Celiac disease (CeD) is associated with psychopathology in children. It is unknown whether this association is present in children with celiac disease autoimmunity (CDA) identified by screening. We examined the associations between subclinical CDA and emotional and behavioral problems in children without previous CeD diagnosis.
In a population-based cohort study of 3715 children (median age: 6 years), blood titers of tissue transglutaminase autoantibodies were analyzed. CDA was defined as a measurement of tissue transglutaminase autoantibodies ≥7 U/mL (n = 51). Children with previous CeD diagnosis or children on a gluten-free diet, were excluded. The Child Behavior Checklist (CBCL) was filled in by parents and was used to assess behavioral and emotional problems of children at a median age of 5.9 years. Multiple linear regression models were applied to evaluate the cross-sectional associations between CDA and CBCL scores. Sensitivity analyses were done in a subgroup of children who were seropositive carrying the HLA antigen risk alleles for CeD.
In basic models, CDA was not associated with emotional and behavioral problems on the CBCL scales. After adjustment for confounders, CDA was significantly associated with anxiety problems (β = .29; 95% confidence interval 0.02 to 0.55; P = .02). After exclusion of children who did not carry the HLA-DQ2 and/or HLA-DQ8 risk alleles (n = 4), CDA was additionally associated with oppositional defiant problems (β = .35; 95% confidence interval 0.02 to 0.69). Associations were not explained by gastrointestinal complaints.
Our results reveal that CDA, especially combined with the HLA-DQ2 and HLA-DQ8 risk alleles, is associated with anxiety problems and oppositional defiant problems. Further research should be used to establish whether behavioral problems are a reflection of subclinical CeD.
Diagnosed celiac disease (CeD) has been linked with several emotional and behavioral problems in children, including anxiety, depression, and aggressive behavior. However, whether this association is already present in children with subclinical CeD identified by screening is unknown.
Screening-based celiac disease autoimmunity is associated with anxiety and oppositional defiant problems in children with subclinical CeD identified by screening, independent of gastrointestinal complaints.
Celiac disease (CeD) is a common immune-mediated chronic disease with an estimated prevalence of nearly 1% in the general pediatric population.1 However, ∼80% of patients are not diagnosed and, thus, untreated.2–4 This can partially be explained by the heterogeneous clinical presentation. Complaints can vary from diarrhea and abdominal distention to nonspecific symptoms such as fatigue or anemia.1 Also, extraintestinal manifestations have been reported,1,5 including emotional and behavioral problems in childhood.6–13 The mechanism of psychological symptoms in CeD remains unclear. Because studies have revealed that psychological symptoms may also emerge before CeD diagnosis in adults, psychological symptoms could only be partly a consequence of disease awareness and the gluten-free diet.11,14 Mechanisms underlying psychological symptoms in unknown CeD may be somatic complaints because of unrecognized CeD,15 nutrient malabsorption,16 or alterations in the gut-brain axis.17,18 Although psychological symptoms in early life predispose to psychopathology in later life and contribute to a lower quality of life,19,20 previous studies have revealed that CeD-related psychopathology resolves under treatment with a gluten-free diet.7,21,22 This emphasizes the need for early identification and, thus, treatment of CeD.19,20,23 However, information on the effect of diagnosis and treatment in patients with celiac disease autoimmunity (CDA) identified by screening is scarce.23,24 To our knowledge, only 1 study has been conducted on psychological manifestations in children with subclinical CeD identified by screening, and this study revealed contradictory results.13
Therefore, we aimed to investigate whether screening-identified CDA was associated with emotional and behavioral problems in children at the age of 6 years.
Methods
Study Design and Subjects
This study was embedded in the Generation R Study, a population-based, prospective cohort study consisting of 9778 pregnant mothers and their offspring in the region of Rotterdam, Netherlands. Children were born between April 2002 and January 2006 and followed from fetal life until adolescence, as described in detail previously.25,26 At 6 years of age, 6690 children visited the research center. Tissue transglutaminase autoantibody (TG2A) titers were available for 4442 children (median age of 6 years [SD 0.55; range 4.93−9.12]). One child per twin pair (n = 59) and children who consulted a doctor because of CeD (n = 10) or were on a gluten-free diet without CeD diagnosis (n = 6) were excluded (Fig 1). Of the remaining 4365 children with TG2A titers available, parents of 3715 children completed the questionnaire about child mental health. Thus, the final sample consisted of 3715 children for analysis. However, the number for analysis for separate Childhood Behavioral Checklist (CBCL) scales was slightly lower because of missing values. Written consent was obtained from parents of all participants. Approval for the study was obtained from the Medical Ethical Committee of Erasmus University Medical Center (Rotterdam, Netherlands).
CDA
CDA was defined as a single positive measurement of TG2A without the presence of any classic CeD symptoms. TG2A measurement was considered positive at ≥7 U/mL, according to the manufacturer’s instructions (Phadia AB, Uppsala, Sweden). At the children’s visit to the research center at 6 years of age, blood samples were drawn and stored at room temperature before being transported to the regional laboratory for storage at −80°C.25 In 2013, when follow-up of the whole cohort was completed, TG2A titers were analyzed by using a fluorescence enzyme immunoassay (EliA Celikey IgA, Phadia ImmunoCAP 250; Phadia AB) at the Department of Immunology of Erasmus University Medical Center. Intra- and interassay coefficients of variability were <10% and 15% respectively.
Measurement of TG2A was performed on a screening basis. Children and caregivers were not aware of the measurement of TG2A. If TG2A titers resulted positive in children without a previous diagnosis of CeD, children were referred to the pediatric gastroenterology department for further follow-up.27
Emotional and Behavioral Problems
Emotional and behavioral problems were assessed by using the CBCL for children age 1.5 to 5 years (median age of 5.9 years).28 The CBCL is a questionnaire that includes 99 items describing child behaviors that are to be completed by the primary caregiver (89% mothers in the current study). CBCL items can be scored on 2 broadband scales: internalizing and externalizing problems. These scales are used to capture emotional and behavioral problems, respectively. The broadband scales cover more specific syndrome scales. The internalizing composite includes the emotionally reactive, anxious and/or depressed, somatic complaints, and withdrawn syndrome scales. The externalizing composite covers attention problems and aggressive behavior. The CBCL score is also used to evaluate Diagnostic and Statistical Manual of Mental Disorders (DSM)–oriented scales, which are based on criteria of disorders described in the DSM. DSM-oriented scales include the affective, anxiety, pervasive development, attention-deficit/hyperactivity, and oppositional defiant problems scales. These scales help to relate the assessment of emotional and behavioral problems to DSM categories. DSM scales are more clinically oriented but are less scientifically substantiated.28 Emotional and behavioral problems were measured on both the empirical syndrome scales and the DSM scales. The CBCL is a widely used instrument and has been demonstrated to have good reliability and validity.28
Covariates
Age and anthropometric data were collected during the children’s visit to the research center. BMI was calculated. We obtained age-adjusted standard deviant scores (SDSs) using Dutch reference growth curves (Growth Analyzer 3.0; Dutch Growth Research Foundation, Rotterdam, Netherlands).29 The HLA antigen risk profile for CeD was defined as the presence of HLA-DQ2.5 (DQA1*05/DQB1*02), HLA-DQ2.2 (DQA1*02/DQB1*02), or HLA-DQ8 (DQA1*0301/DQB1*0302). A genome-wide association scan (Illumia 610K) of child DNA was taken from cord blood samples. Genotyping was done by using EUROarray according to the manufacturer’s instructions (Euroimmun AG, Lübeck, Germany). The tag single-nucleotide polymorphism (SNP) typing method was used to determine if a child carried the HLA-DQ2 (rs2187668, rs2395182, rs4713586, and rs7775228) and/or HLA-DQ8 (rs7454108) allele, as described in detail previously by Monsuur et al30,31 Sex and birth weight were obtained from obstetric records.32 Ethnicity was based on the country where parents were born, according to criteria used in a previous study.33 Data on maternal income and education level were obtained at enrollment through questionnaires. Classification of income and educational level was based on the classification of Statistics Netherlands.34,35 Gastrointestinal complaints over the past 3 months were assessed in children at a median age of 6 years by using parent-reported questionnaires. Obstipation was defined as present if the following symptoms of Rome III were reported in the past year for at least 2 weeks in a row: defecation frequency <3 times a week, predominantly hard feces for the majority of stools, and ≥1 episode of fecal incontinence per week. Mothers were asked if their child had abdominal pain at least once a week (yes or no) and if their child had diarrhea (“very soft/slushy” or “watery” looking feces).36 Maternal antibody titers were measured during the second trimester of pregnancy. Titers were analyzed as previously described.
Statistical Analysis
First, to compare the characteristics of CDA and non-CDA groups, independent t tests, Mann–Whitney U tests, and χ2 tests were performed. Second, given the absence of normality, square root transformation was performed on CBCL scores, and CBCL scores were standardized into z scores.
Associations between child CDA and CBCL scales were evaluated by using linear regression analyses, with CDA (TG2A-negative titers versus TG2A-positive titers) as an independent variable and the continuous CBCL z scores as dependent variables. We created 4 models for regression: a basic model without any adjustment, an adjusted model with adjustment for confounders, and 2 mediation models consisting of the adjusted model with additional adjustment for BMI and a model with additional adjustment for gastrointestinal complaints on top of adjustment for BMI. Potential confounders (sex, HLA antigen risk profile, sociodemographic factors [income, education], ethnicity, maternal TG2A seropositivity, and birth weight) included in the adjusted model were based on associations described in the literature and were included in the model when they attained a >10% change in effect size, as suggested by Mickey and Greenland,37 determined for the CBCL scales separately.
Because it can be argued that BMI is in the pathway between CeD and emotional and behavioral problems,17,38 an additional model was created for BMI. Likewise, because gastrointestinal complaints may be a collider (ie, common effect of exposure and outcome) instead of a confounder, we additionally adjusted for gastrointestinal problems.
A sensitivity analysis was performed to evaluate the robustness of our results. In these analyses, children with TG2A titers with positive results and without the HLA antigen risk alleles were excluded.
To reduce selection bias due to missing data, multiple imputations of covariates (pooled results of 10 imputed data sets) were performed. Covariates with missing values were HLA antigen risk profile, maternal income, maternal educational level, and ethnicity, obstipation, abdominal pain, and stool consistence, as detailed in Supplemental Tables 4 and 5. P < .05 was considered statistically significant. All analyses were performed by using SPSS version 24.0 for Windows (IBM SPSS Statistics, IBM Corporation, Armonk, NY).
Results
Subject Characteristics
Maternal and child characteristics are shown in Table 1. At the age of 6 years, 51 children had CDA (1.4%) (Fig 1). Of children with CDA, 92.2% carried the HLA-DQ2 or HLA-DQ8 allele (Table 1). Female sex, HLA-DQ2 and/or HLA-DQ8 positivity, Western ethnicity, lower BMI, and maternal TG2A titers with positive results were related to CDA (Table 1). Data on the internalizing CBCL composite were available for 3686 children (99.2%), and data on the externalizing CBCL composite were available for 3703 children (99.7%) (Fig 1).
Characteristics According to CDA
| CDA at the Age of 6 y | |||
|---|---|---|---|
| No CDA (n = 3664) | CDA (n = 51) | Pa | |
| Child characteristics | |||
| Age at research center visit, median (range), y | 6.0 (4.9−90) | 6.0 (5.6−8.2) | .90 |
| Female sex, n (%) | 1755 (47.9) | 33 (64.7) | .02 |
| BMI, median (range) | 15.82 (12.04−27.65) | 15.40 (12.99−19.45) | .01 |
| Wt, median (range), kg | 23.4 (13.0−56.0) | 21.0 (16.0−35.6) | .00 |
| Height, mean (SD), cm | 119.5 (5.9) | 118.0 (6.3) | .07 |
| BMI for age SDS, mean (SD) | 0.26 (0.9) | −0.03 (0.8) | .02 |
| Birth wt, median (range), g | 3460 (635−5270) | 3520 (1810−4750) | .75 |
| HLA-DQ2 and/or HLA-DQ8 positivity (yes), n (%) | 1492 (40.7) | 47 (92.2) | .00 |
| Western ethnicity, n (%) | 2533 (69.1) | 42 (82.4) | .04 |
| Gastrointestinal complaints (yes), n (%) | |||
| Obstipation | 702 (19.2) | 6 (11.8) | .20 |
| Diarrhea | 114 (3.1) | 2 (3.9) | .65 |
| Abdominal pain | 4S4 (13.2) | 6 (11.8) | .57 |
| Maternal characteristics | |||
| Income ≥€2200, n (%) | 2173 (59.3) | 38 (74.5) | .05 |
| High educational level, n (%) | 1853 (50.6) | 33 (64.7) | .06 |
| TG2A titers with positive results during second trimester, n (%) | 12 (0.4) | 1 (2.8) | .04 |
| Unknown antibody titers | 896 | 15 | — |
| CDA at the Age of 6 y | |||
|---|---|---|---|
| No CDA (n = 3664) | CDA (n = 51) | Pa | |
| Child characteristics | |||
| Age at research center visit, median (range), y | 6.0 (4.9−90) | 6.0 (5.6−8.2) | .90 |
| Female sex, n (%) | 1755 (47.9) | 33 (64.7) | .02 |
| BMI, median (range) | 15.82 (12.04−27.65) | 15.40 (12.99−19.45) | .01 |
| Wt, median (range), kg | 23.4 (13.0−56.0) | 21.0 (16.0−35.6) | .00 |
| Height, mean (SD), cm | 119.5 (5.9) | 118.0 (6.3) | .07 |
| BMI for age SDS, mean (SD) | 0.26 (0.9) | −0.03 (0.8) | .02 |
| Birth wt, median (range), g | 3460 (635−5270) | 3520 (1810−4750) | .75 |
| HLA-DQ2 and/or HLA-DQ8 positivity (yes), n (%) | 1492 (40.7) | 47 (92.2) | .00 |
| Western ethnicity, n (%) | 2533 (69.1) | 42 (82.4) | .04 |
| Gastrointestinal complaints (yes), n (%) | |||
| Obstipation | 702 (19.2) | 6 (11.8) | .20 |
| Diarrhea | 114 (3.1) | 2 (3.9) | .65 |
| Abdominal pain | 4S4 (13.2) | 6 (11.8) | .57 |
| Maternal characteristics | |||
| Income ≥€2200, n (%) | 2173 (59.3) | 38 (74.5) | .05 |
| High educational level, n (%) | 1853 (50.6) | 33 (64.7) | .06 |
| TG2A titers with positive results during second trimester, n (%) | 12 (0.4) | 1 (2.8) | .04 |
| Unknown antibody titers | 896 | 15 | — |
—, not applicable.
Differences in subject characteristics between the groups were evaluated by using t tests and Mann–Whitney U tests for continuous variables and χ2 tests for categorical variables.
CDA and Emotional and Behavioral Problems
Median CBCL scores per scale for children with and without CDA are provided in Table 2.
Median CBCL Scores According to CDA
| CBCL Scale | No CDA (n = 3664) | CDA (n = 51) |
|---|---|---|
| Median (Interquartile Range) | Median (Interquartile Range) | |
| CBCL syndrome scales | ||
| Internalizing composite | 4.00 (2.00–8.00) | 4.00 (2.00–9.00) |
| Emotionally reactive | 1.00 (0.00–2.25) | 1.50 (0.00–3.00) |
| Anxious and/or depressed | 1.00 (0.00–2.00) | 1.00 (0.00–3.00) |
| Somatic complaints | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| Withdrawn | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| Externalizing composite | 6.00 (2.00–11.00) | 5.00 (3.00–10.43) |
| Attention problems | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| Aggressive behavior | 5.00 (2.00–8.44) | 5.00 (2.00–8.00) |
| Sleep problems | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| DSM-oriented scales | ||
| Affective | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| Anxiety | 1.00 (0.00–3.00) | 2.00 (0.00–3.00) |
| Pervasive development | 2.00 (0.00–3.00) | 2.00 (1.00–4.00) |
| Attention-deficit/hyperactivity | 3.00 (1.00–5.00) | 3.00 (1.00–4.00) |
| Oppositional defiant problems | 2.00 (1.00–4.00) | 2.00 (1.00–4.00) |
| CBCL Scale | No CDA (n = 3664) | CDA (n = 51) |
|---|---|---|
| Median (Interquartile Range) | Median (Interquartile Range) | |
| CBCL syndrome scales | ||
| Internalizing composite | 4.00 (2.00–8.00) | 4.00 (2.00–9.00) |
| Emotionally reactive | 1.00 (0.00–2.25) | 1.50 (0.00–3.00) |
| Anxious and/or depressed | 1.00 (0.00–2.00) | 1.00 (0.00–3.00) |
| Somatic complaints | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| Withdrawn | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| Externalizing composite | 6.00 (2.00–11.00) | 5.00 (3.00–10.43) |
| Attention problems | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| Aggressive behavior | 5.00 (2.00–8.44) | 5.00 (2.00–8.00) |
| Sleep problems | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| DSM-oriented scales | ||
| Affective | 1.00 (0.00–2.00) | 1.00 (0.00–2.00) |
| Anxiety | 1.00 (0.00–3.00) | 2.00 (0.00–3.00) |
| Pervasive development | 2.00 (0.00–3.00) | 2.00 (1.00–4.00) |
| Attention-deficit/hyperactivity | 3.00 (1.00–5.00) | 3.00 (1.00–4.00) |
| Oppositional defiant problems | 2.00 (1.00–4.00) | 2.00 (1.00–4.00) |
In basic models, no significant associations were found between CBCL syndrome scales and CDA (Supplemental Table 6). Adjustment for confounders and additional adjustment for BMI only changed the effect estimates slightly. After adjustment for gastrointestinal complaints, association between CDA and aggressive behavior was borderline significant (adjusted β = .32; 95% confidence interval [CI] 0.00 to 0.65; P = .05). In basic models, no significant associations were found between CDA and the DSM-oriented scales. After adjustment for confounders, CDA was associated with a significantly higher CBCL score on the anxiety scale (adjusted β = .29; 95% CI 0.02 to 0.55; P = .02) and a borderline significantly higher score on the oppositional defiant problems scale (confounder adjusted β = .33; 95% CI 0.00 to 0.65; P = .05) (Table 3). Additional adjustment for BMI had no influence on effect size. Additional adjustment for gastrointestinal complaints revealed a stronger association on the anxiety scale (adjusted β = .31; 95% CI 0.04 to 0.57; P = .01) and oppositional defiant problems scale (adjusted β = .35; 95% CI 0.02 to 0.67; P = .02).
Association of CDA With CBCL Scores
| Adjusted Model 2, β (95% CI) | Adjusted Model 3, β (95% CI)a | Adjusted Model 4, β (95% CI)b | |
|---|---|---|---|
| CBCL syndrome scales | |||
| Internalizing compositec,d,e,f,g,h | .16 (−0.16 to 0.47) | .15 (−0.17 to 0.47) | .21 (−0.10 to 0.52) |
| Emotionally reactivec,d,e,f,g | .22 (−0.06 to 0.49) | .21 (−0.07 to 0.49) | .23 (−0.04 to 0.51) |
| Anxious and/or depressedd,e,f,g,h | .02 (−0.31 to 0.35) | .03 (−0.30 to 0.36) | .06 (−0.27 to 0.38) |
| Somatic complaintsc,d,e,f | .05 (−0.23 to 0.32) | .03 (−0.25 to 0.30) | .09 (−0.17 to 0.35) |
| Withdrawnc,d,g,h | .19 (−0.14 to 0.52) | .18 (−0.15 to 0.51) | .22 (−0.11 to 0.55) |
| Externalizing compositec,d,e,f,g,h | .26 (−0.06 to 0.58) | .26 (−0.06 to 0.58) | .29 (−0.03 to 0.60) |
| Attention problemsc,d,e,f,g | .03 (−0.30 to 0.35) | .03 (−0.30 to 0.35) | .05 (−0.27 to 0.37) |
| Aggressive behaviorc,d,e,f,g,h | .30 (−0.02 to 0.63) | .30 (−0.02 to 0.63) | .32 (0.00 to 0.65)* |
| Sleep problemsc,d,e,f,g,h | .20 (−0.13 to0.53) | .21 (−0.12 to 0.54) | .24 (−0.09 to 0.56) |
| DSM-oriented scales | |||
| Affectived,f,g,h,i | −.07 (−0.40 to 0.26) | −.07 (−0.39 to 0.26) | −.02 (−0.34 to 0.30) |
| Anxietye,f,g | .29 (0.02 to 0.56)** | .29 (0.02 to 0.56)** | .31 (0.04 to 0.57)*** |
| Pervasive developmentc,d,g,h | .16 (−0.17 to 0.49) | .16 (−0.17 to 0.49) | .19 (−0.14 to 0.52) |
| Attention-deficit/hyperactivityc,d,e,f,g,h,i | .24 (−0.08 to 0.57) | .25 (−0.07 to 0.57) | .26 (−0.06 to 0.58) |
| Oppositional defiant problemsc,d,g,h | .33 (0.00 to 0.65)* | .33 (0.00 to 0.65)* | .35 (0.02 to 0.67)** |
| Adjusted Model 2, β (95% CI) | Adjusted Model 3, β (95% CI)a | Adjusted Model 4, β (95% CI)b | |
|---|---|---|---|
| CBCL syndrome scales | |||
| Internalizing compositec,d,e,f,g,h | .16 (−0.16 to 0.47) | .15 (−0.17 to 0.47) | .21 (−0.10 to 0.52) |
| Emotionally reactivec,d,e,f,g | .22 (−0.06 to 0.49) | .21 (−0.07 to 0.49) | .23 (−0.04 to 0.51) |
| Anxious and/or depressedd,e,f,g,h | .02 (−0.31 to 0.35) | .03 (−0.30 to 0.36) | .06 (−0.27 to 0.38) |
| Somatic complaintsc,d,e,f | .05 (−0.23 to 0.32) | .03 (−0.25 to 0.30) | .09 (−0.17 to 0.35) |
| Withdrawnc,d,g,h | .19 (−0.14 to 0.52) | .18 (−0.15 to 0.51) | .22 (−0.11 to 0.55) |
| Externalizing compositec,d,e,f,g,h | .26 (−0.06 to 0.58) | .26 (−0.06 to 0.58) | .29 (−0.03 to 0.60) |
| Attention problemsc,d,e,f,g | .03 (−0.30 to 0.35) | .03 (−0.30 to 0.35) | .05 (−0.27 to 0.37) |
| Aggressive behaviorc,d,e,f,g,h | .30 (−0.02 to 0.63) | .30 (−0.02 to 0.63) | .32 (0.00 to 0.65)* |
| Sleep problemsc,d,e,f,g,h | .20 (−0.13 to0.53) | .21 (−0.12 to 0.54) | .24 (−0.09 to 0.56) |
| DSM-oriented scales | |||
| Affectived,f,g,h,i | −.07 (−0.40 to 0.26) | −.07 (−0.39 to 0.26) | −.02 (−0.34 to 0.30) |
| Anxietye,f,g | .29 (0.02 to 0.56)** | .29 (0.02 to 0.56)** | .31 (0.04 to 0.57)*** |
| Pervasive developmentc,d,g,h | .16 (−0.17 to 0.49) | .16 (−0.17 to 0.49) | .19 (−0.14 to 0.52) |
| Attention-deficit/hyperactivityc,d,e,f,g,h,i | .24 (−0.08 to 0.57) | .25 (−0.07 to 0.57) | .26 (−0.06 to 0.58) |
| Oppositional defiant problemsc,d,g,h | .33 (0.00 to 0.65)* | .33 (0.00 to 0.65)* | .35 (0.02 to 0.67)** |
Values are regression coefficients (95% CI) from linear regression models that reflect differences in CBCL z scores on the CBCL syndrome scale and DSM-oriented scale according to CDA. Estimates are based on multiple-imputed data.
Adjusted for confounders in model 2 and additionally adjusted for BMI.
Adjusted for confounders in model 2 and additionally adjusted for gastrointestinal complaints.
Adjusted for sex.
Adjusted for HLA antigen type.
Adjusted for ethnicity.
Adjusted for maternal income.
Adjusted for maternal education level.
Adjusted for maternal antibody level.
Adjusted for birth wt.
P = .05;
P = .02;
P = .01.
Emotional and Behavioral Problems in HLA-DQ2–Positive and/or HLA-DQ8–Positive Children
For sensitivity analyses TG2A-positive children who were not carrying HLA-DQ2 and HLA-DQ8 alleles (n = 4) were excluded. In the HLA-DQ2–positive and HLA-DQ8–positive children, CDA was borderline significantly associated with a higher score on the aggressive behavior subscale on the syndrome scale after adjustment for confounders (adjusted β = .32; 95% CI 0.00 to 0.65; P = .05). On the DSM-oriented scales, children with CDA who were carrying the HLA-DQ2 and/or HLA-DQ8 alleles had a higher score on the anxiety scale (adjusted β = .38; 95% CI 0.08 to 0.67) and the oppositional defiant problems scale (adjusted β = .35; 95% CI 0.02 to 0.69; P = .04) after adjustment for confounders. Gastrointestinal complaints could partly explain association with the oppositional defiant problems scale (adjusted β = .34; 95% CI 0.00 to 0.68; P = .05). Additional adjustment for BMI had no influence on effect size.
Discussion
This observational, population-based study revealed that subclinical CDA might be associated with anxiety problems in a population-based cohort of 6-year-old children. The association we found was stronger in children with CDA who were carrying the HLA-DQ2 and/or HLA-DQ8 allele. Additional associations were found with oppositional defiant problems and aggressive behavior in the subgroup carrying the HLA-DQ2 and/or HLA-DQ8 allele. However, no associations were found on the empirical broadband scales of internalizing and externalizing problems.
Previous studies have revealed a positive association between psychopathological manifestations, including anxiety problems, and CeD in children with symptoms who were diagnosed with CeD.6–8,11,39 Our results reveal that anxiety problems might even occur in children with CDA identified by screening who are unaware of any disease or positive serology. As far as we know, only 1 other study has been conducted to evaluate psychological symptoms in children identified by screening who are unaware of their CDA status.13 Comparable to our results, Smith et al13 reported associations between CDA and anxiety, depression, aggressive behavior, and sleep problems in children age 3.5 years. However, associations were not confirmed in follow-up at age 4.5 years.13 The inconsistency of the associations could have been caused by the lack of adjustment for confounders. For example, Smith et al13 did not adjust for ethnicity and sociodemographic factors, which we found to be highly associated with emotional and behavioral problems and CDA.
The etiology of emotional and behavioral problems in CeD has yet to be established. The question rises whether emotional and behavioral problems are caused through pathologic pathways of CeD or are a secondary psychological effect of the diagnosis and/or gastrointestinal complaints. The current study involves a population that was unaware of its current CDA status, which ruled out an effect of diagnosis or disease awareness. Furthermore, after additional adjustment for gastrointestinal complaints, the association persisted. Therefore, gastrointestinal complaints related to CeD cannot explain the association between CDA and psychological problems. This suggests that underlying pathologic mechanisms might explain the association between psychopathology and CDA in our cohort. Several hypotheses have been proposed in the literature to explain these mechanisms. First, depression and anxiety disorders have been linked to deficiencies of multiple nutrients such as folate and omega-3 polyunsaturated fatty acid.40–45 These nutrients are essential components in the central nervous system and are involved in brain development, neuroprotection, and neurotransmitter transport.44,46 Children suffering from these nutrient deficiencies have a greater risk of developing behavioral problems.46 CeD is associated with deficiencies of these nutrients that are caused by malabsorption due to villous atrophy.16,47,48 Second, recent studies suggest that the intestinal microbiome may play an important role in the pathogenesis of CeD.49 The intestinal microbiome is suggested to be highly associated with the gut-brain axis,50,51 a system between the brain and gastrointestinal tract that links emotional and cognitive centers of the brain with peripheral functioning of the digestive tract.52 Studies have revealed changes of the microbiome in children with autism.53,54 Additionally, in children with anxiety and depression disorders, probiotics seem to have an influence on psychological well-being, suggesting that the microbiome might have a role in these disorders as well.55 Patients with CeD also have alterations in their intestinal microbiome,17 suggesting that the microbiome might be involved in the psychopathology in patients with CeD. We found stronger associations in the group carrying the HLA-DQ2 and/or HLA-DQ8 allele. Given the high sensitivity and specificity of TG2A, children with CDA who are carrying the HLA-DQ2 and/or HLA-DQ8 haplotype are at increased risk of developing CeD. On the contrary, children without the HLA antigen risk alleles have a negligible chance of developing CeD1,56 and, thus, are not likely to have any of the pathology and mechanisms related to CeD. The stronger associations we found when excluding these children support that CeD-related mechanisms could partially explain anxiety symptoms in patients with CeD instead of a secondary psychological effect of the diagnosis or complaints.
A major strength of our study is the population-based design, including a large cohort of children with data on a wide range of possible confounding variables available.
However, limitations should be taken into account, too. First, although sensitivity and specificity of TG2A are high (>95%),1 information on additional antibody testing, such as antiendomysium and biopsies of all TG2A-positive children, was not available. This precluded definite diagnoses of CeD. From 26 children (51%), we had consent to obtain information through medical records on biopsies. Of these children, 23 showed Marsh grade II or III villous atrophy and were diagnosed with CeD. The remaining 3 had Marsh grade I with TG2A titers >10 ULN and required follow-up. Further research should be used to evaluate whether children with elevated TG2A levels identified by screening do have CeD or develop CeD over time.27 We determined presence of the HLA-DQ2 and/or HLA-DQ8 alleles to enhance the identification of children at risk for developing CeD. Remarkably, we observed that 4 children with elevated TG2A titers did not carry the HLA-DQ2 and/or HLA-DQ8 alleles, whereas CeD is restricted to those who carry these alleles. Two of these children had TG2A titers with strong positive results (>10 ULN), and the other 2 children had titers of 65 U/mL and 11 U/mL, respectively. Two children were of non-Western ethnicity. The absence of the HLA antigen risk alleles in these children could have been due to the use of a tag SNP approach of HLA antigen typing.30 Although the tag SNP approach has high positive and negative predicted values in a population with CeD diagnosis, a slightly higher rate of false-positives and false-negatives may be expected in the general population.30 Another limitation was the use of a parent report on emotional and behavioral problems in children. An additional clinical observation would give less subjective measurements. However, the CBCL was proven to be reliable in previous studies.28,57 As in any observational study, residual confounding should be taken into account. We were not able to adjust for all potential confounding factors, such as having siblings with CeD or presence of other autoimmune diseases. Also, because the etiology of CeD is largely unknown, the possibility exists of unknown confounders related to CeD and emotional and behavioral problems.
Our results reveal that anxiety problems might already be present among children who are screened and unaware of their CDA status. Because it is known that psychological symptoms may improve under treatment with a gluten-free diet,7,22 early diagnosis and treatment of CeD could prevent further development or deterioration of psychopathology in these children. However, information on the effect of diagnosis and treatment on patients identified by screening is scarce.23,24 Further research should be done to determine the effect of treatment on anxiety problems in children with CDA to evaluate if children could benefit from early diagnosis of CeD. Furthermore, our results suggest that clinicians should be aware of CeD in children with reported unexplained anxiety, oppositional defiant problems, and aggressive behavior.
Conclusions
Our results do not support the hypothesis that subclinical CDA is associated with the broadband internalizing” and externalizing scales of the CBCL scale. However, our results suggest that subclinical CDA might be associated with anxiety and oppositional defiant problems in 6-year-old children. Further research should be done to evaluate whether behavioral problems can be improved by a gluten-free diet in children with CDA identified by screening.
Acknowledgments
The Generation R Study is conducted by the Erasmus University Medical Center in close collaboration with the School of Law and Faculty of Social Sciences of Erasmus University Rotterdam, the Municipal Health Service Rotterdam area, the Rotterdam Homecare Foundation, and Stichting Trombosedienst & Artsenlaboratorium Rijnmond. We acknowledge the contribution of children and parents, general practitioners, hospitals, midwives, and pharmacies in Rotterdam.
Drs Wahab and Kiefte-de Jong conceptualized the study, conducted the statistical analysis, interpreted the data, drafted the manuscript, and revised the manuscript on the basis of the coauthors’ comments; Dr Beth coordinated data acquisition on tissue transglutaminase autoantibodies, conceptualized and designed the study, and critically reviewed and revised the manuscript; Ms Derks and Dr Jansen designed the Childhood Behavioral Checklist database, advised on using and analyzing the Childhood Behavioral Checklist data, and critically reviewed and revised the manuscript; Dr Moll designed the Generation R Study, coordinated data acquisition on tissue transglutaminase autoantibodies, and critically reviewed and revised the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2019-1683.
FUNDING: Supported by Erasmus University Medical Center, Erasmus University Rotterdam, the Netherlands Organization for Health Research and Development, and NutsOhra. The sponsors had no role in the design of the study, the data collection and analyses, the interpretation of data, the preparation and review of the manuscript, and the decision to submit the manuscript.
- CBCL
Childhood Behavioral Checklist
- CDA
celiac disease autoimmunity
- CeD
celiac disease
- CI
confidence interval
- DSM
Diagnostic and Statistical Manual of Mental Disorders
- SDS
standard deviant score
- SNP
single-nucleotide polymorphism
- TG2A
tissue transglutaminase autoantibody
Comments
RE: Celiac Disease Autoimmunity and Emotional and Behavioral Problems in Childhood
We read with great interest “Celiac disease autoimmunity and emotional and behavioral problems in childhood” by Wahab et al.1 The authors reiterated the fact that aside from the physical symptoms of Celiac disease (CeD) such as gastrointestinal discomfort and fatigue, CeD is also associated with symptoms of behavioral, emotional and other psychological problems. Additionally, CeD autoimmunity greatly increased the likelihood of a child exhibiting symptoms of anxiety and defiant, disobedient behaviors. Determining why this happens is the next step because as the author states, “psychological symptoms in early life predispose to psychopathology in later life and contribute to a lower quality of life” although currently there is no reliable explanation.1
A recent study that examined the psychological symptoms associated with CeD in children found that there were increased reporting by parents of anxiety, depression and aggressive behavior in younger children as compared to older children.2 This suggests that younger children who can’t verbalize their physical discomfort may express their discomfort through their behavior. However, another one of the findings of the study was that a gluten-free diet, which relieves gastrointestinal discomfort and other associated symptoms, had no effect on the children’s psychological functioning.2 If a gluten-free diet doesn’t resolve the psychological symptoms of CeD then is it possible that there could be another cause of these symptoms?
Perhaps a causative factor to consider regarding CeD is that of adverse childhood events (ACEs). ACEs are traumatic events that are experienced as a child, such as physical, sexual, or emotional abuse, which includes experiences such as witnessing substance abuse or domestic abuse, growing up with a parent who is incarcerated or has a mental illness, experiencing socioeconomic hardships, or having parents who are divorced.3, 4 It is important to note that CeD is an autoimmune disease, and ACES have been shown to have destructive effects on the functioning of the immune system by altering the adaptive stress response and increasing the levels of inflammatory markers in the blood.3 Another recent study found that an increasingly higher ACE score was associated with a higher susceptibility of rheumatoid arthritis, among other autoimmune diseases. ACEs are not only linked to a higher prevalence of autoimmune disorders, but also a higher prevalence of behavioral disorders, such as Attention Deficit-Hyperactivity Disorder (ADHD).4 Children who had experienced two or more adverse events were much more likely to have a moderate or severe diagnosis of ADHD.
Since ACEs are correlated with both a higher prevalence of autoimmune disorders and behavioral issues, this begs the question of which comes first, the autoimmune disorder or the psychological manifestations? Is it possible that children with CeD autoimmunity who exhibit behavioral problems are experiencing these symptoms because of the stress placed on their immune system due to traumatic events in their childhood? These behavioral problems might not be a symptom of the autoimmune disease, but rather a result of experiencing an adverse event in their childhood which may have also triggered the onset of immune dysfunction.
Carly Shultis, Health Studies
Mary Siniscarco, Associate Professor of Health Studies
Utica College
Utica, NY 13502
References
Wahab RJ, Beth SA, Derks IPM, Jansen PW, Moll HA, Jong JCK. Celiac disease autoimmunity and emotional and behavioral problems in childhood. Pediatrics. 2019;144(4):49-51. https://uclibraryts.on.worldcat.org/oclc/8292795513.
Smith LB, Lynch KF, Kurppa K, et al. Psychological manifestations of celiac disease autoimmunity in young children. Pediatrics. 2017;139(3). https://uclibraryts.on.worldcat.org/oclc/6952675439. doi: 10.1542/peds.2016-2848.
Luiz APL, Antico HdA, Skare TL, Boldt ABW, Nisihara R. Adverse childhood experience and rheumatic diseases. Clinical Rheumatology : Journal of the International League of Associations for Rheumatology. 2018;37(10):2863-2867. https://uclibraryts.on.worldcat.org/oclc/7844639334. doi: 10.1007/s10067-018-4200-5.
Brown NM, Brown SN, Briggs RD, Germán M, Belamarich PF, Oyeku SO. Associations between adverse childhood experiences and ADHD diagnosis and severity. Academic Pediatrics. 2017;17(4):349-355. http://www.sciencedirect.com.ezproxy.utica.edu/science/article/pii/S1876.... doi: https://doi-org.ezproxy.utica.edu/10.1016/j.acap.2016.08.013.