OBJECTIVE We have undertaken a study to assess the efficiency of serological tests in the diagnosis of celiac disease (CD) during the period January 1, 1994 to January 7, 1997. Our aim was to evaluate the sensitivity of IgA antiendomysium (EMA) and IgA antigliadin (AGA) with regard to the degree of histological abnormality in biopsy specimens of small intestine in untreated celiac disease patients and first-degree relatives. METHODS The study population comprised 101 cases: 85 untreated celiac patients and 16 first-degree relatives with a mean age of 42 yr (range, 2-76 yrs). Sixteen of 85 were excluded from study because they did not satisfy the study or diagnostic criteria of CD. EMA and AGA have been compared with the degree of villous atrophy (VA) in 69 celiac patients and 16 relatives according to the Marsh criteria of 1992. We divided the Marsh III histology into three subgroups as follows: Marsh IIIa (partial VA), Marsh IIIb (subtotal VA), and Marsh IIIc (total villous atrophy). RESULTS The specificity and positive predictive value of EMA for CD was excellent, because all EMA-positive patients (n = 42) were diagnosed with CD. The sensitivity of EMA, however, differed between CD subgroups; in patients with total VA, the sensitivity of EMA was 100% (17/17). However, in patients with partial VA (Marsh IIIa), the sensitivity of EMA was disappointing, only 9/29 (31%). Three of 72 celiacs with Marsh IIIb and Marsh IIIc had IgA deficiency and were excluded from the study. Elevated AGA has been detected in the sera of 39 of 69 (62%) patients. A combination of EMA and AGA tests showed a sensitivity of 76% (53/69). None of 16 first-degree relatives with Marsh I-II had positive EMA. CONCLUSIONS Interpretation of negative serology needs great awareness. Although EMA sensitivity in total villous atrophy is excellent, in partial villous atrophy the sensitivity of EMA appears to be disappointing. Our experience shows that EMA and AGA have only limited value in screening programs for CD.