered as having CD. Biopsy was not indicated in the remaining 5 patients because they had negative results on second EMA assays at 3 to 6 months.
According to survival curves for positive EMA results and for CD diagnosis (Fig 2), 13.8% and 8.3% of patients with diabetes studied from the time of type 1 diabetes onset and for 6 years thereafter had the risk of developing at least 1 positive EMA assay and biopsy-confirmed CD, respectively.
Clinical and Laboratory Data
Auxological, nutritional, and clinical features of patients with type 1 diabetes with and without CD identified on screening are shown in Table 2. There was a decreased concentration of serum ferritin in patients with CD, but no other differences were noted. Three (18.7%) of the 16 patients with CD had autoimmune disorders: 1 had hypothyroidism attributable to Hashimoto’s thyroiditis, 1 had psoriasis, and 1 had increased thyroid microsomal autoantibodies associated with normal thyroid function.
After introduction of a GFD Table 3, no significant changes were noted in Ht-SDS, serum ferritin, serum folate, and serum albumin when compared with pretreatment values. However, GFD led to significant improvement in BMI-SDS (P = .02), hemoglobin (P = .003), and mean red cell volume (P = .028).
CD occurs more commonly in children3–12 and adults7,9,21 with type 1 diabetes than in the general population. The likely explanation for the frequent simultaneous occurrence is the shared genetic susceptibility provided by the HLA DR3-DQ2 haplotype.3 Most of the previous studies have been cross-sectional surveys, with screening tests performed either at the time of diagnosis of type 1 diabetes5–7,10,11 or at some time thereafter.3,4,9,12,19 It has been shown that CD can develop after the onset of type 1 diabetes5,11,12,22 in patients with no signs of CD-related autoimmunity at the time of diagnosis of type 1 diabetes.5,11,12
The present study was designed as prospective and longitudinal to identify for the first time the occurrence of CD in a large population of children, adolescents, and young adults studied from type 1 diabetes onset until year 6 after the diagnosis of diabetes. Our study confirms a high prevalence of biopsy-proven CD in patients with type 1 diabetes, with the overall prevalence of 6.2% (95% CI: 3.4–9.8). This represents a nearly 20-fold increase over the estimated prevalence of CD (0.38%; 95% CI: 0.24–0.55) in the Italian childhood population.1 These data are similar to5,23,24 or higher than11,12,25–27 that reported in previous studies, in which screening serologic tests were used to assess the prevalence of CD. It is interesting that the prevalence in our longitudinal study is similar to figures from studies in adult patients with type 1 diabetes23,24; because the incidence of CD seems to increase with the duration of type 1 diabetes,24,26 it is reasonable that prospective-longitudinal studies in which patients are screened several times may identify a higher number of celiac patients than childhood cross-sectional surveys, with screening at 1 point only.
It has been reported by others that CD may be present in the absence of EMA positivity28 and that the negative predictive value of EMA testing is in the order of 94%.29 Moreover, children with CD before 2 years of age may not express a positive EMA test30 and could go undetected. These data raise the question of whether our screening program detected all celiac patients among this cohort of individuals with type 1 diabetes. As only 8 individuals were younger than 2 years at the time of type 1 diabetes diagnosis and the combination of EMA and IgA-antigliadin antibody tests leads to a marginal increase of negative predictive value, from 94% to 95%,29 we believe that the present study expresses a reliable figure of CD occurrence in patients with type 1 diabetes. Moreover, most of the individuals who were younger than 2 years underwent subsequent annual testing. For obvious ethical reasons, we did not biopsy any of our antibody negative cases to determine the false-negative rate of detection.
The onset of type 1 diabetes in the group of patients with CD did not occur at an earlier age compared with the individuals with diabetes and without CD, in contrast with previous reports.5,8,31 Moreover, the age of type 1 diabetes onset was not a predictive factor of CD development.
Most of the CD cases are already present at type 1 diabetes onset (10 of 17 patients with CD), either as a preexisting diagnosis (1 case) or detected by serologic screening. This observation is confirmed by common practice and previous findings, showing that only occasionally the CD diagnosis precedes type 1 diabetes onset.3 In fact, most of the patients with CD and type 1 diabetes have few or no symptoms related to malabsorption, and when gastrointestinal complaints are present, they are often mild and appreciated only in retrospect.3,9 We found a similar trend: only 2 of the 16 patients with CD identified by screening complained of symptoms suggestive of CD; 1 had recurrent abdominal pain, and 1 had abdominal distension and loose stool. Despite the severe mucosal damage on biopsy, evidence or consequences of overt malabsorption were lacking as reported in previous studies.3,23,25–27,32 In patients with CD, growth parameters such as Ht-SDS and BMI-SDS, hematologic and biochemical data such as hemoglobin, mean red cell volume, and serum albumin were not different compared with subjects without CD. As a group, decreased serum ferritin was the only sign of malabsorption in patients with CD, as commonly described in either symptomatic or asymptomatic patients.18 Therefore, clinical history and routine biochemical testing at the time of type 1 diabetes onset did not help to predict which patients had CD, indicating that serologic testing may be needed to detect subclinical disease.
It has been suggested that the exposure to gluten in genetically predisposed individuals may be a possible trigger of type 1 diabetes5,33 and other autoimmune diseases.34 Our study showed that the prevalence of autoimmune disorders, mainly thyroid autoimmunity, was in fact increased in patients with CD, but the differences were not statistically significant. This is consistent with the known increased prevalence of thyroid autoimmunity in children and adolescents with type 1 diabetes and also with the increased risk of developing chronic autoimmune thyroiditis.35
The benefits of a GFD in CD have been traditionally viewed in terms of symptom resolution; this was seen in the 2 patients who were symptomatic. BMI-SDS, hemoglobin, and mean red cell volume significantly increased with respect to basal values, but no significant changes were observed in any of the other parameters that we monitored, as previously reported by Acerini et al.32 It is interesting that the improvement of BMI-SDS after a GFD was not statistically significant in 7 patients whose CD diagnosis occurred after onset of type 1 diabetes (BMI-SDS before GFD: 0.23 ± 0.76 vs BMI-SDS after GFD: 0.49 ± 0.61; P = .49; data not shown). Therefore, we can speculate whether the BMI-SDS gain observed in patients with CD on a GFD, which was started at the time of the onset of type 1 diabetes in most of them, could be attributed to the recovery from the onset of diabetes rather than a remarkable improvement in the nutritional status as a result of the therapeutic effect of the GFD.
A number of serious complications are significantly associated with untreated CD, such as gastrointestinal malignancies, autoimmune diseases, osteoporosis, neurologic problems, and reduced fertility.3,18 As there is compelling evidence that a GFD reverses or at least decreases the risk of complications and associated conditions, we advocate that silent CD in patients with type 1 diabetes be treated with a GFD even in absence of symptoms to prevent the development of associated conditions or morbidities.
Seven of 17 patients with CD and type 1 diabetes (41.2% of cases) demonstrated EMA seroconversion and villous atrophy on biopsy during a 6-year follow-up. Although only a small portion of the 273 patients who began the study were analyzed by year 6, we showed that all new cases of CD developed within 4 years after the onset of diabetes, confirming previous reports. Maki et al12 reported a mean time between onset of type 1 diabetes and identification of positive reticulin antibodies of 13 months, and Saukkonen et al11 found that the serologic signs of CD tend to develop within 2 years after type 1 diabetes onset. We therefore suggest that screening for CD in patients with diabetes might be reduced in frequency with increasing duration of type 1 diabetes.
Eleven patients tested weakly positive for EMA in 1 or more assays and normalized during follow-up (Table 1); mucosal biopsy was performed in 4 cases, showing in all but 1 an infiltrative lesion (Marsh type 1). We speculate that these patients have either transient false-positive test, as described for reticulin antibodies11 and for EMA5 in type 1 diabetes patients, or developing or latent CD36 that has not yet manifested in significant mucosal injury. These patients require additional serologic and histologic follow-up.
We showed that >10% of children with new-onset type 1 diabetes had or developed serologic markers for CD within 6 years of the diagnosis of type 1 diabetes. Of these patients, 6.2% had definitive intestinal biopsy changes for CD. Most patients were asymptomatic and were detected because of the serologic screening. Although the majority were diagnosed at time of onset of type 1 diabetes, a significant number were diagnosed up to 4 years later. We therefore recommend that children with diabetes have screening serologic tests for CD at the time of onset of type 1 diabetes and annually for at least several years after the diagnosis of diabetes.
- ↵Fraser-Reynolds KA, Butzner JD, Stephure DK, Trussell RA, Scott RB. Use of immunoglobulin A-antiendomysial antibody to screen for celiac disease in North American children with type 1 diabetes. Diabetes Care.1998;21 :1985– 1989
- ↵Carlsson AK, Axelsson IEM, Borulf SK, et al. Prevalence of IgA-antiendomysium and IgA-antigliadin autoantibodies at diagnosis of insulin-dependent diabetes mellitus in Swedish children and adolescents. Pediatrics.1999;103 :1248– 1252
- ↵Maki M, Huupponen T, Holm K, Hallstrom O. Seroconversion of reticulin autoantibodies predicts coeliac disease in insulin dependent diabetes mellitus. Gut.1995;36 :239– 242
- ↵Maki M, Hallstrom O, Marttinen A, et al. Screening tools for use in coeliac disease. In: Auricchio S, Visakorpi JK, eds. Common Food Intolerances 1: Epidemiology of Coeliac Disease. Basel, Switzerland: Karger;1992:93– 104
- ↵Amman AJ. Selective IgA deficiency. In: Stites DP, Terr AL, eds. Basic and Clinical Immunology. 7th ed. East Norwalk, CT: Lange;1991:329– 332
- ↵Tanner JM, Whitehouse RH, Takaishi M. Standard from birth to maturity for height, weight, height velocity and weight velocity: British children, 1965. Part II. Arch Dis Child.1966;41 :613– 634
- ↵Cole TJ, Freeman JV, Preece MA. Body mass index reference curves for the UK, 1990. Arch Dis Child.1995;73 :25– 29
- ↵Page SR, Lloyd CA, Hill PG, Peacock I, Holmes GKT. The prevalence of coeliac disease in adult diabetes mellitus. Q J Med.1994;87 :631– 637
- ↵De Vitis I, Ghirlanda G, Gasbarrini G. Prevalence of coeliac disease in type I diabetes: a multicentre study. Acta Paediatr.1996;412(suppl) :56– 57
- ↵Maki M, Hallstrom O, Huupponen T, Vesikari T, Visakorpi JK. Increased prevalence of coeliac disease in diabetes. Arch Dis Child.1984;59 :739– 742
- ↵Russo PA, Chartrand LJ, Seidman E. Comparative analysis of serologic screening test for the initial diagnosis of celiac disease. Pediatrics.1999;104 :75– 78
- ↵Stern M, Teuscher M, Wechmann T. Serological screening for celiac disease: methodological standards and quality control. Acta Paediatr.1996;412(suppl) :49– 51
- ↵Bürgin-Wolff A, Gaze H, Hadziselimovic F, et al. Antigliadin and antiendomysium antibody determination for celiac disease. Arch Dis Child.1991;66 :941– 947
- ↵Barera G, Bianchi C, Calisti L, et al. Screening of diabetic children for coeliac disease with antigliadin antibodies and HLA typing. Arch Dis Child.1991;66 :491– 494
- ↵Mckenna MJ, Herskowitz R, Wolfsdorf JI. Screening for thyroid disease in children with IDDM. Diabetes Care.1990;13 :801– 803
- ↵Ferguson A, Arranz E, O’ Mahony S. Clinical and pathological spectrum of celiac disease: active, silent, latent, potential. Gut.1993;34 :150– 151
- Copyright © 2002 by the American Academy of Pediatrics