Non-biopsy diagnosis of celiac disease is possible in children with anti-transglutaminase 2 antibodies (TGA) > 10× the upper limit of normal (ULN) and positive anti-endomysial antibodies (EMA). Similar criteria have been suggested for adults, but evidence with different TGA assays is scarce. We compared the performance of four TGA tests in the diagnosis of celiac disease in cohorts with diverse pre-test probabilities. Serum samples from 836 adults with either clinical suspicion or family risk of celiac disease were tested with four commercial TGA assays, EmA and celiac disease-associated genetics. The diagnosis was set based on duodenal lesion or, in some cases, using special methods. 137 (57%) patients with clinical suspicion and 85 (14%) of those with family risk had celiac disease. Positive predictive value (PPV) for 10×ULN was 100% in each TGA test. The first non-diagnostic investigations were encountered with ULN 1.0×–5.1× in the clinical cohort and 1.3×–4.9× in the family cohort, respectively. Using the assays’ own cut-offs (1×ULN) the PPVs ranged 84–100%. Serology-based diagnosis of celiac disease was accurate in adults using different commercial kits and pre-test probabilities using 10×ULN. The results also suggest that the ULN threshold for biopsy-omitting approach could be lower.