Coeliac disease (CD) is a common immune-mediated enteropathy, whereby increased immunological sensitivity to dietary gluten results in chronic small intestinal mucosal inflammation in genetically susceptible individuals.1 CD affects ~1% of the global population, but despite its rising prevalence, the majority of patients remain undiagnosed.2 Currently, most patients with suspected CD are screened serologically for antibody positivity, in particular for IgA antibodies to tissue transglutaminase 2 (tTG) and antiendomysial IgA antibodies (EMAs).1 As these serological markers are not 100% specific for detecting intestinal lesions compatible with CD, positive coeliac serology is confirmed by duodenal biopsies demonstrating the hallmark pathological changes of mucosal remodelling, such as villous atrophy, crypt hyperplasia and intraepithelial lymphocytosis.3

However, upper GI endoscopy with biopsies is expensive, with costs relating directly to the procedure and indirectly to the processing and analysing of samples. The procedure is poorly tolerated by many patients4 and carries risks, including sedation complications and although rare in current practice, GI haemorrhage and/or perforation.5 Furthermore, variability exists in the identification of the main histological changes of CD, meaning cases can be missed despite duodenal biopsy sampling.6 7 Therefore, the identification of a non-invasive approach to the diagnosis of CD is of great interest.

The 2012 European Society for the Study of Paediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN) guidelines for the diagnosis of CD challenged the necessity for duodenal biopsies in paediatric patients. They suggested that an IgA tTG antibody titre greater than 10 times the upper limit of normal (ULN), in combination with a positive EMA antibody test and compatible human leucocyte antigen (HLA) genotype, is sufficient to support a diagnosis of CD in symptomatic individuals.8 This eliminates the need for gastroscopy and its associated costs/risks in selected paediatric patients.8 The latest guidelines published this year remove the requirement for the presence of symptoms and HLA testing in the diagnostic pathway.9 This underscores the specificity of a serology-based or ‘no-biopsy’ approach for the diagnosis of paediatric CD.

Studies have evaluated whether this strategy can be applied in symptomatic adult CD.10–21 These studies have suggested that tTG levels of ≥10×ULN could be predictive of CD in adults, and the recently published Finnish national guidelines for the diagnosis of CD have incorporated this diagnostic pathway into their practice.22 However, this approach has not been widely adopted into adult clinical practice or guidelines. Adult gastroenterologists have cited the lack of international, multicentre data including testing in low CD prevalence populations and thus intimated caution.23 Furthermore, there are inconsistencies in the histological diagnosis of CD in the published literature.23 Therefore, the aim of the current study was to assess whether a tTG level of ≥10×ULN could be sufficient for detecting duodenal mucosal abnormalities diagnostic of CD and to determine the diagnostic yield of this approach across different adult cohorts. In doing so, we aimed to address the shortcomings within the adult literature.

Altogether this study included 1417 patients; 431/1417 (30%) patients had tTG titres of ≥10×ULN, of which 424 (98%) patients had Marsh 3 lesions on duodenal biopsy and were diagnosed with CD. This study addresses the concerns raised by the adult gastroenterology community and supports a change in guidelines towards a no-biopsy approach for the diagnosis of CD in adults.

We identified Marsh 3 histology as our primary reference standard across all cohorts because (1) small intestinal mucosal remodelling, in keeping with Marsh 3 histology, represents the hallmark pathological changes in CD26; (2) few studies in the adult literature have reported on the predictive capacity of tTG titres of ≥10×ULN against Marsh 3 histology; rather, a clinical diagnosis of CD (based on the constellation of appropriate symptoms, serology and associated risk factors, in the absence of villous abnormalities) has been used as a reference standard by many27; and (3) few studies have included appropriate details regarding the location (including the duodenal bulb) and interpretation of biopsies.28 By contrast, our centre in Sheffield, UK, has published extensively on the matter of duodenal biopsy sampling in adult CD.29–32

In taking this approach, initially, we evaluated the predictive value of tTG titres of ≥10×ULN against the presence of Marsh 3 histology in patients seen in the specialist CD clinic (cohort 1). Most individuals within this cohort were diagnosed with CD, which reflects a high referral bias and pretest probability for CD in this cohort. This aligns with many previous studies, which have also assessed the diagnostic accuracy of tTG titre multiples of the ULN in high pretest probability/disease prevalence populations (CD prevalence >70%).10 12–14 16 18 In accordance with these previous studies, we found that a tTG titre threshold of 10×ULN accurately identified individuals with Marsh 3 lesions.

However, the pretest probability/disease prevalence within a cohort can influence the PPV of a diagnostic test.33 This means that as the disease prevalence increases, the more likely it is that a positive test represents a true positive case.33 The predictive value of the 10×ULN threshold in adults has less commonly been evaluated in more general clinical scenarios where the disease prevalence of CD is considered <10%.11 28 33 Sugai et al previously demonstrated that a tTG titre threshold of around 7×ULN had a 100% PPV for CD in an adult cohort with a CD prevalence of 3.3%.11 More recently, Fuchs et al demonstrated that a tTG titre threshold of 10×ULN was accurate at identifying individuals with Marsh 3 lesions across adult cohorts with different pretest probabilities for CD.28 However, in this study, the low CD pretest probability cohort included only older (>50 years) adults.28 Furthermore, the performance characteristics of the 10×ULN tTG threshold in a more general clinical scenario have been estimated, but not studied.33

Therefore, to confirm and extend these findings, we next sought to determine the predictive capacity of the 10×ULN threshold against the presence of Marsh 3 lesions on duodenal biopsy in a cohort of individuals considered as low suspicion for CD (cohort 2). Low suspicion was determined on the basis that, in those referred for upper GI endoscopy, the referral bias for CD was removed by excluding those who tested positive for CD in primary care. Indeed, the prevalence of CD in this population was 3.2%, which aligns with the local prevalence of CD in symptomatic adults in primary care (3.6%).34 These results demonstrate that tTG titres of ≥10×ULN are equally good at predicting Marsh 3 lesions in the setting of both low and high disease pretest probability/prevalence and complement previous work that this diagnostic pathway can be used in adults with different risk levels for CD.11 28 Importantly, while these data support the use of the biopsy avoidance approach in adults, we still advocate that suspected cases with tTG titres exceeding the 10× threshold are referred to a gastroenterologist for assessment, rather than the diagnosis of CD being made in primary care. In addition, these data were generated in adult cohorts within secondary and tertiary care settings, and thus, these results do not apply to screening the general population for a diagnosis of CD. These are important considerations to ensure that patients receive the correct diagnosis and to prevent the inappropriate institution of a GFD, which has been raised as concerns regarding this approach by some.23 35

One of the major challenges in implementing this diagnostic pathway is the variability and lack of standardisation between tTG assays.36 The inability to compare tTG results accurately between different assays increases the chances of misdiagnosis if a uniform threshold is used. In adults, there is a lack of data evaluating this in a multicentre international setting. To address this, we evaluated the 10×ULN threshold across multiple sites/countries and multiple assays (cohort 3). The PPV of the 10×ULN threshold at detecting Marsh 3 lesions dropped from nearly 99% across cohorts 1 and 2 at our UK centre to 95.2% in a multicentre international setting. This raises the concern that sole reliance on this as a marker for CD will lead to a false-positive diagnosis in around 1 in 20 cases.

Across all cohorts, the diagnostic performance of this pathway is based strictly on the presence of Marsh 3 histology. The PPV in all three cohorts increased to >99%, where both Marsh 2 and Marsh 3 lesions are accepted as CD. It is widely accepted that in the setting of relevant symptoms/high baseline risk of CD, raised serological markers, HLA DQ2/8 positivity and/or response to a GFD, the diagnosis of CD can be established in the absence of villous effacement. Indeed, this situation may represent a false-negative histology rather than a false-positive serology28 due to a variety of reasons, including incorrect histological interpretation, which can vary in 10%–25% of cases between independent evaluations6 7 and/or patchy disease.32 Additionally, individuals with symptomatic CD may present with subtle villous abnormalities that fall within the scope of microscopic enteritis, which may lead to diagnostic inaccuracy.37

In comparison with the datasets that have evaluated the tTG IgA cut-off at ≥10×ULN included in the latest paediatric ESPGHAN guidelines, 2/30 reported a PPV of <95%; 7/30 had a PPV between 95% and 99%; and 21/30 reported a PPV of >99% for detecting Marsh 3 lesions.9 In studies where Marsh 2 or Marsh 3 duodenal lesions were accepted for a diagnosis of CD, the PPV was >97% in all studies.9 Thus, based on the current study, the performance characteristics of this diagnostic pathway in adults with different risk levels for CD or in a multicentre international setting is at least comparable with the studies included within the paediatric guidelines.

The high tTG threshold recommended by the paediatric community in part is included to control for tTG assay variation.36 We found that for some tTG assays, a lower cut-off may retain diagnostic specificity but increase the sensitivity of this approach. In contrast, the 10×ULN threshold may not eliminate all false-positive cases across all assays. Therefore, a uniform threshold may hinder the clinical impact and accuracy of the no-biopsy approach in adult patients. Other studies have reported similar findings of differences in the optimal tTG thresholds and performance characteristics of different assays.14 19 28 This does not imply that one assay is inferior to the other nor that using tTG threshold values in this manner is not appropriate, rather that the threshold ULN cut-off should be specific to the tTG assay, as there is currently no way to accurately standardise tTG results across different assays.38 Therefore, local validation of assay specific thresholds and/or limiting the use of this pathway to certain tTG assays may help to optimise the clinical impact and accuracy of this approach.

The latest ESPGHAN guidelines no longer consider either the presence of symptoms or HLA typing as mandatory in this diagnostic pathway.9 HLA typing has been shown to add no diagnostic value in this pathway in children or adults, and testing is costly and not widely available.9 28 Therefore, the HLA status of individuals was not specifically evaluated in this study. In addition, while each cohort comprised predominantly symptomatic adults, we did not specifically evaluate symptomatology in this study. Current ESPGHAN guidelines recommend confirming tTG titres with EMA positivity in a second blood sample.9 EMA status was not evaluated in all patients in the present study. While EMA positivity has been shown to enhance the predictive capacity of the biopsy avoidance approach, EMA testing incurs a high cost and is labour intensive, and the interpretation is subjective.35 For these reasons, some centres have stopped performing EMA tests.17 Therefore, including EMA testing in this algorithm may lead to poor uptake of this pathway in some areas. Alternative strategies have been suggested, including simultaneous tTG measurements28 or testing for deamidated gliadin peptides (DGPs) alongside tTG titres, the latter of which has shown promise in both adult and paediatric cohorts.11 39 Further adult studies are awaited to address the place for DGPs in this algorithm.

Finally, it is noteworthy that the applicability of the biopsy avoidance approach is not as straightforward in adults as it is in children. Indeed, it should be stressed there are a number of reasons that a biopsy should still be performed in adults, particularly if there are ‘red flag’ signs/symptoms, such as, but not limited to, persistent dyspepsia in those aged ≥55 years, weight loss, severe malnutrition, dysphagia and unexplained iron deficiency anaemia.40–42 Some of these presentations align with those of adult CD; this underscores the importance of assessment by a gastroenterologist before a diagnosis of CD is made in the absence of a duodenal biopsy. In addition, a biopsy should also be considered in cases of discordance between serology and risk, and in borderline cases.27 Moreover, individuals with persistent symptoms despite adherence to a GFD require duodenal biopsy sampling, as tTG titres provide no information regarding histological improvement after introduction of a GFD.43

In conclusion, our data show that anti-tTG titres of ≥10×ULN perform well at identifying adults with intestinal changes associated with CD. This study supports a no-biopsy strategy in adult gastroenterology services. This approach has implications in reducing the cost, risk and caseload associated with diagnostic endoscopy in adult CD. However, local validation of assay-specific thresholds would enable this approach to have the greatest impact on adult patients.