Remember me
Abstract
Background: The prevalence of celiac disease (CD) is relatively high in Saudi Arabia, and little is known about the accuracy of serological markers in the local population. This study aimed to assess the diagnostic performance of various serological markers for detecting CD in Saudi children and adults.
Methods: We conducted a retrospective study of 148 CD patients and 512 controls to assess the diagnostic performances of IgA anti-tissue transglutaminase antibodies (TTG), IgG anti-TTG, IgA anti-deamidated gliadin peptide antibodies (anti-DGP), IgG anti-DGP, and endomysium antibodies (EMA).
Results: Immunoglobulin A (IgA) anti-TTG was the most sensitive test [98.9% (95% confidence interval (CI) 94.1–99.8%)], while EMA was the most specific [100%, 95%CI 98.6–100%]. By applying the criteria of IgA anti-TTG titers ≥10 × upper limit of normal (ULN) and positive EMA, 57.3% of patients could have avoided intestinal biopsy. IgG anti-DGP test had a sensitivity of 85.9% (95% CI = 77.3–91.5%) and a specificity of 93.5% (95% CI = (90.0–95.9%). Titers of IgA anti-TTG, IgA anti-DGP, and IgG anti-DGP were higher in CD patients with the Marsh 3c class than in those with the Marsh 3b and Marsh 3a classes. IgG anti-TTG and IgA anti-DGP had no additional diagnostic value.
Conclusions: IgA anti-TTG and EMA are excellent CD markers in children and adults. The use of IgA anti-TTG titers ≥10 × ULN and positive EMA as criteria for CD diagnosis in children and adults might be a good alternative to intestinal biopsy.
Keywords: Anti-deamidated gliadin peptides antibodies, anti-tissue transglutaminase antibodies, celiac disease, endomysium antibodies, Marsh–Oberhuber classification
How to cite this article: Introduction 
Celiac disease (CD) is an immune-mediated enteropathy with protean manifestations, triggered by the intake of dietary gluten from wheat, barley, and rye. It occurs in genetically predisposed individuals with the human leukocyte antigen (HLA) DQ2 and/or DQ8 haplotypes. CD affects 0.5–3% of the general population worldwide; however, it is still under-diagnosed. Studies suggest that CD prevalence is high and varies between 1.5% and 2.2% in Saudi Arabia.[1],[2]
Classically, CD patients present with malabsorption and diarrhea. However, the occurrence of atypical and asymptomatic presentations is increasing. Symptomatic and untreated diseases are associated with reduced life expectancy.[3] Therefore, early diagnosis is important to avoid complications, such as bone disorders and malignancy.
Intestinal biopsy, which detects villous atrophy, crypt hyperplasia, and intra-epithelial lymphocytosis (IEL), is considered the gold standard for diagnosing CD. However, the availability of highly reliable, non-invasive, and cost-effective serological tests have transformed the diagnosis and management of CD. The availability of such tests led the European Society for Pediatric Gastroenterology, Hepatology, and Nutrition (ESPGHAN) to state that CD can be diagnosed without biopsy if there are clinical signs, strongly positive serology in repeated testing, and the presence of HLA-DQ2 or HLA-DQ8.[4] More recently, the ESPGHAN has given more credit to serology and has recommended, under certain conditions, that diagnosis of CD may be solely based on serology without biopsy or HLA genotyping.[5] Suspension of endoscopic activities on a large scale during the coronavirus disease (COVID-19) pandemic led the British Society of Gastroenterology to propose a non-biopsy protocol, solely based on serology, for diagnosing CD in patients under 55 years.[6]
Traditionally, CD serological diagnosis is based on the detection of several autoantibodies. Identifying tissue transglutaminase (TTG) as the target antigen for Immunoglobulin A (IgA)-endomysium antibodies (EMA) was a turning point in diagnosing CD. Anti-tissue transglutaminase antibodies (anti-TTG) have shown high sensitivity and specificity. The higher the titer of anti-TTG, the greater the likelihood of an actual positive biopsy result.[7] Based on these findings, the difficult standardization of EMA, and the cost (especially when monkey substrate is used), IgA anti-TTG has become the choice test for CD diagnosis. However, EMA is still considered the most specific test for CD.[8] Gliadin is a protein present in gluten. For several years, anti-gliadin antibodies have been used for the serological diagnosis of CD. Modifications of gliadin by the intestinal enzyme TTG 2, through a process known as deamidation, confer better binding of gluten peptides to HLA-DQ2 and HLA-DQ8 molecules. CD patients develop antibodies against deamidated gliadin peptides (anti-DGP), and studies have revealed that these antibodies have better specificity than anti-gliadin antibodies.[9],[10]
CD is more common in patients with IgA deficiency than in the general population.[11] Therefore, to ensure that CD is properly diagnosed in this category of patients, it is recommended to either measure serum IgA levels in all patients or to include both IgA- and IgG-based testing.
The diagnostic performance of anti-DGP, anti-TTG, and EMA have been assessed in several population-based studies, and different testing combinations have been proposed. However, little is known about the diagnostic performance of these tests in the Saudi Arabian population. Therefore, this study aimed to assess the diagnostic performance of these tests among Saudi Arabian patients.
Patients and Methods Patients and controls
This retrospective study of Saudi Arabian patients was performed between January 2018 and December 2021. All patients with gastrointestinal and/or extra-intestinal symptoms suggestive of CD or type I diabetes, a condition known to be associated with CD, were enrolled. All patients with clinical suspicion of CD who underwent intestinal biopsy and had a CD serology workup performed before the intestinal biopsy were included. Exclusion criteria were the previous diagnosis of CD and inadequacy or poor quality of histological samples. Intestinal biopsy was considered the reference standard irrespective of serology results, to avoid overestimating serology accuracy.[12] Patients with intestinal biopsy findings compatible with Marsh–Oberhuber 2 and 3 classifications were considered CD patients, whereas patients with Marsh 0 and 1 were considered controls.[13]
Serological study
Serum IgA anti-DGP, IgG anti-DGP, IgA anti-TTG, and IgG anti-TTG were detected using the enzyme-linked immunosorbent assay (ELISA) (INOVA Diagnostics Inc., San Diego, CA, USA), following the manufacturer's instructions. Twenty units (U)/ml was considered the upper limit of normal (ULN) for all the mentioned tests.
EMA was detected by indirect immunofluorescence using monkey esophagus as substrate (IMMCO Diagnostics, Buffalo, NY, USA), and titers of 1:2.5 or higher were considered positive. A positive result was reported when the connective tissue surrounding the muscle cells fluoresced bright in a honeycomb pattern.
Duodenal biopsy
All subjects underwent esophago-gastro-duodenoscopy with duodenal sampling. Before light microscopy, paraffin-embedded samples were cut and stained with hematoxylin and eosin. The diagnosis of CD was based on the demonstration of normal villi, >30 IEL, crypt hyperplasia equivalent to Marsh 2 or villous atrophy, >30 IEL, and crypt hyperplasia equivalent to Marsh 3 lesions. Marsh 3a, b, and c classes were defined as follows: Marsh 3a: >30 IEL, crypt hyperplasia, and mild villous atrophy; Marsh 3b: >30 IEL, crypt hyperplasia, and marked villous atrophy; Marsh 3c: >30 IEL, crypt hyperplasia, and complete villous atrophy.[13]
Ethical approval
This study was conducted in accordance with the protocol and principles of the Declaration of Helsinki and approved by the Institutional Review Board.
Statistical analysis
Statistical analyses were performed using the (SPSS version 17.0, Chicago, IL, USA). First, data were checked for normality. Then, as appropriate, continuous variables were expressed as means or medians, while proportions were estimated to describe qualitative variables. The Kruskal–Wallis test was used to compare the medians of autoantibody titers. The Chi-square test was used to test the association between categorical data. Several combinations of tests were assessed. The test results for "OR" combinations were considered positive if any test in the combination was positive, and negative if all tests in that combination were negative. For "AND" combinations, the test results were considered positive if all the tests in the combination were positive, and negative if any test in the combination was negative. Youden index (sensitivity + specificity –100%) was used to compare tests' combination performance. Receiver operating characteristic (ROC) curves were created for all quantitative tests (IgA anti-DGP, IgG anti-DGP, IgA anti-TTG, and IgG anti-TTG) by plotting sensitivity against 1-specificity.
Results Demographics
A total of 724 patients with suspected CD who underwent intestinal biopsy were initially assessed for enrolment [Figure 1]. Among them, 148 patients with intestinal biopsy compatible with Marsh classifications 2 and 3 and with data available for CD serology before intestinal biopsy, were enrolled as cases. In addition, 512 patients with a negative intestinal biopsy and CD serology performed before intestinal biopsy, were included as the control group. Among the CD patients, two (1.3%) had IgA deficiency, and five (3.4%) had type 1 diabetes. Fifty-two (35.1%) CD patients were males, and 96 (64.9%) were females. The mean age of the CD patients was 26.2 ± 16 years (minimum = 3; maximum = 73). Among the CD patients, 58 (39.9%) were children aged less than 18 years [29 (50%) males and 29 (50%) females]. The control group comprised 349 (68.2%) male and 163 (31.8%) female individuals with a mean age of 32.7 ± 16 years (minimum = 1, maximum = 85). Furthermore, 92 (18%) children aged less than 18 years were in the control group.
Figure 1: CONSORT flow diagram of study recruitment and serology testingIntestinal biopsy findings
An intestinal biopsy was performed after serology in all patients and controls. Between serology and intestinal biopsy, the time elapsed was 8.3 (± 2.1) months. Among CD patients, 2 (1.3%) were Marsh 2, 57 (38.5%) were Marsh 3a, 67 (45.3%) were Marsh 3b, and 22 (14.9%) were Marsh 3c. Among controls, 476 (93%) were Marsh 0, and 33 (6.4%) were Marsh 1.
There was no association between histological lesions and the sex of the patients and controls (p = 0.4). In addition, the median age was comparable among patients and controls with Marsh 0 (31.6 years), Marsh 1 (21.3 years), Marsh 2 (28.4 years), and Marsh 3 (29.9 years) (p = 0.3).
Diagnostic performance of serological tests
IgA anti-TTG was positive in 145 (98%) of 148 CD patients and negative in 3 (2%) patients. Two of the three patients with negative IgA anti-TTG had selective IgA deficiency. The median IgA anti-TTG blood level was 208 U/ml among CD patients and 3.6 U/ml among controls (p < 0.001). Furthermore, among the positive CD patients, 127 (87.6%) had IgA anti-TTG titers ≥3.5 × ULN, and 81 (55.9%) patients had a titer ≥10 × ULN. In contrast, the IgA anti-TTG antibodies in all controls had titers <10 × ULN. Twenty-nine (5.3%) controls of the 498 tested were IgA anti-TTG positive. Their median IgA anti-TTG titer was significantly lower than that of CD patients (39 U/ml vs. 208 U/ml, P < 0.001). Twenty-two (75.9%) of them were Marsh 0 and 7 (24.1%) were Marsh 1.
IgG anti-TTG was positive in 56 (38.1%) of the 147 patients tested and in 2 (0.4%) controls among 495. The median IgG anti-TTG concentration among the patients was 12 U/ml, whereas it was 1 U/ml among the controls (p < 0.001).
IgA anti-DGP was positive in 94 (78.3%) patients of the 120 patients tested and in 29 (7.1%) of the 409 controls. The median IgA anti-DGP was 66.5 U/ml among CD patients and 5 U/ml among controls (p < 0.001).
IgG anti-DGP was positive in 104 (85.2%) of the 122 patients tested and in 26 (22.8%) of the 414 controls. The median IgG anti-DGP was 71 U/ml among CD patients and 1 U/ml among the controls (p < 0.001).
EMA was positive in 107 (91.4%) of the 117 patients tested, and all 337 controls tested negative for EMA.
Agreement (simultaneous positivity or negativity) between IgA anti-TTG and EMA, IgG anti-DGP, IgA anti-DGP, IgG anti-TTG was 92.2%, 90.5%, 89.4%, and 80.5%, respectively.
Diagnostic performances were determined in patients and controls simultaneously tested for all markers [among 92 (62.1%) CD patients and 278 (54.3%) controls] to allow for direct comparison between the different tests. The sensitivities of the tests were as follows: 98.9% (91/92) for IgA anti-TTG, 88.0% (81/92) for EMA, 85.9% (79/92) for IgG anti-DGP, 82.6% (76/92) for IgA anti-DGP, and 42.4% (39/92) for IgG anti-TTG. Furthermore, the specificities of the tests were as follows: 100% (278/278) for EMA, 99.6% (277/278) for IgG anti-TTG, 93.9% (261/278) for IgA anti-DGP, 93.5% (260/278) for IgG anti-DGP, and 93.2% (259/278) for IgA anti-TTG [Table 1]. The area under the curve was 0.995 (CI 95% 0.991- 0.999) for IgA anti-TTG, 0.979 (CI 95% 0.967-0.990) for IgG anti-DGP, 0.954 (CI 95% =0.929-0.978) for IgA anti-DGP, and 0.928 (CI 95% 0.897-0.960) for IgG anti-TTG [Figure 2].
Figure 2: Roc curves for IgA anti-TTG, IgG anti-DGP, IgA anti-DGP, and IgG anti-TTG. These curves were constructed based on the data obtained from patients with CD (n = 92) and controls (n = 278)
Table 1: Direct comparison of diagnostic performance of celiac disease serological testsThe diagnostic performances of the test combinations are shown in [Table 2] and [Table 3]. For the "OR" combination, the best diagnostic performance was achieved by IgA anti-TTG and EMA with a sensitivity of 98.9% (91/92) and a specificity of 93.2% (259/278). Similarly, for the "AND" combination, the best diagnostic performance was achieved by the combination of IgA anti-TTG and EMA, with a sensitivity of 88% (81/92) and a specificity of 100% (278/278).
Table 2: Diagnostic performance of celiac disease serological tests in children and adults
Table 3: Diagnostic performance of serological tests' combinations for celiac diseaseSerological tests and age
The diagnostic performance of all serological tests were comparable between the children and adults [Table 4]. Among the 117 patients [49 (41.9%) children and 68 (58.1%) adults] who were tested for both IgA anti-TTG and EMA, 68 (58.1%) patients [34 (69.4%) children and 34 (50%) adults] had an IgA anti-TTG titer ≥10 ULN. Among them, 67 (57.3%) [33 (67.3%) children and 34 (50%) adults] also had EMA.
Table 4: Diagnostic performance of serological tests' combinations for celiac diseaseSerological tests and mucosal damage
Titers of IgA anti-TTG, IgA anti-DGP, and IgG anti-DGP correlated with villous atrophy and were highest in patients with Marsh 3c [Figure 3]. The median IgA anti-TTG titers were 190 U/ml, 210 U/ml, and 269 U/ml in patients with Marsh 3a, 3b, and 3c, respectively (p = 0.004). The median IgA anti-DGP titers were 35 U/ml, 81.5 U/ml, and 248 U/ml in patients with Marsh 3a, 3b, and 3c, respectively (p = 0.003). The median IgG anti-DGP titers were 53 U/ml, 77.5 U/ml, and 103 U/ml in patients with Marsh 3a, 3b, and 3c, respectively (p = 0.004). Conversely, the titers of EMA and IgG anti-TTG did not correlate with mucosal damage. The median IgG anti-TTG titers were 8.5 U/ml, 17.5 U/ml, and 21.5 U/ml in patients with Marsh 3a, 3b, and 3c, respectively (p = 0.066). Median EMA titers were 40, 80, and 120 in patients with Marsh 3a, 3b, and 3c, respectively (p = 0.132).
Figure 3: Correlation between serological tests and mucosal damage. Unlike IgG anti-TTG and EMA titers, titers of IgA anti-DGP, IgG anti-DGP, and IgA anti-TTG were correlated with mucosal damage. They were more elevated in patients with Marsh 3c than in patients with Marsh 3b or Marsh 3a Discussion In this study, we assessed the diagnostic performance of several serological tests currently available for CD in both children and adults. One of the major possible biases affecting the assessment of serological tests in CD is ascertainment bias. Indeed, selecting cases and controls based on positive or negative serology may result in falsely high sensitivity and/or specificity. In this study, we attempted to limit ascertainment bias by selecting cases and controls based only on intestinal biopsy findings, regardless of serology results.[12] Moreover, the controls included were patients initially suspected of having CD and not healthy individuals, or patients with other specific diseases.[14],[15] Our findings revealed that when tests were directly compared, IgA anti-TTG and EMA achieved the best diagnostic results in untreated patients with CD. Our findings are in line with those reported in the literature.[12],[16],[17] Anti-DGP antibodies have progressively replaced lesser performing anti-gliadin antibodies. In addition to the important role of deamidation in CD pathogenesis, it has also been shown that there is a cross-linkage between DGP and TTG, which creates common immunogenic epitopes and enhances the antigenic presentation of gliadin. Furthermore, there is a cross-linkage to collagen, and the resulting complex may be the target for EMA.[18],[19] It has been shown that in some populations, the diagnostic performance of IgA anti-DGP is comparable to that of IgA anti-TTG.[10],[20],[21] However, a meta-analysis has shown that even though both IgA anti-DGP and IgA anti-TTG have high specificities (>94% and >96%, respectively), IgA anti-DGP has a lower sensitivity than IgA anti-TTG (88% vs. 93%, respectively) for the diagnosis of typical CD with villous atrophy.[22] In our study, we found similar results with a specificity of 93.9% and 93.2%, respectively, and a sensitivity of 82.6% and 98.9%, respectively.
In our study, IgG anti-DGP achieved a better overall diagnostic performance than IgA anti-DGP and IgG anti-TTG. Saadah et al.[23] found that IgG anti-DGP was comparable to IgA anti-TTG and may be useful as an alternative to IgA anti-TTG in the diagnosis and follow-up of children with CD.[24] They reported that combining IgG anti-DGP with IgA anti-TTG only marginally improved sensitivity in Saudi Arabian children with CD. Our results are consistent with their findings in terms of sensitivity. However, they indicated that IgG anti-DGP had very good specificity and might be a good second-line marker, particularly in cases with low IgA anti-TTG titers. The IgG-based test showed particularly interesting results in patients with CD and IgA deficiency. In our study, the two patients with IgA deficiency had positive IgG anti-DGP and negative IgA-TTG and IgA anti-DGP results, as expected. The 2019 ESPGHAN guidelines for diagnosing CD recommend against using IgA/IgG anti-DGP-IgG antibodies for initial testing. They suggest using an IgG-based test only if total IgA is low/undetectable.[5] Our findings indicate that the IgG anti-DGP test should be performed in situations where IgA levels are low/undetectable and in laboratories where facilities for EMA tests are not available. EMA is detected by an indirect immunofluorescence method using monkey esophagus or human umbilical cord tissue as the antigen. Therefore, it is more labor-intensive and operator-dependent, which limits its use, compared with ELISA-based tests. In our study, IgG anti-TTG had high specificity and was positive in patients with selective IgA deficiency. However, due to its much lower sensitivity (42.4%), this test was outperformed by IgG anti-DGP. Previous studies have shown that IgG anti-DGP has a higher specificity and a better sensitivity than IgG anti-TTG fd,[25],[26] and this seems to be also true in cases of IgA deficiency.[27],[28]
We found that the diagnostic performances of IgA anti-TTG and EMA were high in children and did not differ significantly from those in adults. It was previously reported that serological tests, particularly EMA, are less sensitive in children, especially children under 2 years of age. However, recent data have shown that EMA and IgA anti-TTG perform equally or better than anti-DGP in this population.[27],[28],[29] Unfortunately, in our population, all included children were older than 2 years, which made it impossible to assess the tests in children aged <2 years.
Since none of the serological tests for CD have a reputation of being an ideal test with perfect diagnostic performance, it is recommended to combine two tests.[30] A combination of IgA anti-TTG and EMA is recommended by most authors.[5],[27] Our results have shown that the best combination of two tests is achieved by combining IgA anti-TTG and EMA. Our results are consistent with those reported in the literature.[31],[32],[33]
The most recent guidelines of the ESPGHAN allow diagnosis of CD without an intestinal biopsy in children, with IgA anti-TTG ≥10 ULN and positive EMA taken by a separate blood sample.[5] These guidelines have since been validated in large prospective studies.[34],[35] Based on this evidence, the British Society of Gastroenterology has suggested, during the COVID-19 pandemic, to rule in the CD diagnosis in patients under 55 years with IgA anti-TTG titers ≥10 × ULN and positive EMA.[6] In our study, among the 49 children tested for both IgA anti-TTG and EMA, 34 (69.4%) had an IgA anti-TTG titer ≥10 ULN. Among these 34 patients, 33 (67.3%) were EMA positive. Therefore, if the ESPGHAN guidelines were applied, the diagnosis of CD would be possible in 67.3% of pediatric cases, without subjecting them to intestinal biopsy. In a study of Saudi Arabian pediatric CD patients detected using mass screening, Al-Hussaini et al. reported that an IgA anti-TTG titer >10 ULN correlates with villous atrophy.[36],[37]
Similarly, several studies advocate that serological evidence alone may be sufficient for diagnosing CD in adults.[38],[39] In our population, among the 68 adult patients tested for both IgA anti-TTG and EMA, 34 (50%) had an IgA anti-TTG titer ≥10 ULN, and all of them were EMA positive.
Only intestinal biopsy was considered the reference standard to avoid overestimating serological accuracy. Therefore, in the control group, we included patients classified as having potential CD cases (positive serology and negative histology) in the control group. In our study, twenty-nine (5.3%) controls of the 498 tested were IgA anti-TTG positive. Their median IgA anti-TTG titer was significantly lower than that of CD patients (39 U/ml vs. 208 U/ml, P < 0.001) and most of them (75.9%) were Marsh 0.
We found that IgA anti-TTG titers were correlated with mucosal damage. The titers of these antibodies were highest in patients with the Marsh 3c classification (complete villous atrophy). Several studies have reported a correlation between IgA anti-TTG titers and histological damage. Our findings corroborate those of Alharbi et al. They reported a correlation between IgA anti-TTG titers and histologic damage in 134 Saudi Arabian patients with CD.[40] In our study, mucosal damage was also correlated with IgA anti-DGP and IgG anti-DGP titers but not with EMA and IgG anti-TTG titers. Ziv–Baran et al. did not find any correlation between EMA and histological damage in 432 pediatric CD cases.[41]
Our study has some limitations. First, the sample size was small. Second, our study is retrospective in nature, which prevents association studies between serological markers and the relevant clinical and biological variables. Finally, in most cases, intestinal biopsy findings were not cross-examined. Therefore, possible operator and pathologist biases cannot be excluded.
In conclusion, our study showed that IgA anti-TTG and EMA are excellent markers of CD in both children and adults. Using IgA anti-TTG titers ≥10 × ULN and positive EMA as criteria for CD diagnosis in children and adults might be a good alternative to intestinal biopsy. Furthermore, our results suggest that IgG anti-DGP is also a good marker for CD. Its combination with IgA anti-TTG shows particularly interesting results in CD diagnosis among patients with selective IgA deficiency, and is an alternative test in laboratories where EMA is unavailable.
Financial support and sponsorship
College of Medicine Research Center, Deanship of Scientific Research, King Saud University, Riyadh, Saudi Arabia
Conflicts of interest
There are no conflicts of interest.
References 
Source of Support: None, Conflict of Interest: None
CheckDOI: 10.4103/sjg.sjg_280_22
Comments (0)