Diagnostic accuracy of coeliac serological tests: a prospective study

Abstract

The best way to test serologically for coeliac disease (CD) remains controversial, with endomysial (EMA), transglutaminase (TTG), and gliadin antibodies (AGA) being assessed in various combinations with no apparent standardization. The objective of this study was to evaluate whether TTG-IgA+/-TTG-IgG could be used as a replacement for endomysial antibodies as a reliable screen for CD in patients presenting to a major Australian tertiary referral hospital for assessment of symptoms consistent with CD. Individuals referred for gastroscopic assessment of possible CD were prospectively evaluated by duodenal biopsy assessment. The following diagnostic methods were compared: dual-isotype transglutaminase (TTG-dual), combined-isotype transglutaminase (TTG-IgA+G), TTG-IgA, combined-isotype gliadin antibodies (AGA-IgA+G), AGA-IgA, and endomysial antibody assays. Clinical performance characteristics (sensitivity, specificity, area under the curve for receiver-operating characteristic analysis; AUROC) were assessed for all kits. The correlation between transglutaminase kits was generally good, with the best transglutaminase kit demonstrating high correlation (r=0.86) with endomysial antibodies. A comparison of different types of endomysial antibody assays displayed variable diagnostic performance (sensitivity 61.90-68.42%; specificity 80.00-98.57%; AUROC 0.71-0.83). Sensitivity (90.48-92.31%), specificity (80.77-82.89%) and AUROC values (0.92-0.94) for dual-isotype transglutaminase kits displayed narrow ranges. AGA assays were less sensitive (AGA-IgA: 42.31-46.15%; AGA-IgG: 61.54%) and less specific (AGA-IgA: 85.09-87.73%; AGA-IgG: 82.46-84.09%). Dual-isotype transglutaminase testing was diagnostically equivalent to transglutaminase-IgA (AUROC 0.92 versus 0.91, P=0.33). Our study suggests that transglutaminase screening (using the IgA+/-IgG isotype) is a sensitive and specific alternative to endomysial antibody testing in the serological assessment of CD. On the basis of our findings, AGA antibody testing no longer appears to be an essential part of the diagnostic strategy for adult CD.