Case Finding for Celiac Disease With a Point-of-Care Test Open Access

Celiac disease (CD) is one of the most common chronic gastrointestinal diseases triggered by the ingestion of gluten. Its prevalence is continuously increasing in many countries and currently it affects approximately 1% of the world’s population.^1,2 CD is characterized by the production of specific autoantibodies against transglutaminase type 2 (TGAs) and endomysium (EMAs) during gluten ingestion. Serologic testing identifies most patients with CD using these CD-specific and CD-sensitive antibodies. Nevertheless, it presents a diagnostic challenge because of both underdiagnosis and overdiagnosis.^3,4 

Despite well-implemented guidelines for CD diagnosis, growing awareness among health care professionals, and accurate serologic tests, the diagnosis is often delayed for years. Even worse, CD remains frequently unrecognized and, therefore, untreated.⁵ This is partially because of the absent or nonspecific heterogeneous symptoms and a continuous change in the clinical presentation.⁶ Untreated disease is associated with inflammation and villous atrophy within the small intestine that may lead to malabsorption, comorbidity, and late complications such as chronic anemia, infertility, osteoporosis, and, rarely, malignancy.^7,8 Conversely, overdiagnosis has also become a concern with the rising awareness of CD. Some individuals may be incorrectly diagnosed, resulting in unnecessary dietary restriction. Because it is difficult to diagnose CD promptly and accurately, prevention of the disease is desirable.

Results from large prospective cohorts have shown that most cases of the disease develop very early in life, that primary prevention of CD by dietary interventions/recommendations is not possible, and that early diagnosis and treatment represent the only way to secondary prevention.^9–12 There are 2 approaches to achieve this: mass screening and case finding, but the first one is controversial because of possible ethical implications.¹³ Active case finding refers to liberal diagnostic testing of persons with CD-associated symptoms, for example by primary health care professionals. The existence of reliable, easy-to-use, onsite, rapid point-of-care tests (POCTs) to detect the specific CD antibodies provides a possibility to assess CD in preventive health care settings.^14,15 

The aim of this study was to prospectively assess whether case finding at the Preventive Youth Health Care Centers (YHCCs) in the Netherlands can effectively be implemented for early CD diagnosis.

A prospective controlled observational cohort study was carried out at the 14 locations of the Preventive YHCCs in the Kennemerland region in the Netherlands from February 4, 2019, to January 4, 2022, with 5 months interruption (March 19, 2020 to August 15, 2020) because of the COVID-19 pandemic. Study region Kennemerland is a coastal region in the northwestern Netherlands, in the province of North Holland. It consists of 9 municipalities and has 553 825 inhabitants (male/female ratio 1.05; average age 40.6 years).

In the Netherlands, more than 95% of all children until the age of 4 years visit the Preventive YHCCs, whose goal is to promote and secure the health and safety of all children by, among other methods, early detection or prevention of diseases. Children attending the YHCCs for a standard consultation during the study period were in advance identified by the Preventive YHCC administration and the parents/legal guardians (from this point on called “parents”) and received information about the study. Inclusion criteria for CD testing were (1) age between ages 12 months and 4 years, (2) following a gluten-containing diet, (3) parents having sufficient knowledge of the Dutch language, and 4) having 1 or more CD-associated symptoms according to a standardized questionnaire based on the recommendations of the European Society for Paediatric Gastroenterology, Hepatology and Nutrition for CD testing in children (Supplemental File 1).¹⁶ Informed consent was requested twice: first orally for completing the CD-associated symptom questionnaire and later written (by both parents/guardians) for performing a validated POCT for TGA (immunoglobulin [Ig] A/IgG/IgM) (Celiac Quick Test, BioHit Oyj).¹⁵ Before the COVID-19 pandemic, the questionnaire was presented to the parents of children aged 14 months and 2, 3, and 4 years, and, after the research stopped because of the pandemic, to the parents of children aged 14 months and 4 years, because only these children attended the YHCCs.

For the included children, a separate appointment was made to perform the POCT, which was performed and interpreted by trained nurses according to the manufacturer’s instructions. All POCT results (positive/negative) were photographed and stored in the electronic patient’s file at the Preventive YHCC. If the POCT result was positive, the child was referred within 3 weeks to the Department of Pediatric Gastroenterology of the Leiden University Medical Center (LUMC) for further work-up. We considered the diagnosis of CD as confirmed with TGA titers more than 7 times the upper limit of normal (ULN; Thermofisher; ImmunoCAP250; cut-off of normality 7 U/mL) plus positive EMA results (Aeskuslides; immunofluorescence) or by small bowel biopsies.^3,8,16 If the POCT result was negative, the child was invited to repeat the POCT for TGA during their next yearly visit at the YHCCs. Ethical approval was obtained from the Medical Ethics Committee of Leiden-Den Haag-Delft (METC-LDD; P17.240/NL63291.058.17). The trial was inscribed in Trial Register (number NL7089 at onderzoekmetmensen.nl).

To assess the overall incidence of clinically diagnosed CD in the Netherlands during the study period, the data reported to the Dutch Pediatric Surveillance Unit (DPSU) was used. The DPSU is a unique registry of the Dutch Society of Pediatrics, comprising all Dutch pediatricians including pediatric gastroenterologists, and was functional until April 2021. In the Netherlands, CD is diagnosed by pediatricians/pediatric gastroenterologists. Between April 1, 2019, and April 30, 2021, all Dutch practicing pediatricians were asked monthly by email to report newly diagnosed CD in patients aged 0–18 years. After reporting a new diagnosis of CD, the pediatricians were asked to fill in a standardized questionnaire in a secure application. This questionnaire collected patient information such as gender, age, and (results of) diagnostic tests. Personal data were limited to initials, age, and the first 3 numbers of the zip code to identify the study region. The completed questionnaires were subsequently sent to the investigators at the LUMC where the data were stored and analyzed. Demographic data of the general population were provided by the Dutch Central Bureau of Statistics (CBS).¹⁷ 

The statistical analysis plan was published online on May 11, 2021 before the analyses were performed (Supplemental File 2). Analyses were done using R (version 3.6.2). The incidence rate of CD diagnoses with case finding was calculated by dividing the total number of CD diagnoses (including diagnoses by standard of care) in the study region by the total number of children years aged 1–4 years living in this region according to the CBS.¹⁷ 

The incidence rate of CD diagnosed by the standard of care in the rest of the Netherlands was calculated by dividing the number of CD diagnoses in children aged 1–4 years reported to the DPSU by the total number of children years aged 1–4 years living in the Netherlands outside the study region during the study period.¹⁷ Because the response rate to the DPSU varied in the different Dutch provinces, the incidence of CD was also calculated for the province with the highest incidence rate reported to the DPSU.¹⁸ 

Univariable and multivariable analyses were performed to determine the relation between child characteristics and the risk of CD, employing logistic models with generalized estimation equations to adjust inference for the correlation between observations from patients who participated multiple times.

The results of the study are presented in Figure 1 and Table 1. Of the 16 289 approached parents, 14 917 gave informed consent to fill in the questionnaire on CD-associated symptoms. The characteristics of the children whose parents gave or did not give informed consent were similar in terms of age (2.1 vs 2.2 years), gender (male 51.3% vs 52.0%), and having relatives with CD (7.3% vs 5.8%). Parents of 5301 (35.5%) children reported CD-related symptoms, 201 of this number in 2 or 3 consecutive years (total 5512 questionnaires). Of them, 3102 (58.5%) gave informed consent for a POCT in their children (47.8% female, mean age 2.13 years), with 99 children being tested 2 consecutive years and 1 child 3 years. In total, 3203 POCTs were performed, and 61 were positive: in 58 children at their first test and in 3 during their second test (11, 13, and 26 months after the initial test).

After additional serologic and/or histologic examination at the LUMC, CD was confirmed in 56 children (1.7%, 95% CI, 1.32–2.26; mean age 2.56 years). The diagnosis of CD was made in 53 children based on serology and in 3 children confirmed by histology of the small bowel. Of the remaining 5 children with a positive POCT result, 3 were diagnosed with potential CD (positive TGA <10× ULN and Marsh 1) and in the other 2 CD was ruled out (negative human leukocyte antigen [HLA]-DQ2 and DQ8 and TGA in serum; 3.3%) (Supplemental Table 1). The positive predictive value for CD of the questionnaire was 1.06% (95% CI, 0.81–1.36) and the 1 of the POCTs was 91.8% (95% CI, 81.9–97.3).

With the exception of abdominal distention (more frequent in CD n = 23/56, 41.1% vs n = 824/3115, 26.5%; P = .036), the reported symptoms were similarly frequent among children with and without CD (Table 1). The most frequently reported symptoms among the whole study population were growth retardation, abdominal distention, and irritability (71.3%, 26.7%, and 18.4%, respectively); and vomiting and recurrent aphthous mouth ulcerations were the least frequently reported (2.5% and 3.7%, respectively). Children with confirmed CD were significantly older (mean age 2.56 vs 2.13 years; P = .001) and significantly more girls were diagnosed with the disease (40/56, P = .001). Also, CD was significantly more frequent among children with a family member with CD (n = 13/52, 25.0% vs n = 387/2614, 14.8%; P = .043) (Table 1). In a multivariable analysis, the characteristics of the children with and without CD were significantly different in terms of age (P < .001), gender (P = .001), and having a family member with CD (P = .049), but not in terms of any symptoms (P = .808) and growth restriction (P = .170).

Clinically diagnosed CD was reported to the DPSU in 479 children (mean 7.2 years [95% CI, 11 months-17.7 years]; 66% female). Of them, 150 children were aged 1–4 years (mean age 2.6 years [95% CI, 1.0–3.9 years]; 60% female). Four of them lived in the study region (2 of those were diagnosed during the COVID-19 stop), giving a total of 146 children diagnosed with CD among the control population.

The overall crude incidence rate of CD diagnosed in the case finding region was 1.67 per 1000 person-years (95% CI, 1.27–2.15): 2.57 per 1000 person-years (95% CI, 1.80–3.56) before the COVID-19 stop and 1.09 (95% CI, 0.70–1.62) after the COVID-19 stop. The crude incidence rate was 1.47 (95% CI, 1.10–1.93) at the first test (53 cases) and 0.08 (95% CI, 0.02–0.24) at the second test (3 cases) (P < .001).

The overall incidence of CD in the control population diagnosed according to the standard of care was 0.14 per 1000 person-years, significantly lower than what was found in the case-finding region (P < .001). This significant difference was also present when comparing the incidence of CD diagnoses in the case-finding region with the 1 found by standard of care in the Dutch province with the highest incidence reported to the DPSU (Drenthe: 0.37 per 1000 person-years; P < .001).

As presented in Table 2, the incidences of CD diagnoses by case finding was significantly higher than the 1 found by standard of care in the 3 age groups 1–2, 2–3, and 3–4 years, with incidences of 1.11, 1.50, and 2.36 per 1000 person-years and 0.11, 0.15, and 0.15 per 1000 person-years, respectively (all P < .001).

Our study demonstrates that case finding for CD using a POCT is feasible and effective at the Dutch Preventive YHCCs.¹⁹ The frequency of CD detected with case finding is significantly higher than that diagnosed by standard of care during the same period.

It is known that the incidence of CD diagnoses has increased in recent years, also in the Netherlands, because of guidelines on who to test and increased awareness among physicians.^8,18 However, the diagnosis by standard of care is only possible if physicians are approached by parents/guardians because of their child’s symptoms. In our study, in which all the included children had 1 or more CD-related symptoms, only 3 of the 56 with CD had visited a health care professional, and in none of them was possible CD considered or tested. That case finding is effective to detect undiagnosed CD has been described previously in different populations, both in adults and children.^20–23 Our results also show that the CD-related symptoms reported by the parents, with the exception of abdominal distention, do not discriminate between children with and without CD. However, the standardized CD-associated symptoms questionnaire is effective in case finding for CD among young children from the general population when mass screening is not possible.

The poor predictive value of symptoms for childhood CD has been previously reported, indicating that mass screening is more effective than case finding to detect the disease in all the children.^7,23–25 This may be done by detection of the HLA-CD–compatible individuals (HLA-DQ2 and/or DQ8, about 40% of the general population) in whom further testing for CD-specific antibodies is performed.²³ The effectiveness, costs, and acceptability of this method are currently being investigated in the same setting. In our center, mass screening was not approved by the METC-LDD nor by the Dutch Central Committee on Human Subjects Research because of insufficient evidence to assess the balance of advantages and disadvantages of CD diagnosis in asymptomatic children, as previously suggested.²⁶ However, preliminary results of our inquiry among the general population of parents of young children show that most are positive about mass screening of CD (approximately 70%). Since Italy started in 2024 a national mass screening program in children aged 1–17 years, it is the right time to re-open the discussion on CD mass screening and/or case finding, including their cost-effectiveness.^27–29 

Possible shortcomings of our study were the separate appointment to perform the POCT, which decreased the participation rate by 29%, and the fact that all research projects, including ours, were temporarily halted owing to the COVID-19 pandemic. This also resulted in a reduction of consultations at the YHCCs after the age of 14 months and possibly explains the lower number of detected CD cases after the lockdown.³⁰ However, as well as shown in the TEDDY Study, we found significantly more CD diagnoses after age 1.5 years (Table 1).³⁰ Another possible shortcoming could be the underreporting of newly diagnosed cases by the DPSU. It is known that the response to the DPSU varies in the different Dutch provinces, and thus the incidence of CD was also calculated for the province that reported the highest incidence to the DPSU. Even then, case finding still detects significantly more CD than standard of care. Also, we should take into account that CD may develop later in life, limiting the efficacy of early CD screening/case finding. Prospective studies are needed to assess long-term retesting and the best age to do so in the diagnostic algorithm. A final shortcoming is that we used a nonbiopsy approach at TGA more than 7 times the ULN to reduce the burden of the endoscopy for children. However, it has been shown that titers of 7 times the ULN have a reliable positive predictive value (PPV) for severe changes in the small intestine in CD, but because of intertest and intratest variability, this is not advised.³ 

One strong aspect of our study is its setting: the Preventive YHCCs, which are optimal to detect CD early because they are freely accessible to the entire population up to age 4 years. Previous prospective studies in different countries among high-risk groups (genetic predisposition and CD families) have shown that CD develops at a young age (mean age 4.3 years).^9,12 Our study shows that this is also the case in children from the general population. CD detection at a young age is needed, as undetected CD is associated with important health problems early in life, such as reduced bone mineral density, delayed weight gain, short stature, and behavior problems.^7,8 

Also, the POCT was reliable to detect CD in this setting, with a PPV of 91.8% (95% CI, 81.9–97.28), although it should be noted that serologic testing has higher accuracy. However, serologic determination requires blood collection and organizational challenges, such as laboratory availability or transport of the material, which are expensive and not feasible at the Preventive YHCCs. In addition, the results of serologic testing are not immediately available as it is the case with the POCT. Other strong aspects of our study are the large sample size and the representativity of the region (in terms of gender, age distribution, parental educational level, and socioeconomic status) in which the study took place (Kennemerland) for the Dutch children population aged 1–4 years.¹⁷ 

After the study period, case finding for CD has been implemented in the regular care of the YHCCs in the Kennemerland region, and preliminary data show an even higher prevalence of CD than during the study, with 21 confirmed cases among 811 tested children (2.6%). This higher CD detection rate may be because of, among other things, avoidance of written informed consent by both parents/guardians, which was a clear barrier for participation, and which is compulsory in the Netherlands for research but not for standard of care.

Implementation of case finding for CD using a POCT is effective and feasible in young children at the Dutch Preventive YHCCs, and it detects a significantly higher number of cases compared with regular care. Implementation of early detection of CD as a standard part of all the Preventive Health Care Centers for children will reduce the underdiagnosis of CD and contribute to improving their health and well-being.

We thank Virginia Rodriguez, project manager of GLUTENSCREEN, for her support; Jan Heijdra-Suasnabar for English-language editing; and all the parents and children who participated in this project.

CBS

Central Bureau of Statistics

CD

celiac disease

DPSU

Dutch Pediatric Surveillance Unit

EMA

autoantibody against endomysium

HLA

human leukocyte antigen

Ig

immunoglobulin

LUMC

Leiden University Medical Center

METC-LDD

Medical Ethics Committee of Leiden-Den Haag-Delft

POCT

point-of-care test

PPV

positive predictive value

TGA

autoantibody against tissue transglutaminase type 2

ULN

upper limit of normal

YHCC

Preventive Youth Health Care Center