OBJECTIVE: Antibody screenings and diagnosis of celiac disease (CD) among children with type 1 diabetes have suggested that a considerable proportion of children with CD may, in fact, have preclinical (undiagnosed) symptoms. We aimed to test if a questionnaire would lead to significant case finding in an unselected population of 8- to 9-year-old children.
PATIENTS AND METHODS: The study population included 9880 children aged 8 to 9 years. Before the study, 13 children from the study population were known to have CD. We developed a questionnaire on the basis of 5 simple items suggestive of CD and mailed the questionnaire to the families of all children in the study population who resided in the County of Funen, Denmark. In total, 7029 respondents returned the questionnaire (70%); among them, 2835 children had 1 or more symptoms. These children were invited for a blood test to determine their human serum immunoglobulin A (IgA) anti–tissue transglutaminase antibody (anti-tTG) levels.
RESULTS: Of the 1720 children who were tested for the human serum IgA anti-tTG, 24 participants had a positive result (range: 20 to >150 U). Seventeen of these children underwent an upper endoscopy procedure. Fourteen children had histologic signs of CD (Marsh classification stage III). Fourteen children met the diagnostic criteria for CD. The prevalence proportion of patients who were newly diagnosed with CD was 0.14% (95% CI: 0.08–0.24) (14 of 9967), and the estimate of the minimum total prevalence proportion of children with CD was 0.27% (95% CI: 0.18–0.39) (27 of 9980). The maximal prevalence proportion of patients with newly diagnosed CD was 1.22% (95% CI: 0.76–1.90) (21 of 1720), including those participants who had a positive anti-tTG result but not a final diagnosis of CD. The ratio of known to minimally symptomatic CD was ∼1:1.
CONCLUSION: A number of preclinical and low-grade symptomatic patients with CD may be identified by their responses to a mailed questionnaire.
WHAT'S KNOWN ON THIS SUBJECT:
Unselected screening studies have shown that some children with CD may have low-grade symptoms or silent forms of the disease. The rate of low-symptomatic CD cases may be 6 to 7 times that of symptomatic cases.
WHAT THIS STUDY ADDS:
Physicians may be able to diagnose children with CD by administering a simple 5-question survey; determining a patient's IgA tTG levels; and performing a duodenal biopsy. This process may substantially increase the number of children who are diagnosed with CD; thereby doubling the prevalence proportion.
Because of the diversity in its clinical presentation, the detection of celiac disease (CD) may require different approaches. Patients with classical symptoms such as diarrhea or failure to thrive are usually given the diagnosis because of the severity of symptoms. Doctors may detect the condition through a clinical evaluation of those patients who have mild but still clinically overt symptoms.1,2 Unselected screening procedures that are based on determining celiac antibody levels in serum may help to uncover both symptomatic and nonsymptomatic cases.2,–,5 A ratio of 1 known to 3 to 7 undiagnosed cases has been observed.3,–,5 Unselected population-based screening has been suggested as a routine clinical procedure, but it is not currently advocated.6,7 Therefore, detection according to risk groups may be more effective. For instance, the prevalence of CD is high in first-degree relatives (5%–11%) because of strong heritability.8,–,11 CD is also associated with other diseases, particularly autoimmune diseases such as type 1 diabetes. Thus, antibody screening for CD in these high-risk groups of patients may lead to the detection of a number of patients with previously undiagnosed CD.12,–,24
In a prospective 2-year follow-up study of children with CD and type 1 diabetes, we observed a prevalence of CD of 12.4%.16 The majority of the children with CD had low-grade symptoms that they were not particularly aware of until they improved after starting a gluten-free diet. Studies of children with abdominal pain have only yielded a prevalence of CD of 1% to 3.6%.25,–,27 However, research has shown that searching for CD in groups of children who have specific signs, such as anemia, has reportedly enabled detection of unknown cases.28,29 We hypothesized that CD may not always be identified because the children experience low-grade symptoms for which the parents do not seek medical advice. To our knowledge, no population-based case-finding study of CD from a questionnaire has been published thus far. The aim of our population-based study was to determine if a questionnaire could identify children with low-grade symptoms who would otherwise be classified as having silent CD.
PATIENTS AND METHODS
A questionnaire on the central symptoms of CD was mailed to the families of all of the 8- and 9-year-old children in the County of Funen, Denmark (∼478 437 inhabitants). The children were invited to participate in the study, which comprised 5 diagnostic steps.
Step 1. a simple 5-item questionnaire; if there were 1 or more positive answers, the child was invited to proceed to step 2;
Step 2. serologic testing of human serum anti–tissue transglutaminase immunoglobulin A (IgA) antibody (anti-tTG). If positive antibody determination existed, the child was invited to proceed to step 3;
Step 3. clinical examination, followed by step 4;
Step 4. an upper endoscopy under general anesthesia, including biopsies from the duodenum; if the findings were consistent with CD, the child proceeded to step 5;
Step 5. a gluten-free diet.
All 8- and 9-year-old children living in the County of Funen who were born between September 1, 1997, and August 31, 1999 (n = 10 220) were identified from the Danish Civil Registration System by using a unique personal identification code.30,31 Registrars at the Odense University Hospital allowed us to identify all children who had been diagnosed with CD before the start of the study. Clinical information about these children was obtained retrospectively from hospital records. No primary care pediatricians or local hospitals in the county performed endoscopies of children; hence, no cases were missed. The Regional Committee of Southern Denmark on Biomedical Research Ethics approved this study.
A short questionnaire was developed that covered 5 of the most common symptoms of CD (Table 1).32,–,35 The questionnaire was validated by 22 test interviews of randomly selected parents of children in the above-mentioned age group who were admitted to the Odense University Hospital emergency department with minor injuries. Furthermore, we carried out a pilot study including the families of 198 children aged 5 to 9 years which led to no additional changes in the questionnaire.
The questions referred to the presence of recurrent abdominal pain,36 chronic diarrhea, constipation, and a lack of increase in height and weight according to the parents' opinion. Respondents were instructed to indicate their answers by selecting yes, no, or do not know. The “do not know” response was regarded as a no. The parents were asked to provide consent for their children to participate in additional examinations. Children who had 1 or more symptoms and whose parents had consented to additional examinations were called in to have blood samples drawn. The questionnaire also provided general information for both the children and parents. Reminder letters including both the questionnaire and general information were mailed twice to all nonrespondents.
Blood samples were collected from the study participants at 2 hospitals in the County of Funen and analyzed at Odense University Hospital. Human anti–red cell IgA anti-tTG was measured by using an enzyme-linked immunosorbent assay technique (Inova Diagnostics, Inc, San Diego, CA). The antibodies were measured in units (cutoff of 20 arbitrary units). The assay was accredited according to ISO/IEC 17025 (DANAK, Copenhagen, Denmark).
Human Leukocyte Antigen Typing
Human leukocyte antigen (HLA) typing of HLA-DQA1* and HLA-DQB1* was performed by using a polymerase chain reaction–amplification technique with sequence-specific primers (Invitrogen Corporation, Carlsbad, CA).
Clinical and Laboratory Investigations
Children with positive anti-tTG test results were invited for a consultation at the pediatric outpatient clinic at Odense University Hospital. The children were offered an upper endoscopy procedure under general anesthesia and instructed not to change their normal gluten-containing diet. Endoscopies were conducted by a pediatric gastroenterologist. Biopsies were obtained, including 4 biopsies from the duodenum; 1 each from the duodenal bulb, antrum, corpus, and fundus of the stomach; and 2 from the esophagus. The histopathology of the duodenal biopsies was classified according to the modified Marsh classification criteria.37 All biopsies were evaluated by the same pathologist, who was not blinded to the test results. During anesthesia, blood samples were collected for reanalysis of anti-tTG levels and analysis of anti-gliadin IgA (IgA AGA), antigliadin IgG antibody (IgG AGA), antiendomysium IgA antibody (EMA), HLA-DQ2/DQ8, ferritin, cobalamin, folic acid, liver enzyme, and electrolyte levels and hematologic parameters. If the histopathology of the biopsies was compatible with CD, an experienced dietician instructed the family to start a gluten-free diet. A clinical consultation after 1 to 3 months of a gluten-free diet included retesting the CD antibodies and other blood tests that had previously shown abnormal results. IgA and IgG anti–deamidated gliadin antibodies were measured in all blood samples. The total IgA level was measured in the blood samples as being negative for anti-tTG and positive for deamidated gliadin antibodies.
The prevalence proportion of CD was defined as the number of children who were diagnosed with CD divided by the number of children in the cohort. Different prevalence proportions were calculated by changing the numerator and/or denominator. The numerator was either the total number of children who were diagnosed with CD or the children who were diagnosed in our study. The denominator was the total number of participants included in the study, the number of respondents, the number of children with symptoms, or the number of children who proceeded to provide a blood sample. Ninety-five percent confidence intervals (CIs) of the prevalence proportion were calculated under the assumption that the data were binomially distributed.
To analyze selection bias, we used a logistic regression model of the association between the presence of a symptom (the dependent dichotomized variable) and the number of families who responded to the primary questionnaire or subsequent reminders (the independent continuous variable). We used a logistic regression model to analyze the association of the number of symptoms and the children diagnosed with CD. The number of symptoms was dichotomized to 1 or more than 1 symptom(s) as the dependent variable, and the presence of CD was the independent variable.
We used the Stata statistical package (Stata Corp, College Station, TX) for statistical analyses.
The study population comprised 9980 children (Fig 1). A total of 7029 families completed the questionnaire (70%). One or more symptoms were checked off by 2834 (40%) respondents, with abdominal pain being the most frequent (33% of all respondents) (Table 2). Among the 24 children with a positive anti-tTG result, 17 underwent an upper endoscopy procedure. Eleven children had signs of CD in the duodenum that corresponded to Marsh classification stage III, of whom 2 had Marsh stage I signs and 1 had Marsh stage II signs (8 girls [62%]). Those patients who had Marsh stage I to II signs in the duodenum also had histologic signs that corresponded to Marsh stage IIIB in the duodenal bulb. Thirteen patients had positive EMA test results. Positive results were found in the screenings of 8 patients for HLA-DQ2, 3 for DQ8, and 3 for DQ2 and DQ8. One patient had a Giardia lamblia infestation, and no rebiopsy was performed because the anti-tTG disappeared after treatment with antibiotics. Two patients had normal biopsy results. One of the 3 patients with no histologic signs of CD had positive EMA test results. All of the children had weak symptoms that did not affect their daily lives, and no other symptoms of CD were present. None of the children who had positive serology and came to a consultation belonged to an at-risk group, except for 1 child who had a grandparent with CD. Twelve of the blood sample results were positive for the IgG anti–deamidated gliadin antibody but were negative for the anti-tTG. All 12 samples had normal levels of total IgA in serum.
All of the children who proceeded to step 5 stated that they maintained a gluten-free diet, and all were diagnosed with CD and achieved clinical remission. Despite the clinical remission, 2 children started a normal diet, and their anti-tTG levels did not change. Among the other children with CD, their anti-tTG levels disappeared or declined. Some of the parents reported that their child's mood and behavior improved after initiation of a gluten-free diet, although this symptom was not mentioned at the first consultation.
The study population included 13 children who were diagnosed with CD before the study began. Seven patients had clinical overt CD, and 6 were diagnosed from a screening of high-risk groups (4 patients with type 1 diabetes, 1 with Down syndrome, and 1 with autoimmune hepatitis). At the end of the study period, 27 of the 9980 children had CD, that is, the prevalence proportion of newly diagnosed patients was 0.14% (95% CI: 0.08–0.24) (14 of 9967) and the estimate of the total prevalence proportion was a minimum of 0.27% (95% CI: 0.18–0.39) (27 of 9980). On the basis of the 14 new patients who were diagnosed during the study period, we estimated a prevalence proportion of 0.20% (95% CI: 0.11–0.33) (14 of 7029) among respondents who answered the questionnaire, 0.49% (95% CI: 0.27–0.83) (14 of 2835) among children with 1 or more symptoms, and 0.81% (95% CI: 0.45–1.36) (14 of 1720) among those who proceeded to blood-sample testing. The maximal prevalence proportion of newly diagnosed CD was 1.22% (95% CI: 0.76–1.90) (21 of 1720), including all patients with positive anti-tTG levels but excluding those with a normal biopsy result. Because 14 new cases were identified, the ratio of minimally symptomatic CD to known CD was ∼1:1.
We used a logistic regression model to analyze the respondents' selection bias on the basis of the presence of symptoms. This model resulted in an odds ratio of 0.66–0.89 (P < .05, Wald test) when the first and second reminder was compared with the primary request (Table 2). Thus, symptomatic participants had a higher probability for answering the primary request than the follow-up requests.
We used a logistic regression model to analyze whether the number of symptoms increased the likelihood for CD. The odds for CD increased slightly with an increasing number of symptoms, as shown by the β coefficient of 1.05 (P < .05) in the logistic regression model.
The clinical examinations of those participants who had positive anti-tTG levels showed that weight and height for all patients were in the normal range of ±2 SDs. Weight was 0.25 to 1.00 SD lower than height for 2 patients, whereas weight was 0.25 to 1.00 SD higher than height for 5 patients. Five children had serum ferritin levels that were below normal. Two children had low levels of serum folic acid, and 1 of the 2 children also had a low level of serum ferritin. All of the other blood analyses, except for the CD-related antibody tests (see “Patients and Methods”), were normal.
In this study, we tested the hypothesis that CD may be unidentified because of low-grade symptoms that are otherwise characterized as silent or subclinical CD. We observed a minimal prevalence of low-symptomatic CD of 0.14% (14 of 9967) in a group of 8- to 9-year-old children. All patients who were identified as having CD had recurrent abdominal pain, and the majority of them also had constipation. Before their participation in this study, no more than 3 families had seen a doctor because of these complaints. Hence, most children with low-grade symptoms had not previously obtained medical attention.
Families with symptomatic children were more likely to respond to the primary request (odds ratio: 0.66–0.89) compared with nonsymptomatic children. The primary questionnaire was answered by 56% of families of children with abdominal pain, which was equal to the response rate from the total study population. A high proportion of families with children who were diagnosed with CD answered the primary questionnaire (12 of 14). This result indicates a selection bias of both symptomatic children and children with CD in relation to their response to the questionnaire. Hence, there were likely fewer symptomatic children among the nonrespondents. The odds for CD increased with the number of symptoms but only with a factor of 1.05. This finding supports the inclusion of symptomatic children irrespective of the number of symptoms.
Recurrent abdominal pain was the entrance into the study for the majority of children. One third of the respondents (2297) reported having abdominal pain, which was higher than previously reported in Denmark38,39 and likely resulted because there were fewer restrictions on the item of abdominal pain in our study.
A limitation of this study was the fact that we did not obtain blood samples and determine IgA anti-tTG levels in the nonsymptomatic children. A study design that included an unselected screening would have given a more clear indication of the gains of the questionnaire approach. Neither the reliability nor the validity of the questionnaire was tested. Hence, there may be symptomatic children who did not report their symptoms. Despite the low detection limit of IgA anti-tTG antibody40,41 some children with CD may have missed being diagnosed in this study. Strengths of our study were the large and well-defined population and the high response ratio of 70%.
Previous Danish studies of the prevalence proportion of CD among children have indicated extremely low levels (0.02%) compared with those in neighboring countries.42 We recently performed a national register–based study based on hospital records, which showed a prevalence proportion at 0.07%.43 The majority of children were diagnosed with CD in the first 3 years of life. In addition, we have estimated a high prevalence proportion (12.4%) of CD in diabetic children.16 The results of our study suggest that the prevalence proportion of CD is higher than previously demonstrated. The results are closer to previous results from studies of the prevalence proportion in neighboring countries. However, a recent Swedish study among children born in 1993 during the CD epidemic revealed a prevalence proportion of 3%.44 This difference between Denmark and Sweden is high considering that the proportion in our study is a minimum prevalence proportion. Differences in feeding practices in the first year of life may be a reason among several possibilities that explain this difference.45
Different strategies for finding cases of CD may be used. In a Finnish study of gastrointestinal complaints, a questionnaire was delivered by the school nurse to 422 schoolchildren aged 10 to 11 years in a single school; the response ratio was 96%.27 The inclusion criteria for additional examinations (including blood tests) were abdominal pain, diarrhea, and constipation, as well as a self-reported milk intolerance. A total of 110 children were examined, which led to 5 participants being diagnosed with CD. The response ratio was very high, which could be explained by the personal contact at the initiation of the study. Hence, direct contact with the children may increase the participation rate.
In another study, researchers recruited symptomatic children and adults with CD from across the United States. The prevalence of CD among 2- to 18-year-old children (n = 1326) (symptoms of abdominal pain, constipation, or diarrhea) was 4%.2 Authors of another study that involved primary care pediatricians in Italy used a similar strategy to study 240 children (mean age of the pediatric population: 5.2 years) with symptoms of CD.29 The prevalence proportion ranged from 5.2% to 23.5% depending on the type of symptom. In another Italian study, an educational strategy was used with the medical profession in an attempt to increase the diagnosis of children with undiagnosed CD.1 After an educational intervention and examinations of 447 children, the prevalence proportion of 9- to 14-year-old children was increased from 0.12% to 0.17%. It is obvious that this strategy requires that the patient or caregiver seek medical advice. In conclusion, varying strategies have been tried without leading to a clear indication of the most effective approach for finding patients. An increased awareness of CD in the general population is likely the most effective strategy to increase the ratio of recognized versus unrecognized cases of CD.46
Unselected screening studies are based on serologic testing followed by biopsy and, hence, find cases of CD independently on the basis of presence of symptoms. In a Dutch study, a large population of 2- to 4-year-old children was screened. Half of the children with CD had previously unrecognized CD with typical symptoms such as abdominal pain, diarrhea, constipation, or lassitude.5 In Sweden, a screening study among 2.5-year-olds attending a routine health screening showed a 1:1.6 ratio of asymptomatic to symptomatic children.47 These studies indicate that some of the children who are classified as having silent CD have no symptoms; thus, these patients can only be detected through screening. However, some of the children with the silent form of CD may be diagnosed by a case-finding strategy similar to the one used in our study.
The strategy of a questionnaire in the present population-based study was feasible for identifying children with low-grade CD. The prevalence proportion of CD in 8- to 9-year-old children was at least 0.27% (95% CI: 0.18–0.39). Abdominal pain and constipation were the main symptoms among the newly diagnosed patients with CD. The ratio of known to minimally symptomatic CD was ∼1:1, which is higher than that in screening studies and may indicate that the case finding was not complete. Unselected screening has not previously been considered to be cost-effective.6,48 Therefore, a comparison of the cost-effectiveness of a population-based case-finding strategy and unselected screening is clearly needed.
This study was supported with grants from the Egmont Foundation, Lundbeck Foundation, A. P. Møller and the Chastine Mc-Kinney Møller Foundation, Institute of Clinical Research at the University of Southern Denmark, and the A. L. and D. Rasmussen Foundation.
We thank Henrik Støvring, PhD, for assistance with the statistical analysis.
- Accepted September 18, 2009.
- Address correspondence to Steffen Husby, MD, DMSc, Hans Christian Andersen Children's Hospital, Sdr Boulevard 29, DK-5000 Odense C, Denmark. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- CD =
- celiac disease •
- Ig =
- immunoglobulin •
- anti-tTG =
- human serum IgA anti–tissue transglutaminase antibody •
- HLA =
- human leukocyte antigen •
- EMA =
- IgA antiendomysium •
- IgA AGA =
- antigliadin IgA antibody •
- IgG AGA =
- anti-gliadin IgG antibody •
- CI =
- confidence interval
- Beattie RM
- Greco L,
- Romino R,
- Coto I,
- et al
- Dolinsek J,
- Urlep D,
- Karell K,
- Partanen J,
- Micetic-Turk D
- Hansen D,
- Brock-Jacobsen B,
- Lund E,
- et al
- Luft LM,
- Barr SG,
- Martin LO,
- Chan EK,
- Fritzler MJ
- Ventura A,
- Facchini S,
- Amantidu C,
- et al
- Malig C
- Apley J,
- Naish N
- Wolters V,
- Vooijs-Moulaert AF,
- Burger H,
- et al
- Hansen DG,
- Nielsen C,
- Lillevang ST,
- Husby S
- Toftedal P
- Carlsson AK,
- Axelsson IE,
- Borulf SK,
- Bredberg AC,
- Ivarsson SA
- Copyright © 2010 by the American Academy of Pediatrics