The identification of celiac disease in asymptomatic children: the Generation R Study

This study was embedded within the Generation R study, a prospective population-based cohort study from fetal life until young adulthood, described in detail previously [22, 23]. Children were born between April 2002 and January 2006 in Rotterdam, The Netherlands. At the age of 6 years, 6690 children visited the research center. During this visit, serum samples were collected from 4593 (69%) children and subsequently stored over a time period of 2.5 years [22]. In 2013, after the inclusion of the whole cohort was completed and children were 9 years old, these samples were thawed and analyzed for TG2A levels. Children and parents were not aware of TG2A determination. We excluded 20 children in whom total IgA concentrations were below the detection limit, possibly indicating IgA deficiency. Finally, 4442 children provided data on TG2A levels at 6 years of age [24, 25]. Between November and December 2013, when children were 9 years old, parents of 60 children were informed about a positive TG2A screening result at 6 years of age, and invited for diagnostic follow-up. Of these, eight children were lost to follow-up and one child was diagnosed with CD prior to the initial serology screening at 5 years of age, and therefore excluded. Of the remaining 51 children, two were on a gluten-free diet (GFD), and had been diagnosed with CD prior to the follow-up (between 6 and 9 years of age). Therefore, these children were excluded from clinical retesting (but not from the final analyses). In total, 49 children received diagnostic follow-up within 3 months after initial notification of the elevated TG2A level at 9 years of age (January to April 2014). None of them consumed a low- or gluten-free diet (GFD).

Written informed consent was obtained from all participants. Approval for the study was obtained from the Medical Ethical Committee of Erasmus MC, University Medical Centre Rotterdam, The Netherlands.

TG2A serum levels were measured using a fluorescence enzyme immunoassay (EliA Celikey IgA, Phadia ImmunoCAP 250, Phadia AB, Uppsala, Sweden). The intra- and interassay coefficient of variation (CV) was below 10 and 15%, respectively. Sera with a TG2A level of 7 U/ml or higher were considered to be positive per the manufacturer’s instructions. We further subdivided positive TG2A levels into two categories on the basis of below or above the 10 upper limit normal (ULN) of the test kit (>70 U/ml) [19, 24, 25].

Of 51 children included, two were diagnosed with CD between 6 and 9 years of age, and one child received follow-up elsewhere (Fig. 1). To assess whether clinical symptoms were different between the CD groups (Table 1), these three children were excluded from analyses (but not from the final analyses). From the remaining 48 children visiting the Sophia Children’s Hospital at the median age of 9.9 years, a standardized medical and family history was taken by a pediatric gastroenterologist. The presence of gastrointestinal complaints was assessed systematically after parents were informed about the TG2A positivity at 6 years of age. Information on gluten intake was assessed by diet questions recorded at 6 years of age prior to TG2A screening, and at the moment of follow-up at 9 years of age, by asking whether their child had been on a low- or GFD. If not, the child was considered to be exposed to a normal amount of gluten. Height and weight were measured without shoes and heavy clothing, and body mass index (BMI; kg/m²) was calculated. Age- and sex-adjusted standard deviation scores (SDS) were obtained using Dutch reference growth curves [26]. Prior data on height and weight as recorded in the community health centers were included. A delay in linear growth over time was defined as decrease of >0.5 SD height-for-age relative to the normal SD line. The second TG2A test at 9 years of age was performed using the same fluorescence enzyme immunoassay (exact same kit). In addition, anti-endomysial antibody (anti-EMA) levels were measured, and children were genotyped for HLA DQ2.2, DQ2.5, and/or DQ8. Serological criteria for recommending an upper gastrointestinal (GI) endoscopy are summarized in Fig. 1. All GI endoscopies were performed by two experienced pediatric gastroenterologists (MG or JCE) between March and December 2014. Mucosal biopsies were taken from both the proximal (including the bulb) and distal (2nd or 3rd) part of the duodenum as recommended [27‐29]. Enteropathy was graded according to the Marsh-Oberhuber criteria [12, 13], which can vary from intraepithelial lymphocytosis (IEL; ≥30 lymphocytes/100 enterocytes; grade 1) to more extensive lesions including crypt hyperplasia (grade 2), and various degrees of villous atrophy (grade 3a partial; grade 3b subtotal, grade 3c total). Enteropathy was defined as having Marsh II or greater [19].

Endomysial antibodies (EmA) of IgA isotype were determined by indirect immunofluorescence using commercial monkey esophagus slides, according to the manufacturer’s instructions (Inova Diagnostics, San Diego, CA) (S1).

Presence of CD-associated HLA-DQ haplotypes DQ2.2 (DQA1*02/DQB1*02), DQ2.5 (DQA1*05/DQB1*02) and DQ8 (DQA1*0301/DQB1*0302) was determined by EUROArray, according to the manufacturer’s instructions (Euroimmun AG, Lübeck, Germany) (S1) [30].

Of 4442 screened children (median age, 6.0 years), 60 (1.4%) had increased TG2A levels (≥7 U/ml), of whom 31 children had TG2A levels above 10 ULN (>70 U/ml). Of 60 children, eight were lost to follow-up, and one child was diagnosed with CD at 5 years of age (prior to the initial serology screening) and therefore excluded from analyses. Two children (4%) had received a CD diagnosis in the period preceding retesting (between 6 and 9 years of age) and consumed a gluten-free diet (GFD). Here, CD was detected based on gastrointestinal symptoms by routine clinical care in the Netherlands, of whom one diagnosis was confirmed by biopsy (Marsh 3a/3b) (Fig. 1).

Of 48 children who were retested in the Sophia Children’s Hospital, 31 (65%) were positive for TG2A and EMA, carried HLA DQ2.2, DQ2.5, or DQ8, and were advised to undergo gastrointestinal endoscopy. Of these, 20 had TG2A levels ≥10 ULN (Fig. 1). Of 26 children who underwent GI endoscopy, 20 had Marsh grade 3 [(3a n = 13), 3b (n = 7)]; three children had Marsh grade 2; and three children had Marsh grade 1. Of 23 children with Marsh grade 2–3 lesions, 16 (70%) had TG2A levels ≥10 ULN. All three children with Marsh grade 1 lesions had TG2A levels ≥10 ULN and were advised to continue on a gluten-containing diet, and to consider a second serology and/or intestinal biopsy when clinical complaints compatible with CD occur. A total of ten children were HLA DQ2.2, DQ2.5, and DQ8 negative, or carried the genetic risk type but lacked EMA and TG2A positivity, and were therefore considered not to have CD. The second serology was not conclusive in seven children: they carried the genetic risk type, but had TG2A concentrations <3 ULN, or TG2A levels were negative in accordance with a positive EMA, whereas the initial screening TG2A test result at 6 years of age was positive. Furthermore, these children mentioned gastrointestinal complaints (abdominal pain), and had a first-degree family history of CD (in contrast to the ten children who were considered not to have CD). Hence, these children were advised to be serologically retested in 6–12 months. During the period of follow-up (January 2014 to April 2017), none of the ten children received a CD diagnosis; only one child occasionally mentioned complaints of abdominal bloating, and was on a low-gluten-containing diet (but not a gluten-free diet). The remaining children did not mention CD-associated symptoms while consuming a gluten-containing diet.

In total, 31 of 51 (60.8%) children were considered to have CD, of whom two had developed symptomatic CD within 3 years after the screening test was performed; 10/51 (19.6%) needed follow-up because of inconclusive serology (n = 7) or negative intestinal biopsies (n = 3, of whom all three had levels exceeding 10 ULN); and 10/51 (19.6%) were considered to not have CD (Fig. 1). Thus, the positive predictive value of TG2A screening was 61%, (95% CI 49–75), to detect subclinical CD and 81%, (95% CI 70–91) to detect both potential and subclinical CD.

All 48 children consumed a normal gluten-containing diet at 6 years of age, as well as at the visit to the pediatric gastroenterologist at 9–10 years. Of 28 CD cases who provided data on medical history and physical examination at 9 years of age, nine (32%) children were truly asymptomatic relative to one (10%) of the children without CD diagnosis (p = 0.17). Abdominal pain was reported most frequently (57%), followed by constipation (43%), nausea (29%), and diarrhea (21%), but no significant differences in gastrointestinal symptoms were observed between children with and without CD diagnosis (p = 0.12) (Table 1). In addition, no significant differences were observed in the presence of anorexia, irritability, food allergy, lactose intolerance, eczema, and absenteeism from school between children with and without CD diagnosis (data not shown), nor in family history of CD (Table 1). However, CD cases weighed less (median 29.3 kg; range, 19.2–61.6, p = 0.04) than children who did not have CD (median 36.8 kg; range, 25.9–69.8). In addition, their BMI was lower (mean BMI SDS −0.49, SD 0.92, p = 0.04) compared to children without CD (mean BMI SDS 0.47, SD 1.37) No significant differences in height were observed between children with and without CD diagnosis (p > 0.32) (Table 1). Associations were not substantially different in children without gastrointestinal symptoms [S2].

In addition, no significant differences in gastrointestinal symptoms were observed between children who were TG2A negative (<7 U/ml) and strongly positive (≥10 ULN), but children having TG2A levels exceeding 10 ULN weighed less (median 29.1 kg, range, 19.2–44.0) and had a lower BMI (mean BMI SDS −0.51, SD 0.85), than children who were TG2A negative (median 35.3 kg, range, 25.9–69.8 and SDS BMI 0.21, SD 1.16 resp.) [S3].

We observed a high variability in serum TG2A levels at 6 and 9 years of age (Fig. 2). Of 29 CD cases, 20 children (69%) had high TG2A levels (≥10 ULN) at 6 years of age, which increased to a higher level in the majority (16/20) of cases at 9 years of age. Of ten children who needed follow-up at 9 years of age because of negative serologies, four children (40%) had TG2A concentrations exceeding 10 ULN levels at 6 years of age. Of six children who carried the genetic risk type, but did not have CD, one child had TG2A levels exceeding 10 ULN at 6 years of age (Fig. 2).

A positive linear tendency was observed for TG2A levels in predicting CD diagnosis: children having high TG2A levels (≥10 ULN) at 6 and 9 years of age were more frequently diagnosed with CD than children having TG2A-positive concentrations below <10 ULN (OR 7.7; 95% CI 2.2–27.4) and [OR 10.0 (1.07–93.4) resp.] [S4]. Second, we observed that TG2A levels at 6 or 9 years of age were not related to the presence of severe intestinal enteropathy (Marsh 3) [S5] (Fig. 3).

At the child’s age of 9 years, parents were asked by questionnaire whether their child had ever received a doctors diagnosis of CD. Of 4382 TG2A-negative children at 6 years of age, nine (0.2%) received a doctors diagnosis of CD between 6 and 9 years of age.

The premise of the study was the identification of CD in asymptomatic children. Three children in our study developed symptoms following screening, thereby triggering CD diagnosis in routine clinical practice. One of the strengths of our study is that we used a prospective study design, with a similar protocol for the diagnostic process for all screened TG2A-positive children. Furthermore, all screened TG2A-positive children were of the same age, and analyzed 3 years later by the same laboratory and same test kit. Because all participants were unaware of TG2A determination, we were able to study the development of TG2A levels over 3 years of time. Ten children with negative TG2A did not have CD, despite TG2A positivity at 6 years of age. Explanations for the contradictory test results at 6 and 9 years may be that TG2A levels were transiently high at 6 years, test results at 6 years were false positive (i.e., 1-specificity), or gluten intake decreased over time. Intriguingly, on top of the 1.4% of children with positive TG2A levels at 6 years of age (of whom 61% received a subclinical CD diagnosis), 0.2% of TG2A negative children developed CD within 3 years after the negative TG2A screening result at 6 years of age. The median age of retesting was 9.9 years, with a range between 8.4 and 11.4 years. This large age range is related to the inclusion period of 2.5 years. Another strength of our study is that TG2A measurements were performed in combination with assessment of EMA, genetic risk types, and biopsy specimens at the same time of second serology assessment. Last, all biopsies were evaluated by expert pathologists by using the Marsh–Oberhuber criteria to classify the degree of enteropathy. Nevertheless, several limitations should be taken into account as well. First of all, five children in the CD group were considered to have subclinical CD without biopsy results. Therefore, we did not fully meet the ESPGHAN criteria for CD diagnosis. However, three of them had levels exceeding 10 ULN, they were all EMA positive, and carried the genetic risk type. Hence, CD diagnosis is highly likely. Second, symptoms were assessed only among those with TG2A positivity at 9 years of age, after parents were informed about the initial screening TG2A status, which might have induced bias. However, the presence of gastrointestinal symptoms was evaluated systematically during routine clinical evaluation, preceding second serological (TG2A) assessment, additional EMA assessment, genetic risk typing, and gastrointestinal endoscopies at 9 years of age. Hence, symptom assessment was independent of the outcome, making bias less likely. Furthermore, symptoms were assessed by one experienced gastroenterologist minimizing inter-observer bias. We did not specifically ask for the presence of joint pains or neurological symptoms, but these complaints were not reported on the general question “any complaints”. Hence, the occurrence of joint pains and neurological symptoms seems unlikely.