A retrospective, observational study was conducted that reviewed the clinical records of patients with suspected celiac disease seen at the outpatient service of the gastroenterology department of the Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, within the time frame of January 1, 2015, and December 31, 2020. The STROBE checklist for observational studies was utilized.

Patients of both sexes were included, above 18 years of age, referred to the gastroenterology outpatient service due to chronic diarrhea or clinical suspicion of celiac disease, who met the study protocol that included serum IgA, anti-tTG IgA, and biopsy of the duodenal mucosa. The patients also had to have a subsequent review at which the results of the laboratory tests and histopathologic findings were evaluated. Patients who had no duodenal biopsy, were lost to follow-up, had a previous diagnosis of celiac disease, or whose clinical records were incomplete were excluded. Patients with IgA levels under 66 mg/dL were eliminated from the study.

Demographic, clinical, and biochemical data, histopathologic findings, and the final diagnosis were collected for all patients. The reason celiac disease was suspected was investigated, along with the laboratory data that supported the diagnosis, including iron profile, vitamin profile (folic acid and vitamin B12) and fecal fat, if present.

The anti-tTG IgA antibody levels were quantified through the ELISA (Org 240A, Orgentec) technique in all patients. In some of the cases, the tests for anti-EMA (Euroimmun Anti-Endomysium IIFT, Euroimmun), anti-deamidated gliadin IgA and IgG (Quanta Lite Gliadin IgA/IgG II, Inova Diagnostics), or the HLA DQ2/DQ8 (Olerup SSP HLA DQ2/DQ8 Typing Kit, Olerup) risk-related haplotypes were performed.

A total of six duodenal biopsies were taken from each patient (four from the second part and two from the duodenal bulb). The samples were fixed in 10% formaldehyde and embedded in paraffin. The slices (4 μm-thick) were stained with hematoxylin and eosin (H&E) and independently examined by two specialists in gastrointestinal pathology (JAR and MLS), who were blinded to the serologic marker results and clinical characteristics of each patient. The histopathologic classification was carried out according to the modified Marsh-Oberhuber criteria.

The diagnosis of celiac disease was made based on consistent clinical symptoms associated with elevated levels of serum autoantibodies (anti-tissue transglutaminase IgA, anti-EMA, or anti-deamidated gliadin IgA and IgG) and compatible histopathologic findings (Marsh 2–3). Seronegative celiac disease was considered when there were suggestive clinical data and villous atrophy that responded satisfactorily to a gluten-free diet and there was evident improvement in the histologic characteristics.

For the diagnosis of small intestinal bacterial overgrowth (SIBO), the lactulose breath test was utilized and read as suggestive when an elevation of hydrogen ≥ 20 ppm at 90 minutes or methane ≥ 10 was observed at any time.

Tropical sprue was diagnosed based on 1) compatible clinical symptoms, 2) deficient absorption of at least two different compounds (B12 in blood, fecal fat), 3) compatible histologic findings, 4) rule-out of other causes, and 5) clinical and biochemical response to treatment with tetracyclines and folic acid.

Inflammatory bowel disease was classified based on clinical presentation and endoscopic, histologic, and biochemical findings.

The group with irritable bowel syndrome was defined by the Rome IV criteria, having ruled out other causes.

The diagnosis of drug-induced diarrhea was made in the cases in which there was a temporal relationship between the onset of clinical symptoms and the start of the suspected drug (nonsteroidal anti-inflammatory drugs, immunosuppressants, or antihypertensives from the angiotensin II receptor antagonist group), ruling out other possible causes. The rest of the diagnostic criteria are detailed in Supplementary Table 1.

Regarding clinical criteria, isolated vitamin deficiency was defined as levels of serum folic acid < 2.7 ng/mL, vitamin B12 < 160 pg/mL, iron deficiency (transferrin saturation index below 15%), or weight loss (>5% body weight in 6 months). Fecal fat was determined through near-infrared reflectance spectroscopy (NIRS) (normal < 10%) in a randomly taken stool sample without following a fixed fat diet.

For the descriptive analysis, mean and median were used as measures of central tendency with standard deviation (SD) or 25%–75% interquartile range (IQR) as measures of dispersion, depending on the data distribution. The Student’s t-test was applied for comparing two means, whereas the Mann-Whitney U test was used for comparing medians in variables with nonparametric distribution (determined through the Shapiro-Wilk test). The chi-square test or the Fisher’s exact test were applied to the categorical variables, when necessary.

Receiver operating characteristic (ROC) curves were constructed to determine the optimum cutoff point of anti-tTG IgA antibodies for identifying IVA. The data analysis was carried out using Rstudio software (2023.06.1 + 524).

This work meets the current bioethical research regulations, was authorized by the Research, Bioethics Committee of theInstituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán(GAS-4987-24-25-1,). Likewise(GAS-4987-24-25-1, patient data were collected, inserted into a database, identified by numbers different from those in the clinical records to ensure data confidentiality. The authors declare that this article contains no information that could identify patients, so their informed consent was not deemed necessary.

A total of 367 medical records of patients that met the inclusion criteria were reviewed. One case with documented IgA deficiency was eliminated, resulting in a final total of 366 patients, 244 (66.7%) of whom were women. The median age of patients at diagnosis was 51 years (IQR 34–68). Table 1 shows the most relevant clinical and biochemical characteristics. Chronic diarrhea with abdominal pain and bloating were the most frequent symptoms leading to suspected celiac disease in the majority of cases (n = 214, 58.5%). Other less common symptoms were isolated nutritional deficiencies (n = 123, 33.6%) and isolated weight loss (n = 29, 7.9%). Iron deficiency was identified in 125 patients (34.1%) and steatorrhea in 92 (25.1%). A concomitant autoimmune disease was documented in 112 (30.6%) patients, and the most common one was hypothyroidism (n = 39, 10.7%). Fig. 1 shows the histopathologic findings in the study patients that are related to the final diagnosis. Of the entire group, 107 (31.8%) patients had Marsh 3a-3c villous atrophy.

Figure 1.

Duodenal biopsy (by Marsh-Oberhuber classification) in relation to the final diagnosis. Source: Original figure of our own authorship.

Fifty-three cases were diagnosed with celiac disease, corresponding to 14.5% of the study population total. Of those cases, 36 were women and 17 were men (p = 0.834). The patients with celiac disease were older (58 vs. 49 years, p = 0.012) and presented with higher percentages of iron deficiency (50.9% vs. 31.3%, p = 0.005) and steatorrhea (41.5% vs. 22.4%, p = 0.003) than the non-celiacs.

There were no statistically significant differences in the frequency of autoimmune diseases between celiacs and non-celiacs (p = 0.370). Nineteen of the celiac patients had an associated autoimmune disease; 7 of them had autoimmune hypothyroidism. Only two of the 53 patients with celiac disease were classified as having seronegative celiac disease (3.8%).

Of the 107 patients with IVA, 49 were celiac patients (45.8%). Of those 107 cases, only 3 were classified as Marsh 3c and they were all celiacs.

Nineteen celiac patients had anti-tTG IgA levels between 1-5xULN (3.2–16.0 U/mL). Levels in 8 cases were between 5-10xULN (16−32 U/mL) and 9 were above 10xULN (>32 U/mL). In 7 cases, anti-tTG IgA antibodies were negative (<3.2 mg/dL). Five of those patients were on a gluten-free diet at the time of the approach, and the other 2 were classified as having seronegative celiac disease. A level above 10xULN had 17.9% sensitivity, 100% specificity, 100% PPV, and 74.7% negative predictive value (NPV).

The area under the curve (AUC) was 70% (95% CI 63.5–76.6), with an optimum cutoff point of ≥ 3.3xULN for predicting IVA (45.3% sensitivity, 89.2% specificity) (Fig. 2).

Figure 2.

ROC curve detailing the capacity to predict duodenal villous atrophy based on serum tTG-IgA levels.

Source: Original figure of our own authorship.

Table 2 details the diagnostic accuracy of the different cutoff points of anti-tTG IgA for predicting IVA.

Anti-EMA testing was performed in 218 cases (60%). Only 7 of the cases were positive, 6 of which were associated with anti-tTG IgA antibody levels >10xULN.

Anti-deamidated gliadin IgA antibodies were determined in 117 cases (31.9%) and anti-deamidated gliadin IgG in 213 cases (58.2%). Table 3 shows the relation of the different antibodies with the anti-tTG IgA levels.

Forty-five (60%) of the 75 cases in whom risk-related haplotypes were looked for expressed the HLA DQ2/DQ8 haplotypes and 8 of those patients had celiac disease (17%).

Fig. 1 shows the diagnoses made in the total study population. Marsh 3a-3c IVA was documented in 107 of the 366 patients, and 49 of them were celiacs. SIBO was diagnosed in 25 cases, tropical sprue in 18, inflammatory bowel disease in 5, and drug-induced diarrhea in 5. Three of the drug-induced diarrhea cases were attributed to antimetabolites (azathioprine or mycophenolate mofetil), one case to nonsteroidal anti-inflammatory drugs, and one case to telmisartan. Other less common causes were exocrine pancreatic insufficiency (n = 1), miscellaneous enteropathies (2 cases of eosinophilic enteritis), malignancy (one case of lymphoma), and immunodeficiency (one case of Good syndrome). Supplementary Table 2 details the remaining cases without IVA (n = 259).