Intestinal Biopsy is not Always Required to Diagnose Celiac Disease

A Retrospective Analysis of Combined Antibody Tests

Annemarie Bürgin-Wolff; Buser Mauro; Hadziselimovic Faruk


BMC Gastroenterol. 2013;13(19) 

In This Article


The histology of 149 consecutive patients (104 females, age range at biopsy 0.9–80 years, median age 29 years; 45 males, age range at biopsy 2–73 years, median age 13 years) revealed subtotal or complete villous atrophy, hyperplasia of the crypts, and an increase in intraepithelial lymphocytes (Marsh classification 3a, b, or c lesions). All of these patients recovered after starting a gliadin-free diet and were regarded as having CD. The biopsies of 119 consecutive patients (66 females, age range at biopsy 1.5–72 years, median age 17 years; 53 males, age range at biopsy 2–66 years, median age 7 years) revealed a normal mucosa or a mucosa with slight, nonspecific changes; these patients were considered free of CD and served as controls (The Control group was younger than the group with CD, P=0.0074).

Deamidated Gliadin Peptides Compared With Native Gliadins as Antigens

Sera from 149 patients with CD and 119 control patients were tested for IgG and IgA antibodies against dpgli and ngli proteins. IgG antibody determination for dpgli was superior to that for ngli. Specificity was 68% vs. 92% and sensitivity was 79% vs. 85% for ngli and dpgli, respectively; ppv was 76% vs. 93% and npv was 72% vs. 83% for ngli and dpgli, respectively. For IgA antibody determination, sensitivity was 61% vs. 78% for ngli and dpgli, respectively, while the specificity and ppv remained at a high level of 97% (McNemar's test for significant changes P < 0.00001, Table 1). Because dpgli antigens were clearly superior to ngli, we used only dpgli for further CD-specific antibody determinations.

Antibody Profile in CD and Control Patients

We also determined the levels of IgA anti-tTG and EMA in sera from the 149 CD patients and 119 controls (Table 2). Because the IgA anti-tTG and EMA results were comparable, we have omitted the EMA results; instead, we have shown the IgA anti- tTG, and IgA anti-dpgli, and IgG anti-dpgli antibody levels of each individual and compared them with the histological result. We used a multiple test consisting of three individual tests, which produce a total of eight possible results. We defined the outcome of the multiple tests as positive only when all three individual tests were above the cut-off, and as negative only when all three individual tests were below the cut-off. The majority of the patients (208/268) had either positive (110) or negative (98) results in all three tests. Nearly all patients (109/110) who tested positive for antibodies in all three tests had CD according to histological findings. The ppv was 99% in our population, with a CD frequency of 59% (Table 3). Patients who did not test positive for CD-specific antibodies in any of the three tests were almost all free of CD according to the results of jejunal biopsy (96/98 patients); the npv was 98%. Patients with discordant antibody results (60/268 patients, 22%) could not be defined as positive or negative for CD with the multiple tests and remained unclassified. The likelihood positive ratio (lr+) was 87 and the likelihood ratio negative (lr-) was 0.01 (Table 3). These findings indicate that a biopsy is avoidable if all antibody values are either above or below the cut-off. In patients with discordant antibody results, an intestinal biopsy is necessary to diagnose or exclude CD.

Performance of Single Antibody Tests and Selected Test Combinations

We compared the performance of IgA anti-dpgli, IgG anti-dpgli, IgA anti-tTG, and EMA tests, and calculated the sensitivity, specificity, ppv, npv, lr+, lr-, and efficiency of each test and some of the possible test combinations (Table 3). We also indicated the absolute number of patients whose antibody test results were falsely positive or falsely negative for CD, as well as those who could not be classified based on antibody tests. Most of the following diagnostic tests are multiple tests (compare with Table 2). We defined the outcome of a multiple test as positive only when all individual tests were above the cut-off, and as negative only when all individual tests were below the cut-off. Test combinations containing only IgA antibodies were not considered; they are unsuitable for diagnostic purposes, because of the possibility that some patients may be deficient in IgA.

Currently, biopsies are often performed when a patient's IgA anti-tTG or EMA test is positive. Negative serological results are usually not followed by a jejunal intervention, unless there is a very strong clinical suspicion of CD. However, the data in Table 3 clearly show that single tests are neither specific nor sensitive enough to reduce the number of biopsies in patients with symptoms of CD, although the number of nonclassified patients was zero. Single tests such as the widely used IgA anti-tTG test can give rise to many falsely classified patients.

A combination of two tests also yielded many incorrectly classified patients and is therefore unsuitable for reducing the number of biopsies. The two-test combinations yielded either too many false-positive diagnoses (IgG anti-dpgli + IgA anti-tTG or IgG anti-dpgli + EMA) or too many false-negative diagnoses (IgA anti-dpgli + IgG anti- dpgli), although the number of nonclassified patients was smaller than in combinations with more than two tests. The combination of four tests was optimal: only one patient was falsely positive, no patients were falsely negative, the ppv was 99%, the npv was 100%, the lr+ was 86, and the lr- was 0.00 For practical reasons, a combination of three tests using IgA anti-tTG instead of EMA in combination with IgA anti-dpgli and IgG anti-dpgli (Table 2) may be sufficient to set a standard (ppv 99%, one false-positive result; npv 98%, two false-negative results; and lr+ = 87, lr- = 0.01). A biopsy was avoidable in 208/268 patients (78%), while 60/268 patients (22%) could not be diagnosed with the combination of serological tests because their results were in disagreement (only one or two results above the cut-off, with the remaining result(s) below the cut-off; Table 2).