De Europese richtlijn van The European Society for Paediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN, 2020) adviseert, onder strikte voorwaarden, dat de diagnose coeliakie bij kinderen met een normaal totaal IgA (low for age or <0.2 g/L above the age of 3 years) in een groot aantal gevallen enkel op basis van serologie gesteld kan worden (Husby, 2020). Dit betekent dat duodenumbiopten veelal niet meer noodzakelijk zijn. In deze richtlijn is vastgesteld wat de beste strategie is om de diagnose coeliakie te stellen bij kinderen (rekening houdend met testeigenschappen, kosten en impact voor de patiënt). De ESPGHAN richtlijn is in deze Nederlandse richtlijn overgenomen en wordt hieronder in het Engels weergegeven.

Sub-question 1. Which serological test is most suitable for the initial diagnosis of suspected celiac disease?

Sub-question 2. Should more than one serological test be used and, if so, what should be the sequence of testing?

Sub-question 3. At which cut-off for tTG-IgA (ULNx10, x7, x5) may a diagnosis of CD

safely be done (positive predictive value > 95 %) with omission of biopsies?

Sub-question 4. Is endomysial antibody (EMA-IgA) testing necessary in every case to

diagnose CD without biopsy?

Sub-question 1. Which serological test is most suitable for the initial diagnosis of suspected celiac disease?

Description of studies

Eighteen articles were selected for a detailed QUADAS-2 Analysis, along with 5 prospective and 13 retrospective studies. Three tests were thoroughly evaluated: type-2 (tissue) transglutaminase (tTG-IgA), deamidated gliadin peptide antibodies (DGP-IgG,) and endomysial antibodies (EMA-IgA). In order for a test to be useful at ruling out a disease, it must have high sensitivity and to be useful at confirming a disease it must have high specificity.

Results

The prospective studies (Wolf, 2017; Mubarak, 2012; Oyaert, 2015; Parizade, 2009; Panetta, 2011) were in general large ones with a low risk of bias and of high quality. The largest study (Wolf, 2017) found EMA-IgA to have excellent accuracy. Three out of 5 recent articles showed surprisingly low specificity for EMA-IgA (Panetta, 2011; Mubarak, 2011; Parizade, 2009), with 2 of them coming from the same centre. As the 3 articles provided insufficient information about the technical aspects of serology and histology assessment with an allowed time gap up to 6 months between serology and biopsies (possibly leading to false negative histology due to gluten restriction), it was not possible to decipher the reasons for the discrepant results.

The retrospective studies had higher degrees of bias, in particular as to patient selection, and were judged to be of lower quality. A bivariate binomial meta-regression meta-analysis disclosed similar accuracies for the three antibody species that showed overall significant differences between the tests for both sensitivity (P=0.005) and specificity (P=0.016), with summary receiver operating characteristics curves showing summary points and 95% confidence intervals for tTG-IgA, DGP-IgG, and EMA-IgA. The highest value obtained was for EMA-IgA that had the highest sensitivity, but tTG-IgA had the highest specificity (ESPGHAN guideline 2020, Fig. S2, Supplemental Digital Content).

Level of evidence of the literature

The level of evidence regarding the outcome measure serological diagnostics was not downgraded and remained high.

Sub-question 2. Should more than one serological test be used and, if so, what should be the sequence of testing?

Description of studies

Literature was searched to find whether any combination of tests (either 2 separate tests or a blended test kit for both IgA and IgG detection) is better for initial testing than tTG-IgA plus total IgA. Evidence from studies restricted to young children below 2 or 3 years of age was downgraded, if the diagnosis of coeliac disease (CD) in cases with negative autoantibodies was not confirmed by a gluten challenge. Of 107 studies identified, 10 were of sufficient quality to be considered for the final analysis

Results

tTG-IgA plus DGP-IgG with or without DGP-IgA

The only unbiased prospective study was performed in adults (Horwitz, 2015), in which 2297 unselected participants were screened with tTG-IgA, DGP-IgG, and DGP-IgA. A total of 56 were positive on at least 1 antibody test and duodenal biopsies were performed in 40. Of 8 biopsy-confirmed CD cases, 7 were positive for tTG-IgA, 5 for DGP-IgG and 5 for DGP-IgA, with 4 positives in all 3 tests. False positive results were found in 2 for tTG-IgA, in 5 for DGP-IgG, and in 28 for DGP-IgA. In order to find the only CD-case with negative tTG-IgA, almost 2300 tests for DGP-IgG had to be performed plus 4 unneeded endoscopies. One study (Wolf, 2017) prospectively included children below 18 years of age, referred because of either a positive serology for CD and/or symptoms. A total of 898 children were centrally tested for total IgA, tTG-IgA, DGP-IgG, and EMA-IgA. When tTG-IgA plus total IgA was compared with tTG-IgA and DGP-IgG (tTG-DGP procedure) for initial testing, 592 were diagnosed with CD, 245 as no CD, and 24 had no final diagnosis. The tTG-DGP procedure detected 6 additional CD patients, 5 of which were also negative for EMA IgA, whereas the remaining child was positive for EMA-IgG and tTG-IgG. The tTG-DGP compared with the TG2-IgA procedure resulted in 16 unnecessary endoscopies (negative tTG-IgA but false positive DGP-IgG).

Of 5 retrospective studies, 4 were performed in young children only (Parizade, 2009; Hojsak, 2012; Frulio, 2015; Aberg, 2009), whilst 1 included children and adults (Vermeersch, 2013). These studies, however, have major limitations: the biopsy rate of patients with a positive test result in either tTG-IgA or DPG was low, there was no reference pathologist and no gluten challenge in children below 2 years of age with villous atrophy but negative autoantibodies. In summary, these studies do not support to add DGP to tTG-IgA testing for initial screening.

DGP-IgA with total IgA and AGA-IgA

Vriezinga (2014) reported a European multi-centre placebo- controlled intervention trial in infants at genetic risk for CD (all HLA DQ2 or DQ8 positive). Participants were regularly tested from age 4 months for tTG-IgA and AGA-IgA, and with an IgG-based test in case of low total IgA. Biopsies were offered to those with persistent positive tTG-IgA levels, high or increasing AGA-IgA, and symptoms strongly suggesting CD regardless of serology results. All IgA-competent children with biopsy-proven CD were positive for tTG-IgA, whilst all 17 tTG-IgA negative children biopsied, based on symptoms or AGA-IgA positive results, had either a normal mucosa or a transient enteropathy. Transient positivity of AGA-IgA occurred in a third of infants randomized to early gluten exposure and was not predictive for later CD.

IgA and tTG-IgG

Of 2911 persons (age range 1–80 years) with a positive coeliac serology during a 17-year period, 233 individuals with an isolated positivity for tTG-IgG were identified (Absah, 2017). Biopsies were performed in 178/233 (78%), with a normal histology in 160 (90%), Marsh 1 in 9 (4.5%), villous atrophy because of other diagnosis than CD in 3 (1.5%) and the remaining 6 patients (3%) having histopathology as CD. The authors concluded that tTG-IgG did not add to the accuracy of the diagnosis.

tTG-IgA with total IgA and AGA-IgA

One study (Hojsak, 2012) evaluated whether AGA-IgA testing in addition to tTG-IgA testing improves case finding in children below 2 years of age. Of 4122 children tested, 312 (8%) where tTG-IgA or EMA positive, whilst 85 were only AGA-IgA positive. Clinical data was available in 62 and duodenal biopsy results in 33 of them, leading to CD in 5 children. The remaining 57 children with isolated AGA positivity, including 4 with villous lesions,

received a diagnosis other than CD.

Level of evidence of the literature

The level of evidence regarding the outcome measure serological tests was downgraded to low because of conflicting results (inconsistency) and number of included patients (imprecision).

Sub-question 3. At which cut-off for tTG-IgA (ULNx10, x7, x5) may a diagnosis of CD

safely be done (positive predictive value > 95 %) with omission of biopsies?

Description of studies

A search yielded 44 studies where predictive positive value (PPV) of high tTG-IgA levels were compared with the histological outcome. Of these, only 36 utilised antibody tests suitable for calculating multiples of upper limit of normality (ULN). After narrowing the scope to those paediatric studies, where exact numbers of true positives and false positives could be extracted, 19 retrospective (Donat, 2016; Klapp, 2013; Nevoral, 2013; Trovato, 2015; Oyaert, 2015; Parizade, 2009; Panetta, 2011; Hojsak, 2012; Aita, 2013; Alessio, 2012; Dahlbom, 2010; Wolf, 2014; Lurz, 2009; Gidrewicz, 2015; Olen, 2012; Prause, 2009; Vivas, 2009; Schirru, 2013; Saginur, 2013) and 3 prospective studies (Werkstetter, 2017; Wolf, 2017; Mubarak, 2012) remained for QUADAS-2 and further analysis. These 49 datasets constituted 9 conventional ELISA assays (Biosystems, DiaSorin, Euroimmun, Eurospital, Immco, Inova, Orgentec, Phadia, R-Biopharm), 2 fluorescent immunoassays (Phadia), and 2 chemiluminescence tests (Immulite, Inova).

Results

From the 30 datasets evaluating the tTG-IgA cut-off at ≥10 X ULN, 28 reported >95% PPV and 21 reported >99% PPV. The PPVs were higher in studies where both Marsh 2 and 3 were accepted as CD (all>97% PPV) compared with studies, which required strictly Marsh 3 for CD diagnosis. At cut-off levels 5 to 7.5 X ULN PPV values varied between 92.3% and 100%, still 4/7 datasets showing PPV >99%. At cut-off levels 2 to 4T ULN PPV values varied between 86 and 100%, again with 4/7 datasets still showing PPV >99%. The study of Werkstetter (2017), included 8 tTG-IgA assays in the central head-to-head analysis, which showed a PPV of 99% even at lower cut-offs than 10 X ULN (presented only as graphs), but the same study also demonstrated high inter-test and inter-laboratory variability at these lower ranges. Notably, in local laboratories, a PPV >99% was only reached at 10 X ULN. At the cut-off levels between 3 and 10 X ULN, the test kit and the diagnostic approach (Marsh 3 only or Marsh 2–3 as CD) influenced the clinical outcome. In most retrospective studies, no reference pathologists were involved, and the histology evaluation was not blinded. Interestingly, all three prospective studies providing a blinded reference pathologist yielded excellent PPV values (100%, 99.1%, and 98.9%) suggesting that high tTG-IgA levels strongly support the CD diagnosis and discrepant results occur more likely because of technical difficulties with the histology.

tTG-IgG cut-off levels reliably predicting CD in IgA-deficient persons could not be derived from the literature. Therefore, the 10 X ULN cut-off is not validated for tTG-IgG. Differences in calibrators and slow IgG antibody kinetics warrant special caution with IgA-deficient subjects where levels of EMA and tTG-IgG may remain high for several years after starting a gluten-free diet (Korponay-Szabo, 2003).

Level of evidence of the literature

The level of evidence regarding the outcome measure cut-off for tTG-IgA testing without biopsy was downgraded to low because of conflicting results (inconsistency) and number of included patients (imprecision).

Sub-question 4. Is endomysial antibody (EMA-IgA) testing necessary in every case to

diagnose CD without biopsy?

Description of studies

The recommendations for a no-biopsy approach in patients with high tTG-IgA levels rests on a second serum sample taken for EMA-IgA on a separate occasion on a gluten-containing diet (Giersiepen, 2012). This consideration aims at avoiding mislabeling of samples or technical errors and confirming coeliac autoimmunity using another test assay with high specificity.

EMAs are directed against the transglutaminase 2 (TG2) antigen present in the anatomical endomysium in a tissue section and the EMA test is based on indirect immunofluorescence performed on primate esophageal or human umbilical cord substrate. The test is considered positive if a serum dilution of 1:5 or higher gives a visible binding pattern. The EMA-IgA test performance, however, depends on a subjective interpretation of the results, which may be critical at low titres. The inter-lab variability of EMA titers is highly depending on lab condition. In addition, the test is more time-consuming and expensive than measurement of tTG-IgA. Ten studies were identified for QUADAS2 analysis

Results

In three prospective studies, a total of 1788 symptomatic children were included of whom 1357 had a final diagnosis of CD (Werkstetter, 2017; Wolf, 2017; Mubarak, 2012). In total, 895 out of 1357 had tTG-IgA levels of ≥10 X ULN qualifying for the no-biopsy approach, 4 of which had a negative EMA; 1 of these 4 had a final diagnosis of no CD and 3 remained unclear. Thus, 1 to 4 of 895 patients qualifying for the no-biopsy approach (symptoms with tTG-IgA ≥10 X ULN) need to be tested with EMA-IgA to find 1 case with a final diagnosis of no CD. This yields a ‘‘number needed to test’’ of 224 to 895 to identify a non-CD case among those with tTG-IgA ≥10 X ULN.

In the study by Wolf (2017), 5 patients considered not to have CD (n=2) or unclear diagnosis (n=3) out of 405 patients with tTG-IgA ≥10 X ULN also had a positive EMA-IgA (Downey, 2015). Four out of 5 had <10 X ULN for tTG-IgA at the first sample assayed locally, suggesting transient high levels and a need for 2 separate samples to apply the no-biopsy criteria, or alternatively uncertain histology. In the study by Werkstetter (2017), 4 cases regarded as possible false positives also had tTG-IgA <10 X ULN or were negative in the second sample.

The studies have not formally assessed whether a second tTG-IgA test could serve as an alternative approach to cater for the possibility of transient increases or technical errors.

Seven retrospective studies did not report symptoms (Klapp, 2013; Nevoral, 2013; Panetta, 2011; Mubarak, 2011; Gidrewicz, 2015; Donaldson, 2008; Burgin-Wolff, 2013). Only 1 of these presented stratified tables in categories of tTG-IgA levels and with EMA-IgA for the group with tTG-IgA ≥10 X ULN. Two of the papers presented data suitable to answer the question. In total, 4 out of 565 individuals with tTG-IgA ≥10 X ULN were considered as false positives for a diagnosis of CD based on biopsy. These appeared in the same study, and all had a negative EMA-IgA. The authors reported that in 3 of those children, the tTG-IgA result was normal after 2–5 months whilst still on a gluten-containing diet. Thus, these cases could be due to a transient antibody increase, sample mixing or technical errors. The numbers needed to test with EMA-IgA to avoid a false positive diagnosis was 141.

Level of evidence of the literature

The level of evidence regarding the outcome measure EMA testing was not downgraded and remained high.

The ESPGHAN Guideline for Diagnosing Coeliac Disease 2020 (Husby, 2020) was used to answer this clinical question. For an overview of the search and selection process and details of the search strategy, we refer to the above mentioned ESPGHAN Guideline. Conclusions from the European Society Paediatric Gastroenterology, Hepatology and Nutrition Guideline have been included in this module.

In brief, the ESPGHAN Guideline for Diagnosing Coeliac Disease 2020 has been developed in line with GRADE methodology. Ten focused clinical questions were formulated according to PICO format: Population, Indicator, Comparator, and Outcome. For each question, a bibliographic search was conducted; informative studies, systematically assessed for the risk of bias and clinical applicability, were included in the evidence base; meta-analysed study results were summarized and graded for certainty of evidence; and the implications for clinical practice were discussed and recommendations formulated and graded for strength. GRADE provided a basis for statements and recommendations.